Agonist Anti-cd137 Mab Act on Tumor Endothelial Cells to Enhance Recruitment of Activated T Lymphocytes

Agonist monoclonal antibodies (mAb) to the immune costimulatory molecule CD137, also known as 4-1BB, are presently in clinical trials for cancer treatment on the basis of their costimulatory effects on primed T cells and perhaps other cells of the immune system. Here we provide evidence that CD137 is selectively expressed on the surface of tumor endothelial cells. Hypoxia upregulated CD137 on murine endothelial cells. Treatment of tumor-bearing immunocompromised Rag À/À mice with agonist CD137 mAb did not elicit any measurable antiangiogenic effects. In contrast, agonist mAb stimulated tumor endothelial cells, increasing cell surface expression of the adhesion molecules intercellular adhesion molecule (ICAM)-1, vascular cell adhesion molecule (VCAM)-1, and E-selectin. When adoptively transferred into mice, activated T lymphocytes derived from CD137-deficient animals entered more avidly into tumor tissue after treatment with agonist mAb. This effect could be neutralized with anti–ICAM-1 and anti–VCAM-1 blocking antibodies. Thus, stimulation of CD137 not only enhanced T-cell activation but also augmented their trafficking into malignant tissue, through direct actions on the blood vessels that irrigate the tumor. Our findings identify an additional mechanism of action that can explain the immunotherapeutic effects of agonist CD137 antibodies.

While investigating genes whose expression was related to oncological angiogenesis, as opposed to the angiogenesis that occurs in nonmalignant tissue regeneration, it was found that mouse CD137 RNA probes were a selective marker of tumor related angiogenesis (9).Indeed, immunohistochemical evidence for CD137 expression on human malignant tumor capillaries had been previously obtained by the group of H. Schwarz (10).Moreover, evidence of expression of CD137 on the endothelial lining of atherosclerotic lesions in mice has been published (11).
The mechanism of action has been interpreted in the sense that a stronger CD8 T-cell-mediated immune response is artificially costimulated in treated animals (3,13,29).Direct stimulation of the CD137 receptor on T cells is definitely involved in the immunotherapeutic effects, but the interaction of the agonist antibodies with CD137 expressed on the surface of activated NK cells (15,30) and dendritic cells (5) can also be of importance (3).
If CD137 were expressed at the protein level on tumor endothelial cells, treatment with anti-CD137 mAb could signal into the vascular cells.We hypothesized that if CD137 signaled in the same way as in T cells via TNF receptor associated factor (TRAF) adaptors (1,31,32) toward NF-kB (31,32), proinflammatory molecules could be induced.Among them, receptors that would increase T-cell homing to tumors, such as intercellular adhesion molecule (ICAM)-1, vascular cell adhesion molecule (VCAM)-1, and E-selectin, were postulated to be involved (33).
Our experiments show that adoptively transferred activated CD137-deficient T cells enter more avidly into tumor lesions as a result of enhanced expression of homing receptors on endothelial cells that have received direct stimulation via their surface CD137.

Mice and cell lines
BALB/c wild-type (WT) mice (5-6 weeks old) were purchased from Harlan Laboratories.Rag À/À mice (BALB/c and C57BL/6) and MMTV-neuT mice (34) were purchased from the Jackson Laboratory and bred in our animal facility under specific pathogen-free conditions.CD137 knockout mice were generated at Lexicon Genetics.To generate CD137 À/À mice, 129/Svderived embryonic stem (ES) cells (Lex1) were electroporated with a targeting construct that replaced the entire coding region of CD137 (exons 2-5) with a lacZneo fusion cassette (Supplementary Fig. S2).ES cells that survived neomycin selection (300 mg/mL) were amplified and split into several aliquots.From several hundred ES cell clones, 6 clones were identified that had the proper recombination events at the 5 0 (internal probe) and 3 0 ends (external probe; Supplementary Fig. S2A).One of these clones (IE7) was injected into C57BL/6J blastocysts and yielded chimeric male offspring that were able to pass the CD137 disruption through the germline after mating with C57BL/6J females.The resulting germline mice were of mixed (C57BL/6J Â 129/Sv) background and were expanded as a mixed population for several generations prior to being subjected to speedbackcrossing (using microsatellite markers for BALB/c mice at Charles River Laboratories) for 5 generations to yield mice that were more than 99% BALB/c character.An additional 2 backcrosses were made to inbred BALB/c mice prior to expansion for use in these experiments.Because the entire coding region for CD137 is missing from these mice, there is no CD137 mRNA nor protein present in them (Supplementary Fig. S2B).
All animal procedures were conducted under institutional guidelines that comply with national laws and policies.
The murine endothelial cell line PY-4-1 was kindly provided by V. Bautch (University of North Carolina; ref. 35).The mouse coronary endothelial cells were purified and immortalized by A. Luque (36).In brief, endothelial cells were isolated by positive selection using platelet/endothelial cell adhesion molecule-1 or ICAM-2 (37) antibodies bound to magnetic beads (Dynabeads, Invitrogen).Subconfluent primary cultures were incubated in the presence of polybrene (8 mg/mL; Sigma) with supernatant from packaging cells producing recombinant retroviruses that transduced polyoma middle T antigen (virus concentration, approximately 10-200 pfu/cell).
Primary endothelial cells were isolated from lungs of C57BL/6 mice by collagenase digestion followed by selection using Dynabeads Sheep anti-rat IgG (Invitrogen) coupled to a rat anti-mouse ICAM-2 (BD Pharmingen).

In vivo tumor growth and microPET analysis
A total of 0.5 Â 10 6 CT26, EMT-6, and LLC cells were injected in 100 mL PBS subcutaneously in the flank.Mice were sacrificed when tumor sizes reached 150 mm 2 .Positron emission tomography (PET) analyses showing tumor hypoxia in CT26 established tumors with the radiotracer fluorine-18fluoromisonidazole ( 18 FMISO) synthesized as described (39).
Mice were injected with 18 F-FMISO (14.9 AE 4.9 MBq in 100 mL) in the tail vein.Four hours later, a static 30-minute study (sinogram) was acquired using a Mosaic (Philips) small animal dedicated imaging tomography with 11.9-cm axial field of view (FOV), 12.8-cm transaxial FOV and a 2-mm resolution.
For the assessment of tumor 18 F-FMISO uptake, all studies were exported and analyzed using the PMOD software (PMOD Technologies Ltd.).Regions of interest were drawn on coronal 1-mm-thick small-animal PET images on consecutive slices including the entire tumor.Finally, maximum standardized uptake value (SUV) was calculated for each tumor using the formula: SUV ¼ [tissue activity concentration (Bq/cm 3 )/ injected dose (Bq)] Â body weight (g).

Flow cytometry, antibodies, and tissue immunofluorescence
For flow cytometry analysis of tumor endothelial cells, we placed cleanly excised tumor nodules in Petri dishes and minced them finely with a scalpel blade, and incubated them for 45 minutes at 37 o C in a solution containing Collagenase-D and DNAse-I (Roche) in RPMI.We then took the entire material and passed it through a 70-mm cell strainer (BD Falcon, BD Bioscience) pressing with a plunger, to obtain unicellular cell suspensions.Single cell suspensions were pretreated with FcR-Block (anti-CD16/32 clon 2.4G2; BD Pharmingen).Afterwards, cells were stained with the following antibodies or reagents obtained from BD Pharmingen: CD34-FITC, CD31-PE, CD45.2-APC,CD137-biotin, Syrian Hamster Isotype control-biotin, ICAM-1-biotin, VCAM-1-biotin, E-selectin-biotin, Rat IgG2a, and RatIgG2b biotin conjugated isotype controls.Biotin-conjugated primary antibodies were detected with Streptavidine-APC-Cy7 and 4 0 ,6 diamidino 2 phenylindole was used for dead cell exclusion.Other antibodies used were CD3-FITC, CD8-APC, CD137-PE, and Syrian Hamster Isotype control-PE (BD Pharmingen).
FACS-Canto II and FACSCalibur (BD-Biosciences) were used for cell acquisition and data analysis was carried out using FlowJo software (Tree Star Inc.).For sorting, a FACS-ARIA (BD-Biosciences) was used.
The agonist mouse CD137 clone 1D8 was produced in Chinese hamster ovary (CHO) cells.The heavy and light chain genes for mAb 1D8 were cotransfected into DG44CHO cells maintained in protein-free CHO media (Sigma-Aldrich).Purified anti-CD137 mAb was certified to have less than 0.5 EU/ mg endotoxin level, more than 95% purity, and less than 5% high molecular weight species (Bristol-Myers Squibb).Stock solutions of anti-CD137 mAb were kept at À80 C and were thawed on ice prior to use.Polyclonal rat immunoglobulin G (IgG) lipopolysaccharide (LPS)-free was purchased from Sigma (<0.5 EU/mg endotoxin confirmed in our hands).
M-450 tosylactivated Dynabeads (Dynal) coupled with 1D8 anti-CD137 mAb or Rat IgG were used for in vitro stimulation of endothelial cells.Antibody coupling to the microbeads was done following the manufacturer's instructions.

Directed in vivo angiogenesis assay
Angiogenic responses were compared and quantitated as described previously (40).
In brief, this assay consists of subcutaneous implantation of semiclosed silicone cylinders (angioreactors) filled with CT26 tumor cells embedded in Matrigel (BD Biosciences; Supplementary Fig. S3A).Vascularization within angioreactors was quantified 11 days later by the intravenous injection of the intravascular contrast fluorescein isothiocyanate (FITC)-dextran, followed by spectrofluorimetry on the Matrigel from excised angioreactors (Supplementary Fig. S3B and C).As a positive control of angiogenesis inhibition, the compound 77427 was used (40).

Molecular analyses
Total RNA was extracted from tumor endothelial cells sorted by FACS-Aria (BD Biosciences) using RNAsy Mini Kit (Qiagen).We treated RNA with DNaseI (Gibco-BRL) prior to reverse transcription with Moloney murine leukemia virus (MMLV) reverse transcriptase (Gibco-BRL) in the presence of RNaseOUT (Gibco-BRL).Real-time PCR was done with iQ SYBR green supermix in an iQ5 real-time PCR detection system (Bio-Rad).Copy numbers of CD137 DNA were quantified by quantitative RT-PCR with primers annealing the mouse CD137 cDNA (forward primer: 5 0 -AACATCTGCA-GAGTGTGTGC-3 0 , reverse primer: 5 0 -AGACCTTCCGTCTA-GAGAGC-3 0 , product length: 252 bp).PCR amplification was done under the following conditions: 1 cycle of 3 minutes at 95 C; followed by 45 cycles of 30 seconds at 95 C, 15 seconds at 60 C, 30 seconds at 72 C, 25 seconds at 77 C, 10 seconds at 84 C, and 30 seconds at 85 C (detection temperature of 84 C); followed by a single final extension cycle of 72 C for 3 minutes.Immediately after the PCR, a melting curve was generated by raising the incubation temperature from 55 C to 95 C to confirm amplification specificity.The final PCR product was also analyzed by agarose gel electrophoresis.Samples were analyzed in duplicate and data were normalized by comparison with b-actin as an internal control (forward primer: 5 0 -CGCGTCCACCCGCGAG-3 0 , reverse primer: 5 0 -CCTGGTGCCTAGGGCG-3 0 , product length: 194 bp).The amount of each transcript was expressed according to the formula 2 Ct (b-actin) À Ct(CD137) , in which Ct is the point at which the fluorescence increases appreciably above background fluorescence.Primers for amplification of the 2 mCD137 isoforms are 5 0 -TGTGTGCAGGCTATTTCAGG-3 0 and 5 0 -GAGCTGCTCCAGTGGTCTTC-3 0 (35 cycles of 30 seconds at 95 C, 30 seconds at 60 C, 30 seconds at 72 C) with a 2720 Thermal Cycler (Applied Biosystems) and BioTaq DNA Polymerase (Bioline).Isoform 1 product length: 504 bp; isoform 2 product length: 369 bp.PCR products were visualized with SYBR Safe (Invitrogen) after electrophoresis on 1% agarose gels.
We treated recipient mice with anti-CD137 1D8 (100 mg per mouse) or LPS-free NA/LE Rat IgG as control for 48 hours before T-cell transfer.In some experiments, we treated mice i. p. with a neutralizing antibody against mouse ICAM-1 (50 mg per mouse, clone YN1/1.7.4; Biolegend) and against mouse VCAM-1 (50 mg per mouse, clone M/K-1.9).We harvested tumors 48 hours after T-cell transfer, and analyzed them for CFSE and CD8-APC in a FACS Canto II flow cytometer.Absolute numbers of infiltrating T cells were calculated and referred to the total number of lymphocytes counted per cell suspension (Z1 Coulter Particle Counter, Beckman Coulter).

Statistical analysis
Prism software (Graph Pad Software) was used to analyze the quantitative PCRs, directed in vivo angiogenesis assay, adhesion molecule induction and tumor infiltration data, and to determine statistical significance of the differences between groups by applying unpaired Student's t tests or 2-way ANOVA tests.P < 0.05 was considered significant.

CD137 protein is expressed on endothelial cells in mouse tumors
To ascertain whether CD137 protein was present on tumor endothelial cells, a multicolor staining strategy was performed on unicellular suspensions obtained from syngeneic tumors grafted in mice.Supplementary Figure S1 shows the strategy to gate CD45 À CD31 þ CD34 þ .When those cells from CT26 colon carcinomas were costained with anti-CD137 mAb, surface expression was indeed observed.Such costaining was preserved when tumor cells were grafted in T-and B-celldeficient immunodeficient Rag À/À mice (Fig. 1A).To ensure immunostaining specificity, mice whose CD137 gene had been silenced were used (Supplementary Fig. S2).In this case, CD45 À CD34 þ CD31 þ endothelial cells did not show reactivity to anti-CD137 mAb on multicolor FACS staining.
Moreover, when such CD45 À CD34 þ CD31 þ endothelial cells were sorted by FACS, mRNA encoding for CD137 was detected by quantitative RT-PCR (Fig. 1B).Interestingly, the 2 differentially spliced transcripts were detected in tumor endothelial cells (Fig. 1B) using primers that amplify both cDNAs as they do in spleen T cells activated with concanavalin-A (Fig. 1B).This is of interest because the shorter transcript has been found to encode for a soluble form of CD137, whose function is unclear even in the case of T lymphocytes (41).RT-PCR bands were thoroughly sequenced to confirm CD137 identity (data not shown).
To rule out the possibility that this was a feature exclusive to CT26-derived tumors implanted in BALB/c mice, we conducted similar experiments with grafts of LLC and EMT6 breast carcinoma implanted in C57BL6 and BALB/c mice respectively.In these tumors, specific immunostaining was also documented with anti-CD137 mAb.(Fig. 1C).Moreover, expression of CD137 was found on the CD45 À CD31 þ CD34 þ cells from spontaneous breast carcinomas (Fig. 1C) arising in MMTV-HER-2 transgenic mice (34).Therefore, the expression of CD137 on endothelium is not an exclusive characteristic of transplanted tumors.
In surgically excised CT26 tumors, it was possible to document by tissue immunofluorescence under confocal microscopy that most CD31 þ cells in vascular structures are costained with anti-CD137 further confirming the presence of CD137 in tumor-associated vascular endothelium (Fig. 1D).

Hypoxia induces CD137 on mouse endothelial cells
The expression of CD137 on mouse endothelial cells from tumors raised the question of which microenvironmental factor would drive CD137 expression.Tumors constitute a hypoxic tissue as we observe in CT26 established tumors using a PET technique that visualizes hypoxia as a result of FMISO retention (Fig. 2A).We found that the endothelioma cell line PY-4-1 (35) does not express surface CD137.However, if these cells are cultured for 24 hours in a 1% O 2 hypoxia chamber, immunostaining with anti-CD137 mAbs becomes apparent (Fig. 2B).PY-4-1 is derived from hemangiomas arising in SV40 T transgenic mice.To check whether this could be seen in other mouse endothelial cells, a cell line immortalized in culture from coronary endothelium by SV40 T Ag transfection (42) was used rendering similar results (Fig. 2C).Induction of CD137 expression took place at the transcriptional level as it could be observed in quantitative RT-PCRs (Fig. 2D).
However, it could be argued that CD137 upregulation in response to hypoxia might be caused by the transformed nature of the endothelial cell lines used in Figure 2. To exclude this interpretation, experiments were done on short-passaged primary endothelial cells isolated from mouse lungs that showed upregulation of surface CD137 (Fig. 3A) and CD137 mRNA (Fig. 3B) in response to hypoxia.

Lack of antiangiogenic effects of anti-CD137 mAb
Our findings on selectively CD137 expression on tumor endothelium suggested that perhaps part of the therapeutic effects of anti-CD137 mAb could be attributed to antiangiogenesis or other effects on tumor vascularization.
However, when mice were implanted with mm-sized tubes filled with CT26 cells embedded in Matrigel, anti-CD137 treatment showed a degree of vascularization similar to that found with control antibody.These experiments shown in Supplementary Figure S3 were done in Rag À/À mice to rule out any indirect effects mediated by CD8 þ T cells.The method for vascularization analysis was based on measuring an intravascular fluorescent dye inside the excised Matrigel tubes.The amount of vascularization of the Matrigel implants was similar in CD137 À/À mice indicating a nondefective vascularization, that indeed could be suppressed by the well-described antiangiogenic drug 77427 (40) used as a positive control.

CD137 ligation on endothelial cells increases expression of lymphocyte-homing adhesion molecules
An attractive hypothesis was that if CD137 on endothelial cells were coupled to the same signal transduction machinery as in T-cells, this could drive proinflammatory changes.Indeed, when PY-4-1 endothelioma cells were cultured under hypoxia and treated with 4.5-mm diameter microbeads coated with the agonist anti-CD137 mAb 1D8, expression of E-selectin, VCAM-1, and ICAM-1 were upregulated (Fig. 4A; Supplementary Fig. S4).Agonist anti-CD137 antibody-coated microbeads were chosen to mimic cells expressing CD137L.
If these mechanisms occurred in vivo, the expression of adhesion molecules could facilitate infiltration of T lymphocytes into tissues.To verify whether this mechanism was actually at work in vivo, experiments in BALB/c mice grafted with CT26 tumors were done.Accordingly, mice with 10-mm mean diameter tumors were treated with 1D8 anti-CD137 mAb or control rat IgG.Both antibodies were devoid of LPS.As can be seen in Figure 4B, the immunostaining in CD45 À CD34 þ CD31 þ cells for ICAM-1, VCAM-1, and E-selectin dramatically increased.To exclude events related to T cells and to ensure that the observations were due to CD137 ligation, experiments were carried out in Rag À/À and CD137 À/À mice. Figure 4C shows that no upregulation took place in CD137 À/À mice, but upregulation was readily seen in T/B-cell-deficient Rag À/À mice.
Confocal microscopy examination of tumor tissue samples from Rag À/À mice treated with anti-CD137 mAb showed ICAM-1 and VCAM-specific immunofluorescence associated to capillary walls (Fig. 4D).
Our interpretation is that ligation of CD137 on mouse tumor endothelial cells gives rise to signals that enhance the expression of lymphocyte homing molecules.

Treatment with anti-CD137 mAb increases T-cell homing into tumors as a result of a direct stimulation of endothelial cells
The next question was to ascertain whether CD137 upregulation of homing receptors on endothelial cells actually gave rise to enhanced intratumoral infiltration of activated T lymphocytes.
For this purpose, as depicted in Figure 4A, Con-A T-cell blasts were prepared from the spleen of CD137 À/À syngenic mice (Supplementary Fig. S2) and labeled with CFSE.Rag À/À mice harboring CT26 tumors were used as recipients for the CD137 À/À Con-A T-cell blasts.Mice had been pretreated with anti-CD137 mAb or control antibody as presented in the scheme (Fig. 5A).
As seen in Figure 5B, many more fluorescent cells were recovered from unicellular suspensions of the CT26 tumors 2 days later on treatment with anti-CD137 mAb.Given that the antitumor immunotherapeutic effects of anti-CD137 mAb are mostly attributed to CD8 T cells, it was also important to verify that CD8 þ CFSE þ T cells entered more avidly into the tumor tissue as shown in Figure 5C.Similar data have been observed in LLC tumors transfected to express ovalbumin when transferring anti-OVA TCR transgenic CD8 þ T cells (Supplementary Fig. S5) indicating that tumor antigen-specific T cells can also enter more efficiently to tumors under treatment with anti-CD137 mAb.Other groups have also reported enhanced entrance of tumor antigen-specific T cells on anti-CD137 mAb treatment (43) In the experiments shown in Figure 5B and C, we observed that pretreatment of the mice with a mixture of anti-ICAM-1 and anti-VCAM-1 blocking mAbs almost completely blocked the entrance of CD137 À/À fluorescent T-cell blasts.

Discussion
Traffic of T cells into tumors is a limiting factor for the efficacy of immunotherapy that can spoil the results of cancer vaccines and adoptive T-cell therapies even in the cases when tangible cellular immune responses are present in peripheral blood (44).Local proinflammatory cytokines (45) including TNFa, IFNg, interleukin (IL)-1, and CD40 ligation (46) enhance the expression of homing receptors for activated T lymphocytes on endothelial cells.In this study, we found that CD137 protein and mRNA are functionally expressed by endothelial cells harvested from mouse solid tumors.Importantly, CD137 ligation on tumor endothelial cells triggers proinflammatory changes in the tumor vasculature that recruit T cells toward malignant tissue.

A B
Figure 3. Hypoxia upregulates CD137 in mouse primary endothelial cells from the lung.Primary endothelial cells were isolated from disaggregated mouse lung tissue and cultured for 3 passages before being subjected or not to 1% hypoxia for 48 hours.Surface expression of CD137 (A) and mRNA encoding CD137 (B) were measured as in Figure 2. Previous reports had advanced evidence for the expression of CD137 on vascular cells.This includes human tumor vessels (10) and inflamed atherosclerotic lesions (11,47).Immunohistochemical evidence for CD137 expression in capillaries from a number of human malignant tumors had been reported, although benign neoplasms tended to show no anti-CD137 immunostaining in vessels (10).In mice, CD137encoding mRNA has been found in mouse tumor endothelial cells purified by multistep immunomagnetic procedures from tumor vessels, but not from angiogenic vascular cells obtained from regenerating liver (9).In this study, we provide evidence for functional protein expression at the tumor endothelial cell surface.Expression of CD137 on mouse tumor vessels seems to be a common feature that includes a variety of transplantable tumors and spontaneous carcinomas arising in oncogene transgenic mice.
The tumor microenviroment is known to be hypoxic, as we confirm by PET imaging (39).Hypoxia controls many phenotypic features of endothelial cells.Thus, it could be hypothesized that CD137 expression on endothelial cells could be driven by hypoxia.We have observed that mouse endothelial cells subjected to hypoxia expressed surface CD137, whereas it was absent in normoxic conditions of culture.When analyzing the locus of mouse CD137 (GenBank accession number NC_000070), we have identified several putative hypoxia response elements.The focus of our currently ongoing research (A.Palazon and colleagues, manuscript in preparation) is to specifically explore the role of hypoxia in controlling CD137 expression in immune and nonimmune cells.The fact that CD137 is not expressed in a normoxic atmosphere indicates that CD137 would be selectively upregulated in tissues suffering lack of oxygen supply such as the tumor microenvironment, healing wounds, ischemia-atheroscerosis conditions.Therefore, the importance of these findings goes beyond tumor immunotherapy.The nature of the CD137L þ cells that would naturally stimulate CD137 on the hypoxic endothelial cells remains elusive, but in atherosclerotic lesions has been proposed to be a subset of activated macrophages (47).
Two important conclusions can be drawn from selective expression in tumor vessels: (i) the tumor endothelium would offer CD137 as a selective marker for targeted therapies, and (ii) selective tissue distribution would confine the vascular effects of CD137-stimulation inside the malignancy, thereby avoiding systemic side effects.
Our series of angiogenesis assessments concluded that there were no detectable antiangiogenic effects in vivo by the anti-CD137 mAb.This was consistent with previous results measuring angiogenesis in nude mice implanted with Matrigel plugs containing VEGF and b-FGF (data not shown).In  addition, anti-CD137 mAb have never induced any tumor shrinkage when tumors are implanted in T-cell-deficient mice.For this reason, our angiogenesis experiments have been carried out in Rag À/À to avoid any antivascularization effects indirectly mediated by lymphocytes whose function was going to be otherwise enhanced via CD137.According to these negative results, the primary therapeutic effects of anti-CD137 mAb cannot be attributed to downregulation of angiogenesis or vasculogenesis.What we have observed is that CD137 cross-linking at the surface of endothelial cells gives rise to upregulated expression of VCAM-1, ICAM-1, and E-selectin.These observations are consistent with previous findings in atherosclerotic lesions in mice on treatment with anti-CD137 mAb (11).These homing molecules acting in a concerted fashion with chemokines are key mediators in the recruitment of an inflammatory infiltrate (33).Primed lymphocytes bear activated LFA-1 and VLA-4 integrins that would avidly interact with their counter-receptors, thereby to mediate extravasation.Contrary to what happens with CD40 or proinflammatory cytokine receptors, CD137 is selectively expressed on tumor vasculature and thus the observed proinflammatory changes ought to take place selectively at the tumor lesion.However, these proadhesive effects of agonist anti-CD137 mAb can also occur in atherosclerotic lesions (11) and therefore attention must be paid to cardiovascular events in clinical trials, particularly when considering patients with vascular risk factors.
The ICAM-1 and VCAM-mediated enhanced entrance of lymphocytes into tumors could define an additional antitumor mechanism of action of the immunostimulatory anti-CD137 mAb.Such a mechanism would not be alternative, but complementary to the costimulatory effects of the anti-CD137 agonist antibodies on T cells.The vascular effects could be considered an independent mechanism of action, as we observed migration of CD137 À/À activated lymphocytes that are unresponsive to the agonist anti-CD137 mAb.
We had found a similar antitumoral mechanism of action depending on VCAM upregulation by confining IL-12 expression to tumor nodules by means of gene therapy (48,49).Because agonist antibodies directed to OX40, CD27, and GITR exert immunotherapeutic effects (17), it would be of much interest to ascertain whether any of these members of the TNF-receptor family with homology with CD137 are also functionally expressed, at least in some cases, on tumor endothelial cells.If this is the case, agonist mAb against such molecules could share to some extent this mechanism of action.
In the case of CD137-targeted therapies, a multilayered mechanism of action emerges in which actions on immune cells and endothelium could synergistically contribute to the therapeutic effects.We postulate that vascular and immune mechanisms may act in a mutually cooperative manner, as the most intensely activated lymphocytes will be selectively recruited by adhesion molecules elicited at the lumen of blood vessels irrigating tumor lesions.In turn, the growing T-cell infiltrate costimulated via CD137 would further enhance local inflammation by more intensely producing chemokines and proinflammatory cytokines.In this regard, the effects on tumor endothelium contribute to explain the observed synergies of anti-CD137 mAb treatment with cancer vaccines (17,18) and adoptive transfers of activated T cells (19,50).
Our data are very much in agreement with recent elegant experiments by the group of L. Chen (50) with mouse bone marrow chimeras that advocate an involvement of a CD137expressing nonhematopoietic cell component in the therapeutic mechanism of action of anti-CD137 mAb against B16-OVA derived melanomas.Moreover, we have observed that in a synergistic combination of adoptive transfer of activated tumor specific CD8 T cells plus anti-CD137 mAb treatment, the synergistic curative effects are not observed if the host animal is CD137 À/À indicating that other recipient cells expression CD137 beyond the transferred T cells are absolutely required (Marinez-Forero and colleagues, manuscript in preparation).Further research will establish which lymphocyte subsets are recruited to the tumor microenvironment on CD137 stimulation of tumor endothelium.It might be that not only effector lymphocytes, but also regulatory T cells could be invited to enter the tumor.If this is the case, elimination of Treg cells concomitantly to CD137 therapeutic stimulation may be advisable as recently observed by the group of Ronald Levy (51).
Taking collectively, these data point to a model in which the tumor vascular response to anti-CD137 agonist mAb can be considered an additional and important element in the multilayered mechanisms of action against tumors elicited by these promising immunotherapeutic agents (3).

Figure 2 .
Figure 2. Hypoxia induces CD137 expression in mouse endothelial cells.A, 18 FMISO microPET imaging of mice harboring established CT26 tumors in the right flank.Tumors are gated and SUV in the gated regions are provided under the corresponding pictures.T indicates tumor, B bladder, L liver, and G gut. B, PY-4-1 endothelioma cells were cultured for 24 hours in normoxia and 1% O 2 atmosphere.CD137 surface expression was studied using direct immunofluorescence in normoxic or hypoxic cells as indicated in the figure.Shaded histograms represent staining with isotype-matched control antibody.Experiments are representative of 3 similarly conducted.C, MCoEC cells derived from coronary endothelium immortalized by stable transfection of SV40T were cultured and studied for CD137 as in A. This experiment was repeated twice rendering similar results.D, quantitative RT-PCR studied on RNA extracted from PY-4-1 cells cultured under 1% O 2 hypoxia for the indicated periods of time.

Figure 4 .
Figure 4. CD137 ligation on tumor endothelial cells increases expression of lymphocyte homing receptors.A, PY-4-1 cells cultured on hypoxia for 36 hours and exposed to 4.5-mm diameter latex beads coated with anti-CD137 1D8 mAb (continuous line histograms) or control antibody (dotted histograms) at a 1:10 bead/cell ratio.After culture, cells were studied by FACS for expression E-selectin and VCAM-1 as indicated in the figure.Shaded histograms represent negative control immunostaining with isotype matched irrelevant mAbs.Data are representative of 3 experiments.Mean AE SD of these independent experiments is presented as Supplementary Figure S4.B, BALB/c mice bearing CT26 tumors (8-12 mm in diameter) were treated with RatIgG or aCD137 1D8 mAb.Tumors were excised 36 hours later and cell suspensions were analyzed for the indicated markers on CD45 À CD34 þ CD31 þ -gated endothelial cells.Results represent the mean intensity of immunofluorescence AE SD in 3 experiments with 3 mice per group.C, experiments as in B but done in the indicated strains of mice (Rag À/À and CD137 À/À as indicated).Data represent the mean intensity of immunofluorescence AE SD in 3 experiments with 3 mice per group.D, tissue immunofluorescence experiments visualized by confocal microscopy showing colocalization of CD31 with ICAM-1 or VCAM-1 in CT-26 tumors implanted in syngeneic Rag À/À mice and treated 1 day before with 100 mg of anti-CD137 mAb.

Figure 5 .
Figure 5. Treatment with anti-CD137 agonist mAb induces Tcell homing into tumor lesions.A, schematic representation of the experimental procedures.Rag À/À mice were grafted with CT26 tumors and treated with 100 mg of RatIgG or aCD137 mAb 1D8.One day later, some of the mice were i. p. injected with a mixture of anti-ICAM-1 and anti-VCAM-1 neutralizing mAb.Twenty-four hours later, all mice received 10 7 T-cell blasts activated with 10 mg/mL of Con-A.Con-A blasts were derived from the spleen of syngenic CD137 À/À mice and were labeled with CFSE prior to i.v.adoptive transfer.FACS analysis of tumor infiltrating lymphocytes (TIL) was done 2 days after adoptive transfer.B, quantitation of CFSE þ events in cell suspensions obtained from tumors treated as indicated in A. Data represent mean AE SD from an experiment out of 3 rendering comparable results.C, CD8 þgated CFSE þ events analyzed by FACS multicolor analysis from the same samples as in B.

M
. Jure-Kunkel and J. Dinchuk are full-time employees of Bristol-Myers Squibb.I. Melero has received consultancy honoraria and reagents from Bristol-Myers Squibb.