Tumor Pretargeting with Avidin Improves the Therapeutic Index of Biotinylated Tumor Necrosis Factor α in Mouse Models¹

The antitumor properties of TNF³ and its unique efficacy in selective destruction of tumor-associated vessels are well known (1). Phase I and Phase II clinical trials have shown that the systemic administration of TNF to cancer patients is hampered by dose-limiting toxicity, with the maximum tolerated dose being 8–20-fold less than the efficacious dose in animals (2). For this reason, the clinical use of TNF as an anticancer drug has been limited thus far to local or locoregional treatments. For instance, regional administration of relatively high doses of TNF in combination with melphalan by isolated limb or hepatic perfusion has shown high complete response rates in patients with melanoma and sarcoma of the extremities (3, 4, 5) and regression of bulky hepatic cancers confined to the liver (6). In addition, i.p. administration of TNF resulted in the reduction of malignant ascites (7), whereas local administration at the tumor site has shown promising response rates in Kaposi’s sarcoma, plasmacytomas, ovarian adenocarcinomas, and various metastatic tumors in the liver (8, 9). The positive results of these studies imply that TNF can exert antitumor effects against human cancer if high, locally effective doses are used and the systemic toxicity is kept under control. Thus, several attempts have been made to enhance the antitumor activities of TNF or to reduce its systemic toxicity (reviewed in Ref. 10); unfortunately, thus far, these attempts have had little success.

We have recently described a new approach to localize the TNF action on tumor cells based on sequential incubation of cells with specific biotinylated antibodies, avidin, and biotinylated TNF (11). We observed that this treatment markedly increases the amount and the persistence of biotin-TNF on the cell surface. Furthermore, biotin-TNF bound to avidin is able to trigger cytolytic effects in vitro and decrease the growth rate of tumor cells after injection in mice under conditions in which treatment with nonbiotinylated TNF is almost inactive. Based on these findings and on the fact that tumor pretargeting with biotinylated antibodies and avidin is currently performed in patients to increase the tumor uptake of biotinylated radioisotopes (12, 13, 14, 15), we have hypothesized that tumor avidination strategies might increase the intratumoral homing of systemically administered biotin-TNF and, consequently, its therapeutic index. To address this hypothesis, we studied both the therapeutic and toxic effects of biotin-TNF in mice bearing different tumors (pretargeted or non-pretargeted with avidin). We show that tumor avidination potentiates the antitumor activity of biotin-TNF, with no evidence of increased toxicity.

The Rauscher virus-induced RMA lymphoma of C57BL/6 origin (H-2^b) and the 3-methylcholanthrene-induced WEHI-164 fibrosarcoma of BALB/c origin (H-2^d; ATCC CRL-1751) were maintained in vitro in RPMI 1640 with 5% FCS, 100 units/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B, 2 mm glutamine, and 50 μm 2-mercaptoethanol. Methods for generating RMA cells that express the Thy 1.1 antigen on their surface (RMA-T) have been described previously (11). WEHI-164 cells expressing the Thy 1.1 antigen (WEHI-164-T) were obtained and cultured using the same protocols. In the absence of transcription inhibitors, the WEHI-164-T and WEHI-164 cells (different from the highly sensitive WEHI-164 clone 13 commonly used for in vitro cytolytic assays of TNF) are resistant to <100 ng/ml TNF. In the presence of actinomycin D, the WEHI-164 and WEHI-164-T cells are killed by TNF (LD₅₀, 1–3 ng/ml). This is agreement with the results of previous works (16). A similar behavior was observed also with RMA-T cells.

Murine and human recombinant TNF (5 × 10⁷ units/mg were produced by expressing their cDNAs in Escherichia coli (17, 18). Both products were purified to homogeneity from E. coli crude extracts by ammonium sulfate precipitation, followed by hydrophobic interaction chromatography on phenyl-Sepharose, ion exchange chromatography on DEAE-Sepharose, and gel filtration chromatography on Sephacryl-S-300 (Pharmacia-Upjohn, Cologno Monzese, Italy). TNF biotinylation was carried out as described previously (11). The number of biotins per TNF subunit was measured by electrospray mass spectrometry, as described previously (19). Only conjugates containing monomeric species with zero and one biotin, accounting for 65–70% and 30–35% of the total, respectively (i.e., with about one biotin per trimer on average) were used. The bioactivity of each biotin-TNF conjugate was estimated by standard cytolytic assay using L-M mouse fibroblasts (ATCC CCL1.2; Ref. 20). The protein content was measured using a commercial protein assay kit (Pierce, Rockford, IL). The specific activity of conjugates was 2.0–2.5 × 10⁷ units/mg, and endotoxin content was 0.1–0.2 units/μg, as measured by the Lymulus Amoebocyte Lysate Pyrotest (Difco Laboratories, Detroit, MI).

The anti-Thy 1.1 mAb 19E12 (IgG2a) was purified and biotinylated as described previously (11). Biotin-mAb19E12 was quantified by sandwich ELISA using purified goat antimouse IgG polyclonal antibody in the capturing step and a goat antimouse IgG polyclonal immunoglobulin-peroxidase conjugate in the detection step, according to standard procedures. The activity of various lots of antibody conjugates was checked by FACS analysis of RMA-T cells using streptavidin-R-phycoerythrin (Sigma Chemical Co., St. Louis, MO) as the detecting reagent. The endotoxin content was 0.021 units/μg.

Avidin and streptavidin were purchased from Società Prodotti Antibiotici (Milan, Italy). The endotoxin content of both products was <0.006 units/μg.

The neutralizing antihuman TNF rabbit polyclonal antiserum (primarily IgM and IgG) was purchased from Genzyme (Cambridge, MA), the rat antimurine TNF mAb MP6-XT22 (IgG1) was from PharMingen (San Diego, CA), and the goat antirat IgG-FITC and goat antimouse IgG-FITC were from Southern Biotechnologies (Birmingham, AL).

Coating of RMA-T and WEHI-164-T cells with biotinylated mAb, avidin, and biotin-TNF was carried out by three sequential incubations (10 min each) in the presence of biotin-mAb 19E12, avidin, and biotin-TNF, as described previously (11). The TNF antigen present on the cell surface was then measured by cytofluorometric analysis. To this end, after the third incubation step, the cells were washed with PBS (0.15 m sodium chloride and 0.05 m sodium phosphate) containing 2% FCS (PBS/FCS), incubated on ice with rabbit anti-TNF antiserum (1:1000, 10 min), washed again with PBS/FCS, and further incubated with goat antirabbit-FITC (1:1000, 10 min). After the final washing, the cells were analyzed using a FACScan (Becton Dickinson, Mountain View, CA). For dissociation kinetic studies, two aliquots of the cells were incubated with TNF or biotin-TNF for 1 h, washed, and further incubated in culture medium. At various times, subaliquots of cells were withdrawn and fixed with 0.25% paraformaldehyde for 1 h at 4°C. At the end of the experiment, the cells were analyzed by FACS analysis. The mean fluorescence was used to estimate the amount of TNF bound to the cells at each time.

In vivo studies on animal models were approved by the Ethical Committee of the San Raffaele H Scientific Institute and performed according to the prescribed guidelines. C57BL/6 and BALB/c mice (Charles River Laboratories, Calco, Italy) were challenged with 5 × 10⁴ RMA-T or 7 × 10⁶ WEHI-164-T living cells, respectively, s.c. in the left flank. At various times after tumor implantation, mice were treated by sequential injections of biotinylated antibody, avidins, and biotin-TNF according to a 3-day or a 2-day protocol. In the 3-day protocol, we injected 40 μg of biotin-mAb19E12 (i.p., step 1), 60 μg of avidin and 60 μg of streptavidin after 18 and 19 h, respectively (i.p., step 2), and 1–22 μg of biotin-TNF 24 h later (i.p or i.v., step 3). In the 2-day protocol, step 2 included an additional injection of 5 μg of biotin-HSA (i.p.) 4 h after streptavidin to clear the excess of circulating avidins, and step 3 was performed 1 h later. Throughout this work, steps 1 and 2 are collectively called “pretargeting steps,” whereas step 3 is called the “effector step.” Each compound was diluted with a sterile 0.9% sodium chloride solution. In control experiments, one or more of the above components were replaced with the diluent. Each experiment was carried out with five mice/group. The tumor growth was monitored daily by measuring the tumor size with calipers. The tumor area was estimated by calculating r₁ × r₂ π, whereas tumor volume was estimated by calculating r₁ × r₂ × r₃ × 4/3 π, where r₁ and r₂ are the longitudinal and lateral radii, and r₃ is the thickness of tumors protruding from the surface of normal skin. Animals were killed before the tumor reached 1.0–1.3 cm in diameter.

Tumor sizes are shown as the mean ± SE (five animals/group). Tumor growth curves within each experiment were compared by ANCOVA, considering only the time points after treatment with TNF. The Ps for each of the experiments are shown in the figure legends. The global effect of pretargeting, which was adjusted for experiment and dose and estimated by including all of these variables in an ANCOVA model, was highly significant (P = 0.0001).

To verify that pretargeting can improve the therapeutic index of TNF, we developed two separate models based on modified WEHI-164 fibrosarcoma or RMA lymphoma cells transfected with the Thy 1.1 allele. The Thy 1.1 antigen enables targeting with the anti-Thy 1.1 mAb 19E12 (11).

FACS analysis of transfected cells showed that Thy 1.1 was stably expressed on the cell surface (Fig. 1, A and B). Both transfected cell lines were tumorigenic in syngeneic mice and were resistant to TNF even after a 48-h culture in the presence of 100 ng/ml mouse or human TNF. Moreover, FACS analysis of both cell lines, which were preincubated with mAb 19E12 for 24 h at 37°C, showed that the mAb 19E12/Thy 1.1 complex is not internalized (data not shown).

The levels of Thy 1.1 expression on engineered cells were comparable to that of the HMW-MAA on human Colo 38 cells, a naturally expressed antigen (19), as observed by FACS analysis with mAb 225 (Fig. 1 C).

In a previous study, we showed that the half-life of biotin-TNF on RMA-T cells coated with biotin-19E12 and avidin is about 7 h, 30-fold higher than that of nonbiotinylated TNF (11). Similarly, biotin-TNF persisted for several hours on the surface of WEHI-164-T cells precoated with biotin-19E12 and avidin (half-life, about 8 h; Fig. 2). In contrast, biotin-TNF disappeared rapidly from nonavidinated cells. Thus, in vitro avidination of either RMA-T or WEHI-164-T cells markedly increases the persistence of biotin-TNF on the cell surface.

To assess whether in vivo avidination of tumors can also increase the uptake and local persistence of systemically administered biotin-TNF, we performed a complete targeting experiment in animals bearing 2-day-old tumors. One day later, the tumors were excised, disaggregated, and analyzed by FACS analysis with an anti-TNF antibody. As shown in Fig. 3, the amount of TNF persisting on the pretargeted tumor after 24 h was markedly higher than that of a control in which avidin and streptavidin were omitted.

The antitumor effects of targeted human and murine biotin-TNF were then investigated. Using the WEHI-164-T model and the 2-day targeting protocol (see “Material and Methods”), we found that: (a) the pretargeting steps did not affect tumor growth when started on day 5 after tumor cell implantation (Fig. 4,A) or on day 10 after tumor cell implantation (data not shown); (b) 10 μg of human biotin-TNF alone (without pretargeting) caused significant necrosis in the central part of the tumor (data not shown) and delayed tumor growth for a few days (Fig. 4,B), whereas treatment with 1 and 10 μg of human biotin-TNF induced a complete regression of the tumor in 0% and 40% of animals, respectively; and (c) 1 μg of human biotin-TNF with pretargeting induced an antitumor effect similar to that of a 10-μg dose without pretargeting (Fig. 4 C) and induced complete tumor regression in 20% of animals.

Tumor growth was completely arrested by 2 μg of murine biotin-TNF injected after pretargeting (100% complete regression), whereas it was partially affected by the same amount of nonbiotinylated TNF (40% complete regression; Fig. 5,A). Because murine TNF binds both p55 and p75 TNF-Rs, whereas human TNF binds only the p55 TNF-R (21), the stronger effects observed with murine TNF suggest that both receptors are critically involved in the antitumor activity of targeted TNF. Notably, no potentiation of biotin-TNF occurred when mice were treated 2 days after tumor transplantation, i.e., when the tumors were not yet well established (Fig. 5,B). In nude mice, murine biotin-TNF induced a growth delay in tumors (but not rejection) only after pretargeting (Fig. 5 C).

We then performed a second series of in vivo experiments using the RMA-T lymphoma model. High doses of murine TNF (8–10 μg) administered i.v. to immunocompetent animals bearing established s.c. tumors caused a reduction in tumor mass and hemorrhagic necrosis in the central part of the tumor within 24 h, followed by a significant growth delay for 2–3 days (data not shown). TNF was unable to cause complete regression of this tumor because the living cells remaining around the necrotic area restarted growth 2–3 days after treatment. Animal death was observed at higher doses.

Tumor pretargeting performed according to the 2-day protocol induced a significant increase in the antitumor activity of murine biotin-TNF (3 μg, i.p.; Fig. 6,A). When we used the 3-day protocol, we obtained similar results (Fig. 6, B and C). Of note, in this case, we observed that biotin-HSA can be omitted without compromising the targeting effect, but not when the 2-day protocol is used (data not shown). It is likely that biotin-HSA is necessary in the 2-day protocol to remove the avidin/streptavidin excess from circulation, which may interfere with the homing of biotin-TNF to the tumor. Thus, the 3-day protocol, which is more simple, was used in all subsequent experiments.

Histological analysis of RMA-T tumors 24 h after treatment showed that pretargeting increased the hemorrhagic necrosis induced by biotin-TNF (Fig. 7).

To assess how many times pretargeting increases the therapeutic index of murine biotin-TNF, we compared the toxic and therapeutic effects of various doses of biotin-TNF administered i.v. with or without pretargeting. Pretargeting did not change the lethal effects of murine biotin-TNF (Table 1), whereas it increased the therapeutic effect of biotin-TNF without causing any animal death, even at doses close to the maximum tolerated dose (Fig. 8, A and B). We repeated the experiment several times with different doses and lots of reagents, with similar results. In general, the antitumor activity of biotin-TNF was increased by at least 5 times, whereas lethality was not affected. Only 1 of 20 animals was cured in the group treated with 9 μg of biotin-TNF (Table 1). Thus, in most cases, pretargeted TNF induced a delay in the RMA-T tumor growth but did not induce complete remission. It is noteworthy that the surviving animal rejected a second tumorigenic dose of RMA-T cells without the need for further treatment.

The loss of weight after TNF treatment of animals is a well known sign of systemic toxicity (22). Thus, we further examined the effect of pretargeting on the toxicity of biotin-TNF by measuring the animal weight before and after treatment. As expected, biotin-TNF alone caused a loss of weight 1 day after treatment in a dose-dependent manner, with a 0.25- and 1.8-g weight loss after the injection of 1 and 5 μg of biotin-TNF, respectively (Fig. 9,A). Pretargeting did not increase the weight loss caused by 1 μg of biotin-TNF; however, in agreement with the above results, it did increase the antitumor effect of biotin-TNF (Fig. 9, A and B). These results collectively suggest that pretargeting increases the antitumor effects of biotin-TNF without affecting its toxicity.

The results show that tumor pretargeting with avidin can potentiate the antitumor activity of systemically administered biotin-TNF. We performed the study using two separate animal models based on either transplanted mouse WEHI-164 fibrosarcoma or RMA lymphoma tumors genetically engineered to express the Thy 1.1 antigen on tumor cell membranes. Because both animal strains express the Thy 1.2 allele, the Thy 1.1 antigen enables in vivo tumor avidination by sequential injections of an anti-Thy 1.1 biotinylated antibody (mAb19E12) and avidin. Using these pretargeting models, we found that the activity of 1–2 μg of biotin-TNF, either murine or human, on pretargeted tumors was significantly increased and was comparable to that of 5–10 μg of biotin-TNF on non-pretargeted tumors. Furthermore, pretargeting did not increase the toxicity of biotin-TNF, as judged from animal survival and weight loss after treatment. Thus, pretargeting increased the antitumor activity of biotin-TNF at least 5 times without an apparent change in its systemic toxicity.

The results have also demonstrated that biotin-TNF can persist for several hours on the surface of avidinated cells as well as on tumors, whereas in the absence of pretargeting reagents, the conjugate disappears more rapidly. Moreover, when one or more of the targeting reagents was omitted, no increase in the antitumor activity of biotin-TNF was observed, suggesting that the mechanism for the improved activity is based on increased homing and persistence of TNF at the tumor site.

The question arises as to whether biotin-TNF bound to tumor cell membranes exerts its antitumor effects directly (by killing the targeted cells) or indirectly (by triggering host-related antitumor mechanisms). The following considerations suggest that the latter hypothesis is more likely.

(a) Cytotoxicity experiments carried out with either targeted or nontargeted TNF indicated that WEHI-164-T and RMA-T cells are not killed in vitro by TNF in the absence of transcription inhibitors. However, both tumors can undergo massive hemorrhagic necrosis and growth arrest or delay when targeted with biotin-TNF in vivo.

(b) We observed potentiation of biotin-TNF by pretargeting only with well established and vascularized tumors (tumors > 6–7 days old) but not with freshly transplanted tumors (3-day-old tumors). This strongly suggests that targeted TNF does not exert critical effects against the antibody-targeted cells themselves but rather against other targets present within the established tumors. Similarly, a previous study has shown that Meth A sarcoma is relatively resistant to the cytotoxic action of TNF in vitro; however, in vivo, it can undergo hemorrhagic necrosis upon systemic administration of TNF only 6–7 days after tumor implantation (22). Other studies showed that this phenomenon is related to the toxic effects of TNF on endothelial cells of tumor angiogenic vessels and on a shift to a procoagulant state by the same cells (23, 24, 25, 26). Our observation that targeted TNF does not arrest the growth of freshly transplanted tumors also suggests that the effects on established tumors are related to vessel damage. This view is further supported by the histological finding that pretargeting increases the hemorrhagic necrosis induced by biotin-TNF in vascularized RMA-T tumors.

(c) In nude mice, the overall effects of targeted TNF on the WEHI-164-T tumors were lower than those in immunocompetent animals. T-cell-dependent immunity is therefore critical and is part of the indirect mechanisms.

In summary, although pretargeting increases the binding and persistence of biotin-TNF on the surface of tumor cells in vivo, its ability to potentiate its antitumor effects is more likely related to indirect mechanisms involving changes in host-tumor relationships and/or host-mediated antitumor responses and, to a lesser extent, to direct effects on the targeted cells.

How can TNF on targeted tumor cells affect nontargeted host cells? We have shown in a previous in vitro study that bioactive TNF trimers are released in 1–2 days from the surface of targeted cells through trimer-monomer-trimer transitions (19). Moreover, we have also shown that free TNF can then diffuse and interact with TNF-Rs expressed by targeted cells as well as bystander cells. This may well explain the capability of targeted TNF to trigger indirect effects, e.g., by affecting endothelial cells and cells of the immune system within the tumor compartment.

A number of research groups are investigating various pretargeting approaches based on the avidin-biotin system for increasing the tumor uptake of radioactivity (12, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36). It is noteworthy that various antibodies against natural tumor-associated antigens have already been used, in combination with avidin, for delivering biotinylated radio-labeled conjugates to tumors in patients (12, 13, 14, 15, 37). These antibodies, in principle, could also be exploited to target biotin-TNF in patients. However, the fact that bioactive TNF can be released in the microenvironment of targeted cells (19), together with the evidence of indirect mechanisms of antitumor activity, opens the possibility that antigens other than those expressed on the surface of tumor cells may serve as targets. In view of the marked damage that TNF can cause to endothelial cells of angiogenic vessels (26), one attractive possibility is to directly target vascular proliferation antigens, e.g., angiogenesis-associated fibronectin isoforms (38) or angiogenic vessel endothelium markers (39) 

Several other points have to be considered to further assess the relevance of these results for a clinical application of the pretargeting approach. For instance, monoclonal antibodies as well as avidin and streptavidin are immunogenic (40). Thus, one can foresee that repeated injection of these reagents in patients will be limited by immunoresponses. However, the results obtained by a single treatment of patients with TNF and melphalan by isolated limb perfusion (3, 4, 41) suggest that one or few treatments before the development of an immunoresponse might be sufficient to reduce the tumor burden and to open the way to other therapeutic interventions (e.g., immunotherapy, chemotherapy, angiogenesis inhibition, etc.). Another issue that must be taken into consideration, which is not addressed in this study, is that the combination of TNF with melphalan is required for efficacy, because TNF is poorly active when used alone in locoregional treatments of human cancer (3, 42). Further work is therefore necessary to assess the synergistic effects of pretargeted TNF with melphalan or other chemotherapeutic drugs.

It has been suggested that the therapeutic index of TNF should be increased by more than 5–25-fold to be of value for a systemic use (43). The increase of at least 5-fold observed in our experimental models suggests that if a suitable targeting system is available, then pretargeting could be a realistic possibility to improve the therapeutic index of TNF in patients.

We thank Dr. F. Magni for mass spectrometry analysis of biotin-TNF conjugates and Dr. G. Paganelli for helpful suggestions.