TLP permeabilization of liposomes to Ca²⁺ ions

Loss of Ca²⁺ from TLPs is an early step following engulfment at the cell surface [26]. We tested whether an interaction of virions with a membrane could lead to Ca²⁺ permeability, without contribution of potential cellular factors, using liposomes containing, in their lumen, a soluble, Ca²⁺-sensitive fluorophore (Fluo-4) and, in their bilayer, a fluorescently labelled lipid (Cy5-DOPE) for visualization, as well as biotinylated lipids for attachment to an avidin-coated coverslip. We prepared the liposomes in the absence of Ca²⁺ and used a lipid composition similar to that of mammalian-cell plasma membranes supplemented with 2.5% GD1a, a validated rotavirus attachment factor [19–23]. In the absence of Ca²⁺, we established the background fluorescence of the incorporated Fluo-4 dye. We incubated the liposomes with fluorescently labeled TLPs (Atto 565) in Ca²⁺-containing buffer, then deposited the sample onto a coated coverslip and recorded signal at appropriate wavelengths by total internal reflection fluorescence (TIRF) microscopy (Figs 2A, 2B, S1 and S2). Thus, the intensities of the diffraction-limited spots in each of the three recorded channels reported, respectively, the presence of liposomes, the presence of Ca²⁺-liganded Fluo-4 (and hence permeabilization of the liposome to Ca²⁺), and the attachment of TLPs. In a parallel control experiment, we used ionomycin, a Ca²⁺ ionophore, to permeabilize similarly prepared liposomes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Interaction with rotaviruses makes liposomes Ca²⁺ permeable.

(A) Schematic overview of the TIRF experimental setup. (B) Representative fluorescence images, showing a field of view for the control (left, liposomes only), ionophore (center, liposomes incubated with ionomycin) and TLP (right, liposomes incubated with TLPs) samples. The 640 nm channel (Cy5-DOPE, liposome signal) is in cyan and shifted 7 pixels to the right in all images to visualize colocalization; the 488 nm channel (Fluo-4, Ca²⁺ sensor signal) is in green; the 561 nm channel (Atto 565 NHS ester dye, virus signal) is in magenta and shifted 14 pixels to the right in the TLP sample. The scale bar corresponds to 5 μm. Representative examples of liposomes are boxed (1–6) and displayed beneath the images at higher magnification (scale bar = 0.5 μm) with the three channels shown next to each other. (C) Normalized frequency of the liposome intensity signal in arbitrary units (AU) from the TLP sample (experiment 1). The grey line shows the distribution for internal control liposomes in the TLP sample (liposomes that did not colocalize with a virus particle in the TLP sample) (n = 28,985). The hatched grey bars show the distribution for the liposomes that colocalized with a virus (n = 415). (D) Ca²⁺-positive liposomes (%) for the control sample (white bars), the ionophore sample (grey bars), the internal control in the TLP sample (liposomes in the TLP incubated sample that do not colocalized with TLP, hatched white bars), and TLP sample (liposomes that colocalized with TLP, hatch grey bars) from experiments 1, 2 and 3, respectively. "Ca²⁺-positive liposomes" means percent of the liposomes that had Ca²⁺ signal intensity greater than 1.5 standard deviations from the mean in the corresponding internal control sample (see red dashed lines in S2A and S2C Fig for the cutoffs). n = total number of liposomes in each condition. (E) Average percent (%) Ca²⁺-positive liposomes for the control sample (white bar), the ionophore sample (grey bar), the internal control in the TLP sample (hatched white bar), and TLP sample (hatched grey bar) from the three independent experiments (S1 and S2 Figs). Error bars are the standard error calculated from the mean of the independent experiments. Statistical analysis used a t-test between samples. * = p<0.05; ** = p<0.01; *** = p<0.001.

https://doi.org/10.1371/journal.ppat.1011750.g002

The distribution of intensities for the lipid marker (Cy5-DOPE) showed that the liposomes varied in size over an extended range but that this range was very similar for all samples and experiments (S1B Fig). The larger liposomes were more likely to have a colocalized TLP, suggesting that the particles had associated preferentially with the larger liposomes, as might be expected from their greater cross section (Fig 2C). Some control liposomes, incubated with neither ionomycin nor TLP, showed Ca²⁺ positive signal, corresponding to random "leaks", which varied somewhat in frequency from sample to sample (Figs 2B, 2D and S2). Addition and binding of TLPs yielded a markedly enhanced frequency of Ca²⁺ penetration (Fig 2D), at levels comparable to the enhancement by ionomycin (Figs 2D, 2E and S2). For the overall effect shown in Fig 2D, we calculated the mean number of Ca²⁺-positive liposomes by counting the number of particles that showed a Ca²⁺ signal with an intensity of 1.5 standard deviations above the mean intensity in the corresponding internal control calculated from the TLP-free liposomes in the same field as the TLP-associated liposomes (S1 and S2 Figs). For the data summary in Fig 2E, we averaged the results of the three independent experiments.

We also used the method just described to examine Ca²⁺ permeabilization by recoated particles (rcTLPs)—DLPs recoated in vitro with purified VP7 and VP4. These particles have physical properties and specific infectivities comparable to those of authentic TLPs as purified from infected cells [15,26,27]. As shown by the experiments summarized in Figs 3A and S3, Ca²⁺ permeability induced by rcTLPs is, like other most other properties of these particles, essentially the same as permeability induced by authentic TLPs (Fig 2E).

Ca²⁺ permeabilization requires foot insertion

We tested whether insertion of the VP4 foot into the lipid bilayer is essential for transit of Ca²⁺ into the liposome by eliciting spike reversal with a short exposure to high pH and by comparing results obtained with particles recoated with wild-type (wt) VP4 and with a mutant, "foot-locked VP4", in which a disulfide bond prevents unfolding and outward projection of the foot [18]. Brief incubation of RRV particles at pH 11 followed by return to pH 8 induces complete reversal of wild-type (wt) VP4 on both authentic and recoated TLPs [18,28]. The foot-locked mutant (mut) VP4 transitions to an intermediate structure, in which the projecting domains rearrange, but the foot remains in place [18]. Recoating with this mutant VP4 is even more efficient than with the wild-type protein, perhaps because of stabilization of the foot as the disulfide forms during the recoating step (S4 Fig), but owing to incomplete disulfide formation, the mutant-recoated particles retain some residual infectivity (S4 Fig). Classification of cryo-EM subparticles shows that about one-third of the VP4 spikes do reverse at pH 11 and that the efficient recoating has resulted in essentially complete occupancy (no "empty-position" class), rather than the 40–50% level for wild-type VP4 [18].

Fig 3B and 3C show experiments that compare Ca²⁺ permeabilization by wild-type (wt rcTLP) and disulfide-mutant (mut rcTLP) particles. The substantially more efficient recoating with mutant VP4 results in an even higher proportion of permeabilized rcTLP-associated liposomes than for wild-type VP4. Therefore, to inactivate all reversible VP4 spikes (i.e., those in which the disulfides had failed to form), we pulsed the particles at pH 11 for 30 mins and returned them to pH 8 for the liposome incubation. Fig 3B shows that this treatment inactivated wild-type VP5*/VP8* spikes, and Fig 3C shows that it also eliminated permeabilization by the mutant-recoated particles. Thus, neither the intermediate VP5*, with hydrophobic loops on the β-barrel domain well exposed but the foot still anchored in the particle, nor the fully reversed VP5* has the capacity to allow Ca²⁺ to cross the bilayer. That is, for Ca²⁺ permeabilization, the VP5*/VP8* conformational change needs to occur while the particle is bound to the membrane, probably (as indicated by the cryo-EM analysis that follows) so that coiled-coil formation can provide some of the energy needed to drive part or all of the foot into (and potentially through) the lipid bilayer.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Exposure of rotaviruses to pH 11.0 before liposome incubation prevents Ca²⁺ permeabilization.

(A) Average percent (%) Ca²⁺-positive liposomes, from at least three independent experiments (S3 and S7A Figs), for control liposomes (white bars), ionophore-exposed liposomes (grey bars), internal control in the wt rcTLP sample (liposomes in the wt rcTLP sample that did not colocalize with wt rcTLP: dotted white bars) and wt rcTLP sample (liposomes that colocalized with wt rcTLP: dotted grey bars). (B) Average % Ca²⁺-positive liposomes, from three independent experiments (S5 and S7B Figs), for the internal control and wt rcTLP sample pre-incubated at pH 8.0 (dotted white and dotted grey bars, respectively), and for the internal control and the wt rcTLP sample and wt rcTLP pre-incubated at pH 11.0 (dotted white and dotted grey bars on grey background, respectively). (C) Average % Ca²⁺-positive liposomes, from three independent experiments (S6 and S7C Figs), for the internal control and the mutant rcTLP sample pre-incubated at pH 8.0 (crossed hatched white and crossed hatched grey bars, respectively), and for the internal control and the mutant rcTLP sample pre-incubated at pH 11.0 samples (crossed hatched white and crossed hatched grey bars on grey background, respectively). "Ca²⁺-positive liposomes" designates liposomes that had a Ca²⁺ signal intensity greater than 1.5 standard deviations from the mean in the corresponding control sample (see red dashed lines in S3A, S3C, S5A, S5C, S6A and S6C Figs for the cutoffs). Error bars are the standard error calculated from the mean of the independent experiments. Statistical analysis used a t-test between samples. n.s. = p>0.05; * = p<0.05; ** = p<0.01; *** = p<0.001.

https://doi.org/10.1371/journal.ppat.1011750.g003

Despite permeabilization to Ca²⁺, the liposomes remained impermeable to Fluo-4. Thus, the interaction with rotavirus particles did not, under these conditions, disrupt liposome integrity. We can conclude that binding of an RRV TLP to a membrane containing an appropriate sialic-acid receptor permeabilizes the liposome membrane to Ca²⁺ ions, without otherwise rupturing the lipid bilayer. If passage of Ca²⁺ is through a pore-like structure, then retention of Fluo-4 sets an upper limit of ~15 Å for the pore diameter.

Cryo-EM analysis of TLPs with liposomes

We examined whether the upright-to-reversed conformational change indeed drives the foot into the membrane, as proposed in our earlier paper [18] and inferred in the previous section, by visualizing with cryo-EM the interaction of TLPs with lipid-bilayer membranes (Fig 4A). We prepared liposomes as in the experiments above, but without Fluo-4 and without the membrane label. We used recoated particles (wt rcTLPs), because of previous observations that VP4 on rcTLPs appeared to undergo more readily a spontaneous transition from upright to reversed [18]. We recorded four cryo-EM data sets on a Thermo Fisher Titan Krios operated at 300 kV and equipped with a Gatan Bioquantum energy filter and K3 detector (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Cryo-EM analysis of rotavirus in the presence of liposomes.

(A) Sample preparation scheme and representative micrograph showing rcTLP particles bound to liposomes. Scale bar: 50 nm. (B) Illustration of how template matching was used to locate virions in the micrographs and estimate their z position based on per-particle defocus fitting. (C) Histogram of estimated ice thickness of the cryo-EM samples. For each micrograph that contained at least two particles, we calculated the largest z difference and added 100 nm (approximately two times the particle radius). (D) Mapping of spike positions and their class assignment onto the micrographs of liposome-bound viruses, color-coded as shown in panel (E). The asterisks indicate a cluster of VP5* trimers around the five-fold axis of the virus. Scale bar: 50 nm. (E) Representative reconstructions of the different spike conformations after subparticle classification.

https://doi.org/10.1371/journal.ppat.1011750.g004

We found the position of viral particles on our micrographs by template matching from cisTEM [29,30]. By refining the local defocus for each particle, we obtained its relative z position within the vitreous ice (Fig 4B). Fig 4C shows a histogram of the largest z difference plus 100 nm (approximately two times the radius of a virion), for micrographs from which we picked at least two virus particles. We did not directly measure the ice thickness [31], but we estimated from the observed distribution, that the ice thickness of our samples varied between 100 and 250 nm.

Images in Fig 4D show that many particles appeared to have extended contacts with liposomes, but many of these were with liposomes of radius ≥300 nm and hence substantially flattened in the ice. As a consequence of flattening [32,33], the bilayer would have been forced to curve away from the particle, not just in the plane of the ice but in the directions normal to it, probably peeling the membrane away from the particle during blotting (S8 Fig). Moreover, as the suspension thinned, attached particles would have migrated to the periphery of the flattening liposome in order to remain in solution, distorting the liposome in the process. Thus, the invaginations seen at the more extended contacts appear to be largely within the plane of the image and of limited extent above and below it, and our images probably do not represent the full extent of contact achieved in the preparation. We could detect no evidence of liposome disruption when TLPs bind or when the interacting VP5* has transitioned from upright to reversed conformation.

An icosahedral reconstruction carried out after merging virus images from all four datasets resulted in a density map with an overall resolution of the viral shell of 2.43 Å (S9 Fig and S1 Table). As described in Methods and in S10 and S11 Figs, we then used subparticle classification (S10D and S10E Fig), starting with a total of 4,550,760 masked VP4 subparticles, to separate upright, intermediate, reversed, and empty VP5*/VP8* spikes positions (classifications #1 and #2, S12 Fig and S2 Table) and to distinguish membrane-engaged from non-engaged spikes (classifications #3 and #4, S13 Fig and S3 Table). Two relatively heterogeneous classes showed membrane-engaged reversed spikes, with apparently variable angles of spike contact (S13A Fig, classification #3). A further classification step then yielded two maps from 56,593 and 70,018 subparticles (classes 5 and 6), respectively, with relatively well-defined lipid-bilayer density (S13B Fig, classification #4). Virions that contributed subparticles to these two classes had on average a higher percentage of occupied spike positions (liposome-unbound rcTLPs = 48.9% empty VP5*/VP8* positions, liposome-bound rcTLPs = 36.9% empty VP5*/VP8* positions) (S4 Table), showing that those particles had bound preferentially to liposomes. Even within these classes, variability of curvature at positions at which the bilayer bent away from the particle blurred the membrane profile around the circumference of the contact. Local resolution estimates for classes 5 and 6 showed a well-resolved, virion-bound VP5* β-barrel trimer with its central coiled-coil domain and a poorly resolved lipid bilayer (S14 Fig).

The class 6 reconstruction has pronounced lipid bilayer density on one side of the reversed VP5* and somewhat fuzzier density on the side opposite (Fig 5). The angle between the curving bilayer and the VP5* trimer axis shows that this class represents a reversed spike near the edge of the contact. The side with a well-defined profile faces the nearest fivefold axis (Fig 5A and 5B). A fivefold axis may be close to the center of many of these membrane contacts, because the five surrounding spikes can offer a compromise between a minimally distorted membrane region and a set of membrane contacts stable enough to resist the forces during thinning of the aqueous film, as described above (see diagram in S8 Fig). We verified that many of the subparticles contributing to this class indeed came from reversed VP5* spikes related by a five-fold axes, by mapping the class assignment onto the original micrographs (Fig 4D and 4E). The coiled-coil at the C-terminal end of the VP5* β-barrel domains extends just to the inner lipid headgroup layer, and diffuse density extends beyond it into the bilayer profile, but peeling away of the bilayer leaves nearly all the engaged spikes at the periphery of the contact, and the resolution leaves uncertain the molecular features of the likely membrane-penetrating elements (Fig 5C and 5D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Cryo-EM reconstruction of membrane-bound VP5* in the reversed conformation.

(A) Reconstruction of class 6 (S13B Fig) in context of the full virus. The map was low-pass filtered at 8.0 Å resolution. The outer-shell VP7 is colored yellow; VP5* is colored red, orange, and salmon; membrane density is gray. The asterisk denotes the VP5* trimer on which the subparticle classification was focused. The direction of one of the icosahedral five-fold axis is indicated. (B) View around the five-fold axis showing multiple VP5* trimers interacting with liposome. The asterisk denotes the VP5* trimer on which the subparticle classification was focused. (C) Close-up view of the membrane-bound VP5* trimer. The map is partially cut. (D) Slab views of the reconstruction showing diffuse lipid bilayer and membrane-embedded VP5* foot domain density.

https://doi.org/10.1371/journal.ppat.1011750.g005

The reversed-spike structure (PDB-ID 6WXG) [18] can be docked with confidence into the observed subparticle density (Fig 6). The tips of the VP5* β barrel are hydrophobic loops [15]. Two of the three tips contact the liposome bilayer, while its curvature from flattening has probably eliminated a contact from the third. Mutations that reduce the hydrophobicity of these loops inhibit viral entry [15], and interpretation of tomographic reconstructions of TLPs entering at the thin edges of BSC-1 cells has suggested that these loops interact with membrane directly [18]. Our current reconstruction is consistent with this interpretation and suggests that the interaction might already be present in the proposed intermediate state, seen with the foot-locked mutant, in which the β-barrel domains have shifted into the threefold arrangement found in the reversed structure, but the foot domains have not yet begun to project outward from their positions in the pre-attachment TLP [18].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Interactions between the reversed VP5* trimer and the membrane.

(A) Placement of the rotavirus-bound VP5* trimer structure in reversed conformation (PDB-ID 6WXG) [18] into the density map of the class 6 reconstruction (Fig 5). Density is shown as gray mesh, docked viral protein subunits are shown in ribbon representation and colored yellow (VP7), red, orange and salmon (VP5*). Close-up views are shown for the hydrophobic loops at the tip of one VP5* β-barrel that interact with the proximal membrane leaflet (1), for the hydrophobic loops at the tip of a second VP5* β-barrel that interact with the proximal membrane leaflet (2), and for the trimeric VP5* coiled-coil extending into the membrane (3).

https://doi.org/10.1371/journal.ppat.1011750.g006

Cells, plasmids and constructs

MA104 cells (ATCC, type: MA104 clone 1, ATCC CRL2378.1, cat. # ATCC CRL2378.1) were grown in M199 medium (Thermo Fisher Scientific) supplemented with 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and 10% Hi-FBS (Thermo Fisher Scientific). BSC-1 cells (ATTC) were grown in DMEM (Thermo Fisher Scientific) supplemented with 10% Hi-FBS (Thermo Fisher Scientific). For VP4 expression, we cloned its full-length genomic sequence from serotype P5B [3], NCBI:txid444185) into a pFastbac expression vector. For VP7 expression, we cloned its full-length genomic sequence (G3 serotype, NCBI:txid444185) into a pFastbac expression vector and transformed E. coli DH10α cells. For bacmid generation, VP4 and VP7 vectors were transformed into DH10-Bac cells (ThremoFisher Scientific) and plated onto LB-agar plates supplemented with 50 μg/ml kanamycin, 7 μg/ml gentamycin, 10 μg/ml tetracycline, 100 μg/ml blue-gal and 40 μg/ml β-D-1-thiogalactopyranoside (IPTG).

Reagent preparation

Purification of rotavirus DLPs and TLPs–We grew MA104 cells to confluency in 850 cm² roller bottles in M199 medium supplemented with 10% FBS and 25 mM HEPES. Cells were infected with rhesus rotavirus (RRV, strain G3P5B [3]) at a multiplicity of infection (MOI) of 0.1. After incubation at 37°C for 24 h, the medium containing cells and debris was collected and frozen at -80°C for storage. To purify DLPs, we uncoated TLPs by adding EDTA pH 8.0 to a final concentration of 5 mM. After thawing, we pelleted cells and debris by centrifugation in a Beckman Coulter rotor JS 4.2 at 2,900 g for 30 min at 4°C. Supernatant was then removed and the pellet resuspended in 1 ml ice cold TNE buffer (20 mM Tris, 100 mM NaCl, 1 mM EDTA, pH 8.0) for DLP purification. We concentrated viral particles from the supernatant by ultracentrifugation in a Beckman Coulter rotor Ti-45 rotor at 224,500 g. The supernatant was then discarded and the pellet resuspended in 1–3 ml ice cold TNC and combined with the cell-debris fraction from the first centrifugation. We transferred the resulting suspension to a 15 ml conical tube and added TNC to a final volume of 4 ml. We added 4 ml of 1,1,2-trichloro-1,2,2-trifluoroethane (Freon 113) and inverted the tube 5–10 times. The suspension was centrifuged for 2 min in a Beckman Coulter rotor SX4750 at 200 g at 4°C. We removed the upper aqueous phase, transferred it to a new clean 15 ml conical tube and repeated the Freon extraction two more times. The resulting virus solution was layered on top of a 34.7–60.0% (w/v) CsCl gradient (1.26–1.45 g/ml). CsCl solutions were prepared in either TNE (DLP purification) or TNC (TLP purification). Gradients were spun at 4°C in a Beckman Coulter rotor SW 60 at 406,000 g for 2.5 h. Bands for TLPs and DLPs were harvested by side puncture. TLPs and DLPs were dialyzed against 2 L of TNC or TNE overnight. Viral particles were concentrated by pelleting at 4°C in a Beckman Coulter rotor SW 60 at 406,000 g for 1 h. Supernatants were removed with a 1 ml pipette and pellets resuspended in 100–200 μl of remaining buffer. Concentrations of DLPs were determined with a nanodrop measuring the absorption at 260 nm (extinction coefficient = 4.7028 g^-1cm^-1). TLP concentrations were measured by densitometry of VP6 bands on Coomassie stained SDS-PAGE gels against DLP standards ranging from 0.1 to 1.0 mg/ml.

Purification of recombinant VP7 –We initially obtained baculovirus carrying the genetic information of full-length rhesus rotavirus VP7 (G3 serotype) from bacmid transfected Sf9 cells (Thermo Fisher Scientific). Sf9 cells were then inoculated with baculovirus and passaged in the same cells three times, with an incubation time of 72 h for each passage. Eight flasks of 500 ml Sf9 cells with approximately 2 million cells per ml were then infected with 13 ml of passaged virus stock solution and incubated for 72 h. Cells were harvested at 4°C by centrifugation in Beckman Coulter rotor J 4.2 at 3500 g for 30 min. Supernatant was transferred to a clean 4 L beaker and benzamidine and sodium azide were added to final concentrations of 1 mM and 0.01% (w/v), respectively. 50 ml of a concanavalin A Sepharose (ConA) resin was added to the supernatant and stirred at 200 rpm overnight. Using gravity flow, the resin was packed into a column and washed with five column volumes (CV) of TNC buffer. Protein was eluted with five column volumes of TNC buffer containing 0.6 M α-methyl mannose. Using a peristaltic pump, we loaded the eluate onto 11 ml protein A resin with immobilized antibody m159 [48] (10 mg per ml of protein A resin), specific for trimeric VP7, and equilibrated in buffer A (20 mM Tris, 50 mM NaCl, 0.1 mM CaCl₂, pH 8.0). We washed bound VP7 protein with five CV of buffer, and eluted with five CV of buffer B (20 mM Tris, 50 mM NaCl, 1 mM EDTA, pH 8.0). Buffer was exchanged using a G25 PD10 column equilibrated in 0.1HNC (2 mM HEPES, 10 mM NaCl, 0.1 mM CaCl₂) spiked with 0.1 mM PMSF. Protein was divided into aliquots, which were frozen in liquid nitrogen and stored at -80°C.

Purification of recombinant wild type VP4 and S567C/A590C VP4 (“foot-locked” VP4)–We harvested baculovirus carrying the genetic information of full-length wild type VP4 (P5B[3] serotype) or S567C/A590C VP4 from bacmid transfected Sf9 cells. Sf9 cells were then inoculated with baculovirus and passaged in the same cells three times, with an incubation time of 72 h for each passage. Four flasks of 500 ml Sf9 cells grown in suspension to a density of approximately 2 million cells per ml were then infected with 13 ml of passaged virus stock solutions and incubated for 72 h. Cells were harvested at 4°C by centrifugation in Beckman Coulter rotor J 4.2 at 3,500 g for 30 min. We discarded the supernatant, resuspended the cell pellet in 100 ml lysis buffer (75 mM Tris, 100 mM NaCl, 5 mM EDTA, 7.5% (v/v) glycerol, 1 mM PMSF, 1 mg/ml aprotinin, 1 mg/ml pepstatin, 1 mg/ml leupeptin), and froze the sample in liquid nitrogen for storage at -80°C. After thawing of the cell suspension, cells were lysed on ice with a Branson 450 Digital Sonifier (Branson Ultrasonics, Brookfield, CT) with 40% amplitude, for 5 min with 50% duty cycles. After 5 min incubation on ice, the sonification procedure was repeated. We the centrifuged the suspension 4°C in a Beckman Coulter rotor Ti 45 at 104,000 g for 1 h, collected the supernatant and precipitated VP4 by adding 0.244 g ammonium sulfate per ml of protein solution. The resulting suspension was stirred in a glass flask overnight at 4°C at 200 rpm and then spin at 4°C in Beckman Coulter rotor Ti45 for 30 min at 104,000 g. We discarded the supernatant, resuspended the pellet in 60 ml of TE buffer (20 mM Tris, 1 mM EDTA, pH 8.0) with 1 mM PMSF, and transferred the solution to a Dounce homogenizer for homogenization with 15 strokes. The suspension was then spun at 4°C for 30 min in a Beckman Coulter rotor Ti 45 at 104,000 g. Supernatant was collected and diluted with TE buffer with 1 mM PMSF to a final volume of 900 ml and loaded onto a 5 ml Q Sepharose column (Cytiva) using a peristaltic pump at a flow rate of 5 ml/min. Using an HPLC purification system, the column was washed with two CV of T10NE buffer (20 mM Tris, 10 mM NaCl, 1 mM EDTA, pH 8.0), and bound protein was eluted with 20 CV using a linear gradient from T10NE to T150NE (20 mM Tris, 150 mM NaCl, 1 mM EDTA, pH 8.0) at a flow rate of 1 ml/min. We pooled VP4-containing protein fractions, concentrated them to 500 μl using a 10 kDa MWCO centricon (Millipore Sigma), and exchanged the buffer to HNE with 0.1 mM PMSF using a G25 PD10 column. Protein was distributed to 420 μg aliquots and frozen in liquid nitrogen for storage at -80°C.

Liposome preparation– 2 mg in total lipids of a mixture containing cholesterol, phosphocholine egg extract (egg PC), sphingomyelin (SM), 1,2-dioleyl-sn-glycero-3-phopshoethenolamine (DOPE), 1-palmitoyl-2-oleyo-sn-glycerol-3-phosphoetanolamine (POPE), 1,2-dioleyl-sn-glycero-3-phospho-L-serine (DOPS) and ganglioside GD1a (GD1a) in molar ratio of 50:12.5:10:8.5:8.5:8:2.5, dissolved in approximately 9:1 chloroform:methanol, were dried in a round bottom glass tube under argon flow. Remaining solvent was completely removed by incubating the lipids under high vacuum overnight. The lipids were resuspended in 200 μl of TNC buffer by vortexing for two minutes, in a volume corresponding to 10 mg/ml. We then freeze-thawed the lipid suspension twice using liquid nitrogen and formed liposomes by extrusion though a 0.2 μm filter using 41 passages. We evaluated the size of liposomes by dynamic light scattering using a 1:100 dilution of the liposome stock.

Recoating of DLPs with recombinant VP7 and VP4 –We distributed 45 μg of DLPs in HNE equally among five 1.5 ml conical tubes (2.25 μl per tube). We first added 1 M sodium acetate, pH 5.2 to a final concentration of 100 mM and then added 65 μl VP4 (stored at 1.3 mg/ml) to a final concentration of 0.9 mg/ml in the final reaction volume, resulting in a 33-fold excess of VP4 monomer over a total of 180 sites on DLPs. A 0.1 mg/ml aprotinin solution was added to the samples to a final concentration of 0.2 μg/ml followed by incubation at 37°C for 1 h. Required amounts of VP7 (9 μl stored at 1.0 mg/ml in HNE) to achieve a 2.3-fold excess of VP7 monomer over a total of 780 sites on DLPs were premixed with 0.1 volumes of TC buffer (20 mM Tris, 10 mM CaCl₂, pH 8.0) and 0.1 volumes of 1 M sodium acetate pH 5.2 for 15 min before adding them to the DLP-VP4 mixture. Samples were incubated for 1 h at room temperature and then quenched by the addition of 0.1 volumes of 1 M Tris pH 8.0. Recoated particles from the five tubes were combined, and TNC was added to a final volume of 2.5 ml. rcTLPs were separated from excess VP4 and VP7 by ultracentrifugation at 4°C in a Beckman Coulter rotor TLS 55 at 215,000 g for 1 h. We removed 2.0 ml of the supernatant, returned the volume to 2.5 ml with TNC, and pelleted again. Supernatant was carefully removed so that 100 to 200 μl remained. The rcTLP pellets were resuspended in the remaining buffer, sodium azide added to a final concentration of 0.01%, and the rcTLPs stored at 4°C.

Ca²⁺ permeabilization experiment and TIRF microscopy

Labelling of TLPs and rcTLPs–TLPs or rcTLPs (mutant or wt) were diluted to 1.0 mg/ml in a total volume of 30 μl using HNC (20 mM HEPES pH 8.0, 100 mM NaCl, 1 mM CaCl₂) and 3.33 μl of 1 M NaHCO₃ pH 8.3 added. This solution was mixed with 0.5 μl of 0.0076 mg/ml Atto 565 NHS ester. The reaction proceeded at room temperature for 1 h before quenching with 3 μl of 1 M Tris pH 8.0. The labeled TLPs were then buffer exchanged into a solution containing 20 mM Tris pH 8.0, 100 mM NaCl, and 1 mM CaCl₂ using a Zeba Spin Desalting Column (Thermo Fisher Scientific).

pH shift experiments with rcTLPs–rcTLPs were incubated with 5 μg/ml trypsin for 30 min at 37°C and the reaction quenched by addition of 1 mg/ml aprotinin. The high-pH induced conformational change was achieved by adding 0.7 M N-cyclohexyl-3-aminopropanesulfonic acid (CAPS) buffer at pH 11.0 to a final concentration of 54 mM. Samples were incubated for 30 min at room temperature and neutralized to pH 8.0 by addition of diluted HCl. As a control, we used 20 mM Tris-HCl, 54 mM CAPS pH 8.0 to match the osmolarity of both buffers.

Liposome preparation–A chloroform solution containing 2 mg total lipid mixture of cholesterol, phosphocholine egg extract (egg PC), sphingomyelin (SM), 1,2-dioleyl-sn-glycero-3-phopshoethenolamine (DOPE), biotinylated-DOPE, Cy5-DOPE, 1-palmitoyl-2-oleyo-sn-glycerol-3-phosphoetanolamine (POPE), 1,2-dioleyl-sn-glycero-3-phospho-L-serine (DOPS) and ganglioside GD1a (GD1a) in molar ratio of (40:22.5:10:7.25:0.5:0.25:8.5:8:2.5) was dried in a round bottom glass tube under argon flow. Remaining solvent was removed by incubating the lipids under high vacuum overnight. We then resuspended the lipids in 200 μl of TNE (20 mM Tris pH 8.0, 100 mM NaCl, 0.07 mM EGTA) containing 400 μM Fluo-4 dye by vortexing for two minutes, forming a 10 mg/ml lipid suspension, which we subjected to two freeze-thaw cycles with liquid nitrogen. For the pH shift experiments, the lipids were resuspended in 200 μl of TNE (20 mM Tris-HCl, 100 mM NaCl, 0.07 mM EGTA), 54 mM CAPS, pH 8.0, containing 400 μM Fluo-4 dye, so that the buffer had the same osmolarity that the final preparation of rcTLPs. By including 0.07 mM EGTA and then exchanging the liposomes into a buffer containing 1 mM CaCl₂ before incubating them with TLPs (see below), we ensured that there was no free Ca²⁺ inside the liposomes. Liposomes were formed by extrusion with 41 passages through a 200 nm pore filter. The size of the liposomes was evaluated with dynamic light scattering using a 1:100 dilution of the liposome stock. The liposome suspension was loaded onto a G25 PD10 column equilibrated with TNC (or TNC, 54 mM CAPS pH 8.0 in the case of pH shift experiments) to separate liposomes from non-incorporated fluorophore and for buffer exchange. We collected 500 μl fractions and used peak fractions containing liposomes for further experiments.

Liposome Ca²⁺ permeabilization assay– 4 μl of liposomes were added to 12 μg virus particles in a final volume of 18 μl of TNC (or TNC, 54 mM CAPS pH 8.0) and incubated for 30 min at 37°C with 5 μg/ml trypsin (to activate VP4). Trypsin was inactivated by adding aprotinin to a final concentration of 1 mg/ml. All samples were incubated for an additional 1 h at 37°C. For controls, same amount of liposomes was incubated for the same time and temperature in the absence of virus (negative control) or in the presence of 20 μM ionomycin (positive control).

TIRF measurements– 25 mm circular glass coverslips were cleaned and coated with a 10% (w/v) solution of biotinylated-polyethylene glycol (PEG):PEG in a ratio 1:99 (Laysan Bio, cat. no. mPEG-SCM-5000) [49] and pre-incubated with 0.5 mg/ml neutrAvidin (Thermo Fisher Scientific, cat. no. LF144746) for 15 min at room temperature (RT) before use. Liposome solutions were diluted 500 times in TNC in order to avoid stochastic co-localizations of virus and liposomes, incubated over the coverslip for 10 min at RT, and unbound liposomes were removed before imaging by a single wash with 100 μl TNC. Images of different fields of view were acquired under TIRF illumination using an Olympus IX70 microscope equipped with a Hamamatsu ImageEM camera and excitation from 488, 561 and 640 nm lasers (Coherent, Inc., Salem, NH) operated at 100, 150 and 70 mW, respectively (data from Figs 2, S1 and S2). For each field of view, we recorded three images with 50% power from each laser and 30 ms exposure at 488 nm (Fluo-4 dye, Ca²⁺ sensor signal), 50 ms exposure at 561 nm (Atto 565 NHS ester dye, virus signal), 20 ms exposure at 640 nm (Cy5-DOPE dye, liposome signal). A Zeiss 200M microscope equipped with a Zeiss TIRF slider, a Photometrics Cascade camera and 488, 561 and 640 nm excitation lasers (Coherent, Cobalt and Coherent, Inc., respectively) operated at 100 mW, 50 mW and 40 mW, respectively, was also used for acquiring additional images (data from Figs 3, S3, S5, S6 and S7). The microscopes were controlled by the SlideBook acquisition program (Intelligent Imaging Innovations, Denver, CO), and image processing was carried out using Fiji software [50].

Infectivity assay of TLPs and rcTLPs–Particles to focus-forming unit (FFU) ratios for TLP and rcTLPs were determined by infectious focus assays as previously described [27], and specific infectivities were calculated from concentration measurements on the basis of densitometry of a coomassie-blue stained SDS-PAGE gel. BSC-1 cells were washed with Gibco M199 media (Thermo Fisher Scientific), inoculated with serial dilutions of virus in M199 containing 1 μg/ml trypsin and incubated for 1 h at 37°C. The inoculum was removed and replaced with M199 containing 10% fetal bovine serum and 2.5 μg/ml neutralizing monoclonal antibody M159 to inhibit secondary infection. At 16 h postinfection, cells were washed with phosphate-buffered saline and fixed with methanol. Infectious foci detected by immunoperoxidase staining were counted. Monoclonal antibody M60, which recognizes VP7, was used as the primary detection antibody.

Image analysis–We used previously described custom-made MATLAB (MathWorks) scripts [51] for automatic detection of fluorescent liposomes in the 640 nm channel (Cy5-DOPE dye, liposome signal) images by fitting a 2D Gaussian function to diffraction limited spots for which the intensity was 1.5 above background. We extracted the XY position of each liposome in all images and calculated the integrated fluorescence intensity on a 4x4 pixel area around the XY position for the 3 channels. We calculated the background by measuring the integrated fluorescence intensity on a 6x6 pixel region around the XY position, which was then was subtracted from the integrated intensity on the 4x4 area after area correction for the 3 different channels. The integrated fluorescence intensity after background correction is shown in the figures as “Ca²⁺ signal” and “liposome signal” (arbitrary units, AU). For the samples that were incubated with TLPs and rcTLPs, the liposomes that colocalized with virus were selected by applying a threshold to the integrated fluorescence intensity in the 561 channel (Atto 565 NHS ester dye, virus signal) at the specific XY position of the liposomes; the threshold was calculated from images, taken at the same time, that contained only virus.

Cryo-EM specimen preparation and data collection

Incubation of rcTLPs with liposomes–To rcTLP stock solutions (0.7–1.5 mg/ml), we added liposome stock solutions (10 mg/ml, average diameter 200 nm) to a final concentration of 1.5 mg/ml, followed by the addition of trypsin to a final concentration of 5 μg/ml and incubation for 30 min at 37°C. Trypsin was then inactivated with aprotinin at a final concentration of 1 mg/ml. Samples were incubated at 37°C for an additional 30 min.

Cryo-EM sample preparation–C-flat holey copper carbon grids (CF-2/1-2C, Electron Microscopy Sciences) were glow discharged for 30 s at 10 mA. Using a CP3 cryo-plunger, 4 μl of virus-liposome samples were applied to the grids, blotted for 4 s with at a humidity ranging from 88% to 92% and plunged into liquid ethane having a temperature of -273°C. Grids were stored in liquid nitrogen.

Cryo-EM data collection–We recorded movies with a Titan Krios G3i microscope (Thermo Fisher Scientific, Waltham, MA), operated at 300 kV and equipped with a BioQuantum energy filter and a K3 direct electron detector camera (Gatan, Inc.). Dose-fractionated movies with 50 frames were recorded in counting mode with 0.05 s per frame for a total of 2.5 s with SerialEM v3.7 [52]. The magnification was 60,606, resulting in a pixel size of 0.825 Å per pixel. For some grids, we recorded a total of 27 movies for each stage position (nanoprobe mode with an illuminated area of ~1 μm, 3 exposures per hole from a total of 9 holes).

Cryo-EM data processing

Preprocessing and initial particle stack preparation–We used MotionCor2 [53] with 5×5 patch alignment to obtain summed micrographs from the frames of 33,563 movies that we had recorded from four sample preparations. With CTFFIND4 [54], we estimated contrast transfer function (CTF) parameters from the summed micrographs. We then picked 77,139 rotavirus particles from the micrographs template matching in cisTEM [29,30], with the routines match_template (version 1.00), refine_template (version 1.00), and make_template_result (version 1.00). We carried out template matching at 5 Å resolution with angular search steps of 1.7° and a previously determined reconstruction as reference structure [18], resampled to the correct pixel size. The symmetry for the search was I₂ and the radius of the mask 403 Å. We determined CTF parameter values at particle positions with Gctf [55] and used relion_preprocess [56] for particle extraction and normalization. The box size was 1536×1536 pixels and the radius for background area designation 666 pixels. After 2D classification with relion_refine [56], we kept 75,846 particle images for analysis.

Icosahedral reconstruction–An initial reconstruction calculated with the alignment parameters obtained from template matching had a nominal resolution of 3.89 Å as determined by Fourier shell correlation (FSC) with a cutoff value of 0.143 and after masking the half maps with a spherical shell mask that had inner and outer radii of 222 and 403 Å, respectively (S9 Fig). We then improved particle alignment, particle images, and reconstruction parameters by iterating routines in cisTEM (refine3d version 1.01, reconstruct3d version 1.02) [29] with routines in RELION (relion_motion_refine, and relion_ctf_refine) [57]. The resolution for alignment in the last iteration was 3.4 Å. In relion_ctf_refine, we used per-particle defocus fitting, beam-tilt refinement, symmetric aberration estimation, and fitting of anisotropic magnification distortion for each of the particle stack’s 63 optics groups (—fit_defocus–kmin_defocus 30 –fit_mode fpmfp–fit_beamtilt–kmin_tilt 30 –fit_aniso–odd_aberr_max_n 3 –fit_aberr). Reconstuctions were calculated with relion_reconstruct with the option–fom_weighting, where we assigned a per particle figure of merit, FOM, based on the cisTEM alignment scores, where S_i is the score of particle I, and S_max and S_min are the highest and lowest scores, respectively:

After two iterations, the nominal resolution of the final icosahedral reconstruction of full viral particles converged to 2.43 Å as estimated with relion_postprocess (S9 Fig and S1 Table).

Local analysis of spike positions–For local analysis of individual spike positions, we used protocols similar to those previously described [10,18,37], with some modifications (S10 Fig). Generally the workflow for preparing subparticle stacks for local analysis involved symmetry expansion of the RELION particle star file, followed by signal subtraction with relion_project using a suitable mask to retain density, and sub-particle extraction from the signal-subtracted images with custom Python scripts and IMOD routines [58]. Depending on the analysis, different subparticle stacks were prepared.

For initial classification of spike positions—that is, determining for each of the 60 positions on the virion whether the VP5*/VP8* penetration protein was in an upright, intermediate or reversed conformation, or whether it was entirely absent—we made a 384×384 pixel subparticle stack 1 in which we had signal-subtracted all density except within the volume of a single VP7 trimer plus the volume potentially occupied by VP5*/VP8* in upright and reversed conformations (S10A–S10D Fig). We carried out classifications with refine3d and reconstruct3d from cisTEM [29], keeping the subparticle alignment angles and shifts constant during all iterations. We also turned off per-particle weighting in the reconstruction step by setting the particle weighting factor input parameter of reconstruct3d to 0.0 and by resetting the LogP, SIGMA, and SCORE values in par files to constant values of -2000, 4.1, and 20.1, respectively, and running the reconstruction step again before proceeding to the next iteration. Initial classes were seeded randomly. In a first classification (classification #1), we asked for 16 classes (S2 Table). The resolution limit for classification was 5 Å and the radius of the spherical mask applied to the 2D images was 140 Å. 3D references were masked around the single VP7 trimer and VP5*/VP8* volume. Inspection of the resulting classes from the first classification showed that particles partitioned not only based on spike occupancy and conformation, but also based on radial shifts (up to approximately 5 Å) with respect to the virion center (S1 Movie). As evident from S11 Fig, the main cause for the observed radial shifts appeared to be anisotropic magnification distortion, because the class assignment of subparticles from the two classes with the largest relative shift observed (classes 4 and 13) showed a systematic bias depending on the extraction angle with respect to the virion center. We corrected for anisotropic magnification when we calculated icosahedral reconstructions of full viral particles, but no correction was implemented during subparticle extraction. Other imperfections probably also contributed to the observed shifts; for instance, if data had been collected slightly away from the eucentric height of the microscope. We therefore further classified each of the 16 classes in a second classification (classification #2), aligned all 32 classes by rigid-body fitting a model of the VP7 trimer and updating the alignment parameters in the RELION star file based on the superpositions of the fitted models, and made a final assignment of each subparticle as upright, intermediate, reversed, or unoccupied (S10E, S10F and S12 Figs and S2 Table).

Local analysis of reversed VP5*/VP8* spikes bound to liposomes–Because we observed minor density artifacts in reconstructions calculated from subparticle stack 1 (see above) when contoured at very low level (presumably caused by a bug or array overflow in relion_project when the volume for signal subtraction had a large size of 1536×1536×1536 pixels), we prepared subparticle stack 2 with slightly down sampled data of 1024×1024×1024 pixels for the full virus particle box, corresponding to a pixel size of 1.2375 Å. For this stack, we signal-subtracted all density within a tight-fitting spherical mask except within the volume occupied by VP5*/VP8* at the reference icosahedral position (S10C and S10D Fig). Classifications #3 and #4 were carried out starting with 1,510,279 particles in reversed conformation selected from the signal-subtracted subparticle stack 2 based on the results from classification #1 and #2 (S10G Fig). 2D and 3D masking was with a radius of 124 Å, and the resolution for classification was 5 and 8 Å, respectively. Subparticles from classes 1 and 3 of classification #3 that showed liposome density were merged and subjected to classification #4 (S13 Fig and S3 Table). Final reconstructions of liposome-bound classes were computed from the original (no signal subtraction) subparticle stack 2 (S10H Fig).

Mapping of subparticles in the original micrographs–When we extracted subparticles, we assigned unique viral particle and subparticle identifiers together with the x and y coordinates of the extraction position of each subparticle with respect to the original extraction position of the corresponding virus in the micrograph. We could thus map subparticles from liposome-bound classes onto the original micrographs (Fig 4D).

Figure preparation

We prepared the figures using PyMOL (The PyMOL Molecular Graphics System, Version 2.3 Schrödinger, LLC), ChimeraX [59], BioRender, matplotlib [60] and ImageMagick (ImageMagick Studio LLC, 2023, available at: https://imagemagick.org).