Functional Genetic and Biophysical Analyses of Membrane Disruption by Human Adenovirus

Abstract

The identification of the adenovirus (AdV) protein that mediates endosome penetration during infection has remained elusive. Several lines of evidence from previous studies suggest that the membrane lytic factor of AdV is the internal capsid protein VI. While these earlier results imply a role for protein VI in endosome disruption, direct evidence during cell entry has not been demonstrated. To acquire more definitive proof, we engineered random mutations in a critical N-terminal amphipathic α-helix of VI in an attempt to generate AdV mutants that lack efficient membrane penetration and infection. Random mutagenesis within the context of the AdV genome was achieved via the development of a novel technique that incorporates both error-prone PCR and recombineering. Using this system, we identified a single mutation, L40Q, that significantly reduced infectivity and selectively impaired endosome penetration. Furthermore, we obtained biophysical data showing that the lack of efficient endosomalysis is associated with reduced insertion of the L40Q mutation in protein VI (VI-L40Q) into membranes. Our studies indicate that protein VI is the critical membrane lytic factor of AdV during cellular entry and reveal the biochemical basis for its membrane interactions.

Studies of the cell entry pathways of diverse families of nonenveloped viruses, including reovirus, picornaviruses, rotavirus, and Flock House virus, have revealed several common themes for how these viruses penetrate cell membranes during infection (3, 50). Nonenveloped virus assembly initially generates stable, immature particles that undergo proteolytic processing to form metastable intermediates. During infection, cues from the host cell trigger additional conformational changes that lead to exposure of hydrophobic moieties in the virus that breach the limiting cell membrane (i.e., plasma, endosomal, or endoplasmic reticulum [ER] membrane). Human adenovirus (HAdV) cellular entry closely parallels this general model. Several preproteins in the AdV virion are cleaved by the virally encoded cysteine protease during capsid maturation to prime the virus for entry into and infection of a new host cell (9, 18). Infection is initiated by a high-affinity interaction between fiber and one of several cellular receptors (e.g., CAR, CD46, or sialic acid) (2, 5, 48). Secondary interactions of conserved RGD motifs on the penton base with cellular integrins trigger clathrin-mediated endocytosis and loss of the fiber protein near the cell surface (25, 29, 56). Further stepwise disassembly of the capsid occurs concomitant with endosome acidification as penton base, peripentonal hexons, and the minor proteins IIIa and VI are shed from the incoming virion (19, 39, 45, 57). The virus escapes the endosome, trafficks along microtubules, and docks with the nuclear pore complex, where the viral genome translocates into the nucleus (16, 17, 36, 59).

One of the key steps in nonenveloped virus entry is disruption of the limiting membrane to allow the incoming virus particle access to the cytosol. For AdV, this entry step, which occurs within the endosome, is poorly understood. Nonetheless, it is known that partial disassembly of the viral capsid and release of a membrane lytic factor from the interior of the capsid are required for endosome penetration (19, 43, 57). A temperature-sensitive mutant, AdV2ts1, is properly internalized into cells yet fails to escape the endosome because the membrane lytic factor is sequestered inside a hyperstabilized capsid formed by the absence of maturational proteolytic processing (18, 34, 54). Although this lytic agent was historically believed to be the penton base protein, increasing evidence from recent studies now indicates that the endosomal membrane penetration protein for AdV is the internal capsid protein VI (27, 37, 38, 55, 57).

Protein VI is present in the mature AdV virion in ∼360 copies, and structural and biochemical data suggest that it resides within the sizeable internal cavity of each hexon trimer (26, 35, 41, 47, 51). Like many viral proteins, VI has multiple functions in the AdV life cycle. Preprotein VI is a substrate for the AdV L3/23K protease, and a C-terminal fragment of VI released by this cleavage functions to enhance L3/23K proteolytic activity (28). In addition, VI functions as an adaptor protein for hexon nuclear import during the assembly phase of viral replication (60). Most recently, VI was reported to aid in the microtubule-dependent transport of incoming virions toward the nucleus via a PPxY motif and recruitment of E3 ubiquitin ligases (61). As one of the four different cement proteins of AdV, VI is also thought to play a role in stabilizing the viral capsid (47). The precise location of protein VI inside the outer capsid shell and its copy number have yet to be accurately determined.

A role for protein VI in endosomal membrane penetration has been suggested in large part by a series of in vitro studies. Most notably, the ability of heat-dissociated virions to lyse model liposomes is abrogated upon VI immunodepletion (57). Anti-VI antibodies capable of neutralizing the membrane-permeabilizing activity of VI are also able to reduce AdV permeabilization of endosomal membranes (27). In addition, a recombinant version of VI binds to and disrupts membranes in a pH-independent manner, and this activity depends upon an N-terminal amphipathic α-helix (residues 34 to 54) (27, 57). Furthermore, recombinant VI was recently shown to bind to model membranes in an oblique orientation and likely provokes membrane fragmentation by inducing positive membrane curvature (27).

While these in vitro studies strongly suggest that VI is the endosomalytic factor for AdV, direct evidence of a role for VI in virus cell entry is lacking. We therefore generated mutant AdVs that shed light on the role of this capsid protein in endosome penetration, using a combination of random mutagenesis and recombineering targeted to the amphipathic α-helix of protein VI. We identified a virus encoding a single point mutation in protein VI, AdV VI-L40Q, that exhibits a significant reduction in both virus infectivity and endosome penetration. Cellular trafficking of the L40Q virus is also perturbed, as we observe prolonged residence of the mutant virus in early endosomes concomitant with a significant decrease in nuclear localization. We further show that a recombinant version of VI harboring the L40Q mutation exhibits attenuated membrane binding and significantly altered insertion properties. Taken together, these analyses provide convincing evidence that protein VI is the major mediator of endosome penetration during AdV cellular entry.

MATERIALS AND METHODS

Cell lines and viruses.

Tissue culture reagents were obtained from Invitrogen. A549 cells (ATCC) and 293β5 cells were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (44). The bacterial artificial chromosome (BAC) vector originally used to produce wild-type and mutant AdVs for these studies (pAd5-GFPn1) has been previously described (61). This construct carries the HAdV5 genome and is replication defective, owing to deletion of E1/E3 genes and insertion of a cytomegalovirus (CMV)-driven enhanced green fluorescent protein (EGFP) reporter cassette. Virus stocks were generated by transfecting PacI-linearized pAd5-GFPn1 into E1/E3-complementing 293β5 cells, followed by amplification in serial passages. Purified virus was isolated from cellular lysates by double banding in CsCl gradients and dialyzed into A195 buffer (12). Aliquots were flash frozen in liquid nitrogen and stored at −80°C.

Random mutagenesis.

The construction of AdVs using recombineering has been previously described and was adapted for this protocol (53, 61). A mutagenic PCR product (154 bp) carrying residues 34 to 53 of protein VI (54 bp) plus 50 nucleotides 5′ and 3′ of the helix (100 bp total) was generated using the GeneMorph II random mutagenesis kit (Stratagene) and the following primers: AH-mutF (5′-TGGGAAACTGGCAAGATATCGGCACCAGCAATATGAGCGGTGGCGCCTTC) and AH-mutR (5′-TTATCCCTCAGCATCTGGCCTGTGCTGCTGTTCCAGGCCTTGCTGCCATA). The PCR template used was HAdV5 protein VI. Based on the manufacturer's instructions, 0.1 or 1 ng of template DNA/reaction was used to achieve a mutation rate of 12 to 16 mutations/kilobase of DNA, or ∼1 mutation/154 bp PCR product. The mutagenic PCR product was incorporated into the AdV genome via homologous recombination in bacteria harboring pAd5-GFPn1-ΔAHgalK. This vector contains a substitution of the galactokinase (galK) cassette of pGalK for the nucleotides encoding residues 34 to 53 of VI and was generated via homologous recombination in bacteria harboring pAd5-GFPn1 with a product from PCR amplification of galK using the following primers: AH-galK-F (5′-GGGAAACTGGCAAGATATCGGCACCAGCAATATGAGCGGTGGCGCCTTCCCTGTTGACAATTAATCATCGGCA) and AH-galK-R (5′-TTATCCCTCAGCATCTGGCCTGTGCTGCTGTTCCAGGCCTTGCTGCCATATCAGCACTGTCCTGCT CCTT) (53). The galK-specific sequences are shown in italics, and protein VI sequences 5′ and 3′ of the helix are underlined. DNA from individual clones was isolated and sequenced to identify mutations. Virus stocks were generated as described above and sequenced again to verify the mutations.

Single-round infection.

A549 cells were infected with serial dilutions of AdV or AdV with mutated VI and incubated for 48 h. Cells were harvested, washed, and resuspended in FACS buffer (phosphate-buffered saline [PBS]-1% FBS-0.2% sodium azide), and the percentage of GFP-positive cells was determined by flow cytometry using a BD LSRII (BD Biosciences). Data were analyzed with FlowJo software (Tree Star, Inc.). Curves of the resulting data were used to calculate the 50% infectious dose (ID₅₀) with the nonlinear regression function using GraphPad Prism (La Jolla, CA).

Thermostability assay for AdV uncoating.

AdV thermostability was assessed using a protocol similar to that which has been previously described (43). AdV and AdV VI-L40Q (1.5 μg) were diluted such that final buffer conditions used were 0.05% bovine serum albumin (BSA), 50 mM NaCl, and 7.5 mM HEPES. The virus samples were then heated to 37, 40.8, 43, 46, 49.8, or 52°C for 10 min. After cooling to room temperature, the sample was loaded onto a 30 to 80% Histodenz (Sigma) discontinuous gradient and centrifuged in an SW55ti rotor (Beckman Coulter) for 1.5 h at 209,000 × g (average). Equivalent volumes of supernatant (top) and band (interface) fractions were taken from each gradient. The samples were boiled in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer containing dithiothreitol (DTT), separated by SDS-PAGE, transferred to nitrocellulose, and probed with a polyclonal rabbit antibody to protein VI (a gift from Harold Wodrich, Institute of Molecular Genetics of Montpellier, Montpellier, France) or the monoclonal anti-fiber antibody 4D2 (NeoMarkers/Lab Vision Corporation, Fremont, CA).

Sarcin assay for endosome rupture.

A549 cells (∼20,000 cells/well) in black 96-well plates were incubated in DMEM without cysteine or methionine supplemented with 10% dialyzed FBS [DMEM(−)] 1 h prior to infection. The medium was removed and replaced with 50 μl DMEM(−) containing 0.1 mg/ml α-sarcin (Calbiochem/EMD Biosciences, La Jolla, CA) and 3-fold serial dilutions (0.02 ng to 1000 ng) of AdV, AdV VI-L40Q, or AdV2ts1. After 2 h at 37°C, the cells were washed and transferred into DMEM(−) containing 1 μCi/well [³⁵S]methionine (PerkinElmer) and incubated for 2 additional hours at 37°C. The cells were washed twice with PBS and then incubated with ice-cold 5% trichloroacetic acid (TCA) for 1 h on ice. The plate was washed twice with 100% cold ethanol and then air dried overnight. Samples were solubilized with 10 μl/well of 1% SDS-0.1 N NaOH for 10 min, followed by neutralization with 2 μl/well of 0.6 N HCl. Microscint-20 (Packard) was added to each well, and [³⁵S]Met incorporation was measured using a TopCount (Packard) scintillation counter. The percent [³⁵S]Met incorporation for each data point was determined by dividing the total counts for each experimental well by the total counts in a control well containing [³⁵S]Met and α-sarcin but not AdV.

Immunofluorescence and confocal microscopy.

A549 cells plated on coverslips were incubated with AdV, AdVVI-L40Q, or AdV5-P137L (30,000 particles/cell) in DMEM for 1.5 h on ice (30). The samples were washed twice in DMEM to remove unbound virus and then shifted to 37°C. At the indicated times postinfection, the cells were fixed with 2% paraformaldehyde in PBS for 15 min. The coverslips were washed in PBS with 0.05% Tween 80 (PBST), followed by permeabilization/quenching in 20 mM glycine-0.5% Triton X-100-PBS for 20 min and being washed again. Samples were blocked with 2% BSA in PBST. Early endosomes were visualized by staining with rabbit anti-EEA1 (Cell Signaling Technology) primary antibody followed by an Alexa Fluor 488-conjugated anti-rabbit secondary (Invitrogen). AdV was labeled with a mouse monoclonal anti-hexon antibody (9C12) conjugated to Alexa Fluor 555 (Invitrogen) (52). The nucleus was stained with DAPI (4′,6-diamidino-2-phenylindole). Images were collected on a Zeiss 710 confocal microscope, and Z projections were generated with ImageJ. Prior to colocalization analysis, background noise was subtracted by thresholding the images based on controls without virus for the red channel and cells stained with the Alexa Fluor 488 secondary alone for the green channel (EEA1). The percentage of virus colocalized with EEA1 was calculated as follows: [PA]_R/G/[PA]_R × 100 where [PA]_R/G is the total area of pixels that are both red (virus) and green (EEA1) and [PA]_R is the total area of all red pixels. Nuclear colocalization was calculated the same way.

Generation of VI mutant constructs.

To study the effect of the L40Q mutation on the topology of the protein VI N-terminal amphipathic α-helix when bound to membranes, the L40Q mutation was introduced into 3 versions of pET15b-VI114Δ (encoding residues 34 to 114 of HAdV5 protein VI) in which 2 out of 3 tryptophan residues were previously mutated to phenylalanine (W37, W41, and W59) using the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA) (27). Three different mutants were generated using the primers indicated as follows: VIW37FL40Q, 5′ GCTTTGGCTCGCAGTGGAGCGGCATTAAAAATTTCG 3′ (L40Q/W41); VIW37FL40Q, W41F 5′ GCTTTGGCTCGCAGTTTAGCGGCATTAAAAATTTCG 3′ (L40Q/W59); VIW41FL40Q, 5′ GCTGGGGCTCGCAGTTTAGCGGCATTAAAAATTTCG 3′ (L40Q/W37). Complementary primers are not shown. Altered nucleotides are indicated in boldface.

The single point mutation L40Q was introduced in pET15b-VIhis, a construct encoding the mature form of protein VI (residues 34 to 239 of HAdV5 protein VI), using the following primer: L40Q 5′ GCTGGGGCTCGCAGTGGAGCGGCATTAAAAATTTCG 3′. All mutations were confirmed by sequencing.

Purification of recombinant proteins.

Recombinant proteins were expressed in BL21(DE3) cells. Cultures of cells harboring the plasmid pET15b-VIhis, pET15b-VIL40Qhis, pET15b-VIΔ54 (residues 54 to 239 of HAdV5 protein VI), pET15b-VI114ΔW37/L40Q, pET15b-VI114ΔW41/L40Q, or pET15b-VI114ΔW59/L40Q were grown at 37°C until they reached an optical density at 600 nm of 1.0 (57). The NaCl concentration was then increased by adding an additional 0.9 g NaCl/liter, and protein expression was induced by addition of 1 mM IPTG (isopropyl-α-d-thiogalactopyranoside) for 1 h. Cells were pelleted and resuspended in cell lysis buffer (1% Triton X-100, 25 mM phosphate, 150 mM NaCl at pH 7.5, 0.5 mg/ml lysozyme, 0.1 mg/ml DNase, and 1 mM PMSF [phenylmethylsulfonyl fluoride]), and soluble protein was isolated by centrifugation at 13,000 × g for 15 min at 4°C. Recombinant proteins were purified with Talon cobalt resin by using the manufacturer's protocol (BD Biosciences) and extensively dialyzed into 25 mM phosphate, 150 mM NaCl, and 10% (vol/vol) glycerol at pH 7.5 before being flash frozen in liquid nitrogen. Aliquots were stored at −80°C until use.

Determination of VI in vitro membrane lytic activity.

Liposomes containing 1-palmitoyl-2-oleoylphosphatidylcholine (POPC)-1-palmitoyl,2-oleoylphosphatidylserine (POPS) (75:25 mol%) (Avanti Polar Lipids) and entrapping 100 mM sulforhodamine B (SulfoB) (Molecular Probes) were generated by repeated extrusion of lipid dispersions through 0.2-mm polycarbonate filters 11 times as previously described (27, 57). Liposomes containing entrapped SulfoB were separated from free dye using a Sephadex G-75 column preequilibrated with 25 mM HEPES and 150 mM NaCl buffer, pH 7.5 (HBS). The liposome concentration was determined using a phosphate assay as previously described (13).

Membrane lytic activity of recombinant protein VI was determined by measuring SulfoB fluorescence dequenching upon release from liposomes. The liposomes were diluted in HBS to a final concentration of 10 μM. Different concentrations of VI were then added to the liposomes and incubated for 20 min at 37°C. Fluorescence intensity was measured using the Cary Eclipse fluorescence spectrophotometer (Varian), with an excitation wavelength of 575 nm and an emission wavelength of 590 nm. Maximal dye release was determined by adding Triton X-100 to the liposomes at a final concentration of 0.5% (wt/vol). The percentage of SulfoB released was calculated using the following formula: 100 × [(F_meas − F₀)/(F_tx100 − F₀)], where F_meas is the maximum fluorescence intensity measured, F₀ is fluorescence intensity in the absence of protein, and F_tx100 is the fluorescence intensity in the presence of 0.5% Triton X-100.

Determination of VI membrane affinity.

The binding to liposomes was assessed by monitoring changes in VI intrinsic tryptophan fluorescence upon titration with increasing amounts of liposomes (POPC/POPS, 75:25 mol%) The fluorescence emission spectra from 300 to 480 nm of VI in HBS at 37°C were obtained by selective excitation of tryptophan at 295 nm. The protein is diluted in HBS such that trace amounts of glycerol remaining from protein storage buffer (<0.5%) do not influence VI membrane binding. Increasing amounts of liposomes were added to VI, with mixing for 3 min, and additional spectra were obtained. Spectra of buffer or an equivalent amount of liposomes alone was subtracted from the spectra of each protein/lipid mixture to obtain the corrected spectra. The spectral center of mass, I_λ, for the emission spectra was determined using the Carey Eclipse software. Assuming that this spectral change in tryptophan fluorescence correlates with the amount of protein bound, the fractional saturation of binding sites, θ, was calculated using the following equation: θ = [I_λ(obs) − I_λ(0)]/[I_λ(max) − I_λ(0)], where I_λ(obs) is the spectral center of mass for each protein/lipid ratio and I_λ(0) and I_λ(max) is the spectral center of mass for protein alone and the protein in the presence of saturating amounts of liposomes, respectively. Plotting θ versus protein/lipid molar ratios yielded the resulting binding isotherms.

Determination of tryptophan depth of membrane penetration.

Quenching of tryptophan fluorescence by brominated phospholipids was used to determine the depth of tryptophan penetration into the lipid bilayers (8, 23). The brominated lipids 1-palmitoyl-2-stearoyl(6′,7′-dibromo)-sn-glycero-3-phosphocholine, 1-palmitoyl-2-stearoyl(9′,10′-dibromo)-sn-glycero-3-phosphocholine, and 1-palmitoyl-2-stearoyl(11′,12′-dibromo)-sn-glycero-3-phosphocholine were purchased from Avanti Polar Lipids. Three different liposome preparations were made containing POPC/POPS/(6,7)-Br₂-PC, POPC/POPS/(9,10)-Br₂-PC, or POPC/POPS/(11,12)-Br₂-PC at a 25:25:50 mol% ratio. Recombinant VI single-tryptophan mutants were incubated for 10 min at 37°C with brominated liposomes at a 1:100 molar ratio (protein/lipid) in HBS at pH 7.5, since this ratio results in the saturation of protein VI binding to membranes (see Fig. 6B). The intensity of tryptophan fluorescence was measured at 325 nm upon excitation at 295 nm. The differences in quenching tryptophan fluorescence by the (6,7)-, (9,10)-, or (11,12)-Br₂-PC was used to calculate the location of the residue in the bilayer using the parallax method (8, 23). In the parallax method, the depth of the tryptophan residue was calculated using the following formula: Z_cf = L_cl + [(−ln(F₁/F₂)/πC − L²₂₁]/2L₂₁, where Z_cf represents the distance of the fluorophore from the center of the bilayer, L_cl is the distance of the shallow quencher from the center of the bilayer, L_2l is the distance between the shallow and deep quenchers, F₁ is the fluorescence intensity in the presence of the shallow quencher, F₂ is the fluorescence intensity in the presence of the deep quencher, and C is the concentration of quencher in molecules/Ų. The average bromine distances from the center of the bilayer, based on X-ray diffraction were taken to be 10.8, 8.3, and 6.3 Å for (6,7)-, (9,10)-, and (11,12)-Br₂-PC, respectively (8, 23). Mean Z_cf values were calculated from at least 3 independent experiments.

Circular dichroism.

The secondary structure content of protein VI and VI-L40Q was analyzed by circular dichroism spectropolarimetry. Purified proteins concentrated to 0.2 mg/ml in 25 mM phosphate (pH 7.5) and 150 mM NaCl were analyzed using a Jasco J-820 spectropolarimeter with a 0.1-cm quartz cuvette. Spectra were collected at 25°C in triplicate from 190 to 260 nm in 0.2-nm increments and analyzed for secondary structure content using the CDPro software package (46). The data reported are the average values obtained from Contin, Cdsstr, and Selcon3 analyses.

RESULTS

Generation of VI mutant AdVs.

In vitro membrane lytic activity of recombinant VI protein is largely dependent upon a conserved N-terminal amphipathic α-helix (Fig. 1A and B) (57). Therefore, we reasoned that mutations in this domain would generate AdVs with altered endosomalytic activity. However, initial attempts at directed mutagenesis did not yield replication-competent viral mutants with significantly attenuated infectivity (Table 1). In contrast, random mutagenesis combined with the recently developed system of AdV genome recombineering proved to be a more successful approach (53, 61). We used a bacterial artificial chromosome (BAC) carrying a replication-defective AdV genome harboring an EGFP reporter gene in place of E1/E3 (61). Random mutations of the nucleotides encoding residues 34 to 53 of the amphipathic α-helix of protein VI were generated via mutagenic PCR and inserted into the genome using homologous recombination. In a preliminary screen of 57 candidate clones, we identified 16 unique mutant viral genomes encoding single, double, or triple point mutations in this region (Table 1). Subsequently, we generated stocks of each VI mutant virus by transfecting the linearized viral genome into the E1/E3-complementing 293β5 cell line. We were able to produce viable virus from 15 of the 16 clones; however, a genome harboring three point mutations (S36C/N46D/K52R) in the amphipathic α-helix was likely lethal, as no virus could be rescued.

FIG. 1.

Protein VI structure. (A) Schematic representation of the primary structure of the 250-residue preprotein VI. The AdV L3/23K protease cleaves VI between amino acids 33/34 and 239/240 (arrows) to generate mature VI. The putative N-terminal amphipathic helix (residues 34 to 54) is indicated by a black box. (B) Helical wheel plot of the amphipathic α-helix of VI generated with HELIQUEST (15). Residues are indicated as follows: hydrophobic (white), polar (black), and charged (gray).

TABLE 1.

Summary of protein VI mutants

VI mutanta	Infectivity reductionb
S36C	−
S36R	−
S36N	−
W37A	+
L40Q	+++
L40W	NR
W41A	+
S42A	−
G43A	−
G43S	−
K45R	−
K45R	−
K45A	−
K45E	−
F47S	−
G48C	+
V51I	−
K52M	−
K52E	−
N53D	−
G33S/S36G	−
S36C/V51F	−
W37A/W41A	NR
K45R/F47L	−
S36C/N46D/K52R	NR
W41R/G43D/T50I	−

^aProtein VI mutants generated with random mutagenesis are shown in boldface, while those generated via directed mutagenesis are shown in lightface.

^bAbbreviations: −, no reduction; +, <5-fold reduction; +++, ≥10-fold reduction; NR, virus not recovered.

Effects of protein VI mutations on virus infectivity.

We first evaluated the VI mutant viruses in a single-round infection assay to assess the effect of amino acid substitutions in the amphipathic α-helix on infectivity. The majority of the mutants (14 of 15) did not show a significant difference in infectivity compared to the wild-type virus (Table 1). However, one virus with a single point mutation, AdV VI-L40Q, exhibited a marked decrease (11-fold) in infectivity compared to that of the wild-type virus (Fig. 2A) following infection with an equivalent number of virus particles.

FIG. 2.

Infectivity of VI-L40 mutants is reduced compared to that of wild-type AdV. Nonpermissive A549 cells were infected with serial dilutions of wild-type (WT) or L40Q-1 (A) and WT, L40Q-1, L40Q-2, or L40N (B) mutant AdV vectors. L40Q-1 was generated by random mutagenesis, while L40Q-2 and L40N are directed mutants. For both panels, the percentage of infected cells was determined by flow cytometry. Nonlinear regression analysis of the resulting data was used to calculate ID₅₀ values and, from these, the fold change in infectivity, as reported in the text. The data are represented as the mean percentages of GFP-positive cells resulting from three independent experiments ± standard errors of the means (SEM).

Due to the fact that random mutagenesis and recombineering were used to generate the initial panel of mutants, we were concerned that an unidentified secondary mutation elsewhere in the viral genome might account for the observed reduction in infectivity. Therefore, we used directed mutagenesis and recombineering to introduce the L40Q mutation into the wild-type viral backbone. The infectivity profile of the reconstructed mutant AdV was highly similar to that of the VI-L40Q virus obtained by random mutagenesis (Fig. 2B), thus confirming that the attenuation in virus infectivity is a direct result of the L40Q mutation.

Additional mutations at the L40 position of protein VI further highlighted the importance of this residue in virus infection and provided further validation that the observed reduction in infectivity for AdV VI-L40Q was not due to an unknown secondary mutation. A virus containing a L40N mutation exhibited a reduction in infectivity essentially identical to that of AdV VI-L40Q (Fig. 2B). In addition, an L40E mutation in VI was lethal. Thus, replacement of the hydrophobic leucine at position 40 with a polar residue (Q/N) is partially tolerated by the virus, whereas substitution of a charged acidic residue (E) is not.

AdV VI-L40Q retains normal capsid composition and particle stability.

As VI has a role in virus maturation and capsid stability, we examined whether the L40Q mutation impaired these functions (28, 47). SDS-PAGE analysis of purified virions indicated that the normal repertoire of viral proteins was present in the AdV VI-L40Q capsid, that protein VI was incorporated to wild-type levels, and that proteolytic processing of viral preproteins VI and VII by the AdV L3/23K protease was unaffected by the L40Q mutation (Fig. 3A). In addition, we found no significant alterations in capsid thermostability (Fig. 3B). The dissociation of fiber and protein VI from virions exposed to increasing temperatures was unperturbed in VI-L40Q virus compared to that in wild-type virus, with both proteins being released from the capsid between 46 and 49.8°C. These results are in agreement with previous reports showing that wild-type AdV capsid dissociation occurs between 43 and 49°C (42, 43, 57). Thus, we conclude that VI-L40Q assembles normally and that, unlike AdV2ts1, reduced infectivity of AdV VI-L40Q is not due to capsid stabilization, a situation that would impede cell entry.

FIG. 3.

L40Q does not affect VI encapsidation or capsid thermostability. (A) Equivalent amounts (1.5 μg) of two independent preparations of wild-type or L40Q virus were subjected to SDS-PAGE. Viral proteins were stained with Deep Purple and imaged with a Typhoon fluorimager. (B) Wild-type and L40Q viruses were heated to the indicated temperatures for 10 min and then fractionated on 30-to-80% Histodenz discontinuous gradients. Supernatant (Sup) and band fractions were collected, separated by SDS-PAGE, and probed by immunoblotting for fiber or VI.

The VI-L40Q virus is attenuated for endosome lysis.

We hypothesized that AdV VI-L40Q is defective at an early stage of its life cycle based on the results from a single round of infection; however, the identification of the specific step in the cell entry pathway that is perturbed by the L40Q mutation required further investigation. Therefore, a ribotoxin (α-sarcin) codelivery assay was used to specifically measure the effect of the L40Q mutation on endosome rupture during AdV cell entry. In this assay, the total levels of de novo cellular protein synthesis are used as a quantitative marker for translocation of α-sarcin from the endosome to the cytosol, an event that is dependent upon endosome rupture by AdV. We observed a 10-fold increase in the ID₅₀ for AdV-mediated endosome penetration by the VI-L40Q mutant virus compared to that by the wild-type virus (Fig. 4A and B). The AdV2ts1 mutant exhibited a nearly 100-fold increase in the ID₅₀ required for cytosolic α-sarcin translocation, consistent with its more impaired infectivity. The close correlation between the 10-fold attenuation in endosomalysis and the 11-fold reduction in infectivity strongly suggests that the VI-L40Q mutant virus is specifically defective in endosomal escape.

FIG. 4.

The L40Q mutation attenuates endosomalysis. (A) [³⁵S]methionine incorporation into cellular proteins was measured in the presence of the ribotoxin α-sarcin and increasing concentrations of wild-type AdV, AdV VI-L40Q, or AdV2ts1 (ts1). The data are presented as the percentages of [³⁵S]methionine incorporation compared to those of uninfected control cells. Points indicate the means ± standard deviations (SD) of results of using triplicate samples from a single experiment, representative of three independent experiments. (B) The ID₅₀ value for each independent experiment was calculated by nonlinear regression analysis. The data shown are the mean fold changes in ID₅₀ over the wild type ± SEM.

To substantiate this, we used confocal fluorescence microscopy to ascertain whether the VI-L40Q mutant virus showed reduced accumulation in the nucleus due to an inability to escape the endosome (Fig. 5). As a control, we used AdV5-P137L, which harbors the same point mutation in the AdV protease found in the AdV2ts1 virus and is also defective for endosome escape (30). Incubation of AdV with A549 cells at 4°C revealed strong virus staining at the cell periphery for the wild-type, VI-L40Q, and AdV5-P137L viruses (data not shown). Therefore, we infer that AdV VI-L40Q is not defective for receptor binding. For all three viruses tested, we observed ∼16% nuclear colocalization at the 0-min time point, which we attribute to coincidental colocalization of viruses bound to the outer cell membrane above the nucleus. Nuclear localization of AdV VI-L40Q increased modestly from 0 to 15 min postinfection, with no significant change occurring beyond 15 min postinfection. In contrast, there was nearly twice as much nuclear accumulation of the wild-type AdV particles as those of the VI-L40Q mutant (73% of total visible particles versus 39%, respectively) at 60 min postinfection. We also investigated the colocalization of VI-L40Q particles with a marker for the early endosome, EEA1. A baseline value of ∼12% colocalization for each virus with EEA1 at the 0-min time point, like that for nuclear colocalization, is likely due to coincidental colocalization. Strikingly, the time-dependent colocalization of AdV VI-L40Q with early endosomes closely paralleled that of AdV5-P137L, which is defective for endosome penetration. Both viruses persisted in EEA1-positive vesicles at early times postinfection (15 to 30 min), whereas the wild-type virus had already escaped to the cytosol during this interval. These findings confirm the results from the sarcin codelivery assay and indicate that the VI-L40Q mutant virus is attenuated for endosome escape during cell entry.

FIG. 5.

Cellular trafficking of AdV is altered for the VI-L40Q mutant. (A) A549 cells were infected with AdV, AdV VI-L40Q, or AdV5-P137L and stained for virus (red), EEA1, a marker of early endosomes (green), and the nucleus (blue). Panels are Z projections of the 30-min postinfection time point, and the scale bar represents 10 μm. (B and C) Colocalization of virus with the nucleus (B) or EEA1 (C) was determined as a function of time postinfection. The data shown are the mean percent colocalizations ± SEM from ∼75 cells per time point from three independent experiments.

We also noted that by 60 min postinfection, the percentages of both AdV5-P137L and VI-L40Q mutant viruses colocalized with EEA1 dropped to 36 and 30%, respectively. We reasoned that failure to escape the endosome could result in trafficking of the VI-L40Q mutant virus to lysosomal compartments. However, we found no significant accumulation of the mutant virus in lysosomes (data not shown). It is possible that the VI-L40Q particles that fail to escape the endosome are recycled back to the cell surface in a manner similar to that reported for AdV2ts1 (18). Alternatively, degradation of the hexon epitope in lysosomes may prevent recognition by the anti-hexon antibody used to visualize the virus.

Recombinant VI-L40Q exhibits altered membrane binding and insertion in vitro.

The finding that AdV VI-L40Q has attenuated endosomalytic activity prompted us to investigate the biophysical basis for this using recombinant protein. Previous studies have shown that protein VI lyses model liposomes and that this pH-independent lytic activity is attributed primarily to the N-terminal amphipathic α-helix of the protein (57). In agreement with previous studies, wild-type full-length VI caused a dose-dependent increase in the release of an entrapped fluorophore (sulforhodamine B) from liposomes (Fig. 6A). In contrast VI-L40Q required ∼10-fold-higher protein concentrations to cause the same levels of SulfoB release as the wild-type protein. Consistent with previous results, a VI variant with the entire amphipathic α-helix (residues 34 to 54) deleted (VIΔ54) was drastically attenuated for VI-mediated membrane rupture. These results indicate that VI-L40Q is specifically attenuated for membrane lysis in vitro, consistent with our findings using a cell-based endosome rupture assay.

FIG. 6.

The L40Q mutation attenuates membrane lysis and affinity. (A) Sulforhodamine B-entrapped liposomes were incubated with increasing concentrations of purified VI, VI-L40Q, or VIΔ54. Liposome lysis is reported as the percentage of dye release compared to that of total lysis in the presence of Triton X-100. (B) Membrane binding of various VI constructs was determined by monitoring changes in intrinsic tryptophan fluorescence following incubation with increasing amounts of lipid. The data shown are reported as the means ± SEM from three independent experiments. (C) Representative CD spectra of purified protein VI and VI-L40Q. The data are presented as the molar circular dichroism versus wavelength. Note the two negative peaks indicative of α-helices at 208 and 222 nm for the wild-type protein and the shift to greater β-sheet (215 nm) and random coil (205 nm) values for the VI-L40Q mutant.

We next examined whether the reduced in vitro membrane lytic activity, and by extension, endosome rupture, of VI-L40Q was due to decreased membrane affinity, as measured by changes in intrinsic tryptophan fluorescence. This approach is routinely used for monitoring interactions between proteins and ligands, membranes, or other proteins and relies on the assumption that the fractional spectral change in tryptophan fluorescence correlates directly with the amount of protein bound to its substrate (11). Upon incubation with increasing amounts of lipid, we observed a dose-dependent increase in fractional saturation of binding sites for all VI variants tested (Fig. 6B). However, the ability of VI-L40Q to bind membranes was reduced by 10-fold compared to that of wild-type VI. As expected, VIΔ54 was severely attenuated for membrane binding. From these results, we concluded that the defect in VI-L40Q membrane lytic activity was due primarily to a decrease in membrane affinity.

Previously, we used intrinsic tryptophan fluorescence quenching in model brominated lipid membranes to determine the depth to which conserved tryptophans W37, W41, and W59 penetrate lipid bilayers (27). These results supported a model in which the putative N-terminal amphipathic α-helix adopts an oblique orientation in lipid membranes. We used this technique to determine the effect of the L40Q mutation on VI membrane insertion. Accurate interpretation of data in these experiments requires that only a single tryptophan be present in the protein. Therefore, we generated variants of a truncated form of VI consisting of residues 34 to 114, VI114Δ, in which 2 of the 3 tryptophan residues were mutated to phenylalanine, leaving a single tryptophan residue in each construct. Phenylalanine has no intrinsic fluorescence, yet its lipid interaction properties closely resemble those of tryptophan. Previous studies have demonstrated that replacing any 2 tryptophans with phenylalanine did not alter VI114Δ membrane binding or protein secondary structure (27). Using these mutants with double phenylalanine substitutions, we found that VI114Δ insertion depths agreed very well with previously published results, as summarized in Fig. 7A and Table 2. We then created a series of constructs incorporating the L40Q substitution. Insertion of W59 into model bilayers to a depth of 11.6 Å for VI114Δ-L40Q was comparable to the depth of 10.6 Å observed for VI114Δ (P = 0.19, Student's t test). Remarkably, however, the VI114Δ-L40Q mutant was completely defective for the insertion of tryptophans 37 and 41 into the lipid bilayer, based on a lack of tryptophan fluorescence quenching by brominated lipids and despite the fact that protein was maximally bound to lipid membranes at the protein/lipid molar ratio (1:100) (Fig. 6B) used in this experiment. The data obtained for these residues were similar to results collected with a negative-control protein (BSA), which does not penetrate bilayers (Fig. 7A, right).

FIG. 7.

L40Q alters membrane insertion. (A) Insertion depths of three tryptophan residues (W37, W41, and W59) inserted into brominated phospholipids were measured independently via tryptophan fluorescence quenching. BSA was included as a negative control. (B) Membrane binding model for residues 34 to 114 of protein VI. Wild-type protein VI is shown oriented parallel to the membrane. For L40Q, the extreme N terminus of the protein does not interact with membranes. The model shows one possible conformation of the protein on a lipid bilayer.

TABLE 2.

Distance from the bilayer center

Residue	Distancea
	WT	L40Q
W37	9.9 ± 0.3	ND
W41	11.1 ± 0.5	ND
W59	10.6 ± 0.2	11.6 ± 1.1

^aThe distance of each tryptophan residue from the center of the bilayer (Z_cf [Å]) is calculated by fitting a parametric equation to the data shown in Fig. 7A. The data shown are the mean distances from the bilayer center ± SEM resulting from three independent experiments. ND, not detected.

We also utilized circular dichroism (CD) to better understand the effect of the L40Q substitution on protein VI secondary structure and found that this mutation results in a significant decrease in its overall α-helicity compared to that of the wild-type protein (Fig. 6C and Table 3). While we cannot definitively say that the decrease in α-helical content maps to the N-terminal amphipathic helix, structural alterations in this critical domain would be consistent with observed defects in membrane insertion and disruption. Taken together, the results of these biophysical studies are consistent with models whereby the L40Q mutation prevents the extreme N terminus of the amphipathic α-helix from normal insertion into the membrane (Fig. 7B).

TABLE 3.

Secondary structure of wild-type and mutant VI

VI protein	Secondary structurea
	α-Helix	β-Sheet	Random coil
WT	25 ± 3	45 ± 6	30 ± 3
L40Q	11 ± 2	53 ± 6	36 ± 4

^aThe average secondary structure content of each protein was measured with CD. The data shown are represented as relative percentages ± SEM obtained from separate analyses using CONTIN, CDSSTR, and SELCON (46).

DISCUSSION

While many steps of the AdV cell entry pathway have been described in detail, the precise mechanism involved in AdV endosome penetration and the viral protein responsible for this event have remained elusive. For AdV, penton base, peripentonal hexons, VI, and IIIa are released from the capsid at early times postinfection, and it was assumed that one of these proteins mediates endosomalysis (19). Recent studies strongly suggested protein VI is the AdV membrane lytic factor based on biochemical and immunologic studies (27, 57). We chose a functional genetics approach to evaluate the role of protein VI during AdV entry but were initially hampered by relatively uninformative or lethal mutants produced via directed mutagenesis. In an attempt to use a less biased approach, we developed a novel method to generate random mutations targeted to the amphipathic helix of VI within the context of the full AdV genome. Previously, random mutagenesis of the AdV genome required propagation of the virus in the presence of mutagenic chemicals, and the resulting mutations were incorporated arbitrarily throughout the viral DNA (22, 32, 58). Adaptation of the recently developed system for AdV recombineering allowed us to specifically target a region of protein VI that was important for in vitro membrane lytic activity and that we reasoned would reveal mutant AdVs with altered endosomalytic activity (27, 53, 57, 61). While we did not exhaustively examine the various parameters of our random mutagenesis technique, we were able to incorporate at least a single base mutation in a relatively small DNA fragment (54 bp) in 50% of clones by tuning our mutation rate to the upper limits of the system. This technique could be broadly applicable for investigators wishing to identify functionally important residues in discrete domains of AdV proteins.

Using this random mutagenesis system, we obtained the first definitive evidence that protein VI mediates endosome disruption during AdV infection. We generated a virus encoding a single point mutation in protein VI, L40Q, which exhibits a 10-fold reduction in infectivity. The decrease in infectivity is primarily attributed to impaired endosomal escape based on biochemical and confocal microscopy experiments. In each of our experimental systems, we consistently observed a 10-fold decrease in activity for the AdV VI-L40Q mutant compared to that of the wild-type virus. The lone exception to this was our immunofluorescence studies, in which we recorded only a 2-fold decrease in nuclear accumulation of the mutant virus. However, technical constraints required a high virus particle/cell ratio to achieve quantitative staining. At 30,000 particles per cell, the difference in infectivity for the wild-type and L40Q viruses was greatly reduced (Fig. 2). Nevertheless, our findings clearly indicate that the reduction in infectivity is a direct result of failure of the VI-L40Q mutant virus to escape the endosome. Furthermore, we suggest that failure to disrupt the endosome is due to the decreased association of VI molecules released from the incoming capsid with the inner endosomal membrane.

Prior to this study, the accumulated evidence indicated that partial disassembly in the acidified endosome was a prerequisite for AdV endosomal escape. Monensin and bafilomycin, which block endosomal acidification, inhibit AdV infection (39, 57). When grown at the nonpermissive temperature, the AdV2ts1 virion does not undergo proteolytic maturation, resulting in a hyperstable capsid that fails to undergo disassembly (uncoat) in the endosome (4, 18, 54). Moreover, components of the immune system, including α-defensins, a class of naturally occurring antimicrobial peptides, and neutralizing antibodies stabilize the HAdV5 capsid and prevent uncoating (43, 44, 62). In each of these cases, perturbation of normal capsid disassembly results in a considerable loss of infectivity. In light of work presented here, we attribute the loss of viral infectivity in each of these cases to the sequestration of protein VI within the intact capsid, thereby impeding endosomalysis. In the case of the AdV VI-L40Q mutant virus, the loss of infectivity is not due to hyperstabilization of the capsid, as indicated by assessment of thermostability. Qualitatively, both the fiber and protein VI dissociate from the VI-L40Q mutant virus at the same temperature range as the wild-type virus. Although we cannot formally exclude a model in which VI-L40Q prevents uncoating or fails to dissociate from the capsid in the endosome, we also observed that exposure of the viral DNA in cells infected with either mutant or wild-type virus occurred similarly for both viruses (data not shown), indicating that the mutant virus is capable of undergoing proper disassembly (30).

One question raised by our studies is the fate of AdVs that fail to escape the endosome. While AdVs bound by α-defensins primarily accumulate in lysosomes, the AdV2ts1 virus has been shown to traffic to lysosomes or to recycle back to the cell surface (14, 18, 43). Confocal fluorescence microscopy revealed that the VI-L40Q mutant virus is efficiently internalized into early endosomes. Nevertheless, AdV VI-L40Q escape from early endosomal compartments was markedly reduced compared to that of the wild-type virus, clearly indicating a failure to mediate maximal endosomal escape. In addition, overall nuclear accumulation for the mutant was reduced by 50%. However, we did not observe trafficking of the VI-L40Q virus to lysosomes, suggesting that a portion of the input virus is either recycled back to the cell surface or that the epitope recognized by our monoclonal anti-hexon antibody is rapidly degraded in lysosomes.

The mechanism for membrane disruption by several nonenveloped viruses is thought to involve the formation of pores. For poliovirus, oligomerization of the VP1/VP4 membrane penetration factors, likely at the plasma membrane, generates a channel to allow translocation of the viral RNA genome (10, 49). Translocation of reovirus to the cytosol is also thought to involve the generation of size-selective pores in the endosomal membrane (1, 20). The Flock House virus endosomalytic factor is the gamma peptide, a small amphipathic peptide released from the capsid following proteolytic cleavage during cell entry (31). Biophysical studies revealed that this peptide first orients relatively parallel to the membrane and then likely induces pore formation following oligomerization (6, 7, 21). Interestingly, a previous study showed that AdV protein VI interacts with model membranes in an oblique fashion, similar to that observed for Flock House virus (27). However, we have recently proposed a model for VI-mediated endosome disruption involving gross fragmentation of the endosomal membrane via induction of positive curvature rather than pore formation (27, 57). Evidence for large-scale disruption of the endosomal membrane is supported by the finding that AdV-mediated release of differently sized dextrans from endosomes is not size selective, in clear contrast to that of other viruses that form pores (33). This mechanism of membrane rupture does not involve pore formation but rather occurs when bilayer structure is compromised via increased curvature stress induced when lytic α-helices bind membranes oriented parallel to the membrane surface (40). In the case of the VI-L40Q mutant, membrane binding was found to be reduced by 10-fold compared to that of wild-type protein VI, indicating an overall decrease in membrane affinity. Furthermore, insertion of the VI114Δ-L40Q mutant into bilayers was significantly altered. We observed complete abrogation of membrane insertion for the two tryptophan residues (W37 and W41) nearest to the L40Q mutation. Membrane insertion of a more distant tryptophan residue (W59) that is outside the predicted amphipathic helix, but still within the largely helical N terminus, was statistically the same for the wild type and VI-L40Q. This suggests that the 10-fold decrease in membrane binding can be largely attributed to decreased affinity for the extreme N terminus of VI114Δ-L40Q for lipid bilayers. Thus, we attribute the 10-fold reduction in endosomalysis for AdV VI-L40Q to defective insertion of residues 37 to 41 (minimally) into the endosomal membrane.

Our studies also underscore the conservation of protein VI sequences involved in membrane lytic activity among the different >52 types, representing species A to G, of HAdV. Based on an alignment of published protein VI sequences from all HAdVs, it is apparent that the entire N-terminal half of the protein (residues 1 to 108 of HAdV5 used in this study) is very well conserved (56% identical and 36% similar). The remaining C-terminal residues are somewhat less conserved (18% identical and 15% similar), which may indicate that this largely disordered domain of the protein interacts with other capsid proteins (e.g., hexon) in a type-specific manner. An aliphatic residue at position 40 is highly preferred, as 98% of all members of the family Adenoviridae, including nonhuman AdVs, sequenced to date contain a leucine (45%), isoleucine (30%), or valine (23%) at this position. This correlates well with our mutagenic studies that indicated that AdV tolerance for substitutions at the L40 position is limited. Substitutions of two polar amino acids at this position (glutamine and asparagine) resulted in a comparable loss of viral infectivity, suggesting that hydrophobicity is important for membrane lysis. The observed lethality of the change from L40 to glutamic acid further strengthens this argument.

In summary, our findings shed further light on the precise mechanism of endosomalysis by adenovirus protein VI and the process of membrane perturbation. These studies highlight both the distinct and similar processes by which nonenveloped viruses invade host cells. Such knowledge may allow manipulation of protein VI to either enhance or attenuate cell entry of HAdV vectors as well as provide a foundation for the engineering of artificial cell delivery vehicles (24).