Identification of Amino Acid Residues in BK Virus VP1 That Are Critical for Viability and Growth

Abstract

BK virus (BKV) is a ubiquitous pathogen that establishes a persistent infection in the urinary tract of 80% of the human population. Like other polyomaviruses, the major capsid protein of BKV, virion protein 1 (VP1), is critical for host cell receptor recognition and for proper virion assembly. BKV uses a carbohydrate complex containing α(2,3)-linked sialic acid attached to glycoprotein and glycolipid motifs as a cellular receptor. To determine the amino acids important for BKV binding to the sialic acid portion of the complex, we generated a series of 17 point mutations in VP1 and scored them for viral growth. The first set of mutants behaved identically to wild-type virus, suggesting that these amino acids were not critical for virus propagation. Another group of VP1 mutants rendered the virus nonviable. These mutations failed to protect viral DNA from DNase I digestion, indicating a role for these domains in capsid assembly and/or packaging of DNA. A third group of VP1 mutations packaged DNA similarly to the wild type but failed to propagate. The initial burst size of these mutations was similar to that of the wild type, indicating that there is no defect in the lytic release of the mutated virions. Binding experiments revealed that a subset of the BKV mutants were unable to attach to their host cells. These motifs are likely important for sialic acid recognition. We next mapped these mutations onto a model of BKV VP1 to provide atomic insight into the role of these sites in the binding of sialic acid to VP1.

The human polyomavirus, BK virus (BKV), was first isolated from a renal transplant recipient in 1971 (10). Initial exposure through a fecal-oral or a respiratory route is common in early childhood and results in adult seropositivity of ca. 80% (2, 3, 11, 13, 17, 26). Postinfection, BKV establishes a persistent, but usually silent infection within epithelial tissue of the kidney (19, 22). In the context of an impaired immune system, the lytic nature of this virus in combination with the host inflammatory response can cause multiple diseases such as hemorrhagic cystitis, retinitis, meningoencephalitis, and interstitial nephritis (4, 5, 24, 25). Polyomavirus-associated nephropathy, is a complication in 5% of kidney transplantations induced by elevated BKV replication (23). Nearly half of patients diagnosed with polyomavirus-associated nephropathy lose their renal graft due to excessive necrosis, fibrosis, and breakdown of renal tubule structure, which disables the function of the kidney (12, 23). Currently, reduction in immunosuppressive therapy is used to modulate viral replication and infection (1, 13).

At the cellular level, BKV infection begins by attachment to host cell α(2,3)-linked sialic acid on N-linked glycoproteins [NeuNAc-(α2,3)-Gal-(β1,3)-GlcNAc] or to sialic acid moieties on the gangliosides GD1b and GT1b (7, 20, 28). Whether a specific proteinaceous receptor is required for entry is unknown; however, α(2,3)-sialic acid linkages on O-linked glycoproteins cannot support infection, suggesting that the presence of sialic acid alone is insufficient for infection (7). Virion attachment is followed by cholesterol-dependent, caveola-mediated endocytosis (9). After entry, dynamic mircotubules shuttle the virus to membranous structures contiguous with the endoplasmic reticulum (6, 8). The release of genome into the nucleus initiates transcription of viral genes and replication of the genome.

Like other polyomaviruses, the BKV virion is composed of 360 copies of the major capsid protein, VP1, that form 72 pentamers, each consisting of a single minor capsid protein, VP2 or VP3, positioned within the center of the pentamer. The crystal structures of simian virus 40 (SV40) and the mouse polyomavirus (mPyV) have been determined (18, 29, 32). In addition, the structure of the mPyV VP1 monomer associated with its oligosaccharide receptor is known (30-32). The VP1 monomer consists of antiparallel β-strands that fold into a jelly-roll β-barrel structure with multiple exposed outer loops that nestle tightly with the adjacent monomer (29). In mPyV, both the straight NeuNAc-(α2,3)-Gal-(β1,3)-GalNAc and the branched sialic acid oligosaccharide NeuNAc-(α2,3)-Gal-(β1,3)-[(α2,6)-NeuNAc]-GlcNAc rest flat and in a shallow groove located at BC1 and BC2 loop junction. The receptor interacts with side chains from amino acids on both HI and BC loops (30-32).

By mutating charged amino acids within BC, DE, and HI loops of BKV VP1, we identified motifs that are important for virus proliferation and also for proper virion assembly. Some mutant viruses show no defect in viral protein production, capsid protection of genome, or lytic release from cells but are less infectious than the wild type (WT). A novel binding assay that quantifies the level of viral encapsidated DNA that is bound to the cell surface shows that a subset of BKV mutants cannot attach to host cells. We coupled these data with molecular dynamic (MD) simulations in order to visualize one possible conformation for BKV VP1 and sialic acid receptor binding.

MATERIALS AND METHODS

Modeling of BKV VP1 and computer simulations of the sialic acid-VP1 complex.

Using the SWISS-MODEL protein structure homology-modeling server, a model of BKV VP1 was generated based on 88% amino acid homology and 82% amino acid identity to the crystal structure of SV40. For alignment of BKV VP1 Dunlop strain (accession number P0388) and SV40 VP1 K661 strain (accession number AAAC59343), the location of loops, RG motifs, and mutations are as shown (Fig. 1) (27). Using the mPyV VP1/receptor complex (protein Data Bank entry 1VPS), 20 initial conformations of NeuNAc-(α2,3)-Gal-(β1,3)-GlcNAc receptor/BKV VP1 complex were generated by translations or rotations of the ligand within the tentative binding pocket. Each complex was energy minimized to remove the initial strain, placed centrally in a simulation box of 12 nm and soaked with water (typically 37,700 water molecules), and further energy minimized. To allow for rearrangement of the ligand-receptor complex, each complex was subjected to 100 ps of MD simulation (at 300 K, with an integration time step of 2 fs and a constant pressure of 1 bar) using a GROMACS 3.3 and the OPLS-AA force field (33). The complex with the lowest energy was subjected to an extended MD simulation of 1.3 ns.

FIG. 1.

Alignment of SV40 VP1 K661 strain and BKV VP1 Dunlop strain. Outer VP1 loops of interest are shaded; the BC loop is located at amino acids 57 to 89, the HI loop is located at amino acids 127 to 146, and the DE loop is located at amino acids 269 to 283. Amino acids that are identical in BKV and SV40 are depicted with two dots, conservative amino acids are depicted as a single dot, and nonconserved amino acids are depicted as a space without dots. The RG motifs are boxed. Any amino acid mutated in the present study is underlined. The amino acid switch that represents a change subtype polymorphism is underlined and in boldface.

Cells and plasmids.

Vero cells (American Type Culture Collection [ATCC]) were maintained in a humidified 37°C CO₂ chamber in Eagle minimal essential medium (EMEM; Mediatech) supplemented with 5% heat-inactivated fetal bovine serum (Mediatech) and 1% penicillin-streptomycin (HyClone). The BKV Dunlop genome was cloned out of pBKV(34-2) (ATCC) and into the BamHI site of pUC19 to enable high-copy expression of DNA (Invitrogen). In this new plasmid, BKV-pUC19(WT), the BamHI site separates the coding region of VP1. For virion production to occur, BamHI digestion prior to transfection, as well as host cell repair mechanisms to circularize plasmid after transfection, is required.

Transfection.

Plasmids were digested with 2 U of BamHI (New England Biolabs) for every 1 μg of DNA for 4 h at 37°C to separate the BKV genome from the backbone plasmid. The DNA was incubated at 65°C to inactivate the enzyme. The DNA was transfected into Vero cells by using Lipofectamine and Lipofectamine Plus reagents (Invitrogen) as described in the manufacturer's instructions.

Site-directed mutagenesis.

The QuikChange site-directed mutagenesis kit (Stratagene) was used to generate mutations in BKV-pUC19 plasmid according to the manufacturer's instructions. The primers used for mutagenesis were as follows (5′→3′): R64A, CCAGATGAAAACCTTGCGGGCTTTAGTCTAAAG and GGTCTACTTTTGGAACGCCCGAAATCAGATTTC; R64K, CCAGATGAAAACCTTAAGGGCTTTAGTCTAAAG and GGTCTACTTTTGGAATTCCCGAAATCAGATTTC; G65A, GATGAAAACCTTAGGGCCTTTACTCTAAAGC and CTACTTTTGGAATCCCGGAAATCAGATTTCG; E61A, GAAATGGGGGATCCAGCTGAAAACCTTAGG and CTTTACCCCCTAGGTCGACTTTTGGAATCC; K69A, GGCTTTAGTCTAGCGCTAAGTGCTGCTGAAAATG and ACGAAATCAGATCGCGATTCACGACTTTTAC; E73A, CTAAAGCTAAGTGCTGCAAATGACTTTAGC and GATTTCGATTCACGACGTTTACTGAAATCG; D75A, GTGCTGAAAATGCCTTTAGCAGTGATAG and CACGACTTTTACGGAAATCGTCACTATC; S77D/D74A, GCTGAAAATGCCTTTGACAGTGATAGCCCAGAG and CGACTTTTACGGAAACTGTCACTATCGGGTCTC; D79A, GACTTTAGCAGTGCTAGCCCAGAGAG and CTGAAATCGTCACGATCGGGTCTCTC; E82A, GTGATAGCCCAGAGCGAAAAATGCTTCC and CACTATCGGGTCTCGCTTTTTACGAAGG; R83A, GTGATAGCCCAGAGGCAAAAATGCTTCCCTG and CACTATCGGGTCTCCGTTTTTACGAAGGGAC; K84A, GATAGCCCAGAGAGAGCAATGCTTCCCTG and CTATCGGGTCTCTCTCGTTACGAAGGGAC; H130A, AATGCTTAACCTTGCTGCAGGGTCACAAAAAG and GTACGAATTGGAACGACGTCCCAGTGTTTTTC; H137A) GGTCACAAAAAGTGGCTGAGCATGGTGGAGG and CCAGTGTTTTTCACCGACTCGTACCACCTCC; R281A, GGAACACAACAGTGGGCAGGCCTTGCAAGATATTTTAAG and CCTTGTGTTGTCACCCGTCCGGAACGTTCTATAAAATTC; R281K, GGAACACAACAGTGGAAAGGCCTTGCAAGATATTTTAAG and CCTTGTGTTGTCACCTTTCCGGAACGTTCTATAAAATTC; and G282A, CAACAGTGGAGAGCCCTTGCAAGATATTTTAAG and GTTGTCACCTCTCGGGAACGTTCTATAAAATTC (nucleotides in boldface denote mismatched mutations). To confirm all mutations, sequencing reactions containing 480 ng of plasmid and 8 pmol of primer (5′-TTG TCT GTA AGC GGA TGC-3′) were sent to Genewiz, North Brunswick, NJ.

Indirect immunofluorescence.

Vero cells were fixed with 2% paraformaldehyde (EMB) and permeabilized by using 0.5% Triton X-100 (Shelton Scientific). The cells were incubated with the anti-V antigen (V-Ag) monoclonal antibody PAB597 (which was made against SV40 and cross-reacts with both JCV and BKV VP1) and Alexa Fluor 488-labeled goat anti-mouse antibody (Molecular Probes), counterstained with 0.02% Evan's Blue (red cytoplasmic dye) for 1 min, and visualized by using a Nikon epifluorescence microscope (Eclipse E800). Approximately 8,000 cells were screened for V-Ag expression.

RNA isolation and PCR.

RNA was extracted from cells 4 days posttransfection by using RNeasy minikit (QIAGEN). cDNA was generated by using oligo(dT) primers from iScript select cDNA synthesis kit (Bio-Rad) and 1 μg of extracted RNA.

PCR.

For conventional PCR, 1 μl of cDNA or 1 μl of eluted DNA was mixed with 0.5 μg of each primer (5′-TGA GAC TTG GGA AGA GCA TTG TG-3′ and 5′-TGA AGA TGT AAA AGG GAC AGG AGC-3′) and reagents in Taq polymerase PCR kit (Promega) and then cycled at 94°C for 1 min once; followed by 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 1 min; and held at 4°C. On a 2% agarose gel, 5 μl of each reaction was loaded per well. The bands were visualized using UV setting on ChemiDocXRS (Bio-Rad). For real-time quantitative PCR (RQ-PCR) experiments, the protocol of McNees et al. (21) was followed. Briefly, 10 μl of eluted DNA, 50 μl of iScript TaqMan universal master mix (Bio-Rad), 200 nM TaqMan FAM-MGB BKV specific probe (5′-AGT GTT GAG AAT CTG CT-3′) (Applied Biosystems), and 0.5 μg of each large T-antigen primer (5′-CTT TCT TTT TTT TTT GGG TGG TGT T-3′) and (5′-TTG CCA GTG ATG AAG AAG CAA-3′) were mixed in a final volume of 100 μl. For each condition, triplicate samples containing 25 μl each were loaded onto a 96-well PCR plate (Bio-Rad). For the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) control, 5 μl of eluted DNA was mixed with 50 μl of SYBR green reagent (Applied Biosystems) and 0.5 μg of each primer (5′-GGT AAG GAG ATG GCT GCA TTC G-3′ and 5′-AGG CAA GAA GGC ATG AGG AG-3′) in a total volume of 100 μl and run in triplicate. Both reactions were run at 50°C for 2 min and 95°C for 10 min, followed by 50 cycles of 94°C for 15 s and 60°C for 1 min. The threshold cycle (C_T) value was calculated by using Bio-Rad iQ5 optimal system software.

Infections.

Virus-containing supernatants were collected at 5 and 19 days posttransfection. Vero cells were treated with 200 μl of supernatant for 2 h at 37°C, followed by the addition of fresh media. Three days later, cells were stained for V-Ag.

DNase protection assays.

Four days later after transfection, cells and media were lifted by mechanical scrapping and centrifuged at 4,400 rpm for 10 min. The cell pellet was suspended in 600 μl of medium, three freeze-thaw cycles were conducted, and 60 μl of 2.5% deoxycholic acid (Fisher Biotech) was added, followed by incubation at 37°C for 30 min. After centrifugation at 10,000 rpm for 30 min at 4°C, the supernatant was collected, and the DNA was content determined with a Biophotometer spectrophotometer (Eppendorf). To degrade free DNA not encapsidated within virions, 10 μg of DNA was treated with 250 U of DNase I (DNase I) or water (Invitrogen) for 15 min at room temperature. Both samples were treated with 12.5 μl of 0.5 M EDTA and incubated for 10 min at 65°C to inactivate the DNase I. To release viral DNA (vDNA) from capsids, proteinase K was used as described in the blood and body fluid spin protocol of the QIAamp DNA blood minikit (QIAGEN). DNA was eluted in 40 μl of sterile water. PCR as described above was used to determine level of vDNA.

Cell surface binding assays.

For flow cytometry experiments, labeled BKV was prepared by mixing 50 μg of Alexa Fluor 488 dye (Molecular Probes) with 1 ml of CsCl purified virus harvested 20 days postinfection for 1 h at room temperature and then dialyzed in buffer A (10 mM Tris, 50 mM NaCl, 0.01 mM CaCl₂) for 2 days. Vero cells in suspension were treated with phosphate-buffered saline (PBS) or 0.25 U of neuraminidase (NA)/ml from Vibrio cholerae (Sigma-Aldrich) for 1 h at 37°C and washed three times with PBS. A 1:20 dilution of labeled virus was added to NA- or PBS-treated cells for 1 h on ice, and then the cells were washed three times to remove unbound virus. The fluorescence intensity was determined by using a FACScaliber, and the data were analyzed using CellQuest software (Becton Dickinson). To test mutant binding, two 75-cm² flasks were each transfected with 64 μg of the WT or the mutants R281K, E61A, K69A, H137A, R83A, and K84A. Five days later, the cells and 3 ml of medium were collected, sonicated for 3 min, and treated for 30 min at 37°C with 330 μl of 2.5% deoxycholate acid. The supernatant was layered over 20% sucrose in buffer A containing 0.01% Triton X-100 and spun at 35,000 rpm for 3 h. The pellet was suspended in 2.5 ml of buffer A containing 0.01% TritonX-100 and layered over a gradient of 1.35, 1.32, 1.29, 1.26, and 1.23 g of CsCl/ml (with the highest concentration at the bottom) and spun at 33,000 rpm for 18 h. At the interface where the virus was layered, 0.5 ml was collected and dialyzed in buffer A. To validate viral binding using RQ-PCR, cells were treated with PBS or 0.25 U of NA/ml as described above, washed three times with PBS, and incubated with 20 μl of CsCl-purified WT virus and 480 μl of EMEM for 1 h on ice. The cells were washed with PBS three times to remove unbound virus. To remove any nonencapsidated vDNA, the cell pellet was treated with 250 U of DNase I (Invitrogen) for 15 min at room temperature. The DNase I was inactivated, and then DNA was eluted as described in the DNase protection assay described above. RQ-PCR as described above was conducted. To compare the binding of each mutant versus the amount of virus in each purified preparation, 20 μl of virus was mixed with 480 μl of EMEM and added to Vero cells and either incubated for 1 h on ice (bound) or divided into aliquots into a tube without cells (input). Cells were washed three times to remove any unbound virus. As above, cells or input virus was treated with DNase I, the DNA was eluted, and the expression of vDNA was determined by using RQ-PCR. The normalized amount of vDNA associated with cells was divided by the amount of vDNA input.

RESULTS

Single amino acid point mutations in VP1 BKV alter infectivity. To determine which amino acids of BKV VP1 were potentially important for receptor binding, we generated single amino acid mutations of 17 amino acids on the BC, HI, and DE loops as indicated (Fig. 1, 2, and 6C). An additional set of mutants was generated to mimic polymorphisms seen in the four major BKV subtypes. After the mutations in VP1 were generated, the transfection efficiency and expression of mutant VP1 was investigated. Vero cells were plated on coverslips and transfected with 5 μg of linear BKV-pUC19 (WT) or mutant DNA (R64A, R64K, E73A, D75A, K69A, K84A, H137A, G282A, E82A, H130A, E61A, G65A, D79A, S77D/D75A, R83A, R281A, or R281K). Four days posttransfection VP1-specific mRNA transcripts were found in all cells transfected with WT or mutant DNA by reverse transcriptase-PCR (Fig. 2A). The cells were then fixed and stained with a V-antigen-specific monoclonal antibody (PAB597), and VP1 was found to correctly localize to the nucleus in all of the mutants tested (Fig. 2B). V-antigen-positive cells were scored, and the results were graphed as a percentage of the WT transfected cells. We found reduced VP1 expression for the mutant genomes E82A, H130A, E61A, G65A, and D79A compared to WT level (Fig. 2C). This suggests a potential defect or instability in VP1 protein expression. By scoring T-Ag expression 4 days posttransfection, all mutants were shown to have very similar transfection efficiencies (data not shown).

FIG. 2.

Mutations in the BKV sequence do not disrupt VP1 transcription or translation. Vero cells were transfected with 5 μg of linear WT or mutant DNA. (A) Four days later, RNA was extracted and reverse transcribed into cDNA, and VP1 was amplified by PCR. The VP1 transcript was present for the WT and every mutant. (B) Cells on coverslips were fixed, permeabilized, and stained for V-Ag. All mutants show nuclear VP1 expression. Magnification, ×400. (C) The number of V-Ag expressing cells was counted and compared to the WT. Error bars are standard deviations from three independent experiments.

FIG. 6.

Binding of BKV mutants to host cells. (A) NA treatment of Vero cells reduced the level of BKV binding to cell surface. Vero cells were treated with PBS (shaded and light gray line) or 0.25 U of NA/ml from V. cholerae (dark gray line) and washed with PBS. Cells were incubated with PBS (shaded) or Alexa Fluor 488-labeled virus (light and dark gray lines), and the cell surface fluorescence determined by using flow cytometry. (B) Cells were treated with PBS or NA, washed, and incubated with unlabeled WT BKV that was CsCl purified 5 days posttransfection. After a washing step to remove unbound virus, cells were DNase I treated, and eluted vDNA was evaluate by using RQ-PCR. (C) Cells were incubated with each CsCl-purified mutant virus collected 5 days posttransfection and then washed to remove unbound virus. Virus associated with cells (bound) or virus alone (input) was treated with DNase I, the DNA was eluted, and the vDNA levels were quantified by RQ-PCR. Cell-associated vDNA was normalized by using GAPDH primers. The expression of bound vDNA was divided by input vDNA. These values were compared to WT and converted into a percentage.

To determine whether any of the mutants were defective in viral infectivity and spread, the number of infected cells was monitored over time after the transfection of vDNA. Vero cells plated on coverslips were transfected with 1 μg of WT or mutant DNA and stained for V antigen every 3 days (Fig. 3A). When WT vDNA was transfected into cells, virus was produced and quickly spread throughout the culture, infecting 80% of the cells by 19 days posttransfection (Fig. 3B). The S77D/D75A double mutant that mimics the polymorphism seen in subtypes II, III, and IV propagated like WT, indicating that in vitro subtypes II, III, and IV have fitness comparable to WT (Fig. 3A). All of the positively charged residues, when mutated to alanine, significantly reduced viral spread (K69A, R83A, and K84A) or were nonviable (R281A and R64A) (Fig. 3A). The mutant R281K showed lower infectivity despite conservation of the positive charge. Several mutations produced nonviable virus (E61A, H137A, R64A, D79A, R281A, H130A, G282A, and G65A). Alanine mutations in either amino acid of the two RG motifs of the BC and HI clefts rendered the genome noninfectious. Two histidine residues on the DE loop were also necessary for infectious spread (Fig. 3A). Representative images show the level of infection at 19 days posttransfection for each mutation (Fig. 3B). Table 1 summarizes the growth curves of each mutant. One possibility to account for our observations is that some of the mutations may have affected the anti-V antigen epitope in VP1. To control for this, we also stained the cultures with an antibody to the early protein, T antigen, and obtained identical results, indicating that the antibody epitope was not affected by any of the mutations (data not shown). To directly assess the infectivity of each mutant, media were collected from each culture at 19 days posttransfection and used to infect Vero cells. As expected, the cell culture supernatant isolated from cells transfected with BKV mutants with steep growth curves was more infectious than the supernatants from slower-growing viruses (Fig. 3C).

FIG. 3.

Multiple mutations in VP1 alter the infectious spread of BKV in cell culture. (A) Vero cells were transfected with 1 μg of linearized WT or mutant DNA. Viral spread was assessed by counting V-Ag positive cells every 3 days until day 19 posttransfection. The graph shows the growth curve of each mutant virus. The arrangement of the mutations in the legend (top to bottom) directly correlates with the level of viral infection on day 19 (highest infection to lowest). (B) Representative indirect immunofluorescent image of V-Ag expression (green) on day 19. All cells are stained red using Evan's Blue cytoplasmic dye. Magnification, ×100. (C) Infectious titer of WT and mutant viruses. The supernatant collected from the growth curve experiment shown in panel A was harvested at 19 days posttransfection and used to infect Vero cells. Vero cells were scored 3 days later for expression of V-Ag by indirect immunofluorescence assay.

TABLE 1.

Summary of BKV VP1 mutants

Mutant	Loop location	Alteration typea	Growth phenotype	Protection of DNA from DNase I	Binding to host cells
WT			+++	+	+
R64A	BC	RG motif	−	−	NTb
R64K	BC	RG motif	+++	+	NT
E73A	BC	− charge	++	+	NT
D75A	BC	− charge	+++	+	NT
K69A	BC	+ charge	+	+	−
K84A	BC	+ charge	+	+	−
H137A	DE	+ charge	−	+	−
G282A	HI	RG motif	−	−	NT
E82A	BC	− charge	+++	+	NT
H130A	DE	+ charge	−	−	NT
E61A	BC	− charge	−	+/−	+++
G65A	BC	RG motif	−	−	NT
D79A	BC	− charge	−	−	NT
S77D/D75A	BC	Subtype polymorphism	+++	+	NT
R83A	BC	+ charge	+	+	++
R281A	HI	RG motif	−	−	NT
R281K	HI	RG motif	+	+	+

^a− charge, negatively charged amino acid; + charge, positively charged amino acid.

^bNT, not tested.

A subset of mutant viruses have structural or packaging defects.

VP1 is the main structural protein encapsidating the BKV DNA genome. It is possible that an amino acid mutation in the external loops of VP1 may cause a structural defect within the virion, preventing efficient genome packaging or virion assembly. To test this, we isolated cell culture supernatant 5 days after cells were transfected with 5 μg of WT or mutant DNA. The virus-containing supernatant was treated with either water or DNase I to digest any nonencapsidated DNA. The capsid was then degraded by proteinase K, and the amount of vDNA protected from DNase digestion was determined by PCR (Fig. 4). For WT transfectants the majority of vDNA was protected from digestion, suggesting that the DNA is encapsidated by viral proteins (Fig. 4). WT DNA that was not BamHI digested (WT undigested) prior to transfection does not form virions because the backbone plasmid is positioned in the center of the VP1 sequence. The intensity of the PCR band for WT undigested VP1 was ∼90% reduced upon DNase exposure. This indicated that the concentration of DNase I used degraded most of the input vDNA. The mutants that exhibit WT-like spread (E73A, R64K, E82A, D75A, and S77D/D74A) effectively protected their DNA, suggesting that the vDNA in the supernatant primarily exists within capsids. Most VP1 mutations showing low viability did not effectively protect their genome (R281A, H130A, R64A, G65A, D79A, and G282A). A lack of protection from DNase I suggests that the capsid is malformed. Mutants displaying moderate spread of virus infection (K84A, R83A, K69A, and R281K) and the H137A mutant protect their genome from DNase I as effectively as WT (Fig. 4 and Table 1).

FIG. 4.

Mutant viruses exhibit sensitivity to DNase I. Five days after transfection with 5 μg of linear WT or mutant BKV DNA virions are released into the media supernatant. (A) This supernatant was treated with DNase I to digest free DNA or with proteinase K to digest capsid proteins, and the remaining DNA was PCR amplified. Without DNase I, the total input BKV DNA is amplified. DNase I treatments digest all DNA except that encapsidated by protein. The amount of DNA protected by capsid protein was visualized on agarose gel. (B) Capsid protection of DNA was quantified by first determining densitometry of each band. The reduction in band intensity after DNase I treatment was compared to its no-DNase I control and converted into a percentage.

The mutants K84A, R83A, K69A, R281K, H137A, and E61A exhibited reduced BKV spread that cannot be explained by a defect in VP1 production, localization, or packaging of the genome. To determine whether these mutants have a defect in lytic escape from cells, we indirectly examined the level of virus released into cell culture supernatant. Five days after transfection with WT or mutants, vDNA was isolated from DNase I- and proteinase K-treated supernatants and amplified by using PCR. At this time point, only a single infectious life cycle is possible. We found that the supernatant extracted from the mutants K84A, R83A, K69A, R281K, H137A, E73A, and E61A contained similar levels of vDNA as WT (Fig. 5A). RQ-PCR revealed no statistically significant difference in V-Ag expression (data not shown; P > 0.35 as determined by two-tailed test comparing each mutant to WT, except for E61A [P = 0.14]). This indicates that there is no obvious defect in viral exit and that the supernatant of each mutant contained the same amount of virus. Next, the supernatant was added to Vero cells, and the V-Ag expression was determined 3 days later (Fig. 5B). The supernatants from these mutants were less infectious than WT. Because of this phenotype a defect in BKV binding or early cellular trafficking steps seemed likely.

FIG. 5.

Release and infectivity of mutants. Vero cells were transfected with 5 μg of linear WT or mutant DNA. (A) Five days later, medium supernatant was treated with DNase I, proteinase K, and VP1 DNA amplified via PCR. This revealed that these mutants had amounts of vDNA encapsidated in protein similar to that of the WT. The WT undigested control showed that the amount of DNase I used here could degrade input DNA. (B) Untreated medium supernatant also isolated 5 days posttransfection was used to infect Vero cells. Three days later V-Ag-expressing cells were scored. These mutants showed reduced infectivity compared to the WT.

A subset of mutants are deficient in binding to cells.

NA is an enzyme that cleaves the bond between sialic acid and galactose, which releases sialic acid. Using flow cytometry, Vero cells treated with 0.25 U of NA/ml showed reduced binding of WT Alexa Fluor 488-labeled BKV (Fig. 6A). To evaluate the binding of a limited amount of WT and mutant viruses that result from one round of transfection, we developed a novel PCR-based approach to measure cell-associated virions. To validate the assay and to be sure we were measuring vDNA associated with virions bound to cells and not contaminating vDNA from the initial transfections, we treated cells with 0.25 U of NA/ml or PBS, washed them, and then incubated them with WT CsCl-purified virus collected 5 days posttransfection. The cells were then treated with DNase I to degrade any free vDNA. Using RQ-PCR, the expression of vDNA associated with cells after several wash steps was reduced by ∼60% when cells were treated with NA compared to PBS (Fig. 6B). To test cell surface attachment of BKV VP1 mutants, each CsCl purified mutant virus was either added to Vero cells and allowed to bind for 1 h at 4°C (bound) or placed in a tube without cells (input). After cells were washed to remove any unbound viral particles and treated with DNase I, virion-associated vDNA was evaluated by using RQ-PCR. The input vDNA was compared to the vDNA associated with the Vero cells and converted to a percentage (Fig. 6C). The mutants K84A, K69A, and H137A had significantly less vDNA associated with cells than vDNA input, suggesting a cell surface binding defect. In addition, R83A and E61A mutants seemed to bind to cells better than the WT, suggesting this mutant is more adhesive or attaches more firmly to cells (Fig. 6C and Table 1).

Molecular modeling reveals a stable receptor binding pocket.

Based on the homology model of BKV VP1, the BC loops, which contain multiple charged amino acids, show that positive charges (indicated in yellow) are directed toward the center depression, while the negative residues (indicated in green) project toward the outer loop regions (Fig. 7C). The two RG motifs, common in sialic acid-binding polyomaviruses, are indicated in black (Fig. 7C). The homology-based model indicates that NeuNAc-(α2,3)-Gal-(β1,3)-GlcNAc, found on N-linked glycoproteins (Fig. 7A), lies flat in a shallow groove formed by the BC and HI loops (Fig. 7B and C). Our simulation results shows that Arg-281 forms hydrogen bonds with the sialic acid portion of the receptor, thereby defining the relative positioning of VP1. In addition, the model shows Lys-69 interacting with the galactose ring of the receptor. In contrast, the side chains of Arg-64 and Glu-61 project away from the receptor-binding pocket, forming a salt bridge that contributes to the open topology of the BC loops, as well as receptor accessibility to this region.

FIG. 7.

Oligosaccharide receptor binding pocket on BKV VP1. (A) Structure of α(2,3)-linked sialic acid-containing receptor [NeuNAc-(α2,3)-Gal-(β1,3)-GlcNAc]. Oxygen molecules are labeled in red, nitrogen molecules are labeled in blue, and carbon molecules are labeled in yellow. The carbons of each ring are numbered. The carboxylate group of sialic acid is located off of carbon 1 of the NeuNAc molecule. (B) Generated using Insight II software, a space-filling representation of VP1 illustrates how the sialic acid and galactose molecules of the receptor fragment interact flat and shallow on the surface of VP1. (C) Proposed model of receptor engagement to VP1. By using a SWISS-MODEL protein server, the structure of BKV VP1 was generated based on 88% homology to the crystal structure of SV40 and MD simulations conducted using GROMACS 3.3 software. The carbon atoms of the sialic acid-containing receptor were colored royal blue to enhance contrast. The receptor orientation remains the same as in panel A. All mutated residues are labeled in black text. Within the BC, HI, and DE loops, positive amino acids are yellow, the negative amino acids are green, and the arginine-glycine (RG) motifs are black. Side chains were included for any charged residue on these loops. Numbers denote the amino acid position in VP1. Boldface text indicates the VP1 loops.

DISCUSSION

Site-specific mutagenesis, coupled with infectivity assays, was used to identify amino acids in the major capsid protein of BKV that were important for viability. A summary of these findings are found in Table 1. The VP1 molecules of BKV and SV40 are structurally similar, and alignment reveals 88% amino acid homology and 82% amino acid identity (Fig. 1) (14). SV40 has two KG motifs within the cleft formed by the HI and BC loops instead of the RG motifs that are found in BKV, mPyV, and JCV at similar positions. To address whether lysine is interchangeable with arginine at these locations, we mutated both RG motifs of BKV to KG motifs. The R64K mutation behaved like WT. In contrast, the R281K mutation causes a defect in infectious spread, but the virions are structurally sound. We found that R281K virus could bind to host cells as well as WT, indicating there is no receptor-binding defect. The KG motif mutants can properly assemble into virions and seem to support receptor binding as well as RG motifs.

Here we propose a potential receptor-binding mode in which Arg-281 interacts directly with the carboxylic acid group of sialic acid and largely defines its position on the capsid. The R281A mutation is structurally unstable, making binding of this mutant impossible to test. The conservative R281K mutation can still bind to host cells largely due to the similar charge properties of arginine and lysine. The simulations indicate the receptor can stably interact with a groove found between the HI and BC loops, in accord with the observed importance of the RG motif in sialic acid binding to mPyV. In our model, the Lys-69 motif interacts with the terminal (O₂) hydroxyl group of the galactose ring. In mPyV, the carboxylate group of the sialic acid binds the arginine higher up on the ascending inner region of the BC2 loop (32). According to the homology model, this is unlikely for BKV because the side group of Arg-64 is deeper in the pocket and turned inward, making it inaccessible to the sialic acid receptor. Crystalization studies of BKV in complex with its sialic acid receptor should reveal the accuracy of our model and confirm the strength of this approach.

Because of the low yield of mutant viruses, we devised and validated a novel assay that detects the association of vDNA with cells and compares this to input vDNA. Our binding studies reveal that three VP1 mutations (K69A, K84A, and H137A) have >75% reduced binding compared to WT. In our model, we predict Lys-69 interacts with part of the carbohydrate receptor. However, two mutations in VP1 (K84A and H137A) also have a role in binding to the host cell that is not directly apparent in Fig. 6. The model shows only a fragment of the sialic acid-containing receptor. Depending on the bend of the remaining carbohydrate chain at C6 of GlcNAc, Lys-84 and/or His-137 from an adjacent monomer (not shown in the model) could interact with upstream portions of the receptor, explaining their binding-deficient phenotype. We cannot exclude the possibility that the receptor binds in a completely different orientation. For example, the receptor may lie across the BC loops and interact with Lys-69 and Lys-84 and not the HI loop (Fig. 6C). It is also important to consider a model does not reflect a stepwise process of attachment and that several transient receptor orientations may be occurring before a stable final conformation is achieved. The GD1b and GT1b gangliosides containing terminal α(2,8)-linked and α(2,3)-linked sialic acids also act as cellular receptors in BKV infection (20). Although not modeled here, we predict the α(2,3) sialic acid linkage of GT1b to bind similarly to our predicted model because both receptors contain the terminal fragment, NeuNAc-(α2,3)-Gal-(β1,3). An alternative binding position for the NeuNAc-(α2,8)-NeuNAc-(α2,3)-Gal-(β1,3) fragment of GD1b and GT1b gangliosides might explain the role of other motifs that are not fully understood in our model.

BKV is divided into four subtypes, of which subtype I is the most prevalent. The subtypes are determined by the VP1 sequence between amino acids 61 and 83, which is the variable antigenic region and maps to the BC1 and BC2 loops. Despite the overall homology of amino acid between subtypes, one conspicuous difference between subtype I and subtypes II, III, and IV is a change in the location of a negative amino acid from positions 75 to 77 (15, 16). To address whether subtype I is more prevalent because it is better able to attach to its sialic acid receptor, we generated the double mutant S77D/D75A. This virus behaved identically to WT, suggesting that the BC loops of subtypes II, III, and IV bind to a sialic acid receptor as well as subtype I (Table 1). We predict that the prevalence of subtype I within the human population cannot be explained by an advantage in sialic acid receptor binding.

The roles specific amino acids in the external loops of BKV VP1 play in the infectious life cycle of BKV were extensively evaluated and modeled. Some mutations caused capsid instability that prevented viral spread (R281A, H130A, R64A, G65A, D79A, G282A, and E61A). We believe that these motifs are critical for the structural integrity of the virion. This does not rule out the possibility that these amino acids also play a role in receptor interactions. Other mutants generated structurally sound virus, but the progeny virus failed to be fully infectious. Binding data revealed the reason for this phenotype for some mutants (K69A, H137A, and K84A) was due to defective attachment to host cell. Interestingly, the R83A and E61A mutant seems to bind too tightly to Vero cells. This may prevent productive receptor interactions and explain a reduced infectious spread over time. The reason for the reduced infectivity for other mutants (R281K) remains to be determined.