Tumor cells as cellular vehicles to deliver gene therapies to metastatic tumors

Abstract

A long-pursued goal in cancer treatment is to deliver a therapy specifically to metastases. As a result of the disseminated nature of the metastatic disease, carrying the therapeutic agent to the sites of tumor growth represents a major step for success. We hypothesized that tumor cells injected intravenously (i.v.) into an animal with metastases would respond to many of the factors driving the metastatic process, and would target metastases. Using a model of spontaneous metastases, we report here that i.v. injected tumor cells localized on metastatic lesions. Based on this fact, we used genetically transduced tumor cells for tumor targeting of anticancer agents such as a suicide gene or an oncolytic virus, with evident antitumoral effect and negligible systemic toxicity. Therefore, autologous tumor cells may be used as cellular vehicles for systemic delivery of anticancer therapies to metastatic tumors.

Full text

Tumor cells as cellular vehicles to deliver gene therapies
to metastatic tumors
Javier Garcı
´
a-Castro,
1,5
Jesu
´
s Martı
´
nez-Palacio,
1
Rosa Lillo,
2
Fe
´
lix Garcı
´
a-Sa
´
nchez,
2
Ramo
´
n Alemany,
3
Luis Madero,
4
Juan A Bueren,
1
and Manuel Ramı
´
rez
4
1
Unidad de Hematopoyesis y Terapia Ge
´
nica, CIEMAT/Fundacio
´
n Marcelino Botı
´
n, Madrid, Spain;
2
Centro
de Transfusio
´
n, Madrid, Spain;
3
Institut Catala
`
d’Oncologı
´
a, Barcelona, Spain; and
4
Oncologı
´
a Pedia
´
trica,
Hospital Universitario Nin
˜
o Jesu
´
s, Madrid, Spain.
A long-pursued goal in cancer treatment is to deliver a therapy specifically to metastases. As a result of the disseminated nature of
the metastatic disease, carrying the therapeutic agent to the sites of tumor growth represents a major step for success. We
hypothesized that tumor cells injected intravenously (i.v.) into an animal with metastases would respond to many of the factors
driving the metastatic process, and would target metastases. Using a model of spontaneous metastases, we report here that i.v.
injected tumor cells localized on metastatic lesions. Based on this fact, we used genetically transduced tumor cells for tumor
targeting of anticancer agents such as a suicide gene or an oncolytic virus, with evident antitumoral effect and negligible systemic
toxicity. Therefore, autologous tumor cells may be used as cellular vehicles for systemic delivery of anticancer therapies to
metastatic tumors.
Cancer Gene Therapy advance online publication, 14 January 2005; doi:10.1038/sj.cgt.7700801
Keywords: neoplasm metastasis; delivery system; oncolytic virus; NOD SCID mice
C
ancer is often diagnosed when the disease has already
disseminated and most of the patient deaths are
related to metastatic disease. Current treatment options
for metastatic tumors lack efficacy and metastases
targeting remains a major challenge for curing cancer.
Novel cell- and gene-based therapies have been developed
aimed to deliver an antitumor effect specifically to
metastases. Some strategies rely on the immune system
for selectivity, such as the stimulation of effector immune
cells
1
or the use of monoclonal antibodies that recognize
tumor antigens.
2
Others target angiogenic
3–5
or anaerobic
signals
6,7
that arise at the metastases. Recently, onco-
tropic viruses have been used to selectively kill the tumor
cells with the advantage that once targeting is achieved
the propagation of the virus amplifies the therapy.
8,9
These promising new therapies have fallen behind
expectations mainly because of their limited capacity for
effectively targeting in vivo.
2,10
The process of metastasis is a turning point in the
progression of malignant solid tumors. Million tumor
cells leave the primary tumor, reach the bloodstream and
disseminate through the body.
11
Only a minority will
eventually survive to become the origin of a new cancer
nodule
12
at anatomical sites that are specific for the tumor
type.
13
The dissemination of cancer cells from the primary
tumor and their homing in specific organs involve several
steps: invasion, detachment, circulation, cell adhesion,
motility and invasion again. Cells must express a specific
receptor molecule repertoire (cell adhesion molecules,
chemokine receptors or integrin ligands among others) to
complete this metastatic process.
14–17
We hypothesized that cancer cells may be good
candidates to target established metastases in vivo because
they express the receptor and effector molecules involved
in the metastatic process. Autologous intravenously (i.v.)
injected tumor cells should respond to metastasis-related
cues as the cells leaving the primary tumor did when the
latter formed the metastases. In the present work, we
demonstrated that i.v. injected tumor cells localized on
established metastases and were able of carrying genetic-
based antitumor agents such as suicide genes or oncolytic
viruses to the metastases, delivering the therapy in a
localized and toxic-tempered fashion.
Methods
Cell lines
The human breast cancer cell line MDA-MB-231 and the
human prostate cancer cell line PC3 were maintained in
DMEM supplemented with 10% fetal bovine serum
Received May 31, 2004.
Address correspondence and reprint requests to: Dr Manuel
Ramı
´
rez, MD, PhD, Oncologı
´
a Pedia
´
trica, Hospital Universitario
Nin
˜
o Jesu
´
s, Avenida Mene
´
ndez Pelayo, 65, 28009-Madrid, Spain.
E-mail: mramirezo.hnjs@salud.madrid.org
5
Present address: Servicio de Oncologı
´
a, Hospital Nin
˜
o Jesu
´
s,
Madrid.
Cancer Gene Therapy (2005), 1–9
r
2005 Nature Publishing Group All rights reserved 0929-1903/05 $30.00
www.nature.com
/
cgt

(FBS), glutamine and antibiotics (Life Technologies, Inc.,
Grand Island, NY).
Transduction of tumor cells with adenovectors or
oncolytic adenoviruses
MDA-MB-231 or PC3 cells were transduced with
different adenoviral vectors or oncolytic adenoviruses
depending on the experiments. The adenovectors had
either the lac-Z gene (AdLacZ), or the cytosine deaminase
(CD) gene (AdCD
18
), under the control of the cytome-
galovirus (CMV) promoter. The oncolytic adenovirus
(AdwtFi-GFP) contains the wild-type E1 region and,
after the fiber sequence (L5), the GFP coding sequence
has been inserted with a splicing acceptor from the IIIa
adenovirus gene. In this way, the adenovirus major late
promoter regulates GFP, and viral replication can be
followed by detection of green fluorescent cells.
The cells were transduced in DMEM, under conditions
optimized so that 100% of the cells were infected: 1hour
infection with 100 pfu/cell.
18
Evaluation of bystander effect in vitro
Tumor cells were mixed with different proportions of
AdCD transduced cells, and cultured in the presence of
400 mM 5-fluorocytosine (5-FC, Sigma Chemical Co., St
Louis, MO). The cell viability was evaluated after 4 days
by Trypan-blue (Sigma) exclusion of viable cells and
crystal violet staining. In different experiments, tumor
cells were mixed with different proportions of Ad-
wtFiGFP-infected cells. The cell viability was evaluated
after 6 days by Trypan-blue exclusion of viable cells and
green fluorescence determination.
In vivo studies
We used 6–8-week-old NOD.CB17-Prkdcscid/J (NOD
scid) mice, bred at the CIEMAT Laboratory Animals
Facility (Registration Number 28079-21 A) from breeding
pairs originally obtained from Jackson Lab (Bar Harbor,
Maine). The mice were routinely screened for pathogens,
in accordance with FELASA (Federation of European
Laboratory Animal Science Associations) procedures.
They were housed in microisolator individually ventilated
cages, and allowed water and food ad libitum. All
experimental procedures were carried out according to
European and Spanish laws and regulations and internal
biosafety and bioethics guidelines.
MDA-MB-231 cells were implanted in the mammary
fat pad of female NOD/SCID mice while PC3 cells were
implanted in the subcutaneous flank of male NOD/SCID
mice. At indicated time points, the animals were
euthanized by CO
2
inhalation. Samples from different
organs were collected in 10% buffered formalin, Tissue-
Tek (O.C.T. Compound, Zoeterwoude, The Netherlands)
or immediately frozen.
Detection of tumor cells by RT-PCR
RNA was purified using the TRIzol reagent (Life
Technologies). RT-PCR amplification of a human cyto-
keratin-19 sequence was carried out as published else-
where.
18
b-gal þ cells were detected using the following
lac-Z primers (sense: 5
0
CCGATCGCGTCACACTAC3
0
and antisense: 5
0
CAGATGATCACACTCGGG 3
0
).
Histology and immunohistochemistry
Sections of paraffin-embedded organs were stained with
hematoxilin–eosin, or with the CAM5.2 monoclonal
antibody (anti-human pancytokeratin, Becton Dickinson,
San Jose, CA) using a standard indirect avidin–biotin
horseradish peroxidase method (Vectastain ABC kit,
Vector Laboratories Inc., Burlingame, CA). Color was
developed with diaminobenzidine (DAB, Vector Labora-
tories Inc.) and sections were counterstained with
hematoxylin. Transduced tumor cells were stained with
the anti-b-galactosidase Ab1 antibody (Oncogene, Cam-
bridge, MA) and counterstained with eosin. GFP
þ
cells
were stained with a rabbit antiGFP IgG fraction
(Molecular Probes, Eugene, OR) and counterstained with
hematoxylin.
Cryosections were stained with X-gal (5-bromo-4-
chloro-3-indolyl-b-
D-galactoside, United States Biochem-
ical Corp., Cleveland, OH) and then with the CAM 5.2
antibody, and counterstained with eosin.
Evaluation of lung area, lung metastases and
localization of the i.v. injected tumor cells
Images of the lung cryosections stained with the CAM 5.2
antibody (n ¼ 10, two animals) were digitalized and
analyzed with the IMAGE-PRO Plus 4.5 software (Media
Cybernetics Inc., Carlsbad, CA). Different colors were
assigned to healthy and to metastatic tissues, and the total
lung and metastatic areas were calculated in square pixels
for each color, by an independent researcher (Dr Jose
´
M
Martı
´
nez, departamento de Ciencias de Materiales,
Universidad Polite
´
cnica de Madrid). The associated error
was estimated as less than 4% in triplicate measurements.
The number of b-gal þ cells and its localization in the
lungs were directly scored under a light microscope on
lung cryosections stained with X-gal and the CAM 5.2
antibody (n ¼ 10, two animals). Colocalization was
defined as the presence of a b-gal þ cell directly in
contact with a cytokeratin-positive nodule. All other
situations were not considered as colocalization.
Evaluation of bystander effect in vivo
(a) Cells transduced with the AdCD vector. Metastatic
tumors were generated as explained before. The primary
tumors were excised and a group of mice received five
doses of 3  10
5
transduced cancer cells, two doses per
week. An osmotic pump (Alza Corporation, Palo Alto,
CA) delivering 5-FC at a rate of 0.25 mL/hour 5-FC
(10 mg/mL) was subcutaneously implanted in the animals.
At 10 days after the last i.v. injection, the mice were killed,
the diameters of the remaining primary tumors were
scored and the organs were recovered for histology and
quantification studies. Metastatic burden was evaluated
on DNA samples from the organs by dot blot hybridiza-
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
2
Cancer Gene Therapy

tion with a human specific probe (a kind gift of Dr John
Dick).
(b) Cells infected with the AdwtFiGFP virus. Metastatic
tumors were generated as explained before. The primary
tumors were excised and a group of mice received five
doses of 3  10
5
transduced cancer cells, two doses per
week. At 10 days after the last i.v. injection, the mice were
killed, the fresh organs were analyzed under a fluorescent
light and the lungs were weighed and recovered for
immunohistochemical analysis.
Statistical analysis
Statistical analysis was performed using the Stata
software program (Stata Press, College Station, TX).
Results are expressed as mean7standard error (SE).
Results were considered significant if the P value was
equal to or less than .05. We used the two-sample
Wilcoxon rank-sum test (Mann–Whitney two-sample
statistic) for comparisons.
Results
Tumor cells colocalized on pre-existing metastases
We generated metastatic cancer in mice by implanting
MDA-MB-231 human breast cancer cells in the mam-
mary fat of female NOD/SCID mice. In this model,
human breast cancer cells formed a primary tumor in the
mammary fat, and metastasized preferentially to the local
lymph nodes, lungs, kidneys, liver and other organs (Fig
S1). In a different experiment, we determined whether i.v.
injected tumor cells would localize on pre-existing
metastases. In mice with metastatic disease, we injected
i.v. a single dose of MDA-MB-231 cells transduced with
an adenoviral vector carrying the lac-Z transgene.
Transduction conditions were optimized so that all the
injected cells expressed the transgene.
18
At 24 hours after
i.v. injection, we determined the localization of the
transduced cancer cells. Immunohistochemical analysis
showed b-gal þ cells in micrometastatic nodules of the
lungs and kidneys (Fig 1). Microscopic examination
(serial sections of 5 mm) showed that 46% of the b-gal þ
cells in the lungs colocalized with micrometastases. A
total of 19% of the total lung area was tumoral and an
average of 20% of the lung metastases had b-gal þ cells
after a single i.v. injection of 3  10
5
transduced tumor
cells. We did not detect b-galactosidase activity (X-gal
staining) in the preparations we studied from the kidney
(10 sections), liver (10 sections) and cerebrum (10
sections). Using a different detection technique, we
confirmed the presence of b-gal þ cells in the primary
tumor and in organs with metastases by RT-PCR of the
lac-Z transgene (Fig S1). These experiments demonstrated
that i.v. injected tumor cells targeted metastatic lesions. In
addition, injected tumor cells carried a heterologous
molecule to the metastases. Preliminary experiments using
the same approach with PC3 (human prostate) cancer
cells in male NOD/SCID mice confirmed that i.v. infused
cancer cells localized on established metastases.
Ex vivo manipulated tumor cells exerted a bystander
toxic effect in vitro and in vivo on the metastases
In the light of these findings, we used the tumor cells for
cancer treatment. MDA-MB-231 or PC3 cancer cells were
transduced with an adenoviral vector carrying the
Escherichia coli enzyme cytosine deaminase (CD)
(AdCD). The cells expressing the CD transform the
prodrug 5-FC into the anticancer drug 5-fluoruracil (5-
FU), which eliminates the transduced cells and the cells
around them (bystander effect). Transduced cells were
mixed with different proportions of untransduced cells
and cultured in the presence of 5-FC. The viability of the
mixture decreased when 5% of cells in the starting
cultures were transduced (Fig S2). We next studied
whether the ex vivo manipulated tumor cells have
therapeutic effect in vivo. We generated metastatic cancer
in mice (as explained) and, when the primary tumor
reached a major axis of 10 mm, we surgically eliminated as
much tumor as possible before delivering the treatment.
For treatment, we implanted an osmotic pump delivering
0.25 mL/hour 5-FC (10 mg/mL) for 4 weeks in a first group
of mice, and i.v. injected them with five doses of MDA-
MB-231 cells transduced with the AdCD vector (two
doses per week, 3  10
5
cells per dose). At 10 days after
the last i.v. injection, the mice were killed and analyzed.
The macroscopic examination showed that the treated
mice had a significant smaller primary tumor volume than
the untreated group (received only surgery). Large
infiltrated lymph nodes, local and distant to the primary
tumor, could be seen in the animals of the control group
but were absent in the treated mice. Histological
examination of the lungs showed abundant large cancer
nodules in the untreated mice while the treated mice had
no visible nodules (Fig 2a). Moreover, we quantified the
amount of cancer in the organs prone to metastasis and
found that the treatment had diminished the metastatic
burden in the lungs and in kidneys when compared with
the control group (Fig 2b). We did not detect tumor cells
in the control mice that received only transduced tumor
cells i.v. and the implantation of the 5-FC osmotic pump
(not shown).
Ex vivo manipulated tumor cells carried an oncolytic
adenovirus to the metastases
Aiming to validate this experimental approach with a
different antitumor agent, the i.v. injected cancer cells
were infected with oncolytic adenoviruses. We used a
replication-competent adenovirus, AdwtFiGFP. Follow-
ing the sequence encoding the fiber (L5), this vector
contains an adenovirus IIIa splicing acceptor and the
GFP coding sequence (such an L6 unit). Therefore, GFP
is expressed as another late protein upon replication. We
expect this adenovirus to replicate in tumor cells, lyse
them and spread to the surrounding tumor cells. Viral
replication can be followed by detection of green
fluorescent cells. We used it as a proof of concept that
tumor cells may transport oncolytic viruses to the
metastases. In our model, the AdwtFiGFP cannot
replicate in murine cells but it does in the human tumor
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
3
Cancer Gene Therapy

cells. In preliminary in vitro experiments to assess the
bystander effects of this strategy, MDA-MB-231 and
PC3-infected cells were mixed with different proportions
of uninfected cells. The infection kinetic was followed by
periodical analyses under a fluorescence microscope. The
viability of the mixtures was determined after 6 days and
decreased when as low as 5% of the infected cells were
present in the culture (not shown). Longer culture times
ended up with a complete mortality of the mixtures. In a
set of mice with metastatic cancer, after surgical elimina-
tion of their mammary fat pad tumors (as above), we
injected i.v. cells carrying the AdwtFiGFP adenovirus.
The mice received five doses of infected cells (two doses
per week, 3  10
5
cells per dose) and 10 days after the last
dose were killed and their fresh organs examined under
fluorescent light (Fig 3a). We found nodules expressing
the GFP in the lung and liver of three out of seven mice,
and confirmed that the images corresponded to tumor
Figure 1 I.V. injected tumor cells localized on established metastases. (a) Lung cryosections stained with X-gal and the CAM5.2 antibody (anti-
human pancytokeratin) showed b-gal þ cells (i.v. injected) in or by established metastases. (b) Above: a low-power view showing the
hematoxilyn–eosin and CAM 5.2 antibody stain for the tumor nodule. Below: three different paraffin-embedded sections of the metastasis stained
with an anti-b-galactosidase antibody. Positive cells were detected at different levels within the tumor nodule.
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
4
Cancer Gene Therapy

nodules by RT-PCR of the human cytokeratin-19 gene.
Green nodules were not seen in control mice (received
surgical treatment) or mice injected only with the infected
cells (the AdwtFiGFP does not replicate in murine cells),
indicating that the green nodules corresponded to
metastases where AdwtFiGFP was actively replicating.
Immunohistochemical staining with an anti-GFP anti-
body revealed the presence of GFP-positive tumor cells in
the middle of GFP-negative micrometastatic nodules (Fig
S3). These results showed that i.v. injected tumor cells
initiated an oncolytic infection and that the lytic process
propagated to the surrounding tumor mass. This onco-
lytic infective process was active long after the last i.v.
injected cells had died (in vitro cell lysis happened within
96 hours), and resulted in a significant reduction in the
lung metastases of the treated mice compared to the
control mice (Fig 3b). Our results indicate that effective
targeting could be achieved with this delivery system and
virus replication in metastases was revealed by the
presence of GFP.
Figure 2 Ex vivo manipulated tumor cells exerted a bystander toxic effect on the metastases. (a) Hematoxilin–eosin colored sections of paraffin-
embedded lungs showed the presence of abundant metastases in the control mice but not in the treated ones. (b) Tumor volumes were
estimated by the formula V ¼ (p/6)a
2
b, where ‘‘a’’ was the short axis and ‘‘b’’ the long axis (left). Metastatic burden was quantitated as percentage
of human DNA by dot blot with a human-specific probe on DNA samples from the lungs and kidneys (right) (n ¼ 5 mice per group).
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
5
Cancer Gene Therapy

The therapeutical effect of i.v. injected tumor cells has
no detectable systemic toxicities
Microscopic examination of histological sections of the
organs from the treated mice did not show evidence that
the treatment had been toxic to the animals. We further
evaluated the toxicity associated to the treatment with i.v.
injected tumor cells by analyzing the renal and hepatic
functions. For this experiment, we did not generate any
Figure 3 Ex vivo manipulated tumor cells carried an oncolytic adenovirus to the metastases. (a) Upper: a cancer nodule in the lung surface of a
treated mouse is shown (left), and the same nodule is under fluorescent light (right). Bottom: a cancer nodule in the lung surface of a control
mouse (left). The same nodule under fluorescent light (right). (b) Lungs were weighed after killing. Weight increase was calculated in comparison
to healthy mice (n ¼ 7 mice per group).
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
6
Cancer Gene Therapy

metastatic cancer in the mice before the i.v. injections, but
mice received the same treatments as above. Biochemical
parameters of both renal (urea and creatinine) and
hepatic (transaminase enzymes, total proteins, bilirrubin
and gGT enzyme) functions were not affected by the
treatment (Table 1).
Discussion
We show here that i.v. injected tumor cells home in
established metastases. Our preliminary quantification
indicated that half of the b-gal þ cells in the lungs
localized in established metastases. Given that only 19%
of the total lung area were occupied by metastases, this
number suggests that tumor targeting was not a random
process. It has already been reported that intraperitone-
ally injected ovarian carcinoma cells preferentially loca-
lized at sites of locoregional metastatic ovarian cancer,
19
showing that injected cancer cells interacted with estab-
lished tumors. In addition to the lungs, we found b-gal þ
cells in different sections of a kidney metastasis (Fig 1).
Thus, localization cannot be solely explained by anato-
mical factors (route of injection and first pass effect). We
foresee three steps at which targeting may take place.
First, i.v. injected tumor cells may respond to the
chemokines involved in the preferential homing of
metastatic cells in organs.
16
Second, i.v. injected tumor
cells may attach to the vascular endothelium similarly to
intravascular micrometastases.
20
Last, i.v. injected tumor
cells should be able to participate in the crosstalk between
the cancer cells and factors from the host tissues that
governs the extravasation of the metastases.
21,22
Support-
ing this, in vitro experiments have shown that noninvasive
cancer cells can progress through three-dimensional
matrices following the track of invasive tumor cells,
which suggests that tumor cell migration and invasion
may be modulated by signals generated by other tumor
cells.
23
Infiltrating central nervous system tumors give rise
to signals, which can be read and followed by specific
neural cell types,
24
underscoring the role of tumor–host
signals for recruiting other cells into the metastases
environment. It is reasonable to assume that transducing
tumor cells with genes implicated in the metastatic process
would likely enhance the targeting efficacy.
Different cell types have been used as vector carriers for
systemic cancer gene therapies.
25
These cellular vehicles
act like shuttles for delivering vectors to the sites of tumor
growth much more efficiently than the vectors per se
regularly accomplish. Injection of oncolytic adenoviruses
to the circulation is followed by viral sequestration in the
liver.
8
The effective targeting of metastases when injecting
monoclonal antibodies is just 0.001–0.01% of the total
injected dose.
2
Therefore, there is a growing need of
vehicles that enhance the efficacy of new oncolytic viruses
or monoclonal antibodies.
2,10
Allogeneic tumor cells have
been tested as vehicles in locoregional tumors
19
but not as
systemic carriers. We present here two examples of how
ex vivo manipulated autologous tumor cells may deliver a
localized anticancer effect in vivo after systemic delivery.
Using autologous rather than allogeneic tumor cells will
prevent an immune rejection against the anticancer
cellular vehicles. In the case of AdCD-transduced cells,
locally produced 5-FU would be responsible for the
killing of the metastases, although a systemic effect of 5-
FU may also participate. The highest levels of the drug
would be reached at the site where it is produced, that is,
the injected cells. High levels of 5-FU from transduced
tumor cell should reach the metastases immediately after
being produced. The CD/5-FC system does not require
intercellular gap junctions, often lost in tumor cells,
26
for
bystander effect. This gives advantage over other suicide
gene therapy systems, such as the HSV-TK. Antitumor
immunity has been associated with the destruction of
tumor cells with the CD/5-FC gene therapy
27
and may
enhance the therapeutic benefits. As we used immunode-
ficient mice we could not evaluate this possibility. We
used the AdwtFiGFP adenovirus as a model of oncolytic
virus. Most of the advances on oncolytic adenoviruses
have been related to replication selectivity through
mutations or promoter insertions but a major hurdle
remains with regard to tumor targeting.
10
Our studies
focus on the delivery and detection of viral replication in
the targeted metastases and we used this kind of virus, a
replicating vector with no such mutations or promoter
insertions, as a proof of concept. Tumor cells present a
major advantage as vehicles for oncolytic viruses since
they are the primary cell types where the viruses replicate.
Candidate anticancer agents that would fit in our strategy
are suicide genes, antiangiogenic genes, oncolytic viruses,
monoclonal antibodies or drugs.
Safety is likely the most important concern when
considering new therapies. Our preliminary results about
Table 1 Biochemical evaluation of liver and kidney functions after treatment
ALT gGT Bilirubin Protein Creatinine Urea
(a) acute phase (10 days after last i.v. injection) and (b) chronic phase (50 days after last i.v. injection)
(a) Untreated 18.776 4.773 0.4470.1 4.7470.2 0.3170.05 42.771.1
AdCD treated 1477 5.374 0.3470.2 4.1270.6 0.370.02 38.372.9
(b) Untreated 13.471.3 4.471.9 0.570.2 570.4 0.370.02 4975.3
AdCD treated 9.576.6 7.271.9 0.470.2 4.770.3 0.470.05 39.374.2
ALT: alanine transaminase (U/L); gGT: gamma glutamil transferase (U/L); bilirubin (mg/dL); protein: (g/mL); creatinine: (mg/dL); urea:
(mg/dL).
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
7
Cancer Gene Therapy

systemic toxicities are encouraging, however, the potential
danger related to the infusion of live tumor cells back into
the patient needs to be addressed in further research. It is
well known that metastasis is a very inefficient pro-
cess,
12,22
even so, injected tumor cells may cause an
unwanted bystander effect on healthy cells, since targeting
is not totally specific. Risks can be greatly minimized
before infusionly optimizing the transduction efficiency
up to 100% as shown here, selecting the transduced cells
and inactivating the cells (lethal irradiation, chemicals).
Loading cancer cells with selective antitumor agents will
also enhance safety, that is, oncolytic viruses provide a
self-destructive mechanism in this strategy.
We consider our strategy as a treatment for the
metastatic disease rather than for the primary tumor,
even though we detected that i.v. injected tumor cells
localized in the primary tumor. Surgery of the primary
tumor is a standard first-line treatment for patients with
cancer, although it stimulates neoangiogenesis and
growth of metastases.
28
In this sense, our model resembles
the clinical situation of many patients with solid tumors.
The strategy shown is rationally different from the actual
anticancer therapies (surgery, chemoradiotherapies) and
they could be employed synergistically. We anticipate that
tumor cells may be loaded with different anticancer
agents, as unique or combined therapeutic elements, and
deliver its beneficial effect locally with less toxicity. In
summary, we show here that autologous tumor cells may
be used as a flexible tool for the treatment of the
metastatic disease. Further research is needed for addres-
sing crucial points before its use in patients. We have
designed experiments to rule out the ability of lethally
irradiated MDA-MB-231 cells to target metastatic le-
sions. Chasing experiments with a different tumor cell line
will help in clarifying the target specificity. In addition,
the experiments will be performed in immunocompetent
models with murine tumors in order to determine the role
of the immune system.
Acknowledgments
We thank Isabel de los Santos, Pilar Herna
´
ndez and
Sergio Garcı
´
a for technical assistance, and Dr Jose
´
M.
Martı
´
nez for his help with analyzing image files. This
work was supported in part by Fundacio
´
n Leucemia
Linfoma (MR) and Fundacio
´
n Oncohematologı
´
a Infantil
(JGC, LM and MR).
References
1. Rosenberg SA. Progress in human tumour immunology and
immunotherapy. Nature. 2001;411:380–384.
2. Carter P. Improving the efficacy of antibody-based cancer
therapies. Nat Rev Cancer. 2001;1:118–129.
3. Arap W, Pasqualini R, Ruoslahti E. Cancer treatment by
targeted drug delivery to tumor vasculature in a mouse
model. Science. 1998;279:377–380.
4. Hood JD, Bednarski M, Frausto R, et al. Tumor regression
by targeted gene delivery to the neovasculature. Science.
2002;296:2404–2407.
5. Lyden D, Hattori K, Dias S, et al. Impaired recruitment
of bone-marrow-derived endothelial and hematopoietic
precursor cells blocks tumor angiogenesis and growth.
Nat Med. 2001;7:1194–1201.
6. Sznol M, Lin SL, Bermudes D, et al. Use of preferentially
replicating bacteria for the treatment of cancer. J Clin
Invest. 2000;105:1027–1030.
7. Dang LH, Bettegowda C, Huso DL, et al. Combination
bacteriolytic therapy for the treatment of experimental
tumors. Proc Natl Acad Sci USA. 2001;98:15155–15160.
8. Alemany R, Balague C, Curiel DT. Replicative adenoviruses
for cancer therapy. Nat Biotechnol. 2000;18:723–727.
9. Bell JC, Garson KA, Lichty BD, et al. Oncolytic viruses:
programmable tumour hunters. Curr Gene Ther. 2002;2:
243–254.
10. Shayakhmetov DM, Li ZY, Ni S, et al. Targeting of
adenovirus vectors to tumor cells does not enable efficient
transduction of breast cancer metastases. Cancer Res.
2002;62:1063–1068.
11. Liotta LA, Kleinerman J, Saidel GM. Quantitative relation-
ships of intravascular tumor cells, tumor vessels, and
pulmonary metastases following tumor implantation.
Cancer Res. 1974;34:997–1004.
12. Fidler IJ. Metastasis: quantitative analysis of distribution
and fate of tumor embolilabeled with 125 I-5-iodo-2’-
deoxyuridine. J Natl Cancer Inst. 1970;45:773–782.
13. Nicolson GL. Organ specificity of tumor metastasis: role of
preferential adhesion, invasion and growth of malignant
cells at specific secondary sites. Cancer Metastasis Rev.
1988;7:143–188.
14. Liao F, Li Y, O’Connor W, et al. Monoclonal antibody to
vascular endothelial-cadherin is a potent inhibitor of
angiogenesis, tumor growth, and metastasis. Cancer Res.
2000;60:6805–6810.
15. Ruoslahti E, Giancotti FG. Integrins and tumor cell
dissemination. Cancer Cells. 1989;1:119–126.
16. Muller A, Homey B, Soto H, et al. Involvement of
chemokine receptors in breast cancer metastasis. Nature.
2001;410:50–56.
17. Felding-Habermann B, O’Toole TE, Smith JW, et al.
Integrin activation controls metastasis in human breast
cancer. Proc Natl Acad Sci USA. 2001;98:1853–1858.
18. Lillo R, Ramirez M, Alvarez A, et al. Efficient and non-
toxic adenoviral purging method for autologous transplan-
tation in breast cancer patients. Cancer Res. 2002;62:
5013–5018.
19. Coukos G, Makrigiannakis A, Kang EH, et al. Use of
carrier cells to deliver a replication-selective herpes simplex
virus-1 mutant for the intraperitoneal therapy of epithelial
ovarian cancer. Clin Cancer Res. 1999;5:1523–1537.
20. Al-Mehdi AB, Tozawa K, Fisher AB, et al. Intravascular
origin of metastasis from the proliferation of endothelium-
attached tumor cells: a new model for metastasis. Nat Med.
2000;6:100–102.
21. Liotta LA, Kohn EC. The microenvironment of the
tumour–host interface. Nature. 2001;411:375–379.
22. Cameron MD, Schmidt EE, Kerkvliet N, et al. Temporal
progression of metastasis in lung: cell survival, dormancy,
and location dependence of metastatic inefficiency. Cancer
Res. 2000;60:2541–2546.
23. Horino K, Kindezelskii AL, Elner VM, et al. Tumor cell
invasion of model 3-dimensional matrices: demonstration of
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
8
Cancer Gene Therapy

migratory pathways, collagen disruption, and intercellular
cooperation. FASEB J. 2001;15:932–939.
24. Aboody KS, Brown A, Rainov NG, et al. Neural stem cells
display extensive tropism for pathology in adult brain:
evidence from intracranial gliomas. Proc Natl Acad Sci
USA. 2000;97:12846–12851.
25. Harrington K, Alvarez-Vallina L, Crittenden M, et al. Cells
as vehicles for cancer gene therapy: the missing link between
targeted vectors and systemic delivery? Hum Gene Ther.
2002;13:1263–1280.
26. Holder JW, Elmore E, Barrett JC. Gap junction function
and cancer. Cancer Res. 1993;53:3475–3485.
27. Mullen CA, Coale MM, Lowe R, et al. Tumors expressing
the cytosine deaminase suicide gene can be eliminated in vivo
with 5-fluorocytosine and induce protective immunity to
wild type tumor. Cancer Res. 1994;54:1503–1506.
28. O’Reilly MS, Holmgren L, Shing Y, et al. Angiostatin: a
novel angiogenesis inhibitor that mediates the suppression
of metastases by a Lewis lung carcinoma. Cell. 1994;79:
315–328.
Supplementary information accompanies the paper on Cancer Gene Therapy website (http://www.nature.com/cgt).
Delivering antitumor therapies using tumor cells
J Garcı´a-Castro et al
9
Cancer Gene Therapy