Tumor Cell Surface Expression of Granulocyte-Macrophage Colony-Stimulating Factor Elicits Antitumor Immunity and Protects from Tumor Challenge in the P815 Mouse Mastocytoma Tumor Model

Preclinical tumor models using genetically modified tumor cells to secrete cytokines have been used in efforts to augment the immune response against tumor-associated Ags (1, 2). One of the more promising cytokines for the induction of potent antitumor activity is GM-CSF. This 24-kDa glycosylated cytokine has paracrine as well as autocrine effects on a number of cell types, including monocytes, dendritic cells, eosinophils, and neutrophils (3, 4). In preclinical and clinical studies, GM-CSF secreted from tumor cells has been shown to be a potent stimulator of antitumor responses (5, 6, 7, 8). The consensus from these studies suggests that GM-CSF stimulates APC such as dendritic cells (DC)² to generate potent immune responses.

DC are the most potent APC in the immune system (9, 10, 11) and are able to prime naive T cells almost 30–100 times more efficiently than B cells (12, 13). Langerhans cells (LC) are immature DC that reside in the epidermis and continually sample Ag encountered in this compartment. Once LC receive the appropriate stimulus (e.g., GM-CSF, TNF-α, TGF-β, and LPS), they mature into DC (14) and migrate to lymph nodes, appearing within 24 h and peaking at 2 days after Ag uptake (15, 16), when they initiate the activation of naive T cells. These characteristics have made LC and DC the focus of intense research and attractive targets for immunotherapy. Immunotherapeutic approaches using DC include Ag pulsing of autologous DC (17, 18), transfection of DC with plasmids encoding Ags (19), and fusion of DC to tumor cells (20, 21). Taken together, these studies strongly suggest that to optimally induce an immune response using LC and DC, two requisites must be met. First, LC must be in close proximity to the appropriate Ag(s), and second, LC must receive the appropriate signals to cause maturation and migration of Ag-loaded cells to the lymph nodes to activate naive T cells.

We report here the expression and use of a novel form of GM-CSF anchored to the surface of the mouse P815 mastocytoma line through fusion with a heterologous transmembrane domain. We tested the hypothesis that tumor cells modified to express membrane-bound GM-CSF (mbGM-CSF) would effectively target tumor Ags to DC and provide an effective immune response against the unmodified parental tumor cells.

DBA/2 female mice, 8–10 wk old, and BALB/c nu/nu mice were purchased from Charles Rivers Laboratories (Wilmington, MA).

The following Abs were purchased from PharMingen (San Diego, CA). Anti-GM-CSF Ab MP1-22E9 (a rat anti-mouse GM-CSF mAb), anti-CD8 Ab 53-6.7, 28.14.8 (a mouse anti-L^d Ab), SF1-1.1 (a mouse anti-K^d Ab), 34-2-12 (a mouse anti-D^d Ab), and an isotype-matched control Ab IgG2a.

P815, a mouse (H-2^d) mastocytoma derived from the DBA/2 mouse strain, was a gift from Dr. David M. Kranz (University of Illinois, Urbana, IL). Clones 1D1 and 1D6 are mbGM-CSF-positive cells derived from P815 transfected with the vector made as described below. The clone 1D6.1E5 is a subclone derived from the 1D6 cell line by limiting dilution cloning.

For expression of an mbGM-CSF molecule, the pHOOK-1 plasmid vector was used for cloning (Invitrogen, Carlsbad, CA). Standard DNA cloning techniques were used for construction of the vectors.

The pHOOK-1 plasmid vector originally contained the coding sequence for a single-chain Ab located between the murine κ-chain signal peptide and the platelet-derived growth factor receptor (PDGFR) transmembrane domain coding sequences. The gene encoding the single-chain Ab was removed by cutting with restriction enzymes ApaI and SalI. The resulting vector was treated with calf intestinal alkaline phosphatase (Life Technologies, Gaithersburg, MD) to remove the terminal phosphate groups.

The cDNA of murine GM-CSF was derived from PCR using BALB/c mouse mRNA from spleen cells stimulated with 4 μg/ml Con A for 2 days. The cells were lysed in RNAzol (Life Technologies), and the total RNA was extracted. Oligo(dT) primers were used to prime the synthesis of cDNA from target mRNA. The following PCR primers were used with Taq polymerase in a standard PCR reaction using a Perkin-Elmer Thermocycler (Norwalk, CT): 5′Apa mseGM-CSF, 5′-GCTAGGGCCCTAGCACCCACCCGCTCACCCATCACT-3′; and 3′Sal mseGM-CSF, 5′-ACCGCGGTCGACTTTTTGGACTGGTTTTTTGCATTCAAAGGGG-3′. The resulting PCR fragment was purified and cloned into compatible sites in pHOOK-1 using T4 ligase (Life Technologies).

Electroporation was used for transfecting the plasmid construct into P815. Briefly, cells were grown in log phase using standard tissue culture methods in RPMI 1640 supplemented with 10% FBS and antibiotics. Cells (5 × 10⁶) were electroporated at a voltage of 250 V in the presence of 50 μg linearized plasmid vector. Cells were then incubated with 800 μg/ml G418 (Life Technologies), and subclones were screened by FACS for the presence of GM-CSF on the surface of the cells (as described below).

Cells (10⁶) were washed once with 2% FBS in PBS. The cells were resuspended in 50 μl of wash buffer containing 40 μg/ml rat anti-mouse GM-CSF Ab MP1-22E9 and were incubated on ice for 30 min. Cells were washed twice with wash buffer and were resuspended in wash buffer containing a fluorescein-labeled secondary Ab (goat anti-rat, mouse absorbed; Kirkegaard & Perry Laboratories, Gaithersburg, MD) and were incubated on ice 30 min. After two washes with wash buffer, the cells were resuspended in 500 μl of PBS. In some experiments cells were fixed with 4% paraformaldehyde in PBS. For flow cytometric analysis, a Becton Dickinson FACSort was used (Becton Dickinson, San Jose, CA).

P815 or clone 1D6.1E5 cells (10⁷) were incubated with 250 μg/ml mitomycin C for 30 min at 37°C. The cells were extensively washed with PBS, pH 7.2, containing 5% (v/v) FBS, resuspended in RPMI 1640 medium supplemented with 10% FBS and antibiotics, and then added to round-bottom wells at the amounts indicated. In test wells, 3 × 10⁴ DBA/2 bone marrow cells were added, and the cultures were incubated for 2 days at 37°C in 5% CO₂ in a humidified incubator. The wells were then pulsed with 1 μCi/well [³H]thymidine and harvested the next day using a Cambridge Technology PHD cell harvester (Watertown, MA).

Wild-type P815 cells (10⁶) or mbGM-CSF clones at a concentration of 20 × 10⁶ cells/ml were injected intradermally into the hind flanks of DBA/2 mice. In other experiments, tumor cells were injected intradermally into BALB/c nu/nu mice. Tumors were measured three times per week over the course of the experiment starting 7–10 days after the injections. Tumor sizes are expressed as a product of the longest diameter and the shortest diameter as measured by a calibrated micrometer. Five to ten mice per group were injected.

Wild-type P815 cells (10⁶) or mbGM-CSF clone 1D6.1E5 were irradiated with 20,000 rad in the presence of complete tissue culture medium using a JLSheperd and Associates model 109-85 Irradiator with a ⁶⁰Co source. After irradiation, cells were washed with complete medium once, then extensively with PBS before injection into one flank of DBA/2 mice (10 mice/group). After 15 days the mice were boosted with the same number of cells in the same manner. Mice receiving irradiated wild-type cells were designated the control group. Five days after the boost, both groups of mice were challenged s.c. in the opposite flank with 10⁶ live wild-type P815 cells. Tumors were measured as described above.

Groups of five mice were vaccinated intradermally in the hind flank with 10⁶ irradiated tumor cells with or without mbGM-CSF. Mice were boosted intradermally, and spleens were removed 5 days after the boost. Spleen cells were incubated with irradiated P815 cells at a ratio of 10:1 (spleen cells:tumor cells) for 5 days in complete medium. After stimulation with tumor cells, spleen cells were harvested, separated from dead cells and debris by Nycodenz density centrifugation (Accurate Chemical & Scientific Corp., Westbury, NY), then depleted of CD4⁺ cells by Dynal magnetic bead separation (Dynal, Lake Success, NY). Target cells were prepared by adding 150 μCi of ⁵¹Cr to 3 × 10⁶ P815 cells in log phase growth for 1 h at 37°C with occasional agitation. The effector cells were then incubated at the indicated E:T cell ratios with ⁵¹Cr-labeled P815 cells for 4 h in a humidified 37°C incubator. In Ab inhibition experiments, anti-CD8 Abs or a mixture of anti-class I Abs were used with the final concentrations of Abs indicated. Supernatants were then evaluated for ⁵¹Cr release. The percent specific lysis was calculated as follows:

The mouse GM-CSF gene consisting of 372 nucleotides was amplified from cDNA derived from Con A-stimulated mouse splenocytes, placed under control of the CMV promoter downstream of a murine Ig κ-chain signal sequence, and fused to the sequence of the PDGFR transmembrane domain (Fig. 1). We took advantage of the commercial vector, pHOOK-1, which was originally designed to express a hapten-specific single-chain Ab anchored to the plasma membrane through the PDGFR transmembrane domain (22). Cells transfected with pmbGM-CSF were selected under drug treatment, and subclones were derived from limiting dilution cloning. FACS analysis demonstrated that mbGM-CSF was expressed at levels comparable to that of the endogenous class I molecule, K^d, while isotype-matched control Abs failed to stain cells (Fig. 2). For experiments in which transfected cells were irradiated to halt cell division, we observed that 20,000 rad did not alter the levels of mbGM-CSF on the surface of cells (data not shown). It is interesting to note that P815 cells express approximately 10³ GM-CSF receptors/cell (23); however, we did not observe any significant coagulation or cell-cell clumping compared with those of wild-type cells (W.S.H. and K.A.L., unpublished observations).

One of the important effects of GM-CSF is the stimulation of DC, resulting in their maturation marked by a rapid change from an Ag-sampling cell to APC. For efficient delivery of Ag to naive T cells, we believe it necessary for mbGM-CSF to be biologically active. Therefore, we determined whether mbGM-CSF would be able to mediate a signal to cells that are normally responsive to soluble GM-CSF. We chose to stimulate syngeneic bone marrow cells, which express the GM-CSF receptor and proliferate in the presence of soluble GM-CSF in a dose-dependent manner (24). In Fig. 3, cells expressing mbGM-CSF stimulated the proliferation of bone marrow cells, while the control wild-type P815 cells did not (Fig. 3,A). This stimulation was specifically inhibited by the addition of an anti-GM-CSF mAb in a dose-dependent manner, whereas an isotype control Ab had no effect (Fig. 3 B). GM-CSF also could be detected by ELISA at low levels in supernatants of membrane-bound clones (<8 pg/10⁶ cells/24 h). However, these levels were orders of magnitude below that required for stimulation of bone marrow cells and approximately 4000-fold less than that reported to be biologically relevant in the elicitation of an antitumor immune response in mouse tumor models (25).

We wanted to ensure that any differences we might observe in tumor growth in in vivo experiments were not due to differences in the intrinsic growth rates of the transfected cells vs wild-type cells. Therefore, the growth rates of both wild-type P815 and transfected clones were compared in vitro. Fig. 4,A shows that the growth rates of 1D6.1E5 cells (a subclone of 1D6) were essentially the same as those of wild-type cells. Also, it made no difference whether the transfected cells were growing in the presence or the absence of the G418 selection drug. Growth rates were also determined to be equal in T cell-deficient BALB/c nu/nu nude mice (Fig. 4 B). These data suggest that in the absence of a significant T cell response, cells expressing mbGM-CSF grow as solid tumors at the same rate as wild-type cells. An observation was made in all the in vivo studies with both nude and normal DBA/2 mice that there was a characteristic reduction in the mean tumor size of mice after 18–20 days. This apparent reduction was the result of the death of mice with the largest tumor burdens. We noted that the mean values of tumor size rose again, which represented the continued growth of tumors in the remaining mice.

Normally, P815 cells will grow as solid tumors when injected into a syngeneic host. It was of interest to determine whether there would be a difference between the growth of tumors initiated by transfected clones and wild-type cells. Nonirradiated P815 cells or clones bearing mbGM-CSF (designated 1D1 and 1D6) were injected intradermally into the flanks of DBA/2 mice. Initially, both mbGM-CSF cells and wild-type P815 cells grew similarly in the host. However, on day 12, tumors caused by the mbGM-CSF cells were quickly rejected, while wild-type P815 cells produced tumors that grew to significantly large sizes (Fig. 5). Indeed, for clone 1D6, 100% of animals became tumor free, while clone 1D1 showed low, but measurable, numbers of tumors. These animals were later challenged with live wild-type P815 in the opposite flank and were able to reject these tumors. It was interesting that the FACS comparisons between these two clones showed that clone 1D6 expressed a significantly higher density of mbGM-CSF than did 1D1 (data not shown), which may explain the differences observed between the responses to 1D6 and 1D1 injections. For these reasons, we continued experiments with a subclone of 1D6 termed 1D6.1E5. Preliminary experiments from our laboratory using a transfected P815 cell line expressing a membrane-bound form of IL-4 (associated with a Th2 cytokine response) demonstrated that mbIL-4-expressing cells were not rejected as efficiently as cells expressing mbGM-CSF (W. Soo Hoo and J. R. Kohrumel, unpublished observations).

Prevention of the growth of nonirradiated tumor cells may be a function of innate immunity responding to the proinflammatory effects of GM-CSF as opposed to the longer lasting effects of an adaptive immune response. To evaluate whether the mbGM-CSF-expressing cells (1D6.1E5) could elicit a protective and systemic antitumor response, irradiated cells were used to vaccinate mice before challenge with nonirradiated wild-type P815 cells. DBA/2 mice were injected intradermally with 10⁶ cells in the left flank and boosted 15 days later with the same number of cells in the same flank. Five days after the last vaccination, the mice were challenged s.c. with nonirradiated wild-type tumor cells in the opposite flank. Although all mice developed palpable tumors in the first 2 wk, only animals vaccinated with mbGM-CSF cells were able to reject their tumors completely. By 30 days after the initial challenge with wild-type tumor cells, the treated group did not show any signs of tumor growth and remained tumor free throughout the remainder of the experiment. The control group, however, grew large tumors, and 50% of the mice died (Fig. 6, inset). In another experiment, mice were prevaccinated with half the dose (5 × 10⁵ cells), and a group was added that received no vaccination before challenge. Once again, the majority of mice vaccinated with the mbGM-CSF cells were able to completely reject their tumors, while no significant difference was observed between either the mice vaccinated with wild-type P815 cells or those receiving no vaccination (Fig. 7).

In light of the results of tumor growth in nude mice and the kinetics of tumor rejection in other experiments, it seemed reasonable to assume that the antitumor activity observed in normal mice was due in large part to a proliferation of tumor-specific T cells. To test this idea, mice were immunized with 10⁶ irradiated mbGM-CSF clone 1D6.1E5 or wild-type P815 cells. Five days after a boost with the same number of cells, splenocytes were isolated from all mice and given one round of stimulation by irradiated wild-type P815 cells. The question was then asked whether CTL from these preparations could kill wild-type P815 tumor cells. In Fig. 8,A, mice that have been vaccinated with mbGM-CSF cells produce significantly more anti-P815 CTLs than mice receiving the wild-type vaccination. Anti-CD8 Abs (at a final concentration of 125 μg/ml) could inhibit the specific killing (Fig. 8,B), indicating that the killing was CD8⁺ CTL dependent. Furthermore, when a mixture of anti-class I Abs (L^d, K^d, D^d, each at 42 μg/ml final concentration) was used, they also inhibited the killing of P815 target cells, although to a slightly lesser degree (Fig. 8,B). Isotype control Abs failed to inhibit CTL killing of P815 targets (data not shown). In a number of experiments, the mbGM-CSF cells consistently elicited CTL in a manner superior to P815 wild-type cells (Table I).

To our knowledge, this is the first report of a soluble cytokine genetically engineered to be expressed in a membrane-bound form. There are examples of cytokines expressed as membrane-anchored proteins; however, these represent natural alternative splice forms (Refs. 26 and 27; reviewed in 28). Earlier work by Tao and Levy showed that GM-CSF fused to a single protein Ag was a potent immunogen and that the GM-CSF component of the fusion protein was biologically active (29). Also, the three-dimensional structure of GM-CSF described by Diedrichs et al. suggested that the molecule could be fused through its carboxyl terminus without hindering the putative receptor contact points (30). In light of this, we thought it reasonable that anchoring the GM-CSF molecule to the cell surface could be accomplished while maintaining its biological activity. Thus, we have taken advantage of the high affinity interaction between GM-CSF and its cognate receptor CD116/CDw131 (K_d = 120 pM) (31) to target tumor Ags to epidermal LC through direct cell-cell interaction.

In this study we have demonstrated that a novel mbGM-CSF on the surface of the P815 mastocytoma cell line can generate an antitumor immune response in syngeneic DBA/2 hosts. The mbGM-CSF molecules are expressed at levels comparable to those of class I molecules (Fig. 2) and are able to stimulate bone marrow cells in an in vitro analysis demonstrating that the membrane-bound molecules have retained biological activity (Fig. 3). Based on our assays using soluble recombinant mouse GM-CSF, we calculate that a response yielding a stimulation index of 20 was equivalent to 2.6 × 10¹⁰ molecules of soluble GM-CSF. The number of 1D6.1E5 cells that produced the equivalent stimulation was 3 × 10⁵ cells. Assuming that the activity/binding affinity of mbGM-CSF is the same as that of soluble GM-CSF and that each mbGM-CSF molecule on the surface has an opportunity to bind receptor, we estimate that there are 10⁵ molecules of mbGM-CSF/cell. In studies using ¹²⁵I-radiolabeled anti-L^d mAbs, the P815 tumor line expressed approximately 6 × 10⁵ molecules of L^d on the surface (40).

The P815 tumor cell line is known to be moderately immunogenic in the syngeneic host (32), and repeated vaccinations with wild-type P815 cells result in some antitumor immunity. For this reason, we determined that the most appropriate controls for these studies are mice vaccinated with wild-type cells. In the experiment in which nonirradiated mbGM-CSF P815 cells were injected into mice, tumors grew during the first 10–12 days. After this time, however, the tumors were quickly rejected until no palpable tumor was detected (Fig. 5), while wild-type cells grew to large sizes, resulting in 50% mortality. One explanation for this initial growth is that the priming of the immune response through the interaction of GM-CSF and DC results in a short lag period before sensitized CTL can be generated in sufficient numbers. In contrast, Nakajima et al. demonstrated that rejection of P815 cells transfected to express CD40L was immediate (i.e., no tumor growth was observed), and they showed that this rejection was due to a significant contribution of NK cells (33). When the mbGM-CSF cells were injected into BALB/c nude mice, the transfected cells grew at the same rate as wild-type cells. This is in contrast to the results seen in the Nakajima report in which CD40L-P815 cells suppressed tumor growth in nude mice, suggesting that the mechanism of antitumor immunity using mbGM-CSF is different from that used by the CD40-CD40L system. Our results are in agreement with those obtained by Sampson et al., who reported that the contribution of NK cell activity elicited by soluble GM-CSF from modified B16 melanoma cells is measurable, but relatively minor (6). This further suggests that the immunity elicited by mbGM-CSF cells is T cell dependent and that very little, if any, innate immunity is involved.

The rejection of live mbGM-CSF cells demonstrated the elicitation of an immune response directed to the modified tumor cells. This is in contrast to an initial observation in a study by Dranoff et al. in which they report that injection of live tumor cells secreting soluble GM-CSF grew progressively, inducing lethal toxicity and hepatosplenomegaly (5). However, when used as an irradiated vaccine, the GM-CSF-secreting cells induced strong antitumor immunity. From this observation one might speculate that the mbGM-CSF may be safe, since injection of live cells did not produce any lethal toxicity. Taken together with the data we have shown in athymic mice (Fig. 4 B), we hypothesized that the rejection of live mbGM-CSF cells is mediated by a T cell-dependent systemic immune response and that this response to 1D6.1E5 cells may extend to the parental P815 cells.

The final goal of this study was to show that CTL generated using a vaccine composed of irradiated mbGM-CSF cells could elicit an immune response directed to the unmodified, parental tumor cell line both in vitro and in vivo. When used as a cell vaccine, the mbGM-CSF P815 cells were consistently superior to similar vaccinations of wild-type P815 cells in the elicitation of tumor-specific CTL (Fig. 8 and Table I) against the challenge of unmodified tumor cells. These observations are in agreement with the idea that GM-CSF is a potent cytokine adjuvant for the elicitation of antitumor responses. Since the major tumor Ags (e.g., p1A and p2Ca) of P815 are known, we plan to examine the specificity of CTL generated to the mbGM-CSF P815 cells with regard to their recognition of specific peptide Ags. The demonstration of tumor-specific CTL does not rule out the possibility that a relevant humoral immune response was also elicited. Future studies will investigate whether cells or sera from immunized animals can be adoptively transfered to convey protection from a tumor challenge or induce eradication of established tumors.

The mbGM-CSF differs from the current cytokine secretion paradigms in two distinct ways. First, contact with mbGM-CSF by GM-CSF receptors on dendritic cells requires direct physical contact with the cellular vaccine. In strategies where cytokines are secreted into the extracellular milieu, it may be tempting to speculate that DC would receive maturation signals coming from a gradient concentration of soluble GM-CSF without being close enough to take up specific Ags optimally. Also, it has been noted by others that the use of secreted, soluble cytokines requires relatively high rates of secretion for prolonged periods (∼36 ng/10⁶ cells/24 h) (24). Depending on the mode of transfection and the clones isolated, this may prove to be a serious limitation. The second distinction of this strategy lies in the ability to engage multiple GM-CSF receptors on a given DC. Stimulation of LC with GM-CSF causes maturation of LC in culture (34, 35, 36), resulting in the up-regulation of costimulatory molecules and increased expression of class I and class II molecules (37, 38). An additional signal is provided by the local production of TNF-α, which is responsible for the rapid migration of these cells to lymph nodes (39). We postulated that a high avidity cell-cell contact with multiple GM-CSF receptors on LC would cause the LC to experience a stronger signal transduction via the GM-CSF receptor α- and β-chains for efficient Ag uptake and rapid migration to lymph nodes. One of the aims of this system was to optimize the transduction of the signal that leads to the maturation and migration of DC. By increasing the avidity (i.e., the number of receptor-ligand interactions) between the tumor cell and the APC, we propose that the signal transduction to the DC is optimized.

The data shown here provide evidence that specific, systemic antitumor immunity can be elicited by tumor cells expressing GM-CSF on their surface. We believe that the adjuvant effect provided by mbGM-CSF is a result of the Ag-presenting DC in physical contact with the source of Ag (tumor cell), thus efficiently stimulating the antitumor response. Whether mbGM-CSF is more efficacious than secreted GM-CSF remains to be tested, and we are conducting experiments in other tumor models to directly compare the membrane-bound and secreted forms of GM-CSF.

We thank René Aleman and Erin Gouveia for their technical support in the in vivo tumor experiments and Joji P. Dively for additional FACS analysis support.