Fig 1.
Overview on circRNA biogenesis, experimental setup of HSV-1 infection time-courses, and NEAT1 transcripts and circRNAs.
(a) circRNA biogenesis occurs via back-splicing connecting the 3’ end of a downstream exon with the 5’ end of an upstream exon (indicated by dashed gray lines) or stabilization of lariats. CircRNAs originating from lariats are also denoted as circular intronic RNAs (ciRNAs) [17]. (b) Experimental setup for the total RNA- and 4sU-seq time-courses in WT and Δvhs HSV-1 infection. Fig adapted from [18] (CC BY 4.0 license, © the authors). Two biological replicates were obtained for each time-point. (c) Genomic coordinates of the NEAT1_1 (polyadenylated) and NEAT1_2 (non-polyadenylated, stabilized by triple helix) transcripts (blue) and NEAT1_2 circRNAs observed in HSV-1 or influenza A virus (IAV) infection. The hsa_circ_0003812 circRNA induced strongly in HSV-1 and IAV infection is shown in red, while other much less abundant NEAT1_2 circRNAs are indicated in orange.
Fig 2.
vhs-dependent enrichment of circRNAs in HSV-1 infection.
(a) Outline of our pipeline for de novo circRNA detection (see also methods). It combines two different algorithms, circRNA finder [44] and CIRI2 [45], and retains only circRNAs identified independently by both algorithms. circRNA finder is based on analysis of chimeric reads determined with the RNA-seq mapper STAR [49]. Chimeric reads are reads that align to two distinct regions of the genome in manner not consistent with “normal” linear transcripts. CIRI2 is based on identifying pairs of clipped read alignments (i.e., local alignments for substrings of the read, determined with BWA [50]) for the same read where a downstream part of the read aligns upstream of an upstream part of the read (= back-spliced junction (BSJ) reads). Both methods use further filtering steps to remove false positive results. (b) Number of circRNAs identified for each sample of total RNA for WT and Δvhs infection after adjusting the required number of reads for detecting a circRNA to the number of reads mapping to the host genome in each sample. Thus, a threshold of 2 reads was used for the sample with the lowest number of reads and higher thresholds were used for samples with higher number of host reads. Numbers of identified circRNAs with a threshold of 2 reads for each sample (as outlined in (a)) are shown in S1A Fig for total RNA and in S1B Fig for 4sU-RNA. (c, d) Scatterplots comparing normalized circRNA counts (normalized to the number of linear junction reads mapped to the host genome) between mock and 8 h post infection (p.i.) WT infection (c) and mock and 8 h p.i. Δvhs infection (d). Linear regression analysis across all circRNAs (red line) was used to estimate the enrichment of circRNAs relative to linear mRNAs in HSV-1 infection compared to mock infection. The regression estimate for the enrichment is shown on the bottom right. The gray line indicates the diagonal, i.e., equal values on the x- and y-axis. The five most highly expressed circRNAs are marked by name. Corresponding scatterplots for 2, 4 and 6 h p.i. are shown in S2 Fig. (e) Percentage of expressed circRNAs (= normalized circRNA count >0 in uninfected cells) at each time-point of WT HSV-1 infection for which at least one exon within the genomic region of the circRNA shows differential exon usage for the corresponding gene (determined with DEXSeq, multiple testing adjusted p-value ≤0.005). (f) Boxplots showing the distribution of log2 fold-changes for exons located within circRNAs. For each circRNA, only the exon with the maximum absolute log2 fold-change is shown.
Fig 3.
Induction of a NEAT1_2 circular and linear splicing in WT, Δvhs and ΔICP27 infection.
(a) Heatmap of normalized circRNA counts (normalized to the number of linear junction reads mapped to the host genome) for all identified NEAT1_2 circRNAs in total and 4sU-RNA time-courses of WT and Δvhs infection and nucleoplasmic, chromatin-associated, cytoplasmic and total RNA for 8 h p.i. WT, Δvhs and ΔICP27 infection. Columns represent individual circRNAs, which are ordered according to their genomic coordinates from the most 5’ to the most 3’. The hsa_circ_0003812 NEAT1_2 circRNA is marked by a blue rectangle. (b, c) Sashimi plots showing NEAT1 read coverage (overlay of both replicates) and circular (red) and linear (same color as read coverage) splice junctions as arcs connecting acceptor and donor splice site in (b) selected samples of the total RNA time-courses of WT and Δvhs infection and (c) chromatin-associated RNA in mock, WT, Δvhs and ΔICP27 infection. Number of junction reads are annotated to arcs separately for the two replicates. Junctions are only shown if at least 4 reads align by at least 10 nt on both sides of the junction. Genomic coordinates of NEAT1 transcripts and the HSV-1-induced circRNA are shown at the bottom. Sashimi plots of the full total RNA and 4sU-RNA time-courses and nucleoplasmic and total RNA from the subcellular fractions experiment are shown in S13 and S15 Figs. (d) Heatmap of linear splicing rates (= number of reads for a linear junction / number of exon-intron reads crossing the corresponding acceptor and donor splice sites) for the WT and Δvhs infection time-courses and the subcellular fraction experiment (dark red = high, white = low, see color bar on the right). Columns represent splice junctions, which are ordered according to their genome positions, first by the donor and then by the acceptor splice site. Counts include only junction and exon-intron reads overlapping the junction or exon-intron boundary, respectively, by at least 10 nt on either side. A pseudocount of 0.01 was used to avoid division by zero. Rectangles on top indicate whether the splice junction is within the genomic region or downstream of the hsa_circ_0003812 circRNA. Colors on the left indicate the virus and time-point of infection (light colors early in infection, darkest colors 8 h p.i, cyan = WT, blue = Δvhs, gold = ΔICP27). Numbers in brackets at the bottom refer to the four most frequent splice junctions outlined in (e). (e) Schematic overview on the circular and the four most frequent linear NEAT1_2 splice junctions induced in HSV-1 infection and sequences around splice sites (“|” indicates exon-intron boundaries in sequences). The height of boxes indicates the relative abundance of each exon (= never or rarely spliced out) and intron (= frequently spliced out) by 8 h p.i. WT infection. Genomic coordinates on chromosome 11 are indicated for the start and end of exons. Numbers in brackets on top of introns indicate the ranking of corresponding splice junctions according to their splicing rate shown in (d). This ranking is also indicated in (d) at the bottom.
Fig 4.
Ectopic co-expression of ICP22 and ICP27 is sufficient for induction of NEAT1_2 splicing.
Sashimi plots as in Fig 3 showing read coverage and circular and linear splice junctions in (a) T-HFs-ICP22/ICP27 cells without and with Dox-induced co-expression of ICP22 and ICP27 as well as mock, WT-F and ΔICP22 infection at 12 h p.i., (b) T-HFs-ICP22 cells without and with Dox-induced ICP22 expression and (c) mock and WT KOS infection and ICP27 overexpression. A minimum read count of 4 was again required for (a), while in (b) and (c) all junction reads are shown covering at least 10nt on either side of the junction. For an explanation of sashimi plots, see caption to Fig 3. Circular splice junctions are marked in red, linear splice junctions in the same color as read coverage.
Fig 5.
Induction of NEAT1_2 circular and linear splicing in IAV infection.
(a) Heatmap of normalized circRNA counts (normalized to number of linear splice junctions mapped to the host genome) for the three most highly expressed NEAT1_2 circRNAs identified in H1N1, H3N2 and H5N1 infection of HTBE and MDM cells or time-matched controls. Columns represent individual circRNAs, which are ordered according to their genomic coordinates from the most 5’ to the most 3’. The NEAT1_2 hsa_circ_0003812 also induced in HSV-1 infection is marked by a blue rectangle. IAV strain and time-points of infection are color-coded on the left (white = time-matched control, cyan = H1N1, green = H3N2, purple = H5N1, darker colors indicate later time-points). (b) Heatmap of linear splicing rates (= number of reads for a linear junction / number of exon-intron reads crossing the corresponding acceptor and donor splice sites) in IAV infection and time-matched controls (dark red = high, white = low, see color bar on the right). Columns represent individual splice junctions, which are ordered according to their genome positions. Counts include only junction and exon-intron reads overlapping the junction or exon-intron boundary, respectively, by at least 10 nt on either side. A pseudocount of 0.01 was used to avoid division by zero. Colors on the left indicate time-points of infection as in (a). Colored rectangles on top indicate whether linear splice junctions were also observed in HSV-1 infection (cyan) or are within (blue) or downstream (gray) of the hsa_circ_0003812 circRNA region. Numbers in brackets at the bottom refer to the introns outlined in the schematic in Fig 3E.
Fig 6.
NEAT1_2 splicing and read-through upon CDK7 inhibition.
(a) Heatmap of linear splicing rates (= number of reads for a linear junction / number of exon-intron reads crossing the corresponding acceptor and donor splice sites) upon treatment by DMSO/control, THZ1 or Act-D or knockdown (KD) of CDK7 or MED1 (dark red = high, white = low, see color bar on the right). Columns represent individual splice junctions, which are ordered according to their genome positions. Counts include only junction and exon-intron reads overlapping the junction or exon-intron boundary, respectively, by at least 10 nt on either side. A pseudocount of 0.01 was used to avoid division by zero. Colored rectangles on top indicate whether linear splice junctions were also observed in HSV-1 (cyan) or IAV (purple) infection or are within (blue) or downstream (gray) of the hsa_circ_0003812 circRNA region. Numbers in brackets at the bottom refer to the introns outlined in the schematic in Fig 3E. (b, c) Sashimi plots as in Fig 3 showing read coverage and linear splice junctions in (b) UM-Chor1 cells after 4, 8, 12 and 24 h THZ1 or control treatment (3 replicates each) and (c) CDK7 and MED1 knockdown in VCaP and LNCaP cells (2 replicates each). All junctions are shown that are covered by at least 4 reads aligning to at least 10 nt on either side of the junction. (d) Boxplots showing distribution of %read-through (for calculation see methods) after THZ1 treatment in example studies or CDK7 or MED1 knockdown in VCaP and LNCaP cells.