Induction of Antitumor Immunity via Intratumoral Tetra-Costimulator Protein Transfer¹

One cancer vaccination strategy of interest entails the administration of “immunogenic” tumor cells to elicit systemic antitumor immunity. Such immunogenic tumor cells can be engineered in a number of ways, including via the neoexpression of immune costimulators on their surfaces (reviewed in Ref. 1). Enforced expression of surface T-cell-directed costimulators, such as B7–1 (CD80; Ref. 2), or dendritic cell activators, such as CD40L (3, 4), confers on certain tumor cell lines the capacity to elicit systemic antitumor immune responses. However, in the face of idiosyncratic results obtained for different tumor cell lines when costimulators are used individually (5), we (6) and others (7, 8, 9, 10) have advocated the use of more than one costimulator in combination to additionally augment the immunogenic potential of engineered tumor cells. Nonetheless, to date, only a limited number of costimulator combinations have been explored. Moreover, whereas costimulator combinations have been selected with their synergistic and/or additive effects in mind, the opportunity remains for rationally designing costimulator combinations that would simultaneously invoke multiple antitumor immune cell effectors.

In most studies to date, enforced expression of costimulators on tumor cell surfaces has been achieved through gene transfer. However, gene transfer has limitations in this context, both experimentally and clinically. For one, it does not enable fine control of protein product expression, thus precluding the titration of costimulator mixtures at optimal levels. A second and more fundamental limitation of gene transfer is that it is poorly suited for the combinatorial expression of multiple gene products in the same cell, even when viral vectors are invoked. The challenge increases with the number of gene products to be expressed. Protein transfer offers advantages over gene transfer in both regards, because protein transfer is well suited for the simultaneous delivery of well-defined amounts of multiple proteins onto individual cells.

Previously, we (11, 12) and others (13) reported the engineering of immunogenic tumor cells by painting on their surfaces GPI³-modified derivatives of the B7–1 (CD80) costimulator. More recently, we reported an even simpler two-step protein transfer method that uses precoated pal-prot A as artificial docking sites for Fc_γ1 fusion proteins, such as B7–1·Fc_γ1 (14). The present study builds on this latter protein transfer method in two ways. First, protein transfer is invoked here for the first time to express complex costimulator arrays, with up to four costimulator·Fc_γ1 derivatives (B7–1·Fc_γ1, Fc_γ1·4–1BBL, CD48·Fc_γ1, and Fc_γ1·CD40L) now being used in combination to simultaneously drive more than one antitumor immune effector pathway. Second, protein transfer is accomplished here for the first time in situ, rather than ex vivo, via intratumoral injection of costimulator·Fc_γ1:pal-prot A conjugates. The data point to the efficacy in an animal tumor model for this unique “tetra-costimulator, intratumoral” protein transfer strategy.

Female DBA/2J mice, 6–8 weeks old, were purchased from The Jackson Laboratory (Bar Harbor, ME) and used experimentally under protocols approved by the University of Pennsylvania Laboratory Animal Regulatory Committee.

The murine cell lines Ag104A (sarcoma) and T50 (bladder carcinoma) were gifts from Drs. Lieping Chen (Bristol-Myers Squibb Pharmaceutical Research Institute, Seattle, WA) and Abraham Hochberg (Hebrew University, Jerusalem, Israel), respectively. The murine mastocytoma cell line P815 and the T-lymphoma cell line L5178Y-R were purchased from the American Type Culture Collection (Manassas, VA). The Ag104A, T50, and P815 cell lines were maintained in DMEM supplemented with 10% FCS, whereas the L5178Y-R cell line was maintained in DMEM supplemented with 10% FCS, 0.11 g/liter sodium pyruvate, and 1.125 g/liter sodium bicarbonate. For in vivo studies, an L5178Y-R cell bank was generated from which cell cultures were freshly started and maintained for <3 weeks before inoculation into mice.

FITC-mouse antihuman CD40L mAb (TRAP1), rat antimouse B7–1 (1G10), and FITC-goat antirat IgG were purchased from BD PharMingen (San Diego, CA). Rat antimouse CD48 was purchased from Serotec Ltd. (Oxford, United Kingdom).

The strategy for assembling chimeric expression cassettes encoding murine B7–1·Fc_γ1 and CD48·Fc_γ1 mirrored that reported previously for human B7–1·Fc_γ1 (14). The coding sequences for the extracellular domains of murine B7–1 (V38-T247; Swiss-prot accession number: Q00609) and murine CD48 (F23-R261; Swiss-prot accession number: P18181) were linked in-frame to a coding sequence for the Fc_γ1 domain of human IgG1 within our expression constructs pmB7–1·Fc_γ1/REP7 and pmCD48·Fc_γ1/EE14, respectively. Because 4–1BBL and CD40L are type II membrane proteins, the Fc_γ1 domain was instead appended to their NH₂ termini. Hence, in the pFc_γ1·m4–1BBL/EE14 and pFc_γ1·mCD40L/EE14 expression constructs, the coding sequence for the Fc_γ1 domain of human IgG1 was linked in-frame to the extracellular domains of mouse 4–1BBL (R104-E309; Swiss-prot accession number: P41274) and human CD40L (H47-L261; Swiss-prot accession number: P29965), respectively. The fusion proteins were produced in 293 (for B7–1·Fc_γ1) or Chinese hamster ovary (CD48·Fc_γ1; Fc_γ1·4–1BBL; and Fc_γ1·CD40L) cell transfectants and purified by protein A-agarose chromatography, as described previously (14). Purified recombinant proteins were analyzed by SDS-PAGE on precast 3–8% NuPAGE gels, as per the manufacturer’s protocol (Invitrogen Corporation, Carlsbad, CA).

Procedures for generating pal-prot A (15) and using it for protein transfer (14) have been detailed previously. However, whereas the earlier studies called for the sequential (two-step) addition of pal-prot A followed by Fc_γ1 fusion proteins, protein transfer is accomplished here in one step by using preformed conjugates of pal-prot A and each of the Fc_γ1 fusion proteins (or control human IgG1). These preformed conjugates were generated by combining the components at a 1:1 ratio (w/w) in PBS/0.1% BSA (transfer buffer) at ambient temperature for 30 min. P815 and L5178Y-R tumor cells, grown in suspension, were suspended in transfer buffer after three washes in the same buffer. Ag104A and T50 tumor cells, grown as adherent cells, were detached with 5 mm EDTA in PBS before being suspended in the transfer buffer. Protein transfer was accomplished by coincubating 2 × 10⁶ cells with 12 μg of each of the preformed pal-prot A:Fc_γ1 fusion protein (or IgG1) conjugates at 37°C for 1 h in 0.2 ml of the transfer buffer, supplemented with 0.1% sodium azide, 10 mm 2-deoxyglucose (to prevent endocytosis), 1 mm phenylmethylsulfonyl fluoride, 1 μm leupeptin, and 1 μm pepstatin (to prevent proteolysis). For simultaneous transfer of more than one Fc_γ1 fusion protein, each costimulator·Fc_γ1 fusion protein was conjugated with pal-prot A separately, as described above, and the resultant conjugates were then combined at equal mass ratios and transferred to the cells under the same conditions described above.

Cells coated with control FITC-human IgG were analyzed without additional immunostaining by flow cytometric methods described previously (14). To detect cell surface-associated murine B7–1·Fc_γ1 or CD48·Fc_γ1, cells were immunostained with rat antimouse B7–1 mAb (1G10; BD PharMingen) or rat antimouse CD48 mAb (Serotec) as primary antibody, respectively, and FITC-goat antirat IgG as secondary antibody (BD PharMingen). Of note, rat and goat antibodies were chosen here because neither of them binds to protein A. Because neither rat or goat antibodies are available for the detection of 4–1BBL or CD40L, biotinylated derivatives of their corresponding Fc_γ1 fusion proteins were used in experiments aimed at monitoring protein transfer efficiencies. Fc_γ1·4–1BBL and Fc_γ1·CD40L were biotinylated with EZ-Link Sulfo-NHS-LC-Biotin, as per the manufacturer’s protocol (Pierce, Rockford, IL) and were then detected with FITC-avidin (BD PharMingen).

For experiments monitoring the distribution of pal-prot A injected intratumorally, pal-protein A was biotinylated and injected in a PBS solution containing 80 μg/ml of the biotinylated protein at 50 μl/tumor. The tumors were excised 1 or 18 h later, cross-sectioned, stained with avidin-peroxidase and DAB, and counter-stained with H&E.

For experiments monitoring in vivo protein transfer efficiency of B7–1·Fc_γ1:pal prot A conjugates, each of the two protein elements were combined at 1:1 (w/w) ratio in PBS to a total protein concentration of 160 μg/ml, without the other components included in the conjugate mixtures described above for the ex vivo protein transfer experiments. The conjugates were injected into L5178Y-R tumors grown intradermally at 50 μl/tumor. Cells were recovered from excised tumors 1 h after injection of the conjugates, treated with ACK lysing buffer (BioWhittaker, Walkersville, MD) to remove blood cells, stained with rat antimouse B7–1 mAb and FITC-goat antirat IgG, and analyzed by flow cytometry.

Pal-prot A conjugates of B7–1·Fc_γ1, CD48·Fc_γ1, Fc_γ1·4–1BBL, and Fc_γ1·CD40L were generated separately by combining each of the two components at a 1:1 ratio (w/w) in PBS, to a total protein concentration of 160 μg/ml. To constitute the tri- and tetra-costimulator solutions, equal volumes of the component Fc_γ1 fusion protein:pal-prot A conjugates were combined. The pal-prot A control solution contained pal-prot A alone and was prepared in PBS at 80 μg/ml.

L5178Y-R tumors in syngeneic DBA/2J mice (Jackson Laboratory) were treated as follows. Mice (identified individually by earmarks) were injected intradermally (via a 26-gauge needle) with 5 × 10⁵ L5178Y-R tumor cells on the right flank (day 0). Starting at day 4, when the L5178Y-R tumors generally reached a size of 25–50 mm², the tumors were injected (again via a 26-gauge needle) with one of the conjugate-containing solutions (50 μl/tumor) once daily for each of 4 consecutive days. Cured mice were rechallenged either intradermally (5 × 10⁵ L5178Y-R tumor cells) on the opposite (left) flank or i.p. (10⁴ L5178Y-R tumor cells) at least 5 weeks after the initial tumor inoculation. All of the mice were monitored daily, and tumor sizes were measured three times per week with a caliper. Mice were euthanized when they became moribund or when their tumors exceeded 400 mm² in size.

Bulk splenocytes, prepared 3–6 weeks after tumor rechallenge, were cultured in 24-well plates (10⁷ cell/2 ml/well) with mitomycin C-treated L5178Y-R tumor cells (2 × 10⁵/well) as stimulators in RPMI 1640 containing 10% FCS, 15 mm HEPES, and 50 μm β-mercaptoethanol. After 5 days, viable cells were harvested and used as effectors in a standard JAM assay (16), with [³H]labeled L5178Y-R and (syngeneic) P815 tumor cells as the specific and nonspecific tumor targets, respectively.

Pal-prot A-based Fc_γ1 fusion protein transfer was performed previously as a two-step procedure, with the pal-prot A being precoated onto cells before the addition of the second, Fc_γ1-containing fusion protein component (14). To simplify the procedure and adapt it for direct in vivo use, we preassembled the pal-prot A and Fc_γ1-containing components to enable one-step protein transfer. To this end, pal-prot A and FITC-human IgG were preconjugated and then combined with each of four different tumor cell lines, two adherent (Ag104A and T50) and two nonadherent (L5178Y-R and P815). This one-step protein transfer approach yielded efficient membrane incorporation in all of the cases (Fig. 1), albeit with differences among the lines that did not correlate with adherent versus nonadherent growth properties. As expected, conjugates containing native (nonpalmitated) protein A did not attach to cells (Fig. 1), confirming that the membrane incorporation is lipidation-dependent.

Next, the stability of incorporated conjugates at the cell surface was determined. L5178Y-R and Ag104A cells, coated with FITC-human IgG:pal-prot A conjugates, were subsequently incubated at 37°C for a 6-day period and serially monitored by flow cytometry. Conjugates were remarkably stable at cell surfaces, with only a marginal decrement in MFI over the 6-day period (Fig. 2). Of note, because the FITC-labeled human IgG alone did not bind to these metabolically blocked cells (Figs. 1 and 2 ), the cell-associated fluorescent signal being detected by flow cytometry is indicative of human IgG that has been tethered to cell membranes via pal-prot A. Taken together, these data established that efficient, reproducible, and stable membrane incorporation is achievable by the modified one-step method.

Moving beyond IgG:pal-prot A test conjugates, we next applied the one-step method to costimulator protein transfer per se. Fc_γ1 fusion protein derivatives of four costimulators, directed at lymphoid (B7–1, CD48, and 4–1BBL) and APC (CD40L) effectors, were produced (Table 1). The Fc_γ1 component was positioned at the COOH termini of those fusion proteins corresponding to type I membrane proteins (B7–1 and CD48), whereas it was placed at the NH₂ termini of those fusion proteins corresponding to type II membrane proteins (4–1BBL and CD40L). This variable positioning of the Fc_γ1 component serves to distance it from the functional regions of the appended domains. The chimeric expression cassettes were incorporated into either the pREP7β EBV episomal expression vector (for B7–1·Fc_γ1) or the pEE14 amplification/expression vector (for CD48·Fc_γ1; Fc_γ1·4–1BBL; and Fc_γ1·CD40L), which in turn, enabled quantitative expression in 293 and Chinese hamster ovary cell transfectants, respectively. The fusion proteins were affinity-purified to near homogeneity by protein A chromatography, as visualized by SDS-PAGE under reducing conditions (Fig. 3,A). Of note, the native molecular masses for B7–1 (17), CD48 (18), CD40L (19), 4–1BBL (20), and human IgG1 Fc (21) have all been reported previously to be larger than expected because of extensive glycosylation. The SDS-PAGE masses of the four Fc_γ1 fusion proteins generated here (Fig. 3 A) correspond well to what would be predicted based on the reported native masses of their component parts.

To document that the costimulator·Fc_γ1 proteins can be transferred to cells by the one-step method, pal-prot A conjugates were formed individually with each fusion protein and subsequently transferred to L5178Y-R tumor cells, either individually (not shown) or simultaneously (Fig. 3,B). In both instances, efficient membrane incorporation of proteins was detected by flow cytometry. Significantly, the tetra-costimulator combination yielded a uniform population of cells that are all coexpressing each of the four costimulators (Fig. 3 B), thus establishing the feasibility of combinatorial protein transfer.

The functionality of these four costimulator Fc_γ1 fusion proteins was formally demonstrated in appropriate bioassays. To test the activity of B7–1·Fc_γ1, CD48·Fc_γ1, and Fc_γ1·4–1BBL, the costimulator proteins were “painted” on mitomycin C-inactivated L5178Y-R cells, which in turn costimulated proliferation of purified syngeneic splenic T cells (not shown). In the case of Fc_γ1·CD40L, the soluble protein stimulated the proliferation of purified human B cells in the presence of human IL-4; this stimulation could be blocked with TRAP1, an anti-CD40L mAb (not shown).

As a first step toward applying the protein transfer method to tumor therapy, we determined the feasibility of intratumoral costimulator·Fc_γ1 protein transfer. The murine T lymphoma line L5178Y-R was chosen for these studies, because it readily forms tumors when injected either intradermally or i.p., is highly metastatic (generally killing mice in <2 weeks with liver metastases), and expresses levels of MHC class I (H-2K^d) that are comparable with those on splenocytes from syngeneic DBA/2J mice (data not shown). The L5178Y-R cells do not express native B7–1, B7–2, or CD40L, and express only low levels of CD48 (data not shown).

On intradermal seeding, L5178Y-R cells form well-circumscribed tumors that are amenable to intratumoral injection. Biotinylated pal-prot A and a nonlipidated control, biotinylated protein A, were injected separately into intradermally seeded L5178Y-R tumors, and the injected tumors were excised 1 h later, sectioned transversely across the needle track, stained with H&E, and counter-stained with avidin-peroxidase with DAB as chromogen to detect immobilized biotin (Fig. 4,A). H&E sections showed lymphoid neoplasm infiltrating the dermis and subcutis. A band-like necrotic zone was present at the base of tumors right above the deep fascial plane in both injected and noninjected tumors. The needle track was identified in injected tumors. Necrosis around the needle tract was noted, although there was no significant difference in necrosis around the needle track between tumors injected with biotin-pal-prot A or biotin-protein A. The avidin-biotin complex product was visualized as brown staining under the light microscope. In noninjected tumors as well as tumors injected with the biotin-protein A, only RBCs showed light brownish color because of nonspecific reactions with DAB (Fig. 4, A–F).

In marked contrast to these controls, tumors injected with biotin-pal-prot A additionally displayed intense staining of tumor cells, extending to a 400 × 600 μm² section across the needle track (Fig. 4, G–I). Significant levels of tumor cell staining were still observed even 18 h after injection with biotin-pal-prot A (not shown). These histochemical data indicate that intratumorally injected pal-prot A coats tumor cells in situ, penetrates to a significant depth beyond the needle track, and persists on tumor cells many hours after injection.

Next, B7–1·Fc_γ1 was combined with pal-prot A, and the resulting conjugates were injected into intradermal L5178Y-R tumors. After injection (1 h), cells were recovered from an excised (∼1 mm) strip of tumor tissue encompassing the needle track, and these suspended cells were analyzed for B7–1 epitope expression by immunofluorescence and flow cytometry. As shown in Fig. 4 B, 33% of the cells from B7–1·Fc_γ1:pal-prot A-injected but not noninjected or negative control B7–1·Fc_γ1:prot A-injected tumors were B7–1⁺ cells. In addition to validating the intratumoral protein transfer strategy, these immunofluorescence data confirmed that the transferred proteins are accessible for molecular recognition (in this case, via mAbs).

With the intratumoral, “multicostimulator” protein transfer strategy well in hand, we next investigated its antitumor therapeutic utility. To start, we established that pal-prot A, when injected into the tumor bed on its own (4 μg within a 50-μl injection volume in intradermal L5178Y-R tumors), did not influence tumor growth or animal survival (data not shown). With this as baseline data, we moved on to assess the therapeutic efficacy of various combinations of costimulator·Fc_γ1:pal-prot A conjugates, with up to four costimulators in the mix. Of note, the previous experience with costimulator combinations is limited (with only one report of more than two costimulators; Ref. 10), and it is largely confined to ex vivo manipulation of cancer vaccine cells by gene transfer. Here we incorporated up to four costimulators, leveraging protein transfer and intratumoral delivery.

In a first set of tumor cure experiments performed in the L5178Y-R T lymphoma model, we evaluated the antitumor effects of three different costimulator combinations: a tetra-costimulator combination encompassing activators of T cells, NK cells, and dendritic cells (B7–1·Fc_γ1, CD48·Fc_γ1, Fc_γ1·4-BBL, and Fc_γ1·CD40L); a tri-costimulator combination directed at T cells and NK cells only (B7–1·Fc_γ1, CD48·Fc_γ1, and Fc_γ1·4-BBL); and the dendritic cell activator Fc_γ1·CD40L alone. Of note, the human form of CD40L is active in mice (22). Each of these costimulator permutations (injected as mixtures of pal-prot A conjugates) yielded beneficial effects with respect to tumor growth (Fig. 5,A) and animal survival (Fig. 5,B), but their relative efficacies varied. In a typical experiment, 6 of 12 mice injected with the tetra-costimulator combination showed complete tumor regression as compared with 4 of 12 mice for each of the other two groups (Fig. 5,A). Animal survival data mirrored these differences. Hence, 6 of 12 tetra-costimulator-injected animals were alive at day 35 in contrast with 4 of 12 for each of the other two groups (Fig. 5 B). In contrast with the treated animals, control animals injected with pal-prot A alone all died with metastatic tumor by day 33.

Three additional tumor cure experiments were performed focusing on tetra-costimulator, intratumoral protein transfer. The aggregated survival data, as summarized in Table 2, showed that 45% (18 of 40) of tetra-costimulator-treated animals were cured as opposed to 6% (2 of 34) of control animals (either untreated or treated with pal-prot A alone).

Having documented local tumor regression in a substantial percentage of tetra-costimulator-injected tumors, we next looked for evidence of systemic antitumor immunity. To this end, cured mice were rechallenged with L5178Y-R tumor cells ranging from 5 to 8 weeks after the first inoculation of the tumor cells. For this rechallenge, tumor cells were administered at sites distant from the original tumor, either intradermal or i.p. As summarized in Table 3, all (11 of 11) of the animals resisted rechallenge with the tumor cells. It is especially notable that 6 of 8 animals subjected to i.p. tumor rechallenge 6–9 months after the initial tumor cell inoculation were resistant to tumor. These rechallenge experiments point to a persistent, systemic antitumor immunity that is evoked by intratumoral costimulator inoculation.

To additionally support the existence of systemic antitumor immunity, we recovered splenocytes from tetra-costimulator-treated mice 3 weeks after tumor rechallenge and looked for recall responses against the L5178Y-R tumor cells after 5 days of in vitro restimulation with tumor cells (Fig. 6). The bulk splenocytes from cured, tetra-costimulator-treated mice showed strong CTL activity and efficiently lysed the tumor cells in a standard JAM assay. The CTL activity was L5178Y-R-specific, with significantly less lysis of another syngeneic tumor line, P815. Furthermore, naïve mice failed to develop any cytolytic activity against the L5178Y-R tumor cells during the in vitro restimulation phase, indicating that this response is dependent on an immunizing effect resulting from previous treatment of the tumor. Thus, tumor-specific CTL can be recovered from a secondary lymphoid compartment that is by definition distal from the costimulator-coated tumor cells at the injection site.

The present study documents the utility of intratumoral, tetra-costimulator protein transfer for cancer vaccination. This constitutes the first example of intratumoral protein transfer, which is accomplished here via the intratumoral inoculation of preformed molecular conjugates consisting of Fc_γ1-chimerized costimulators linked to membrane-anchoring, pal-prot A. Remarkably, murine L5178Y-R T lymphomas painted with immunostimulatory proteins in situ by this method elicit not only local tumor regression but also systemic tumor-specific immunity.

One distinguishing feature of this study relates to the number of costimulators being used. The preponderance of preclinical studies, as well as clinical trials to date, have dealt with enforced expression of individual costimulators. The more limited multicostimulator literature has been mostly confined to no more than two costimulators at a time (7, 8, 9), with one group recently invoking up to three (10). Here we have used four costimulators, leveraging the unique advantages of protein transfer for achieving combinatorial surface protein expression. Even more significantly, costimulators were chosen here not simply for their synergistic action in activating singular effectors but additionally for their collective capacity to invoke multiple antitumor immune effectors. The capacity to deliver complex arrays of immunostimulatory proteins to tumors via protein transfer opens the door to more elaborate combinatorial surface protein transfer schema in the pursuit of ever more effective cancer vaccination strategies.

Whereas costimulator-expressing tumor vaccine cells are intended to function as nonprofessional “tumor APC” (1, 23), some have argued the efficacy of such vaccine cells can be attributed, at least in part, to their triggering of cross-priming of tumor antigens by bystander professional APC (24). The immune-activating proteins in this study were chosen with both of these activation pathways in mind. Hence, whereas the triad of B7–1, 4–1BBL, and CD48 amplify the ability of tumor cells to directly prime lymphoid cells, CD40L was added to the mix to promote the activation of, and cross-priming by, professional APC, particularly dendritic cells (25, 26, 27, 28, 29, 30). Significantly, whereas the lymphoid-directed costimulator triad and CD40L each showed therapeutic efficacy on its own, the combination yielded optimal antitumor effects. This finding implies that both direct priming and cross-priming mechanisms may be operative in this context.

Within the costimulator triad, the choice of costimulators was prompted by their reported synergies and subset selectivity. Pairwise combinations of B7–1/4–1BBL and B7–1/CD48 synergistically activate T cells (7, 9, 31). Moreover, these three costimulators may preferentially influence different lymphoid subsets. Thus, B7–1 and 4–1BBL are preferentially linked to the CD4⁺ and CD8⁺ T-cell subsets, respectively (32, 33). Furthermore, murine CD48 additionally activates NK cells (34, 35). Given the higher order complexity of tetra-costimulator combinations, sorting out the precise contributions of individual costimulators within the mix to the activation of various antitumor effectors will require elaborate additional experimentation.

The preponderance of cellular cancer vaccine literature deals with tumor or dendritic cells that have been engineered ex vivo, often by approaches that that would be hard to implement clinically. Intratumoral delivery methods offer clinical advantages, at least for those tumors that can be readily accessed, because these methods bypass cumbersome ex vivo cellular manipulation. Whereas intratumoral transfer of cytokine genes has been evaluated by others (36), the intratumoral delivery of costimulator proteins has not. The observed efficacy of this latter therapeutic approach is encouraging and now provides motivation for undertaking additional studies directed at mechanistic questions, including the identification of the site(s) where the triggering of lymphoid and APC effectors occurs.

Early on, our group advocated protein transfer as a tool for engineering both tumor cell and dendritic cell vaccines (1). Subsequent studies by us (11, 12, 37, 38) and others (13) validating this concept centered around the use of GPI-modified costimulator and MHC derivatives as “protein paints.” However, GPI proteins have some limitations. Because the GPI moiety is appended to the COOH termini of proteins, this post-translational modification cannot be conferred to type II membrane proteins, a class of proteins that includes important costimulators of the tumor necrosis factor family such as 4–1BBL and CD40L. This constraint, along with the poor yields obtained when scaling-up the production of certain GPI derivatives, prompted our recent development of a two-component protein transfer method that combines pal-prot A and Fc_γ1-derivatized proteins (14). This approach not only enables the transfer of type II membrane proteins, as demonstrated here, but it invokes components that can each be produced in quantity as recombinant soluble proteins. Of note, whereas the original report (14) applied the two components to cells in two distinct steps, the present study simplifies their use by preparing them as molecular conjugates, which can be applied in a single step. Moreover, when it comes to combinatorial protein transfer, there is no need to produce each protein as a lipidated derivative, which is the case with GPI proteins.

The use of a membrane-anchoring “bridge” protein (in this case, protein A) that can be palmitated and used subsequently with multiple costimulator·Fc_γ1 fusion protein partners, is preferable to the use of costimulator·Fc_γ1 fusion proteins that have been directly palmitated. There are several reasons. First, excess palmitation can interfere with costimulator protein activity. Consequently, the palmitation reaction would have to be titrated for each new costimulator protein (because of variability in the number and distribution of target lysine residues within these proteins) and even for different preparations of the same protein. In contrast, protein A can be palmitated in bulk reactions, and each large preparation can be evaluated afterward for membrane-binding and Fc-binding activities. Second, because directly palmitated costimulator proteins would have palmitate residues attached randomly, the membrane topology of these proteins would vary considerably once anchored to membranes. Third, because protein A has five potential Fc-binding sites per molecule, it is likely that more than one costimulator·Fc_γ1 fusion protein is engaged by each membrane-anchored, pal-prot A molecule. Hence, it is possible that our protein A-based method results in costimulator aggregates, thereby increasing the “functional valency” of bound costimulators. This may be especially significant here, because our Fc_γ1 fusion protein derivatives of B7–1 and 4–1BBL lack their native homodimerization and trimerization motifs, respectively, and were detected as monomers by nonreducing SDS-PAGE.⁴ In contrast, directly palmitated costimulator·Fc_γ1 fusion proteins would not have the added valency conferred by protein A.

The set of protein transfer tools continues to expand. Another two-component protein transfer method has been reported, which uses nitrilotriacetic acid di-tetradecylamine and polyhistidine-tagged costimulators in combination (39). Nonlipidation-based methods, such as the use of a transmembrane protein sequence to directly insert transmembrane-derivatized fusion proteins to tumor cells, have also been explored (40). Together, these various protein transfer methods, along with the expanding array of known costimulators, now offer a growing set of options for tumor cell engineering, both ex vivo and in vivo.

We thank Dr. Tzvete Dentchev for providing essential training for the animal work.