A Common Mechanism for Cytoplasmic Dynein-Dependent Microtubule Binding Shared among Adeno-Associated Virus and Adenovirus Serotypes

Abstract

During infection, adenovirus-associated virus (AAV) undergoes microtubule-dependent retrograde transport as part of a program of vectorial transport of viral genome to the nucleus. A microtubule binding assay was used to evaluate the hypothesis that cytoplasmic dynein mediates AAV interaction with microtubules. Binding of AAV serotype 2 (AAV2) was enhanced in a nucleotide-dependent manner by the presence of total cellular microtubule-associated proteins (MAPs) but not cytoplasmic dynein-depleted MAPs. Excess AAV2 capsid protein prevented microtubule binding by AAV serotypes 2, 5, and rh.10, as well as adenovirus serotype 5, indicating that similar binding sites are used by these viruses.

A key requirement for viral infection is the efficient transport of genetic material from the cell surface to the nucleus. The infection cycle of adeno-associated viruses (AAVs) is of special interest since, as a dependovirus, it has jettisoned some viral functions that can be provided by other viruses (4, 14, 21, 25, 34, 36). As a result, AAV carries fewer genes than many other viruses and utilizes a relatively small capsid (∼26 nm) (1, 29). Whereas larger viruses such as adenovirus or herpesvirus (∼100 nm capsid or nucleocapsid) require the molecular motor, cytoplasmic dynein, during infection (9, 17, 19, 20, 28, 32, 33), the small size of the AAV viral capsid raises a question as to whether interaction with molecular motors is utilized. Based on the observation that AAV requires an intact microtubule cytoskeleton to traffic to the nucleus (24), we hypothesized that AAV capsids would associate with microtubules via an interaction with cytoplasmic dynein.

To investigate the potential interaction of AAV capsids with microtubules, binding of Cy3 fluorophore-conjugated AAV serotype 2 (AAV2) with endogenous microtubules and microtubule-associated proteins (MAPs) isolated from A549 cells was evaluated by using a previously published method (15). Briefly, recombinant AAV2 was produced by using a two-plasmid transfection system as previously described (6). Covalent conjugation of AAV2 to Cy3 fluorophore (Amersham, Arlington Heights, IL) utilized a method previously applied to both adenovirus and adeno-associated virus and shown to have no significant effect on viral infectivity (2, 16, 22, 24). Cell lysate was prepared from A549 epithelial carcinoma cells (CCL-185; American Type Culture Collection, Rockville, MD) by sonication and low-speed centrifugation. Taxol (40 μM) was added to polymerize microtubules which were pelleted with associated MAPs over a glycerol cushion (100,000 × g for 40 min). The microtubule pellet containing microtubules and MAPs was resuspended in a physiological buffer and incubated with Cy3-conjugated AAV2 capsids (10⁹ particles) for 40 min at 22°C, followed by centrifugation to create the final supernatant and pellet. Equal proportions of supernatant and pellet were evaluated by using a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel and fluorescence scanning. Parallel gels were stained with Coomassie blue staining (Bio-Rad, Hercules, CA) or transferred to nitrocellulose for immunoblotting. When mixed with microtubules and MAPs derived from A459 cells, the majority of the Cy3-AAV2 capsid pelleted with microtubules and MAPs (Fig. 1 and Table 1). When Taxol was omitted from the experiment, microtubules failed to form and both tubulin and AAV remained in the supernatant demonstrating that AAV pelleting could only occur in association with polymerized microtubules (data not shown). A Coomassie blue stain of the gel revealed that high-molecular-mass MAPs (>200-kDa band in the Coomassie blue stain) copelleted with microtubules (55-kDa tubulin band in the Coomassie blue stain) (Fig. 1). MAP-free microtubules (prepared by adding 500 mM NaCl during pelleting of microtubules from the A549 lysate) failed to interact with Cy3-AAV2 efficiently, and supplementation of MAP-free microtubules with a single, purified MAP (bovine brain Tau; 10 μg per reaction; Cytoskeleton, Inc., Denver, CO) failed to reestablish Cy3-AAV2 binding to microtubules, indicating that enhancement of AAV-microtubule association was not a general property of microtubule binding proteins (Table 1). Whereas microtubules continued to pellet in the presence of high salt concentrations (data not shown, see reference 15), both MAPs and Cy3-AAV2 were released to the supernatant (Table 1).

FIG. 1.

Analysis of MAP-dependent AAV interaction with microtubules. To evaluate AAV2-microtubule interaction, a previously published microtubule binding assay was performed (15). Cy3-conjugated AAV2 was combined with MAP-containing or MAP-depleted microtubules from A549 cell lysate, incubated, and then centrifuged to pellet microtubules. Supernatant (S) and pellet (P) were then analyzed by gel electrophoresis. A fluorescence scan (top) and a Coomassie blue stained gel (bottom) show that MAPs enhance AAV2-microtubule interaction and pelleting.

TABLE 1.

Quantitative assessment of viral binding to microtubules

Condition(s)	% Capsid in pellet	Pa
MAP-dependent binding
MAP-containing cell lysate + microtubules	74 ± 3	Control
MAP-containing cell lysate + microtubules + 500 mM NaCl	1.2 ± 0.2	<0.01
MAP-depleted cell lysate + microtubules	27 ± 7b	<0.01
MAP-depleted cell lysate + microtubules + Tau	23 ± 8b	<0.01
Anticytoplasmic dynein-depleted cell lysate + microtubules	26 ± 7	<0.01
Irrelevant immunoglobulin G-depleted cell lysate + microtubules	73 ± 3	>0.1
MAP-containing cell lysate prepared in 10 mM ATP + microtubules	47 ± 3	<0.01
MAP-containing cell lysate prepared in 10 mM AMP-PNP + microtubules	77 ± 7	>0.1
Competition studies
Cy3-AAV2 + naive Sf9 cell lysate	90 ± 4	Control
Cy3-AAV2 + pAAV2-Sf9 cell lysate	1 ± 2	<0.01
Cy3-AAV2 + pNull-Sf9 cell lysate	91 ± 3	>0.1
Cy3-AAV5 + naive Sf9 cell lysate	95 ± 5	Control
Cy3-AAV5 + pAAV2-Sf9 cell lysate	1 ± 1	<0.01
Cy3-AAV5 + pNull-Sf9 cell lysate	97 ± 1	>0.1
Cy3-AAVrh10 + naive Sf9 cell lysate	95 ± 3	Control
Cy3-AAVrh10 + pAAV2-Sf9 cell lysate	1 ± 0.4	<0.01
Cy3-AAVrh10 + pNull-Sf9 cell lysate	95 ± 3	>0.1
Cy3-adenovirus serotype 5 + naive Sf9 cell lysate	68 ± 6	Control
Cy3-adenovirus serotype 5 + pAAV2-Sf9 cell lysate	3 ± 1	<0.01
Cy3-Adenovirus serotype 5 + pNull-Sf9 cell lysate	63 ± 5	>0.1

^aControl, condition followed by P values for comparator conditions.

^bP > 0.1 for comparison of these two values.

The role of cytoplasmic dynein in mediating AAV2 capsid interaction with microtubules was assessed by using two methods: (i) evaluation of AAV2-microtubule interaction after depletion of cytoplasmic dynein from cell and (ii) demonstration of the nucleotide sensitivity of AAV2-microtubule binding MAP-containing microtubule pellets (see reference 15 for immunoprecipitation and nucleotide sensitivity assay methods). After removal of cytoplasmic dynein from cell lysates, AAV no longer interacted efficiently with microtubules (Table 1). Cell lysates supplemented with 10 mM ATP failed to support efficient AAV2-microtubule interaction, whereas unsupplemented lysates or lysates supplemented with 10 mM AMP-PNP, a nonhydrolyzable analog of ATP, allowed AAV2 to pellet with microtubules, indicating that AAV2-microtubule binding was sensitive to high ATP concentration, a property of cytoplasmic dynein (23, 26) (Table 1).

We further hypothesized that cytoplasmic dynein would bind to AAV capsids in the absence of microtubules. To test this hypothesis for AAV2 as well as AAV5 (the most distantly related human AAV) (12) and AAVrh.10 (a nonhuman primate family member) (12), AAV5 and AAVrh.10 were purified (6, 7), and either Cy3-AAV2, Cy3-AAV5, or Cy3-AAVrh.10 capsids (2 × 10⁹ particles) were added to A549 cell lysate (without Taxol to prevent microtubule formation) and incubated 4 to 8 h at 4°C prior to immunoprecipitation with anticytoplasmic dynein antibodies as described in Kelkar et al. (15). The anticytoplasmic dynein immunoprecipitate contained cytoplasmic dynein, dynamitin, and AAV2, AAV5, or AAVrh.10 capsid proteins (Fig. 2). When the immunoprecipitation was carried out with an irrelevant primary antibody, or when cell lysate was omitted from the experiment, no dynein, dynamitin, or AAV capsid proteins were detected in the immunoprecipitate.

FIG. 2.

Coimmunoprecipitation of cytoplasmic dynein, dynamitin, and AAV. Cy3-conjugated AAV2, AAV5, or AAVrh.10 capsids were added to cell lysate, incubated, and immunoprecipitated by using a monoclonal anticytoplasmic dynein antibody (15). Immunoprecipitated AAV proteins were analyzed by gel electrophoresis and fluorescence scanning. Cytoplasmic dynein and dynamitin were evaluated by immunoblotting. (A) Coimmunoprecipitation of AAV2 and the dynein complex. (B) Coimmunoprecipitation of AAV5 and the dynein complex. (C) Coimmunoprecipitation of AAVrh.10 and the dynein complex.

The preceding data demonstrated that AAV2, AAV5, and AAVrh.10 interact with cytoplasmic dynein in solution in the absence of microtubules. Previously, we showed that adenovirus serotype 5 capsid interacted with microtubules in a similar manner (15). In order to determine whether the AAV serotypes and adenovirus have distinct or mutually exclusive binding mechanisms, a virus-microtubule binding competition assay was established in which Cy3-AAV2, Cy3-AAV5, and Cy3AVrh.10, and adenovirus serotype 5 were combined with A549-derived microtubules and MAPs in the presence of excess AAV2 capsids proteins (Fig. 3 and Table 1). The source of the excess AAV2 capsid protein was a lysate of Sf9 cells (Invitrogen, Carlsbad, CA; cultured and infected according to the manufacturer's instructions) obtained at 72 h postinfection with a baculovirus expressing the AAV2 cap gene (35). Sf9 cell lysate was prepared by scraping and suspending cells in PEM buffer and lysing them by sonication and centrifugation to clarify the lysate. MAP-enriched microtubules from A549 cells were incubated with cell lysate derived from naive Sf9 cells, AAV2 capsid encoding baculovirus-infected Sf9 cells, and null baculovirus-infected Sf9 cells (40 min at 22°C) with the addition of Cy3-AAV2, Cy3-AAV5, Cy3-AAVrh.10, or Cy3-Ad 5 capsids (10⁹ particles). Microtubules incubated in the presence of naive baculovirus Sf9 cell lysate or lysate derived from cells infected with a null baculovirus efficiently bound Cy3-AAV2, Cy3-AAV5, Cy3-AAVrh.10, or Cy3-adenovirus serotype 5, with a majority of the capsid protein signal found in the microtubule-associated pellet fraction. However, lysate containing excess AAV2 capsid proteins failed to pull down Cy3-AAV2, Cy3-AAV5, Cy3-AAVrh.10, or Cy3-adenovirus serotype 5, with the majority of each capsid protein found in the supernatant (Fig. 3 and Table 1).

FIG. 3.

Common mechanism for microtubule interaction among divergent AAVs. MAP-containing microtubules were isolated from A549 cells as described above and incubated with an Sf9 cell lysate derived from naive Sf9 cells or from Sf9 cells infected with a baculovirus expressing the AAV2 cap gene (producing VP1, -2, and -3) or a null baculovirus. Cy3-AAV2, Cy3-AAV5, Cy3-AAVrh.10, or Cy3-adenovirus were incubated with the Sf9 lysates and pelleted. Supernatant (S) and pellet (P) were then analyzed by gel electrophoresis. (A) AAV2 fluorescence scan; (B) AAV5 fluorescence scan; (C) AAVrh.10 fluorescence scan; (D) adenovirus fluorescence scan. The Sf9 lysate containing AAV2 VP1, -2, and -3 inhibited virus-microtubule interaction, whereas control Sf9 cell lysates did not.

Although fluorophore-conjugated AAV capsids were used in the majority of the experiments described here, these data likely reflect the native interaction of AAV capsids with microtubules since Cy3-AAVs were competitively inhibited by excess, unconjugated AAV2 capsid protein (Fig. 3) and since immunoprecipitation of cytoplasmic dynein pulled down unconjugated AAV2, as detected by immunoblotting (data not shown). These data do not test the hypothesis that AAV-microtubule interaction might be affected by changes in the AAV capsid that could occur after endocytosis, such as lysosomal processing or ubiquitination.

The microtubule cytoskeleton has been implicated in the intracellular retrograde transport of AAV, as well as several other viral capsids or nucleocapsids that function to deliver viral genomes to the nucleus (5, 10, 18, 27). In the present study, biochemical evidence demonstrated that cytoplasmic dynein associates with several serotypes of AAV in solution and that cytoplasmic dynein is required for AAV binding to microtubules. Furthermore, these AAV serotypes and adenovirus either use the same molecular mechanism to interact with cytoplasmic dynein or share closely located binding sites, making the binding of two different viruses mutually exclusive. Recent studies have shown that specific purified recombinant subunits of the cytoplasmic dynein motor complex can bind to recombinant capsid and tegument proteins of viruses (herpesvirus, rabies virus, lyssavirus, and African swine fever virus [10]). Whole viral capsids, including adenovirus and canine parvovirus, also have dynein binding ability (15, 30, 31). AAV and adenovirus do not share amino acid homology in their capsid proteins, nor do they share homology with other viral capsid proteins shown to interact with cytoplasmic dynein subunits. These comparisons suggest that the mechanism of dynein interaction that is shared by adenovirus and AAV either results from physical or chemical aspects of the capsid other than specific amino acid motifs or represents the parallel evolution of viral capsids based on the functional requirement for intracellular motility. Whereas AAV, a dependovirus, has evolved to depend on adenovirus or herpesvirus helper functions to execute essential steps during replication, AAV infection per se does not require helper functions (3, 8, 11, 13). It stands to reason that AAV evolved and retained the ability to interact with microtubule-based motors to overcome the diffusional barrier presented by the cytoplasm.