Importance of Vp1 Calcium-Binding Residues in Assembly, Cell Entry, and Nuclear Entry of Simian Virus 40

Abstract

For polyomaviruses, calcium ions are known to be essential for virion integrity and for the assembly of capsid structures. To define the role of calcium ions in the life cycle of the virus, we analyzed simian virus 40 (SV40) mutants in which structurally deduced calcium-binding amino acids of Vp1 were mutated singly and in combination. Our study provides evidence that calcium ions mediate not only virion assembly but also the initial infection processes of cell entry and nuclear entry. Mutations at Glu48, Glu157, Glu160, Glu216, and/or Glu330 are correlated with different extents of packaging defects. The low packaging ability of mutant E216R suggests the need to position the Glu216 side chain for proper virion formation. All other mutants selected for further analysis produced virus-like particles (VLPs) but were poorly infectious. The VLPs of mutant E330K could not attach to or enter the cell, and mutant E157A-E160A and E216K VLPs entered the cell but failed to enter the nucleus, apparently as a result of premature VLP dissociation. Our results show that five of the seven acidic side chains at the two calcium-binding sites—Glu48 and Glu330 (site 1), Glu157 and Glu160 (site 2), and Glu216 (both sites)—are important for SV40 infection. We propose that calcium coordination imparts not only stability but also structural flexibility to the virion, allowing the acquisition or loss of the ion at the two sites to control virion formation in the nucleus, as well as virion structural alterations at the cell surface and in the cytoplasm early during infection.

The capsid of simian virus 40 (SV40), like those of other small DNA viruses in the polyomavirus family, is composed of 72 pentamers of the major capsid protein Vp1 arranged on a T=7d icosahedral lattice (2, 19, 26). The architecture of the SV40 capsid, resolved at atomic resolution (19), provides clues to aspects of capsid assembly and stability. Each Vp1 monomer comprises a core β-barrel structure with a jelly-roll topology, an amino-terminal extension, and a long carboxy-terminal arm. The five monomers in a pentamer are intimately associated via interlocking secondary structures. Interaction between pentamers (in six different modes, α, α′, α", β, β′, and γ) is made through the insertion of carboxy-terminal arms into the cores of neighboring pentamers. Several lines of evidence have suggested that this interpentamer interaction is strengthened by calcium ion chelation and disulfide bonding (4, 6, 19, 32). Structural refinement on SV40, in which divalent calcium ions (Ca²⁺) were replaced with trivalent gadolinium ions (Gd³⁺), has identified two probable sites of calcium ion coordination per Vp1 monomer on the capsid (32) (Fig. 1). Site 1 consists of the Glu216 side chain and Ser213 carbonyl oxygen of one monomer, the Glu46 and Glu48 side chains of a second monomer from the same pentamer, and the Glu330 side chain (C-terminal arm) of a third monomer from a neighboring pentamer. Site 2 consists of the Glu157, Glu160, and Glu216 side chains and Lys214 carbonyl oxygen of the first monomer and the Asp345 side chain (C-terminal arm) of the third monomer. Thus, each pair of calcium ions is expected to tie together two different pentamers by interacting with mostly acidic amino acid residues contributed by three Vp1 chains. All but one (Glu46) of the seven acidic residues are conserved in the polyomavirus family (25). It is suspected that besides contributing to capsid integrity, calcium ion-mediated interactions play a role in various processes of virus dissemination, including cell entry, intracellular trafficking, disassembly, and uncoating, as well as in capsid assembly. We have previously shown that mutating Cys254, a residue located on a conserved loop near the calcium sites, can severely reduce the formation of infectious particles. That this defect is observed for substitution into certain side chains (leucine, glycine, serine, and arginine) but not into alanine is consistent with the local structure of Cys254, including the putative calcium-binding residues, being important for particle assembly and infectivity (17). Mutating certain acidic residues of the calcium-binding sites also affects the shape or stability of virus-like particles (VLPs) formed in Vp1-expressing insect cells (12).

FIG. 1.

Calcium-binding side chains of Vp1. At left is a ribbon diagram of an SV40 Vp1 pentamer, with one of the Vp1 monomers highlighted in white. An invading arm from a neighboring pentamer is shown in pink. Two blue circles represent the twin Ca²⁺ sites. At right, an enlarged view of the Ca²⁺-binding sites shows the coordination of the two Ca²⁺ ions by glutamate and aspartate side chains from three Vp1 monomers on two different pentamers. Glu157, Glu160, and Glu216 are from the highlighted monomer. Glu46 and Glu48 (E46 and E48; indicated by *) are from a neighboring monomer in the same pentamer. Glu329, Glu330, and Asp345 (E329, E330, and D345; indicated by **) are from a second pentamer, and the helix αC** is from a third pentamer. The left-hand image is adapted from reference 32 with permission of the publisher.

In this study, we tested whether the seven putative calcium-binding acidic residues play a role in the viral life cycle. We report that five acidic side chains—Glu48 and Glu330 of site 1, Glu157 and Glu160 of site 2, and Glu216, shared by both sites—contribute to the formation of infectious virions. Our analysis included mutation of the residues into lysine and arginine as well as into alanine. We expected that the substituted basic side chains would either perturb the local structure of the sites and disrupt assembly or lead to the formation of poorly infectious particles with new salt bridges in place of calcium atoms. Both categories of defects were observed. Mutant E216R was ineffective in packaging into VLPs. Mutant E330K assembled VLPs that cannot bind or enter the cell. Mutants E216K and E157A-E160A made VLPs that appear to prematurely dissociate following cell entry, thus failing to target the viral DNA to the nucleus. Our findings support the idea that calcium coordination at both of the calcium sites is important not only for virion assembly but also for initiation of infection at the stages of cell entry and nuclear entry of the viral genome. A model for calcium ion-dependent, sequential structural alterations that are involved in the entry of the virion into the cell and into the cell nucleus is presented.

MATERIALS AND METHODS

Plasmid construction.

Oligonucleotides for PCR, for linkers, and for sequencing were synthesized by Genosys (The Woodlands, Tex.). All mutations created were confirmed by dideoxynucleotide sequencing analysis. DNA nucleotide sequences are given in capital letters except for mutated SV40 nucleotides, which are highlighted in lowercase. Relevant restriction sites are underlined.

Mutagenesis of the Vp1 calcium-binding acidic residues was performed within pBS-based plasmids. pBS-Vp1-E46A, pBS-Vp1-E48A, and pBS-Vp1-E46A-E48A were made by inserting into pBS-Vp1 (17) an AccI-to-AflII linker in which the Glu46 codon, the Glu48 codon, or both codons, respectively, were replaced with gct. pBS-Vp1-E157A, pBS-Vp1-E160A, and pBS-Vp1-E157A-E160A were constructed by inserting into pBS-Vp1 the XbaI-to-PstI PCR fragment generated using pBS-Vp1 as a template, the sense primer 5′-GGAGTAGCTCTAGAATGAAGATG-3′, and the antisense primer 5′-AACACACCCTGCAGwxCCAAAGGyzCCCCACCAACAGCAAAAAATG-3′, where wx and yz are CT and ag (E157A), ag and TT (E160A), or ag and ag (E157A- E160A), respectively. pBS-Vp1-E216A, pBS-Vp1-E216K, and pBS-Vp1-E216R were made by inserting into pBS-Vp1 the BstBI-to-MluI PCR fragment generated using the antisense primer 5′-TGCCCATCCACGCGTTGTGTtCTgCgGTTAATcAGGTCACTTAACAAAAAGGA-3′ and the sense primer 5′-GGGTTCCTGATCCTTCGAAAAATxyzAACACTAGATATTTTGGAACCTACACAG-3′, where xyz represents Gcg (Ala), aAg (Lys), or cgc (Arg), respectively. pBS-Vp1-E329A, pBS-Vp1-E330A, pBS-Vp1-E330K, pBS-Vp1-E330R, pBS-Vp1-E329A-E330A, pBS-Vp1-E329A-E330K, and pBS-Vp1-E329A-E330R were made by inserting into pBS-WT (modified pBluescriptII containing the whole wild-type Vp1 coding region) (12) respective PstI-to-XbaI fragments, which were generated via two consecutive rounds of PCR using pBS-Vp1 as the initial PCR template. The first round of PCR used the following sense and antisense primer pairs: 5′-CCTCTCAAGTAGcGGAGGTTAGGGTTTATGAGGACACAG-3′ and 5′-CTAACCTTACAGGAGAGTTCATCgCCTCCAATCC-3′ (E329A); 5′-CCTCTCAAGTAGAGuvwGTTAGGGTTTATG-3′ and 5′-CATAAACCCTAACxyzCTCTACTTGAGAGG-3′ (E330A, E330K, and E330R), where uvw and xyz represent Gct and agC, aAG and CTt, or aga and tct, respectively; and 5′-CCTCTCAAGTAGcGuvwGTTAGGGTTTATGAGGACACAG-3′ and 5′-CCTAACxyzCgCTACTTGAGAGGACATTCCAATC-3′ (E329A-E330A, E329A-E330K, and E329A-E330R), where uvw and xyz represent the same codons as for the corresponding E330 single mutants. The second round of PCRs used the primer pair E160Q-Sense (12) and WT3′-XbaI (5′-CCGGTCTAGATCACTGCATTCTAGTTGTGGTTTG-3′). pBS-D345A, pBS-D345K, pBS-D345R, and pBS-D345N were made by inserting into pBluescriptII the BamHI-to-PstI PCR fragment generated using pUCVP1 (15) as a template, the anitsense primer M13 Reverse, and the sense primers 5′-CTGGGGATCCAgccATGATAAGATAC-3′ (D345A), 5′-CTGGGGATCCAaagATGATAAGATAC-3′ (D345K), 5′-CTGGGGATCCAcgcATGATAAGATAC-3′ (D345R), and D345N-Sense (12), respectively.

To make nonoverlapping SV40 plasmids (NO-pSV40) containing mutations in Glu46, Glu48, Glu157, Glu160, Glu216, Glue329, and Glu330, suitable restriction fragments of the Vp1 coding region from respective mutant pBS-Vp1s were inserted into NO-pSV40-BSM. To make NO-pSV40-D345A, NO-pSV40-D345K, NO-pSV40-D345R, or NO-pSV40-D345N, pBS-D345A, pBS-D345K, pBS-D345R, or pBS-D345N was sequentially subjected to insertion with the following: the PstI-to-KpnI fragment of NO-pSV40-BSM containing the SV40 ori and the N-terminal coding region for the large T antigen, the KpnI-to-EcoRI fragment of NO-pSV40-BSM encoding Vp2/3 and the N-terminal portion of Vp1, and the EcoRI-to-EcoRI fragment of pBS-WT encoding the rest of Vp1. NO-pSV40-E157Q-E160Q-D345N and NO-pSV40-E329Q-E330Q-D345N were made by inserting the EcoRI-to-EcoRI Vp1-encoding fragments from pBS-mtE and pBS-mtF (12), respectively, into NO-SV40-D345N. All NO-SV40 viral genomes were prepared from their respective NO-pSV40 plasmids by digestion with BamHI and recircularization with T4 DNA ligase as described previously (11).

Antibodies.

Polyclonal rabbit anti-Vp1 serum (14) and affinity-purified polyclonal rabbit anti-Vp3 immunoglobulin G (IgG) (21) have been described previously. Monoclonal mouse anti-importin-α (anti-Rch-1) and anti-importin-β (anti-karyopherin-β) antibodies were obtained from Transduction Laboratories. Protein A-Sepharose beads were obtained from Amersham-Pharmacia.

Assays for viability, replication, and Vp1 production.

Viability was determined by plaque formation assays using serial dilutions of lysates prepared by freeze-thawing NO-SV40 DNA-transfected CV-1 monkey kidney cells at 72 h posttransfection, as described previously (17, 18). For select NO-SV40 mutants, plaques were lifted from the culture dish, resuspended in 0.1 to 1.0 ml of TD buffer, and lysed by freeze-thawing. A 5-μl aliquot of the plaque lysate was incubated with 5 μl of trypsin-EDTA (Gibco-BRL) at 37°C for 15 min and at 99.9°C for 10 min to release viral DNA from virions in the lysate. A 1-μl volume of the trypsin digest was serially diluted over a 100-fold range with 10 mM Tris-HCl (pH 8.0), and 1 μl of each dilution was subjected to semiquantitative PCR (22) followed by agarose gel electrophoresis and ethidium bromide staining to determine the concentration of the viral DNA. The amount of DNA in individual plaques ranged from 200 to 400 pg. Aliquots of the original plaque lysates containing equal amounts of viral DNA were then serially diluted and reassayed for plaque formation. Throughout this assay, mutants such as E157A were found to retain the characteristic plaque number and size relative to the wild type.

The extent of viral DNA replication was determined by gel quantitation of DpnI-resistant viral DNA extracted from 1 × 10⁶ transfected cells, and Vp1 production was assessed by anti-Vp1 Western blot analysis of 5 × 10⁴ transfected cells, as described previously (17, 18).

DNase I treatment and packaging assay.

The methods used for the DNase I treatment and packaging assay are similar to those described previously (17), with minor modifications. CV-1 cells harvested at 72 h after transfection with each NO-SV40 DNA were lysed by freeze-thawing in serum-free culture medium or by sonication in hypotonic buffer at a concentration of 10⁷ cells per ml. For DNase I treatment, an aliquot of the lysate was cleared of cellular debris and digested with 500 U of DNase I per ml at 37°C for 30 min. To determine the extent of viral DNA packaging, 20 μl of transfected lysate (corresponding to 2 × 10⁵ cells) was treated with DNase I and the DNase I-resistant viral DNA was extracted, digested with BamHI, and quantitated by Southern blot analysis followed by phosphorimaging. The value obtained was expressed relative to that derived from a non-DNase I-treated but otherwise identically processed lysate aliquot, which was taken to be 100%.

VLP analysis and preparation by sucrose sedimentation.

Sonicated lysates prepared from 5 × 10⁶ to 2 × 10⁷ transfected cells were treated with DNase I and fractionated in a 5 to 32% continuous sucrose gradient, as described previously (17). For each of the 17 fractions, half was subjected to Southern blot analysis of viral DNA content and 1/50 was subjected to Western blot analysis of Vp1 content. For infection assays, the two or three peak VLP fractions (based on viral DNA content, usually between fractions 4 and 8 from the bottom) were pooled and used as a source of VLPs. The VLP concentration was determined by extracting viral DNA from an aliquot of these preparations and analyzing it by Southern blot analysis alongside known amounts of viral DNA.

For electron microscopy analysis, 5 ml of sonicated lysate, prepared from 4 × 10⁸ cells transfected with NO-SV40-E330K, was centrifuged at 10,000 × g at 4°C for 10 min. The resulting supernatant was pelleted through a 20% sucrose cushion in 10 mM HEPES (pH 7.5) at 35,000 rpm at 4°C in an SW50 rotor. The pellet was resuspended in 0.2 ml of fetal bovine serum supplemented with a protease inhibitor cocktail (1 μg of aprotinin/ml, 1 μg of leupeptin/ml, 10 μg of pepstatin/ml, 1 mM phenylmethylsulfonyl fluoride), further sedimented through a 5 to 32% continuous sucrose gradient as above, and separated into 30 fractions. The peak VLP fractions, determined from anti-Vp1 Western blot analysis of 2 μl from each fraction alongside known amounts of Vp1, were pooled and pelleted through a 20% sucrose cushion again. The pellet was resuspended in 0.1 ml of 10 mM HEPES (pH 7.5)-10 mM NaCl-protease inhibitor cocktail. The VLP concentration in this purified preparation was again quantitated by anti-Vp1 Western blot analysis.

Cell entry assay.

The cell entry assay method was similar to that described previously (22), with minor modifications. TC7 monkey kidney cells on 100-mm dishes were infected with approximately 1,000 VLPs per cell, incubated at 4°C for 1 h and then at 37°C for 4 h, and harvested by either scraping or trypsin treatment. An aliquot of harvested cells was extracted for viral DNA, which was digested with BamHI and quantitated by Southern blot analysis followed by phosphorimaging. Another aliquot of the cells was assayed for Vp1 by anti-Vp1 Western blot analysis.

Analysis of internalized VLPs by immunoprecipitation.

The immunoprecipitation procedure was carried out essentially as described previously (22). Briefly, TC7 cells were infected with approximately 1,000 VLPs per cell, incubated at 4°C for 1 h and then at 37°C for 6 h, and harvested by trypsin treatment. An aliquot of the cytoplasmic fraction, prepared as the supernatant from the homogenization followed by low-speed centrifugation of 5 × 10⁵ harvested cells, was reacted with each specified antibody, and the immune complexes were collected by reaction with protein A beads alone or with protein A beads bound with rabbit anti-mouse IgG. After extensive washing, the immunoprecipitates were mixed with NO-pSV40ΔNcoI control DNA before the collective DNA was extracted, amplified by semiquantitative PCR, and quantitated by Southern blot analysis.

Electron microscopy.

A 10-μl volume of purified virions (4 × 10¹⁰ particles/ml) or VLP preparation (2 × 10¹⁰ particles/ml) was allowed to adhere for 60 s on carbon-coated copper grids that had been freshly glow-discharged (to make the grid hydrophilic), and excess sample was removed by touching it with the edge of a Whatman no. 4 filter paper wedge. While still wet, the grids were washed three or four times successively with a drop of 1% aqueous uranyl acetate each time, and the final drop was left on the grids for 40 to 45 s. Excess stain was removed by the filter paper wedge, and the grids were allowed to air dry and stored under desiccation until used for electron microscopy observation. The samples were viewed under a Hitachi H6000 electron microscope, operating at 75,000 eV with a condenser aperture of 200 μm and an objective aperture of 50 μm, at an apparent magnification of 30,000.

RESULTS

Viability of single alanine substitution mutants of site 1 and site 2 acidic residues.

To test if the seven Vp1 acidic residues of presumed calcium-binding sites 1 and 2 (Glu46, Glu48, Glu157, Glu160, Glu216, Glu330, and Asp345) play a role in the viral infection cycle, these residues were individually mutated in the nonoverlapping, infectious viral genome, NO- SV40. Glu329 is not among the inferred metal-coordinating residues but was included in the analysis because of its possible influence on the adjacent Glu330. The viability of the mutants was determined by plaque formation assays using lysates of mutant viral DNA-transfected cells. As shown in Table 1, single alanine substitutions had a range of effects on plaque formation efficiency. Mutants E46A, E216A, E329A, and D345A were comparable to the wild type in terms of the infectious titer (PFU per unit lysate) and the average size of the plaques. Mutants E157A, E160A, and E330A gave 50- to 80-fold-reduced titers with concomitant reductions in plaque sizes. Mutant E48A showed a 70,000-fold reduction in PFU and a similar small-plaque phenotype. Thus, of all acidic side chains mapped to the calcium-binding sites, that of Glu48 at site 1 is the most critical to viability.

TABLE 1.

Viability of Vp1 calcium-binding site mutants

Label	NO-SV40	Titer (PFU)a	Plaque diam (mm)
	Wild type	(1.8 ± 1.1) × 108	6.6 ± 2.2
a	E46A	2.0 × 108	5.5 ± 0.7
b	E48A	2.5 × 103	1.9 ± 1.2
c	E157A	2.3 × 106	2.0 ± 0.3
d	E160A	3.6 × 106	1.6 ± 0.4
e	E216A	2.8 × 107	5.8 ± 1.3
f	E329A	5.0 × 107	4.8 ± 1.3
g	E330A	2.4 × 106	1.2 ± 0.4
h	D345A	3.5 × 107	6.4 ± 2.3
i	E46A-E48A	6.2 × 102	1.7 ± 0.9
j	E157A-E160A	1.0 × 103	1.2 ± 0.3
k	E329A-E330A	6.6 × 102	1.7 ± 0.5
L	E216K	3.1 × 102	2.0 ± 0.9
m	E216R	5.0 × 102	1.7 ± 0.8
n	E330K	0b
o	E330R	7.5	1.0 ± 0.0
p	D345K	4.0 × 107	6.5 ± 2.0
q	D345R	4.0 × 107	5.9 ± 2.2
r	E329A-E330K	0b
s	E329A-E330R	0b
T1	E157Q-E160Q-D345N	1.2 × 106	2.1 ± 0.7
T2	E329Q-E330Q-D345N	3.5 × 102	2.2 ± 0.6

^aPFU contained in the lysate of one 60-mm dish of cells that were transfected with the respective NO-SV40 DNA. The value represents the average from five experiments for the wild-type sample and the average from two experiments for mutants a through T2.

^bNo plaques detected in one-eighth of the lysate harvested from one dish of transfected cells.

Effect on viability of basic side chain substitution at Glu216 and Glu330 or Asp345.

The minimal to moderate effect on the viability of alanine substitution at Glu216, Glu330, or Asp345 (Table 1) suggests that the small, neutral alanine side chains are largely compatible with the structural environment of the original glutamic acids. On the other hand, a bulky, basic side chain of lysine or arginine placed at residue 216, 330, or 345 could exert a much greater effect on viability for two reasons. One is that the alignment of the calcium-binding residues would be perturbed, compromising calcium binding and diminishing the yield of particles. The other is that the basic side chain might form salt bridges with other acidic side chains, thereby producing particles that do not bind calcium. Such particles may lack the potential to undergo the appropriate conformational changes or dissociation during subsequent infection.

Basic side chain substitutions at Glu216 and Glu330 proved highly detrimental to viability. Mutants E216K, E216R, and E330R had substantially (>300,000-fold) reduced PFUs and a small-plaque phenotype, whereas mutant E330K failed to produce any measurable infectious units (Table 1). In contrast, mutants D345K and D345R were as viable as their alanine-substituted counterpart. These results suggest that the Glu330 side chain at site 1 and the Glu216 side chain which coordinates calcium ions at both sites 1 and 2 can influence the local structure of the calcium-binding sites. The Asp345 side chain at site 2 is dispensable for productive infection.

Viability of double and triple mutants of site 1 and site 2 acidic residues.

The substantial viability of most single alanine mutants (Table 1) could indicate that their original acidic side chains collectively coordinate the twin calcium ions and that losing individual side chains does not significantly compromise the calcium coordination. We therefore combined alanine mutations for pairs of residues that are close in amino acid sequence and structure (glutamic acids 46 and 48, 157 and 160, and 329 and 330). As seen in Table 1, each double mutant preserved the small-plaque phenotype of the single-mutant member with the smaller plaque size. For mutant E46A-E48A, the PFU value approximated that of the less viable single-mutant member, E48A. For mutants E157A-E160A and E329A-E330A, the PFU values reflected a multiplicative accumulation of PFU reductions sustained by the single-mutant counterparts. Thus, the reductions in infectious titer were at least 200,000-fold for the three double mutants. Adding the E329A mutation to the nonviable or nearly nonviable basic-residue-substituted mutants, E330K and E330R, yielded two nonviable double mutants, as might be expected (Table 1).

We also tested the viabilities of two triple glutamine-asparagine substituted mutants, E157Q-E160Q-D345N and E329Q-E330Q-D345N, whose mutant Vp1s have been previously expressed and found to form VLPs in insect cells (12). The relative unimportance of the side chain identity at residue 345 makes it reasonable to compare the viability of the triple mutants with that of their alanine-substituted double-mutant counterparts, E157A-E160A and E329A-E330A. Whereas mutant E329Q-E330Q-D345N had a similar titer to mutant E329A-E330A, mutant E157Q-E160Q-D345N had a 3-log-unit-higher titer than did mutant E157A-E160A (Table 1). This difference could be due to the smaller disruptive effect of replacing the more structurally similar glutamines, rather than alanines, by glutamic acids in mutant E157Q-E160Q-D345N and is consistent with the greater-than-wild-type resistance of E157Q-E160Q-D345N and E329Q-E330Q-D345N VLPs to calcium-chelating agents (12).

The collective results show that five of the seven putative calcium-binding residues of Vp1—Glu48 and Glu330 from site 1, Glu157 and Glu160 from site 2, and Glu216 from both sites—contribute to viral viability, presumably by jointly coordinating calcium ions at respective sites. Glu46 and Asp345, on the other hand, are nonessential. Glu329 can apparently substitute for Glu330 in the absence of the latter residue. The low viability of basic-residue-substituted Glu216 and Glu330 mutants may be due to either the disruption of the calcium-binding site structure or the displacement of site 1 calcium.

Viral DNA replication and Vp1 production by mutants.

We proceeded to dissect the effects of calcium-binding-site mutations on various processes of the infection cycle. We began by examining viral DNA replication and Vp1 production. The amount of replicated viral DNA extracted from transfected cells was similar for the wild type and all mutants, as judged by the intensity of DpnI-resistant viral DNA bands following agarose gel electrophoresis and ethidium bromide staining (data not shown). The amount of Vp1 in transfected cell lysates, as detected by Western blot analysis, was also comparable for the wild type and the mutants, with the exception of mutant E157Q-E160Q-D345N (Fig. 2, lane T1). The low steady-state level of this mutant Vp1 was reproducible and was not investigated further.

FIG. 2.

Vp1 production by mutants. Cells transfected with wild-type (lanes W) or mutant (lanes a through T2 as designated in Table 1) NO-SV40 DNA were subjected to anti-Vp1 Western blot analysis as described in Materials and Methods.

Packaging efficiency of mutants: DNase I resistance assay.

We further tested whether the mutants could package viral DNA into particles, using a DNase I resistance assay. The fraction of intracellular viral DNA that was packaged into some protective structure should remain after the nuclease digestion of the transfected lysate and was quantitated by Southern blot analysis. As seen in Fig. 3, the percentages of DNase I resistance for the mutants range from 59 to 5%. A plot of these percentages against PFU (Table 1) in Fig. 4 shows distinct clustering of the mutants into five groups, which we label as wild-type-like, I, II, III, and IV.

FIG. 3.

Extent of packaging by mutants. An aliquot of wild-type or mutant NO-SV40-transfected lysates was treated with DNase I, and the remaining, nuclease-resistant DNA was quantitated and expressed as a percentage of total viral DNA in the aliquot, as described in Materials and Methods. Each bar, with standard deviation marked, represents the average value from two to four experiments, as indicated on the right of the graph.

FIG. 4.

DNase I-resistant packaging as a function of the number of PFU of the mutants. Percent DNase I-resistant viral DNA from Fig. 3 is plotted against PFU from Table 1 for NO-SV40 wild type (Wt) and mutants. Clusters of mutants are boxed and designated as one of the following groups: Wt-like, I, II, III, and IV. The names of the mutants are listed near each boxed group in descending order of DNase I resistance. Site 1 mutants are plotted as dots; site 2 mutants are plotted as dots surrounded by a single circle; and mutants containing both site 1 and site 2 mutations are plotted as dots surrounded by a double circle.

The wild-type-like group had wild-type-like PFUs and plaque sizes and packaged viral DNA either at a wild-type level of 60% ± 9% (D345A and D345K) or in a somewhat lower range of 47 to 30% (E46A, E216A, E329A, and D345R) (Fig. 3 and 4). Group I mutants also packaged in the 47 to 30% range, except that they had 50- to 150-fold-reduced PFUs and a small-plaque phenotype (E157A, E160A, E330A, and E157Q-E160Q-D345N) (Table 1; Fig. 3 and 4). Thus, group I mutants appeared to form reasonable quantities of particles that had lower infectivities than wild-type particles.

Group II mutants had substantially (70,000- to 600,000-fold) lower PFUs than did the wild type, along with small-plaque phenotypes, and packaged from 36 to 21% (E48A, E46A-E48A, E157A-E160A, E216K, E329A-E330A, and E329Q-E330Q-D345N) (Table 1; Fig. 3 and 4). The group III mutant E216R is in the same PFU range as group II, except that its packaging level was extremely low, about 5% (Fig. 3 and 4). Thus, group II mutants appeared to make particles that were even more poorly infectious than group I particles whereas the mutant E216R was significantly defective in packaging viral DNA into a nuclease-resistant structure.

Group IV mutants were nonviable (E330K, E329A-E330K, and E329A-E330R) or nearly so (E330R) but were surprisingly capable of packaging into a DNase I-resistant form (24 to 40%) (Fig. 3 and 4). Thus, group IV mutants are expected to assemble almost exclusively noninfectious particles.

The above results show that mutants of the calcium-binding site residues are, in general, defective to various extents in the assembly of virus particles. For mutants of the wild-type-like group and groups I and II, there is a reasonable correlation between packaging efficiency and viability. The group III mutant E216R has the largest defect in packaging. It indicates that the positioning of the Glu216 side chain relative to that of its α-carbon is important for packaging and virion assembly and suggests that the local structure of Glu216 is susceptible to disturbance by the insertion of a bulky, positively charged side chain. The packaging-competent group IV mutants are expected to be defective in initiating new infections.

VLP formation by mutants.

To confirm that the nuclease-protected fraction for the mutants represented VLPs, the DNase I-resistant materials of select mutants from groups II, III, and IV were analyzed by sedimentation in sucrose gradients. As seen in Fig. 5A, the wild-type-transfected sample had a peak of viral DNA (in the form of nicked open circular, linear, and covalently closed circular species) in fractions 4 and 5, coinciding with the sedimentation position for purified virions, and Vp1 was found mostly in the same fractions.

FIG. 5.

VLP formation by mutants. DNase I-treated, transfected cell lysates for wild-type (A) and mutant E48A (B), E216R (C), and E330K (D) NO-SV40s were sedimented through 5 to 32% sucrose gradients, and the 17 fractions collected from the bottom of the gradients were analyzed for viral DNA (top half of each panel) or for Vp1 (bottom half of each panel), as described in Materials and Methods. The amount of lysates sedimented was equivalent to 1 × 10⁷ (A), 2 × 10⁷ (B), 4 × 10⁷ (C), and 1.5 × 10⁷ (D) transfected cells. A solid arrowhead above the junction of fractions 4 and 5 indicates the position for purified virions sedimented in a parallel gradient. Three bars to the right of each viral DNA half panel mark the positions (from top to bottom) of the open circular, linear, and closed circular forms of the viral DNA. The profiles of mutants E157A-E160A and E216K resemble that of mutant E48A (B), and the profile of mutant E330R resembles that of mutant E330K (D).

For group II mutants E48A, E157A-E160A, and E216K, the mutant DNA, along with a notable amount of the mutant Vp1, sedimented mostly in fractions 4 through 6, consistent with the formation of VLPs (Fig. 5B). The viral DNA found in fractions 1 through 3 (near the bottom of the gradient), together with some Vp1, may represent particles of somewhat different shape or size from wild-type virions or could be aggregated VLPs. The substantial amount of Vp1 that was found in fractions 7 through 9 without much cosedimented viral DNA may represent mostly capsid protein aggregates or Vp1 assembly intermediates.

The group III mutant E216R had a strikingly different sedimentation profile (Fig. 5C). Viral DNA was nearly absent in fractions 2 through 11, whereas Vp1 was distributed broadly throughout the gradient. This pattern suggests that mutant E216R formed capsid protein aggregates, packaging intermediates, and even VLPs, but most of the viral DNAs in these complexes was susceptible to DNase I, consistent with a low DNase I resistance level of only 5% (Fig. 3). The elevated amount of mutant DNA in fraction 17 was most probably due to the overloading of the mutant sample onto the sucrose gradient relative to other samples, but it could also signify that some of the DNA that was protected during the nuclease digestion was released from the unstable structures during sedimentation.

For group IV mutants E330K and E330R, the cofractionation of a majority of the viral DNA with Vp1 between fractions 3 and 9 suggests the formation of VLPs (Fig. 5D). The distribution of viral DNA and Vp1 is shifted toward the right (top of the gradient) relative to that of the wild type (Fig. 5A) and may indicate a difference in the size or shape of mutant E330K and E330R VLPs compared to wild-type virions. However, electron microscopy revealed the E330K VLP to be almost identical in appearance and diameter (47.9 ± 3.4 nm [Fig. 6A ]) to wild-type SV40 particles (47.8 ± 2.7 nm [Fig. 6B]).

FIG. 6.

Electron micrograph of mutant E330K VLPs. Mutant E330K VLPs (A) and wild-type SV40 virions (B) were visualized by electron microscopy together with polystyrene spheres (diameter, 91 nm) as an internal particle standard. Bars, 10 nm. The diameter of the SV40 virion is 47.8 ± 2.7 nm (n = 337), and that of the E330K VLP is 47.9 ± 3.4 nm (n = 257).

The collective results showed that at least some of the DNase I-resistant materials for group II and IV mutants represented VLPs, supporting the idea that the poor plaque-forming ability of these mutants is due to a block in reinfection processes.

Cell attachment and entry by mutant E157A-E160A, E216K, and E330K VLPs.

We further analyzed two group II mutants, E157A-E160A and E216K, and one group IV mutant, E330K, to test the ability of their VLPs to initiate new infections. Two sources of VLPs were assayed: peak VLP fractions from the sucrose gradients above (Fig. 7A and B) and crude transfected cell lysates (Fig. 7C). The former type of VLP preparations for mutants E157A-E160A, E216K, and E330K contained all three capsid proteins, Vp1, Vp2, and Vp3, at similar ratios to those observed for wild-type particles (Fig. 7D). Therefore, all three mutant VLPs had a wild-type-like capsid protein composition.

FIG. 7.

VLP composition and infection processes for mutants E157A-E160A, E216K, and E330K. (A to C) Cell attachment and internalization by mutant VLPs. Cells grown on 100-mm dishes were infected with 1,000 particles per cell for 4 h, with peak sucrose fractions containing wild-type particles or mutant E157A-E160A (E157/160A), E216K, or E330K VLPs. Cells in one dish were harvested by scraping (representing “Cell associated” viral DNA or Vp1), those in another dish were harvested by trypsin treatment (“Internalized” viral DNA or Vp1). (A) Viral DNA was extracted from one-eighth of the input particles (“Input”) or from half of the respectively harvested infected cells, linearized, and detected by Southern blot analysis. (B) One-hundredth of the input particles (“Input”) and one-twentieth of each type of harvested cells were analyzed for Vp1 Western blot analysis. (C) Cells were infected with transfected cell lysates instead of peak VLP fractions and processed for viral DNA detection by Southern blot analysis, as above. (D) Capsid protein composition of mutant VLPs. Peak sucrose fractions corresponding to wild-type particles or various mutant VLPs were analyzed for viral proteins by Western blotting using anti-Vp1 (top panel) or anti-Vp3 (bottom panel) antibody. Bands corresponding to Vp1, Vp2, and Vp3 are marked. (E) T-antigen expression in mutant VLP-infected cells. Cells grown on coverslips were infected for 20 h with peak sucrose fractions corresponding to wild-type particles or various mutant VLPs. The number of T-antigen-positive cells was determined by immunofluorescence microscopy. Each bar, with marked standard deviation, represents the average from three sets of experiments, in each of which approximately 2,000 cells were counted.

We first asked whether the VLPs could attach to cells and become internalized following infection. To distinguish cell surface-bound and internalized VLPs, cells were harvested at 4 h postinfection by two different methods: by scraping, for quantitating the VLPs that had become cell associated either at the surface or in the interior, and by trypsin treatment, for quantitating only the VLPs that had been internalized. When peak VLP fractions were used for the infections, mutant E157A-E160A and E216K VLPs attached to cells just as well as wild-type particles did, as judged by the amount of cell-associated viral DNA (Fig. 7A) and Vp1 (Fig. 7B). The amounts of their internalized viral DNAs were also comparable to that of the wild-type sample (Fig. 7A and B). Infection with the crude lysates for mutants E157A-E160A and E216K gave essentially the same results (Fig. 7C). In contrast, little or no viral DNA or Vp1 became cell associated or internalized following infection with mutant E330K VLPs (Fig. 7A and B). However, cell-associated mutant E330K DNA was detected when crude mutant E330K lysate was used for infection (Fig. 7C), suggesting that non-particle-derived viral DNA could enter cells together with viral proteins in the lysate. These results indicate that mutant E157A-E160A and E216K particles are capable of cell attachment and entry whereas mutant E330K particles are defective in cell attachment.

T-antigen expression following infection of mutant E157A-E160A, E216K, and E330K VLPs.

Since mutant E157A-E160A and E216K VLPs could enter the cell, we then asked if they could deliver their viral DNA to the nucleus and initiate viral gene expression. At 20 h postinfection, 46.2% ± 11.7% of the wild-type-particle-infected cells became positive for the large T antigen (Fig. 7E). However, only a small portion of E157A-E160A (0.43% ± 0.05%) or E216K (0.065% ± 0.02%) VLP-infected cells was T-antigen positive (Fig. 7E). Not surprisingly, no T-antigen-positive cells resulted from E330K VLP infection (Fig. 7E), consistent with the nonviability and defect in cell attachment of this mutant. These results show that internalized mutant E157A-E160A and E216K VLPs are blocked at some event preceding T-antigen expression, such as delivery of the viral genome to the nucleus.

State of internalized mutant E157A-E160A and E216K VLPs in the cytoplasm.

We have previously shown that once internalized in the cytoplasm, the SV40 particle undergoes some alteration in capsid structure such that the nuclear localization signals (NLSs) of the virion, the internal Vp3 NLSs, become exposed for recognition by cellular importins (22). Thus, the defect of mutant E157A-E160A and E216K VLPs could be either due to an inability to undergo this critical alteration or due to a premature disassembly of the VLPs. These two possibilities can be distinguished by analyzing the coimmunoprecipitation of internalized, VLP-associated viral DNA with anti-capsid protein antibodies. Poor viral DNA coprecipitation with anti-Vp3 but normal coprecipitation with anti-Vp1 would suggest failure to undergo a structural alteration, whereas reduced coprecipitation with both anti-Vp1 and anti-Vp3 would suggest premature disassembly. In addition, immunoprecipitation with anti-importin antibody would reveal whether the internalized VLPs could be recognized by the importins.

Cytoplasmic lysates prepared from wild-type-particle- or mutant VLP-infected cells were reacted with individual antibodies, and coimmunoprecipitated viral DNA was detected via semiquantitative PCR. Control antibody immunoprecipitation yielded no detectable viral DNA, as expected (Fig. 8, lane 2). Although similar amounts of viral DNA were present in wild-type and mutant VLP-infected cytoplasms (lane 1), much less of the mutant viral DNA than wild-type viral DNA was coimmunoprecipitated by either anti-Vp1 or anti-Vp3 (lanes 3 and 4). For example, whereas 52.4% of the internalized wild-type DNA was detected in association with Vp1, only 2.8 and 14.6% of the internalized mutant E157A-E160A and E216K viral DNAs, respectively, were detected. Not surprisingly, the mutant DNAs were also poorly coimmunoprecipitated by mixed anti-importin-α and -β antibodies (lane 5). These results are consistent with the interpretation that mutant E157A-E160A and E216K particles fail to sustain adequate structural integrity in the infected cytoplasm and hence cannot mediate the nuclear transport of the viral DNAs via interaction with the importins.

FIG. 8.

Immunoprecipitation analysis of internalized particles. The cytoplasmic fraction was prepared from cells infected with wild-type particles (top panel) mutant E157A-E160A particles (middle panel), or mutant E216K particles (bottom panel) as described in the legend to Fig. 7A. Aliquots of the cytoplasmic fractions were reacted with anti-mouse IgG (lane 2, Cont), anti-Vp1 (lane 3, Vp1), anti-Vp3 (lane 4, Vp3), and mixed anti-importin-α and anti-importin-β antibodies (lane 5, Imps), as described in Materials and Methods. The coimmunoprecipitated (IP) viral DNA was purified from the immune complexes in the presence of a fixed amount of control DNA and detected via semiquantitative PCR. The expected amplification product of the NO-SV40 genome is 2.2 kbp (arrow, viral DNA), and that of the standard DNA is 1.7 kbp (arrowhead, cont. DNA). For Input (lane 1), one-fifth as much cytoplasmic lysate as the amount used for each immunoprecipitation was purified for viral DNA in the presence of the control DNA and detected by PCR.

DISCUSSION

In the present study, we analyzed SV40 mutants in which the Vp1 side chains mapped by crystallographic studies to the two calcium-binding sites (Fig. 1) (19, 32) were mutated. A DNA transfection system and assays that probe for particle formation and the effect of particle infection in the next host have allowed us to dissect the defects of the mutants at various stages of the SV40 life cycle (e.g., particle formation and particle entry into the cell and nucleus). We show that a subset of those seven acidic side chains—Glu48 and Glu330 of site 1, Glu157 and Glu160 of site 2, and Glu216, which is a part of both sites—are important for SV40 infection. Our in vivo study provides the first evidence for the involvement of calcium ion-mediated interactions not only in virion assembly but also in cell entry and delivery of the viral genome into the nucleus.

Our data suggest that Glu48, Glu330, Glu157, Glu160, and Glu216 are the primary side chains that coordinate calcium. We have also found Glu329, which is not among the predicted calcium-binding residues, to be important in the absence of the adjacent Glu330. All six residues are well conserved in the polyomavirus family. As for the two acidic residues that we found to be unimportant for SV40 infection, Glu46 is not well conserved and Asp345 does not assume an ordered structure for γ-type Vp1 monomers on the capsid (32). Thus, Glu330 of site 1 could be the only residue on the invading C-terminal arm that normally contacts calcium ions. This C-terminal contribution, however, is essential for calcium coordination, as suggested by our analysis below. It concurs with the prediction that calcium would not bind to both sites 1 and 2 unless the invading arm is present in the pentamer core (32).

For SV40 and the closely related murine polyomavirus and human JC virus, the importance of calcium binding in virion stability and assembly is well documented. Experimental evidence includes the in vitro dissociation and/or reassembly properties of virions (4, 5, 6), of VLPs from bacterially expressed Vp1 (10, 29), and of VLPs formed in insect cells (12) and yeast cells (8). The observation that VLPs unexpectedly form in the cytoplasm when polyomavirus Vp1-expressing insect cells are treated with a calcium ionophore (20) implies that VLP formation is controlled by the availability of calcium ions in specific intracellular compartments. These studies, however, do not address the possible role of the metal ions in the earlier phase of infection, during virion entry into the cell and into the nucleus.

The SV40 virion is presumed to be a carrier of calcium ions. The metal ions, jointly coordinated by mostly acidic Vp1 residues, map near the base of Vp1 pentamers at a stoichiometry of two ions per Vp1 chain (Fig. 9A and B). If all available binding sites on the capsid are filled, the total 720 metal ions would amount to 5 to 10 μM, or 100 times the estimated average level of free intracellular calcium, 0.1 μM. It is not known whether all available binding sites are occupied in SV40 or how the metal ions can be extensively accumulated during particle formation in the nucleus. Later, such as during the infection of a new host cell, release of the calcium ions from the particle may occur at different stages in a controlled mannner. Although the number of calcium ions bound per virion or mutant VLP is not known, the results presented here suggest that the calcium ions carried by an SV40 virion participate in multiple steps of the infection cycle.

FIG. 9.

Location and function of calcium-binding sites. (A and B) Positions of Glu330 and Glu160 in SV40. The α-carbon positions for Glu330 (site 1) and Glu160 (site 2) are marked in red and yellow, respectively, in a cross section that includes parts of pentamers (A) as viewed from the top, with a cluster of seven pentamers (a hexavalent one surrounded by six others) highlighted by enclosure in a white circle (B). (C) Proposed functions of Vp1 calcium-binding amino acids in infection: cell attachment through viral DNA nuclear import. The model shows an SV40 virion undergoing cell attachment through MHC class I (“Y”), internalization via caveolae (crisscrossing double lines) on the plasma membrane (straight double lines), structural alteration in the cytoplasm to expose the Vp3 NLS (Vp3-NLS), recognition by importin-α and importin-β, and nuclear entry through a nuclear pore complex embedded in the nuclear envelope. Glu330 is important for attachment to the cell. The Glu157-Glu160 pair and Glu216 function in the controlled and selective structural alteration to achieve a nuclear import-competent state. Panel C modified from references 13 and 23, with permission of the publishers.

Model.

Calcium coordination links the C-terminal arms of neighboring pentamers with the core of each pentamer, thereby helping to stabilize the virion particle by reinforcing interpentamer interactions (19, 32). Our working model for the role of calcium binding in SV40 infection is as follows. The capsid structure held together by calcium ions has the potential to change after loss or gain of the ions at some locations of the capsid. Consequently, appropriate structural alterations can occur during different steps of the infection cycle. During virion assembly in the nucleus, calcium ion addition drives the formation of the capsid structure from Vp1 pentamers, together with Vp2 and Vp3 and the viral minichromosome, and stabilizes the particle structure. During cell attachment, the virion undergoes a conformational change, perhaps via calcium ion release, to achieve receptor binding. Such a structural alteration can also underlie internalization via the caveolar pathway (1, 24). Furthermore, nuclear entry of the viral DNA may depend on a calcium-mediated structural alteration of the virion, without dissociating the capsid proteins from the DNA, so that the NLSs of the virion, the Vp3 NLSs, are exposed and recognized by the importins (22). In short, calcium ion coordination can contribute to different processes of the SV40 infection cycle by imparting stability to the capsid structure while allowing for the flexibility of that structure to alter on cue. That is, biologically significant conformational change appears to occur at early stages of infection. This model is consistent with the results we report here.

Regulation of virion conformation by calcium ions, also thought to occur in the early steps of infection, has been documented for several T=3 icosahedral plant viruses. For example, when divalent cations are removed and the pH is raised above neutrality, tomato bushy stunt virus (16, 27) and cowpea chlorotic mottle virus (3, 30) undergo swelling, a reversible change in capsid structure that results from the loss of the positively charged metal chelated at interfaces of capsid protein subunits. We envision that controlled or local release of calcium ions from the virion under certain extra- or intracellular conditions would also trigger a defined conformation change in SV40, unlike the complete disruption of virions or VLPs after treatment with divalent cation-chelating and -reducing agents in vitro (4, 6, 29).

Viability of mutants.

Viability, measured as the number of PFU detected in each viral DNA-transfected cell lysate and the size of plaques, allows us to rate the impact of the calcium-binding site mutations on the overall ability to make infectious particles. The rating reveals that five of the seven putative calcium-binding residues (Glu48, Glu157, Glu160, Glu216, and Glu330), plus Glu329, are important, because mutating them singly or in combination produced the lowest infectious titers observed in this study, along with a small-plaque phenotype (Table 1 and Fig. 4, group II mutants E48A, E157A-E160A, E329A-E330A, and E216K and group III mutant E216R), or produced no detectable infectious titer at all (group IV mutants E330K and E330R). The defects of these mutants lie in packaging and particle formation and/or in stages of reinfection (see below).

Glu46 and Asp345, which can be mutated while maintaining a wild-type-like viability, are likely to play only a minor role during infection. It may be that Glu46 and Asp345 do not actually contact calcium ions, contrary to structural prediction (32). The equivalent viability of mutants D345A, D345K, and D345R suggests that the Asp345 side chain does not contribute to the local structure of the capsid; i.e., the region of the C-terminal arm encompassing residue 345 may be somewhat flexible. An alternative possibility, that the infectious D345K or D345R particles contain rigid salt bridges between the basic side chain at residue 345 and the acidic side chain at residue 157, 160, or 216, is unlikely in the light of the poor infectivity of the analogous mutants E330K and E330R (Table 1).

Packaging and VLP assembly.

The extent of packaging into DNase I-resistant structures in transfected cells gave us a measure of the ability of the mutants to assemble into VLPs. That the wild-type-like mutant D345R packaged 31% of the DNA, or roughly half the wild-type level (Fig. 3 and 4), suggests that viability is not greatly affected by this level of packaging reduction per se. In fact, only 6 of 21 mutants analyzed packaged below 30% (E48A, E329A-E330A, E157A-E160A, E216K, E216R, and E330R). Significantly, two of the six are doubly mutated and three contain a basic-residue substitution. Conceivably, losing one calcium-binding side chain may not greatly affect the calcium coordination affinity at a particular site because of the presence of other side chains, but losing two could have a significant effect. If an acidic residue were replaced with arginine, the basic, bulky side chain could disrupt the structure of one or both calcium sites, hence disrupting calcium ion binding and possibly other interactions normally made by the calcium site residues with other residues. These observations support the idea that calcium coordination at two sites is a driving force in virion assembly.

Among the mutants that packaged least efficiently, all but mutant E216R made particles that sedimented in similar positions to wild-type particles, although group II mutants appear to form some particle aggregates or assembly intermediates besides VLPs (Fig. 5). The VLPs of the group IV mutant E330K are indistinguishable from wild-type virions (Fig. 6). Mutant E157A-E160A, E216K, and E330K VLPs also had normal capsid protein compositions (Fig. 7D). Therefore, the primary reason for the low viabilities of group II and IV mutants is the low infectivity of the VLPs, which are blocked at some stage of reinfection.

The single group III member, mutant E216R, packaged viral DNA poorly and was poorly viable. Its sucrose gradient profile suggests that it assembled the mostly capsid protein aggregates or nucleoprotein complexes whose viral DNA was accessible to DNase I (Fig. 5C). The detrimental effect on assembly of the arginine substitution at residue 216 may stem from the unique location of Glu216, which links calcium ions at both sites 1 and 2 and might have an influence on the local capsid structure. The bulky basic side chain of arginine can perturb both sites 1 and 2, possibly abolishing the coordination of any calcium ions. Although the Glu216 side chain may normally make contacts with both ions, substitution by the small, neutral side-chain alanine (as in the wild-type-like mutant E216A) is well tolerated (Table 1). Therefore, the positioning of the Glu216 side chain relative to that of the α-carbon may influence minichromosome packaging and virion assembly. The phenotype of mutant E216R is reminiscent of the poorly packaging and poorly viable Vp1 cysteine mutant, C254L (17). Cys254, residing on the short, highly conserved GH loop near the calcium-binding sites, also appears to be in a highly structurally sensitive local environment that is perturbed by certain substituted side chains at that residue (17).

Virus attachment to cells, Glu330.

The phenotype of group IV mutants, including E330K, E330R, E329A-E330K, and E329A-E330R, is intriguing: VLPs with the same morphology and composition as virions were effectively made but were largely noninfectious (Fig. 3, 5, and 6A). In the mutant E330K particle, the substituted lysine side chain at residue 330 may be positioned such that it actually fits in the space normally occupied by the Glu330 side chain and the calcium ion. The lysine side chain probably makes salt bridges with other acidic side chains of site 1, Glu48 and Glu216, thus permanently excluding calcium from site 1. This new arrangement would allow E330K VLPs to assemble effectively while chelating only half of the calcium ions chelated by wild-type particles. Future in vitro assays could test whether these VLPs are more stable than wild-type virions in the presence of calcium-chelating agents.

Mutant E330K VLPs turned out to be defective in adsorbing to cells (Fig. 7A and B). Conceivably, the salt bridges in place of calcium ion at site 1 could have prevented the E330K VLP from undergoing the calcium-dependent structural shift necessary for binding to the cell receptor, the major histocompatibility complex (MHC) class I molecule (7, 31), leading to the adsorption defect. If the structurally unresponsive mutant particle failed to expose the myristylated N terminus of Vp2, which is thought to insert into the plasma membrane to facilitate virion entry (28, 33), this could contribute to the adsorption defect. Glu330 residues, marked in red in Fig. 9A and B, lie on the C-terminal arms of neighboring pentamers and are situated near the base of the pentamers. The results obtained with mutant E330K suggest a role of the calcium ions in supporting the dynamic alteration of capsid structure that takes place at the pentamer base during cell attachment (Fig. 9C). For example, the virion may release site 1 calciums to achieve the conformation for the MHC class I interaction. This scenario is part of our working model that cell attachment requires a particle structural flexibility conferred by the coordination of calcium ions. Analyzing the structural difference between the mutant E330K VLP and the wild-type virion could help delineate the virus site or epitope responsible for attachment to MHC class I.

Nuclear import of infecting virus, calcium site 2.

For the two group II mutants, E157A-E160A and E216K, the particles could attach to and enter cells (Fig. 7A and B) but were ineffective at delivering their viral genomes into the nucleus, as seen by the low incidence of T-antigen expression (Fig. 7E). Little internalized DNA of the mutants was in complex with the capsid proteins or with the importins (Fig. 8). It is striking that not only Vp1, but also Vp2 and Vp3, have dissociated from the DNAs. Since the structural proteins promote the nuclear entry of SV40 DNA (21) and since the Vp3 NLS is necessary for virion DNA entry into the nucleus (22), it is not surprising that mutant E157A-E160A and E216K DNAs, with few associated structural proteins, were not recognized by the importins or targeted to the nucleus. Thus, the Glu157- Glu160 and Glu216 mutations appear to weaken the affinity of calcium binding at site 2 sufficiently to cause a premature dissociation of the mutant particles once internalized into the cell, though the VLPs still assembled in the nucleus of the mutant DNA-transfected cells (Fig. 3 and 5). It is not known what conditions in the caveolae or the cytoplasm or which aspect of the attachment and entry process might have compromised the integrity of the mutant VLPs. The infectivity of mutant E157A-E160A and E216K particles is more than 1,000-fold lower than that of Vp3 NLS-null mutant particles, also blocked in viral genome nuclear targeting (22). This difference suggests that the premature dissociation of particles is an even greater obstacle to viral DNA nuclear entry than is the loss of functional Vp3 NLSs in the particles.

Our study thus shows that effective nuclear targeting of the viral DNA cannot be achieved when the particle dissociates. We propose that a selective series of conformational changes in the particle is necessary for the exposure of Vp3 NLSs to cytoplasmic importins (Fig. 9C). Such a model is consistent with the result of a preliminary analysis in which the SV40- infected cytoplasm was fractionated by sucrose gradient sedimentation and then reacted with anti-importin antibodies. Whereas a majority of the antibody-precipitable viral DNA sedimented near the viral chromatin position, some was found in virion fractions (A. Nakanishi, unpublished data). Still unknown is the extent of the structural alteration or the composition of the internalized particle. We seek clues to the possible location of a structural alteration by building an atomic model of the SV40 Vp1-Vp3 complex based on the structure of the homologous mouse polyomavirus Vp1-Vp2 complex (9). From this model, the distance between the Vp3 Leu178 side chain, which is within the defined Vp1-contacting structure, and the calcium-binding Vp1 Glu48 side chain, is 13 Å. The Vp3 NLS, 21 residues downstream from Leu178, is inferred to be near the pentamer base. Since Glu157, Glu160, and Glu216 are all situated in calcium site 2 near the base of pentamers, a structural change in the wild-type particle could be brought about through a depletion of calcium ions from site 2 to permit the exposure of the Vp3 NLSs. Such a particle would then be competent for importin recognition (Fig. 9C). Potential calcium-binding-site mutants that have a permanent bond in place of site 2 calcium, analogous to the site 1 mutant E330K, would help test this scenario. Again, calcium coordination by the SV40 particle is likely to be a key to the structural flexibility necessary for the processes involved in initiating a new infection cycle.

In summary, we have shown that certain acidic side chains from the two structurally deduced calcium-binding sites contribute to the formation of infectious SV40. The potential to chelate calcium ions at both sites is important for virion assembly in the nucleus. The cell attachment defect of the site 1 mutant, E330K, and the DNA nuclear import defect of the site 2 mutants, E157A-E160A and E216K, due to premature particle dissociation in the cytoplasm, imply that local structural alterations occurring at the base of Vp1 pentamers are important in the initiation of an infection cycle. Thus, the proper formation of the virion structure and the ability of the virion to undergo structural alterations on cue at the cell surface or in the cytoplasm are essential for infectivity and are probably regulated by calcium coordination by residues at the two calcium sites (Fig. 9C). These structural alterations set the stage for the nuclear entry of the infecting SV40 through interaction with importin complex and for launching the viral gene expression program.