Control of large, established tumor xenografts
with genetically retargeted human T cells
containing CD28 and CD137 domains
Carmine Carpenitoa, Michael C. Milonea,b, Raffit Hassanc, Jacqueline C. Simoneta, Mehdi Lakhala, Megan M. Suhoskia,
Angel Varela-Rohenaa, Kathleen M. Hainesa, Daniel F. Heitjand, Steven M. Albeldae, Richard G. Carrolla,b,
James L. Rileya,b, Ira Pastanc,1, and Carl H. Junea,b,1
aAbramson Family Cancer Research Institute,bDepartment of Pathology and Laboratory Medicine,dDepartment of Biostatistics and Epidemiology, and
eDepartment of Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA 19104; andcLaboratory of Molecular Biology,
National Cancer Institute, Center for Cancer Research, National Institutes of Health, Bethesda, MD 20892-4264
Contributed by Ira Pastan, December 23, 2008 (sent for review October 14, 2008)
Mesothelin is a cell-surface molecule over-expressed on a large
fraction of carcinomas, and thus is an attractive target of immu-
notherapy. A molecularly targeted therapy for these cancers was
created by engineering T cells to express a chimeric receptor with
high affinity for human mesothelin. Lentiviral vectors were used to
CD28, and CD137 (4–1BB). When stimulated by mesothelin, lenti-
virally transduced T cells were induced to proliferate, express the
antiapoptotic gene Bcl-XL, and secrete multiple cytokines, all fea-
tures characteristic of central memory T cells. When transferred
intratumorally or intravenously into NOD/scid/IL2r??/?mice en-
grafted with large pre-established tumors, the engineered T cells
reduced the tumor burden, and in some cases resulted in complete
eradication of the tumors at low effector-to-target ratios. Incor-
cells for multifunctional cytokine secretion and enhanced persis-
tence of T cells. These findings have important implications for
adoptive immunotherapy of cancer, especially in the context
of poorly immunogenic tumors. Genetically redirected T cells
have promise of targeting T lymphocytes to tumor antigens,
confer resistance to the tumor microenvironment, and providing
adoptive immunotherapy ? chimeric receptor ? mesothelin
I molecules is very low, no T cells are available that have high
avidity for tumor specific antigens, or no T cells that have the
desired specificities remain in the patient after chemotherapy.
To address these problems, tumor antigen-specific T cells have
been engineered by introduction of chimeric immunoreceptors
(CIR)—or ‘‘T bodies’’—that have antibody-based external re-
ceptor structures and cytosolic domains that encode signal
can function to retarget T cells in vitro in an MHC-unrestricted
manner. Several pilot clinical trials to test this concept have
recently been reported (2–4). In all of the trials, safety has been
documented. However, the principal issue revealed in the initial
trials has been poor in vivo persistence and expression of the
transgene. Approaches to remedy these shortcomings currently
involve improved receptor design, the incorporation of costimu-
latory signaling domains into the signaling module, and reducing
the immunogenicity of the T-body construct (5).
Mesothelin is a glycosyl-phosphatidyl inositol-linked mem-
brane glycoprotein expressed in a variety of cancers (6). Immu-
nohistochemistry studies have shown that mesothelin is over-
expressed in virtually all pancreatic ductal adenocarcinomas and
mesotheliomas (7), and in a high percentage of epithelial ovarian
cancers, as well as non-small cell carcinomas of the lung (8, 9).
principal limitation of cancer immunotherapy is that most
solid tumors are poorly antigenic, expression of MHC Class
Because mesothelin has recently been reported to bind to
CA125/MUC16, one possible biological role of this membrane
protein may be linked to heterotypic cell adhesion, which may
facilitate the metastatic spread of mesothelin-bearing cancer
cells (10). In earlier studies, tumor cells expressing mesothelin
have been shown to be killed by MHC Class I-restricted T cells
(11). In addition, mesothelin-specific antibodies and antibody-
based immunotoxins have been shown to cause tumor regression
in preclinical and clinical studies (12, 13). Here we report that
lentiviral vectors can efficiently generate T cells that specifically
target mesothelin. The mesothelin-retargeted T cells eradicate
large pre-established mesothelioma xenografts in NOD/scid/
IL2r?–/–mice at low effector-to-target (E:T) ratios, and the
incorporation of costimulatory domains enables the prolonged
persistence of the engineered T cells in tumor-bearing mice
following intratumoral or i.v. administration.
Chimeric receptors were designed that contain the SS1 scFv that
recognizes human mesothelin (Fig. 1A). SS1 was chosen because
it has a high binding affinity to mesothelin (Kd? 0.7 nM) (14),
and because it has been found to be safe in patients when
administered as a recombinant immunotoxin (13). We created a
series of SS1 scFv-based chimeric receptors that contain the
TCR-? signal-transduction domain with the CD28 and CD137
(4–1BB) intracellular domains in tandem. Chimeric receptors
that contained a truncated form of the TCR-? intracellular
domain (SS1-?z) or an anti-CD19 scFv (anti-CD19-z) were
designed as controls for signal transduction and binding speci-
Because of the increased efficiency of transduction and ex-
pression, and the possible decreased potential for adverse ef-
fects, such as insertional mutagenesis compared to retroviruses,
we used lentiviral-vector technology to express the fusion con-
structs in primary human T cells using clinically validated
techniques (15). The cDNA sequences containing the various
fusion constructions were cloned into a third-generation lenti-
viral vector in which the CMV promoter was replaced with the
EF-1? promoter (16). Lentiviral vector supernatants transduced
primary T cells with high efficiency (Fig. 1B). In preliminary
M.L., M.M.S., A.V.-R., and K.M.H. performed research; R.H. and I.P. contributed new
reagents/analytic tools; C.C., D.F.H., S.M.A., and R.G.C. analyzed data; and C.C. and C.H.J.
wrote the paper.
The authors declare no conflict of interest.
1To whom correspondence may be addressed. E-mail: or firstname.lastname@example.org or
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2009 by The National Academy of Sciences of the USA
March 3, 2009 ?
vol. 106 ?
experiments, the SS1-transduced T cells were found to have
sustained proliferation in the presence of various cell lines that
express mesothelin, while culture in the absence of mesothelin-
expressing feeder cells failed to sustain T-cell proliferation (data
of CD4 and CD8 T cells, and expression was stable for at least
2 months in culture (not shown). T-cell expansion during culture
contained SS1 scFv fused with either costimulatory domain than
for T cells expressing the SS1 TCR-? signaling domain only (data
not shown), consistent with previous reports that have tested T
cells with other chimeric receptors in vitro (17–22). Further-
more, the SS1 scFv fused with costimulatory domains enriched
the population of T cells to near purity during culture on
mesothelin-expressing tumors (data not shown).
To investigate the antitumor potential of the transduced T
cells, effector function was measured in standard51Cr release
assays using mesothelin-negative K562 cells, K562.meso (a de-
rivative engineered to express mesothelin), and primary tumor
lines isolated from patients with ovarian cancer and malignant
mesothelioma (Fig. 1C). T cells transduced with SS1 scFv
efficiently lysed K562.meso but did not kill parental K562 cells.
Importantly, the SS1-transduced T cells were also highly cyto-
toxic for carcinoma cells that express mesothelin, killing
OvCa68.4 and M108 cell lines derived from patients with ovarian
that did not express mesothelin (OvCa61.4). Transduction of
OvCa61.4 cells with a mesothelin-expressing lentiviral vector
rendered them susceptible to SS1 scFv lysis (data not shown).
The inclusion of CD28 and CD137 costimulatory domains in
tandem or in triplicate with TCR-? generally failed to increase
in vitro cytotoxicity above that of T cells expressing SS1-TCR-?
only. The killing was efficient, with plateau lysis occurring at a
10:1 E:T ratio during a 4-h assay, and at ?1:1 E:T ratio during
a 48-h culture [supporting information (SI) Fig. S1], suggesting
that the redirected T cells were capable of serial killing. More-
over, the lysis was specific because T cells transduced with GFP
or an irrelevant anti-CD19 scFv chimeric receptor showed no
cytotoxic activity against the same target cells, excluding allo-
reactivity or nonspecific lysis. Furthermore, T cells expressing a
truncated TCR-? intracellular domain (SS1-?z) also failed to kill
mesothelin-expressing targets, demonstrating the requirement
for an intact TCR-? signaling domain.
Studies in mouse tumor models indicate that antitumor effects
of redirected CD8 T cells will likely be sustained by CD4 cells
(23). Therefore, we transduced primary human CD4 and CD8 T
cells and measured cytokine secretion after exposure to me-
sothelioma cells (Fig. S2A). The release of inflammatory cyto-
secreted Th1 cytokines. In contrast, the T cells secreted low or
undetectable amounts of IL-4, IL-5, IL-10, and IL-17 (data not
shown), consistent with a predominant Th1 cytokine bias under
these culture conditions. The pattern of cytokine release ob-
served was consistent with the known signal transduction prop-
erties of the various signaling domains, confirming the modular
nature of the domains and that they can function in cis on the
by CD4 T cells, and this was dependent on the CD28 domain.
Both CD4 and CD8 T cells released IFN-? and IL-6, and either
the CD28 or the CD137 signaling domain was sufficient. There-
fore, in contrast to cytotoxicity, where TCR-? was sufficient,
cytokine secretion was more pronounced in primary T cells
expressing a costimulatory signaling domain. Mesothelin-
independent cytokine secretion was not observed.
An emerging concept in the vaccine field indicates that the
presence of multifunctional T cells is associated with improved
for T-cell-based cancer therapies (24). Transduced T cells were
stimulated with mesothelin-expressing tumor cells and the CD4
and CD8 T-cell subsets stained for intracellular IFN-?, TNF-?,
IL-2, and GM-CSF (Fig. S2B). Cytokine production was re-
stricted to the fraction of T cells expressing chimeric receptors
(data not shown). The incorporation of the CD28 signaling
domain increased the fraction of cytokine secreting CD4 and
CD8 T cells early in culture; however, after 20 h of culture, the
fraction of responding cytokine-secreting cells was similar for all
signaling constructs. However, the fraction of polyfunctional T
cells was highly dependent on the presence of costimulatory
TCR-? domain were monofunctional. In contrast, highly multi-
functional T cells secreting all 4 cytokines were primarily ob-
served in the cells expressing the BB:TCR-? or 28:BB:TCR-?
notable for supporting multifunctional CD4 and CD8 T cells, a
The expression of antiapoptotic genes can confer resistance to
the toxic effects of the tumor microenvironment (25). Most of
the T cells expressing CD28, and to a lesser extent the CD28:BB-
signaling domains, expressed Bcl-XLafter culture with the M108
mesothelioma cell line (Fig. S2C). In contrast, T cells expressing
BB:TCR-?, or a truncated TCR-? had lower or undetectable
levels of Bcl-XL. The expression of Bcl-XLin T cells expressing
CD28 and CD137 signaling domains was not constitutive, as T
cells cultured in media (see Fig. S2C) or with mesothelin-
negative tumor cells (data not shown) did not express detectable
levels of Bcl-XL.
As an initial test of in vivo specificity of the SS1 constructs, a
modified Winn assay was done using A431 cells transduced with
mesothelin (26). Rag2?–/–mice were injected s.c. on opposite
flanks with either A431.meso or A431 parental tumor cells.
engineered T cells. (A) Schematic representation of the mesothelin-binding
domain and a specificity control receptor with an anti-CD19 scFv were also
constructed. (B) Expression of the SS1 scFv fusion proteins was examined on
human primary CD4 T cells. Transduction efficiencies are indicated with mean
fluorescence intensity of the transduced populations in parentheses. (C) Cell
surface expression of mesothelin on K562, K562.meso, OvCa61.4, OvCa68.4,
either the mouse anti-human mesothelin antibody CAK1 (solid black line) or
an isotype control (dotted line) followed by staining with a PE-conjugated
goat anti-mouse Ig. The cytolytic activity of the chimeric receptors on primary
of triplicate wells from 1 of at least 3 separate experiments.
Generation and cytolytic activity of antimesothelin lentiviral vector-
Carpenito et al.
March 3, 2009 ?
vol. 106 ?
no. 9 ?
T cells transduced with SS1 chimeric receptors were coinjected
with the tumor cells at a 1:2 E:T ratio (Fig. S3). SS1 receptors
containing the CD28 and CD137 domains were able to inhibit
tumor growth in a mesothelin-specific manner. However, the
TCR-? group was only able to delay A431.meso tumor growth.
In contrast, T cells transduced with the truncated TCR? or with
GFP had no effect on tumor growth. Even at a high E:T ratio
(8:1), the control T cells had no effect on tumor growth,
confirming an allogeneic effect was not contributing to the
antitumor effects (data not shown).
To further explore the potential antitumor efficacy of the SS1
scFv constructs, we developed a rigorous xenograft model using
M108 tumor cells. When implanted into the flanks of NOD/
to express mesothelin (Fig. S4). Groups of NOG mice were
injected in the flanks with serially passaged primary M108 tumor
cells (Fig. 2A). On weeks 6 and 7 after establishing the tumor,
when the tumor burden was ?500 mm3, the mice were treated
with intratumoral injections of 15 ? 106T cells (?70%–80%
Fig. 2A). Statistical modeling of the tumor growth curves
revealed significant differences among the treatment groups
(P ? 0.0001 by a Wald test that all groups gave equal log-tumor
volume curves). The 7 curves separated themselves into 4
groups: the 3 control treatments (saline, GFP, and ?z) were not
significantly different from each other and showed continued
growth beyond the time of T cell transfer; the TCR-? treatment
slowed growth or tumor decline in others (Fig. 2B); the BBz
treatment showed relatively slow tumor decline; and the 28z and
28BBz treatments showed rapid tumor decline and were statis-
tically indistinguishable from each other.
indicate the magnitude and reproducibility of the antitumor
costimulatory domains, with 1 mouse reaching a tumor volume
of 1,000 mm3before tumor regression to subpalpable levels. The
mice were killed when tumors reached volumes ?2,000 mm3and
tumors were photographed (Fig. S5). Tumors from mice treated
with control T cells were highly vascularized, while many of the
small tissue masses harvested from the 28z or 28BBz mice and
several of the BBz mice were fibrotic or necrotic masses.
However, small regions of tumor with abnormal morphology
were still evident, as determined by histology. Thus, T cells
expressing mesothelin-specific chimeric receptors containing
CD28 and 4–1BB domains were able to control or eradicate
large established tumors after intratumoral injections. There was
no significant contribution of alloreactivity to the antitumor
effects because mice treated with T cells expressing GFP or a
truncated TCR-? receptor had tumor growth that was equivalent
to mice injected with saline.
In the previous experiment, vigorous and specific antitumor
effects were noted in mice treated with intratumoral injections
of the engineered T cells, directly assessing their potential
antitumor effects in the vascularized tumor microenvironment.
We next tested whether the T cells would be effective using
several routes of administration (Fig. 2C). Tumors were estab-
lished with M108 cells, and after the tumors reached a mean
volume of ?500 mm3, injections of 10 ? 106T cells transduced
with SS1–28:BB:TCR-? (?90% positive) on weeks 6 and 7 were
given by i.v., i.p., or intratumoral routes. For these experiments,
the CD28:BB:TCR-? construct was chosen because it has dis-
played the most potent and reproducible antitumor effects, as
well as enhanced engraftment properties (see below). Following
i.v. and intratumoral injections, a potent antitumor effect was
again observed, with a more rapid reduction in tumor mass
following the intratumoral route of administration. We analyzed
tumor volumes on the natural log scale. For log-tumor volume,
the ranking (lowest to highest) was intratumoral SS1–28BBz, i.v.
SS1–28BBz, i.p. SS1–28BBz, intratumoral anti-CD19–28BBz,
and saline. The overall F test for comparison of the means was
significant at P ? 0.0001, and in pairwise comparisons each
treatment was not significantly different (at P ? 0.05) from the
adjacent treatments, but was significantly different from the
others. Thus, we conclude that SS1–28BBz eradication of M108
tumor is antigen-specific because the anti-CD19–28BBz group
displayed no antitumor activity, and that intratumoral appears to
be the superior route of administration, marginally faster than
i.v. (?7 day delay in response) but significantly better than i.p.
Because surprisingly short-term engraftment has been ob-
served in the pilot clinical trials testing chimeric receptors (2–4),
and because long-term engraftment of adoptively transferred T
cells correlates with antitumor effects (27, 28), we next deter-
(A) Human primary M108 tumors were established in the flanks of NOG mice.
After 6 weeks, when the tumors reached a volume of ?500 mm3, the mice
were treated with 2 intratumoral injections of 15 ? 106T cells (?70%–80%
chimeric receptor-positive) on days 46 and 53 (arrows). Results are expressed
as a mean tumor volume (mm3? SEM) with n ? 8 for all groups, and are
representative of 2 experiments. The plots for BBz, CD28z, and CD28BBz are
significantly different from the control treatment groups (P ? 0.0001 by a
Wald test that all groups gave equal log tumor volume curves). (B) The
tumor-growth curves of the individual mice in each group are shown. (C)
M108-bearing NOG mice were treated with T lymphocytes expressing the
28BBz chimeric receptors (against mesothelin or CD19) via intratumoral (IT),
tumors reached a mean volume of ?500 mm3, 2 injections of 10 ? 106T cells
(?90% CIR-positive) on days 43 and 49 (arrows) were performed. The results
are expressed as a mean tumor volume (mm3? SEM) with n ? 8 for the saline
and intratumoral injection groups, and n ? 7 for the i.v. and i.p. groups.
Mesothelin retargeted T cells eradicate large pre-established tumors
www.pnas.org?cgi?doi?10.1073?pnas.0813101106Carpenito et al.
mined the persistence of the lentiviral vector-engineered T cells
in tumor-bearing mice. We first examined the mice from the
experiment shown in Fig. 2A. Peripheral blood from M108-
bearing NOG mice treated with intratumoral injections of SS1
scFv-transduced T cells on days 46 and 53 after tumor injection
was obtained on day 73: that is, 20 days after the last adoptive
T cell transfer, and quantified for the presence of CD4 and CD8
T cells (see Fig. 3A). For all conditions tested, the CD8 T cell
counts were higher than the CD4 T cell counts. Analysis of
variance of the CD4 data indicates significant differences among
the treatment arms (P ? 0.0001). In pairwise comparisons of the
treatment groups, adjusted for multiplicity by the Tukey method,
than BBz and GFP, while the BBz, 28z, Zeta, and ?z groups were
indistinguishable. Analysis of the CD8 data also revealed sig-
nificant differences among the treatment arms (P ? 0.0001). In
Tukey-adjusted pairwise comparisons, the cell counts in mice
higher (P ? 0.05) than GFP, and the 28BBz, BBz, Zeta, 28z, and
?z groups were indistinguishable. It was unexpected that the
T-cell counts in the mice given T cells transduced with the 28z
were not higher than T cells expressing TCR-? only, given their
in Fig. S2). Furthermore, the persistence of T cells in peripheral
blood did not have a direct correlation with tumor burden
(compare Fig. 2 A and B with Fig. 3A).
Finally, we measured the persistence of T cells in vivo in the
mice given 28BBz as a function of the route of administration.
In the experiment shown in Fig. 2C, mice were injected with T
engraftment. Tumor volumes and peripheral blood persistence
on day 65 are plotted for each individual mouse in each group
(see Fig. 3B). By inspection of the plots, the presence of
SS1–28:BB:z T cells generally correlates inversely with tumor
burden in that the groups of mice given T-cell injections by
intratumoral or i.v. routes all had substantial levels of T cells in
peripheral blood and the lowest tumor burdens, while mice given
SS1–28:BB:z by i.p., or anti-CD19–28:BB:z by intratumoral
administration, had high-tumor burdens and low levels of T-cell
engraftment. The overall F test comparing the mean T-cell
counts across groups was significant (P ? 0.0001). The group
with the highest mean T-cell count was i.v. SS1–28BBZ. In
Tukey-adjusted pairwise comparisons of the means, i.v. SS1
administration gave a significantly higher (P ? 0.05) count than
i.p. SS1 administration and intratumoral anti-CD19 scFv admin-
istration. Mice treated with intratumoral SS1 also had signifi-
cantly higher (P ? 0.05) T-cell counts than the intratumoral
anti-CD19 group, suggesting that tumor antigen drives the
expansion of the adoptively transferred T cells in vivo. Further-
more, in addition to a low-level T-cell engraftment, mice injected
with T cells expressing a nonbinding anti-CD19 scFv 28:BB:z
receptor had tumor burdens equivalent to mice injected with
saline, indicating that the specific triggering of the T cell by the
for high-level engraftment and antitumor effects observed for the
T cells expressing the SS1 scFv 28:BB:z receptor. Thus, long-term
systemic persistence of SS1 scFv 28:BB:z engineered T cells is
driven by tumor antigen and occurs in tumor-bearing mice follow-
ing either intratumoral or i.v. administration, with i.v. statistically
equivalent to intratumoral but significantly better than i.p.
Retargeted T cells are particularly attractive for the therapy of
carcinomas, where the available endogenous T-cell receptor
repertoire may be limiting because of tolerance and previous
cytotoxic chemotherapy, and MHC Class I expression may be
decreased. However, while initial clinical studies demonstrate
feasibility with the retargeted T cells, poor in vivo persistence
and expression of the transgene have been documented, and the
therapy has had less clinical activity than expected (2–4). Our
studies have addressed all of these issues. Mesothelin proved to
be an attractive target as the retargeted T cells efficiently and
cytes after treatment of M108 tumor. (A) Peripheral blood from M108-bearing
NOG mice treated with intratumoral injections of SS1 chimeric receptor-
data indicates significant differences among the treatment arms (P ? 0.0001).
Results are expressed as a mean absolute count per ?L of peripheral blood ? SD
Fig. 2C were bled and analyzed for peripheral T cell persistence on day 65 by a
FACS Trucount assay. Absolute T-cell count per ?L of peripheral blood is shown
for individual mice in each group with corresponding tumor volumes directly
was significant (P ? 0.0001). In Tukey-adjusted pairwise comparisons of the
higher T-cell count than i.p. SS1 administration.
CD28 and 4–1BB signals enhance the persistence of human T lympho-
Carpenito et al.
March 3, 2009 ?
vol. 106 ?
no. 9 ?
specifically killed a variety of tumors that express mesothelin.
The engineered T cells could kill tumor cells at ?1:10 E:T ratio
in vitro, and at an unprecedented 1:40 E:T ratio in vivo (see
domain was most correlated with persistence of the T cells in
vivo, as the CD28 domain was predicted to be most efficient by
the preponderance of in vitro measures of cytokine function and
The eradication of large, long-term pre-established tumors by
immunotherapy has rarely been reported. Most preclinical mod-
els in a therapeutic setting have tested tumors that have been
implanted for less than 2 weeks before initiation of therapy (29).
We have presented evidence that T cells expressing SS1 fused to
large, well-established tumors of 500 to 1,000 mm3. Given that
the experiments in Fig. 2A used 2 injections of T cells containing
a total of 20 to 24 ? 106T cells transduced with the SS1
construct, and assuming that a 1,000-mm3tumor mass contains
at least 1 ? 109cells, we estimate that the T cells are able to
eradicate tumors at an initial E:T ratio of ?1:40 in vivo.
Therefore, our preclinical data would support treating patients
with tumor burdens of at least 1 ? 1012cells, because our current
large-scale manufacturing can routinely produce 1 ? 1011trans-
duced T cells during a 10-day culture (15). Tumor cells of 1 ?
1012is a clinically relevant tumor burden, representing 1 kg of
extent this tumor eradication can be attributed to serial killing
of tumor cells by the infused T cells and to the tumor-induced
proliferation of the retargeted T cells and the cytotoxic effects
of the daughter cells.
It is likely that several mechanisms account for the enhanced
efficiency of the redirected T cells observed in the present study.
First, previous studies have generally used T cells transduced
with retroviruses. The high efficiencies associated with lentivi-
ral-mediated transduction, combined with robust in vitro cell-
expansion methods, make the rapid expansion of large numbers
of therapeutically relevant number feasible. In the present study
we have used lentiviral vectors, which have a higher transduction
permitting the use of the T cells early at a time when we have
shown previously that the average telomere length of the cul-
tured T cells is actually longer than at the start of culture (30).
We attribute this to the previous demonstration that the anti-
CD28-driven culture system induces telomerase activity (31) and
preserves central memory cells (32). The use of T cells with
longer telomeres may be important, as the cells have a more
extensive replicative capacity and in adoptive transfer studies
with tumor-infiltrating lymphocytes, there is enhanced antitu-
mor efficacy in patients infused with ‘‘younger’’ tumor-
infiltrating lymphocytes (33). Second, in comparison to previous
reports, our transgenes are brightly expressed on the surface of
the T cells, which may be the result of the lentiviral vector design,
the use of EF-1? as an internal promoter (34), and the use of
CD28 costimulation, which increases the expression level of
transgenes (35). Although a number of studies have shown that
lentiviral vectors are superior in hematopoietic stem cells (36),
only one has directly compared the expression of transgenes in
human T cells following retroviral or lentiviral transduction, and
the lentiviral approach was superior (37). Third, we demon-
strated that incorporation of the costimulatory domains in cis
enhanced the persistence of T cells, confirming and extending
recent work indicating that costimulatory domains expressed in
trans in T cells can enhance the effects of retargeted T cells (38).
Our study suggests that in vitro experiments can be misleading in
predicting the efficacy of the engineered T cells in vivo. Further-
more, engineered cells expressing the CD137 signaling domain
were more likely to be multifunctional and persist in tumor-bearing
mice. See the SI Text for a discussion of these issues.
Previous studies have shown that T cells in the tumor micro-
environment have a number of signaling perturbations (39). One
study has shown that chimeric receptors with a CD28 signaling
domain enhances the resistance of T cells to regulatory T cells
(40). While we have not yet directly tested the effects of
regulatory T cells on T-effector cells expressing our chimeric
receptors, our data would suggest that the combination of
CD137 and CD28 are more effective than either CD137 or CD28
alone, based on the antitumor effects and persistence in tumor-
bearing mice. Although CD28 is sufficient for in vivo antitumor
activity and the addition of CD137 is required for engraftment,
the combination of the 2 signals may provide the most effective
therapy and long-term protection.
We tested several routes of administration, and found that the
intratumoral and i.v. routes were nearly equivalent and distinctly
superior to the i.p. route. The delay in tumor regression with the
i.v. route compared to the intratumoral route may be at least
partially explained by soluble mesothelin in the plasma of the
mice (data not shown), reflecting that patients with mesothelin-
expressing tumors have been documented to have circulating
mesothelin (41). We expect to primarily test the i.v. route in
phase I clinical studies; however, the intratumoral route may
have certain advantages. First, trafficking of T cells into solid
tumors can be rate-limiting (42) and direct injection eliminates
trafficking as a variable. Second, the intratumoral route may
increase the therapeutic index in situations where there is a
potential for on-target, off-organ toxicity. In specific, the
mesothelin-redirected T cells may cause peritonitis or pleuro-
pericarditis, because of patterns of tissue-specific expression (6).
We do not expect that this will be a major problem because
previous clinical studies with mesothelin antibodies and immu-
noconjugates have not yet revealed unmanageable toxicity (12,
13). Finally, recent studies have shown the safety and feasibility
of intratumoral injection of T cells (43). In summary, chimeric
T-cells targeted to mesothelin and possessing intracytoplasmic-
signaling domains from TCR?, CD28, and CD137 appear highly
effective at treating large, well-established tumors in mice.
Clinical translation of this approach to tumors, such as mesothe-
lioma and ovarian cancer, is currently underway.
Generation of Antimesothelin T-Body Molecules. Chimeric antimesothelin scFv-
fusion proteins were generated as described in the SI Text.
Animal Experiments. Xenograft tumors were established by serial passage in
mice of primary pleural effusion cells from mesothelioma patient M108. All
injections were s.c., and performed in the presence of a 50% solution of
allowed to grow in NOG mice for 6 to 7 weeks. The route, dose, and timing of
T-cell injections is indicated in the individual figure legends. Tumor dimen-
sions were measured with calipers, and tumor volumes calculated using the
formula V ?1⁄2 ? L ? W ? W, where L is length (longest dimension) and W is
bleeding and stained for the presence of human CD45, CD4, and CD8 T cells.
After gating on the human CD45? population, the CD4?and CD8? subsets
were quantified using TruCount tubes (BD Biosciences).
Statistical Analysis. All results were expressed as means ? SD or SEM, as
indicated. Tumor-volume data were transformed to the log scale before
analysis. Additional information for methods used is provided in the SI Text.
ACKNOWLEDGMENTS. We thank Gwendolyn Binder, Chrystal Paulos, and
Elizabeth Jaffee for the generous gift of mesothelin vector, and Ronghua Liu
work was supported by National Institutes of Health Grants 1R01CA120409,
5P50CA083638, and 2P01CA066726, the Alliance for Cancer Gene Therapy
and in part of by the Intramural Research Program of the National Institutes
of Health, National Cancer Institute, Center for Cancer Research.
www.pnas.org?cgi?doi?10.1073?pnas.0813101106Carpenito et al.
1. Eshhar Z, et al. (1996) The T-body approach: Potential for cancer immunotherapy. Download full-text
Springer Semin in Immunopathol 18:199–209.
2. Kershaw MH, et al. (2006) A phase I study on adoptive immunotherapy using gene-
modified T cells for ovarian cancer. Clin Cancer Res 12:6106–6115.
3. Park JR, et al. (2007) Adoptive transfer of chimeric antigen receptor re-directed
cytolytic T lymphocyte clones in patients with neuroblastoma. Mol Ther 15:825–833.
4. Till BG, et al. (2008) Adoptive immunotherapy for indolent non-Hodgkin lymphoma
5. Sadelain M, Riviere I, Brentjens R (2003) Targeting tumours with genetically enhanced
T lymphocytes. Nat Rev Cancer 3:35–45.
6. Hassan R, Bera T, Pastan I (2004) Mesothelin: a new target for immunotherapy. Clin
Cancer Res 10:3937–3942.
7. Argani P, et al. (2001) Mesothelin is overexpressed in the vast majority of ductal
serial analysis of gene expression (SAGE). Clin Cancer Res 7:3862–3868.
8. Chang K, et al. (1992) Characterization of the antigen (CAK1) recognized by mono-
clonal antibody K1 present on ovarian cancers and normal mesothelium. Cancer Res
9. Ho M, et al. (2007) Mesothelin expression in human lung cancer. Clin Cancer Res
10. Rump A, et al. (2004) Binding of ovarian cancer antigen CA125/MUC16 to mesothelin
mediates cell adhesion. J Biol Chem 279:9190–9198.
11. Thomas AM, et al. (2004) Mesothelin specific CD8? T cell responses provide evidence
of in vivo cross-priming by antigen presenting cells in vaccinated pancreatic cancer
patients. J Exp Med 200:297–306.
12. Hassan R, et al. (2007) Preclinical evaluation of MORAb-009, a chimeric antibody
targeting tumor-associated mesothelin. Cancer Immun 7:20–29.
13. Hassan R, et al. (2007) Phase I study of SS1P, a recombinant anti-mesothelin immuno-
toxin given as a bolus I.V. infusion to patients with mesothelin-expressing mesotheli-
oma, ovarian, and pancreatic cancers. Clin Cancer Res 13:5144–5149.
14. Li Q, Verschraegen CF, Mendoza J, Hassan R (2004) Cytotoxic activity of the recombi-
nant anti-mesothelin immunotoxin, SS1(dsFv)PE38, towards tumor cell lines estab-
15. Levine BL, et al. (2006) Gene transfer in humans using a conditionally replicating
lentiviral vector. Proc Natl Acad Sci USA 103:17372–17377.
16. Parry RV, et al. (2003) CD28 and inducible costimulatory protein Src homology 2
IL-2 expression in primary human CD4 T lymphocytes. J Immunol 171:166–174.
17. Haynes NM, et al. (2002) Single-chain antigen recognition receptors that costimulate
potent rejection of established experimental tumors. Blood 100:3155–3163.
18. Maher J, et al. (2002) Human T-lymphocyte cytotoxicity and proliferation directed by
a single chimeric TCR zeta/CD28 receptor. Nat Biotechnol 20:70–75.
19. Finney HM, Akbar AN, Lawson ADG (2004) Activation of resting human primary T cells
CD137 in series with signals from the TCR zeta chain. J Immunol 172:104–113.
20. Friedmann-Morvinski D, et al. (2005) Redirected primary T cells harboring a chimeric
receptor require costimulation for their antigen-specific activation. Blood 105:3087–
and supports clonal expansion of primary human T cells. Mol Ther 12:933–941.
22. Hombach A, et al. (2001) Tumor-specific T cell activation by recombinant immunore-
ceptors: CD3 zeta signaling and CD28 costimulation are simultaneously required for
efficient IL-2 secretion and can be integrated into one combined CD28/CD3 zeta
signaling receptor molecule. J Immunol 167:6123–6131.
by gene-modified CD4? T helper-1 and CD8? T cells. Cancer Res 67:11428–11437.
of T lymphocytes: elucidation of intracellular apoptotic events. Blood 95:2015–2023.
26. Winn HJ (1961) Immune mechanisms in homotransplantation. II. Quantitative assay of
27. Greenberg PD, Cheever MA (1984) Treatment of disseminated leukemia with cyclo-
phosphamide and immune cells: tumor immunity reflects long-term persistence of
tumor-specific donor T cells. J Immunol 133:3401–3407.
J Immunol 175:7046–7052.
29. Yu P, Rowley DA, Fu YX, Schreiber H (2006) The role of stroma in immune recognition
and destruction of well-established solid tumors. Curr Opin Immunol 18:226–231.
30. Weng N-P, et al. (1997) Tales of tails: regulation of telomere length and telomerase
activity during lymphocyte development, differentiation, activation, and aging. Im-
munol Rev 160:43–54.
31. Weng N-P, Levine BL, June CH, Hodes RJ (1996) Regulated expression of telomerase
activity in human T lymphocyte development and activation. J Exp Med 183:2471–
32. Bondanza A, et al. (2006) Suicide gene therapy of graft-versus-host disease induced by
central memory human T lymphocytes. Blood 107:1828–1836.
33. Shen X, et al. (2007) Persistence of tumor infiltrating lymphocytes in adoptive immu-
notherapy correlates with telomere length. J Immunother 30:123–129.
34. Serafini M, Bonamino M, Golay J, Introna M (2004) Elongation factor 1 (EF1alpha)
promoter in a lentiviral backbone improves expression of the CD20 suicide gene in
primary T lymphocytes allowing efficient rituximab-mediated lysis. Haematologica
35. Costello E, et al. (2000) Gene transfer into stimulated and unstimulated T lymphocytes
by HIV-1-derived lentiviral vectors. Gene Ther 7:596–604.
36. Miyoshi H, et al. (1999) Transduction of human CD34? cells that mediate long-term
engraftment of NOD/SCID mice by HIV vectors. Science 283:682–686.
37. Zhou X, et al. (2003) Lentivirus-mediated gene transfer and expression in established
human tumor antigen-specific cytotoxic T cells and primary unstimulated T cells. Hum
Gene Ther 14:1089–1105.
38. Stephan MT, et al. (2007) T cell-encoded CD80 and 4–1BBL induce auto-and transco-
stimulation, resulting in potent tumor rejection. Nat Med 13:1440–1449.
from tumor-bearing mice. Science 258:1795–1798.
40. Loskog A, et al. (2006) Addition of the CD28 signaling domain to chimeric T-cell
receptors enhances chimeric T-cell resistance to T regulatory cells. Leukemia 20:1819–
42. PalmerDC, etal.(2004)Vaccine-stimulated,adoptivelytransferredCD8?Tcellstraffic
indiscriminately and ubiquitously while mediating specific tumor destruction. J Im-
with a HLA-A2 restricted MART-1 T-cell receptor: A phase I trial in metastatic mela-
noma. Clin Cancer Res 12:1229–1236.
Carpenito et al.
March 3, 2009 ?
vol. 106 ?
no. 9 ?