A recombinant single-chain antibody interleukin-2 fusion protein

Recombinant interleukin-2 (rIL-2) therapy has been shown to be of value in the treatment of some cases of melanoma and renal cell carcinoma. However, its use can be limited by severe systemic toxicity. Targeting rIL-2 to the tumor should improve the antitumor immune response and decrease the systemic toxicity. With this aim, we have employed recombinant DNA techniques to construct a single-chain antibody interleukin-2 fusion protein (SCA-IL-2). The protein used in this model system consists of the variable domains of the antilysozyme antibody D1.3 fused to human IL-2 and is expressed inE. coli. It retains antigen-binding specificity and has the full biological activity of rIL-2. This approach can be taken to generate SCA-IL-2 proteins that bind to appropriate cellular antigens. In vivo administration of tumor-binding SCA-IL-2 should result in a localized high concentration of rIL-2 in the tumor tissues, maximizing the antitumor response while keeping systemic side effects to a minimum.

Interleukin-2 (IL-2) is a 15 kDa cytokine produced by T helper cells that stimulates cytotoxic T lymphocytes and NK cells (Gillis et al., 1978). Bacterially produced recombinant IL-2 has been used clinically in the treatment of melanoma and renal cell carcinoma to stimulate cancer patients' immune systems (Rosenberg et al., 1989). Recent pre-clinical studies indicate that achieving a prolonged high dose of IL-2 in the tumour can result in the induction of a long lasting anti-tumour response leading to the rejection of an otherwise lethal tumour (Fearon et al., 1990). However the in vivo efficacy is limited by difficulties in maintaining prolonged high doses in the tumour and by the severe systemic toxicity associated with high dose IL-2 therapy (Rosenberg et al., 1989).
To achieve a selective and prolonged concentration of IL-2 in the tumour it is an attractive idea to target it there via an antibody delivery system. IL-2 has been successfully incorporated into a number of fusion proteins. Fusion proteins consisting of IL-2 linked to bacterial toxins have been produced in bacteria and have been demonstrated to be toxic to IL-2 receptor bearing cells (Lorderboum-Galski et al., 1988;Williams et al., 1987). With the aim of concentrating IL-2 activity in the tumour, two different antibody-IL-2 fusion proteins have already been described. A Fab'-IL-2 fusion protein, whilst only partially retaining IL-2 activity, has been shown to increase the T-cell mediated killing of antigenbearing tumour cells in vitro (Fell et al., 1991). A larger IgG-IL-2 fusion protein appears to retain full IL-2 activity and is likewise able to increase effector cell mediated killing in vitro (Gillies et al., 1992). There are potential problems with these molecules in that they are produced in expensive mammalian expression systems and that their large size may result in poor tumour penetrance and prolonged blood residues in vivo. A smaller antibody-IL-2 fusion protein based on a bacterially produced antibody fragment may give economic and therapeutic advantages. As a delivery system the smaller single chain antibody (SCA), comprising linked variable heavy (VH) and variable light (VL) chain antibody domains shows great promise (Huston et al., 1988). Where tested SCAs demonstrate good tissue penetration (Yokota et al., 1992), rapid renal clearance of non-localised protein and potentially low immunogenicity (Colcher et al., 1990). Recent advances with in vitro selection should allow the rapid and economic production of SCA of any required specificity (Clackson et al., 1991;Marks et al., 1991). In this preliminary investigation the feasability of producing a functional SCA-IL-2 was tested using the antilysozyme SCA D1.3. The advantages of using this antibody are that it is well characterised, it is secreted well and it is easy to detect by Elisa and to affinity purify. As we employed a SCA unsuitable for cytotoxicity assays we are not able to add to the already strong evidence for the benefits of concentrating IL-2 in tumours. However we are able to demonstrate for the first time that this fusion protein, in addition to retaining antigen binding ability, possesses the immunostimulatory actions of IL-2 when tested with lymphocytes bearing the high affinity IL-2 receptor. Furthermore the SCA-IL-2 fusion molecule described here retains the ability to stimulate cells expressing low affinity IL-2 receptors as measured by its proliferative effects on human peripheral blood lymphocytes.

Materials and methods
Plasmid assembly Single colonies of E.coli containing plasmid pSV-HIL2-0 (Gift of Professor W. Friers, University of Ghent) were suspended in 500 fcl of water, boiled for 5 min and cleared by centrifugation in a microfuge. Aliquots (10 cLI) of the supernatant were subjected to polymerase chain reaction (PCR) amplification according to the manufacturers instructions (Perkin Elmer Cetus, Norwalk, USA) in the presence of 25 pmol each of oligonucleotide primers IL-2/7 and IL-2 3'. The reaction underwent 30 cycles of denaturation (94°C, 1 min), annealing (50°C, min) and extension (72C, 1 min). Oligonucleotide primer IL-2/7, (5'-ACCAAGCTCGAGATC-AAACGGGAACAAAAACTCCCTACTTCAAGTTCT-3') direct incorporation of an Xho I site and the seven carboxyterminal amino acids of the VL chain domain of the D1l.3 SCA plasmid (pSWsFVDI.3myc, McCafferty et al., 1990; Gift of Dr E.S. Ward, LMB, Cambridge) fused to sequence encoding amino acids 2-7 of human IL-2. Primer IL-2 3', (5'-TTCTCGAATTCGAGCTGGATCCTTATTAAGTCAG-TGTTGAGATGAT-3'), directs the incorporation of an EcoRI site downstream of the termination codon of human IL-2. The 440 bp amplified product was isolated from a 1.5% agarose gel, digested with Xho I and EcoRI and ligated between the Xho I and EcoRI sites of plasmid pSWsFVD 1.3myc to generate plasmid pSCA-IL-2/7. This plasmid bears a chimeric gene encoding a single chain antibody-IL-2 fusion protein (SCA-IL-2) under lac transcriptional control. Figure  1 shows plasmids used, the nucleotide sequence and the deduced amino acid sequence at the fusion junction.  Expression and partial purification of SCA-IL-2 protein Cultures (500 ml) of E.coli K12 KS476 (Stauch et al., 1989; Gift of Professor J. Beckwith, Harvard) transformed with plasmid pSCA-IL-2/7 were grown overnight at 37°C in 2 x TY broth supplemented with appropriate antibodies and, to ensure growth in repressing conditions, 1% glucose. Cells were harvested, washed twice in sterile standard phosphate buffered saline (PBS pH 7.6) at 27°C, suspended in fresh growth media containing appropriate antibiotics, supplemented with 0.1 mM isopropyl-b-D-thiogalactopyranoside (IPTG), and shaken for 16 h at 25°C to permit accumulation of the fusion protein. After expression, cells were harvested, the bacterial growth supernatant was filtered (0.22 mm filter) and applied at room temperature to a lysozyme-Sepharose column. After washing with PBS, bound protein was eluted as described previously (Ward et al., 1989). Prior to use the fusion protein was dialysed exhaustively against PBS and stored at -20°C.

ELISAs
For serological detection of SCA-IL-2 fusions proteins, enzyme linked immunosorbent assays (ELISAs) were employed. Flat bottomed Dynatech Immulon 96 well plates were coated overnight at 25C with hen egg lysozyme (300 jig ml-'), goat polyclonal anti-IL-2 antibody (501g ml-', British Biotechnology Oxford) or other proteins (BSA, KLH, reconstituted milk powder, or insulin at appropriate concentrations) applied in 50 mM bicarbonate buffer pH 9.6. Unoccupied sites were blocked with a 1% solution of milk powder in PBS for 30 min at 25°C. Bacterial supernatants, affinity purified material or recombinant IL-2 (rIL-2, Boehringer Mannheim, Germany) were diluted in PBS/l % milk powder and incubated (30 min, 25°C) in appropriate wells. After three washes in PBS, bound protein was detected with either polyclonal anti IL-2 or DMS1 (Smith et al., 1983), a murine monoclonal antibody that recognises the receptor binding site of human IL-2 (Gift of Professor K.A. Smith, Dartmouth Medical School, USA). After a further three washes, bound antibody was detected with a species specific HRP conjugated antibody (DAKO, Copenhagen, Denmark). ABTS was added to generate a colour change that was monitored at 405 nm.

Western blots
Affinity purified protein samples were electrophoresed through 15% 30:1 acrylamide: bis-acrylamide gels essentially as described in Laemmli (Laemmli, 1970) and transferred electrophoretically to a nitrocellulose membrane (Towbin et al., 1979). The membrane was blocked (30 min, 25°C) in a 1% solution of milk powder in Tris-buffered saline/0.0025% Tween 20 (TBST). Proteins were detected by incubation (1 h, 25°C) with either rabbit anti-SCA serum (Gift of E.S. Ward) or mouse monoclonal DMS1. After five washes in TBST, bound first step antibody was detected with anti-rabbit or anti-mouse AP conjugated antibody (Amersham, Bucks, UK) and revealed by incubation with a solution of NBT and BCIP (Promega, Madison, USA), according to the manufacturers recommendations.
FACS analysis CTLL-2 cells (Gillis & Smith, 1977) deprived of IL-2 for 12 h were seeded into Nunc 96 well plates at 105 per well in 200 flI volumes. To the cells was added either a 1:15 dilution of fusion protein, approximating to an activity of 100 U IL-2 ml-', an equivalent dilution of native SCA, rIL-2 at 100 U ml-' or a mixture of SCA and rIL-2. For competitive inhibition, cells were exposed to 5000 U ml-' of rIL-2 or pXho 1 primer IL-2n IL-2 gene I TNF at 10 pg ml-1 for 10 min prior to the addition of SCA-IL-2 as above. After incubation, cell associated SCA epitopes were detected by incubation with rabbit anti-SCA serum and then the bound rabbit anti-SCA antibodies were detected with FITC-conjugated anti-rabbit IgG (Sigma). After fixing with 1% paraformaldehyde, cell surface fluourescence was measured using a Becton Dickinson FACScan. Cells were washed five times between steps with RPMI to remove unbound material, and all incubations were for 30 min at 40C.
Bioactivity assays CTLL-2 cells (Gillis et al., 1978), which bear the high affinity IL-2 receptor, were maintained in RPMI media supplemented with 10% foetal calf serum (heat inactivated) and 10 U 1-rIL-2. For assay, cells were washed in media and deprived of IL-2 for 4 h, after which they were seeded into 96 well plates at 5 x 103 per well. Dilutions of fusion protein or rIL-2 were added and the cultures incubated for 18 h at 37°C in a 5% CO2 atmosphere. Then to each well 0.5 glCi of 3H-thymidine (Amersham) was added. After a further 4 h incubation cells were harvested onto glass fibre filters, dried and the incorporated radioactivity counted. For inhibition assays, fusion protein or rIL-2 at five times the concentration that produced 50% maximal stimulation of CTLL-2 cells was incubated with dilutions of goat anti-IL-2 antibody (30 min, 37°C) prior to addition to the CTLL-2 cells, for assay as described above.
Peripheral blood lymphocytes were obtained by venepuncture from healthy donors and prepared by differential centrifugation using Lymphoprep (Nycomed, Oslo, Norway). After washing in RPMI media and seeding into tissue culture plates at 105 cells per well, appropriate dilutions of fusion protein or rIL-2 were added. Following incubation (36 h, 37°C, 5% C02), 0.5 tCi of 3H-thymidine was added to each well, and after 4 h further incubation cells were harvested and incorporated radioactivity was measured.

Results
Expression and affinity chromatography of SCA-IL-2/7 protein SCA-IL-2/7 protein expression was induced by addition of 0.1 mM IPTG to transformed cultures of E.coli K12 KS476. Figure 2a shows a Western Blot of material affinity-purified from culture growth medium detected with the anti-IL-2 antibody DMS1; this reveals a single band with an apparent molecular weight of 46kDa. When probed with anti-SCA serum, a 46 kDa band was still apparent, but a number of degradation products were revealed ( Figure 2b). As DMS1 binds the carboxyl end of IL-2 and does not recognise any of the degradation products we assume that proteolysis has removed at least the terminal carboxyl section of the fusion protein in these degradation products. The ability of SCA-IL-2/7 fusion protein to bind lysozyme is demonstrated in Figure 3, where affinity purified material was allowed to bind immobilised lysozyme and was detected with polyclonal anti-IL-2. Furthermore, against the panel of immobilised protein antigens tested in Figure 4 there is no evidence of the non-specific stickiness sometimes associated with antibody fragments.
In Elisas ( Figure 5) in which SCA-IL-2/7 or rIL-2 are immobilised on the polyclonal anti-IL-2, the dose response curves generated with DMS1 are similar for both the fusion protein and rIL-2. An estimate of the serological activity of the IL-2 activity of the SCA-IL-2 sample used in these experiments is 1200-1500 U ml-'. SCA-IL-2/7 Culture supernatants 1:2 Figure 3 Detection of SCA-IL-2/7 protein by ELISA. Bacterial culture supematants of cells transformed with pSWsFVDI.3myc encoding anti-lysozyme single chain antibody (SCA), pSCA-IL-2/ 7 encoding Single chain antibody IL-2 fusion protein and pUCl9 were diluted 1:2 and applied to microtitre plates previously coated with hen egg lysozyme. Bound IL-2 epitopes were detected using anti-IL-2. 0.3-0.2-0.1 significant rise in fluorescence compared with the negative control, whilst free SCA either alone or with free rIL-2 give no increase on the background value. The specificity of the fusion protein interaction with the cells is demonstrated by the reduction of fluorescence almost to background levels in the presence of excess free rIL-2/ This competitive inhibition of SCA-IL-2 binding by free rIL-2 demonstrates that the fusion protein binds to the cells through specific receptorligand interactions. Competition with an excess of a nonspecific protein (TNF) had no effect on the level of SCA-IL-2 binding (Data not shown).
Results from sequential Elisas and western blots also indicate that the fusion protein is stable under the conditions of the biological assays described (Data not shown). Stimulation of CTLL-2 cells and peripheral blood lymphocytes (PBLs) by SCA-IL-2/7 protein In the conditions employed, half-maximal stimulation of IL-2 dependent CTLL-2 cells occurred at a concentration of approximately 0.4 U ml-' rIL-2 (Figure 7). SCA-IL-2/7 protein gave a similar maximal stimulation and dose response curve. The fusion protein sample achieved a similar half-maximal stimulation at a 5120 fold dilution, giving activity of the affinity purified material of approximately 1500 U ml-' rIL-2 against cells bearing the high affinity IL-2 receptor. This figure is in close agreement with the estimate derived from the serological assay. Goat anti-rIL-2 sera inhibited the proliferative effects of both rIL-2 and SCA-IL-2/7 protein in a similar manner, giving complete inhibition at 25 mg ml-' (Figure 8).
The effects of rIL-2 and SCA-IL-2/7 protein on peripheral blood lymphocytes, that bear the low and intermediate affinity forms of the IL-2 receptor, are shown in Figure 9. The dose response curve illustrates that SCA-IL-2 protein acted in a similar manner to IL-2 and the activity of the affinity purified material corresponds to approximately 1600 U ml-' rIL-2, again in close agreement with previous estimates. Native D1.3 SCA prepared by identical methods had no proliferative action in either assay (Data not shown).

Discussion
In this study we report the construction, expression and characterisation of a novel fusion protein, SCA-IL-2/7, that retains both the antigen binding characteristics of the parent single chain antibody and the immuno-stimulatory actions of IL-2. The genetic construct encodes the pelB leader sequence (Lei et al., 1987) that directs the expressed protein to the bacterial periplasm, where the oxidising environment should permit the formation of intramolecular disulphide bonds. Whilst the majority of material produced by the bacteria was degraded by proteases to give a product similar to the native SCA, a significant quantity of functional material was obtained from the culture supernatant and partially purified by affinity chromatography. The yield from the current system is very low, but it is apparent from experience with similar Fv fragments that yields can be optimised to produce a many fold improvement (Better et al., 1990).
We have demonstrated that the fusion protein retains antigen binding ability and that it is able to stimulate lymphoid cells bearing IL-2 receptors. When tested by serological and 30000- *1 biological assays the fusion protein and IL-2 gave similar dose response curves: incorporation of IL-2 into this fusion protein appears not to reduce its activity as it has in a previously described Fab'IL-2 fusion (Fell et al., 1991). It may be that the smaller SCA-IL-2 protein permits efficient receptor complex internalisation. This fusion protein interacts with and stimulates cells bearing either high or low affinity IL-2 receptors. The dual specificity may be allowed by the relatively long linker between the antibody domains and the IL-2 moiety allowing the amino-terminal end of IL-2 freedom to interact with the low affinity receptor. Work with diptheria toxin-IL-2 fusion proteins has demonstrated the importance of this mobility in allowing this interaction to occur effectively (Williams et al., 1987;Kiyokawa et al., 1991).
We have demonstrated that the SCA-IL-2 protein is stable under the conditions of the purification and assay and the FACS results demonstrate that it is intact when it interacts with the IL-2 receptor on the CTLL-2 cell's surface. In addition, work in progress demonstrates that the fusion protein, like IL-2, is able to cause an increase in vascular permeability (Savage et al., 1993). Although IL-2 induced permeability on a systemic scale can lead to toxicity (Lotze et al., 1986), as a localised event it can enhance the passage of itself and other therapeutic macromolecules into the tumour and ease access for effector cells (LeBerthon et al., 1991;Hennigan et al., 1991). Our aim is to target, via antibody variable domains, IL-2 activity to cells of the immune system in the area of the tumour. Accumulation of IL-2 around any inappropriately targeted normal cells should result in little cytolytic action, since effector cells do not interact significantly with normal cells. This approach should therefore much reduce the requirement for highly selective tumour associated antigens. This contrasts with the potentially detrimental effects of radionucleideor toxin-conjugated antibodies binding normal cells. Whilst the specificity of the current fusion protein is only suitable for preliminary in vitro experiments, recent advances in in vitro selection of antibody variable regions should allow the rapid generation of SCA directed against appropriate cellular targets.
Encouraging pre-clinical results with rIL-2 have only been partially supported by clinical experience. Poor clinical responses are in part owing to failure to achieve long lasting therapeutic concentrations in target tissues and also the systemic toxicity associated with large doses. Targeting of IL-2 by an antibody-derived fusion protein should allow concentration and prolonged action of rIL-2 within the area of the tumour whilst minimising systemic toxicity. Since the SCA-IL-2 described here may interact the low affinity IL-2 receptors on resting PBLs it is unlikely that it will localise effectively following intra-venous administration. However regional or direct intra-tumoural administration may result in accumulation, prolonged residence and an increased antitumour immune response in the tumour. For IV administration and localisation it will be preferable to have a form that interacts only with the high affinity IL-2 receptor, as expressed on NK cells and activated T-cells. Work is in progress to determine if shortening the linker in the SCA-IL-2 construct will produce a protein that will selectively stimulate cells bearing the high affinity receptor. SCA-IL-2 fusion proteins may provide an effective method of targeting therapeutic doses of rIL-2 to tumours or other targeted cells whilst significantly reducing systemic toxicity.