D2001 Nature Publishing Group 0929-1903/01/$17.00/+0 www.nature.com/cgt

Autocrine expression of both endostatin and green fluorescent provides a synergistic antitumor effect in a murine neuroblastoma model Andrew M. Davidoff,1 Margaret A. Leary,2 Catherine Y.C. Ng,1 William W. Spurbeck,1 Pascale Frare,2 Marc Vanhove,2 Arthur W. Nienhuis,2 and Elio F. Vanin2

1Department of Surgery, St. Jude Children's Research Hospital and University of Tennessee College of Medicine, Memphis, Tennessee; and 2Department of Hematology/Oncology, St. Jude Children's Research Hospital, Memphis, Tennessee.

Modalities that act through different mechanisms can often provide synergistic antitumor activity for the treatment of refractory tumors when used in combination. Here we report a gene therapy approach in which the genes for the inhibitor, endostatin, and the marker protein and potent immunogen, green fluorescent protein (GFP), were delivered to murine neuroblastoma cells prior to inoculation of the tumor cells into syngeneic immunocompetent mice. Although the effect of either angiogenesis inhibition or immunomodulation alone resulted in only a modest delay in tumor growth, when these approaches were used in combination, prevention of the formation of appreciable tumors was effected in 15 of 24 (63%) mice. The combination of endostatin and GFP expression elicited a strong immune response that was T cell±mediated and was reactive against both GFP and tumor cell line±specific antigens. This afforded treated mice protection against subsequent tumor challenge with unmodified tumor cells. These results suggest that antiangiogenic and immunotherapy strategies, when used in a gene therapy±mediated approach, can act synergistically in an effective multimodality anticancer approach. Cancer Gene Therapy (2001) 8, 537±545

Key words: Gene therapy; immunotherapy; antiangiogenesis; endostatin; green fluorescent protein.

ngiogenesis appears to be a fundamental requirement for An alternative to chronic angiogenesis inhibition would be Atumor growth, invasion, and metastasis.1±3 Evidence to combine this approach with a cytotoxic modality. also exists to suggest that inhibition of tumor-associated Immunotherapy may be one such complementary tumoricidal angiogenesis can retard tumor growth, prevent spread, and approach. It has been used successfully in preclinical models even cause tumor regression.4±6 Several naturally occurring for the eradication of minimal residual disease with long-term inhibitors of angiogenesis have been identified including antitumor immune surveillance resulting in chronic tumor endostatin, a 20 kDa C-terminal fragment of XVIII.7 suppression.13 ± 15 However, rarely does it induce rejection of Endostatin has been shown to inhibit endothelial cell large tumors. The generation of an antitumor immune proliferation and migration in vitro, at least in part, through response has often been successfully established through a its ability to induce endothelial cell apoptosis. 8 In addition, it gene therapy±mediated approach in which tumor cells are has been shown to be among the most potent inhibitors of modified to express foreign antigens that elicit a strong tumor-induced angiogenesis in vivo and, consequently, to immune response.16,17 Immunity to unmodified (parental) inhibit tumor growth in a variety of different murine tumor cells may be established through ``crosspriming'' to models.7,9 ± 11 Antiangiogenic therapy can even reduce an tumor-specific antigens. One potential immunogenic foreign existing tumor burden by limiting the necessary blood supply. antigen is the jellyfish green fluorescent protein (GFP). This However, tumors may remain dormant but viable with this commonly used reporter molecule has been shown to be cytostatic treatment approach and may resume unregulated immunogenic in murine tumor models, inducing a classic T growth once the limitations of angiogenesis inhibition are cell±mediated cytotoxic response.18 removed.11,12 Therefore, long-term expression of an angio- In this set of experiments, murine neuroblastoma cells genesis inhibitor such as endostatin is likely to be required for were modified, using a retroviral vector, to express both the successful treatment of cancer. endostatin for local angiogenesis inhibition, together with GFP, which has the potential for immune stimulation. Growth of these gene-modified tumor cells was then Received May 16, 2001. observed both in vitro and in vivo. These two complementary Address correspondence to Dr. Andrew M. Davidoff, Department of approaches, acting through different mechanisms, effected a Surgery, St. Jude Children's Research Hospital, 332 North Lauderdale, synergistic anticancer response when used in combination in Memphis, TN 38105. E-mail address: [email protected] a murine model of neuroblastoma.

Cancer Gene Therapy, Vol 8, No 7, 2001: pp 537±545 537 538 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP

MATERIALS ANDMETHODS were added to the 50 end of the endostatin cDNA using a complimentary oligonucleotide approach requiring two Cell lines steps. Before the addition of oligonucleotides, the amplified The murine neuroblastoma cell line NXS2, provided by Dr. endostatin was cloned into pSL1180 (Pharmacia, Piscat- R. Reisfeld (La Jolla, CA), was maintained in RPMI-1640 away, NJ) by ligating the 625-bp SpeI±EcoRV fragment medium (Bio-Whittaker, Walkersville, MD) supplemented from pCR2.1-Endo into pSL1180 cut with NheI and EcoRV with 10% fetal bovine serum (Bio-Whittaker), 100 units/ to give 80-30Endo. mL penicillin±100 g/mL streptomycin (Gibco BRL, For the endostatin containing signal peptide 1, the Grand Island, NY) and 2 mM L -glutamine (Gibco BRL). oligonucleotides (shown below) were annealed and ligated NXS2 was derived by hybridizing the C1300 murine (A/J into 80-30Endo cut with PflMI and MluI: strain) neuroblastoma-derived cell line with murine dorsal root ganglion cells from C57BL/6J mice and selecting for 50 -CGCGTCTCCTGGGCAGAGCCACACCAT- high dopamine and GD-2 expression.19 This clone has been CACCACCACCA TATGCATACTCATCAG- shown to have major histocompatibility class I antigen GACTTTCAGCCAGTGCTCCACCTG-30, and expression of the A/J mice, i.e., H2Kk -positive/H2Kb - 50 -GTGGAGCACTGGCTGAAAGTCCTGAT- negative.19 293T cells20 were maintained in Dulbecco's GAGTATGCATAT GGTGGTGGTGATGGT- modified Eagle's medium (DMEM; Bio-Whittaker) with GTGGCTCTGCCCAGGAGA-30. the same supplements as in the growth medium for NXS2. Human umbilical vein endothelial cells (HUVECs) were The resulting plasmid, 80-30EndoSP1, was then digested obtained from Clonetics (Walkersville, MD) and were with XhoI and MluI, and the second set of oligonucleotides maintained in endothelial growth medium (EGM). (shown below) was annealed and ligated to the digested plasmids: Antibodies 5 0 -TCGAGCCTGGGGAGATGGCGCC- Recombinant endostatin was generated from Escherichia CAGGTGGCACCTCCTGG ATGTGCTCAC- coli as described by Boehm et al11 using a bacterial CAGTTTGGTCTTGCTGCTGGTGGCA-30, and expression plasmid for the murine protein obtained from 5 0 -CGCGTGCCACCAGCAGCAAGAC- Dr. J. Folkman (Boston, MA). Goat polyclonal antiserum to CAAACTGGTGAGCACAT CCAGGAGGTGC- murine endostatin was then raised (Panigen, Blanchardville, CACCTGGGCGCCATCTCCCCAGGC-30. WI). Three monthly injections of 1 mg of the recombinant protein, the first of which was in Freund's complete adjuvant The resultant plasmid was termed 80-EndoSP1. while the latter were in incomplete adjuvant, were For the endostatin containing signal peptide 2, the administered by subcutaneous injection to a goat. The serum oligonucleotides (shown below) were annealed and ligated subsequently harvested has demonstrated reactivity with into 80-30Endo, which had been cut with NarI and PflMI to murine endostatin in Western blot analysis. give 80-30EndoSP2:

Cloning and expression of mouse endostatin 50 -CGCCTTGTGCCTCACCATCACCACCACCA- The cDNA for mouse endostatin corresponding to the C- TATGCATACTC AT CAGGACTTTCAGC- terminal 184 amino acids of mouse collagen XVIII was CAGTGCTCCACCTG-30, and isolated by polymerase chain reaction amplification of 50 -GTGGAGCACTGGCTGAAAGTCCTGAT- cDNA from mouse liver, a site known to express high levels GAGTATGCATATG GTGGTGGTGATGGT- of collagen XVIII. The 50 primer (50 -TGCTCATACTCAT- GAGGCACAAGG-30. CAGGACTTTCAGCC-30 ) spanned the sequence encoding the N-terminal amino acids of endostatin, whereas the 30 The DNA encoding the 50 end of signal peptide 2 was then primer spanned the termination codon (50 -AGCTGGCA- inserted into 80-30EndoSP2 by ligating the annealed GAGGCCTATTTGGAGAAAG-30 ). The amplified poly- oligonucleotides (shown below) to the 1.6-kb NarI±NspV merase chain reaction products were then cloned into fragment from 80-30EndoSP2 and the 3.1-kb XhoI±NspV pCR2.1 (Invitrogen, Carlsbad, CA) and the insert was fragment from pSL1180: sequenced. A signal peptide was then added to ensure that the 5 0 -TCGAGCAATTGACTGCCCT- endostatin would be secreted. Mouse collagen XVIII has GATGGCTCCCGACCCCAGCA been shown to have two potential signal peptides.21 Peptide GACGCCTCTGCCTGCTGCTGCTGTTGCTGC- 1 is MAERWHLLDVLTSLVLLLVARVSWA-EP and pep- TCTCCTGC-30,and tide 2 is MAPDPSRRLCLLL-LLLLSCRL-VP. Two amino 50 -CGGCAGGAGAGCAGCAACAGCAGCAG- acids following the signal peptide cleavage site (EP for CAGGCAGAGGCGTCTGCTGGGGTCGG- signal peptide 1 and VP for signal peptide 2) were included GAGCCATCAGGGCAGTCAATTGC-30. in order to ensure that efficient cleavage would occur. In addition, a six-histidine tag was incorporated immediately 30 This plasmid was termed 80-EndoSP2. to the signal peptide in order to facilitate detection and The XhoI±ClaI fragments from 80-EndoSP1 and 80- purification. Both the six-histidine tag and a signal peptide EndoSP2 were then cloned into pSP73 (Promega, Madison,

Cancer Gene Therapy, Vol 8, No 7, 2001 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP 539

WI) digested with XhoI and ClaI to give 73-EndoSP1 and Endostatin cDNA) and BamHI (site 30 to the Endostatin 73-EndoSP2, respectively. In order to make CMV- cDNA) to give 80-30Endo(P): EndoSP1-pA, two separate ligation steps were required. First, the BamHI±EcoRV fragment from pAVs6,22 which 5 0 -GCTTCATGACCTCTTTCCCCAAA- contains the SV40 polyadenylation site, was ligated to 80- TAGGCCTCTGCCAGCTC G-30, and EndoSP1 digested with EcoRV and BamHI to give 80- 50 -GATCCGAGCTGGCAGAGGCCTATTTGGG- EndoSP1-pA. Then the CMV promoter from pCEP4 GAAAGAGGTCATGAAGC-30. (Invitrogen) was cloned 50 to 80-EndoSP1-pA using a three-way ligation among the BstBI±XhoI fragment from The 30 end of this mutated Endostatin was used to replace pSL1180 (the backbone of the plasmid), the BstBI±PmaCI the 30 end of the Endostatin within the MSCV-Endo-I-GFP fragment from 80-EndoSP1-pA (contains the EndoSP1 by ligating the 237-bp SapI±BamHI fragment from 80- cDNA and SV40 polyadenylation site), and the SalI±PvuII 30Endo(P) into MSCV-Endo-I-GFP, which had been fragment from pCEP4 (contains the CMV promoter). CMV- partially digested with SapI and then to completion with EndoSP2-pA was then made by replacing the XhoI±EcoRV BamHI. fragment from CMV-EndoSP1-pA, which contains the EndoSP1 cDNA, with the XhoI±EcoRV fragment from 80- Tumor cell line transduction EndoSP2, which contains the EndoSP2 cDNA. NXS2 tumor cells were transduced with conditioned medium containing high-titer, VSV-G (vesicular stomatitis Transcription/translation/translocation assay virus) ±pseudotyped MSCV-Endo, MSCV-Endo-I-GFP, Evaluation of in vitro transcription, translation, and MSCV-Endo(P) -I-GFP, or MSCV-I-GFP vector particles, translocation was performed using the TNT T7 coupled supplemented with 6 g/mL polybrene. These conditioned reticulocyte lysate system kit according to Promega media were derived by cotransfection of 293T cells with the Technical Bulletin 126. Endo1 and Endo2 were respective retroviral vector plasmids and two helper synthesized in the presence of [ 35S]methionine (1200 plasmids, one containing the gag and pol retroviral genes Ci/mmol; Amersham, Piscataway, NJ) in the absence or and the other containing the gene for the VSV-G envelope, in the presence of canine pancreatic microsomal mem- as described previously.26,27 GFP-expressing transduced branes (Promega) according to the manufacturer's cells were then sorted based on GFP expression by FACS instructions. After 90 minutes, the proteins were resolved (Becton-Dickinson, Bedford, MA) to recover highly on SDS polyacrylamide gel under reducing conditions and transduced cells. These subpopulations were then expanded, visualized by fluorography using the Amplify reagent with GFP expression being confirmed by FACS and (Amersham). endostatin expression being confirmed by western and ELISA (R&D Systems, Minneapolis, MN). Retroviral vector construction MSCV-Endo, a retroviral vector containing the endostatin Inhibition of HUVEC migration expression cassette, was made by ligating the 741-bp XhoI± Migration assays were performed in 6.5-mm tissue MunI fragment from 80-EndoSP1 into MSCV-X, a MSCV- culture±treated transwell plates (Costar, Corning, NY). based vector that contains part of the polylinker from Eight-micron polycarbonate membranes were placed over a pSL1180, digested with XhoI and MunI. MSCV-Endo-I- bottom chamber containing 2% that had been GFP was made by ligating the 1463-bp SpeI±BsaBI allowed to settle at room temperature for 30 minutes. fragment from MSCV-Endo into MSCV-I-GFP(M) 23 cut Excess gelatin was then aspirated and replaced with an with SpeI and StuI. In these constructs, the expression of the equal volume of DMEM+0.1% BSA and incubated for 1 endostatin cassette, containing the signal peptide and six-His hour at 378C. Excess DMEM was aspirated from the tag, is under the control of the MSCV LTR (long terminal bottom wells and replaced with EGM supplemented with repeat), and in the latter vector, the GFP is linked to an VEGF (10 ng/mL; R&D Systems) along with either internal ribosome entry site from encephalomyocarditis virus endostatin or the proline-mutant endostatin (10 ng/mL), or 30 of the endostatin cDNA.24 an equal volume of PBS. HUVECs (passage 5) were An additional retroviral vector was constructed in which trypsinized and resuspended at 1Â106 cells/mL. The 5Â104 the penultimate amino acid of the endostatin protein being HUVECs were then added to each well in the upper expressed was proline rather than serine, as occurs in the chamber after preincubating the cells with either endostatin wild-type protein. This was done in an attempt to create a or the proline-mutant endostatin (10 ng/mL) for 30 more stable endostatin protein that might be resistant to minutes at 378C. Endostatin and the proline-mutant endo- carboxypeptidase hydrolysis.25 This MSCV-Endo(P) -I- statin were purified from NXS2-Endo-I-GFP and NXS2- GFP vector was made by cloning a 715-bp SphI (site within Endo(P) -I-GFP conditioned media using Ni2+ ±NTA the Endostatin cDNA) ±ClaI (site 30 to the Endostatin beads as described (QiaExpressionist Handbook; Qiagen, cDNA) fragment into pSL1180 to give 80-30Endo. The Santa Clarita, CA). The plates were then incubated for 6 mutation (serine to proline) was then introduced by hours at 378C with 5% CO2 to allow the cells to migrate. annealing the two complimentary oligonucleotides shown Nonmigrated cells on the upper surface of the filter were below and ligating the annealed product into 80-30Endo, removed by wiping with a cotton swab. The cells were then which had been partially digested with XmnI (site within the fixed with 10% formalin, stained with Harris' hematoxylin,

Cancer Gene Therapy, Vol 8, No 7, 2001 540 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP

Figure 1. Processing of endostatin with two different signal peptides in vitro. (A) Endostatin synthesized from Endo-SP1 in the absence (1) and presence (2) of canine microsomes and from Endo-SP2 in the absence (3) and presence (4) of canine microsomes. (B) Western blot analysis of supernatants from 293T cells transfected with plasmids encoding en- dostatin and the two different signal peptides (lane 1, recombinant endosta- tin; lane 2, medium control; lane 3, CMV- Endo2; lane 4, CMV-Endo-1).

and after washing, the migrated cells were counted. The antibody was administered by intraperitoneal injection 4 assays were run in triplicate. days prior to tumor cell challenge and then every other day for the duration of the experiment. After tumor-bearing mice Murine tumor models were sacrificed, successful depletion of approximately 99% Tumors were established in either syngeneic, immunocom- of T cells was confirmed by FACS analysis of splenocytes petent A/J mice (Jackson Laboratory, Bar Harbor, ME) or (not shown). C.B-17 SCID (Charles River Laboratory, Wilmington, MA) by subcutaneous injection of 2Â106 cells in 200 L Statistical analysis PBS. Tumor measurements were performed in two dimen- Results are reported as means‹S.E.M. for tumor volumes. A sions with calipers twice weekly and volumes calculated as Student's t test was used to analyze statistical differences width2ÂlengthÂ0.5. among groups in the HUVEC migration assays and between tumor growth curves. A P value less than .05 was considered Cytotoxicity assays to be statistically significant. Spleens were harvested from mice that had been challenged with either NXS2, NXS2-Endo-I-GFP, or NXS2-I-GFP RESULTS cells. These spleens were then minced into single cell suspensions in PBS plus 2% fetal calf serum. Splenocytes Endostatin processing were resuspended in ammonium chloride for 5 minutes at The ability of the two potential signal peptides to be cleaved room temperature, after which they were washed and again at the membrane was first tested by translating in vitro± resuspended in PBS plus serum. Splenocytes were then transcribed RNA in the presence of canine microsomes. The cocultured for 6days with irradiated tumor cells, either results shown in Figure 1A indicate that the EndoSP1 NXS2-Endo-I-GFP or NXS2-I-GFP. Following this construct was efficiently cleaved by the microsomes, coculture, the splenocytes were recovered and incubated with 51 resulting in a smaller polypeptide. The radioactive poly- Cr-labelled, unmodified NXS2, NXS2-Endo-I-GFP, or peptides both in the presence and absence of microsomes for NXS2-I-GFP target cells at effector:target ratios of 40:1, EndoSP1 are of the expected molecular weight. On the other 20:1, 10:1, and 5:1 in a V-bottomed microtiter plate (Costar). hand, the EndoSP2 construct was not cleaved. The minor One set of reactions was performed in the presence of anti- kb bands in both lanes are due to translational starts at internal CD3 antibody (145-2c11, H-2 ) from PharMingen, San methionine. Diego, CA. After 4 hours of incubation at 378C, the This inability of signal peptide 2 to be cleaved in vitro is supernatant was harvested and radioactive release was also consistent with the results obtained when 293T cells measured. Maximal release of the incorporated chromium were transiently transfected with CMV-EndoSP1-pA and was determined by detergent lysis of the neuroblastoma cells. CMV-EndoSP2-pA (Fig 1B). Supernatants from the cell Percent-specific cytotoxicity was determined from the cultures were analysed by Western blot analysis with a goat formula: % Specific cytotoxicity= [ ( Experimental release À polyclonal antisera. Endostatin was found to be present in the Spontaneous release) / (Maximum releaseÀSpontaneous supernatant from 293T cells transfected with CMV- release)Â100]. EndoSP1-pA, but not CMV-EndoSP2-pA, suggesting that signal peptide 2 is not functioning as a signal peptide. T-cell depletion Therefore, signal peptide 1 was used in all of the remaining Mice were depleted of T lymphocytes with anti-CD4 endostatin constructs. The recombinant endostatin shown in (Gk1.5) and anti-CD8 (2.43) monoclonal antibodies28,29 Figure 1B was synthesized in E. coli and serves as a control. kindly provided by Dr. P. Doherty (Memphis, TN). The It has eight more amino acids at the N-terminus than the antibody-containing ascites were diluted 1:5 and 1:10, endostatin from the transfected 293T cells and is, therefore, respectively, in PBS. Five hundred microliters of either slightly larger.

Cancer Gene Therapy, Vol 8, No 7, 2001 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP 541

Stable expression of murine endostatin from neuroblastoma cells Murine neuroblastoma cells engineered to express either wild-type murine endostatin or a mutated form of the endostatin protein were generated by transduction with the MSCV-Endo, MSCV-Endo-I-GFP, and MSCV-Endo(P) - I-GFP retroviral vectors. Cell populations with high degrees of transduction from the latter two vectors were selected for by FACS sorting for high levels of GFP expression. Expression of endostatin was confirmed by both Western analysis (data not shown) and ELISA. The level

Figure 3. Inhibition of VEGF-stimulated HUVEC migration in vitro by recombinant endostatin and a mutant endostatin as compared to PBS control (*P<.02; **P<.33).

of expression was approximately 1250 ng/106 cells/24 hours from the NXS2-Endo-I-GFP cells and 2000 ng/106 cells/24 hours from the NXS2-Endo(P) -I-GFP cells. Unsorted NXS2-Endo cells expressed 750 ng/106 cells/24 hours. These levels of endostatin expression by the transduced tumor cells in vitro were noted to be stable following continuous passage of cells for greater than 3 months (data not shown); no silencing of the promoter in the tumor cells or selection for nontransduced cells was observed. NXS2 cells were also transduced with the MSCV-I-GFP. FACS sorting for high levels of GFP expression was performed to select for cells with high degrees of transduction in this cell line also. Following sorting, each of these three transduced cell lines had similar mean levels of GFP expression (Fig 2) with >99% of the cells in each of the lines expressing GFP. The ability of the endostatin protein being expressed to inhibit HUVEC migration in response to VEGF was evaluated to test the functional activity of the protein in vitro. This has proven to be the most reproducible assay for us and others30 when assessing endostatin function. Wild-type and mutant endostatin were purified from gene- modified neuroblastoma-conditioned medium. The wild- type endostatin caused an 80% inhibition of VEGF-induced endothelial cell migration (P<.02), whereas the mutant form of endostatin resulted in only a small and not Figure 2. FACS analysis for GFP expression in (A) unmodified statistically significant (P<.33) inhibition in this assay NXS2 neuroblastoma cells and those transduced with (B) MSCV-I- (Fig 3). GFP (mean=552), (C) MSCV-Endo-I-GFP (mean=479), and The doubling time for neuroblastoma cells expressing (D) MSCV-Endo(P)I-GFP (mean=1465) retroviral vectors. endostatin plus GFP was compared to that for GFP alone

Cancer Gene Therapy, Vol 8, No 7, 2001 542 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP

growth of either of the other cell lines. In fact, 15 of 24 (63%) mice failed to develop any evidence of a tumor either by palpation or histologic examination (data not shown). These mice were followed for a minimum of 3 months without the subsequent development of a tumor. Tumors that did grow from NXS2-Endo-I-GFP cells had a significant delay in the initiation of appreciable tumor growth compared to that observed for the unmodified tumor cells. These tumors were not derived from nontransduced clones as they were shown to express GFP by histologic examination and endostatin by ELISA (data not shown). Additional experiments were then performed in which NXS2 cells modified to express the mutant, nonfunctional

Figure 4. Growth curve in vitro for unmodified parental NXS2 cells (±  ± ) and subpopulations modified with MSCV-Endo-I-GFP (±  ± ), MSCV-Endo(P) -I-GFP ( ± ~ ± ), and MSCV-I-GFP retroviral vectors ( ± & ±). expressing tumor cells and unmodified neuroblastoma cells. No differences in the growth curves were observed with each cell line having a doubling time of approximately 20 hours (Fig 4). Although endostatin purified from the gene- modified tumor cells had a profound effect on endothelial cell migration in vitro, there was no effect on the intrinsic growth rate of the endostatin-expressing tumor cells in vitro.

Tumor growth in mice The effect of autocrine expression of endostatin on in vivo tumor cell growth was evaluated by comparing the growth of NXS2-Endo-I-GFP tumor cells with NXS2-I-GFP tumor cells and unmodified NXS2 cells. The initial experiments were performed in SCID mice that are deficient of B and T lymphocytes in order to exclude any possible effect of anti- GFP immunity. The growth curves for the unmodified and GFP-expressing cells were identical (Fig 5A). The Endo-I- GFP±expressing cells grew at a marginally slower rate than either control and gave rise to tumors that were smaller by 25% than controls after 27 days. These experiments were then performed in syngeneic, immunocompetent A/J mice. The growth curves are shown in Figure 5B. Although there was now a slight difference observed in the growth rate of the unmodified tumor cells compared to those engineered to express GFP alone, it was Figure 5. Tumor growth in vivo of unmodified parental NXS2 cells not statistically significant. In addition, growth of other cells (±  ± ) and subpopulations modified with MSCV-Endo ( ± 5 ±), with higher GFP expression [e.g., NXS2-Endo(P) -I-GFP] MSCV-Endo-I-GFP ( ± & ± ), MSCV-Endo(P) -I-GFP ( ± 4 ±), and MSCV-I-GFP ( ±  ± ) retroviral vectors in (A) C.B-17 SCID was not slowed, suggesting that there was no effect of mice (n =13 for each group; *P <.44) and (B) syngeneic, expression of the weakly immunogenic GFP protein on immunocompetent A/J mice (n=13 for each group except the tumor growth. The NXS2-Endo-I-GFP tumor cell growth cohort that received cells modified with MSCV-Endo-I-GFP, where was significantly restricted, however, compared to the n=24; *P<.02 compared to unmodified tumor cells).

Cancer Gene Therapy, Vol 8, No 7, 2001 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP 543 endostatin plus GFP were inoculated into A/J mice. The growth of these tumor cells was not significantly different from that observed previously with the GFP-expressing tumor cells (Fig 5B), in spite of a higher mean level of GFP expression. This finding suggests that functional endostatin was required for the growth-inhibitory effect. Similarly, cells expressing endostatin alone showed no perturbation of their growth in vivo. Although expression of endostatin from these cells was lower than from the NXS2-Endo-I-GFP cells, probably because they contained no selectable marker with which highly transduced cells could be selected, this result highlights the contribution of the anti-GFP immune effect.

Analysis of the immune response to endostatin/ GFP-expressing NXS2 cells To investigate the role of T cell±mediated cytotoxicity in the inability of the NXS2-Endo-I-GFP cells to grow in the immunocompetent A/J mice, standard cytotoxicity assays were performed. Spleens were harvested from A/J mice that had failed to develop tumors by 1 month following Figure 7. Tumor growth of NXS2-Endo-I-GFP cells in immuno- subcutaneous inoculation with NXS2-Endo-I-GFP neuro- competent A /J mice ( ±  ± ) and in mice depleted of CD4 cells blastoma cells. Following coculture of these splenocytes for (± 4 ± ) or CD8 cells ( ± 5 ±). n=7 for each group. 6days with irradiated NXS2-Endo-I-GFP cells, these splenocytes showed significant killing not only of NXS2-I- Splenocytes harvested from mice that had been given GFP cells but also the unmodified parental cell line, which unmodified or GFP-expressing NXS2 cells, and subse- does not express GFP (Fig 6). These data suggest that quently grew tumors, showed no killing of target cells ``crosspriming'' of cytotoxic T cells (CTL) for a tumor- following coculture with either of these tumor cell lines. The associated antigen expressed by the unmodified tumor cell importance of T cell±mediated antitumor activity in vivo was line had occurred. This killing was blocked nearly confirmed by depleting mice of either CD4 + or CD8 + cells completely by the addition of an anti-CD3 antibody. prior to tumor cell inoculation. Depletion of either of these T cell subpopulations permitted growth of NXS2-Endo-I- GFP cells in 100% of the mice (Fig 7). One hallmark of immune-mediated tumor growth failure is the resistance of a subsequent tumor challenge. Therefore, 10 A/J mice in whom a tumor had failed to grow following initial inoculation with NXS2-Endo-I-GFP cells were rechallenged with tumor cells in the opposite flank 1 month later. Half of the mice received unmodified NXS2 cells, whereas the other half received cells expressing GFP alone. None of the 10 mice subsequently went on to develop a tumor during a follow-up period of at least 2 months. These results further support the hypothesis that immune reactivity was contributing to the prevention of growth of the NXS2- Endo-I-GFP cells in the syngeneic, immunocompetent A/J mice.

DISCUSSION

Endostatin has been shown to be among the most potent naturally occurring inhibitors of tumor-induced neovascu- larization and tumor growth.7 Administration of recombi- Figure 6. Cytotoxicity of splenocytes from A/J mice that had failed to nant protein to mice has caused regression of large primary grow NXS2-Endo-I-GFP tumor cells against unmodified ( ±  ±) tumors and prevented the spread of metastatic dis- and GFP-expressing ( ±  ± ) NXS2 cells following coculture of the ease.7,10,30 Problems have arisen, however, in the produc- splenocytes for 6 days with NXS2-Endo-I-GFP cells. Inhibition of tion of an active form of endostatin, particularly when cytotoxicity with an anti-CD3 antibody is shown for the effector:target ratio of 20:1 ( ± 4 ± ). Also shown is the lack of cytotoxic effector attempting to manufacture the large quantities of recombi- cells generated under the same conditions from splenocytes nant protein required for clinical trials. One way to harvested from mice injected with NXS2-I-GFP cells ( ± 5 ±), circumvent the nuances of protein production is to have following coculture with NXS2-I-GFP cells. the protein generated in situ by the host's own cells. This

Cancer Gene Therapy, Vol 8, No 7, 2001 544 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP can be done through a gene therapy approach in which an When endostatin was expressed together with GFP, expression cassette for endostatin is introduced into host however, a strong immune response was elicited. This was target cells from which protein can be generated. Yoon et not due to greater GFP expression in these cells as the al31 have demonstrated the efficacy of this strategy in amount of GFP expression in the NXS2-Endo-I-GFP, which tumor cells themselves were modified to express NXS2-Endo(P) -I-GFP, and NXS2-I-GFP cells was this . This approach was shown to approximately the same. The cytotoxicity assays performed result in the significant reduction of tumor-induced on splenocytes from mice that had failed to grow NXS2- neovascularization and consequent restriction of in vivo Endo-I-GFP tumors confirmed the presence of activated tumor growth (73±91% at 3 weeks). The reduction of cytotoxic T lymphocytes. These CTLs were able to kill not endostatin gene-modified tumor cell growth in immuno- only GFP-expressing tumor cell targets but also unmodified, deficient mice was more modest in our tumor model (25% parental cells in vitro. This finding suggests that cross- at 4 weeks). priming of the immune system to an antigenic determinant Although the most direct way to ensure that endostatin is inherent to the tumor cells had occurred. That this was expressed at all tumor sites is to modify the tumor cells mediated, at least in part, by a classic T cell response was themselves, this approach is ultimately not a practical one for confirmed in vitro by blocking the cytotoxicity with a T cell± patients who present with metastatic disease. Therefore, gene neutralizing antibody and in vivo by depleting mice of CD4 therapy approaches have also been tested in which and CD8 cells. In addition, the T cell±mediated response therapeutic systemic levels of endostatin were achieved. provided lasting immunity; mice that had rejected an original This strategy has had some success (40±70% tumor volume challenge with NXS2-Endo-I-GFP cells were protected reduction),32 ± 34 but in no instance was tumor growth from a subsequent challenge from both GFP-expressing completely prevented. In addition, the approaches have used NXS2 cells and unmodified NXS2 cells. either naked DNA or adenoviral vectors as delivery vehicles Thus, although neither autocrine expression of endostatin that generate only transient expression of endostatin. nor GFP alone resulted in long-term tumor suppression, a Because antiangiogenic therapy may only be cytostatic, strong synergistic effect was observed when the angio- long-term expression of this inhibitor is likely to be required genesis inhibitor and immunogen were expressed in for sustained anticancer efficacy. Although there are delivery combination. The majority of immunocompetent A/J mice systems that can direct long-term transgene expression, failed to grow tumors when inoculated with NXS2-Endo-I- concerns exist regarding chronic angiogenesis inhibition GFP cells and were protected from a subsequent tumor cell because angiogenesis is a fundamental part of many normal, challenge. The mechanism for this synergy is uncertain. It physiologic processes, such as reproduction and wound may be that the contribution of endostatin is through its healing. antiangiogenic effect on tumor neovasculature such that We have found an alternate approach in which the tumor growth is slowed sufficiently to permit a more cytostatic effects of local angiogenesis inhibition were effective immune response. Alternatively, endostatin may combined with the cytotoxic effects of an immune-mediated have some other mode of action that more directly affects anticancer strategy. It has been reported recently that the immune effector cells. It is unlikely that this synergy is due to commonly used marker protein, GFP, can be immunogenic an immune response against endostatin as the endostatin when expressed by tumor cells in murine tumor models. used in this study is a normally expressed murine protein. Stripecke et al18 found that nearly all immunocompetent, Additionally, NXS2 cells modified to express endostatin syngeneic mice challenged with certain GFP gene-modified without GFP were not rejected by immunocompetent A/J leukemia and lymphoma cell lines rejected the tumor cells mice. It is clear, however, that functional endostatin is through a T cell±mediated immune response. Although our required to achieve this synergistic effect. When a nonfunc- results differed somewhat from these reports in that the tional, mutant endostatin was expressed by the NXS2 cells, GFP-expressing NXS2 cells were not rejected, we did the growth of these tumor cells was unperturbed even though observe a slightly decreased growth rate for these modified these cells had a higher mean expression of GFP than either cells compared to unmodified NXS2 cells in immunocom- the NXS2 cells expressing functional endostatin plus GFP or petent, syngeneic A/J mice. This difference was not GFP alone. Further work is ongoing in an attempt to gain a statistically significant, however, and no anti-GFP CTLs better understanding of the mechanism of this synergy could be detected in mice challenged with the GFP- between the antiangiogenic activity of endostatin and the expressing cells. It is unclear why, in our study, the GFP- immunogenicity of GFP. expressing tumor cells were still poorly immunogenic. Synergy between treatment modalities that act through Perhaps, the degree of anti-GFP response is strain- or different mechanisms has often been observed and exploited tumor-related, or perhaps the level of GFP expression in our in anticancer strategies. Specifically, synergy between experiments was lower than that which generated significant angiogenesis inhibition and other approaches such as anti-GFP immunity in the other studies. Nevertheless, even conventional , radiation therapy, and immu- in our study, it is likely that the GFP-expressing tumor cells notherapy has been shown recently to be effective. This often had a slightly increased immunogenicity given that the in provides the added benefit of permitting the use of lower vivo growth difference between the unmodified and GFP- doses of each agent to achieve antitumor efficacy. These expressing tumor cells in the syngeneic, immunocompetent lower doses are often far below the levels at which A/J mice was completely eliminated when the cells were undesirable side effects of each individual modality might grown in immunodeficient C.B-17 SCID mice. be observed. This synergy between angiogenesis inhibition

Cancer Gene Therapy, Vol 8, No 7, 2001 DAVIDOFF, LEARY, NG, ET AL: SYNERGISTIC ANTITUMOR EFFECT OF ENDOSTATIN AND GFP 545 and immunotherapy may be exploitable in anticancer 17. Greten TF, Jaffee EM. Cancer vaccines. Blood. 1999;17:1047± vaccine strategies that currently appear to be efficacious 1060. only in settings of minimal residual disease. 18. Stripecke R, Villacres MdC, Skelton DC, et al. Immune response to green fluorescent protein: implications for gene therapy. Gene Ther. 1999;6:1305±1312. 19. Lode HN, Xiang R, Varki NM, et al. Targeted interleukin-2 ACKNOWLEDGMENTS therapy for spontaneous neuroblastoma metastases to bone marrow. J Natl Cancer Inst. 1997;98:1586±1594. This work was supported by grants from the Assisi 20. Dubridge RB, Tang P, Hsia HC, et al. Analysis of mutation in Foundation of Memphis 94-000, Grant IRG-87-008-09 human cells by using an Epstein-Barr virus shuttle system. Mol from the American Cancer Society, Cancer Center Support Cell Biol. 1987;7:739±745. CORE Grant P30 CA 21765, and American Lebanese Syrian 21. Rehn M, Pihlajaniemi T. Identification of three N-terminal Associated Charities. ends of type XVIII collagen chains and tissue-specific differences in the expression of the corresponding transcripts. J Biol Chem. 1995:270,9:4705±4711. 22. Smith TAG, Mehaffey MG, Kayda BD, et al. Adenovirus- REFERENCES mediated expression of therapeutic plasma levels of human factor IX in mice. Nat Genet. 1993;5:397±402. 1. Folkman J. The role of angiogenesis in tumor growth. Semin 23. Persons DA, Allay JA, Allay ER, et al. Enforced expression of Cancer Biol. 1992;3:65±71. the GATA-2 transcription factor blocks normal hematopoiesis. 2. Bicknell R, Harris AL. Mechanisms and therapeutic implica- Blood. 1999;93:488±499. tions of angiogenesis. Curr Opin Oncol. 1996;8:60±65. 24. Migita M, Medin JA, Pawliuk R, et al. Selection of transduced 3. Pluda JM. Tumor-associated angiogenesis: mechanisms, CD34 + progenitors and enzymatic correction of cells from clinical implications, and therapeutic strategies. Semin Oncol. Gaucher patients, with bicistronic vectors. Proc Natl Acad Sci 1997;24:203±218. USA. 1995;92:12075±12080. 4. Folkman J, Ingber D. Inhibition of angiogenesis. Semin Cancer 25. Skidgel RA, Erdos EG. Cellular carboxypeptidases. Immunol Biol. 1992;3:89±96. Rev. 1998;161:129±141. 5. Bicknell R, Harris AL. Anticancer strategies involving the 26. Persons DA, Mehaffey MG, Kaleko M, et al. An improved vasculature: vascular targeting and the inhibition of angiogen- method for generating retroviral producer clones for vectors esis. Semin Cancer Biol. 1992;3:399±407. lacking a selectable marker gene. Blood Cells, Mol, Dis. 6. Scott PAE, Harris AL. Current approaches to targeting cancer 1998;24:167±182. using antiangiogenesis therapies. Cancer Treat Rev. 1994;20: 27. Yang Y, Vanin EF, Whitt MA, et al. Inducible, high-level 393±412. production of infectious murine leukemia retroviral vector 7. O'Reilly MS, Boehm T, Shing Y, et al. Endostatin: an particles pseudotyped with vesicular stomatitis Virus G endogenous inhibitor of angiogenesis and tumor growth. Cell. envelope protein. Hum Gene Ther. 1995;6:1203±1213. 1997;88:277±285. 28. Sarmiento M, Glasebrook AL, Fitch FW. IgG or IgM 8. Dhanabal M, Ramchandran R, Waterman MJF, et al. Endostatin monoclonal antibodies reactive with different determinants on induces endothelial cell apoptosis. J Biol Chem. 1999;274: the molecular complex bearing Lyt 2 antigen block T cell- 11721±11726. mediated cytolysis in the absence of complement. J Immunol. 9. Dhanabal M, Ramchandran R, Ruediger V, et al. Endostatin: 1980;125:2665±2672. yeast production, mutants, and antitumor effect in renal cell 29. Dialynas DP, Quan ZS, Wall KA, et al. Characterization of the carcinoma. Cancer Res. 2000;59:189±197. murine T-cell surface molecule, designated L3T4, identified by 10. Perletti G, Concari P, Giardini R, et al. Antitumor activity of monoclonal antibody GK1.5: similarity of L3T4 to the human endostatin against carcinogen-induced rat primary mammary Leu-3/T4 molecule. J Immunol. 1983;131:2445±2451. tumors. Cancer Res. 2000;60:1793±1796. 30. Yamaguchi N, Anand-Apte B, Lee M, et al. Endostatin 11. Boehm T, Folkman J, Browder T, et al. Antiangiogenic therapy inhibits VEGF-induced endothelial cell migration and tumor of experimental cancer does not induce acquired drug growth independently of zinc binding. EMBO J. 1999;16: resistance. Nature. 1997;390:404±407. 4414±4423. 12. O'Reilly MS, Holmgren L, Chen C, et al. induces 31. Yoon SS, Eto H, Lin C, et al. Mouse endostatin inhibits the and sustains dormancy of human primary tumors in mice. Nat formation of lung and liver metastases. Cancer Res. Med. 1996;2:689±692. 1999;59:6251±6256. 13. Weber JS, Jay G, Tanaka K, et al. Immunotherapy of a murine 32. Blezinger P, Wang J, Gondo M, et al. Systemic inhibition of tumor with interleukin-2. J Exp Med. 1987;166:1716±1733. tumor growth and tumor metastases by intramuscular 14. Lord EM, Frelinger JG. Tumor immunotherapy: cytokines and administration of the endostatin gene. Nat Biotechnol. 1999; antigen presentation. Cancer Immunol Immunother. 1998;46: 17:343±348. 75±81. 33. Sauter BV, Martinet O, Zhang WJ, et al. Adenovirus- 15. Cheung NK. Immunotherapy. Neuroblastoma as a model. mediated gene transfer of endostatin in vivo results in high Pediatr Clin North Am. 1991;38:425±441. level of transgene expression and inhibition of tumor 16. Hock RA, Reynolds BD, Tucker-McClung CL, et al. Human growth and metastases. Proc Natl Acad Sci USA. 2000;97: class II major histocompatibility complex gene transfer into 4802±4807. murine neuroblastoma leads to loss of tumorigenicity, im- 34. Feldman AL, Restifo NP, Alexander R, et al. Antiangiogenic munity against subsequent tumor challenge, and elimination of gene therapy of cancer utilizing a recombinant adenovirus to microscopic preestablished tumors. J Immunother. 1995;17: elevate systemic endostatin levels in mice. Cancer Res. 12±18. 2000;60:1503±1506.

Cancer Gene Therapy, Vol 8, No 7, 2001