Chimeric form of tumor necrosis factor-α has enhanced surface expression and antitumor activity

Abstract

Tumor necrosis factor (TNF)-α is a type-II transmembrane protein that is cleaved by TNF-α-converting enzyme (TACE/ADAM-17) to release soluble TNF, a cytokine with potent antitumor properties whose use in clinical applications is limited by its severe systemic toxicity. We found that human cells transfected with vectors encoding TNF without the TACE cleavage site (ΔTACE-TNF) still released functional cytokine at substantial levels that varied between transfected cell lines of different tissue types. Vectors encoding membrane-associated domains of CD154, another TNF-family protein, conjoined with the carboxyl-terminal domain of TNF, directed higher-level surface expression of a functional TNF that, in contrast to ΔTACE-TNF, was resistant to cleavage in all cell types. Furthermore, adenovirus vectors encoding CD154-TNF had significantly greater in vivo biological activity in inducing regression of established, syngeneic tumors in mice than adenovirus vectors encoding TNF, and lacked toxicity associated with soluble TNF. As such, CD154-TNF is a novel TNF that appears superior for treatment of tumors in which high-level local expression of TNF is desired.

Introduction

Tumor necrosis factor (TNF)-α was initially described as a 17-kDa soluble cytokine that could induce local and systemic tumor regression.¹ However, clinical trials revealed that the soluble cytokine had a low therapeutic index, owing to its toxicity when present at therapeutic concentrations in the systemic circulation.^2–4

Soluble TNF is released from the surface of cells expressing the 26-kDa proform of TNF through the TNF-α-converting enzyme (TACE/ADAM-17),^5,6 a matrix metalloproteinase that recognizes a cleavage site in the extracellular domain of the full-length TNF molecule.⁷ Deletion of this cleavage site results in the generation of a TNF molecule that has enhanced membrane stability.^7–9 Mice made transgenic for this membrane-stabilized form of TNF produced significantly higher levels of interleukin-12 than wild-type mice or TNF-deficient mice after administration of lipopolysaccharide, suggesting that the transmembrane form of TNF might affect cellular immune responses in vivo and have quantitatively and qualitatively distinct functions from that of secreted TNF in vitro and in vivo.^9–11 However, cells expressing such truncated forms of TNF lacking the TACE cleavage site still could release active TNF.⁹ Moreover, enzymes other than TACE could cleave the 26-kDa TNF proform at sites other than that recognized by TACE, to release a soluble TNF.¹² As such, the interpretation that membrane TNF has biological activity distinct from that of soluble TNF has to be tempered in light of the fact that soluble TNF still might be released in the microenvironment of cells expressing such modified forms of TNF.

To address the hypothesis that membrane-bound TNF elicits a biological response distinct from that of the soluble molecule, we generated chimeric forms of TNF that are highly resistant to release from the cellular membrane. This was accomplished by fusing the active ligand-binding portion of the TNF molecule with the transmembrane and proximal extracellular domains of TNF family proteins that primarily are expressed as cell-surface molecules (for example, CD154 or CD70). We generated adenovirus vectors encoding wild-type TNF (Ad-wtTNF), TNF lacking the defined TACE site (Ad-ΔTACE-TNF) and chimeric forms of TNF with CD154 (Ad-CD154-TNF) or CD70 (Ad-CD70-TNF), and examined cells infected with such constructs for expression of functional cell-surface TNF relative to that of the soluble cytokine. Furthermore, we compared the ability of CD154-TNF versus wtTNF to induce an antitumor response against syngeneic fibrosarcoma and lymphoma in vivo.

Materials and methods

Cell lines and mice

HeLa (human cervical carcinoma), HT1080 (human fibrosarcoma), A549 (human lung), HCT-15 and COLO-205 (human colon), DU-145 (human prostate), RPMI-8226 (human myeloma), HT1376 (human bladder), L929 (mouse fibroblast), WEHI-164 (mouse fibrosarcoma) and A20 (mouse B cell lymphoma) were obtained from ATCC, Manassas, VA. The cell line 293AC2, a subclone of 293, was obtained from Molecular Medicine Inc. (San Diego, CA). All cell lines except for 293AC2 were cultured in RPMI-1640 containing 10% fetal calf serum (FCS) and 2mM L-glutamine and maintained in a 5% CO₂ incubator. 293AC2 cells were cultured in DMEM containing 10% FCS and 2mm l-glutamine and maintained in a 10% CO₂ incubator.

Female BALB/c and FVB/N mice, 6–8 weeks of age, were purchased from Charles River Laboratories (Wilmington, MA) and housed at the University of California, San Diego (UCSD) animal facility. All animal studies were approved by the Animal Subjects Committee and the Biosafety Committee of the UCSD School of Medicine and were performed in accordance with institutional guidelines.

Chimeric gene construction

We generated chimeric genes using cDNA encoding either the mouse or human protein. Experiments conducted with human cells used constructs encoding the human protein and studies performed with mouse tumor-cell lines or in vivo used constructs encoding the murine protein.

The human CD154-TNF chimera was designed as follows. DNA encoding a fragment of human CD154 spanning the intracellular, transmembrane and partial extracellular region adjacent to the transmembrane region of CD154 was amplified from the full-length CD154 cDNA by PCR using the following primers (5′-GACAAGCTTATGATCGAAACATACAACC and 5′-TCAGGATCCTCATCTTTCTTCG). EcoRI and BamHI restriction enzyme sites were added at the 5′ and 3′ ends of this fragment. DNA encoding the extracellular region of human TNF distal to the TACE cleavage site was also PCR-amplified from the full-length TNF cDNA using the following primers: 5′-CCGGATCCTGTAGCCCATGTTGTAGCAAACC and 5′-GTTTCTAGATTCACAGAGCAATGATTCCAAAG. XbaI and BamHI restriction enzyme sites were added to the ends of this fragment. The CD154 and TNF DNA fragments were digested with the restriction enzymes described above followed by coligation into the EcoRI and XbaI sites of pcDNA3 (Invitrogen, San Diego, CA).

The mouse CD154-TNF chimera was constructed as follows. A mouse CD154 fragment homologous to the human fragment described above was released from a plasmid containing full-length mouse CD154¹³ with EcoRI and BamHI. This was ligated into pcDNA3 with the PCR-amplified extracellular domain of murine TNF generated with the following primers: 5′-CCGGATCCTGTAGCCCATGTTGTAGCAAACC and 5′-GTTTCTAGATTCACAGGGCGATGATTCCAAAG. The putative matrix metalloproteinase cleavage site of CD154 was then deleted from this construct using Stratagene’s QuickChange site-directed mutagenesis kit according to the manufacturer’s instructions with the following primers: 5′-CAAAGAAGAGAAAAAAGAAGATGAGGATCCTGTAGCCC and 5′-GGGCTACAGGATCCTCATCTTCTTTTTTCTCTTCTTTG.

The ΔTACE-TNF constructs were created by deletion of the nucleotides corresponding to the first 12 amino acids of the mature TNF, that is, amino acids 77–88 of the full-length human and 80–91 of the mouse protein. Using the QuickChange site-directed mutagenesis kit (Stratagene, San Diego, CA), we deleted the nucleotide sequence 5′-GTCAGATCATCTTCTCGAACCCCGAGTGACAAGCCT or 5′-CTCAGATCATCTTCTCAAAATTCGAGTGACAAGCCT, from the human or mouse TNF coding sequences, respectively. All clones were validated by nucleic acid sequence analysis. The proteins encoded by each contruct was tested for its ability to bind specific anti-TNF monoclonal antibodies (mAbs), as assessed by flow cytometry.

Adenovirus synthesis

pcDNA3 plasmids containing the transgene of interest were digested with the restriction enzymes NruI and SmaI to release a DNA fragment containing the CMV promoter from pcDNA3, the transgene and the polyadenylation signal from pcDNA3. Following gel purification of this fragment by electrophoretic separation in a 1% agarose gel, the DNA fragment was ligated into the EcoRV site of the adenoviral shuttle vector MCS (SK) pXCX2. This plasmid is a modification of the plasmid pXCX2 such that the pBluescript polylinker sequence has been cloned into the E1 region (JR Tozer, UCSD, unpublished data, September 1993). A quantity of 5 μg each of the transgene containing MCS (SK) pXCX2 plasmid and JM17 plasmid were then cotransfected into 293AC2 cells using the calcium phosphate Profection Kit from Promega according to the manufacturer’s instructions. Following transfection, the cells were cultured for 5 days to allow for homologous recombination and viral synthesis. Total cells and supernatant were then harvested and freeze–thawed thrice to release cell-associated adenovirus. Clonal isolates of the adenovirus were then obtained by plaque purification as previously described.

Large-scale adenovirus preparations were prepared by successively infecting increasing quantities of 293AC2. The large-scale crude adenovirus preparations were then purified over cesium chloride step gradients. This method makes use of a cesium chloride gradient for concentrating virus particles through a step gradient, with the densities of 1.45 and 1.20 g cm⁻³, in which 293AC2-expanded virus samples are centrifuged for 2 h in a SW40 rotor (Beckman, Brea, CA) at 25 000 r.p.m. at 4 °C The virus band was isolated using a 27-gauge needle and syringe and desalted using a Sephadex G-25 DNA grade column (Pharmacia, Piscataway, NJ). The virus was desalted against phosphate-buffered saline containing 10% glycerol and stored at −70 °C. The final titer of the virus was determined by anion-exchange HPLC.

Flow cytometry

Cells were washed and suspended in staining buffer, composed of phosphate-buffered saline containing 3% FCS, 0.05% sodium azide and 10 μg ml⁻¹ propidium iodide. Cells were incubated with saturating amounts of fluorochrome-conjugated mAb for 30 min at 4 °C. Antibodies utilized included phycoerythrin-conjugated antibody specific for TNF (clone Mab11, BD Biosciences, San Diego, CA) and phycoerythrin-conjugated isotype control antibody (clone MOPC-21). Stained cells were washed twice with staining buffer and analyzed using a FACSCaliber flow cytometer (Becton Dickinson, San Jose, CA). Dead cells and debris were excluded from the analysis by light scatter and by gating out cells that failed to exclude propidium iodide. We determined the percentages of cells that expressed the transgene product and the mean fluorescence intensity ratio of transgene-expressing cells to assess the relative expression levels of the transgene. The mean fluorescence intensity ratio is calculated by dividing the mean fluorescence intensity of cells that were stained with the fluorochrome-conjugated antigen-specific mAb by the mean fluorescence intensity of cells that were stained with a control fluorochrome-conjugated mAb of the same isotype but of irrelevant specificity.

Transient transfection

A total of 5 × 10⁵ cells were transiently transfected with 2 μg of plasmid, previously purified using Qiagen’s (Valencia, CA) Maxi DNA purification kit, using Lipofectamine 2000 (Invitrogen, San Diego, CA) according to the manufacturer’s instructions.

ELISA quantitation of soluble TNF

Two days following infection of cells with adenovirus, the cell supernatant was harvested and cleared of debris by centrifugation for 10 min at 200 g. In some assays, cells were infected with adenovirus in the presence of 100 μM of either the generalized matrix metalloproteinase inhibitor GM6001 (Calbiochem, San Diego, CA) or its negative control (Calbiochem). Soluble TNF was measured by a sandwich enzyme-linked immunosorbent assay (ELISA). Capture (clone Mab1, unlabeled) and detection (clone Mab11, biotinylated) antibodies for human TNF were both obtained from BD Biosciences. Capture (clone G281-2626, unlabeled) and detection (clone MP6-XT3, biotinylated) antibodies for mouse TNF were also obtained from BD Biosciences. Specific quantities of TNF were calculated on the basis of titrations of a known quantity of recombinant soluble TNF (Biosource International). Horseradish peroxidase avidin-biotin conjugated substrate (ABC kit, Vector Laboratories, Burlingame, CA) and TMB peroxidase substrates (Kirkegaard Perry Laboratories) were used for enzymatic color development according to the manufacturers’ instructions. The 450 nm optical densities of developed ELISA plates were measured using a multiwell plate reader.

Bioassay for soluble TNF in culture supernatant

This assay was based on a previously described assay for measuring TNF activity.¹⁴ Briefly, L929 cells were plated in a 96-well flat-bottomed tissue culture plate at 4 × 10⁴ cells per well followed by addition of serial dilutions of cell supernatant previously cleared of dead cells and debris by microcentrifugation for 5 min at 400 g. In addition, actinomycin D (Sigma, St Louis, MO) was added to a final concentration of 1 μg ml⁻¹. The final culture volume was 100 μl per well. Following 18 h incubation in a 5% CO₂ 37 °C incubator, 50 μl per well of XTT reagent (Roche Applied Science, Indianapolis, IN) was added followed by another hour incubation period to allow color development. The 450 nm optical densities of developed plates were measured using a multiwell plate reader.

Assay for bioactivity of cell-surface TNF

The functional activity of membrane-bound TNF was determined in a coculture assay using L929 cells and HeLa cells infected with recombinant Ad virus encoding TNF. HeLa cells were infected with Ad virus at a multiplicity of infection (MOI) of 10 for 24 h, the infected cells were washed with PBS, detached from the culture plate with PBS containing 10 mM EDTA, and washed once with complete media. HeLa cells were labeled with CFSE (Vybrant CFDA SE Cell Tracer Kit, Molecular Probes, Eugene, OR) according to the manufacturer’s recommendations. A total of 5 × 10⁴ HeLa cells and an equal number of L929 cells were then plated in a 12-well culture plate, either mixed together or, for transwell separation experiments, HeLa cells were added to the upper chamber of a 0.4 μm transwell insert to prevent direct cell–cell contact with the L929 cells. Actinomycin D (Sigma) was added to the culture medium for a final concentration of 1 μg ml⁻¹. Following 18 h culture, the cells were stained with 10 μg ml⁻¹ propidium iodide and the viability of L929 cells was determined by flow cytometry and gating on the CFSE^negative/PI^bright cells. The antibody neutralization experiments were similarly performed with the following exceptions: 5 × 10³ HeLa cells transduced at an MOI of 0.3 were plated with 5 × 10⁴ L929 cells. To neutralize TNF activity, the purified mouse IgG₁ antihuman TNF mAb (clone MAb1, BD Biosciences) was added to a final concentration of 10 μg ml⁻¹. Replicate control wells had 10 μg ml⁻¹ of mouse IgG₁ of irrelevant specificity.

ELISPOT-assay

Ninety-six-well PVDF filter plates (Multiscreen IP, Millipore, Bedford, MA) were coated overnight at 4 °C with an antimouse interferon (IFN)-γ antibody (clone R4-6A2, BD Biosciences, San Diego, CA) at 4 μg ml⁻¹ in sterile PBS. The plates were washed two times and then blocked with RPMI (Roswell Park Memorial Institute) 1640 medium containing 10% FCS, for at least 1 h at 37 °C. Splenocytes from treated mice were isolated. A total of 2 × 10⁵ splenocytes were then plated out into separate wells of an enzyme-linked immunosorbent spot assay (ELISPOT) plate and then either cultured alone or cocultured with mitomycin C (Sigma Chemical Co.) treated A20 stimulator cells in a 1:1 cell ratio at a total culture volume of 200 μl. The plates were incubated for 24 h at 37 °C in a humidified atmosphere of 5% CO₂. The plates were then washed three times with PBS, followed by three washes with PBS containing 0.05% Tween-20. To each well, 100 μl of biotinylated anti-IFN-γ mAb (clone XMG1.2, BD Biosciences, San Diego, CA) was added at a final concentration of 2 μg ml⁻¹ in PBS/10% FCS, and the plates were incubated for 2 h at room temperature. Following three washes with PBS/0.05% Tween-20, 100 μl of a 1:500 dilution of horseradish peroxidase-conjugated streptavidin in PBS/10% FCS was added to each well. The plates were incubated for 60 min at room temperature and then washed three times with PBS/0.05% Tween-20 followed by three washes with PBS. Fresh 3-amino-9-ethylcarbazole (AEC) substrate was prepared by dissolving 4 mg of 3-amino-9-ethylcarbazol (Sigma Chemical Co.) in 1 ml of dimethylformamide and then diluting this into 9 ml of sodium acetate buffer, pH 5.0, that contained 5 μl of freshly added 30% hydrogen peroxide. A total of 100 μl of the sterile-filtered AEC substrate was then added to each well and the plates were incubated for 5 min or more at room temperature. The substrate solution was discarded and the plates were rinsed with water and air-dried. After the plates were completely dry, spots were counted using the Immuno-Spot image analyzer and software (Cellular Technology Ltd, Cleveland, OH), and triplicate wells were used to calculate the average number of spots ± the standard error (s.e.).

Animal and tumor studies

WEHI-164 and A20 cell doses have been titrated in preliminary experiments such that subcutaneous implantation of 3 × 10⁶ WEHI-164 or 1 × 10⁵ A20 cells, respectively, would induce palpable tumors in BALB/c mice in approximately 1 week in 100% of the injected animals. For WEHI-164 tumor studies, mice were injected subcutaneously into the left, shaved flank with 3 × 10⁶ cells in 100 μl of PBS. Approximately 8 days later, when tumors reached a mean diameter of 7 mm, tumors were injected three times in 4-day intervals with 10⁹ plaque forming units (PFU) adenovirus (in 100 μl). For A20 tumor studies, mice were injected subcutaneously into the left, shaved flank with 1 × 10⁵ A20 cells in 100 μl of PBS. When tumors reached a mean diameter of 4–5 mm, mice received a one-time intratumoral injection of 5 × 10⁸ PFU adenovirus in 100 μl of PBS. Tumor progression was monitored biweekly over the course of the experiment by measuring the length and width using calipers. Tumor volumes were calculated by the following formula: volume = 0.5 × length × (width)². Mice were sacrificed when tumor length exceeded 20 mm. The log-rank test was performed for survival analysis and differences in tumor volume were compared using Student’s t-test. P-values of <0.05 were considered significant. Analyses were performed with Prism software (GraphPad, San Diego, CA). For in vivo toxicity studies, mice were injected twice intraperitoneally with 10⁹ PFU of adenovirus at a 48 h interval. Blood was collected from animals 24 h after the second injection.

Results

Generation of Soluble TNF

We generated expression vectors encoding wtTNF (pwtTNF) or ΔTACE-TNF (pΔTACE-TNF), TNF lacking the putative TACE cleavage site (Figure 1). Cells transfected with pΔTACE-TNF unexpectedly released substantial amounts of functional TNF, although at levels that were significantly less than that of cells transfected with pwtTNF. The culture supernatants of HT1080 cells transfected with either pwtTNF or pΔTACE-TNF, but not control pcDNA3, induced apoptosis of L929 cells, a TNF-sensitive mouse fibroblast cell line¹⁴ (Figure 2a). Upon further investigation into the expression of ΔTaCE-TNF, we found that soluble TNF was released from various human cell lines infected with adenovirus encoding ΔTACE-TNF (Ad-ΔTACE-TNF) (Table 1). Regardless of the cell line used, we found that soluble TNF was released by Ad-ΔTACE-TNF-infected cells, although at significantly lower levels (P<0.05, Student’s t-test) than that released from Ad-wtTNF-infected cells (Table 1). No detectable TNF was produced from any of the cell lines before transduction or after transduction with a control adenovirus vector. Interestingly, the difference in the amounts of soluble TNF released by cells expressing comparable levels of wtTNF versus ΔTACE-TNF varied for the different cell lines tested and ranged from an approximate 3-fold difference for the lung cell line A549 to greater than a 200-fold difference for the bladder cell line HT1376. Furthermore, addition of the broadly active matrix metalloproteinase inhibitor GM6001 to HeLa cells transfected with Ad-ΔTACE-TNF did not prevent these cells from releasing soluble TNF (11 000 pg ml⁻¹ ± 100 with inhibitor versus 52 000 pg ml⁻¹ ± 500 without inhibitor, mean ± s.e., n = 3). Collectively, these studies indicate that ΔTACE-TNF still has a cleavage site(s) that allows for release of soluble TNF from cells transduced with this truncated form of TNF.

Figure 1.

Chimeric TNF constructs. Scale representation of the chimeric constructs of CD154, TNF and CD70 indicating the membrane anchoring domains (MAD) and ligand binding domains (LBD). The boundaries of TNF-binding domain are identical in the chimeric constructs. The human and murine sequences display a high degree of similarity and differ in length by no more than two amino acids for homologous sequences.

Figure 2.

Expression of TNF by transfected cells. (a) Secretion of soluble TNF from ΔTACE-TNF-transfected cells. HT1080 cells were transfected with pcDNA3 plasmids encoding vector alone (square), wtTNF (circle), or ΔTACE-TNF (triangle) and incubated for 2 days. Serial dilutions of cell supernatants were then incubated with L929 cells and soluble TNF bioactivity determined by the measurement of L929 apoptosis using the XTT colorimetric assay. (b) HT1080 cells were infected with adenovirus as indicated for two days. We examined the culture supernatants for soluble TNF by ELISA. (c) Cell-surface expression of TNF by Ad-infected HeLa cells. HeLa cells were infected with the indicated adenovirus and then analyzed for TNF surface expression by flow cytometry (open histograms). Shaded histograms represent cells stained with isotype control antibody. The percentage of surface-TNF-positive cells is indicated for each histogram. (d) The relationship of soluble TNF versus the mean fluorescence intensity ratio of TNF surface expression was plotted for HT1080 cells (left graph) and HeLa cells (right graph) infected at increasing MOI with adenovirus encoding wtTNF (closed squares), ΔTACE-TNF (open squares), CD70-TNF (open triangles) and CD154-TNF (closed circles). Data points from four different MOI for each TNF construct were used to determine the linear functions plotted on the graph.

Table 1.

Soluble TNF generation following Ad-TNF infection

Name	Type	Soluble TNFa
		wtTNF	ΔTACE-TNF	CD154-TNF
HT1080	Fibrosarcoma	43100±3000	4000±100	<40
A549	Lung	48200±2500	14 200±900	ND
HeLa	Cervical	640000±600	52000±500	<40
RPMI-8226	Myeloma	21 400±1600	4500±200	<40
HT1376	Bladder	60200±300	300±10	<40

Abbreviation: ND, not determined.

^aExpressed as pg ml⁻¹ ± s.d.

Cell lines were infected at an MOI of 10 for 2 days followed by analysis of soluble TNF by ELISA.

To explore this possibility, we generated a chimeric construct that fuses the extracellular carboxyl-terminal domain of TNF with the membrane spanning portion of CD154 that is devoid of putative matrix metalloproteinase cleavage sites.¹⁵ We compared the generation of soluble TNF by cells infected with Ad-CD154-TNF to cells infected with either Ad-wtTNF or Ad-ΔTACE-TNF (Figure 2b). To control for nonspecific TNF secretion by cells, we also examined cells that either were infected with control adenovirus encoding the irrelevant transgene β-galactosidase (Ad-LacZ) or that were not infected (noninfected). We found that Ad-CD154-TNF-infected HT1080 cells produced significantly less soluble TNF (<40 pg ml⁻¹, below detection limit of ELISA assay, P<0.05, Student’s t-test) than cells infected with adenovirus encoding either wtTNF (43 100 ± 3000 pg ml⁻¹) or ΔTACE-TNF (4000 ± 100 pg ml⁻¹) (Table 1). Infection of other cell lines with Ad-CD154-TNF, including HCT-15 and C0L0–205 (human colon), DU-145 (human prostate) and 293AC2 (human kidney) generated cells that had high-level cell-surface expression with negligible release of soluble TNF (data not shown).

Expression of surface membrane TNF

We also examined for the cell-surface expression of TNF on cells transduced with Ad-wtTNF, Ad-ΔTACE-TNF or Ad-CD154-TNF (Figure 2c). In addition, we analyzed cells infected with adenovirus encoding a different chimeric molecule composed of the TNF extracellular domain fused to the amino-terminal portion of CD70 (Ad-CD70-TNF), comprised of a shorter extracellular stalk region (5 amino acids) than CD154-TNF (65 amino acids), to determine if expression results were specific for CD154-TNF. Each gene construct shares the same ligand-binding portion of the TNF molecule. At MOI ratios of 1–10, Ad-CD154-TNF-infected cells expressed higher cell-surface levels of TNF than cells infected with adenovirus encoding any one of the other Ad-TNF constructs (Figure 2c, and data not shown).

We performed linear regression analyses on the relationship between released TNF and cell-surface TNF observed for cells infected at various MOI with each of the Ad-TNF constructs. In each case, we could define a linear relationship between the relative expression levels of soluble TNF versus cell-surface TNF for each cell type following infection at various MOIs. We compared the different Ad-TNF constructs by calculating the slope of the line that defined each relationship, with steeper slopes indicating production of greater amounts of soluble cytokine relative to that of cell-surface TNF. We found that the relative steepness of slopes for the cells infected with each of the Ad-TNF constructs had the following relationship: Ad-wtTNF>Ad-ΔTACE-TNF>Ad-CD70-TNF>Ad-CD154-TNF. For example, the slopes for HT1080 cells infected with the Ad-TNF constructs were as follows: Ad-wtTNF (slope = 1300; R² = 0.8) >Ad-ΔTACE-TNF (slope = 290; R² = 0.8) >Ad-CD70-TNF (slope >15; R² = 0.9) >Ad-CD154-TNF (slope = 0; R²>0.9) (Figure 2d). Likewise, the slopes for infected HeLa cells were as follows: Ad-wtTNF (slope = 4600; R²>0.9) >Ad-ΔTACE-TNF (slope = 46; R²>0.9) >Ad-CD70-TNF (slope >10; R²>0.9) >Ad-CD154-TNF (slope = 0; R²>0.9) (Figure 2d). Furthermore, these relationships are consistent with the values reported for the fold difference in amount of soluble TNF released from cell lines infected with Ad-wtTNF compared with cells infected with either Ad-ΔTACE-TNF or Ad-CD154-TNF (Table 1).

In vitro activity of CD154-TNF

To examine whether the chimeric CD154-TNF construct was functional, we cocultured TNF-sensitive L929 cells with HeLa cells that were infected with adenovirus encoding either wtTNF or CD154-TNF (Figure 3a). We separated the L929 and HeLa cells from direct cell–cell contact using a 0.4 μm transwell insert. As such, apoptosis of L929 should result only from diffusion of soluble TNF from the HeLa cells across the membrane to the L929 cells. As expected, Ad-wtTNF-infected HeLa cells induced apoptosis of L929 even when separated by a transwell membrane. In contrast, Ad-CD154-TNF-infected HeLa cells did not induce apoptosis of L929 cells more than control-infected HeLa cells when separated from the L929 cells across the transwell membrane (Figure 3a). The observed activity was due to TNF signaling because the killing of L929 cells could be blocked specifically by adding a neutralizing mouse antihuman TNF mAb to the culture medium before the assay (data not shown).

Figure 3.

In vitro and in vivo toxicity of soluble and cell-surface TNF. (a) Contact-dependent killing of L929 cells by Ad-TNF-infected HeLa cells. HeLa cells were infected with adenovirus encoding the indicated transgene (GFP, wtTNF or CD154-TNF) and then either stained with CFSE and plated out with an equal number of L929 cells, to allow cell–cell contact (coculture, filled bars) or HeLa cells were placed in the upper well of a transwell filter insert (transwell, empty bars) to prevent direct cell–cell contact with L929 cells. Following 18 h coculture, L929 cells were analyzed for viability by flow cytometry as described in Material and methods. (b) Hemorrhagic tumor necrosis after injection of Ad-TNF. BALB/c mice with WEHI-164 tumors were injected intratumorally with 10⁹ PFU Ad-blank, Ad-wtTNF or Ad-CD154-TNF. Pictures were taken 1 week after injection and are representative for each group. (c) Toxicity after systemic injection of Ad-TNF. Groups of eight FVB/N mice were injected intraperitoneally twice in a 48 h interval with 10⁹ PFU adenovirus encoding wtTNF (closed triangle), CD154-TNF (closed diamond) or adenovirus containing no transgene (open square). Survival of animals is plotted over time.

Although the portion of CD154 used for constructing CD154-TNF is not known to be responsible for signaling, we specifically tested for the ability of CD154-TNF to make a functional interaction with CD40. We infected HeLa cells with Ad-CD154 or Ad-CD154-TNF, confirmed high-level expression of each transgene by flow cytometry 24 h post-infection and subsequently cocultured BALB/c splenocytes with the HeLa cells overnight. Measuring CD40 surface expression on splenocytes by flow cytometry revealed that CD154, but not CD154-TNF, decreased the amount of detectable CD40 expression (data not shown), indicating that the CD154 portion of the TNF fusion protein does not functionally interact with CD40.

In vivo toxicity of Ad-CD154-TNF versus Ad-wtTNF

We examined the local and systemic toxicity of Ad-CD154-TNF relative to that of Ad-wtTNF. First, we injected 10⁹ PFU Ad-wtTNF, Ad-CD154-TNF or Ad-blank into WEHI-164 fibrosarcoma tumors 5 mm in diameter that developed from subcutaneous injections of tumor cells into the hind flanks of BALB/c mice. Injection with Ad-wtTNF resulted in the formation of hemorrhagic ulcers surrounding the tumor within 2 days after the injection in all the treated animals (Figure 3b). The cutaneous ulcers of Ad-wtTNF-treated mice were most pronounced around the injection site and regressed approximately 2–3 weeks after the injection. Similar cutaneous ulcers were observed when Ad-wtTNF was injected subcutaneously into the flank of mice that were not bearing tumors (data not shown). However, mice that received intratumoral injections or subcutaneous injections of Ad-blank or Ad-CD154-TNF did not develop any cutaneous ulcerations (Figure 3b).

To examine the relative toxicity of each of the adenovirus vectors, we injected 10⁹ PFU of either virus vector into the peritoneal cavity of FVB/N mice. Injection of mice with Ad-wtTNF resulted in systemic reactions 12 h after the injection, including lethargy, ruffled fur and inflammation around the eyes. Death occurred in 25% of the treated animals within 48 h of injection (Figure 3c). In contrast, mice injected with Ad-CD154-TNF or Ad-blank did not demonstrate any signs of toxicity. We injected the surviving mice in each group with the same vector 48 h after the first injection. All the mice that received a second injection of Ad-wtTNF died, whereas none of the animals injected twice with Ad-CD154-TNF or Ad-blank displayed any signs of systemic toxicity (Figure 3c). Twenty-four hours after the second injection, we detected TNF in the sera of mice treated with Ad-wtTNF (13 300 ± 800 pg ml⁻¹) but not in the sera of mice injected with Ad-CD154-TNF or Ad-blank (data not shown).

Antitumor activity of Ad-CD154-TNF versus Ad-wtTNF

Having established the decreased systemic toxicity resulting from the elimination of soluble TNF, we tested whether CD154-TNF could induce an antitumor response in vivo. In one model system, BALB/c mice were each injected subcutaneously with a sufficient number (3 × 10⁶) of syngeneic WEHI-164 fibrosarcoma cells to form palpable tumor nodules within 1 week after injection. These tumor nodules each were injected three times at 4-day intervals with 10⁹ PFU of Ad-CD154-TNF, Ad-wtTNF or Ad-GFP, a control adenovirus encoding green fluorescence protein. In contrast to animals treated with Ad-GFP, the animals treated with injections of either Ad-wtTNF or Ad-CD154-TNF experienced an antitumor effect (Figure 4a). However, mice injected with Ad-wtTNF developed hemorrhagic ulcers around the injection site (Figure 3b). In spite of this, the injected tumors continued to grow in most of the Ad-wtTNF-treated animals. Injection with Ad-CD154-TNF, however, caused the tumors to regress in most animals. Moreover, the tumors of Ad-CD154-TNF-treated animals had a significantly smaller average size than the tumors of animals treated with Ad-wtTNF at 5 weeks (89 ± 80 mm³ or 710 ± 280 mm³, respectively) and 6 weeks (220 ±190 or 1230 ± 390 mm³, respectively) after tumor challenge (P<0.05, Student’s t-test, n = 8 for all groups) (Figure 4a). Sixteen weeks following the initial tumor challenge, a significantly greater proportion of the mice injected with Ad-CD154-TNF survived (80%) than did mice treated with Ad-wtTNF (47%) (P = 0.02, log rank test) (Figure 4b).

Figure 4.

Intratumoral injection of WEHI-164 tumor nodules. BALB/c mice were injected subcutaneously with 3 × 10⁶ WEHI-164 tumor cells at day 0. On days 8, 12 and 16, tumor nodules were injected with 10⁹ PFU adenovirus encoding GFP (open triangles, n = 15), wtTNF (closed triangles, n = 28), CD154-TNF (closed diamonds, n = 27) or saline (open squares, n = 8). (a) Mean tumor size and (b) animal survival are plotted over time. Asterisks indicate statistically significant differences between CD154-TNF and wtTNF treatment (Student’s t-test or log-rank test, respectively).

We also examined the activity of these vectors using the murine A20 lymphoma tumor-model system. A20 cells do not express TNF receptor 1 (TNFR1) and are not directly sensitive to TNF-induced apoptosis (data not shown). For this, BALB/c mice were each injected subcutaneously with 1 × 10⁵ syngeneic A20 lymphoma B cells, resulting in the development of subcutaneous lymphoma nodules of 5 mm in diameter within 10 days in each animal. These nodules were each injected with 5 × 10⁸ PFU Ad-blank, Ad-wtTNF or Ad-CD154-TNF, or were left untreated. Again, subcutaneous injection of Ad-wtTNF caused cutaneous hemorrhagic ulcers around the injection site. However, in contrast to the WEHI-164 model system, the animals injected with Ad-wtTNF did not experience an antitumor effect relative to that of mice injected with the control adenovirus vector, Ad-blank. In contrast, the A20 tumors regressed in most animals treated with Ad-CD154-TNF and/or had significantly slower tumor growth compared with that of tumors injected with Ad-wtTNF (44 ± 21 versus 362 ± 121 mm³ at week 4.5, P = 0.012, Student’s t-test, Figure 5a). Furthermore, a significantly higher percentage of mice treated with Ad-CD154-TNF survived than did animals treated with Ad-wtTNF (75 versus 33%, P = 0.019, log rank test, Figure 5b).

Figure 5.

Intratumoral injection of A20 lymphoma nodules. BALB/c mice were injected subcutaneously with 1 × 10⁵ A20 tumor cells at day 0. On day 10, tumor nodules remained untreated (closed square, n = 10) or were injected with 5 × 10⁸ PFU adenovirus carrying no transgene (Ad-blank, open square, n = 16), Ad-wtTNF (open triangle, n = 15) or Ad-CD154-TNF (closed diamond, n = 16). (a) The mean tumor size (±s.e.) and (b) animal survival are plotted over time. Asterisks indicate statistically significant difference between CD154-TNF and wtTNF treatment (Student’s t-test or log-rank test, respectively). (c) Immune response to A20 cells in treated mice analyzed by IFN-γ ELISPOT assay. Splenocytes were isolated from three mice in each treatment group approximately 2 weeks after injection of Ad. Splenocytes were then cocultured with an equal number of mitomycin C-treated A20 cells in a 96-well filter plate and the number of IFN-γ-secreting spleen cells was determined by ELISPOT assay. Bars depict the mean number (±s.d.) of IFN-γ-producing cells per 2 × 10⁵ splenocytes. The asterisk indicates a statistically significant difference (P<0.05, Bonferroni t-test) in the mean number of spots between the Ad-CD154-TNF group and all other treatment groups. (d) Mice with complete tumor regression after treatment with Ad-wt-TNF (open circle, n = 4) or Ad-CD154-TNF (cross, n = 7) were rechallenged with 2 × 10⁵ A20 cells in the contralateral flank. A group of naïve mice (closed squares, n = 5) also received 2 × 10⁵ A20 cells. The mean tumor size (± s.e.) for each group is depicted over time.

We hypothesized that injection of A20 tumors with the various adenovirus vectors enhanced development of an A20-specific cellular immune response. To test this, we isolated splenocytes from tumor-bearing mice 2 weeks after they were injected with adenovirus vector. Splenocytes were cocultured with mitomycin-C-treated A20 cells and the numbers of IFN-γ-secreting splenocytes were enumerated in a 24-h ELISPOT assay. We found that mice injected with Ad-CD154-TNF had a significantly higher number of IFN-γ-secreting, A20-specific splenocytes (P = 0.016, Bonferroni t-test) than mice injected with Ad-wtTNF or Ad-blank, or mice that did not receive any injections with adenovirus (Figure 5c).

Eleven weeks following the initial injection of A20 cells, mice that had rejected the tumor following injection of Ad-wtTNF or Ad-CD154-TNF were challenged with lethal inoculums of A20 tumor cells. In contrast to control mice, the animals that previously had rejected A20 tumors never developed tumor nodules and remained free of tumor (Figure 5d), demonstrating that such animals had protective immunity against subsequent challenge with A20.

Discussion

Tumor necrosis factor is an attractive candidate molecule for cancer therapy.^16,17 TNF can directly induce apoptosis of tumor cells by interaction with TNFR1 (p55).¹⁸ In addition, TNF is a potent immune-stimulatory molecule capable of inducing antigen-presenting cells to express high levels of immune costimulatory molecules and of enhancing both cellular and humoral immune responses.¹⁹⁻²¹ However, the use of TNF has been limited secondary to the serious toxicity of TNF in the systemic circulation.²²

Some investigators have devised strategies to mitigate the problems associated with the systemic release of soluble TNF through techniques such as hypothermic isolated limb perfusion of cells transduced in situ,^23,24 physically linking soluble TNF to the tumor-cell membrane,²⁵ using TNF in the form of a fusion molecule,^26,27 or by driving expression of TNF in transduced cells with weak, tissue-specific or radiation-inducible promoters.²⁷⁻³⁰ In addition, others have used vectors encoding a truncated TNF that lacks the TACE cleavage site.⁷⁻¹⁰ However, such strategies still have the potential to release soluble TNF into the systemic circulation.

The problem associated with use of the truncated TNF is highlighted by the data presented in this study. Indeed, although cells transduced with Ad-ΔTACE-TNF released significantly less soluble TNF than did cells transduced with Ad-wtTNF, some transduced cells still released significant amounts of soluble cytokine (for example, A549, Table 1). Because the TNF released by Ad-ΔTACE-TNF-transduced cells could induce apoptosis of cells sensitive to TNF-mediated apoptosis (Figure 2a), we conclude that TNF does have sites other than the TACE cleavage site that can be cleaved to release a functional soluble cytokine. This notion is supported by data from Marr et al.,³¹ who observed systemic toxicity in mice after injection of tumors with Ad-ΔTACE-TNF.

In light of the results with Ad-ΔTACE-TNF, we hypothesized that global modifications of TNF would be necessary to anchor TNF to the plasma membrane. Unlike TNF, many members of the TNF family are cell-surface molecules. Among these is CD154, a molecule expressed on the plasma membrane of activated T cells.^20,32 Although CD154 can be cleaved by metalloproteinases into a soluble molecule,³³ deletion of this site generates a stable cell-surface protein¹⁵ (and data not shown). As such, we predicted that the membrane-proximal domain of CD154 lacking this solitary cleavage site could provide a stable base for the presentation of the functional carboxyl terminal domain of TNF. In addition, structural analysis of TNF family members indicate a highly conserved tertiary structure of trimeric molecules.^34,35 Structural studies indicate that receptor-ligand interactions occur in the extracellular portion corresponding to the soluble form of TNF. In contrast, the intracellular, transmembrane and proximal extracellular domains of TNF family-member proteins are assumed not to participate in receptor binding, although direct structural evidence for this does not formally exist. This being the case, receptor binding and specificity might not be dependent on the amino-terminal domain. As such, we predicted that a chimeric TNF molecule should preserve TNF receptor binding, although at the same time potentially acquiring the more stable cell-surface expression properties of CD154.

Indeed, we found that a chimeric TNF construct composed of the extracellular, TNF receptor-binding portion of TNF fused to the amino-terminal region of CD154 enables stabile, high-level cell-surface expression of TNF. Importantly, high-level surface expression without production of soluble cytokine was observed for this chimeric molecule in all cell types tested. This is unlike what we observed for ΔTACE-TNF.

The differences in the amount of TNF released by cells infected with Ad-CD154-TNF versus those infected with Ad-wtTNF or Ad-ΔTACE-TNF are not due to differences in susceptibility to infection by each of the adenovirus vector preparations. Indeed, we observed that levels of surface TNF expression by cells infected with Ad-CD154-TNF at an MOI was lower than that required to effect even low-level surface expression of TNF with Ad-wtTNF or Ad-ΔTACE-TNF. Moreover, linear regression analyses comparing the levels of soluble versus cell-surface TNF revealed different slopes for each TNF construct between the different cell lines, as well as differences in the slopes between different TNF constructs. In all cases, infection with Ad-CD154-TNF resulted in high-level surface expression of TNF with negligible release of soluble TNF, resulting in a slope of zero (Figure 2d). Others have reported that removal of the TACE site is sufficient to abrogate detectable release of soluble TNF from murine hepatic carcinoma cells and that soluble and membrane-bound forms elicit distinct antitumor effects.¹⁰ The CD154-TNF represents a novel type of TNF that, in contrast to ΔTACE-TNF, can be expressed on the cell surface of a variety of different cell types without detectable release of TNF.

The observation that the ligand specificity of TNF was maintained when the carboxyl-terminal portion of TNF was conjoined with the transmembrane and proximal membrane domains of another member of the TNF family supports the notion that the intracellular, transmembrane and extracellular stalk region of TNF family members do not directly interact with their cognate receptors. However, such domains still might play a role in the metabolic fates of the expressed protein. This is supported by our studies using CD70-TNF, a construct that we hypothesized should also anchor TNF to the plasma membrane. Although the short stalk domain of CD70 apparently resisted proteolytic cleavage, we could not effect high-level surface expression of CD70-TNF. Even for cells infected with Ad-CD70-TNF at high MOI, the TNF surface expression was much lower than cells similarly infected with Ad-CD154-TNF. This indicates that the composition rather than the length of the stalk region may be important in governing the relative level of cell-surface expression.

Tumor necrosis factor can exert its antitumor effects in several ways. First, TNF can have a direct cytotoxic effect on TNF-sensitive tumor cells, which involves signaling through TNFR1 (p55).¹⁸ However, the antitumor activity of Ad-wtTNF or Ad-TNF-CD154 that we observed in our study cannot be only because of the direct cytotoxic effects of TNF on tumor cells, as A20 cells do not express TNFR1 and do not undergo apoptosis in response to soluble TNF in vitro (data not shown). On the other hand, TNF can induce apoptosis of TNFR1-expressing vascular endothelial cells.^36,37 Conceivably, some of the antitumor effect(s) of local TNF expression could be due to disruption of the tumor’s microvasculature. TNF also could act as a factor in the development of an antitumor immune response. TNF can stimulate maturation and activation of antigen-presenting cells, such as dendritic cells.³⁸ Moreover, TNF coupled with the membranes of apoptotic tumor cells appears more effective than soluble TNF in cross-priming T cells to tumor-associated antigens.³⁹ As such, tumor cells transduced to express CD154-TNF could be more efficient than tumor cells transduced to express wtTNF in inducing antitumor immunity.

In our studies, we observed a modest attenuation of tumor growth in the groups of animals injected with control adenovirus vectors lacking a relevant transgene. The effect was observed with both WEHI-164 and A20 tumors (see Figures 4 and 5). This could reflect inflammatory and innate immune responses to infection with adenovirus, which could be deleterious to tumor cell growth in vivo.^40,41 The tumor growth kinetics observed in animals injected with control adenovirus, however, were significantly greater than those observed in animals injected with adenovirus encoding TNF, particularly CD154-TNF. Nevertheless, it is possible that effects on tumor cell growth observed with adenovirus encoding TNF or CD154-TNF might be less dramatic if we employed other methods to effect tumor-cell expression of TNF or CD154-TNF.

Finally, Ad-CD154-TNF was less toxic than Ad-wtTNF. Injection of the latter into the skin or tumor nodules of experimental animals induced hemorrhagic ulcers that were not observed in animals injected with Ad-CD154-TNF, presumably due to the release of soluble TNF that can cause substantial destruction of endothelial cells, ultimately resulting in hemorrhagic tissue necrosis.^42,43 Moreover, even when delivered systemically by injection into the peritoneum, Ad-CD154-TNF appeared well tolerated, in contrast to Ad-wtTNF. These results are encouraging because CD154-TNF, in addition to enhanced biological activity over that of soluble TNF, could make TNF a viable candidate for use in clinical oncology. The use of TNF for human cancer treatment has been limited due to its poor systemic tolerance, resulting in an unacceptably low therapeutic index. However, the results reported here suggest that CD154-TNF could have a much higher therapeutic index in tumor-bearing patients, allowing TNF to play an effective role in anticancer therapy.