Supraagonistic, bispecific single-chain antibody purified from the serum of cloned, transgenic cows induces T-cell-mediated killing of glioblastoma cells in vitro and in vivo

Abstract

Here we characterize the antitumor activity of a recombinant bispecific single-chain antibody isolated from the serum of cloned transgenic cows. The antibody, termed r28M, is directed to a melanoma-associated proteoglycan, also expressed on glioblastoma cells, and to human CD28. Bound to tumor cells, r28M induced exceedingly efficient supraagonistic T-cell activation via the CD28 molecule without an additional stimulus via the TCR/CD3 complex. In vitro, T cells and NK cells contributed to tumor cell killing after r28M-mediated activation of peripheral blood mononuclear cells. However, NK activity depended on T-cell-derived cytokines. In vivo, r28M markedly inhibited the growth of human glioblastoma cells in nude mice. The serum half-life of the protein after i.v. injection was approximately 6 hr. Thus, r28M is unique not only in inducing supraagonistic CD28-mediated T-cell activation against tumor cells in vitro and in vivo, it also meets 2 additional requirements that are critical for clinical application: a relatively long serum half-life and the possibility of obtaining large amounts of active material from cloned transgenic livestock. © 2005 Wiley-Liss, Inc.

Under physiologic conditions, effective T-cell activation requires stimulation not only of the antigen-specific T-cell receptor (TCR)/CD3 complex but also of receptors for costimulatory signals. These molecules are therefore viewed as amplifiers of a specific immune response and as promising targets for the manipulation of immune reactions.1 Although a considerable number of such receptors on T cells have been identified to date, the CD28 molecule and its ligands, CD80 (B7-1) and CD86 (B7-2) on antigen-presenting cells, still constitute the prototype of a costimulatory pathway for naive T cells.2

In recent years, it has become clear that under certain conditions, the CD28 molecule may induce effective T-cell activation on its own, without a first signal via the TCR/CD3 complex. This phenomenon was originally described by Tacke et al.3 using particular superagonistic antibodies against rat CD28. Recently, the same group has generated superagonistic antibodies directed to human CD28, which bind to a particular epitope of the CD28 molecule.4

We previously reported that a recombinant bispecific single-chain antibody, r28M, directed to a melanoma-associated proteoglycan (MAPG) and human CD28 induces effective tumor cell-restricted T-cell activation without stimulation of the TCR/CD3 complex.5 However, tumor cell killing by r28M activated peripheral blood mononuclear cells (PBMCs) as measured in a 4-hr [⁵¹Cr]-release assay was largely mediated by non-T cells and thus the role of the T cells in the effector phase of r28M-mediated T-cell activation remained unclear. It was also unclear whether r28M-mediated T-cell activation and tumor cell killing would be effective under conditions occurring in vivo.

Recently, we have succeeded in isolating the protein from the serum of cloned transgenic cows.6 Here we use this purified protein to characterize its activity against glioblastoma cells not only in vitro but also in a xenogeneic mouse tumor model.

Material and methods

Cell culture

Cells were grown at 37°C in RPMI-1640 medium supplemented with 10% heat-inactivated fetal calf serum, 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin and 1 mM sodium pyruvate in a humidified atmosphere containing 5% CO₂. The human glioblastoma cell line U87MG was kept in DMEM and tested regularly for mycoplasma contamination using a PCR-based detection kit (VenorGeM; Minerva Biolabs, Berlin, Germany). PBMCs were isolated from heparinized blood of healthy donors by density gradient centrifugation (Ficoll Hypaque; Pharmacia, Freiburg, Germany).

T cells were purified from PBMCs using a mixture of MACS beads coated with CD4 and CD8 antibodies. Cells were then passed over a magnetic LD column and the flow-through was used as T-cell-depleted population. For some experiments, r28M-activated cells were further depleted on a second LD column after incubation with anti-CD56-coated MACS beads. For positive selection, T cells eluted from the first column were separated a second time using an LS column and eluted according to the instructions of the manufacturer (Miltenyi Biotech, Bergisch Gladbach, Germany).

Fab fragments of parental antibodies against the MAPG antigen and CD28 were obtained from intact antibodies purified from hybridoma supernatant by protein A affinity chromatography. The antibodies were digested with pepsin, modified with DTNB and purified by size exclusion chromatography as described.5

Purification of r28M protein from cow serum

The generation of cloned transgenic cows carrying the r28M gene and a partial purification procedure from cow serum has been previously described.6 Briefly, coagulated blood of transgenic cows was sedimented by centrifugation at 400g. The serum was pooled, sterilized by microfiltration and either stored at −80°C or used immediately for purification. To this end, serum was precipitated twice at 14% polyethylene glycol 3350 (Sigma, Taufkirchen, Germany) and centrifuged (5,000g for 30 min at 4°C). It was then diluted in PBS/0.1 M glycine containing 1% Triton X114 and passed over a Protein L-sepharose column (Perbio, Bonn, Germany). Bound material was stepwise-eluted with a 0.1 M citrate/0.1 M glycine buffer system. r28M eluted at pH 2–3. The protein concentration was determined by spectrophotometry and a colorimetric assay (Nanoquant; Roth, Karlsruhe, Germany) based on the Bradford procedure.

Gel filtration chromatography was performed using a PC 3.2/30 Superdex 200 column (SMART System; Pharmacia, Freiburg, Germany) and the following calibration proteins (Pharmacia): thyroglobulin (669 kDa; bovine thyroid), ferritin (440 kDa; horse spleen), catalase (232 kDa; bovine liver), aldolase (158 kDa; rabbit muscle), albumin (67 kDa; bovine serum), ovalbumin (43 kDa; hen egg), chymotrypsinogen A (25 kDa; bovine pancreas) and ribonuclease A (13.7 kDa; bovine pancreas).

In vitro assays

Target cell-dependent T-cell proliferation and tumor cell killing induced by the bispecific antibody fragment r28M were measured as described.5 Briefly, r28M was incubated in triplicates in 96-well microtiter plates with x-irradiated (120 Gy) U87MG tumor cells (5,000 cells per well) and PBMCs (10⁵ cells per well) from healthy donors. During the last 16 hr of a 3-day incubation period, [³H]-thymidine was added (18.5 kBq per well). Cells were harvested on a filtermate and incorporated [³H]-thymidine was measured in a scintillation counter (MicroBeta, Wallac, Finnland). Tumor cell killing was determined as described5 under similar experimental conditions. In this case, viable rather than irradiated tumor cells were used. The assay time was 5 days and the target effector ratio 1:40. The number of viable, adherent tumor cells was determined after washing the plates and adding the tetrazolium salt WST (Roche Diagnostics, Penzberg, Germany).

Alternatively, the lytic activity of cells preactivated with r28M for 4 days in the presence of tumor target cells was measured using a 5-hr and 20-hr [⁵¹Cr]-release assay and U87MG cells as targets. Determination of IL-2 secreted by activated PBMCs was performed by ELISA using a matched pair of capture and biotinylated antibodies purchased from Pharmingen (Heidelberg, Germany).

Animal experiments

All animal work was carried out in accordance with the NIH guidelines Guide for the Care and Use of Laboratory Animals and with paragraph 8, section 1 of the Law for Treatment of Animals in Germany (ethical vote for the applied nude mice xenograft model N3/03 by the Regierungspräsidium Tübingen, Germany).

For determination of the in vivo antitumor activity of r28M, 5 × 10⁴ U87MG human glioblastoma cells were stereotactically implanted into the right striatum of athymic mice (CD1 nu/nu; Charles River, Sulzfeld, Germany). Mice (6 animals per group) were treated at day 7 after inoculation by intralesional application of 5 × 10⁶ freshly isolated human PBMCs and 5 μg of the r28M protein dissolved in carrier solution (PBS substituted with 3.3% human albumin). The r28M batch used for therapy contained less than 0.5 EU of pyrogen per mg of purified protein as determined by the Limulus Amebocyte Lysate test (QCL-1000; BioWhittacker, Heidelberg, Germany). Control mice were injected with identical amounts of PBMCs or PBMCs and a mixture of Fab fragments (5 μg) of the parental anti-MAPG and anti-CD28 antibodies.

To determine the in vivo half-life of r28M molecules, 50 μg of purified protein in a final volume of 150 μl PBS were injected i.v. in C57BL/6 mice. Collection of blood samples was started 30 min after injection and was repeated over 47 hr. Blood was drawn by retroorbital sinus punction and concentration of the r28M protein was determined by flow cytometry. To this end, CD28-positive Jurkat cells were incubated with 1:10 diluted serum samples, washed and stained with phycoerythrin-labeled F(ab′)₂ fragments of a polyclonal goat antimouse IgG antibody (Dianova, Hamburg, Germany). Analysis of stained cells was performed in a FACSCalibur flow cytometer equipped with the Cellquest software (Becton Dickinson, Heidelberg, Germany). For quantitation, a standard curve was generated using different concentrations of purified r28M protein diluted 1:10 in serum of C57BL/6 mice.

Results

Figure 1(a) shows the schematic structure of the bispecific single-chain molecule r28M consisting of 2 monospecific V_H-V_L single-chain F_V fragments linked by a particular sequence (L2) derived from the upper elbow region of the human C_H1 domain. As observed with material purified from cell culture supernatant of transfected Sp2/0 cells,5 the protein isolated from the serum of cloned cows was found to be in a partially dimeric form as revealed by gel filtration analysis (Fig. 1b).

Figure 1. (a) Schematic representation of the r28M structure. Two monospecific single-chain F_V fragments with a glycin serin linker (L1) between their V_H/V_L regions are connected by a second linker (L2) to form a bispecific single-chain molecule with a calculated molecular weight of 57.37 kDa. (b) Gel filtration of the r28M protein isolated from the serum of cloned transgenic cows on a PC 3.2/30 Superdex 200 column. Elution volumes and molecular weights (kDa) of calibration proteins are also depicted (filled circle). Purified r28M yielded 2 peaks with elution volumes well in accordance with calculated molecular weights of monomeric and dimeric versions of the protein (filled square).

In Figure 2(a), T-cell proliferation induced in PBMCs by various concentrations of r28M in the presence of MAPG-positive and -negative target cells is depicted. Proliferation is antigen-restricted, that is, it requires crosslinking of target and effector cells by the bispecific antibody. Unrestricted proliferation in the absence of MAPG-positive or any target cells was observed in some experiments at antibody concentrations ≥ 1 μg/ml. It varied from background values (500–1,000 cpm) to approximately 30,000 cpm (data not shown), when PBMCs from different donors were used. In all experiments, however, T-cell proliferation was target cell-restricted over at least a 2 log concentration range. Likewise, if IL-2 secretion rather than proliferation was measured, a pronounced target cell restriction was observed (Fig. 2b).

Figure 2. (a) Target cell-restricted T-cell proliferation induced by r28M. PBMCs were incubated with the indicated concentrations of r28M in the presence of irradiated MAPG-positive U87MG glioblastoma cells (filled circle) or MAPG-negative SKBr3 mammary carcinoma cells (open circle). Proliferation under control conditions is shown as bars on the right. In (b), PBMCs were incubated with the indicated concentrations of r28M with (filled circle) and without (open circle) U87MG glioblastoma cells. After 24 hr, the amount of secreted IL-2 was measured by ELISA.

That T-cell activation, as depicted in Figure 2, leads to tumor cell killing is shown in Figure 3. After 5 days of coculture with PBMCs and the r28M protein, U87MG glioblastoma cells were killed almost completely. MAPG-negative SKBr3 cells were not affected (Fig. 3a), although the sensitivity of these cells toward killing by activated NK and T cells is comparable to that of U87MG (data not shown). If T-cell-depleted PBMCs were used in this assay, no significant killing of the U87MG cells was observed, whereas purified T cells were highly active (Fig. 3b). This is in marked contrast to what was previously observed if cells were separated after (and not before) a 4-day activation phase with r28M and then tested for lytic activity in a chromium-release assay. In such experiments, activated PBMCs depleted of T cells were highly active against a variety of target cells.5 Figure 3(c) demonstrates that T-cell-depleted preactivated cells are also capable of killing U87MG and that this activity is largely mediated by CD56⁺ cells as shown by further depletion with anti-CD56-coated beads. In contrast, depletion of non-T cells with anti-CD14-coated beads had no significant effect on the lytic activity of such cells (data not shown).

Figure 3. Killing of tumor cells by PBMCs and r28M. In (a), MAPG⁺ U87MG glioblastoma cells (filled circle) and MAPG⁻ SKBr3 mammary carcinoma cells (open circle) were incubated with PBMCs (E:T ratio 40:1) and r28M. After 5 days, the viability of the tumor cells was assessed using the WST dye. In (b), the assay was performed with PBMCs from another donor, U87MG cells as targets and the following PBMC subpopulations purified by magnetic cell separation: purified T cells (filled circle), T-cell-depleted PBMCs (open circle). In (c), PBMCs were cultivated in the absence (−) or presence (+) of r28M (0.15 μg/ml). After 4 days, PBMCs were depleted of T cells and CD56⁺ cells as indicated and tested for lytic activity in a 5-hr (black bars) and 20-hr (dashed bars) [⁵¹Cr]-release assay. Target effector ratio 1:40. Results depicted in (a–c) are representative for 2–6 similar experiments with PBMCs from different healthy donors.

Next, we examined whether the process of r28M-mediated T-cell activation and tumor cell killing is effective in preventing growth of tumor cells in vivo. To this end, we inoculated U87MG cells into the striatum of nude mice. At day 7 after inoculation, we injected a single dose of r28M antibody together with freshly isolated PBMCs. This treatment resulted in a highly significant suppression of tumor growth. At day 50, survival was 100% in the r28M group and 0% in the control groups. Control animals injected with PBMCs alone or with PBMCs plus a mixture of Fab fragments derived from parental anti-MAPG and anti-CD28 antibodies did not survive longer than mice injected with PBS (Fig. 4).

Figure 4. Survival of nude mice after intracerebral inoculation of U87MG cells and various treatments. Seven days after inoculation of tumor cells into the striatum of nude mice, animals received carrier solution (PBS with 3.3% human albumin; filled circle); 5 × 10⁶ PBMCs diluted in carrier solution (open circle); a mixture of anti-MAPG and anti-CD28 Fab fragments (5 μg) and 5 × 10⁶ PBMCs (filled triangle); 5 μg r28M antibody and 5 × 10⁶ PBMCs (open triangle).

Figure 5 shows the time-course of serum concentration in 3 mice after i.v. injection of 50 μg of r28M. The half-life of this protein in normal C57BL/6 mice is around 6 hr. Two days after injection, the protein is clearly detectable in mouse serum at concentrations that are biologically highly active (0.4 μg/ml; Figs. 2 and 3).

Figure 5. Serum elimination of the r28M protein. Concentration of r28M protein in the serum of 3 C57BL/6 mice at various time points after i.v. injection of 50 μg of the protein. Concentration of r28M was determined using flow cytometry with CD28-positive Jurkat cells, which were incubated with 1:10 diluted serum samples. Purified protein diluted 1:10 in mouse serum was used as a standard.

Discussion

One of the aims of this study was a further characterization of the lytic activity induced in PBMCs after tumor cell-restricted supraagonistic CD28 stimulation with the bispecific r28M antibody. We have previously reported that non-T cells significantly contribute to the lytic activity of r28M-activated PBMCs toward melanoma cells. Here we confirm this finding using glioblastoma cells as targets and show that the lytic activity of the non-T-cell population is largely mediated by CD56-positive NK cells. However, these cells cannot act independently of T cells. If they are separated from T cells and then used as effector cells in a long-term cytolytic assay, they are not active. This finding rules out that r28M directly stimulates CD28, which may be expressed on some NK cells.7, 8 Rather, NK cells seem to become activated by T-cell-derived cytokines, e.g., IL-2. In fact, we found that the amount of IL-2 generated during r28M-mediated T-cell activation well exceeds the concentrations necessary to induce lymphokine-activated killer (LAK) activity in both cell populations. It is noteworthy that the secretion of IL-2 and other cytokines such as IFN, TNF and IL-4 (data not shown) is strictly dependent on the presence of the U87MG target cells over a wide range of antibody concentration (Fig. 2b). Thus, one may speculate that upon clinical application of r28M, a cytokine release syndrome as observed after treatment with anti-CD3 containing bispecific antibodies9 may be avoidable.

That the process of T-cell activation and tumor cell killing is effective not only in vitro but also under in vivo conditions was shown in a nude mouse model. Treatment of such mice with r28M and PBMCs 7 days after inoculation of U87MG glioblastoma cells resulted in a marked delay of tumor growth.

A major concern for the clinical application of antibodies in a nonphysiologic recombinant format is their serum half-life. The scarce data available so far indicate relatively rapid elimination of bispecific single-chain antibodies with a half-life of 30–120 min.10, 11, 12 In marked contrast, the half-life of r28M is considerably longer. The concentrations detectable more than 2 days after injection of a single dose of 50 μg in the serum of normal mice are still more than a factor of 10 above those required for the induction of T-cell proliferation or tumor cell killing in vitro. The relatively long serum half-life of r28M may be attributable at least in part to its tendency to form dimers. In fact, we noted that more than 50% of the material purified from the serum of cloned cows is in a dimeric form, as revealed by analytical size exclusion chromatography (Fig. 1b). This is in accordance with previously published results using material isolated from tissue culture supernatant.5 Thus, concerns that a short half-life precludes clinical application of bispecific single-chain antibodies do not appear valid as far as the r28M antibody is concerned.

A merely technical but nevertheless often insurmountable hurdle for the clinical application of bispecific antibodies is the expensive production procedure. For antibodies in a recombinant format, protein concentrations in conventional tissue culture systems rarely exceed 5 μg/ml, not enough for economical large-scale production. A possible solution to this problem might come from transgenic plants or animals, which produce the protein of choice. All experiments described in this report were performed with material purified from the serum of cloned transgenic cows. The amount of protein in the serum of these animals reached 50–100 μg/ml, which is considerably higher than concentrations obtained in conventional culture of transfected cells (1–2 μg/ml). In some of our experiments, the activity of serum-derived material was compared to that of protein isolated from cell culture supernatant and found to be not different (data not shown). Thus, a reliable production system for the r28M protein is available. We are now in the position to initiate clinical pilot studies in which the protein is used to treat patients with metastatic malignant melanoma or glioblastoma.

Acknowledgements

The authors express their special thanks to Elke Hoffmann, Beate Pömmerl and Claudia Falkenburger for technical assistance.