A Genetically Encoded Multifunctional TRAIL Trimer Facilitates Cell-Specific Targeting and Tumor Cell Killing

When a cell is instructed to die in a coordinated fashion, it is considered to undergo apoptosis, or programmed cell death (1). Apoptosis is an important mechanism during early development but is also vital to maintain homeostasis of the immune system throughout life (2). Apoptosis is induced by two mechanisms, namely, the extrinsic or death receptor pathway and the intrinsic or the mitochondrial pathway (3–6). The extrinsic death pathway is triggered by ligand-receptor interactions that lead to intracellular signaling events, which ultimately result in the death of the target cell. One such ligand is represented by a member of the tumor necrosis factor (TNF) superfamily, TNF-related apoptosis-inducing ligand (TRAIL or Apo2L; refs. 7, 8). TRAIL interacts with at least five endogenous receptors, four of which are cell membrane associated [TRAIL-R1/death receptor 4 (DR4), TRAIL-R2/DR5, TRAIL-R3/decoy receptor 1 (DcR1), and TRAIL-R4/DcR2], whereas the fifth receptor, osteoprotegerin, constitutes a fluid-phase protein found in the circulation (9), with the potential to support tumor growth by blocking TRAIL binding to its activating cell surface receptors (10). Although TRAIL binds to all of these receptors, only ligation with DR4 and DR5 has been shown to induce apoptosis. Importantly, TRAIL-based therapeutics have been shown to be more effective inducers of apoptosis in cancer cells than in normal cells (11, 12). The mechanism underlying this tumor selectivity, however, is not fully elucidated. The intrinsic cancer selectivity represents one of the main reasons for the attractiveness of this molecule as an anticancer drug (13).

Endogenous TRAIL exists as a homotrimer, a critical requirement for its biological function. Various expression systems have been explored, and much progress has been made to obtain biologically active TRAIL proteins such as nontagged and tagged (FLAG, His, etc., with or without tag-mediated crosslinking), inclusion of trimerization domains such as a leucine zipper and/or an isoleucine zipper, and stabilization of the trimers with cations (i.e., zinc; ref. 14). In all previous reports, the building block for trimeric, recombinant TRAIL has been expressed from monomer-encoding cDNA sequences and has required self-association into functionally active trimers.

Capitalizing on the inherent antitumor properties of TRAIL, we have developed and tested a novel TRAIL expression construct, designated TR3. Additional modifications showed its potential as an investigative tool and as a platform on which to build cell-targeted anticancer therapeutics.

Recombinant human TRAIL (aa 114-281) was purchased from Enzo Life Sciences (formerly BIOMOL International). Antihuman TRAIL polyclonal antibody (rabbit) was obtained from Peprotech. FITC-conjugated antihuman IgG was purchased from Sigma, the Annexin-V/FITC apoptosis kit was purchased from Biosource, and cell viability was determined using a luciferase-based readout (CellTiter-Glo, Promega).

The wild-type, NH₂-terminal ectodomain of human TRAIL (TR) described in this study contains amino acids 91 to 281 (domain I; compare Fig. 1A, including the striped box; the white box represents aa 114-281 of rTRAIL). A 594 bp DNA fragment amplified by PCR from a human U937 cDNA library (15) was inserted via BsiWI (5′) and HindIII (3′) into sT-DAF (16). This basic TRAIL plasmid, designated pTRBgl, contains an additional BglII site immediately upstream of TRAIL's native stop codon for subsequent cloning purposes. It also contains a signal peptide to ensure secretion of the protein. Following linearization of pTRBgl with BglII and HindIII, the slightly smaller PCR-derived domains I′ (5′ BamHI and 3′ HindIII), containing amino acids 108 to 281 of native TRAIL, were added stepwise resulting in pTR2 (intermediate) and pTR3, respectively. The NH₂-terminal enhanced green fluorescent protein (eGFP) extension of TR3 was introduced as an EcoRI/BsiWI PCR fragment into pTR3, resulting in the eGFP-TR3 fusion protein (Supplementary Fig. S1A). Similarly, scFv-TR3 was generated by inserting the 735 bp BsiWI scFv Ter-119 fragment from sT-DAF into the BsiWI-linearized pTR3 plasmid. The chimera comprising the complement regulatory proteins DAF [short concensus repeat 1 (SCR1)] and CR1 (SCRs 15–17), targeted to the mouse RBC membrane via scFv Ter-119, has been described earlier (Ter-CR1, D.S. and J.P. Atkinson, unpublished data). This cDNA served as a PCR template to generate a scFv-DAF-CR1 fragment that was flanked by the restriction sites XhoI (5′) and SnaBI (3′). This first fragment was then combined with a SnaBI/HindIII PCR-derived TR3 fragment and the XhoI/HindIII backbone of pSBC-1 (17), resulting in the spacer-containing scFv-S-TR3 form (Supplementary Fig. S1A). All PCR-derived DNA fragments were verified by DNA sequencing (Washington University Sequencing Core).

Human embryonic kidney cells [HEK293T, American Type Culture Collection (ATCC), CRL-11268] were used for protein generation. They were maintained in DMEM (Invitrogen) containing 10% FCS (Harlan). Media were supplemented with L-glutamine (Sigma), nonessential amino acids (BioWhittaker), and penicillin and streptomycin (Cellgro, Mediatech). The human T-cell line Jurkat (ATCC, TIB-152) and the pancreatic cancer cell line BxPC3 (ATCC, CRL-1687) were maintained in RPMI1640 medium (Invitrogen), supplemented with 10% FCS, L-glutamine, and penicillin/streptomycin.

The recombinant TR3 forms and soluble death receptor 5 (DR5-Fc, kindly provided by Thomas Griffith, University of Iowa) were prepared by transient expression in HEK293T cells using Gibco Opti-Mem serum-free medium and TransIT-293 (Mirus, MIR2700) transfection reagent, as per the manufacturer's instructions. To obtain concentrated TR3 protein stocks, the supernatants were applied to centrifugal filter devices with a 10 kDa molecular cutoff (Centricon Plus-20, Millipore). DR5-Fc was purified using Protein A columns as per the manufacturer's instructions (Pierce). Protein concentration was determined with a spectrophotometer using bovine serum albumin (New England Biolabs) as a standard.

C57BL/6 wild-type (WT) mice were used as an erythrocyte source and as recipients for recombinant TRAIL forms. Blood was collected from the tail vein using heparinized glass capillaries for coating experiments with scFv-(S)-TR3 and for the generation of plasma. Six- to eight-week-old male nude mice (nu/nu; Harlan) were used as hosts for tumor xenografts. Human BxPC3 tumor cells (5 × 10⁵ cells/animal) were injected s.c. into the right flanks of the mice along with 40 times of mouse RBCs [2 × 10⁷, effector:target (E:T) ratio = 40] in a total volume of 50 μL of FCS-free culture medium. The tumor size was measured with calipers using the following formula: Volume = 0.5 (length × width²) (ref. 18). Procedures involving mice were approved by the Washington University Animal Studies Committee and conducted in accordance with the guidelines for the care and use of laboratory research animals established by the NIH.

Several concentrations of mouse blood and TR3-containing culture supernatants were used as described in the text and figure legends. Briefly, 1 to 4 μL of mouse whole blood (∼1–4 × 10⁷ RBCs) were incubated with up to 1.5 mL of the filtered, HEK293T culture supernatant for 2 hours at room temperature, washed, and then processed for flow cytometry or subjected to coculture experiments with human Jurkat and BxPC3 cells. Supernatants from nontransfected HEK293T cells served as a control.

The killing capacity of our novel TR3 forms was routinely assessed using a morphology-based fluorescence-activated cell scan (FACS) assay using Jurkat reporter cells. This procedure gave identical results compared with a standard Annexin-V/propidium iodide staining protocol (Supplementary Fig. S2). Unless otherwise stated, the simplified protocol was used to compare the cell killing activities of the various TRAIL forms. A FACS-based viability assay was similarly used when RBCs were present during apoptosis measurements (Supplementary Fig. S3). Data acquisition was done on a FACScan flow cytometer (Becton & Dickinson). The data were analyzed with FlowJo software (Version 7.2.5, Tree Star). Cell viability of adherent cells was determined by crystal violet staining after fixation with 2% paraformaldehyde and CellTiter-Glo (Promega) according to the manufacturer's instructions. Data were recorded with a luminescence plate reader (Molecular Devices, SpectraMAX-Gemini, SoftMax Version 5 software).

Linear regression analysis of RBC-mediated Jurkat and BxPC3 cell killing was calculated using Microsoft Excel software. Tumor growth curves are expressed as means ± SE. A nonparametric, one-tailed t-test was done to evaluate the statistical significance of the two curves and individual data points using GraphPad Prism (V 4.02) software.

The recombinant, mature form of human TRAIL described here (TR3) comprises three consecutive extracellular TRAIL domains (with >99% TRAIL-specific amino acid sequence) fused together in a head-to-tail configuration (Fig. 1A). The single-chain character of the fusion protein was verified by Western blot analysis. Under reducing conditions, commercially available recombinant TRAIL (rTRAIL, aa 114-281) exhibits a molecular weight of 18 kDa, and TR3 has a molecular weight of ∼61 kDa, consistent with its calculated size (Fig. 1B).

To obtain evidence for its cell-killing activity, we treated TRAIL-sensitive human Jurkat cells with TR3 or rTRAIL. A FACS-based cell-killing assay was established that takes advantage of the morphologic changes (cell shrinkage) that are induced during apoptosis (Supplementary Fig. S2). We found that this new drug candidate revealed a strong apoptosis-inducing capacity (Fig. 2A, top), nearly identical to rTRAIL (Fig. 2A, bottom). To confirm that the killing activity was mediated by the death receptor pathway, we did a blocking experiment using DR5-Fc. DR5-Fc reduced the killing capacity of TR3 to background levels in a dose-dependent fashion (Fig. 2B, top), similar to that of rTRAIL (Fig. 2B, bottom, and Supplementary Fig. S4). The faster response of TR3 to DR5-Fc solely reflects the higher starting concentration of rTRAIL chosen for this particular experiment and requires a higher dose of soluble receptor to achieve rTRAIL blockade. In any case, this critical inhibition experiment suggests that TR3 assumes a native conformation, capable of interacting both with soluble DR5 and, more importantly, with cell-bound DR5 (the only death receptor expressed from Jurkat T cells; data not shown), through which it stimulates the extrinsic death pathway and ultimately induces target cell death. Finally, to obtain additional evidence for the unaltered specificity profile of TR3, we treated nontransformed, immortalized human pancreatic ductal epithelial cells with TR3 and rTRAIL. We found that both reagents did not cause cell death on these normal cells, which is consistent with the notion that TRAIL preferentially kills transformed target cells (data not shown).

To assess the physicochemical properties of TR3 using defined in vitro conditions, we evaluated (a) its stability during storage at physiologic temperature and (b) the consequences of repeated freeze/thaw (F/T) cycles. When maintained at 37°C, commercially available rTRAIL readily lost a significant fraction of its initial killing potential (Fig. 3A; >95% reduction after 30 min). In contrast, TR3 was unaffected by this treatment and retained 100% of its activity following a 6-hour incubation period at 37°C (Fig. 3A). We also assessed the effect of repeated F/T cycles on the stability of TR3 compared with rTRAIL. We noted a rapid loss of activity of rTRAIL following only one F/T cycle ranging between 90% and 95% compared with the nonfrozen control (Fig. 3B). No residual killing activity was generally detected past the third F/T cycle. In contrast to rTRAIL, TR3 retained 100% of its biological activity even when cycled 10 times (Fig. 3B). Finally, when TR3 and rTRAIL were injected i.v. into WT C57BL/6 mice and plasma was collected and tested for biological activity at 1 minute (baseline) and 10 minutes postinjection, we found that rTRAIL lost a significant fraction of its activity within the first 10 minutes in circulation (Fig. 3C; 49.0 ± 8.4% reduction), whereas TR3 maintained its bioactive form longer (Fig. 3D; 13.8 ± 1.5% reduction). These results show that covalently linked TRAIL monomers (TR3) have more favorable pharmacokinetic characteristics than noncovalently associated rTRAIL. The improved stability of TR3 over rTRAIL, especially in vivo, may prove to be a significant advance for researchers as well as clinicians.

Building on the improved pharmacologic stability of TR3, we addressed next how additional genetically encoded alterations (domain additions) would affect the performance of TR3. The goal of such an approach was to direct TRAIL activity more specifically to a predefined target to further increase its cell specificity and reduce its off-target effects following systemic application. Examples of an extremely functionally diverse but structurally homogeneous class of cell targeting devices are single-chain antibody fragments (scFv). Encoded by single genetic fusion sequences themselves, they are versatile entities that would allow TR3 to be targeted to a selected site and would therefore provide a secondary specificity besides the natural afficity of TRAIL for the death-inducing receptors. As a proof of concept, we introduced a scFv to the NH₂-terminus of TR3 that recognizes glycophorin A on the mouse RBC (scFv-TR3; Supplementary Fig. S1A). This model antigen was chosen because of the lack of endogenous TRAIL receptors on the RBC membrane, and it would better facilitate our ability to interpret the results of target cell binding studies (see below). In anticipation of potential steric constraints, we designed an additional RBC-targeted TR3 form to create a larger distance between the scFv and TR3 (scFv-S-TR3; Supplementary Fig. S1A). This elongated spacer comprised four SCRs, ∼60 amino acid–containing globular domains of the human complement regulatory proteins decay accelerating factor (DAF, CD55) and complement receptor 1 (CR1, CD35). To show the broad applicability of our platform technology, we also added eGFP as another fusion partner of TR3 (eGFP-TR3; Supplementary Fig. S1A).

All fusion constructs were expressed in HEK293T cells and exhibited their expected molecular weights determined by Western blot analysis (scFv-TR3 at 85 kDa, scFv-S-TR3 at 110 kDa, and the eGFP-tagged TR3 at 85 kDa; Supplementary Fig. S1B). To evaluate the impact of the NH₂-terminal extensions of each fusion protein, the fluid-phase killing capacity of the additional TR3 variants was tested and compared with the parental molecule. Interestingly, we found that all TR3 forms were capable of inducing apoptosis on DR5-expressing Jurkat reporter cells (Supplementary Fig. S1C). This result was encouraging from the perspective of designing future therapeutic drugs because we found that attachment of a 50 kDa heterologous, non-TRAIL sequence (scFv-S-TR3) did not sterically hinder the killing capacity of TR3.

Although the ultimate objective was to target TR3 directly to a tumor cell membrane, we believed it was necessary to ask if scFv-TR3, and its spacer variant scFv-S-TR3, could be selectively attached to a cell membrane that lacks a TR3 binding partner. Again, the RBC was chosen because it represents a death receptor–deficient cell type. The respective TRAIL forms were incubated with C57BL/6 mouse whole blood, washed, and assessed for the presence of TR3 on the RBC membranes using purified DR5-Fc. Indeed, FACS analysis revealed a strong, homogeneous signal peak for both RBC-targeted TR3 variants (Fig. 4A). In addition, scFv-S-TR3 was also detected with an antihuman CR1 monoclonal antibody (clone 3D9), which binds to the spacer region of the fusion protein (Supplementary Fig. S1D and E). The fact that a soluble DR5 receptor was capable of binding RBC-bound TR3 suggests that the death receptor–interacting domains of these large fusion proteins were folded correctly and should be also accessible by membrane-expressed DR5 on a given target cell. To prove this hypothesis, we used a rosetting assay, in which TR3-coated murine RBCs (compare Fig. 4A for coating levels) were mixed with eYFP-expressing Jurkat target cells and were allowed to sediment at ambient temperature. We anticipated that the fusion proteins would facilitate bridging of the two cell types and consequently form aggregates (rosettes). Shortly after mixing, we identified mouse RBCs (red) tightly attached to their green target cells (Fig. 4B, arrows). Importantly, this was true only for the spacer version of TR3 (scFv-S-TR3), because we did not observe a rosetting phenomenon in the absence of the spacer (scFv-TR3; data not shown). Moreover, once the RBCs attached to their targets, we observed programmed cell death as documented by the appearance of apoptotic bodies (Fig. 4B, arrowhead).

We next sought to quantitatively investigate the RBC-mediated target cell killing and study the impact of the spacer domain of scFv-S-TR3 with respect to TRAIL receptor recognition once immobilized to the RBC surface. The relative copy numbers of the RBC-bound TR3 forms were determined by DR5-Fc staining (compare Fig. 4A for TR3 coating levels). We specifically chose to attach more copies of the spacer-deficient scFv-TR3 to the RBCs because we had already shown that only the spacer variant was capable of forming rosettes, suggesting that it would likely be a more powerful inducer of target cell apoptosis. A FACS-based Jurkat killing assay in the presence of RBCs was used similar to what has been described above (Supplementary Fig. S3). As anticipated, the presence of medium-treated control RBCs had no effect on the viability of the target cells (Fig. 5A). In contrast, we found that both RBC-targeted TR3 forms elicited cell death in this coculture system (Fig. 5A). In this comparative killing assay using only limited TR3 copy numbers attached to the RBC surface (<5,000 estimated copies/RBC), the spacer variant scFv-S-TR3 was more than twice as effective as the nonspacer form scFv-TR3 (based on the slopes of the killing curves) although we provided a much reduced copy number of the spacer variant attached to the RBC membrane (only 60% relative to scFv-TR3). It is worth mentioning that it was only possible to do these binding and killing experiments because of the improved stability of the TR3 molecule, since the coating process was done for several hours at ambient temperature, conditions under which rTRAIL would have lost its bioactivity.

To rule out that the RBC-mediated target cell killing was not induced by contaminating fluid-phase fusion proteins, we carried out a similar experiment but spatially separated effector and target cells using a transwell system. Under these conditions, target cell killing was completely abolished, confirming that TR3-decorated RBCs induce apoptosis via cell-cell contact (data not shown).

In the experiments described above, we coated the RBCs prior to coculture with their Jurkat target cells. In this setting it was difficult to compare the results with regard to the exact number of TR3 molecules bound to the RBC membrane with nonbound, fluid-phase TR3. To address this issue, increasing numbers of naïve RBCs were added first to the target cells. Then, functionally equivalent doses of RBC-targeted scFv-S-TR3 and, as a control, nontargeted TR3 (with identical fluid-phase killing activities), were added to the mixtures. As expected, in the presence of non-RBC–targeted TR3, increasing numbers of RBCs did not improve (or reduce) target cell killing (Fig. 5B, killing curve remains horizontal). However, increasing concentrations of naïve RBCs during the treatment period with scFv-S-TR3 resulted in an increased target cell killing of >60% at the highest RBC concentration used, likely via in situ RBC coating (Fig. 5B). Interestingly, although the drug input was the same under these different conditions (baseline, fluid-phase killing for both drugs in the absence of RBCs ∼50%), enhanced target cell killing was only possible in the presence of a “native” solid matrix, i.e., the RBC membrane. Experiments are currently planned to investigate the nature of the enhanced killing capacity of surface-immobilized TR3 in greater detail. Together, these results suggest that it should be feasible to decorate, via single-chain antibody fragments, a tumor cell surface with biologically active TR3 potentially augmenting the native killing capacity of TRAIL while decreasing systemic toxicity.

To assess the ability of RBC-targeted TR3 to kill human tumor cells, we used a murine xenotransplantation model of pancreatic cancer. We first confirmed the previously reported TRAIL sensitivity of BxPC3 cells by cocultivation with TR3-decorated mouse RBCs in a similar tissue culture system described above for the Jurkat T-cell line. We found that BxPC3 cells were killed by TR3-coated RBCs in a dose-dependent fashion (Fig. 6A).

To assess the killing capacity of TR3-decorated RBCs, we had to ensure that effector and target cells would have access to each other. Therefore, BxPC3 cells were mixed on ice (to prevent immediate rosette formation) with scFv-S-TR3–coated mouse effector RBCs at an E:T of 40 (with 5 × 10⁵ target cells/animal) and immediately injected into the right flanks of male nude mice. It was expected that in this particular model, at an E:T of 40, a similar fraction of the tumor targets would be eliminated (∼60%; Fig. 6B, inset) within the first 24 hours after implantation due to direct and required cell contact with TR3-coated RBCs. We expected that the surviving cells (∼40%) would engraft and eventually contribute to tumor formation. This is indeed what we observed. Eight days postinoculation, the control animals developed measurable tumor masses that continued to grow exponentially (Fig. 6B). In contrast, mice that received a mixture of BxPC3 and TR3-coated RBC effector cells presented with a ∼30-day delay in tumor progression, after which the growth characteristics were identical to the tumors in the control animals.

TRAIL is a member of the TNF superfamily and is well known for its ability to cause cancer-selective apoptosis. A number of different approaches have been utilized in the past to produce biologically active TRAIL trimers, and are all based on the expression of monomeric cDNAs. In this study, recombinant human TRAIL trimers (TR3 family) were generated based on a single polypeptide format. We showed the potent apoptosis-inducing activity of TR3, similar to rTRAIL but with an enhanced stability profile compared with the latter. We further believe that the genetic approach to trimerization will extend to other TNF family members. In fact, we found that a similar concept has been reported for TNF, and similar to TR3, recombinant TNF trimers also showed increased stability compared with their noncovalently associated form (19). However, the most important finding of our current work is the fact that TR3 can be further genetically modified while TRAIL activity remains fully preserved.

In this report, we show the feasibility of such modifications by incorporating cell-targeting epitopes to the parental TR3 molecule. As an example, we have shown that an antibody fragment (scFv) with specificity for mouse RBCs to the NH₂-terminus of TR3 allowed us to deliver bioactive TRAIL to a native cell membrane in a stoichiometrically controlled fashion. This latter point represents another key feature of the TR3 platform technology and means that one targeting molecule (scFv) could deliver a prearranged, bioactive TRAIL trimer (TR3) to a defined target site. Therefore, changing the scFv fragment to one that specifically targets a cancer cell may reduce some of the off-target toxicity associated with other rTRAIL formulations (9). A similar strategy has been recently applied using a monomeric TRAIL-encoding scFv fusion construct (20–22), which generated predominantly monomeric as well as dimeric and only a minor fraction of biologically active trimeric complexes (22). Therefore, we believe that the steric control provided by our assembly strategy (scFv:TR3 = 1:3) would significantly expand on the size and nature of attachments to TR3 than we would envision to encounter by combining three TRAIL molecules, each carrying a cell targeting antibody fragment.

Another important modification we explored was the elongated spacer domain, inserted between the targeting scFv and TR3. This configuration was expected (a) to better enable binding to the RBC via scFv and (b) to elevate the TR3 domain farther away from the RBC surface, thereby better allowing an interaction with target cell–expressed death receptors. Although we do not present evidence for the spacer being advantageous for optimal target cell binding, we found that, when RBC bound, this spacer substantially enhanced the activity of the drug compared with its spacer-deficient analog. The fact that such a spacer-containing TR3 molecule was capable of bridging two unrelated cell types (TR3-decorated RBCs and Jurkat/BxPC3 cells) and could induce apoptosis via cell-cell contact, suggests that a future anticancer therapeutic based on this design should be as capable of bridging tumor-specific antigens and DR4/5 located on the same cell (cis-effect) as well as adjacent cells (trans-effect) causing apoptosis of both tumor cells.

Finally, the nearly complete absence of artificial linker sequences of TR3 and the incorporation of spacer sequences with human origin, human DAF and CR1 domains (scFv-S-TR3), has the theoretical advantage of reducing potential immunogenicity of a novel therapeutic based on this technology. However, this aspect and immunogenicity due to generation of neoepitopes need to be evaluated in a small animal model, i.e., using mouse TRAIL for the construction of murine TR3 and the respective murine-derived spacer sequences.

Secreted TRAIL, expressed from a monomeric cDNA in mammalian cells, is mostly inactive (23). This has been attributed mainly to the formation of interchain disulfide bridges, which in turn causes this mixture of TRAIL monomers, dimers, and trimers to have reduced affinity for its cell surface receptors. We have been able to confirm these results and only the enforced trimer formation via an isoleucine-zipper domain could restore TRAIL activity (data not shown). Therefore, trimer enforcement via generation of TR3 resembles more closely an isoleucine zipper–TRAIL configuration in that formation of these disulfide bridges is likely being disabled because TR3 is biologically highly active when produced from mammalian cells. These findings have important consequences if biochemical postproduction manipulations are not an option. One example represents gene therapy in which the patient's own (mammalian) cells are the source of the therapeutic protein. Although such a concept has been investigated (24), it seems likely that TR3 could represent a more powerful alternative over TRAIL produced from a monomer-encoding cDNA.

A potential activity-enhancing modification relates to the location where the additional fusion partners of TR3 were attached. We initiated our studies by fusing these domains to the NH₂-terminus of TR3. However, we have recently shown that switching the scFv from the NH₂- to the COOH-terminus of an RBC-targeted mouse complement regulator resulted in a substantial improvement of its activity (25). Theoretically, those types of variations could also improve the properties of tumor-targeted TR3 variants, considering that both termini of the TR3 fusion protein should be accessible in light of several reported crystallographic studies on noncovalently associated TRAIL trimers (26–29). Furthermore, TRAIL binding specificity can be tailored by genetic engineering (by introducing several amino acid substitutions) toward either of the two death-inducing receptors DR4 or DR5 (30), depending on which death receptor would represent the more promising therapeutic target. These types of alterations could be easily applied to TR3.

One unexpected observation requires further exploration. The addition of nontargeted TRAIL forms (TR3 and rTRAIL) to the Jurkat target cells consistently resulted in a plateau in killing at ∼40% to 60%. This plateau effect has been considered a limitation of TRAIL-based monotherapies (31). With TRAIL immobilized on a solid matrix (the RBC surface), this plateau was consistently overcome. Our initial observations indicate that there may be cell signaling advantages to a surface-based TR3 delivery platform which warrant further investigation.

In summary, we present a new method to generate recombinant human TRAIL (TR3-family) based on a single polypeptide format featuring potent apoptosis-inducing activity and enhanced stability. Importantly, TR3 can be further modified in a stoichiometrically controlled fashion without interfering with TRAIL function. To achieve higher target cell specificity and/or killing activity, we are currently adopting this platform to a variety of different tumor-associated antigens. This approach will take advantage of two ligand-receptor interactions facilitated by a single, multifunctional drug-tumor Ag:scFv and TR3:death receptor. As a consequence, we predict that the resulting multidomain therapeutic would have greater affinity for the tumor and a reduced affinity toward normal host cells. The predicted advantages of such a concept would be a stronger and more sustained induction of the death receptor pathway in the targeted tumor cells. As TRAIL has been shown to augment the effects of standard therapies, a tumor-targeted TR3 might further enhance the effectiveness of other chemotherapies while limiting off-target toxicities to the patients.

No potential conflicts of interest were disclosed.

We thank Dr. Tom Ferguson for helpful comments, Dr. Thomas Griffith for providing the DR5-Fc expression plasmid, Jesse Gibbs for technical assistance, and Kim Trinkaus and Jingqin “Rosy” Luo (Washington University Biostatistician Core) for statistical advice.

Grant Support: NIH grants 5P30CA9184208 and 1R21CA150945 to D. Spitzer and W.G. Hawkins.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.