Human cytomegalovirus forms phase-separated compartments at viral genomes to facilitate viral replication

Abstract

Human cytomegalovirus (HCMV) replicates its DNA genome in specialized replication compartments (RCs) in the host cell nucleus. These membrane-less organelles originate as spherical structures and grow in size over time. However, the mechanism of RC biogenesis has remained understudied. Using live-cell imaging and photo-oligomerization, we show that a central component of RCs, the UL112-113 proteins, undergo liquid-liquid phase separation (LLPS) to form RCs in the nucleus. We show that the self-interacting domain and large intrinsically disordered regions of UL112-113 are required for LLPS. Importantly, viral DNA induces local clustering of these proteins and lowers the threshold for phase separation. The formation of phase-separated compartments around viral genomes is necessary to recruit the viral DNA polymerase for viral genome replication. Thus, HCMV uses its UL112-113 proteins to generate RCs around viral genomes by LLPS to ensure the formation of a pro-replicative environment.

Keywords: Herpesviridae; LLPS; human cytomegalovirus; human herpesvirus 5; liquid-liquid phase separation; live-cell imaging; membrane-less organelle; molecular virology; phase transition; replication compartment; viral replication.

Graphical abstract

Figure 1

Pre-replication compartments are fluid biomolecular condensates formed by LLPS (A) Schematic of the UL112-113 gene and its four protein products. (B) Time-lapse images of MRC-5 cells infected with HCMV-mNeonGreen-UL112 (MOI = 10). (C) Aspect ratios of PRCs and RCs at 24 and 48 hpi ± PAA treatment for 48 h. An aspect ratio (width:height) close to 1 indicates roundness. n = 30 per condition. (D) PRCs fusing in MRC-5 cells infected with HCMV-mNeonGreen-UL112 (MOI = 1) at 24 hpi. (E) Aspect ratios of fusing mNeonGreen-UL112-113 foci over time from (D). (F) MRC-5 cells were infected with HCMV-mNeonGreen-UL112 (MOI = 1) for 24 or 48 h and treated with 2% 1,6-HD for 90 min (pulse) and imaged by spinning disk microscopy. The medium was replaced and cells were imaged for another 30 min (chase). (G) MRC-5 cells were infected with HCMV-mNeonGreen-UL112 (MOI = 1) for 24 h, 48 h, or treated with PAA for 48 h and analyzed by FRAP. (H) FRAP curves show means ± SD of n cells. (I) Mobile fraction calculated from the data in (H). (J) Half-time recovery calculated from the data in (H). The data shown are representative of three independent experiments (B, C, F–I) or three individual replicates (D and E). Yellow dotted lines indicate nuclear boundaries. Red dotted lines indicate areas of analysis. Scale bars, 10 μm and 2 μm (insets). ∗∗∗ p < 0.001.

Figure 2

UL112-113 forms liquid compartments in transfected cells (A) HEK293A cells were transfected with a plasmid expressing mNeonGreen-UL112-113. The four UL112-113 isoforms were detected with an anti-mNeonGreen antibody. (B) Maximum intensity projections of cells transfected as in (A), fixed after 48 h, and analyzed by cLSM. (C) Aspect ratios (width:height) of UL112-113 foci from (B). (D) Time-lapse, maximum intensity projections of cells treated as in (A) depicting droplet fusion. (E) Aspect ratios of fusing droplets marked with a red dotted line in (D). (F) HEK293A cells were co-transfected with mNeonGreen-UL112-113 and mCherry-M45 plasmids. Cells were imaged during treatment with 2% 1,6-HD (pulse) and after its removal (chase). (G) Cells transfected as in (F), before and after a 5 min pulse with 2% PG. (H) HEK293A cells were transfected with plasmids encoding mNeonGreen-UL112-113, mCherry-Nucleolin as a positive and mCherry-M45 as a negative control and analyzed 2 days post-transfection by FRAP. (I and J) Mobile fractions and FRAP curves of cells transfected as in (H). Curves show means ± SD of n cells from two independent experiments. (A–J) The data shown are representative of three independent experiments. Scale bars, 10 μm and 2 μm (insets). ∗∗∗p < 0.001.

Figure 3

Clustering of the UL112-113 IDR is sufficient for LLPS (A) Binary classification of phase separation in vitro (from Figure S3B) as a function of UL112-113 (x axis) and KCl concentrations (y axis). White dots, no phase separation; green dots, phase separation. (B) Representative images of UL112-113 droplets. LLPS was induced by mixing purified UL112-113 proteins with a low salt (75 mM KCl) buffer. Treatment with 500 mM KCl or 5% 1,6-HD was used to inhibit phase separation. 2 μg/μL of soluble mNeonGreen was mixed with a low salt buffer. Scale bar, 100 μm. (C) Schematic illustrating the corelet system (adapted from [Bracha et al., 2018]). iLID-EGFP-ferritin forms a 24-mer corelet (green star). sspB-mCherry (red crescent), is conjugated to IDRs from different proteins. sspB fusion proteins bind to corelets upon blue light activation, leading to LLPS (blue). (D) Schematic of constructs. mCherry-sspB was fused to the IDRs of UL112-113 p43, p84, HNRPA1c as a positive control, or left unfused as a negative control. (E) Photoactivation of corelet-expressing cells as depicted in (D). (A–C and E) The data shown are representative of three independent experiments. Scale bar, 10 μm.

Figure 4

Exogenous DNA stimulates UL112-113 phase separation (A) T-REx-293 cells expressing mNeonGreen-UL112-113 in a Dox-dependent manner were either left untreated, induced for 21 h with 2 μg/mL Dox and then infected with a replication-incompetent HSV-1 (MOI = 10) for 3 h, or transfected with plasmid DNA, incubated for 24 h, and then induced with 2 μg/mL Dox for another 24 h before fixation and imaging. MRC-5 cells infected with HCMV-mNeonGreen-UL112 (MOI = 1) for 5 h were used for comparison (MRC-5 infection). Maximum intensity projections are shown. (B) Mean summed pixel intensities of cells in (A). Red are intensities from cells without UL112-113 dots, blue with dots. (C) Time-lapse of UL112-113 behavior in induced T-REx cells with or without exogenous DNA. Cells were mock-transfected (−DNA) or transfected (+DNA) with an mCherry plasmid 24 h before induction with 2 μg/mL Dox for the indicated times. (D) Maximum intensity projections of cells transfected as in (C) induced with 1–4 μg/mL Dox and imaged 24 h post-induction. (E) Cells were manually binned into three categories. 0, disperse, no droplets; 1, metastable, some small droplets with high nuclear UL112-113 background signal; 2, phase separation, large droplets and reduced nuclear background. n = 50 cells per condition. (F) HEK293A cells expressing iLID-EGFP-ferritin and p84IDR-mCherry-sspB globally photoactivated for 5 min. (G) Cells from (F) photoactivated locally resulting in local LLPS. (H) In vivo phase diagram of p84IDR-mCherry-sspB given as core fluorescence intensity and core-to-IDR ratio (Bracha et al., 2018). Solid red circles, LLPS after 10 min of photoactivation; empty red circles, no LLPS; blue circles, LLPS after local activation see (G). Data are pooled from three independent experiments. (A–F) The data shown are representative of three independent experiments. Scale bar, 10 μm. ∗∗∗p < 0.001, ∗p < 0.05, ns not significant.

Figure 5

UL112-113 droplets form at viral genomes and facilitate viral DNA replication (A) Association of UL112-113 droplets with viral genomes. HFF were infected with EdU-labeled or unlabeled HCMV-mNeonGreen-UL112 (MOI = 10). Cells (24 hpi) were fixed, Click-labeled with AF555-Picolyl-azide, and counterstained with Hoechst 33342. The inset depicts an EdU-labeled genome associated with a UL112-113 droplet. (B) Quantification of EdU-labeled genomes associated with UL112-113 droplets. n = total number of genomes counted in 44 cells. (C) Effect of UL112-113 droplet association on genome replication. MRC-5 cells were infected with EdU-labeled HCMV-mNeonGreen-UL112 (MOI = 1). At 24 hpi, cells were pulsed for 6 h with 100 μM BrdU and fixed. EdU-labeled incoming genomes were Click-labeled and nuclei were counterstained with Hoechst 33342. BrdU was detected by indirect immunofluorescence. Upper inset, replicating genome; lower inset, non-replicating genome. (D) Quantification of genome area as a function of UL112-113 association. Incoming EdU viral genome replication was detected through incorporation of BrdU. EdU- and BrdU-labeled summed areas were scored as UL112-113 positive or negative. n = number of incoming EdU-labeled genomes analyzed. (A–D) The data shown are representative of three independent experiments. Scale bars, 10 μm and 2 μm (insets). ∗∗∗p < 0.001, ∗∗p < 0.01, ns not significant.

Figure 6

Phase separation of UL112-113 facilitates recruitment of the viral DNA polymerase accessory factor UL44 to genomes (A) Influence of inhibition of UL112-113 LLPS on genome replication. MRC-5 cells were infected with HCMV-mNeonGreen-UL112 (MOI = 1). At 24 hpi, cells were pulsed for 2 h with 10 μM EdU to label newly synthesized DNA. At the same time, cells were treated with PAA (250 μg/μL), 2% 1,6-HD, or left untreated. Cells were Click-labeled with AF555-Picolyl-azide (EdU), counterstained with Hoechst 33342, and UL44 was detected by indirect immunofluorescence. (B–D) Influence of 1,6-HD on UL112-113 foci and genome replication. Number of UL44 positive foci, area of genome labeling, and number of UL44 positive foci under conditions as in (A). Only cells expressing UL112-113 and UL44 were analyzed. n = number of foci per nucleus analyzed. (E) Schematic of mCherry-tagged UL44 constructs. UL44 consists of an ordered N-terminal region (aa 1–290), a disordered C-terminal IDR (aa 291–424), and an NLS (aa 425–433). (F) HEK293A cells were co-transfected with UL44 and UL112-113 plasmids as indicated in (E) and imaged 48 h post-transfection. Intensity plots illustrate signal profiles along the white dotted lines. (A–D and F) The data shown are representative of three independent experiments. ∗∗∗p < 0.001, ∗∗p < 0.01.