Membrane deformation and scission by the HSV-1 nuclear egress complex

Abstract

The nuclear egress complex (NEC) of herpesviruses such as HSV-1 is essential for the exit of nascent capsids from the cell nucleus. The NEC drives nuclear envelope vesiculation in cells, but the precise budding mechanism and the potential involvement of cellular proteins are unclear. Here we report that HSV-1 NEC alone is sufficient for membrane budding in vitro and thus represents a complete membrane deformation and scission machinery. It forms ordered coats on the inner surface of the budded vesicles, suggesting that it mediates scission by scaffolding the membrane bud and constricting the neck to the point of scission. The inward topology of NEC-mediated budding in vitro resembles capsid budding into the inner nuclear membrane during HSV-1 infection and nuclear envelope vesiculation in NEC-transfected cells. We propose that the NEC functions as minimal virus-encoded membrane-budding machinery during nuclear egress and does not require additional cellular factors.

Fig. 1. Purification of soluble NECs

a, Schematic view of HSV-1 UL31 and UL34 and constructs used here. NLS = nuclear localization sequence, BS = binding site, DN = dominant negative mutation, TM = transmembrane region. b, The purified NEC complexes are homogeneous as shown by 12% SDS-PAGE and Coomassie staining. Panels from two different gels are shown side by side (full gels are shown in supplementary figure 7a).

Fig. 2. The soluble NEC binds model membranes

a, Membrane binding was determined using coflotation with 100 nm LUVs. 5-mL sucrose gradients were fractionated into 5 parts, from top (1) to bottom (5), and visualized by Western Blot using anti-UL31 or anti-UL34 antibodies (full western blots are shown in supplementary figure 7b). b, Membrane binding was quantified using co-sedimentation with MLVs and Coomassie staining (full gels are shown in supplementary figure 7c). IP = input, PP = protein pellet, P = pellet, S = supernatant. c, At least 40% of acidic lipid is required for NEC220 binding. As an additional negative control, liposomes were incubated with a PreScission protease (Prsc) that does not bind membranes. The ~30% “background” signal in the absence of liposomes is due to protein precipitation at the speed used in centrifugation. d, Interaction with MLVs depends on initial salt concentration. No binding was observed in the presence of NaCl concentrations higher than 175 mM. Addition of 1 M NaCl after complex formation did not abrogate the interaction. e, Deletion of 50 but not 40 residues from the N terminus of UL31 reduces binding. The NEC185 binds worse than the NEC220 and combination of both truncations (NEC185-Δ50) lowers the binding affinity further. Experiments in c-e were done in triplicate and the reported values represent averages of the results of three individual experiments. Error bars represent the standard errors of measurement. The statistical analysis used is the Student t test. One asterisk indicates p values smaller than 0.05 and two asterisks indicate p values smaller than 0.005. The asterisks above each sample represent the significance compared to the background, whereas the asterisks above each line represent the significance between these two samples after subtracting the individual background levels.

Fig. 3. NEC220 binding to GUVs drives membrane invagination and scission

GUVs containing 2% DOPE-ATTO594 (red) were visualized alone or in the presence of the NEC220-SNAP labeled with SNAP Surface 488 (green). a, Free GUVs are spherical. b, Upon addition of the NEC220, membrane invaginations and protein-bound intraluminal vesicles inside intact GUVs were observed. Images show membrane invaginations wherever the NEC220 is bound. c, GUVs containing a Nickel-chelated lipid were used with NEC246-His to mimic the membrane anchor. These GUVs were less acidic, but a similar scenario as with the soluble NEC220 could be observed. Wherever the protein is bound, the GUVs are flattened. d, In the presence of NEC220-DN, GUVs are mostly spherical even after prolonged incubation. e-f, When GUVs (red) were sequentially incubated with non-permeable fluorescent dye Cascade Blue and the unlabeled NEC220 or His-tagged NEC220, intraluminal vesicles containing Cascade Blue were observed. g, Quantification of these vesicles show that 12% of all GUVs contain ILVs when NEC220 was added, whereas there are no Cascade Blue filled ILVs in the absence of any protein or in the presence of NEC220-DN. NEC220-His shows a similar behavior with less acidic GUVs as the soluble NEC220 (13% ILVs). The reported values represent averages of the results of at least two individual experiments. Error bars represent the standard errors of measurement from at least two individual experiments, with a count of at least 67 GUVs per sample and experiment (The number of vesicles counted in total are as follows: no protein (104, 111), NEC220 (83, 75, 119), NEC220-DN (77, 136), NEC220-His (67, 86, 109). The statistical analysis used is the Student t test. One asterisk indicates p values smaller than 0.01. Two asterisks indicate p values smaller than 0.005. Asterisks located above the black line indicate significance between these samples and asterisks above each data point indicate the significance compared to ‘no protein’ control.

Fig. 4. The NEC220 but not the NEC220-DN mutant forms ordered arrays on the inner surface of the LUVs

a, Representative cryoelectron micrographs of LUVs (400 nm and 800 nm) alone, in the presence of NEC220, or in the presence of NEC220-DN. b, The NEC220 forms ordered arrays on the inner surface of the LUVs. c, The NEC220-DN mutant forms clusters of spikes mainly on the outer surface of the LUVs. d, A model of how membrane binding and oligomerization of the NEC leads to membrane invagination and scission, based on fluorescence and cryoelectron microscopy. Presumed steps are numbered. The “mother” vesicle eventually breaks down, releasing intact vesicles coated with the NEC on the inside. The NEC220-DN mutant does not proceed beyond step 2 (blue asterisk).

Fig. 5. NEC220 forms hexagonal honeycomb arrays on LUVs

a, Cryo-EM image of a representative vesicle with close-up views of two types of observed arrays, rings arranged in a honeycomb pattern and “fences”. b, A membrane patch from a broken LUV shows a single-layer NEC220 array. c, A close-up view of the hexagonal array. 2D averaging enhances the 6-fold symmetry. A model of NEC220 has been placed over the electron densities to highlight the protein densities. NEC220 spike is depicted as a magenta cylindrical stem topped with a blue sphere. d, 3D-averaged cryo-EM images in top and side views are shown side-by-side with models of NEC220 placed over the electron densities to highlight the protein densities. A model of a section of the NEC220 hexagonal array in a slanted view is also shown. e, The model of the membrane-bound inner NEC220 coat. UL31 has been assigned to the blue “sphere” part of the spike on the basis of the crosslinking data. UL34 has been assigned to the “stem” part because in full-length NEC, UL34 anchors the complex to the membrane. These assignments are approximate and are not based on any direct structural data. The precise arrangement of UL31 and UL34 within the membrane-bound NEC spike is as yet unknown.

Fig 6. NEC220 is crosslinked into high-molecular-weight species in the presence but not absence of acidic LUVs

a-c, Bis(sulfosuccinimidyl)suberate (BS³) was added at 0, 25, 50, 250, 500, 2500-fold molar excess. Samples were analyzed by 12% SDS-PAGE and Coomassie staining.