Structural Basis of Vesicle Formation at the Inner Nuclear Membrane

Abstract

Vesicular nucleo-cytoplasmic transport is becoming recognized as a general cellular mechanism for translocation of large cargoes across the nuclear envelope. Cargo is recruited, enveloped at the inner nuclear membrane (INM), and delivered by membrane fusion at the outer nuclear membrane. To understand the structural underpinning for this trafficking, we investigated nuclear egress of progeny herpesvirus capsids where capsid envelopment is mediated by two viral proteins, forming the nuclear egress complex (NEC). Using a multi-modal imaging approach, we visualized the NEC in situ forming coated vesicles of defined size. Cellular electron cryo-tomography revealed a protein layer showing two distinct hexagonal lattices at its membrane-proximal and membrane-distant faces, respectively. NEC coat architecture was determined by combining this information with integrative modeling using small-angle X-ray scattering data. The molecular arrangement of the NEC establishes the basic mechanism for budding and scission of tailored vesicles at the INM.

Graphical abstract

Figure 1

The NEC in the Replication Cycle of Herpesvirus (A) Schematic of the stages of vesicle-mediated herpesvirus capsid nuclear egress, consisting of (i) primary envelopment by the NEC (green) at the INM and (ii) fusion of the vesicle with the ONM, resulting in de-envelopment to release the capsid into the cytoplasm. (B–F) Developmental stages of the NEC coat in HSV-1-infected Vero cells (moi: 10, 16 hr p.i.) analyzed by electron cryo-microscopy of vitreous sections (CEMOVIS). (B) Projection image taken after pre-irradiation; nominal section feed: 30 nm; compression: 47%, corrected. (B′) Magnification of the yellow box marked in (B) (arrows: NEC coat). (C–E) Slices of tomographic reconstructions (C and D: nominal section feed, 100 nm; compression, 13%, 3D-corrected; E: nominal section feed, 50 nm; compression, 26%, 3D-corrected; asterisk, ILV; Movies S1 and S2). (F) Projection image taken after pre-irradiation; nominal section feed, 30 nm; compression, 47%, corrected. Scale bar, 200 nm (B and F) and 100 nm (B′–E). cyt, cytoplasm; INM, inner nuclear membrane; NP, nuclear pore; nuc, nucleus; ONM, outer nuclear membrane.

Figure 2

Nuclear Ultrastructure in PrV pU_L31/pU_L34-GFP Co-Expressing Cells (A) Slices through a 3D volume of a porcine epithelial-like embryonic kidney cell determined by live-cell 3D-SIM at 37°C (∼120 nm lateral resolution). The cell nucleus, as well as features of the nuclear envelope (arrow, tubular invagination; asterisks, vesicle clusters or “speckles”), are highlighted by stably co-expressing PrV pU_L31 and pU_L34, the latter fused to GFP. Thickness of XZ and YZ projections is indicated by dashed lines in the XY projection (for 3D volume, see Movie S3). Scale bar, 5 μm. (B) A slice of a CEMOVIS tomographic reconstruction of the nuclear periphery of a proteinase-K-detached BK cell cryo-immobilized after 2 days standard cultivation (nominal section feed, 100 nm; compression, 13%, 3D-corrected) depicts the typical size range of ILVs in an invagination of the INM (arrow, “stalk” region; asterisks, membrane crevasses). Scale bar, 200 nm. cyt, cytoplasm; nuc, nucleus.

Figure 3

Ultrastructural Characterization of a Cluster of ILVs by CryoFIB/ET (A–C) Slices through an electron cryo-tomogram of a lamella prepared by cryoFIB, interpreted schematically in (C). Vesicles not shown in the experimental map slices are depicted semi-transparently (Movie S6). ILVs are tightly surrounded by the INM (black arrows) and are closely related in size, exhibiting diffuse contents. The NEC coat (green) appears to nearly cover the entire inner surface of the vesicle membrane (white arrows). A fuzzy layer of density attributed, at least partly, to the C-terminal GFP of the type II transmembrane protein construct pU_L34-GFP, surrounds each vesicle and projects into the intraluminal/perinuclear space (ILS). Inspection of tangential slices (example: black star) suggests that imperfections in the lattice arrangement of the NEC coat do occur but that these likely represent only a small fraction of the total vesicle surface area. Red asterisks highlight a near spherical vesicle from which measurements of NEC coat parameters were taken. Scale bar, 100 nm. nuc, nucleus. (D) Boxplot of the distribution of vesicle sizes (vesicle inner diameter: dashed line in inset vesicle) measured in 3D from BK cells prepared for tomographic electron imaging by CEMOVIS (red) and cryoFIB (blue). The distributions share a median of ∼100 nm. VM, vesicle membrane.

Figure 4

Sub-tomogram Averaging of the NEC Coat (A) Sub-tomogram average (∼3.5 nm resolution, Figure S2) viewed in characteristic “6-2-6” and “3-2-3” cross-section, oriented such that the slice passes through a 2-fold axis and intercepts adjacent 6-fold (Movie S7) or 3-fold axes, respectively. Scale bar, 10 nm. VM, vesicle membrane. (B) Tangential slices through membrane proximal (MP) and membrane-distal (MD) layers (Movie S8). Each layer corresponds to p6 symmetry with lattice spacing of ∼11 nm. However, the MP layer exhibits a distinct arrangement of density that, at the available resolution, appears to correspond to a p6 lattice with a spacing of ∼6 nm and offset from the MD lattice by 30°. Scale bar, 10 nm. (C) Surface views rendered from the vesicle exterior and the vesicle interior show characteristic features of the MP and MD layers, respectively (Movie S9). Symbols indicate symmetry axes (6-fold, 3-fold, and 2-fold).

Figure 5

Interpretation of the NEC Coat Based on the Sub-tomogram Average The initial architectural model of the NEC coat (lower-right) made from hexameric interactions of pU_L31/34 heterodimers (hexameric unit cell; upper-middle) is shown alongside characteristic cross-sectional views of the NEC coat average (“6-2-6” and “3-2-3” views shown in Figure 4A and “6-2-6” in Movie S7). An arch keystone is marked with a white asterisk. pU_L34 (magenta) and pU_L31 (purple) make up the MP and MD layers, respectively.

Figure 6

Stoichiometry of the pU_L31/34 Heterodimer within the NEC Coat, Based on SAXS Data (A–F) (A) SAXS scattering curve for soluble PrV NEC (red circles; Figure S3) with the fit of the theoretical scattering curve calculated from the ab initio model (green line) and the respective size distribution (inset). (B) The ab initio heterodimeric model derived from the simulated annealing bead modeling of the 1D-SAXS curve. (C) Surface views of the SAXS-based hexameric model ([pU_L31/34]₆) accounting best for the cryoEM-derived density. One hexamer is rotated to show four views from different directions. (D–F) Four copies of [pU_L31/34]₆ fitted into the cryoEM map (transparent gray surface, compare Movie S10) are shown in both cross sections (D and E) as well as semi-transparent surface view from the vesicle exterior (F). Note the missing density at the center of each hexamer (the arch keystone, instances marked with a black asterisk in D and E), as the model is based on a SAXS model of a heterodimer with truncated pU_L34 (Figure S3B). (G–I) The EM map of the NEC coat with four fitted hexameric SAXS-based models of soluble heterodimeric NEC is viewed from the side and sliced to remove density up to the “6-2-6” section passing through the map’s center (G; for full hexamers, see Movie S10). The SAXS-derived model accounts for the EM density archways in all regions except the arch itself, thereby serving to validate the stoichiometry of our initial architectural model based on prediction of protein occupied volume. Tangential slices through the map at radii corresponding to MP (H) and MD (I) layers show that the “fitting search” using the SAXS-derived shape model is able to reproduce the characteristic features of the NEC coat (i.e., the two-layered arrangement) but also suggest that interactions between heterodimers occur predominantly across the 2-fold axis within both MD and MP layers, associated with pU_L31, and pU_L34, respectively.

Figure 7

Architectural Basis for Constrained Curvature Formation (A) A schematic model of the NEC where characteristic layers are represented as they appear in tangential slices (Figure 4B and Movie S8). Alignment of pink and purple points will result in formation of curvature essentially defined by the radial separation of each layer (parameter “h”). (B) A “6-2-6” cross-section view of the NEC coat average is modeled according to parameters (h, V_MD, and V_MP; compare Equation 1 in the Supplemental Experimental Procedures) measured from the experimental average. Magenta (MP) and purple (MD) circles highlight that hexagonal layers of characteristic repeat distances (V_MD and V_MP) interact via the pU_L31/34 heterodimer interface to induce a defined curvature. The basis for determination of the exact radial position of the two layers is given in Figure S4, and an arch keystone is marked with a white asterisk. (C) The NEC coat diameter between opposite MD layers (d_MD) is plotted as a function of a (the arc quotient), and experimentally determined vesicle sizes from cryoFIB-prepared samples are shown as colored circles. The vesicle modeled in (B) is represented in red (see also red asterisk in Figures 3A–3C), while the mean vesicle size is indicated in green and a coated capsid by the respective symbol in gray, each with their respective a values. These values are plotted assuming that V_MD is constant, while V_MP is hypothesized to vary, owing to flexibility within coat. (D) The resulting model in the context of an entire vesicle cross-section produced by extrapolation as described in the Supplemental Experimental Procedures. pU_L34 (magenta) and pU_L31 (purple) make up the MP and MD layers, respectively.

Figure S1

Nuclear Structure in PrV pU_L31/pU_L34-GFP Co-expressing Cells, Related to Figure 2 (A–D) Soft X-ray cryo-microscopy/tomography (cryoXM/T) of BK cells (A) Slice of a soft X-ray tomogram, taken with a 40 nm zone plate objective (yellow box and a magnification of a similar slice in the same area: inset). CryoXM/T directly images frozen-hydrated unstained samples and gives about five times higher resolution than 3D-SIM while still allowing the entire nucleus to be observed. Note the high absorption contrast, revealing details of nuclear substructure like nucleoli. Infoldings of the INM were clearly detectable as racemose, tree- or mushroom-shaped clusters, mostly near the nuclear envelope (yellow box and inset). These clusters correlated with fluorescent ‘speckles’ containing pU_L34-GFP/NEC (Hagen et al., 2012). (A’) Side view of the same tomogram. Asterisks in (A) and (A’) mark a tubular invagination of both nuclear envelope membranes running through the nucleus (orange lines: sectional planes). We observed such infoldings in almost all nuclei (28 out of 31; from 28 tomograms). The tubes were often the origin of INM-infoldings. Such infoldings of the INM or of both membranes of the nuclear envelope have been observed in many eukaryotic cells, and are termed nucleoplasmic reticulum (NR) type I or type II, respectively (Malhas et al., 2011). (B) Cells grown only 15 hr on grid (instead of 2 days) had less tubes but exhibited more vesicle-like infoldings of the INM. In larger vesicles, smaller spherical vesicles were detected, exhibiting the size of ILVs (yellow box and an un-binned magnification of a similar slice in the same area: inset). (C) Tubular invaginations of the nuclear membranes, here visualized with inverse contrast in a rendered volume of a tomographic reconstruction, occurred only rarely in non-transfected control cells (asterisk), suggesting induction/enhancement of their formation by co-expression of the NEC components pU_L31 and pU_L34. (D) Tubular infoldings of the entire nuclear envelope crossed the nucleus preferentially, but not invariably, along its smallest extension (D’, Movie S4). The side view of these tubes revealed the interconnection of type I and type II NR (D’’), as shown in detail for one sub-volume of the tomographic reconstruction (yellow box in D, slice in D’’, zoom and schematic interpretation in D’’’). Scale bar is 2 μm (A-D’) and 500 nm (insets; D’’’). cyt, cytoplasm; ILV, intraluminal vesicle; INM, inner nuclear membrane; ncl, nucleolus; NEC, nuclear egress complex; NR, nucleoplasmic reticulum; nuc, nucleus; ONM, outer nuclear membrane

Figure S2

Resolution Assessment of the NEC Coat Map, Related to Figure 4 Fourier shell correlation calculated between two half-datasets of 150 particles each. The resolution of the map is estimated to be between 35 and 40 Å. At a nominal defocus of −6 μm the first node of the CTF is expected at 1/34 Å^-1.

Figure S3

Overview of Protein Constructs, Related to Figure 6 (A) Sequence motifs for full-length pU_L31 (Uniprot: G3G955) and pU_L34 (Uniprot: G3G8X8) (PrV) are illustrated. pU_L31 exhibits a N-terminal nuclear localization signal (NLS, orange), while pU_L34 contains a C-terminal trans-membrane (TM, blue) domain (LC: low complexity domain, yellow; Pro Rich: proline rich domain, red). The NEC coat produced by co-expression and in situ assembly of the pU_L31/34 heterodimer (including a C-terminal GFP-tag at pU_L34) in porcine kidney cells was visualized by 3D-SIM, cryoXM/T, CEMOVIS and cryoFIB/ET. The GFP tag did not (substantially) affect membrane interactions/curvature as concluded from the following two lines of evidence: (i) Nuclei of HSV-1 infected host cells (native full length pU_L31/pU_L34 without any fluorescent tag) exhibited similar intraluminal vesicles (and NEC coats) as the cell model with pU_L34-GFP. (ii) The GFP-cryoEM densities, that topologically face the perinuclear space, were not structured regularly and did not follow the hexameric structure of the membrane-tethering/transmembrane C-terminal part of pU_L34 (Figure 4A, Movie S7). (B) Truncated recombinant constructs (pU_L31ΔNLS and pU_L34t179) are illustrated as used for bacterially co-expression for SAXS experiments, as well as for the subsequent SAXS-EM fitting search. The soluble NEC expressed for SAXS measurements did not contain the C terminus of pUL34, and hence no GFP. (C) The amino acid sequence of pU_L34 (PrV, Uniprot: G3G8X8) is shown and highlights five regions of interest. The transmembrane domain (TM) is colored black, and three low complexity regions (presumably highly flexible), are colored brown, red, and orange. These are in addition to a proline rich region (colored blue).

Figure S4

Characteristic Layers of the NEC CryoEM Map and Curvature Modeling, Related to Figure 7 (A and B) The ‘raw’ NEC coat average map (A; Movie S7), and radially projected NEC coat map (B) are centered on a 2-fold axis, and viewed along a ‘6-2-6’ cross-section. Any given 2-fold axis is neighbored by only two 6-fold axes, and will therefore participate in only one ‘6-2-6’ plane – the direction of curvature in the present model. For each map (A and B), tangential slices at vesicle membrane distal (MD, lower, purple) and vesicle membrane proximal (MP, upper, magenta) layers are shown. These layers are separated radially by 5.13 nm (parameter ‘h’) (Figure 7). Layers have characteristic hexagonal (‘honeycomb’) patterns of protein density (Movie S9). The MD layer density at 3.5 nm resolution does not reveal clearly distinct aggregates of globular domains, but instead appears as a continuous layer of protein density. The MP layer appears to be a distinct (more finely spaced) hexagonal lattice with each presumed pU_L34 unit appearing to form a vertex of the unit cell, with apparent connections to neighbors across the 2-fold axes. These layers are interpreted as ‘hypothetical lattices’ (colored appropriately) with characteristic center-to-center (CTC) distances (or ‘spacing’) as illustrated (not to scale). Observing the slices in reciprocal space as power spectra (i.e., squared amplitudes of structure factors) allowed detection of two primary reflections corresponding to ∼10 nm (MD layer) and ∼6 nm (MP layer, indicated by arrows), giving confirmation of the uniquely differing arrangement of densities between these layers and confirming that the MP lattice is offset by 30° in relation to the MD lattice. Radial average ‘profile’ curves for MD (purple) and MP (magenta) shown rightmost allow lattice parameters to be measured accurately. The relations annotated allow arc length parameters V_MD and V_MP to be approximated, and applied to estimate the vesicle radius (Figure 7C). Measurements from reflections observed from the curved NEC coat map (A) are inherently underestimated – owing to the curvature itself causing signal from multiple radial layers to contribute to the measurement, however the ratio of these is preserved and therefore used to estimate the arc quotient (‘a’) as shown. The MD lattice is established by pU_L31 interactions (at 2-fold and 3-fold interfaces) and are used as a ruler (providing the true V_MD), and therefore used to scale V_MP^∼ as estimated from the curved NEC coat map. Calculations are shown as inset. Slices are 0.57 nm thick.