Group D Adenoviruses Infect Primary Central Nervous System Cells More Efficiently than Those from Group C

Abstract

Group C adenovirus-mediated gene transfer to central nervous system cells is inefficient. We found that wild-type group D viruses, or recombinant adenovirus type 2 (Ad2) (group C) modified to contain Ad17 (group D) fiber, were more efficient in infecting primary cultures of neurons. Together with studies on primary vascular endothelial cells and tissue culture cell lines, our results indicate that there is not a universally applicable adenovirus serotype for use as a gene transfer vector.

Gene transfer for the correction of inborn errors of metabolism or neurodegenerative disease in the central nervous system (CNS) has been accomplished with recombinant adenoviral vectors (7, 9, 14, 25). However, high particle doses are required for efficacy in mice and rats and for the infection of large numbers of cells in monkeys (4, 7, 8, 10, 16). Moreover, the delivery of a high particle load induces an immune response after in vivo application (2, 5, 22). Thus, gene transfer to brain tissues with adenovirus type 2 (Ad2) or Ad5 vectors is inefficient, as it is in endothelia, smooth muscle cells, and differentiated airway epithelia (12, 33, 34). Methods which improve the efficiency of adenovirus-mediated gene transfer to the CNS could reduce the particle load required to achieve sufficient levels of transduction. Improved efficiency could then reduce toxicity and increase the therapeutic index.

The coxsackievirus B-adenovirus receptor and major histocompatibility complex class I α2 domain are recently identified cellular receptors for Ad2 and Ad5 fiber (3, 19, 30), the only serotypes of adenovirus thus far tested in brain tissues. The inefficiency of gene transfer to brain tissue in vivo, and to CNS cells in culture, could be due to a lack of specific receptors for Ad5 fiber, similar to what was found in adenovirus-mediated gene transfer to differentiated human airway epithelia (24, 34). We hypothesized that gene transfer to CNS cells could be enhanced by circumventing the requirement for interaction between Ad5 fibers and Ad5 receptors.

One strategy to accomplish fiber-independent transduction of cells used complexes of adenoviruses with lipids or polymers. Fasbender and colleagues showed that mixing poly-l-lysine or cationic lipids with adenovirus prior to application enhanced gene transfer into cells which were refractory to adenovirus-mediated gene transfer (12). In a different approach, bispecific antibodies or modified fiber knob sequences were used to redirect the binding of adenoviruses to different cellular receptors (11, 21, 28, 33).

As an alternative to altering the native tropisms of Ad2 and Ad5, we hypothesized that other naturally occurring adenoviruses could possess fibers or other capsid proteins which would direct improved infection efficiency in brain cells. We tested the ability of different adenoviral serotypes, representative of groups A to F, to infect primary CNS cell cultures. Our results suggest that adenovirus vectors containing Ad17 fiber may be more appropriate vectors for infection of CNS cells in particular, as well as other primary cells.

Comparative analysis of adenoviral serotypes.

Fetal rat CNS cells were isolated from the upper layer of the cortex of D16-17 rat fetuses, seeded onto collagen-coated plates, and cultured in Eagle’s minimal essential medium (MEM) supplemented with 10% fetal bovine serum, 30 mM dextrose, 2 mM l-glutamine, and 20 mM KCl. Cultures were maintained for 2 to 3 days prior to adenovirus application. Cultures contained more than 90% neuronal cells (17) based on immunohistochemistry assays (data not shown). Wild-type virus lysates were purchased from the American Type Culture Collection, propagated in 293 cells in Dulbecco’s MEM supplemented with fetal bovine serum, 10% and purified by standard methods (10). Titers (in PFU per milliliter) were determined by end point dilution as previously described (18, 26) except that viruses were incubated for 3 days rather than 30 min. Purified viruses were placed onto primary neuronal cultures at a multiplicity of infection (MOI) of 50 PFU/cell for 30 min, and infection was monitored by immunohistochemistry for hexon production after 48 h. Briefly, cells were fixed with acetone-methanol 48 h after infection, followed by the addition of polyclonal fluorescein isothiocyanate-labeled antihexon antibody (Chemicon International Inc., Temecula, Calif.). Hexon-positive cells were counted by using low-magnification fluorescence photomicrographs of the monolayers (Leitz MZWild fluorescent photomicroscope). The following serotypes from all six groups of adenovirus were tested: 31 (group A); 3, 7, and 14 (group B); 2 and 5 (group C); 9, 17, 19, 26, and 30 (group D); 4 (group E); and 41 (group F).

The results, summarized in Table 1, demonstrated that serotypes from groups A and B infected only 5% of cells. Interestingly, group C viruses, which are currently the most commonly used backbone for gene transfer to animals and humans, also showed minimal infection. We found that viruses from groups D, E, and F were the most transfer efficient, with 20 to 60% of the cells in neuronal cultures testing positive for hexon production.

TABLE 1.

Results of serotype screening in primary cell cultures

Adenovirus group	Serotype	% Expressiona
		CNS cells	HUVEC
A	31	1–5	0
B	3	1–5	60
	7	0	20
	14	0	20
C	2	5	0
D	9	20	20
	17	60	60
	19	20	20
	26	20	20
	30	60	>90
E	4	20	0
F	41	40	0

^aExpression based on antihexon immunoreactivity determined by immunohistochemical methods. 

To test if the improved infection noted after application of groups D, E, and F serotypes was a general phenomenon, we did similar experiments on primary vascular endothelial cell cultures (Table 1). Human umbilical vein endothelial cells (HUVEC) were harvested and cultured in M199 medium supplemented with 1% l-glutamine, basal medium Eagle vitamin solution, and basal medium Eagle amino acids. Again, all serotypes were applied at an MOI of 50 for 30 min. Unlike primary neuronal cultures, HUVEC were infected most efficiently by group B in addition to group D viruses. Serotypes from groups E and F showed very low to no evidence of infection under these conditions, as did viruses from groups A and C.

Infection of primary CNS cell cultures with Ad2βgal2(17f) and Ad2βgal2.

The cellular attachment of adenovirus is mediated primarily through fiber. Therefore, we next tested whether fiber 17 sequences were responsible for the increased rate of infection. To determine this, a chimeric Ad2-Ad17 virus was generated (Fig. 1). The plasmid pAdORF6 (1) was cut with NdeI and BamHI to remove the Ad2 fiber-coding and polyadenylation signal sequences (bp 31076 to 32815). An NdeI-BamHI fragment containing the Ad17 fiber-coding sequence (bp 30983 to 32035) was generated by PCR and ligated along with simian virus 40 polyadenylation signal into NdeI-BamHI-cut pAdORF6 to generate pAdORF6fiber17. This plasmid was cut with PacI and ligated to PacI-cut Ad2βgal2 (1) genomic DNA to generate Ad2βgal2(17f). Ligated DNAs were transfected into 293 cells to generate the Ad2βgal2(17f) virus. Plaques were isolated and expanded, and Hirt DNAs were analyzed by NdeI and XbaI restriction to confirm the structure of the virus. The chimeric virus is similar to Ad2βgal2 (1) except that the Ad2 fiber sequences from bp 20624 to 32815 have been deleted and replaced with those from Ad17.

FIG. 1.

Schematic representation of the construction of Ad2 virus containing Ad17 fiber sequences. The final construct contains Ad2 backbone sequences with all but the first 17 amino acids of the fiber molecule removed. Replaced in frame is the Ad17 fiber from amino acid 18 to the COOH terminus. A simian virus 40 (SV40) polyadenylation signal was added to the 3′ end of the fiber sequence as indicated. The E4 ORF6 sequences are indicated for orientation. CMV, cytomegalovirus; β-gal, β-galactosidase; MLT, major late transcription.

Cultured primary cells from fetal rat cortex were infected at an MOI of 10 with Ad2βgal2 or Ad2βgal2(17f), and β-galactosidase activity was assessed initially by histochemical staining. Figure 2A shows a representative example of the differences in both the numbers of cells infected and the relative levels of enzyme expression following Ad2βgal2(17f)- and Ad2βgal2-mediated gene transfer. In parallel studies, β-galactosidase activity was assayed with a Galacto-Lite kit (Tropix, Inc., Bedford, Mass.) and a luminometer (Monolight 2010; Analytical Luminescence Laboratory, San Diego, Calif.) according to the manufacturers’ recommendations and normalized to total protein concentration (protein assay reagent; Bio-Rad Laboratories, Hercules, Calif.). β-Galactosidase activity assays after Ad2βgal2(17f) infection showed that it is approximately sevenfold more efficient than Ad2βgal2 (Fig. 2B).

FIG. 2.

Transgene expression after Ad2βgal2(17f) and Ad2βgal2 infection of primary cortical cultures. Cells were infected at an MOI of 10 for 2 h at 37°C, virus solution was removed, and cells were cultured in standard media for an additional 48 h. (A) β-Galactosidase histochemistry; (B) β-galactosidase activity assay. LU, relative light units. Data represent averages ± standard errors of the means.

To further enrich for neurons, some cultures were treated with cytosine β-d-arabinofuranoside (17). Treatment of cultures with cytosine β-d-arabinofuranoside did not significantly reduce the levels of total enzyme activity, reflecting the fact that the cultures contained mostly neuronal cells (data not shown). Nonetheless, combined X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) histochemistry and immunohistochemical staining for astrocytes (mouse anti-glial fibrillary acidic protein; Biogenix, San Ramon, Calif.) and oligodendroglia (mouse anti-myelin/oligodendrocyte monoclonal antibody; Chemicon) followed by incubation in secondary antibodies conjugated to horseradish peroxidase and horseradish peroxidase substrate (Vector Labs, Burlingame, Calif.) indicated that glia were also capable of being infected with Ad2βgal2(17f) (Fig. 3). There was no obvious preference for oligodendroglia or astrocytes in these culture conditions. Together, these results demonstrate that Ad17 fiber does not direct infection specifically to neurons.

FIG. 3.

Colocalization of β-galactosidase activity and cell-specific markers. Primary cortical cultures were infected with Ad2βgal2(17f) or Ad2βgal2 as described for Fig. 2. Two days after infection cells were histochemically stained for β-galactosidase followed by immunostaining with antibodies specific for astrocytes (glial fibrillary acidic protein [GFAP]) and oligodendrocytes (OLIGO). The photographs are selective and do not represent the cultures in general, which contained more than 90% neurons (17).

We also compared infection of Ad2βgal2(17f) to Ad2βgal2 in cultures of HUVEC and three cell lines (NIH 3T3, HeLa, and HEK293). NIH3T3 and HEK293 cells were cultured on 96-well plates in Dulbecco’s MEM (high glucose) supplemented with 10% fetal calf serum. HeLa cells were cultured on 96-well plates in Eagle’s MEM supplemented with 10% fetal calf serum and 10 mM nonessential amino acids. Figure 4 shows that Ad2βgal2(17f) was 21-fold more efficient than Ad2βgal2 as a gene transfer vector in HUVEC. However, Ad2βgal2(17f) was either equivalent (in NIH 3T3 cells) or less efficient (in HEK293 and HeLa cells) at infecting the cell lines tested. Collectively, the data suggest that the increased infection efficiencies observed with wild-type Ad17 and Ad2βgal2(17f) viruses could be attributed, in part, to properties of the Ad17 fiber.

FIG. 4.

β-Galactosidase activity in primary vascular endothelia or various cell lines infected with Ad2βgal2(17f) or Ad2βgal2 (MOI = 50). Data represent averages ± standard errors of the means. RLU, relative light units.

Various serotypes of human adenovirus show specific tissue tropisms and infect different cells (20). However, only group C adenoviruses have been used as vectors for gene transfer. Thus, the current approach to adenovirus-mediated gene transfer has followed a “one-size-fits-all” strategy. Because gene transfer to brain tissue by Ad5 is not optimal, we tested the hypothesis that different serotypes might be more efficient.

The screening of 12 different serotypes from groups A to F on primary cerebral cortical cells isolated from fetal rats showed that group D adenoviruses were most infectious, with the other groups following in the order F > E > C, A, and B. Our results with primary rat neuronal cells may not necessarily reflect which serotype is most efficient for primate brain tissue. Similarly, the improved tropism for primary vascular endothelia of group D compared with group C viruses may not be characteristic of endothelia in general. However, our studies on primary cells, together with our observations that Ad2 and Ad5 were more efficient than the chimeric viruses for infection of the tissue culture cell lines tested, suggest that there could be a wide variation in receptors capable of mediating adenovirus uptake in different cell types, and in different species.

Adenovirus fiber binds to cellular receptors initiating infection, and altering the fiber’s moieties can change its natural tropism (3, 13, 15, 21, 29). In addition to increased infection of group D serotypes of primary rat neuronal cells and human vascular endothelia, we noted improved gene transfer efficiency with a recombinant Ad2βgal2(17f) chimera. This virus differs from the parental Ad2-based vector only in its fiber region. Together, the data suggest that Ad17 fiber is in part responsible for the improved infection.

Studies have also demonstrated the importance of integrins on the host cell surface in adenovirus infection; both α_vβ₃ and α_vβ₅ mediate internalization of the virus in tissue culture cell lines through interaction with the penton base proteins (31, 32). In addition to its interactions with fiber, increased efficiency of gene transfer with Ad2βgal2(17f) may be partially attributable to improved interaction of cell surface integrins with Ad2 penton base. The Ad17 fiber is 116 amino acids smaller than Ad2 and is approximately one-third shorter in the shaft region. As has been suggested for Ad9 (27), the short shaft length may allow capsid proteins to come into close association with the cell surface, facilitating entry.

In summary, viruses modified to contain Ad17 fiber appear to be better vectors for gene transfer to CNS cells, suggesting that they could be used at lower titers both in culture and in vivo. With lower doses applied, direct cellular toxicity could be reduced and in vivo immune responses may be attenuated (6). In brain tissue, this could translate to extended expression, particularly in the case of nonimmunogenic transgenes (23).

Nucleotide sequence accession number.

The sequence of bp 31001 to 32053 of Ad17 has been deposited in GenBank by D. Armentano under accession no. AF108105.