ARTICLE doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.

com on IDEAL

Human and Mouse IFN-
Gene Therapy Exhibits Different

Anti-tumor Mechanisms in Mouse Models
Xiao-Qiang Qin,* Carla Beckham, Jennifer L. Brown, Matvey Lukashev, and James Barsoum
Biogen, Inc., 14 Cambridge Center, Cambridge, Massachusetts 02142, USA

*To whom correspondence and reprint requests should be addressed. Fax: (617) 679-3200. E-mail: xiao-qiang_qin@biogen.com.

Previously, we suggested that local human interferon-
(IFN-
) gene therapy with replication-
defective adenoviral vectors can be an effective cancer treatment. Clinical trials to treat cancers
with adenovirus expressing the human IFN-
gene (IFNB1) has been planned. As a continued effort
to explore the mechanisms of action of human IFN-
gene therapy that can occur in the clinical
setting, we tested mouse IFN-
gene therapy in human xenograft tumors in both ex vivo and in
vivo models. Delivery of the mouse IFN-
gene (Ifnb) caused tumor inhibition; this effect was
dependent on the indirect anti-tumor activities of IFN-
, notably a stimulation of natural killer
cells. IFN-
does not show cross-species activity in its anti-proliferative effect and mouse IFN-

does not cause as significant an anti-proliferative effect on mouse tumor cells as human IFN-

causes on human tumor cells. Therefore, we believe that mouse models using either human IFN-

or mouse IFN-
gene transfer do not capture all aspects of the action of adenovirus-mediated
human IFN-
gene therapy that may be present in the clinical setting. Due to its multiple mech-
anisms of action, human IFN-
gene therapy may be effective in treating human cancers that are
either sensitive or resistant to the direct anti-proliferative effect of IFN-
.

Key Words: IFN-
gene delivery, tumor regression, mouse tumor model,
anti-proliferative effect, immune stimulation, NK cells

INTRODUCTION
The type 1 interferons (IFNs), the -family and
, share
receptor components and have multiple anti-tumor
activities [1–3]. At appropriate concentrations, they
induce the direct anti-proliferative effects that are man-
ifested as cytotoxicity/apoptosis and inhibition of cell
cycle progression in S phase for solid tumor cells or in
G1 phase in Daudi Burkitt’s lymphoma cells. It is not
clear whether IFNs induce cytotoxicity/apoptosis and
cell cycle inhibition by activation of the same signal
transduction pathways. However, it seems that these two
effects are biologically separable because we found that
IFN-
induced the cytotoxicity effect in cells that were
insensitive to its cell cycle effect (X.-Q.Q., unpublished
data). IFN- and IFN-
also were observed to have dif-
ferences in their receptor binding and signaling [4–7]. In
certain types of human carcinoma cells, IFN-
showed
a more marked anti-proliferative effect than IFN-
[8–10]. Clinically, there are data that suggest a correla-
tion between the in vitro sensitivity of malignant cells to
the direct anti-proliferative effect induced by IFNs and
the in vivo clinical effects in patients receiving IFN pro-
tein therapy [11].
IFNs can also exhibit an indirect anti-tumor effect
through immune stimulation. They induce major histo-
compatibility complex class 1 expression, cause an increase
in cytotoxic T lymphocyte activity, enhance the genera-
tion of T helper cells, and activate macrophages and nat-
ural killer (NK) cells both in vitro and in animal models
[12,13]. Stable gene transfer of IFN- in mouse tumor cells
blocked the tumorigenic potential and caused the infil-
tration of immunological effector cells such as T lympho-
cytes [14–18]. Another indirect anti-tumor activity of IFNs
is the inhibition of angiogenesis. IFN- blocks endothelial
cell migration in vitro and inhibits lymphocyte- and tumor-
induced angiogenesis. IFN-
downregulates expression of
the gene encoding angiogenic factor basic fibroblast
growth factor and inhibits angiogenesis in vivo [19–24].
Clinical experience with IFN protein therapy of cancer,
especially in the treatment of solid tumors, has generally
been disappointing. Convincing effect has been observed
only in limited cancer types such as hairy cell leukemia,
chronic myelogenous leukemia, and melanoma [11,25].
Pharmacokinetic studies have indicated that IFN-
has an
extremely short half-life in the blood system after systemic
protein administration [26,27]. This suggests that par-

356 MOLECULAR THERAPY Vol. 4, No. 4, October 2001

Copyright © The American Society of Gene Therapy
1525-0016/01 $35.00
 doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
ARTICLE

A B

FIG. 1. Human IFN-
gene therapy led to regression of human s.c. xenograft U87 MG tumors in mice and prolonged mouse survival. (A) The mean diameters
of tumors after intratumoral injection with H5.010CMVhIFN-
at 1  1011 particles (open triangle), 1  1010 particles (open circle), and 1  109 particles (filled
diamond), or with PBS (filled circle), or with H5.010CMVlacZ at 1  1011 particles (open, inverted triangle), 1  1010 particles (multiplication sign), or 1  109
particles (filled triangle). *P < 0.001 for tumors treated with H5.010CMVhIFN-
at 1  1011 particles or H5.010CMVhIFN-
at 1  1010 particles versus treated
with PBS; **P < 0.05 for tumors treated with H5.010CMVhIFN-
at 1  109 particles versus treated with PBS. (B) Kaplan–Meier plot showing the percentage
survival of mice after the various treatments, indicated by the same symbols as in (A).

enteral IFN-
protein treatment, through intravenous
(i.v.), subcutaneous (s.c.), or intramuscular (i.m.) injec-
tions, may fail to have an anti-tumor effect due to insuf-
ficient delivery or a lack of sustained delivery of the pro-
tein to the tumor site.
We found previously that human IFN-
gene therapy
inhibits tumor formation and causes regression of estab-
lished tumors in immune-deficient mice, and that the
human IFN-
-mediated anti-proliferative activity alone
can be sufficient to cause tumor regression [28]. We pro-
posed that local IFN-
gene therapy with replication-
defective adenovirus vectors can be an effective treatment
for cancer. Others have demonstrated that IFN-
gene
therapy can also abrogate tumorigenicity and eradicate
established tumors by the collective activation of IFN-
-
induced indirect anti-tumor activities [24,29–31]. Here,
we further evaluated the effect of both ex vivo and direct
local mouse IFN-
gene therapy on various human
xenograft tumors and found that human and mouse IFN-

gene therapy demonstrated two different mechanisms
of the anti-tumor activity of IFN-
gene therapy, both of
which might be efficacious in the clinical setting.
RESULTS
Mouse IFN-
Gene Therapy Leads to Tumor
Inhibition in ex Vivo and in Vivo Human
Xenograft Tumor Models
Previously, we demonstrated that local human IFN
gene
transfer can be an effective treatment for solid tumors in
mouse models [28]. We are now planning clinical trials to
test the safety and efficacy of adenovirus-mediated human
IFN-
gene therapy in human cancers. Malignant glioblas-
toma will be the disease target of our first clinical trial
because it is a local cancer amenable to direct intratumoral
therapy and the disease lacks effective therapeutic alter-
natives. We therefore evaluated gene therapy with aden-
ovirus expressing human IFN-
in a human xenograft
model. We treated preestablished s.c. U87 MG human
glioma tumors in nude mice with a single, intratumoral
injection of phosphate-buffered saline (PBS) or adenovirus
expressing either human IFN-
or the negative control
gene, lacZ. Human IFN-
gene therapy led to tumor regres-
sion and significantly prolonged animal survival (Figs. 1A
and 1B). We observed complete tumor regression and
100% long-term mouse survival at viral doses of 1  1011
and 1  1010 total particles per mouse.
We next tested the effect of mouse IFN-
gene therapy
in human xenograft tumors. First, we used an ex vivo gene
delivery model to compare mouse IFN-
with human IFN-

gene therapy. Previously, a very potent bystander effect
was demonstrated, as adenovirus-mediated human IFN-

gene transduction in as few as 1% of human breast carci-
noma MDA-MB-468 cells was sufficient to block tumor
formation [28]. Therefore, we infected these cells in vitro
with an adenovirus that has deletions of the E1 and E3
regions and a temperature-sensitive mutation of E2A,
encoding mouse Ifnb (H5.110CMVmIFN-
) or human
IFNB1 (H5.110CMVhIFN-
) at a multiplicity such that
nearly all cells should be transduced. We collected the
infected cells and mixed them with uninfected cells imme-
diately before implantation to assess the bystander effect.
Equal numbers of infected cells, mixtures in which 10%
or 1% of cells were infected with H5.110CMVmIFN-
, or

MOLECULAR THERAPY Vol. 4, No. 4, October 2001 357

Copyright © The American Society of Gene Therapy
 ARTICLE doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL

FIG. 2. Ex vivo mouse IFN-
gene therapy in

human IFN-
gene delivery.
MDA-MB-468 tumors. Uninfected and
infected tumor cells were mixed at various Although human IFN-
gene deliv-
ratios immediately before implantation into ery blocked tumor formation, 1%
mice. Uninfected cells (open square), or 100% mouse IFN-
gene transduction
H5.110CMVmIFN-
transduced cells (open resulted in the formation of tumors
triangle), 10% (open, inverted triangle) or 1%
(filled, inverted triangle) H5.110CMVmIFN-

which subsequently regressed.
transduced cells, or 1% H5.110CMVhIFN-
We next examined whether direct
transduced cells (open circle) were injected treatment of tumors with
s.c. into the flanks of nude mice. Mean tumor H5.110CMVmIFN-
leads to their
diameter was plotted versus time post-tumor
regression. We injected pre-
cell implantation. *P < 0.01 for mice
implanted with 10% or 1% established s.c. MDA-MB-468 tumors
H5.110CMVmIFN-
transduced cells, or 1% in nude mice with H5.110CMVmIFN-
H5.110CMVhIFN-
transduced cells, versus
, H5.110CMVhIFN-
, or
with uninfected cells at day 28. H5.110CMVlacZ, all at the dose of 5
 1010 viral particles/mouse, or PBS.
As in our previous study,
H5.110CMVhIFN-
caused rapid
tumor regression, resulting in com-
a mixture in which 1% of cells were infected with plete tumor regression in four of five mice within 1 week
H5.110CMVhIFN-
were s.c. implanted into Balb/c nude after the treatment. In this specific experiment, one of five
mice. Consistent with previous observations, human IFN- mice failed to respond to H5.110CMVhIFN-
treatment, pos-

gene transduction of 1% of cells effectively blocked sibly due to a less efficient virus injection. It should be noted
tumor formation (Fig. 2). In contrast, we observed the for- that in several other experiments performed with
mation of small tumors in all mice in the 10% and 1% H5.110CMVhIFN-
, we often saw complete tumor regression
H5.110CMVmIFN-
treatment groups at end of the first in all mice treated, resulting in 100% animal survival.
week after cell implantation. However, all tumors disap- H5.110CMVmIFN-
treatment also caused complete tumor
peared within 30 days (Fig. 2). We also observed that 100% regression (Fig. 3A). However, in contrast to the very rapid
cell infection with H5.110CMVmIFN-
effectively blocked regression observed following gene delivery of human IFN-
tumor formation (Fig. 2). This result may not necessarily
, the regression was not significant until 2 weeks after the
be due to a mouse IFN-
effect because cells that were treatment (Fig. 3A). Once tumors regressed, they did not
100% infected with the control virus (H5.110CMVlacZ) recur over the observation time of 2 months. We further
did not form tumors [28]. Therefore, ex vivo gene delivery tested mouse IFN-
gene therapy in established human
of mouse IFN
into as few as 1% of human tumor cells xenograft tumors derived from prostate cancer PC-3 cells.
can inhibit tumorigenesis, albeit less efficiently than Direct intratumoral treatment with H5.110CMVmIFN-
also

A B

FIG. 3. Direct in vivo treatment of established tumors derived from human MDA-MB-468 cells (A) or PC-3 cells (B). Preestablished tumors were treated with PBS
(open square), H5.110CMVmIFN-
(open circle), H5.110CMVhIFN-
(open triangle), or H5.110CMVlacZ (open, inverted triangle) at 5  1010 viral parti-
cles/mouse. Measurements represent mean tumor diameter. P < 0.01 for tumor diameter difference between mouse IFN-
or human IFN-
gene therapy ver-
sus PBS treatment in both experiments.

358 MOLECULAR THERAPY Vol. 4, No. 4, October 2001

Copyright © The American Society of Gene Therapy
 doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
ARTICLE

A B

FIG. 4. Local mouse IFN-
and human IFN-
gene therapy led to suppression of both local i.p. tumors and metastasis induced by the i.p. injection of MDA-MB-231 cells
in nude mice. (A) A representative mouse that developed large i.p. tumors in the H5.010CMVlacZ treatment group (right) and a mouse that was cured with IFN-
gene
therapy. (B) Kaplan–Meier plot showing the percentage survival of mice that were treated with PBS (filled, inverted triangle), H5.010CMVmIFN-
(filled circle),
H5.010CMVhIFN-
(filled square), or H5.010CMVlacZ (filled triangle) at 2  1010 particles/mouse (2 weeks after i.p. tumor cell injections) over the observation period
of 90 days. P < 0.005, H5.010CMVmIFN-
or H5.010CMVhIFN-
treatment versus PBS treatment. There is no statistical difference between H5.010CMVlacZ treatment
and PBS treatment.

caused tumor regression (Fig. 3B). In both tumor types,
tumor regression induced by H5.110CMVmIFN-
was slower
than H5.110CMVhIFN-
mediated regression, even though
delivery of both Ifnb and IFNB1 led to significant long-term
animal survival (data not shown).
To confirm these observations in a different model, we
implanted human breast carcinoma MDA-MB-231 cells into
nude mice by intraperitoneal injection. Injection of these
cells led to development of both local (i.p.) tumors and the
subsequent metastasis to liver and lung tissues after 2 to 3
weeks. Mice that received i.p. administration of PBS or the
control LacZ adenovirus developed large i.p. tumors (Fig.
4A). Within 3 months post treatment, virtually all the mice
were euthanized due to local i.p. tumor burden including
ascites and symptoms of metastasis such as jaundice, loss
of activity, and/or difficulty in breathing (Fig. 4B). However,
i.p. treatment with an adenovirus deleted for the E1 and E3
regions, encoding mouse Ifnb or human IFNB1
(H5.010CMVmIFN-
and H5.010CMVhIFN-
, respectively),
significantly prolonged survival (Fig. 4B). In this experi-
ment, human IFN-
gene delivery led to a greater animal
survival rate than mouse IFN-
gene delivery (Fig. 4B).
When the human IFN-
and mouse IFN-
protein levels in
serum were examined by ELISA 3 days after treatment with
H5.010CMVmIFN-
and H5.010CMVhIFN-
, no detectable
mouse or human IFN-
was found in these mice. This fur-
ther supports our previous observation that local gene deliv-
ery leads to a high level of IFN-
locally, but with only
barely detectable or undetectable levels in circulation.
The Mechanism of the Mouse IFN-
Mediated
Anti-tumor Effect
The difference in anti-tumor efficiency led us to speculate
that after gene delivery human IFN-
and mouse IFN-
act
through different mechanisms. In these tumor models,
human IFN-
gene therapy mainly executed a direct anti-
proliferative effect [28]. To investigate whether mouse IFN-

can cause an anti-proliferative effect in human tumor
cells, we treated MDA-MB-468 cells (Fig. 5) and human
hepatoma cells (Huh7; data not shown) with equal amounts
of mouse IFN-
and human IFN-
protein, and examined
cell viability by both light microscopy and an MTT assay.
No anti-proliferative activity was observed with mouse IFN-

, even at 30,000 units/ml, whereas human IFN-
induced
a significant decrease in cell viability at 100 units/ml (Fig.
5 and data not shown). Because mouse IFN-
does not cause
any anti-proliferative effect on human cells, it is unlikely
that the tumor regression caused by mouse IFN-
gene deliv-
ery was due to an IFN-mediated direct anti-proliferative
effect.
The modes of IFN-
anti-tumor action manifest three
aspects: anti-proliferative effect, immune stimulation, and
inhibition of angiogenesis. We investigated whether the
mouse IFN-
stimulated immune function was critical to the

MOLECULAR THERAPY Vol. 4, No. 4, October 2001 359

Copyright © The American Society of Gene Therapy
 ARTICLE doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
anti-tumor effect in these models. Mice bearing s.c. MDA-
MB-468 tumors were pretreated with PBS or anti-asialo GM1
antiserum, which depletes NK cells. We then treated these
mice with intratumoral injection of PBS, H5.110CMVmIFN-

, H5.110CMVhIFN-
, or H5.110CMVlacZ at 5  1010 par-
ticles/mouse. Treatment with PBS or H5.110CMVlacZ did
not cause tumor regression in either the PBS or anti-asialo
GM1 antisera pretreatment groups. Treatment with
H5.110CMVmIFN-
caused tumor regression and signifi-
cantly prolonged animal survival in the PBS pretreatment
group. However, anti-asialo GM1 antisera pretreatment sig-
nificantly reduced the H5.110CMVmIFN-
anti-tumor effect
in terms of both tumor size (Fig. 6A) and mouse survival
(Fig. 6B). In contrast, human IFN-
gene delivery led to
complete tumor regression in both the presence and the
absence of the anti-asialo GM1 antisera pretreatment (Figs.
6A and 6B). These data indicate that NK cell activity had a
role in the H5.110CMVmIFN-
-mediated tumor regression.
It should be noted that the anti-asialo GM1 antisera pre-
treatment did not completely abolish the anti-tumor effect
of H5.110CMVmIFN-
. The lack of complete abolition of
the anti-tumor effect might indicate that NK cell depletion
was not complete or that the mouse IFN-
induced anti-
tumor effect did not rely solely on the stimulation of NK
cells. Similar results were observed in tumor modeling per-
formed in SCID/beige mice, which lack NK cells. Human
IFN-
gene therapy was still effective in SCID/beige mice;
however, the anti-tumor activity of mouse IFN-
gene ther-
apy was reduced, although not completely eliminated, in
these mice (data not shown). Although the differential anti-
tumor effect of H5.110CMVmIFN-
in SCID/beige and nude
mice may result from other genetic differences of these two
mouse species, it is also consistent with a model in which
the anti-tumor effect of mouse IFN-
gene therapy is due to
an indirect activity mediated, at least in part, by NK cells.
To confirm that mouse IFN-
gene delivery enhanced
the NK cell activity, we collected splenocytes from mice
360
FIG. 5. Mouse IFN-
failed to cause a significant anti-
proliferative effect on MDA-MB-468 cells at doses rang-
ing from 100 unit/ml to 30,000 units/ml (gray bars),
whereas human IFN-
displayed potent anti-proliferative
activity (black bars), as determined in an MTT assay.
that received intratumoral
H5.110CMVmIFN-
treatment and exam-
ined NK cell function. Increased NK
cytolysis was observed in these mice com-
pared with control mice (Fig. 7). In the
same experiment, anti-asialo GM1 antis-
era pretreatment blocked NK cell stimu-
lation (Fig. 7). In addition, we also tested
the splenocytes of some mice in the
MDA-MB-468 ex vivo mouse IFN-
gene
delivery experiment. Enhanced NK cell
activity was observed following implan-
tation of H5.110CMVmIFN-
transduced
cells (data not shown).
We next tested whether mouse IFN-
gene delivery
might have an effect on an untreated tumor at a distinct
site. We implanted MDA-MB-468 cells on both flanks of
nude mice. Mice bearing tumors of roughly equivalent size
on each side were selected for the experiment. We then
treated tumors on the right flank with H5.110CMVmIFN-

or H5.110CMVlacZ at 1  1011 particles/mouse or PBS
and monitored tumor growth on both flanks. We found
that mouse IFN-
gene therapy caused the regression of
both the treated tumor and the untreated tumor on the
opposite flank (Fig. 8).
Type 1 IFNs including IFN-
have been shown to be
inhibitors of angiogenesis and can inhibit tumor growth
by blocking tumor vascularization [19–24]. We investi-
gated whether mouse IFN-
gene transfer caused an
observable effect on the blood vasculature in tumors. At 2
weeks after treatment with H5.110CMVmIFN-
or
H5.110CMVlacZ, we harvested tumors to examine the
microvascular density and morphology. No significant dif-
ference was noted in the vasculature following treatment
with these two viruses (data not shown). Recently, mouse
IFN-
gene therapy was also examined in a mouse col-
orectal cancer liver metastases model. No significant reduc-
tion in tumor blood vessels was observed after mouse Ifnb
gene delivery [32].
DISCUSSION
We previously demonstrated that local human IFN-
gene
therapy with adenovirus vectors led to tumor regression [28].
Our study supports the testing of local human IFN-
gene
therapy in cancer patients. Clinical trials testing adenovirus
expressing human IFN-
in cancers will be initiated soon.
Here, we have examined ex vivo and direct in vivo mouse
IFN-
gene therapy in human xenograft tumor models. The
growth of these human tumor cells in vitro is completely
MOLECULAR THERAPY Vol. 4, No. 4, October 2001
Copyright © The American Society of Gene Therapy
 doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
ARTICLE

FIG. 6. NK cell depletion caused an inhibitory effect on the anti-tumor activ-

ity of mouse IFN-
, but not human IFN-
gene therapy. (A) The mean diam-
A
eters of MDA-MB-468 tumors in mice that had tail vein PBS pretreatment and
intratumoral treatment with PBS (multiplication sign), H5.110CMVmIFN-

(open square), H5.110CMVhIFN-
(open triangle), H5.110CMV-lacZ (open cir-
cle), or in the mice that had anti-asialo GM1 antibody pretreatment and then
intra-tumor treatment with PBS (filled triangle), H5.110CMVmIFN-
(filled
square), H5.110CMVhIFN-
(filled, inverted triangle), or H5.110CMV-lacZ
(open circle). *The P value for the diameter difference of tumors treated with
H5.110CMVmIFN-
in the presence or absence of anti-asialo GM1 antibody
is < 0.05 as calculated with the t test. (B) Kaplan-Meier plot showing the per-
centage survival of mice over the observation period of 95 days. Treatments
are indicated with the same symbols as in (A).

insensitive to mouse IFN-
. Both ex vivo and in vivo mouse

IFN-
gene delivery caused significant anti-tumor effects.
These effects were dependent on indirect anti-tumor activi-
ties, notably but not exclusively a stimulation of NK cells.
This is consistent with a previous study in which NK cell acti-
vation was implicated in the rejection of PC-3 cells stably
expressing mouse IFN-
in mice [24]. However, our data fur-
ther indicate that direct local IFN-
gene therapy can also B
lead to an activation of the innate immune system, result-
ing in tumor regression. This underscores the notion that
direct local IFN-
gene therapy with adenovirus vectors
might be an effective treatment for human cancers that are
either sensitive or resistant to the IFN-
induced direct anti-
proliferative activity.
Available animal tumor models for exploring IFN-

gene therapy generally use human tumor cells in immune-
deficient mice or mouse tumor cells in immune-competent
mice. We believe that animal modeling with immune-
competent mice may not be representative of human IFN-

gene therapy in the clinical setting. Because only tumors
of mouse origin grow in the immune-competent mice and
IFN-
does not show cross-species activity, only mouse
IFN-
gene therapy can be tested in these mice. Human
IFN-
and mouse IFN-
seem to have biological differences
in terms of anti-proliferative activity. We noticed that
mouse IFN-
does not cause as significant an anti-prolif-
erative effect on the mouse tumor cells as human IFN-
vated systemically through immune organs such as spleen.
does on human tumor cells, as measured by MTT assay and The NK activation in spleen can result from vector dis-
cell cycle analysis in multiple mouse cell lines (X.-Q.Q., semination into the spleen or the circulation of a low level
C.B., and Amie Dergay, unpublished data). Clearly, animal of IFN-
that is released by the vector-transduced tumor
modeling with immune-competent mice can reveal the cells. Therefore, the mechanisms of action of IFN-
gene
indirect anti-tumor mechanisms of IFN-
gene therapy therapy in a clinical setting will be very likely to include
such as T-cell-mediated immune responses. On the other both IFN-
-induced direct and indirect effects. We think
hand, our previous modeling with human xenograft that human IFN-
gene therapy could generate a combi-
tumors in immune-deficient mice directly tested human natorial effect resulting from both the direct and the indi-
IFN-
gene therapy in tumors of human origin. It did not rect anti-tumor activities, which could be highly effica-
focus on its indirect anti-tumor mechanisms of action. cious and have a long-lasting anti-tumor effect.
Other results [24,29–31] have indicated that IFN-
gene Although we did not find differences in the microvas-
therapy can also cause an effective anti-tumor effect by cular density or morphology between H5.110CMVmIFN-
means of stimulation of both innate and adoptive immu-
treated tumors and control tumors, we cannot rule out
nity. Our current study further confirms that local IFN-
the possibility that inhibition of angiogenesis is involved
gene delivery can lead to the activation of NK cells. The in the anti-tumor effect. It is possible that mouse IFN-
did
NK cells can be activated locally in the tumors and/or acti- not cause a direct disruption of blood microvasculatures

MOLECULAR THERAPY Vol. 4, No. 4, October 2001 361

Copyright © The American Society of Gene Therapy
 ARTICLE doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
FIG. 7. NK cell cytotoxicity assay showing that mouse IFN-
gene delivery led
to the activation of NK cells and that the activation can be blocked by the anti-
asialo GM1 antibody treatment. NK cell cytotoxicity was measured as the spe-
cific release of the carboxyfluorescein molecule BCECF as described [35]. Each
treatment group contained three mice. The effector cells were acquired from
the mixed spleen tissues of the three mice. Effector to target ratios used were
100:1, 50:1, 25:1, and 12.5:1 in triplicate. The data shown were derived from
the 100:1 samples. The lower effector to target ratios were not observed to
have cytotoxic activity. The error bars are SE, which were calculated from the
triplicate samples.
but caused a blockage of tumor vascularization. This block-
age effect could then cause an inhibition of tumor growth
and, therefore, an anti-angiogenesis effect could occur
despite the lack of observable differences in the microvas-
cular density between tumors treated with the mouse IFN-

virus and the controls. Likewise, we could not conclude
that other IFN-
effects such as activation of macrophages
were not involved in the tumor regression observed in our
study. The fact that NK cell deletion did not completely
abolish the tumor regression effect mediated by mouse
IFN-
gene therapy may suggest the involvement of other
effector cells such as macrophages. However, we can con-
clude that NK cell activation had a major role in our mod-
els as mouse IFN-
gene therapy activated NK cells and NK
cell depletion substantially reduced the anti-tumor effect.
We also conclude that another major indirect IFN-
anti-
tumor activity, the T-cell-mediated immune response, was
not necessary for an overt anti-tumor effect because we
used immune-deficient mice that lack T cells.
We have tested the effect of mouse IFN-
gene therapy
in s.c. human xenograft tumor models and demonstrated
tumor inhibition that was independent of the IFN-induced
direct anti-proliferative effect as well as indirect activities
involving T cells. By depleting NK cells, we found that
IFN-
gene therapy causes tumor regression at least in part
362
FIG. 8. Mouse IFN-
gene delivery into tumors on one flank caused regres-
sion of tumors in both flanks. Mice bearing established MDA-MB-468 tumors
on both flanks were treated with an intratumoral injection of PBS,
H5.110CMVmIFN-
, or H5.110CMVlacZ (both at 1  1011 particles/mouse)
in the right flank tumor. The sizes of tumors in both flanks were measured.
The mean tumor diameters of the left and right flanks in the PBS group (open
and filled squares, respectively), in the H5.110CMVmIFN-
group (open and
filled circles, respectively), and in the H5.110CMV-lacZ group (open and filled
triangles, respectively) were plotted versus time post virus treatment. There is
a significant difference between tumors in each flank of the mice that were
treated with H5.110CMVmIFN-
versus PBS-treated tumors (P < 0.05).
through the activation of NK cells. We further argued that
mouse models with either human IFN-
or mouse IFN-

gene delivery could not capture all aspects of the action
of IFN-
gene therapy in the clinical situation. One con-
cern in the clinical usefulness of IFN-
gene therapy is
that cancer cells resistant to the direct anti-proliferative
activity of IFN-
may arise. This work suggests that IFN-
-
resistant cancer cells may be eradicated by the indirect
activities of IFN-
on the host immune system, even in the
absence of direct anti-proliferative activity.
MATERIALS AND METHODS
Recombinant adenoviruses. The production and purification of the sec-
ond-generation adenoviruses H5.110CMVhIFN-
and H5.110CMVlacZ
were described previously [28]. These viruses have deletions of the E1 and
E3 genes and a temperature-sensitive mutation of E2a. The second-gener-
ation adenovirus encoding mouse Ifnb (H5.110CMVmIFN-
[32]) and the
first-generation adenovirus (having deletions only in the E1 and E3 regions)
encoding lacZ, human IFNB, and mouse Ifnb (H5.010CMVlacZ [33],
H5.010CMVhIFN-
, and H5.010CMVmIFN-
, respectively) were used.
H5.110CMVhIFN-
and H5.110CMVlacZ were previously described as
H5.110hIFN-
and H5.110lacZ, respectively [28].
Cell culture. Human breast carcinoma cells MDA-MB-468 and MDA-MB-
231, prostate carcinoma cells PC-3, and glioblastoma cells U87 MG were
obtained from the American Type Culture Collection (Rockville, MD).
Huh7, a line of human hepatoma cells, has been described [34]. MDA-MB-
MOLECULAR THERAPY Vol. 4, No. 4, October 2001
Copyright © The American Society of Gene Therapy
 doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
468, PC-3, and Huh7 cells were maintained as adherent cultures in MEM
containing 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml peni-
cillin, and 100 g/ml streptomycin. U87 MG cells were maintained as
adherent culture in Eagle’s minimal essential medium containing 2 mM L-
glutamine and Earle’s BSS adjusted to contain 1.5 g/l sodium bicarbonate,
0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, and 10% fetal
bovine serum. Mouse Lymphoma Yac-1 cells were purchased from
American Type Culture Collection (Rockville, MD). These cells were main-
tained as nonadherent cultures at 37C and 5% CO2 in RPMI 1640 sup-
plemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml
penicillin, and 100 g/ml streptomycin.
MTT assay. To assess the effect of IFN-
on cell survival, we carried out
an MTT assay using the MTT-based In Vitro Toxicology Assay kit (Sigma).
Human or mouse IFN-
was diluted in DMEM medium without phenol red
containing 10% fetal bovine serum, 2 mM glutamine, 100 units/ml peni-
cillin, and 100 g/ml streptomycin, and was aliquoted in a 96-well plate.
The human breast carcinoma cells MDA-MB-468 were then added to the
plate at 1000 cells/well. The final concentrations of human or mouse IFN-

was 0, 100, 300, 1000, 3000, 10,000, and 30,000 units/ml. The cells were
incubated with IFN-
for 5 days at 37C, 5% CO2, after which the MTT
reagent was added per the manufacturer’s protocol. The MTT formazan
crystals were allowed to completely dissolve. The O.D. of each well was then
read at 570 nm. Eight data points for each condition were used for acquir-
ing the average and the error bar.
Quantitation of IFN-
proteins by ELISA. Sera were collected through
retro-orbital bleeding 3 days after i.p. injection of H5.010CMVmIFN-
and
H5.010CMVhIFN-
in mice. ELISA for human IFN-
protein was carried out
as described [28]. Quantitation of mouse IFN-
protein by ELISA was per-
formed essentially the same as that for the human IFN-
ELISA, with the
following exceptions. We coated 96-well plates overnight at 4C with 10
g/ml anti-mouse IFN-
antibody (MA-MB-7, a monoclonal antibody pre-
pared in Biogen) in coating buffer containing 50 mM sodium bicarbon-
ate/carbonate and 0.2 mM MgCl2 (pH 9.6). The plates were blocked with
0.5% nonfat dry milk in PBS. The standards for mouse IFN-
were pur-
chased from Access Biomedical (San Diego, CA). Serum samples were
diluted in 0.5% nonfat dry milk with 0.05% Tween-20 in PBS. All the suc-
cessive steps are as in the human IFN-
ELISA.
NK cell assay. Mice were treated with anti-asialo GM1 antibody starting
approximately 1.5 weeks after s.c. implantation of MDA-MB-468 cells. We
administered 50 l anti-asialo GM1 antibody (30 mg/ml) to mice via tail
vein injections as suggested by the manufacturer’s instructions (Wako
BioProducts, Richmond, VA). The animals were injected with the anti-
asialo GM1 antibody every 4 to 5 days for four times. Mice were euthanized
for measuring NK activity 7 days after the intratumoral injections of PBS
or viruses. Each treatment group included three mice. NK activity was
measured based on a described assay [35] with slight modifications. Briefly,
the effector cells were collected by breaking apart mouse spleens with
tweezers in RPMI-1640 medium and passing them through 20-G needles
several times. The cells were then washed with PBS and resuspended in
RPMI-1640 media at 2.0  107 cells/ml. Yac-1 cells were used as targets and
their medium was changed 18 h before NK assay to assure that the cells
were in log growth. The labeling of the target cells was performed as
described [35]. After incubating the effector and target cells for 2.5 h at 37C
and 5% CO2, 100 l of each assay well were transferred to a multiscreen
96-well filtration plate (Millipore, Bedford, MA). Cells were further sepa-
rated from medium by vacuum filtration. Fluorescence was measured at
485/530 excitation/emission on the Cytofluor 4000 (PerSeptive Biosystems,
Framingham, MA).
In vitro viral transduction. The viral transduction was carried out as
described [28]. After the transduction, cells were collected with
trypsin/EDTA solution and washed twice with PBS.
Animal models. Female BALB/c nu/nu from Charles River Laboratory were
used in both the ex vivo and in vivo s.c. tumor model experiments. All mice
were maintained in the pathogen-free Biogen animal facility. A total of
approximately 2  106 cells in 100 l of PBS were implanted s.c. into the
right flank or both right and left flanks of mice to generate tumors for both
ex vivo and in vivo experiments. For the direct intratumoral treatment, 100
MOLECULAR THERAPY Vol. 4, No. 4, October 2001
Copyright © The American Society of Gene Therapy
ARTICLE
l of PBS containing various doses of adenoviral vectors was injected
directly into the center of the 4- to 7-mm diameter tumors in a single injec-
tion. The tumor size was measured in length and width using calipers and
presented as the average tumor diameter (mm). Mean tumor diameter is
the average of the tumor width plus length, averaged for all mice in a
group. There were usually five mice per group. For NK cell depletion exper-
iments, anti-asialo GM1 antibody administration was started approximately
1 week after tumor cell implantation. We administered 50 l anti-asialo
GM1 antibody (30 mg/ml) to mice via tail vein injections as suggested by
the manufacturer’s instructions (Wako BioProducts, Richmond, VA). The
animals were treated every 4 to 5 days for four times before the tumors were
treated with various viruses or PBS controls. For generating i.p. tumors and
metastasis, 100 l of PBS containing 5  106 human breast carcinoma cells
MDA-MB-231 was i.p. injected into nude mice. Subsequently, the mice
were subjected to i.p. treatments with PBS, H5.010CMVmIFN-
,
H5.010CMVhIFN-
, or H5.010CMVlacZ at 2 weeks after the tumor cell
injections. We treated 10 mice in each group. Animal death was defined
by euthanizing mice due to local tumor burden including severe ascites
and/or symptoms of metastasis such as loss of activity and/or difficulty in
breathing. For testing the effect of mouse IFN-
gene delivery into a tumor
on one side on the growth of tumors at both sides, a large number of mice
was implanted with MDA-MB-468 cells on both flanks and only mice bear-
ing a tumor on each side were selected for the experiment.
ACKNOWLEDGMENTS
We thank James Wilson and Guang-Ping Gao (Institute for Human Gene Therapy,
University of Pennsylvania) for providing adenovirus vectors, and Paula Hochman
and Gerry Majeau (both from Biogen) for providing mouse IFN-
protein.
RECEIVED FOR PUBLICATION MAY 8; ACCEPTED AUGUST 7, 2001.
REFERENCES
1. Lengyel, P. (1982). Biochemistry of interferons and their actions. Annu. Rev. Biochem. 51:
251–282.
2. Gresser, I. (1989). Antitumor effects of interferon. Acta Oncol. 28: 347–353.
3. Belardeli, F., and Gresser, I. (1996). The neglected role of type 1 interferon in the T-cell
response: implications for its clinical use. Immunol. Today 17: 369–372.
4. Gordon, I., Stevenson, D., Tumas, V., and Natham, C. (1983). Mouse interferon recep-
tors: difference in their response to A and
interferons. J. Gen. Virol. 64: 2777–2780.
5. Abramovich, C., et al. (1994). Differential tyrosine phosphorylation of the IFNAR chain
of the type I interferon receptor and of an associated surface protein in response to IFN-
 and IFN-
. EMBO J. 13: 5871–5877.
6. Abramovich, C., Chebath, J., and Revel, M. (1994). The human interferon -receptor
protein confers differential responses to human interferon-
versus interferon-
sub-
types in mouse and hamster cell transfectants. Cytokine 6: 414–424.
7. Erlandsson, L., et al. (1998). Interferon-
is required for interferon- production in mouse
fibroblasts. Curr. Biol. 8: 223–226.
8. Sica, G., Fabbroni, L., Castagnetta, L., Cacciatore, M., and Pavone-Macaluso, M. (1989).
Antiproliferative effect of interferons on human prostate carcinoma cell lines. Urol. Res.
17: 111–117.
9. Rosenblum, M. G., et al. (1990). Growth inhibitory effects of interferon-
but not inter-
feron- on human glioma cells: correlation of receptor binding, 2,5-oligoadenylate syn-
thetase and protein kinase activity. J. Interferon Res. 10: 141–151.
10. Johns, T. G., et al. (1992). Antiproliferative potencies of interferons on melanoma cell
lines and xenografts: higher efficiency of interferon-
. J. Natl. Cancer Inst. 84: 1185–1190.
11. Einhorn, S., and Grandér, D. (1996). Why do so many cancer patients fail to respond
to interferon therapy? J. Interferon Cytokine Res. 16: 275–281.
12. Tough, D. F., Borrow, P., and Sprent, J. (1996). Induction of bystander T cell prolifera-
ton by viruses and type 1 interferon in vivo. Science 272: 1947–1950.
13. Rogge, L., et al. (1997). Selective expression of an interleukin-12 receptor component
by human T helper cells. J. Exp. Med. 185: 825–831.
14. Ferrantini, M., et al. (1993). 1-Interferon gene transfer into metastatic friend leukemia
cells abrogated tumorigenicity in immunocompetent mice: antitumor therapy by means
of interferon-producing cells. Cancer Res. 53: 1107–1112.
15. Ferrantini, M., et al. (1994). IFN-1 gene expression into metastatic murine adenocar-
cinoma(TS/A) results in CD8+ T cell-mediated tumor rejection and development of anti-
tumor immunity: Comparative studies with IFN- producing TS/A cells. J. Immunol. 153:
4604–4615.
16. Kaido, T., et al. (1995). IFN-1 gene transfection completely abolishes the tumorigenicity
of murine B16 melanoma cells in allogeneic DBA/2 mice and decreases their tumori-
genicity in syngeneic C57BL/6 mice. Int. J. Cancer 60: 221–229.
17. Sarkar S., et al. (1995). Injection of irradiated B16 melanoma genetically modified to
363
 ARTICLE doi:10.1006/mthe.2001.0464, available online at http://www.idealibrary.com on IDEAL
secrete IFN- causes regression of an established tumor. Int. J. Oncol. 7: 17–24.
18. Tüting, T., et al. (1997). Interferon- gene therapy for cancer: retroviral transduction of
fibroblasts and particle-mediated transfection of tumor cells are both effective strategies
for gene delivery in murine tumor models. Gene Ther. 4: 1053–1060.
19. D’Amore, P. A. (1992). Mechanisms of endothelial growth control. Am. J. Respir. Cell Mol.
Biol. 6: 1–8.
20. Sidky, Y., and Borden, E. (1987). Inhibition of angiogenesis by interferons: effects on
tumor- and lymphocyte-induced vascular responses. Cancer Res. 47: 5155–5161.
21. Dvorak, H. F., and Gresser, I. (1989). Microvascular injury in pathogenesis of interferon-
induced necrosis of subcutaneous tumors in mice. J. Natl. Cancer Inst. 81: 497–502.
22. Dinney, C. P. N., et al. (1998). Inhibition of basic fibroblast growth factor expression,
angiogenesis, and growth of human bladder carcinoma in mice by systemic interferon-
 administration. Cancer Res. 58: 808–814.
23. Singh, R. K., et al. (1995). Interferons  and
down-regulate the expression of basic
fibroblast growth factor in human carcinomas. Proc. Natl. Acad. Sci. USA 92: 4562–4566.
24. Dong, Z., et al. (1999). Suppression of angiogenesis, tumorigenicity, and metastasis by
human prostate cancer cells engineered to produce interferon-
. Cancer Res. 59:
872–879.
25. Gutterman, J. U. (1994). Cytokine therapeutics: lessons from interferon-. Proc. Natl.
Acad. Sci. USA 91: 1198–1205.
26. Salmon, P., Le Cotonnec, J. Y., Galazka, A., Abdul-Ahad, A., and Darragh, A. (1996).
Pharmacokinetics and pharmacodynamics of recombinant human interferon-
in healthy
male volunteers. J. Interferon Cytokine Res. 16: 759–764.
364
27. Fierlbeck, G., et al. (1996). Pharmacodynamics of recombinant IFN-
during long-term
treatment of malignant melanoma. J. Interferon Cytokine Res. 16: 777–781.
28. Qin, X.-Q., et al. (1998). Interferon-
gene therapy inhibits tumor formation and causes
regression of established tumors in immune-deficient mice. Proc. Natl. Acad. Sci. USA 95:
14411–14416.
29. Dong, Z., et al. (1998). Expression of interferon-
in UV2237m mouse fibrosarcoma cells
suppressed their tumorigenicity and metastasis in syngeneic mice. Cancer Immunol.
Immunother. 46: 137–146.
30. Lu, W., Fidler, I. J., and Dong, Z. (1999). Eradication of primary murine fibrosarcomas
and induction of systemic immunity by adenovirus-mediated interferon
gene therapy.
Cancer Res. 59: 5202–5208.
31. Xie, K., et al. (1997). Abrogation of tumorigenicity and metastasis of murine and human
tumor cells by transfection with murine interferon
gene: possible role of nitric oxide.
Clin. Cancer Res. 3: 2283–2294.
32. Tada, H., et al. (2001). Systemic IFN-
gene therapy results in long-term survival in mice
with established colorectal liver metastases. J. Clin. Invest. 108: 83–95.
33. Gao, G. P., Yang, Y., and Wilson, J. M. (1996). Biology of adenovirus vectors with E1
and E4 deletions for liver-directed gene therapy. J. Virol. 70: 8934–8943.
34. Nakabayashi, H., Taketa, K., Yamane, T., Oda, M., and Sato, J. (1985). Hormonal con-
trol of -fetoprotein secretion in human hepatoma cell lines proliferating in chemically
defined medium. Cancer Res. 45: 6379–6383.
35. Wierda, W.G., Mehr, D.S., and Kim, Y. B. (1989). Comparison of fluorochrome-labeled
and 51-Cr- labeled targets for natural killer cytotoxicity assay. J. Immunol. Methods 122:
15–24.
MOLECULAR THERAPY Vol. 4, No. 4, October 2001
Copyright © The American Society of Gene Therapy