Available online at www.sciencedirect.

com
R

Virology 309 (2003) 282–293 www.elsevier.com/locate/yviro

A mosaic adenovirus possessing serotype Ad5 and serotype Ad3 knobs

exhibits expanded tropism
Koichi Takayama,a Paul N. Reynolds,a,b Joshua J. Short,a Yosuke Kawakami,a
Yasuo Adachi,a Joel N. Glasgow,a Marianne G. Rots,a,c Victor Krasnykh,a,d
Joanne T. Douglas,a,d and David T. Curiela,d,*
a
Division of Human Gene Therapy, The University of Alabama at Birmingham, Birmingham, AL 35294-2172, USA
b
Chest Clinic, Royal Adelaide Hospital, 275 North Terrace, Adelaide, South Australia 5000, Australia
c
Department of Therapeutic Gene Modulation, University Center of Pharmacy, University of Groningen,
Deusinglaan 1, 9713 AV Groningen, The Netherlands
d
Department of Medicine, Pathology, and Surgery, and the Gene Therapy Center, The University of Alabama at Birmingham,
Birmingham, AL 35294-2172, USA

Received 12 July 2002; returned to author for revision 22 November 2002; accepted 2 December 2002

Abstract
The efficiency of cancer gene therapy with recombinant adenoviruses based on serotype 5 (Ad5) has been limited partly because of
variable, and often low, expression by human primary cancer cells of the primary cellular-receptor which recognizes the knob domain of
the fiber protein, the coxsackie and adenovirus receptor (CAR). As a means of circumventing CAR deficiency, Ad vectors have been
retargeted by utilizing chimeric fibers possessing knob domains of alternate Ad serotypes. We have reported that ovarian cancer cells possess
a primary receptor for Ad3 to which the Ad3 knob binds independently of the CAR–Ad5 knob interaction. Furthermore, an Ad5-based
chimeric vector, designated Ad5/3, containing a chimeric fiber proteins possessing the Ad3 knob, demonstrates CAR-independent tropism
by virtue of targeting the Ad3 receptor. Based on these findings, we hypothesized that a mosaic virus possessing both the Ad5 knob and
the Ad3 knob on the same virion could utilize either primary receptor, resulting in expanded tropism. In this study, we generated a dual-knob
mosaic virus by coinfection of 293 cells with Ad5-based and Ad5/3-based vectors. Characterization of the resultant virions confirmed the
incorporation of both Ad5 and Ad3 knobs in the same particle. Furthermore, this mosaic virus was able to utilize either receptor, CAR and
the Ad3 receptor, for virus attachment to cells. Enhanced Ad infectivity with the mosaic virus was shown in a panel of cell lines, with
receptor profiles ranging from CAR-dominant to Ad3 receptor-dominant. Thus, this mosaic virus strategy may offer the potential to improve
Ad-based gene therapy approaches by infectivity enhancement and tropism expansion.
© 2003 Elsevier Science (USA). All rights reserved.

Keywords: Adenovirus; Ad3; Ad5; CAR; Chimeric vector

Introduction
Cancer gene therapy using adenoviral vectors is a prom-
ising treatment alternative (Gomez-Navarro et al., 1999). To
date, most adenoviral vectors have been based on serotype
5 of species C, because of its ability to infect a wide variety
of tissue and cell types, its ease of production to high titers,
and its in vivo stability. There are, however, some limita-
tions associated with the use of adenovirus serotype 5 (Ad5)
vectors for cancer gene therapy. One disadvantage is related
to the expression profile of the primary adenovirus receptor,
CAR,1 which binds to the knob domain of the Ad5 fiber

* Corresponding author. Division of Human Gene Therapy, 901 South
19th Street, BMR2 508, University of Alabama at Birmingham, Birming-
ham, AL 35294-2172. Fax: ⫹1-205-975-7476.
E-mail address: david.curiel@ccc.uab.edu (D.T. Curiel).
1
Abbreviations used: CAR, coxsackie and adenovirus receptor; Ad5,
serotype 5 adenovirus; Ad3, serotype 3 adenovirus; Ad5/3, Ad5 containing
a chimeric fiber protein possessing the Ad3 knob; AdMS, mosaic adeno-
virus; Ad7, serotype 7 adenovirus; Ad5/7, Ad5 containing a chimeric fiber

0042-6822/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved.
doi:10.1016/S0042-6822(03)00067-9
 K. Takayama et al. / Virology 309 (2003) 282–293
Fig. 1. Strategy for mosaic virus generation. The virus genome of Ad5- and Ad5/3-based vectors were transferred into 293 cells by coinfection of both viruses.
The cells infected by both viruses express two types of fibers, the Ad5 wild-type fiber and the Ad5/3 chimeric fiber. Both types of fibers assemble on the
virion at random and generate mosaic viruses.
protein (Bergelson et al., 1997; Henry et al., 1994; Louis et
al., 1994; Tomko et al., 1997). Since the efficiency of Ad5
gene transfer primarily depends on the level of CAR, cells
or tissues which express low levels of CAR may be refrac-
tory to gene therapy with Ad5-based vectors (Cripe et al.,
2001; Hemmi et al., 1998; Kaner et al., 1999; Kelly et al.,
2000; Li et al., 1999; Miller et al., 1998; Nalbantoglu et al.,
2001; Pickles et al., 1998; Zabner et al., 1997).
Consequently, strategies to modify Ad tropism to cir-
cumvent CAR deficiency have been developed, based on
the use of bispecific retargeting molecules or genetic capsid
modifications (Krasnykh et al., 2000). One genetic approach
to the modification of Ad tropism involves the incorporation
of targeting ligands in capsid proteins, most commonly the
fiber. An alternative approach exploits the fact that while
Ad5 binds to CAR, the fiber knobs of other Ad serotypes
recognize distinct cellular receptors. This has led to the
hypothesis that efficient, CAR-independent gene transfer
could be accomplished by substituting the fiber proteins, or
simply the fiber knob domains, from an Ad5 vector back-
bone with the homologous proteins from an alternative
serotype—a process known as “pseudotyping” (Havenga et
al., 2001; Krasnykh et al., 1996; Miyazawa et al., 1999;
Stevenson et al., 1997; Von Seggern et al., 2000; Zabner et
al., 1999). In this regard, we have reported the successful
generation of Ad5 virions containing fibers consisting of the
tail and shaft domains of Ad5 fiber and the knob domain of
Ad3 (Ad5/3) (Krasnykh et al., 1996). Ad3 is assigned to
protein possessing the Ad7 knob; hCAR, human CAR; VP, viral particle;
FBS, fetal bovine serum; MLP, major late promoter; RLU, relative light
unit; PVDF, polyvinylidene difluoride; TBST, Tris-buffered saline plus
Tween.
283
species B in adenovirus classification schemes and binds to
a distinct non-CAR receptor that has not yet been identified
(Defer et al., 1990; Gall et al., 1996; Stevenson et al., 1995).
As expected, the chimeric Ad5/3 adenovirus showed altered
tissue tropism and improved infectivity in primary ovarian
cancer cells and various cancer cell lines, characterized by
the levels of CAR, compared to Ad5 (Davidoff et al., 1999;
Kanerva et al., 2002; Stevenson et al., 1997).
Considering these lines of evidence, we hypothesized
that a mosaic virus displaying both the Ad5 and the Ad3
knobs on the same virion could utilize either the CAR
and/or the Ad3 receptor, resulting in enhanced infectivity.
In this study, we have generated such a mosaic adenovirus.
Compared to both the Ad5 and the Ad5/3 vectors possessing
knobs with a single receptor specificity, this mosaic Ad has
significantly enhanced infectivity in various cancer cell
lines and, thus, may offer the potential to improve Ad-based
cancer gene therapy approaches.
Results
Generation of mosaic virus possessing Ad5 and Ad3
knobs
A set of four mosaic adenoviruses designed to display
both the Ad5 and the Ad3 fiber knobs was generated by
coinfection of 293 cells with Ad5luc1 and Ad5/3luc1 at
particle ratios of 90:10, 70:30, 50:50, 30:70 (Fig. 1). DNA
was extracted from each crude virus solution and quantified
by real-time quantitative PCR. An amount of each mosaic
virus preparation, corresponding to 1 ⫻ 108 copies of the
E4, was used to infect 2 ⫻ 105 HeLa cells. The parental
284 K. Takayama et al. / Virology 309 (2003) 282–293

primers amplified only Ad5/3luc1 and AdMSluc1. The use

of the MLP-specific primers indicated that equal amounts of
viral DNA were employed as template. These results there-
fore confirm that the population of mosaic vectors,
AdMSluc1, contained two separate viral genomes, derived
from Ad5luc1 and Ad5/3luc1, which were encapsidated by
the mosaic capsids.

Mosaic virions incorporate both Ad5 and Ad5/3 fibers

Since the two viral genomes incorporated into the mosaic

virus population differ only in the knob region of the fiber,
we tested our prediction that AdMSluc1 would contain two
fiber proteins, the wild-type Ad5 fiber and the chimeric Ad5
fiber with the Ad3 knob domain, Ad5/3. To examine fiber
incorporation during mosaic virus assembly, we performed
Fig. 2. Mosaic virus generation by coinfection of Ad5 and Ad5/3 at various Western blotting analysis using the monoclonal antibody
ratios. HeLa cells were infected by crude virus lysate of Ad51uc1, Ad5/
(MoAb) 4D2 that recognizes the common Ad5 shaft tail
31uc1, mixtures of Ad51uc1⫹Ad5/3luc1 at various ratios (open bar), or
four kinds of mosaic viruses, AdMSluc1 (closed bar). Luciferase activity domain of both types of fiber. As shown in Fig. 4, this
was measured after 48 h incubation and is expressed as relative light units analysis of Ad5luc1 and Ad5/3luc1 revealed a single band
(RLU) per milligram of protein. Each point represents the mean of three corresponding to each fiber monomer protein (lanes 1 and
experiments ⫾ SD. 7). Fibers from Ad5luc1 and Ad5/3luc1 exhibited different
electrophoretic mobility via SDS–PAGE and were identi-
fied easily by MoAb 4D2. In contrast, AdMSluc1, and a
vectors, Ad5luc1 and Ad5/3luc1, were employed as con-
70:30 mixture of Ad5luc1⫹Ad5/3luc1, displayed two bands
trols, together with various mixtures of these two vectors.
corresponding to the Ad5 wild-type fiber and the Ad5/3
The luciferase activities were measured after 48 h, as shown
chimeric fiber (lanes 3 and 5). In both cases, the intensity of
in Fig. 2. Ad5/3luc1 showed 4.2 times higher luciferase
the upper band, corresponding to the Ad5/3 chimeric fiber,
activity in HeLa cells than Ad5luc1, while the luciferase
activities of the various mixtures of Ad5luc1 ⫹ Ad5/3luc1
ranged between those of 100% Ad5luc1 and 100% Ad5/
3luc1, increasing with the Ad5/3luc1 content of the mixture.
Surprisingly, each of four mosaic viruses showed a higher
luciferase activity than an equal number of particles of
Ad5/3luc1 alone. The mosaic virus generated by coinfecting
293 cells at an Ad5:Ad5/3 ratio of 70:30 showed the highest
luciferase activity in HeLa cells, 2.8 times and 12 times
higher than Ad5/3luc1 and Ad5luc1, respectively. This ob-
servation encouraged us to perform quantitative analysis
using the purified mosaic virus.

The mosaic virus population contains two separate viral

genomes

Based on the preliminary results, we selected the ratio of

70% Ad5luc1 and 30% Ad5/3luc1 for a large-scale produc-
tion of the mosaic virus (AdMSluc1). The mosaic
AdMSluc1, Ad5luc1, and Ad5/3luc1 were purified by two
rounds of cesium chloride density centrifugation. The virus
particle number (VP) of each purified virus was determined
from the absorbance of 260 nm. PCR was performed with
primers specific for either the Ad5 or the Ad3 knob, while
primers specific for the major late promoter (MLP) were
used to quantify the viral genome. As shown in Fig. 3, the
Ad5 knob-specific primers amplified only the Ad5luc1 and
AdMSluc1 viral genomes, while the Ad3 knob-specific
Fig. 3. Amplication of viral DNA by PCR. DNA extracted from 1 ⫻ 108
VP of each virus was amplified by PCR with primer pairs corresponding to
major late promoter region, Ad5 knob region, and Ad3 knob region. Each
region was amplified in 25 or 28 cycles of PCR. The template DNA was
extracted from Ad5luc1 (lane 1), a mixture of 70% Ad5luc1 ⫹ 30%
Ad5/31uc1 (lane 2), AdMSluc1 (lane 3), or Ad5/3luc1 (lane 4). All PCR
products showed the correct size fragment confirmed by size marker (lane
M) and PCR product from each backbone plasmid as a positive control
(lane C).
 K. Takayama et al. / Virology 309 (2003) 282–293 285

Fig. 4. Western blotting for fiber protein. 2 ⫻ 108 VP of each purified virus sample were dissociated in SDS sample buffer and heated at 96°C for 3 min.
Viruses, bound to immobilized MoAb 1D6.14 on a 96-well plate, were also solubilized in SDS buffer after the binding reaction. All samples were separated
on a 4 to 12% gradient polyacrylamide gel followed by electrotransfer onto PVDF membrane. After being blocked with TBST-casein, the membrane was
treated with MoAb 4D2, followed by treatment with goat anti-mouse IgG-HRP conjugate. The color reaction was developed by incubation of the membrane
with Sigma Fast diaminobenzidine. Lane 1, Ad5luc1 without treatment; 2, Ad5luc1 with MoAb treatment; 3, Ad5luc1⫹Ad5/31uc1 mixture without treatment;
4, Ad51uc1⫹Ad5/31uc1 mixture with treatment; 5, AdMS1uc1 without treatment; 6, AdMSluc1 with treatment; 7, Ad5/3luc1 without treatment; 8,
Ad5/3luc1 with treatment.

was weaker than the lower band, corresponding to the Ad5
wild-type fiber. AdMSluc1 also seemed to have more Ad5
wild-type fiber than chimeric fiber, similar to the mixture of
Ad5⫹Ad5/3. Following densitometric analysis of band in-
tensity, lanes 1, 3, 5, and 7 were similar, indicating that
equivalent numbers of VPs contained a similar total amount
of fiber protein.
These viruses were also applied to an ELISA plate for
specific virus binding to an immobilized anti-Ad5 knob
MoAb, 1D6.14. After washing, virus still bound to the wells
was solubilized in SDS buffer and analyzed by SDS–PAGE
with each intact virus as a control. The lysate of Ad5luc1
bound to the immobilized MoAb showed the same band as
intact Ad5luc1 (lanes 1 and 2) and the lysate from the well
incubated with Ad5/3luc1 showed no band (lanes 7 and 8),
indicating that MoAb 1D6.14 is able to bind the Ad5 knob
specifically without any cross-reaction with the Ad5/3 chi-
meric fiber. We obtained additional evidence about the
specificity of MoAb 1D6.14 from the lysate of the Ad5luc1
⫹ Ad5/3luc1 mixture. After the binding reaction with the
MoAb 1D6.14, only the Ad5 fiber protein was detected in
the well incubated with Ad5luc1 ⫹ Ad5/3luc1 mixture
(lanes 3 and 4). Finally, we treated the AdMSluc1 in the
same way and detected both fiber bands in the mosaic virus
lysates following binding to the ELISA plate (lanes 5 and
6). Based on these results, we concluded that the majority of
the virions in the AdMSluc1 population contain both the
Ad5 wild-type fiber and the Ad5/3 chimeric fiber within the
same particle.
Infectivity profile of mosaic adenovirus in various cell
lines
The mosaic virus demonstrated enhanced infectivity
on HeLa cells compared with Ad5luc1 and Ad5/3luc1, as
shown in Fig. 2. However, it was unclear whether the
infectivity enhancement was specific for this cell line. To
address this question, we examined the infectivity profile
of AdMSluc1 in various cell lines, as compared to
Ad5luc1 and Ad5/3luc1. We selected nine cell lines in-
cluding ovarian cancer, non-small-cell lung cancer, pan-
creatic cancer, glioma, cervical cancer, and 293HEK
cells, based on their expression of the primary receptors
for Ad5 (CAR) and Ad3. After infection with each virus,
the luciferase activities in this panel of cell lines were
measured. In our previous work, we reported that
SKOV3.ip1 and OV-4 express the Ad3 receptor domi-
nantly, while 293HEK cells express CAR as the domi-
nant receptor (Kanerva et al., 2002). Consistent with our
previous results, SKOV3.ip1 and OV-4 showed signifi-
cantly higher luciferase activity with Ad5/3luc1 than
Ad5luc1, as shown in Fig. 5. Conversely, some cell lines
such as C33A, H157, and 293HEK cells were more
susceptible to Ad5luc1 infection, presumably due to their
dominant CAR expression. The luciferase activity of the
Ad5luc1 and Ad5/3luc1 mixture was also examined in
each cell line and showed intermediate values between
those for Ad5luc1 and Ad5/3luc1. No significant en-
hancement of infectivity was shown by this virus mixture.
However, the luciferase activity with AdMSluc1 was superior
to levels noted with Ad5luc1 and Ad5/3luc1 in all cell lines
tested. AdMSluc1 showed a 300-fold increase in luciferase
activity compared to Ad5luc1, in OV-4 cells. Likewise,
AdMSluc1 showed 5.9 times higher luciferase activity than
that of Ad5/3luc1 in C33A cells. These data revealed that
the mosaic virus infected cells more efficiently than
Ad5luc1 and Ad5/3luc1, regardless of the dominant knob
receptor. Furthermore, the mosaic virus luciferase activity
was higher than the sum total of luciferase activities with
Ad5luc1 and Ad5/3luc1 in all cells tested, indicating that the
presence of Ad5 and Ad3 knobs on the same virion creates
a synergistic increase in Ad infectivity: we term this phe-
nomenon the “mosaic effect.”
Analysis of mosaic Ad binding using purified Ad3 and
Ad5 knob protein
Although we showed an infectivity enhancement with
the mosaic virus in various cell lines, we wished to
determine the manner in which the mosaic Ad utilized
CAR and/or the Ad3 receptor during infection. To clarify
this point, we performed infection-blocking experiments
with purified Ad5 knob and/or Ad3 knob. For these
experiments, we selected two cell lines: C33A as a CAR-
286 K. Takayama et al. / Virology 309 (2003) 282–293

Fig. 5. Infectivity profile of Ad5, Ad5/3, and mosaic virus in various cell lines. Ovarian cancer cells, SKOV3.ipl and OV-4, pancreas cancer cells, Panc-I,
lung cancer cells, NCI-H157 and NCI-H358, uterine cervical cancer cells, HeLa and C33A, glioma cells, U118, and 293HEK cells were infected by Ad5luc1
(open bar), Ad5/3luc1 (closed bar), a mixture of Ad5luc1⫹Ad5/3luc1 (gray bar), or AdMSluc1 (checked bar) at 100 VP/cell. The luciferase activity was
measured after 48 h incubation and expressed as relative light units (RLU). Each point represents the mean of three experiments ⫾ SD.

dominant cell line and SKOV3.ip1 as an Ad3 receptor-

Fig. 6. FACS analysis for CAR expression in tumor cell lines. 2 ⫻ 105 of
SKOV3.ipl, HeLa, and C33A cells were incubated with primary MoAb
RmcB to human CAR. The cells were then washed with 1 ml PBS/BSA/
azide and incubated with the secondary FITC-labeled goat anti-mouse IgG.
Then 1 ⫻ 104 cells were analyzed by flow cytometry. Expression levels of
CAR and negative control are shown as areas filled in gray and black,
respectively.
dominant cell line. In fact, a higher percentage of C33A
cells than SKOV3.ip1 cells are CAR-positive by FACS
analysis, as shown in Fig. 6, and a rough correlation
was shown between CAR expression and luciferase ac-
tivity with Ad5luc1 in these cell lines. With C33A cells,
purified Ad5 knob protein blocked gene transfer of mo-
saic Ad significantly at the lowest concentration of 0.1
␮g/ml, and this blocking effect increased in a dose-
dependent manner, as shown in Fig. 7A. Purified Ad3
knob protein did not show any suppressive effect even at
the highest concentration of 100 ␮g/ml, but the combi-
nation of Ad5 and Ad3 knob protein blocked the mosaic
virus infection more strongly than Ad5 knob alone. On
the other hand, purified Ad3 knob blocked the mosaic
virus infection in SKOV3.ip1 cells in a dose-dependent
manner, as shown in Fig. 7B. However, infectivity of the
mosaic virus was not blocked by Ad5 knob, a finding
consistent with our previous results (Kanerva et al.,
2002). Further, the addition of Ad5 knob to Ad3 knob
showed no additional effect. These data demonstrated
that the mosaic virus is able to bind to CAR-dominant
cells with Ad5 wild-type fiber and to Ad3 receptor-
dominant cells with Ad5/3 chimeric fiber.
 K. Takayama et al. / Virology 309 (2003) 282–293 287

Fig. 7. Mosaic virus infectivity in the presence of purified Ad3 and/or Ad5 knob protein. 1 ⫻ 105 (A) C33A cells and (B) SKOV3.ipl cells were preincubated
with the indicated concentrations of each recombinant knob protein for 10 min followed by the AdMSluc1 infection at 100 VP/cell. The knob proteins used
for competition were as follows: closed square: Ad3 knob; closed diamond: Ad5 knob; closed triangle: Ad3 ⫹ Ad5 knob containing each knob protein at
the indicated concentration. A luciferase assay was performed after 48 h incubation, and the luciferase activity is expressed as relative light units (RLU). Each
point represents the mean of three experiments ⫾ SD.

Virus infectivity is not enhanced by free fiber protein
We next sought to determine if the engagement of one
Ad receptor serotype by its cognate knob was sufficient to
improve the infectivity of the virus bearing the other sero-
type knob, resulting in an improvement in infectivity. We
therefore examined the infectivity enhancement of Ad5luc1
and Ad5/3luc1 in the presence of the respective free knob
proteins. HeLa cells were infected with Ad5luc1 or Ad5/
3luc1 in the presence of various concentrations of Ad5 or
Ad3 knob protein. As shown in Fig. 8, the luciferase activ-
ities with Ad5luc1 and Ad5/3luc1 infection were suppressed
by recombinant Ad5 knob and Ad3 knob, respectively, in a
dose-dependent manner. At a concentration of 100 ␮g/ml of
each knob protein, the luciferase activities of the relevant
adenovirus were below 5% of those without knob protein.
Importantly, the luciferase activity of Ad5luc1 showed no
change at the highest concentration of Ad3 knob protein.
Likewise, Ad5/3luc1 infectivity was also not affected by
Ad5 knob protein. We repeated the same experiment with
each virus at 1000 VP/cell and obtained similar results (data
not shown). Thus, these studies confirmed that each knob
protein alone could not induce the infectivity enhancement.
Furthermore, this was also consistent with the previous
result that the mixture of both viruses did not show any
infectivity enhancement regardless of the mixture ratio, as
shown in Fig. 2.
Cytoplasmic tail domain of CAR is not involved in the
mosaic effect
We investigated the possibility of a contribution of
CAR–Ad3 receptor interaction to mosaic Ad infectivity
enhancement. We hypothesized that any receptor interac-
tion resulting in positive cooperativity of mosaic infectivity
would require the intracellular domain of CAR. To directly

Fig. 8. Infectivity of Ad5 and Ad5/3 in the presence of purified Ad3 or Ad5 knob protein. 1 ⫻ 105 HeLa cells were preincubated with the indicated
concentrations of (A) purified Ad5 knob protein or (B) purified Ad3 knob protein for 10 min followed by Ad5luc1 (closed diamond) or Ad5/3luc1 (closed
square) infection at 100 VP/cell. Luciferase assay was performed after 48 h incubation, and the luciferase activity is expressed as relative light units (RLU).
Each point represents the mean of three experiments ⫾ SD.
288 K. Takayama et al. / Virology 309 (2003) 282–293
Fig. 9. Correlation between mosaic virus infectivity and CAR expression level. 1 ⫻ 105 parental U118 cells (open bar), modified U118-hCAR cells (gray
bar), and U118-hCAR-tailless cells (closed bar) were infected by Ad5luc1, Ad5/3luc1, and AdMSluc1 at 100 VP/cell. Luciferase assay was performed after
48 h incubation. Each luciferase activity with Ad5luc1 or AdMSluc1 is expressed as the ratio to the Ad5/3 luciferase activity.
address this issue, we utilized the CAR-negative U118 hu-
man glioma cell line and two genetic variants expressing
either human CAR (U118-hCAR) or an hCAR deletion
mutant lacking the intracellular domain (U118-hCAR-tail-
less). Importantly, parental U118 cells and the two hCAR
derivatives appear to express the Ad3 receptor equally, as
evidenced by similar luciferase activities following Ad5/
3luc1 infection (data not shown). Initially, we compared the
luciferase activities of Ad5luc1 and Ad5MSluc1 in all three
cell lines, expressed as a percentage of Ad5/3luc1 luciferase
activity for purposes of normalization. As expected,
Ad5luc1 transgene activity in the hCAR and hCAR-tailless
expressing cell lines was increased 11-fold and 16-fold,
respectively, compared to the CAR-negative parental line as
shown in Fig. 9. Likewise, mosaic virus luciferase activity
increased 6-fold in U118-hCAR cells and 7.5-fold in the
U118-hCAR-tailless line versus parental U118 cells. Impor-
tantly, we observed no reduction in AdMSluc1 infectivity in
the U118-hCAR-tailless cells versus those expressing full-
length CAR. These findings suggest that the CAR cytoplas-
mic domain does not contribute to the infectivity enhance-
ment of mosaic virus. In other words, CAR and the Ad3
receptors do not interact with each other through their cy-
toplasmic tails to enhance the mosaic virus infectivity.
Discussion
A major obstacle to be overcome in Ad5-based cancer
gene therapy has been the paucity of the primary receptor,
CAR, on human primary tumor cells. Variable expression of
CAR has been documented in many cancer types including
glioma, melanoma, bladder cancer, rhabdomyosarcoma,
neuroblastoma, and ovarian cancer (Cripe et al., 2001;
Davidoff et al., 1999; Fechner et al., 2000; Hemmi et al.,
1998; Kanerva et al., 2002; Kelly et al., 2000; Li et al.,
1999; Miller et al., 1998). Moreover, it was also reported
that CAR protein is down regulated in highly tumorigenic
prostate cancer cell lines (Okegawa et al., 2000). Due to
variable expression of CAR on human primary cancer cells,
the utility of Ad5 as a cancer gene therapy vector is com-
promised, limiting overall efficiency of cancer gene therapy.
On this basis, systems to circumvent intervening tumor-
associated CAR deficiency are required.
In this regard, native Ad5 tropism can be modified to
circumvent CAR deficiency and to enhance the adenovirus
infectivity. One approach is pseudotyping, i.e., retargeting
Ad by creating chimeric fibers possessing knob domains
derived from alternate serotypes which bind to receptors
other than CAR. To this end, we have constructed nonrep-
licating Ads containing chimeric fibers with the tail and
shaft domains of Ad serotype 5 and the knob domain of
serotype 3 (Krasnykh et al., 1996). Our previous work has
revealed that a distinct Ad3 receptor exists in ovarian cancer
cells based on a novel knob binding assay and that the
Ad5/3 chimeric vector is retargeted to the Ad3 receptor
(Kanerva et al., 2002). Furthermore, competition assays
with purified Ad5 and Ad3 knob protein suggested that the
knob regions of Ad5 and Ad5/3 interact with their cognate
receptor independently. These observations led us to make
a dual-knob mosaic virus possessing both Ad5 and Ad3
knob regions. In this regard, coinfection of 293 cells with
Ad5 and Ad5/3 gave rise to a population of viruses pos-
sessing both the Ad5 wild-type fiber and the Ad5/3 chimeric
fiber on the same virion.
From competition assays with purified knob, we con-
firmed that the mosaic virus can attach to either cellular
receptor, CAR or Ad3 receptor. As expected, this mosaic
virus enhanced the Ad infectivity in various cell lines. The
lucifease activity mediated by the mosaic virus exceeded the
sum total of that with Ad5luc1 and Ad5/3luc1 in all cells
tested, suggesting synergistic effect on infection. Interest-
 K. Takayama et al. / Virology 309 (2003) 282–293
ingly, recombinant Ad5 knob protein did not block the
mosaic virus infectivity in HeLa cells despite the fact that a
significant suppression effect was observed with anti-Ad5
knob antibody. Taken together, the mosaic effect seemed to
be synergistic rather than additive. Since the affinity of each
knob for its cognate receptor is unchanged, the coexistence
of both knobs on the same virion likely contributes to the
synergistic infectivity enhancement.
One explanation for the synergistic infectivity enhance-
ment is that binding of the Ad3 knob to the Ad3 receptor
triggers an alteration in the CAR–Ad5 knob affinity, with
the interaction between the two receptors inducing confor-
mational or functional alterations of CAR. Previous studies
with mutated CAR have reported that the cytoplamic tail
functions as a sorting domain for basolateral targeting in
epithelial cells (Cohen et al., 2001) and is not essential for
adenoviral infection (Van’t Hof and Crystal, 2001). How-
ever, other recent reports have suggested that overexpressed
CAR acts as a tumor inhibitor in prostate cancer (Okegawa
et al., 2000) and might have an unknown function other than
sorting. In this study, the comparison between the U118-
hCAR cells and U118-hCAR-tailless cells indicated that
CAR and the Ad3 receptor do not seem to interact with each
other via cytoplasmic domain.
Another possible explanation of the mosaic effect is a
positional advantage of the Ad5 knob on the mosaic virus
anchored to the Ad3 receptor. If the Ad3 receptor and CAR
colocalize on the cell surface, the Ad5 knob may easily
access to CAR close to the Ad3 receptor, allowing compe-
tition with recombinant Ad5 knob to be overcome.
The pathway of intracellular trafficking of Ad5 differs
from that of Ad3 (Defer et al., 1990). Ad5 uses the cyto-
plasmic route preferentially after internalization, while Ad3
is sequestered within phagosomes. A study with an Ad5/7
chimeric vector possessing the Ad7 fiber on the Ad5 capsid
reported that this Ad5/7 chimera traffics via the lysosomal
pathway as well as Ad7 (Miyazawa et al., 1999). If the fiber
protein modulates the virus trafficking pathway, the mosaic
virus may utilize both pathways to the nucleus. This is
another possible reason for synergistic enhancement of
transgene expression. The precise mechanism for the mo-
saic effect observed in this study will be elucidated by the
detailed analysis of Ad3 receptor once this has been cloned.
As the result of the presence of both the Ad5 and the Ad3
knobs on the same virion, the mosaic virus displays ex-
panded tropism covering the infectivity spectrum of both
Ad5 and Ad3 and may be promising for cancer gene therapy
approaches involving gene transfer to cells or tissues refrac-
tory to Ad5 and Ad3 infection. Transgene expression from
various preexisting Ad5-based vectors should be enhanced
by this mosaic formation strategy in combination with Ad5/
3-based vector. An Ad5/7 chimeric vector may also be
available for mosaic virus formation since, similar to Ad3,
Ad7 belongs to subgroup B. This mosaic virus strategy is an
easy and feasible method for infectivity enhancement and
tropism expansion.
289
Concerning the mosaic variation, the ratio of Ad5 and
Ad3 knob number on the virion seems to be important, as
shown in Fig. 2. Since the mosaic virus population is het-
erogeneous, a part of the population may not contribute to
the infectivity enhancement due to the deviated ratio of
knob. This means that the data in this study may represent
the total effect of a heterogeneous virus population with
various knob ratios. Therefore, selection of the mosaic virus
with optimal knob ratio could potentially improve the mo-
saic effect more intensely. If the spatial relationship be-
tween both knobs and their receptors is involved in the
mosaic effect, alignment of each knob on the virion is
responsible for it.
In conclusion, we have described a mosaic adenovirus
generated by coinfection of 293 cells of Ad5 and Ad5/3
chimeric virus. This mosaic virus displays expanded tro-
pism and enhances the Ad infectivity in various cancer cell
lines. Thus, mosaic virus strategy has a possibility to im-
prove the effect of cancer gene therapy in various contexts.
Materials and methods
Cell culture
Human cervical cancer cell lines HeLa and C33A, hu-
man non-small-cell lung cancer cell lines NCI-H157 and
NCI-H358, human pancreatic cancer cell line Panc-1, hu-
man breast cancer cell line AU-565, and the human glioma
cell line U118 were obtained from the American Type
Culture Collection (Manassas, VA). The 293 human trans-
formed embryonal kidney cell line was purchased from
Microbix (Toronto, Ontario, Canada). Human ovarian ade-
nocarcinoma cell line SKOV3.ip1 was obtained from Dr.
Janet Price (M.D. Anderson Cancer Center, Houston, TX).
Human ovarian adenocarcinoma cell line OV-4 was ob-
tained from Dr. Timothy J. Eberlein (Harvard Medical
School, Boston, MA). All cell lines were cultured in the
recommended media at 37°C in a humidified atmosphere of
5% CO2.
Construction of cell lines expressing hCAR and an hCAR
truncation mutant that lacks the cytoplasmic domain
To assess the role of the cytoplasmic domain of CAR in
infection by the mosaic Ad, we engineered a truncated form
of human CAR (hCAR) lacking this domain. The 2.4-kb
BamHI-NotI fragment carrying the hCAR cDNA was first
subcloned from pcDNAI.hCAR (obtained from Robert W.
Finberg, Harvard Medical School, Boston, MA) into the
mammalian expression vector pcDNA3 (Invitrogen, Carls-
bad, CA). A truncation mutant designated hCAR-tailless, in
which a stop codon was inserted into the cDNA at the
position corresponding to amino acid residue 261 (Freimuth
et al., 1999), was constructed by PCR mutagenesis using
pcDNA3-hCAR as the template. The PCR product was
290 K. Takayama et al. / Virology 309 (2003) 282–293
inserted into the BamHI and NotI sites of pcDNA3 to give
pcDNA3-hCARtailless. The integrity of the construct was
verified by DNA sequencing.
Human U118 glioma cells, which are refractory to Ad
infection due to a paucity of CAR although they express the
␣v integrins necessary for virus internalization (Miller et al.,
1998), were stably transfected with pcDNA3-hCARtailless
or with pcDNA3-hCAR as a control. Individual single-cell
clones of U118-hCAR-tailless and U118-hCAR were iso-
lated and expanded by selection in the presence of 400
␮g/ml G418.
Recombinant adenoviruses
Two replication-incompetent Ad vectors containing a
firefly luciferase transgene cassette in place of the deleted
E1 region were used. Ad5luc1 was generated in our labo-
ratory and has been described previously (Krasnykh et al.,
2001). The genome of Ad5/3luc1 (Ad containing chimeric
fibers with the tail and shaft domains of Ad serotype 5 and
the knob domain of serotype 3) was constructed by homol-
ogous DNA recombination in Escherichia coli using the
previously described plasmid pNEB.PKF5/3 in a scheme
described by Dmitriev et al. (1998). The vector of interest
was rescued by transfecting 293 cells with the resultant Ad
genome. The viruses were propagated on 293 cells and
purified by two rounds of cesium chloride density centrif-
ugation. The VP concentration was determined spectropho-
tometrically, using a conversion factor of 1.1 ⫻ 1012 viral
particles per absorbance unit at 260 nm (Maizel et al.,
1968), and standard plaque assays on 293 cells were per-
formed to determine infectious particles (Mittereder et al.,
1996). The ratio of VP:infectious particles was 20, 61, and
15 for Ad5luc1, Ad5/3luc1, and AdMSluc1, respectively.
Generation of mosaic adenovirus
Mosaic adenoviruses, designated AdMSluc1, were gen-
erated by coinfection of 293 cells with both Ad5luc1 and
Ad5/3luc1. Monolayers of 293 cells in T75 flasks were
coinfected with a total of 2 ⫻ 109 particles of Ad5luc1 and
Ad5/3luc1 in mixtures containing 100, 90, 70, 50, 30, 10, or
0% of Ad5luc1. After infection for 3 h at 37°C, the infection
medium was replaced with fresh complete medium, and the
293 cells were incubated for 72 h. The 293 cells were
harvested and resuspended in 1 ml of DMEM:F12 medium
containing 2% FBS and subjected to four freeze-thaw cy-
cles. The crude viral lysates were used in the initial exper-
iments. Based on the results of these preliminary experi-
ments, a ratio of Ad5luc1:Ad5/3luc1 of 7:3 was chosen for
coninfection of 293 cells for the large-scale preparation of
the mosaic adenovirus, AdMSluc1. Purification and titering
were performed as described above.
PCR amplification of viral genome
For 20 min 1 ⫻ 108 VP of Ad5luc1, Ad5/3luc1, or
AdMSluc1 were incubated in 0.1% SDS and 1 mM EDTA
at 56°C to extract the viral DNA. The DNA was diluted
1/1000 in TE and 1 ␮l of this solution was used as a
template for PCR. Each viral genome was amplified in a
PCR reaction mixture (Qiagen, Valencia, CA) containing 50
nM primer pairs with 25 or 28 cycles of denaturation (94°C,
1 min), annealing (50°C, 1 min), and extension (72°C, 1
min). The sequences of the primers were as follows:
Ad5knob-sense: 5⬘ AGTGCTCATCTTATTATAAGA3⬘;
Ad3knob-sense: 5⬘CGCACATCCTATGTTATG3⬘; knob-
antisense: 5⬘CACCACCGCCCTATCCTGAT3⬘; MLP-sense:
5⬘GGTTAATTAAGCATGTCCCTGACTCGCAT3⬘; MLP-
antisense: 5⬘TTGCGCGTGCACCTGGTGCCCGACGA3⬘.
The Ad5luc1 and Ad5/3luc1 backbone plasmids were used as
positive control templates.
To quantify the viral genome DNA in the crude virus
solutions, the extracted DNA was amplified by real-time
PCR as reported previously (Adachi et al., 2001). The viral
DNA was isolated from 200 ␮l of each crude viral solution
using a Blood DNA kit (Qiagen). Viral DNA was eluted
with 100 ml of elution buffer [10 mM Tris–Cl (pH 8.5)].
One microliter of eluted sample was analyzed by real-time
PCR using a Light Cycler (Roche) to evaluate the Ad5 E4
copy number. Oligonucleotides corresponding to the sense
strand of the Ad5 E4 region (5⬘-TGACACGCATACTCG-
GAGCTA-3⬘: 34,885–34,905 nt), the antisense strand of the
E4 region (5⬘-TTTGAGCAGCACCTTGCATT-3⬘: 34,977–
34,958 nt), and a TaqMan probe (5⬘-CGCCGCCCATG-
CAACAAGCTT-3⬘: 34,930 – 34,951 nt) were synthesized
and used as primers and probe for real-time PCR analysis.
The PCR conditions were as follows: 35 cycles of denatur-
ation (94°C, 20 s), annealing (55°C, 20 s), and extension
(72°C, 30 s). An Ad backbone vector pTG3602 (Transgene,
Strasbourg, France) was used to generate a standard curve
for Ad E4 DNA copy number.
Western blotting
Western blotting for fiber protein was performed as de-
scribed previously (Pereboev et al., 2001; Seki et al., 2002).
Briefly, 2 ⫻ 108 purified Ad virions (Ad5luc1, Ad5/3luc1,
or AdMSluc1) were diluted in 2⫻ SDS sample buffer (Bio-
Rad, Hercules, CA) and heated at 96°C for 3 min. Virus
samples were separated on a 4 to 12% gradient polyacryl-
amide gel followed by electrotransfer onto a polyvinylidene
difluoride (PVDF) membrane. After blocking with TBST-
casein [Tris-buffered saline (10 mM Tris–HCl, pH 7.4, 150
mM NaCl) plus Tween 20 to 0.05% and casein to 0.5%],
membranes were treated with mouse monoclonal antibody,
MoAb 4D2 (1:2000 dilution). Then, the membrane was
incubated with goat anti-mouse IgG-HRP conjugate
(DAKO, Carpinteria, CA). The color reaction was devel-
oped by incubation of the membrane with Sigma Fast dia-
 K. Takayama et al. / Virology 309 (2003) 282–293
minobenzidine (Sigma). MoAb 4D2 directed against the tail
domain of Ad5 fiber protein was generated at the University
of Alabama at Birmingham Hybridoma Core Facility (Dmi-
triev et al., 2000; Karayan et al., 1994).
Recombinant fiber knob protein production
Recombinant Ad5 and Ad3 fiber knob proteins with
N-terminal 6⫻ His tags were expressed in E. coli using the
pQE30 expression vector (Qiagen). To increase protein
yields, the expressed Ad3 fiber knob protein was purified
from inclusion bodies using the BugBuster Protein Extrac-
tion Reagent (Novagen, Madison, WI) inclusion body pro-
tocol as recommended by the manufacturer. The protein was
further purified using the Talon metal affinity resin (Clon-
tech, Palo Alto, CA) as recommended by the manufacturer.
The concentration of the purified proteins was determined
by Bio-Rad DC protein assay. The ability of each knob
protein to form a homotrimer was verified by Western blot
of unboiled samples. The primary antibody used for detec-
tion was the monoclonal anti-polyhistidine clone HIS-1
antibody (Sigma) and the secondary antibody was peroxi-
dase-conjugated goat antimouse IgG (DAKO).
Virus binding to immobilized monoclonal anti-Ad5 knob
antibody
Monoclonal antibody, MoAb 1D6.14, which specifically
recognizes the trimeric Ad5 fiber knob protein, has been
described previously (Douglas et al., 1996). Briefly, anti-
knob MoAbs were generated by established methods after
immunization of BALB/c mice with Ad5, followed by two
rounds of immunization with purified recombinant Ad5
knob. Sensitized lymphocytes were fused with P3-X63-
Ag8.653 cells. The reactivity of the hybridoma supernatants
with trimeric Ad5 knob was determined in an ELISA. The
ability of the hybridoma supernatants to neutralize Ad5
infection was determined in an endpoint cytopathic effect
assay. Reactivity with the Ad fiber shaft can therefore be
excluded.
MoAb 1D6.14 or control mouse IgG (Sigma) were di-
luted in 50 mM NaHCO3 (pH 9.6) at a concentration of 5
␮g/ml and immobilized on a Nunc-Maxisorp ELISA plate
overnight. After unbound MoAb was removed, each well
was washed three times with TBS-T (50 mM Tris–HCl, pH
7.6, in normal saline). Each well was then filled with 200 ␮l
of DMEM:F12 media containing 10% FBS to block non-
specific binding at room temperature for 2 h. After blocking,
each well was washed three times with TBS and used for
virus binding. Placed in the well and incubated at RT for 2 h
were 8 ⫻ 107 or 8 ⫻ 108 VP of Ad5luc1, Ad5/3luc1,
AdMSluc1, or a mixture of Ad5luc1 ⫹ Ad5/3luc1 in 80 ␮l
of DMEM:F12 containing 2% FBS. Then, 10 ␮l of the virus
solution in each well was used to infect 1 ⫻ 105 HeLa cells
to examine the luciferase activity. The wells were then
washed with TBS three times and viruses bound to the well
291
were solubilized in 1⫻ SDS buffer and applied to a gradient
polyacrylamide gel. Western blotting was performed as
described above.
Competitive binding assay
To investigate the ability of recombinant Ad5 and Ad3
knobs to block adenovirus infection, Ad5luc1 and Ad5/
3luc1 were added to cells in the presence of purified knob
proteins. Monolayers of C33A, SKOV3.ip1, and HeLa cells
in 12-well plates were preincubated with increasing concen-
trations of Ad5 or Ad3 knob in 500 ␮l DMEM:F12 con-
taining 2% FBS for 10 min at room temperature. Ad5luc1,
Ad5/3luc1, or AdMSluc1 were added at 100 or 1000 VP/
cell, diluted in 100 ␮l of DMEM:F12 containing 2% FBS,
followed by 1 h incubation at room temperature. The cells
were washed once with DMEM:F12 containing 2% FBS
and maintained in complete media. After 48 h of incubation
at 37°C, the cells were lysed and luciferase assays per-
formed as described below. In the experiment with the
anti-Ad5 knob monoclonal antibody, various amounts of
1D6.14 were preincubated with 1 ⫻ 108 VP of each virus at
RT for 30 min in a total volume of 20 ␮l HBS. Then, the
virus–antibody complexes were diluted to 1 ml in DMEM:
F12 containing 2% FBS, and 100 ␮l aliquots were added to
1 ⫻ 105 cells in 12-well plates.
Ad-mediated gene transfer assays
In 12-well plates 1 ⫻ 105 cells were infected for 3 h at
37°C at 10, 100, and 1000 VP/cell by adding Ad5luc1,
Ad5/3luc1, or AdMSluc1 diluted in 500 ␮l of DMEM:F12
containing 2% FCS. Cells were washed once with DMEM:
F12 containing 2% FCS and incubated for 48 h in complete
medium. The luciferase activity in the cell lysate was de-
termined with a luciferase assay kit (Luciferase Assay Sys-
tem, Promega) and a FB12 luminometer (Zylux Corp.).
Background luciferase activities were subtracted from the
readings.
Determination of receptor expression by flow cytometry
Cells grown in T75 flasks were washed with PBS, har-
vested by incubating with 0.53 mM EDTA in PBS, and
resuspended in PBS containing 0.1% sodium azide and 1%
BSA Cells (2 ⫻ 105) were incubated with 1 ml of a 1:50
dilution of primary MoAb RmcB to human CAR for 1 h at
4°C. The MoAb RmcB was produced using a hybridoma
purchased from ATCC (Fechner et al., 2000). An isotype-
matched normal mouse IgG1 was used as negative control.
The cells were washed once with 1 ml of PBS/BSA/azide
and incubated with 1 ml of 1:100 dilution of the secondary
FITC-labeled goat anti-mouse IgG (Jackson Immunore-
search Laboratories, West Grove, PA). Ten thousand cells
were analyzed immediately by flow cytometry at the Uni-
versity of Alabama at Birmingham FACS Core Facility.
292 K. Takayama et al. / Virology 309 (2003) 282–293
The FITC-positive cell population for each cell line was
determined by gating cells incubated with buffer only (neg-
ative control) at 1%.
Acknowledgments
We thank Aleksandr Pereboev, Larisa Pereboeva,
Mikhali Shakhmatov, and Lioudmila Kaliberova for excel-
lent technical support. This work was supported by National
Cancer Institute Grant CA83821, The CapCURE Founda-
tion, The Lustgarten Foundation, United States Department
of Defense Grant 991018, and the American Cancer Soci-
ety.
References
Adachi, Y., Reynolds, P.N, Yamamoto, M., Wang, M., Takayama, K.,
Matsubara, S., Muramatsu, T., Curiel, D.T., 2001. A midkine promoter-
based conditionally replicative adenovirus for treatment of pediatric
solid tumors and bone marrow tumor purging. Cancer Res. 61, 7882–
7888.
Bergelson, J.M., Cunningham, J.A., Droguett, G., Kurt-Jones, E.A., Krithi-
vas, A., Hong, J.S., Horwitz, M.S., Crowell, R.L., Finberg, R.W., 1997.
Isolation of a common receptor for coxsackie B viruses and adenovi-
ruses 2 and 5. Science 275, 1320 –1323.
Cohen, C.J., Gaetz, J., Ohman, T., Bergelson, J.M., 2001. Multiple regions
within the coxsackieviurs and adenovirus receptor cytoplasmic domain
are required for basolateral sorting. J. Biol. Chem. 276, 25392–25398.
Cripe, T.P., Dunphy, E.J., Holub, A., Vasi, N.H., Mahller, Y.Y., Collins,
M.H., Snyder, J.D., Krasnykh, V., Curiel, D.T., Wickham, T., DeGre-
gori, J., Bergelson, J.M., Currier, M.A., 2001. Fiber knob modifications
overcome low, heterogeneous expression of the coxsackievirus-adeno-
virus receptor that limits adenovirus gene transfer and oncolysis for
human rhabdomyosarcoma cells. Cancer Res. 61, 2951–2960.
Davidoff, A.M., Stevenson, S.C., McClelland, A., Shochat, S.J., Vanin,
E.F., 1999. Enhanced neuroblastoma transduction for an improved
antitumor vaccine. J. Surg. Res. 83, 95–99.
Defer, C., Belin, M.-T., Cailet-boudin, M.-L., Boulanger, P., 1990. Human
adenovirus-host cell interactions: comparative study with members of
subgroups B and C. J. Virol. 64, 3661–3673.
Dmitriev, I., Kashentseva, E., Rogers, B.E., Krasnykh, V., Curiel, D.T.,
2000. Ectodomain of coxsackievirus and adenovirus receptor geneti-
cally fused to epidermal growth factor mediates adenovirus targeting to
epidermal growth factor receptor-positive cells. J. Virol. 74, 6875–
6884.
Dmitriev, I., Krasnykh, V., Miller, C.R., Wang, M., Kashentseva, E.,
Mikheeva, G., Belousova, N., Curiel, D.T., 1998. An adenovirus vector
with genetically modified fibers demonstrates expanded tropism via
utilization of a coxsackievirus and adenovirus receptor-independent
cell entry mechanism. J. Virol. 72, 9706 –9713.
Douglas, J.T., Rogers, B.E., Rosenfeld, M.E., Michael, S.I., Feng, M.,
Curiel, D.T., 1996. Targeted gene delivery by tropism-modified adeno-
viral vectors. Nat. Biotechnol. 14, 1574 –1578.
Fechner, H., Wang, X., Wang, H., Jansen, A., Pauschinger, M., Scherubl,
H., Bergelson, J.M., Schultheiss, H.P., Poller, W., 2000. Trans-comple-
mentation of vector replication versus Coxsackie-adenovirus-receptor
overexpression to improve transgene expression in poorly permissive
cancer cells. Gene Ther. 7, 1954 –1968.
Freimuth, P., Springer, K., Berard, C., Hainfeld, J., Bewley, M., Flanagan,
J., 1999. Coxsackievirus and adenovirus receptor amino-terminal im-
munoglobulin V-related domain binds adenovirus type 2 and fiber knob
from adenovirus type 12. J. Virol. 73, 1392–1398.
Gall, J., Kass-Eisler, A., Leinwand, L., Falck-Pedersen, E., 1996. Adeno-
virus type 5 and 7 capsid chimera: fiber replacement alters receptor
tropism without affecting primary immune neutralization epitopes.
J. Virol. 70, 2116 –2123.
Gomez-Navarro, J., Curiel, D.T., Douglas, J.T., 1999. Gene therapy for
cancer. Eur. J. Cancer 35, 867– 885.
Havenga, M.J., Lemckert, A.A., Grimbergen, J.M., Vogels, R., Huisman,
L.G., Valerio, D., Bout, A., Quax, P.H., 2001. Improved adenovirus
vectors for infection of cardiovascular tissues. J. Virol. 75, 3335–3342.
Hemmi, S., Geertsen, R., Mezzacasa, A., Peter, I., Dummer, R., 1998. The
presence of human coxsackievirus and adenovirus receptor is associ-
ated with efficient adenovirus-mediated transgene expression in human
melanoma cell cultures. Hum. Gene Ther. 9, 2363–2373.
Henry, L.J., Xia, D., Wilke, M.E., Deisenhofer, J., Gerald, R.D., 1994.
Characterization of the knob domain of the adenovirus type 5 fiber
protein expressed in Escherichia coli. J. Virol. 68, 5239 –5246.
Kaner, R.J., Worgall, S., Leopold, P.L., Stolze, E., Milano, E., Hidaka, C.,
Ramalingam, R., Hackett, N.R., Singh, R., Bergelson, J., Finberg, R.,
Falck-Pedersen, E., Crystal, R.G., 1999. Modification of the genetic
program of human alveolar macrophages by adenovirus vectors in vitro
is feasible but inefficient, limited in part by the low level of expression
of the coxsackie/adenovirus receptor. Am. J. Respir. Cell Mol. Biol. 20,
361–370.
Kanerva, A., Mikheeva, G.V., Krasnykh, V., Coolidge, C.J., Lam, J.T.,
Mahasreshti, P.J., Baker, S.D., Straughn, M., Barnes, M.N., Alvarez,
R.D., Hemminki, A., Curiel, D.T., 2002. Targeting adenovirus to the
serotype 3 receptor increases gene transfer efficiency to ovarian cancer
cells. Clin. Cancer Res. 8, 275–280.
Karayan, L., Gay, B., Gerfaux, J., Boulanger, P.A., 1994. Oligomerization
of recombinant penton base of adenovirus type 2 and its assembly with
fiber in baculovirus-infected cells. Virology 202, 782–795.
Kelly, F.J., Miller, C.R., Buchsbaum, D.J., Gomez-Navarro, J., Barnes,
M.N., Alvarez, R.D., Curiel, D.T., 2000. Selectivity of TAG-72-tar-
geted adenovirus gene transfer to primary ovarian carcinoma cells
versus autologous mesothelial cells in vitro. Clin. Cancer Res. 6,
4323– 4333.
Krasnykh, V., Belousova, N., Korokhov, N., Mikheeva, G., Curiel, D.T.,
2001. Genetic targeting of an adenovirus vector via replacement of the
fiber protein with the phage T4 fibritin. J. Virol. 75, 4176 – 4183.
Krasnykh, V., Dmitriev, I., Navarro, J.G., Belousova, N., Kashentseva, E.,
Xiang, J., Douglas, J.T., Curiel, D.T., 2000. Advanced generation
adenoviral vectors possess augmented gene transfer efficiency based
upon coxsackie adenovirus receptor-independent cellular entry capac-
ity. Cancer Res. 60, 6784 – 6787.
Krasnykh, V.N., Mikheeva, G.V., Douglas, J.T., Curiel, D.T., 1996. Gen-
eration of recombinant adenovirus vectors with modified fibers for
altering viral tropism. J. Virol. 70, 6839 – 6846.
Li, Y., Pong, R.C., Bergelson, J.M., Hall, M.C., Sagalowsky, A.I., Tseng,
C.P., Wang, Z., Hsieh, J.T., 1999. Loss of adenoviral receptor expres-
sion in human bladder cancer cells: a potential impact on the efficacy
of gene therapy. Cancer Res. 59, 325–330.
Louis, N., Fender, P., Barge, A., Kitts, P., Chroboczek, J., 1994. Cell-
binding domain of adenovirus serotype 2 fiber. J. Virol. 68, 4104 –
4106.
Maizel, J.V.J., White, O., Scharff, M.D., 1968. The polypeptides of ade-
novirus. I. Evidence for multiple protein components in the virion and
a comparison of types 2, 7A, and 12. Virology 36, 115–125.
Miller, C.R., Buchsbaum, D.J., Reynolds, P.N., Douglas, J.T., Gillespie,
G.Y., Mayo, M.S., Raben, D., Curiel, D.T., 1998. Differential suscep-
tibility of primary and established human glioma cells to adenovirus
infection: targeting via the epidermal growth factor receptor achieves
fiber receptor-independent gene transfer. Cancer Res. 58, 5738 –5748.
Mittereder, N., March, K.L., Trapnell, B.C., 1996. Evaluation of the con-
centration and bioactivity of adenovirus vectors for gene therapy.
J. Virol. 70, 7498 –7509.
Miyazawa, N., Leopold, P.L., Hackett, N.R., Ferris, B., Worgall, S., Falck-
Pedersen, E., Crystal, R.G., 1999. Fiber swap between adenovirus
 K. Takayama et al. / Virology 309 (2003) 282–293
subgroups B and C alters intracellular trafficking of adenovirus gene
transfer vectors. J. Virol. 73, 6056 – 6065.
Nalbantoglu, J., Larochelle, N., Wolf, E., Karpati, G., Lochmuller, H.,
Holland, P.C., 2001. Muscle-specific overexpression of the adenovirus
primary receptor car overcomes low efficiency of gene transfer to
mature skeletal muscle. J. Virol. 75, 4276 – 4282.
Okegawa, T., Li, Y., Pong, R.C., Bergelson, J.M., Zhou, J., Hsieh, J.T.,
2000. The dual impact of coxsackie and adenovirus receptor expression
on human prostate cancer gene therapy. Cancer Res. 60, 5031–5036.
Pereboev, A., Pereboeva, L., Curiel, D.T., 2001. Phage display of adeno-
virus type 5 fiber knob as a tool for specific selection and validation.
J. Virol. 75, 7107–7113.
Pickles, R.J., McCarty, D., Matsui, H., Hart, P.J., Randell, S.H., Boucher,
R.C., 1998. Limited entry of adenovirus vectors into well-differentiated
airway epithelium is responsible for inefficient gene transfer. J. Virol.
72, 6014 – 6023.
Roelvink, P.W., Lizonova, A., Lee, J.G., Li, Y., Bergelson, J.M., Finberg,
R.W., Brough, D.E., Kovesdi, I., Wickham, T.J., 1998. The coxsack-
ievirus-adenovirus receptor protein can function as a cellular attach-
ment protein for adenovirus serotypes from subgroups A, C, D, E, and
F. J. Virol. 72, 7909 –7915.
Seki, T., Dmitriev, I., Kashentseva, E., Takayama, K., Rots, M.G., Suzuki,
K., Curiel, D.T., 2002. Artificial extension of the adenovirus fiber shaft
inhibits infectivity in coxsackievirus and adenovirus receptor-positive
cell lines. J. Virol. 76, 1100 –1108.
293
Stevenson, S.C., Rollence, M., Marshall-Neff, J., McClelland, A., 1997.
Selective targeting of human cells by a chimeric adenovirus vector
containing a modified fiber protein. J. Virol. 71, 4782– 4790.
Stevenson, S.C., Rollence, M., White, B., Weaver, L., McClelland, A.,
1995. Human adenovirus serotypes 3 and 5 bind to two different
cellular receptors via the fiber head domain. J. Virol. 69, 2850 –
2857.
Tomko, R.P., Xu, R., Philipson, L., 1997. HCAR and MCAR, the human
and mouse cellular receptors for subgroup C adenoviruses and group B
coxsackieviruses. Proc. Natl. Acad. Sci. USA 94, 3352–3356.
Van’t Hof, W., Crystal, R.G., 2001. Manipulation of the cytoplasmic and
transmembrane domains alters cell surface levels of the coxsackie-
adenovirus receptor and changes the efficiency of adenovirus infection.
Hum. Gene Ther. 12, 25–34.
Von Seggern, D.J., Huang, S., Fleck, S.K., Stevenson, S.C., Nemerow,
G.R., 2000. Adenovirus vector pseudotyping in fiber-expressing cell
lines, improved transduction of Epstein-Barr virus-transformed B cells.
J. Virol. 74, 354 –362.
Zabner, J., Chillon, M., Grunst, T., Moninger, T.O., Davidson, B.L.,
Gregory, R., Armentano, D., 1999. A chimeric type 2 adenovirus vector
with a type 17 fiber enhances gene transfer to human airway epithelia.
J. Virol. 73, 8689 – 8695.
Zabner, J., Freimuth, P., Puga, A., Fabrega, A., Welsh, M.J., 1997. Lack of
high affinity fiber receptor activity explains the resistance of ciliated
airway epithelia to adenovirus infection. J. Clin. Invest. 100, 1144 –
1149.