Sensitization of Tumor Necrosis Factor α-resistant Human Melanoma by Tumor-specific in Vivo Transfer of the Gene Encoding Endothelial Monocyte-activating Polypeptide II Using Recombinant Vaccinia Virus¹

TNF-α³ was initially described and named for its ability to induce necrosis in certain types of malignant tumors (1). The mechanism by which the cytokine exerts this effect is still not completely understood; however, experimental and clinical evidence shows that the effects of TNF-α on tumor-related vasculature are the most important component of its antitumor activity (2, 3, 4).

Clinical responses induced by systemic TNF-α are unusual in melanoma patients (5, 6), and its therapeutic use never came close to its initial expectations, based on preclinical data. The reasons for this are mainly related to the toxicity of the cytokine (7, 8, 9). TNF-α has numerous proinflammatory activities inducing capillary leak syndrome, shock, and coagulation disorders following systemic administration (10, 11), at doses far below those required to mediate antitumor effects (12).

There is, however, still clinical use of TNF-α in treatment settings in which systemic toxicity can be avoided by anatomical isolation and perfusion of a cancer-bearing organ or region of the body (13, 14). Several groups, including ours, have described the usefulness of this approach in isolated limb and hepatic perfusion for extremity melanoma, sarcoma, or unresectable liver tumors (15, 16, 17, 18). Dramatic responses can be observed quite frequently (19), and the procedures have been well defined and can eliminate systemic exposure to TNF-α (14). Currently, clinical trials are underway to define the exact contribution of TNF-α to these treatments as well as the optimal schedule and dosage (20, 21). Most, if not all, of the observed antitumor effects of the cytokine are achieved indirectly via its effect on the vasculature, with increased capillary permeability, erythrostasis, and endothelial cell necrosis (22, 23).

There have been several approaches to develop TNF-α-related gene therapies, mostly by introducing the gene for the cytokine into effector cells or tumor cells using an adoptive immunotherapy or tumor vaccine approach (24, 25, 26, 27, 28, 29). Most reported in vivo results were disappointing, either because of low effectiveness or subsequent toxicity (30, 31, 32).

The tumor-derived cytokine EMAP-II is an important factor in the activation of tumor neovasculature (33). EMAP-II is translated as a M_r 34,000 precursor protein and cleaved to a M_r 22,000 mature protein (34). High EMAP-II expression leads to induction of tissue factor in endothelial cells and correlates directly with the TNF-α sensitivity of tumors (35). Unfortunately, whereas both the EMAP-II gene and mRNA is present in all eukaryotic cells (36), protein expression is down-regulated in mature tissues as well as in most tumors. Actual protein expression levels differ considerably among tumors, even of the same histological type, and their respective sensitivities toward in vivo treatment with TNF-α depend on this expression. Both intratumoral injection of recombinant EMAP-II protein into TNF-resistant tumors (37) and stable in vitro transduction of low EMAP-II-expressing melanomas (35) rendered those tumors TNF-α sensitive in vivo, adding further evidence to the direct relationship between EMAP-II expression and TNF-α sensitivity.

Although these observations helped elucidate the mechanisms of TNF-α activity and its interaction with EMAP-II, their potential clinical applicability is limited, mainly because most tumors are low EMAP-II expressers. We hypothesized that the in vivo introduction of the EMAP-II gene into TNF-α-resistant tumors might lead to increased EMAP-II production by these tumors, thus providing a window of opportunity for systemic TNF-α treatment. By gene transfer leading to overexpression of EMAP-II specifically within tumors, we would be able to artificially increase the therapeutic window for TNF-α between these tumors and normal tissue, thus facilitating the use of nontoxic treatment dosages. The vector system used here is a recombinant vaccinia virus that is severely attenuated by the disruption of its TK gene (38).

Plasmids and viruses carrying the EMAP-II gene or a reporter gene were synthesized, and in vitro experiments provided evidence of successful and selective gene transfer and expression in tumor cells. We then characterized the expression of a reporter gene in a murine in vivo model of s.c. human melanoma in tumors compared with other tissues over time. Finally, we describe the in vivo application of the sensitizing gene therapy and subsequent systemic TNF-α treatment and its effect on in vivo tumor growth.

Pmel and 1286 are primary human melanoma lines derived from patients treated at the National Cancer Institute (Bethesda, MD). The human embryonic kidney cell line 293, the ovarian cancer line HeLa, and the monkey kidney cell line CV-1 (American Type Culture Collection, Manassas, VA) were used for generation, amplification, and titration of recombinant vaccinia viruses. All cell lines were cultured in DMEM supplemented with 10% fetal bovine serum, 2 mm l-glutamine, 1% penicillin/streptomycin, 1% gentamicin, and 0.2% fungizone (Biofluids, Rockville, MD) at 37°C in a 5% CO₂ incubator.

HUVECs (Clonetics, Walkersville, MD) were maintained at 37°C in a 5% CO₂ incubator in EGM-2 medium (basal endothelial cell media EBM enriched with hydrocortisone, fibroblast growth factor, vascular endothelial growth factor, insulin-like growth factor-1, ascorbic acid, epidermal growth factor, GA-1000, and heparin) available from Clonetics. They were passaged no more than six generations and cryopreserved at regular intervals.

The EMAP-II gene was amplified from human monocyte cDNA via PCR using Pfu polymerase (Stratagene, La Jolla, CA) and using primers derived from the published GenBank sequence (accession no. U10117). After ligation into the pCRII cloning vector (Invitrogen, Carlsbad, CA), the correct gene sequence was confirmed by automated sequencing (ABI Prism, Perkin-Elmer Applied Biosystems, Norwalk, CT). The human EMAP-II gene was then cloned by directional cloning into the vaccinia virus shuttle vector vCB023-II, resulting in pMPhEII. The human EMAP-II is therein driven by the optimized synthetic early-late promoter. The construction of vCB023-II is described in detail elsewhere (39). Briefly, the XbaI site of pSC65 (kindly provided by B. Moss, NIH, Bethesda, MD) within the flanking region of the TK gene was abolished with Klenow and dNTP filling. The plasmid was then amplified with primers to create a BSSH II site between the TK flanking regions, allowing for the insertion of the multiple cloning site of pBluescript KS II (+) (Stratagene, La Jolla, CA), which was used to insert gpt as well a promoter cassette consisting of a vaccinia virus p7.5 early/late and a synthetic early/late promoter (40, 41) in the sense and antisense directions, respectively, immediately upstream.

A plasmid containing the firefly luciferase gene was constructed in a similar way using the luciferase gene from pGEM-luc (Promega, Madison, WI), as described previously (39).

Recombination of vaccinia viruses (strain Western Reserve) was performed using CaCl precipitation (42). Recombinant vaccinia viruses were subsequently isolated by mycophenolic acid selection and designated vvLuc for the reporter gene-containing virus and vvEMAP for the virus containing the EMAP-II gene. Recombinant viruses were amplified in HeLa S3 cells using spinner flasks as described (42) and titered in standard plaque forming assays using CV-1 cells.

Total RNA was extracted from cell cultures infected with the viral constructs using RNEasy (Qiagen, Chatsworth, CA). cDNA was transcribed by reverse transcriptase reaction, and PCR was performed. For vvEMAP, the following primers were used: 5′-AACTGAAACAAGAGCTAATT-3′ and 5′-CAGGCTCTCCTGGGAAAG-CA-3′. Amplification of cDNA was performed using a GeneAmp thermocycler (Perkin-Elmer) using 25 cycles of a 94°C denaturing step (60 s), a 52°C annealing step (30 s), and a 70°C extension step (90 s).

To produce tumor-conditioned medium, we plated ∼5 × 10⁶ tumor cells per 75-cm² tissue culture flask (Corning Costar, Cambridge, MA) in 15 ml of serum-free medium 199 (NIH Media Unit, Bethesda, MD). After 24 h, Pmel tumor cells were transfected with vvEMAP or vvLuc (control virus) at a multiplicity of infection (MOI) of 0.1 for 20 h. Conditioned medium was then collected and vacuum-filtered across a 0.45-μm membrane (Corning) to remove any cellular debris and used immediately.

HUVECs were plated at 5 × 10⁵ cells per well of a six-well plate (Corning). After 48 h, the cells were washed twice with sterile PBS and treated, in triplicate, with 2 ml of tumor-conditioned medium, medium conditioned by tumors infected with the viral constructs, EGM-2 medium alone (negative control), or EGM-2 medium plus 150 ng/ml IL-1β (R&D Systems, Minneapolis, MN) as a positive control. In addition, one group of HUVECs was treated with vvEMAP-conditioned medium plus 200 μg/ml polyclonal EMAP-II rabbit IgG as an EMAP-II-blocking antibody. The treated endothelial cells were cultured at 37°C for 18 h, washed twice with sterile PBS, harvested in 500 μl of PBS per well using mechanical scrapers, lysed using three rapid freeze-thaw cycles, and stored at −70°C.

Cell suspensions were thawed at 37°C and centrifuged at 2000 × g for 5 min. The cell pellet was then resuspended in 200 μl of 50 nm Tris, 100 nm NaCl, and 0.1% BSA. A one-stage coagulation assay was performed by adding 100 μl of Factor VIII-deficient human plasma (George-King Biomedical, Overland Park, KS) to 100 μl of endothelial cell suspension and incubating for 2 min at 37°C. The reaction was catalyzed by the addition of 100 μl of 30 nm CaCl₂; clotting time was measured using an Amelung micro-coagulation analyzer (Sigma, St. Louis, MO). Standard curves were generated using dilutions of recombinant human tissue factor (American Diagnostica, Greenwich, CT). The assay sensitivity for TF was <10 pg/ml.

Female athymic NCR-nude mice, 8–12 weeks old, were obtained from the NIH small animal facility and housed in a barrier care room. Human tumors were inoculated in the right flank region by s.c. injection of 10⁶ tumor cells suspended in 200 μl of sterile HBSS. Animals had unlimited access to food and water and were housed at a maximum of five mice per cage. All animal protocols were approved by the Animal Care and Use Committee and conducted in strict compliance to the guidelines established by the NIH Animal Research Advisory Committee.

For the reporter gene experiments, animals were injected i.v. with vvLuc when tumors reached ∼300 mm³ in size, typically 3–4 weeks after tumor inoculation. Cohorts of animals were sacrificed at various time points and tumors and normal tissues (liver, kidneys, spleen, uterus, ovary, bowel, heart, lung, muscle, skin, and brain) were harvested, immediately frozen, and stored at −70°C.

For the in vivo treatment experiments tumor sizes were measured 21–28 days after the injection of tumor cells. Animals were ear-tagged and randomized at that time point, and identical average tumor sizes (close to 300 mm³) were selected for the actual experimental cohorts. The viral constructs were injected via tail vein injection in 200 μl of HBSS. Daily tumor measurements using a caliper were performed by blinded investigators, and tumor volumes were calculated according to the formula: volume = width² × length × 0.52 (43). On days 4 and 6 after virus injection, treatment groups received the maximum tolerable dose for systemic administration, 7 μg of TNF-α (Knoll Pharmaceutical, Whippany, NJ), via tail vein. Activity of a given TNF-α lot was ensured using concomitant in vivo treatment of TNF-α-sensitive 1286 tumors (data not shown). Control groups received tail vein injections of the vehicle alone (200 μl of 0.9% NaCl solution-0.5% BSA). Animals were continuously measured in a blinded way on a daily basis until tumor sizes reached 1000 mm³. At that point, animals had to be sacrificed according to protocol and guidelines to avoid prolonged pain and suffering.

Luciferase activity was assayed using a commercially available assay system (Promega). Frozen tissue samples were briefly thawed, homogenized, and lysed in 750 μl of 1× reporter gene lysis buffer. Samples were centrifuged for 5 min at 13,000 rpm to pellet the cellular debris. Ten μl of supernatant were added to 100 μl of luciferase assay reagent in an 8 mm × 50 mm disposable cuvette and immediately placed into a luminometer (model TD-20/20; Turner, Sunnyvale, CA). After a 2-s delay, light emission was measured for 10 s. Each sample was measured in duplicate, and mean values were recorded. The concentration of total protein in each sample was then determined using a bicinchoninic acid protein assay kit (Pierce, Rockford, IL) using a BSA standard. All measurements were performed in triplicate. The final firefly luciferase activity is expressed in rlu (= light units per mg protein).

Tumor and tissue specimens were harvested from animals sacrificed at time points given and immediately fixed with 10% paraformaldehyde in 0.1 m phosphate buffer. Paraffin embedding, sectioning, and H&E staining were performed at Paragon Biotech (Baltimore, MD). Immunohistochemistry for EMAP-II was performed using a polyclonal chicken anti-EMAP-II antibody at a dilution of 1:4000 (Aves Laboratories, Tigard, OR) and horseradish peroxidase detection. Slides were examined using light microscopy and read by an investigator blinded to the treatment groups.

Data were analyzed using ANOVA with repeated measures and Tukey post hoc tests using Systat (Evanston, Illinois) statistical software and are given as means ± SE. When appropriate, Student‘s paired t test (two-sided) was used, and means ± SD are reported.

We developed a functional bioassay of EMAP-II activity based on the induction of tissue factor in human endothelial cells by EMAP-II-transfected tumor cells. HUVECs were treated with medium conditioned by tumor cells infected with vvEMAP and vvLuc. As shown in Fig. 1, a 20-h exposure of HUVECs to medium conditioned by Pmel infected with vvEMAP leads to an 8-fold increase in the production of tissue factor by these cells, as compared with treatment with conditioned medium from control Pmel or from Pmel infected with vvLuc (P < 0.001). The addition of anti-EMAP-II antibody abrogated the activating effect of medium conditioned with vvEMAP-infected cells (nonsignificant P versus Pmel controls; P < 0.01 versus Pmel plus vvEMAP).

The tumor model used for our in vivo studies was generated by s.c. injection of nude athymic mice with TNF-α-resistant Pmel human melanoma cells. Tumors were grown to an approximate size of 300 mm³, and 10⁶ pfu of vvLuc in 200 μl of HBSS were administered systemically via lateral tail vein injection. HBSS injections were used as negative controls.

Groups of mice (n = 5) were sacrificed at days 1, 3, 5, 7, 9, 12, 14, 16, 18, and 21 after virus injection. Tumor and tissue samples were harvested and assayed for their luciferase content and normalized to the total protein content of samples. Fig. 2 shows the results expressed in rlu per mg protein. In general, tumor samples had a gene expression that was ∼100 times higher than those of other tissues, providing evidence of tumor-selective delivery of the foreign gene. Maximum gene expression in tumors was observed between days 4 and 7, and the difference in luciferase activity between tumors and the next closest gene-expressing organs was maintained for up to 21 days. The observed pattern in gene expression proves the existence of a “therapeutic window,” encouraging the usefulness of the approach in the treatment setting.

The human melanoma model in nude mice was used for in vivo treatment experiments. According to the gene expression pattern observed in the reporter gene experiments, treatment with TNF-α was scheduled on days 4 and 6 after systemic delivery of vvEMAP. Control cohorts included mice treated with PBS or vvLuc. TNF-α was administered to animals at a dose of 350 μg/kg body weight via lateral tail vein injection. Negative treatment controls received tail vein injections of the vehicle alone. Treatment and tumor measurement were performed in a blinded fashion.

Fig. 3 shows the tumor volume of animals treated with TNF-α after pretreatment with vvEMAP or vvLuc or PBS. Starting on day 7, a significant difference in tumor volumes was observed (day 7, P = 0.04). Whereas control tumors pretreated with vvLuc or PBS continued to grow uninhibitedly despite TNF-α treatment, TNF-α abrogated any further growth in the previously TNF-resistant Pmel tumors, indicating the functionally effective overexpression of the sensitizing EMAP-II gene in these tumors. This difference was observed until the end of the experimental period and increased markedly over time (day 14, P < 0.0001). The effect was observed irrespective of the number and timing of TNF-α treatments (data not shown). There was no difference between vvLuc- and PBS-pretreated tumors (nonsignificant P at all time points). All negative control subgroups treated with vehicle instead of TNF-α, irrespective of vector pretreatment showed continued tumor growth (data not shown).

Successfully treated tumors showed the typical macroscopic and microscopic effects of TNF-α. As an example, Fig. 4,A shows hemorrhagic necrosis macroscopically observed in vvEMAP/TNF-α tumors as compared with control vvLuc/TNF-α tumors (Fig. 4,B). Using light microscopy, we determined the average percentage of tumor necrosis in vvEMAP/TNF-α tumors to be ∼75–85% versus 20–30% in controls (Fig. 4, C and D, respectively). Even in treatment animals, however, a small rim of viable tumor cells survived and led to eventual regrowth of tumors.

Histopathological analysis of other organs, particularly the ovaries, did not show necrosis or other changes.

Immunostaining for the presence of EMAP-II protein in tumors treated with vvEMAP showed a marked cytoplasmic and membrane-bound staining pattern (Fig. 4,E), as compared with controls treated with vvLuc (Fig. 4 F).

We have constructed a recombinant vaccinia virus expressing the human EMAP-II gene and shown the feasibility of tumor-specific gene delivery in a murine model of s.c. human melanoma using this vector system. Functional protein secretion was achieved in vitro and in vivo and led to the sensitization of a previously TNF-α-resistant tumor toward TNF-α treatment.

The clinical use of TNF-α has been disappointing, mainly because it is impossible to deliver the necessary antitumor doses of the cytokine to patients without severe toxic side effects. Systemic TNF-α has been very effective in the treatment of established murine tumors (44), but mice tolerate as much as 50-fold higher relative TNF-α dosage than humans (45). The specific targeting of tumors with the cytokine is impossible in most clinical settings, including generalized melanoma, due to the difficulty of tumor-specific delivery of the agent (46). A tumor-specific sensitization gene therapy approach provides a unique opportunity to artificially render established tumors sensitive to the systemically administered TNF-α without increasing the vulnerability of healthy tissues and organs, thus facilitating systemic treatment with tolerable doses of the cytokine.

One important basis for the success of this approach is the unique ability of recombinant vaccinia viruses to specifically transfect tumors. It is not yet fully understood why systemic application of vaccinia viruses leads to preferential expression of the foreign gene in tumors (47). Whereas at least some gene expression can also be observed in other organs, the observed 100-fold increased expression of the genes transferred within tumors as compared with other tissues provides a window of opportunity for many treatment approaches, including the delivery of a “sensitizer” toward subsequent treatment, as demonstrated. Other approaches used in this context include delivery of genes encoding for prodrug converter enzymes in a so-called suicide gene therapy approach (48) and genes increasing the immunogenicity of tumors via expression or overexpression of antigens (49), costimulatory molecules (47, 50, 51), or cytokine gene transfer (52, 53).

Hypothetical reasons for the preferential gene expression within tumors include the relatively large size of the vector (>300 nm) in conjunction with the potentially leakier tumor neovasculature and a decreased immune clearance of the viral construct within tumors due to local immunosuppressive factors (54).

In our view, however, it is most likely that the insertion of the foreign genes into the locus of the viral TK gene, leading to the disruption of that gene, which plays a major role in the tumor specificity of the vector (55). Human tumor cells are metabolically much more active than benign tissue, providing an abundant pool of functional nucleotides that can compensate for the absence of the viral TK. Human melanoma, however, seems to be less permissive for vaccinia virus infection when compared directly to murine tumors (39). We have elsewhere demonstrated differences in gene expression of up to 4 orders of magnitude between tumors and other tissues when a syngeneic murine tumor is used in mice (39).

It is difficult to predict whether this tumor specificity will be preserved or even enhanced when the vector system is used in the clinical setting. Thus far, recombinant vaccinia viruses have mainly been used in Phase I clinical trials to induce specific immune responses against generalized tumors (56). The large amount of viral proteins of recombinant vaccinia viruses, particularly compared with other viral vectors, yields a massive immune response that induces long-lasting immunity against antigens encoded by the viral vectors (57). This, however, may limit the potential clinical use in a drug or cytokine-sensitizing gene therapy approach; repeated application of the same viral vector may likely lead to rapid destruction of the vector (58). At least transient replication and gene expression seem to be possible, even in preimmunized patients.

The high immunogenicity of the vector system also provides a potential safety benefit by virtually eliminating the likelihood of systemic infections with vaccinia virus in immunocompetent human hosts. In addition, although they are replication competent, the recombinant viral constructs we are using are severely attenuated by the disruption of their TK gene (59). Furthermore, the smallpox eradication program has established the safety of even intact vaccinia viruses (60). To date, there has been no vector-associated clinical toxicity reported in several Phase I cancer or other gene therapy trials using recombinant vaccinia viruses, whether administered intratumorally or i.v. (61, 62).

Although we were able to demonstrate dramatic responses and tumor regressions in our in vivo model, our sensitization/treatment concept did not eliminate all viable tumor cells, leading to eventual regrowth of responding tumors. Intervening earlier in the course of the disease may circumvent this problem. Furthermore, lowering the number of viable tumor cells below a certain threshold might even lead to complete clinical remissions and cures (63) because the host’s natural antitumor activity is much more effective with limited tumor burden. There is, however, some indication that the recombinant vaccinia viruses that we use as vectors do not exert their tumor-tracking ability in an adjuvant setting, most likely because established tumors and tumor vasculature are needed, to some extent, for infection and amplification.

A previous attempt to sensitize human tumors toward TNF-α treatment by transfecting human tumor cells with the M_r 55,000 TNF receptor gene in vitro was reported recently (64). Although that attempt provided promising results, the concept has not yet been used for established nonsensitive tumors in vivo.

EMAP-II has been described as a tumor-derived cytokine initially by its ability to induce tissue factor procoagulant activity (65) and up-regulating the leukocyte adhesion molecules P-selectin and E-selectin. Its properties are multiple and have not yet been completely elucidated. (a) It has obvious proinflammatory effects and can be found in experimental autoimmune encephalomyelitis, neuritis, and uveitis (66). (b) It regulates attraction of monocytes/macrophages to the sites of inflammation (34). Recently, EMAP-II mRNA was found most abundant at sites of tissue remodeling in mouse embryos, suggesting that cells heading toward apoptotic cell death overexpress EMAP-II to attract macrophages for removal of dead cells (67). In fact, the release of mature EMAP-II was found coincident with apoptosis in vitro and in vivo.

Direct injection of recombinant EMAP-II leads to sensitization of TNF-α-resistant tumors (37). Conclusive evidence for the causal importance of EMAP-II expression for TNF-α response in human melanoma was provided by stable retroviral transduction of TNF-α resistant human melanoma cells with the EMAP-II gene in vitro, resulting in a phenotypic change of these tumors in vivo, rendering them susceptible to TNF-α treatment (35). The specific in vivo delivery of the gene into established non-EMAP-II-expressing TNF-α-resistant tumors and their subsequent regression and growth inhibition translates those findings into a clinically more realistic perspective because most tumors in patients will either express EMAP-II only at a very low level or not at all.

Although the exact mechanism of how EMAP-II sensitizes tumors toward TNF-α remains unclear, there is evidence that TNF receptor I is up-regulated in endothelial cells by EMAP-II in a dose-dependent fashion (68, 69), thus triggering death of endothelial cells in the presence of TNF-α and, subsequently, inducing ischemic tumor necrosis.

TNF-α exerts a very potent antitumor effect when the dose of the cytokine and the sensitivity of the targeted cells combine in a favorable manner. We have shown that modern gene transfer techniques provide us with the opportunity to artificially improve the local effectiveness of TNF-α in human melanoma while avoiding toxic side effects in other organs. Our results may lead to reevaluation of the role of TNF-α in clinical tumor therapy. The pretreatment of human tumors with recombinant vaccinia virus encoding for EMAP-II may sensitize those tumors toward TNF-α or at least increase the therapeutic window in moderately TNF-α-sensitive tumors, helping to reduce the necessary doses of the agent and, consequently, reduce toxicity and side effects.

We thank Dr. Steven Rosenberg for helpful comments and Barbara Owen for technical assistance with the manuscript.