- Email: [email protected]
A single-chain bispecific Fv2 molecule produced in mammalian cells redirects lysis by activated CTL
Moleculur Iwin~nology, Vol. 33, No. 2. pp. 21 I-219, 1996 Published by Elsevier Science Ltd Printed in Great Britain
Pergamon 0161~5890(95)00087-9
A SINGLE-CHAIN MAMMALIAN CAROLINA Experimental
BISPECIFIC Fv, MOLECULE PRODUCED CELLS REDIRECTS LYSIS BY ACTIVATED CTL R. JOST, JULIE DAVID
Immunology
A. TITUS, ISTVAN M. SEGAL*
Branch, National
(First received 3 Murch
Cancer Institute,
KURUCZ
IN
and
Bethesda, MD 20892, U.S.A.
1995; accepted in recised,form
15 MUJ 1995)
Abstract-Single-chain Fv (sFv) molecules consist of the two variable domains of an antibody (Ab) connected by a polypeptide spacer and contain the binding activities of their parental antibodies (Abs). In this paper we have attached the C-terminus of 2Cl I-sFv (anti-mouse CD3 c-chain) to the N-terminus of OKT9-sFv (anti-human transferrin receptor [TfR]) through a 23 amino acid inter-sFv linker consisting primarily of C,, region residues from 2Cl1, to form a single-chain bispecific Fv, [bs(sFv)J molecule. The bs(sFv), was expressed in COS-7 cells, and was secreted at the same rate as the two parental sFvs. The secreted protein had both anti-CD3 and anti-TfR binding activities. Essentially all of the secreted bs(sFv)z molecules bound TfR and the binding affinity of the bs(sFv), was comparable to that of OKT9 sFv and Fab. Thus, the attachment of the inter-sFv linker to the N-terminus of OKT9-sFv did not impair its binding function. The bs(sFv), retained both binding specificities after long-term storage at 4 C or overnight incubation at 37’C. It redirected activated mouse CTL to specifically lyse human TfR+ target cells at low (rig/ml) concentrations and was much more active than a chemically cross-linked heteroconjugate prepared from the same parental mAbs. Because bs(sFv), molecules secreted by mammalian cells are homogeneous proteins containing two binding sites in a single polypeptide chain, they hold great promise as an easily obtainable, economic source of a bispecific molecule suitable for in cico use. Kql words: Bispecific antibody,
single-chain
antibody,
INTRODUCTION Bispecific antibodies (bsAbs) are proteins that have two different binding specificities within the same molecule. They can be used in any application requiring the crosslinking of two different molecules, an important one being the redirection of lysis by CTL (Perez et al., 1985; Staerz et al., 1985; Segal and Snider, 1989). By linking CD3/TCR complexes on CTL with cell surface components on target cells, bsAbs induce CTL to lyse target cells regardless of their initial specificity. In particular, bsAbs with specificity for the TCR and a tumor Ag redirect CTL to specifically lyse tumor cells. In vivo treatment of tumor-bearing mice with such bsAbs results in the eradication of the tumor or slows its growth (Garrido et al., 1990; Mezzanzanica et al., 1991; Van Dijk et al., 1991; Weiner er a/., 1994; Weiner and Hillstrom, 1991;
*Author to whom correspondence should be adressed at: Building 10, Room 4Bl7, National Institutes of Health, Bethesda, MD 20892, U.S.A. Abbreviations: sFv, single-chain Fv, TfR, transferrin receptor, bs(sFv),, single-chain bispecific Fv,, bsAb, bispecific antibody, Fc,.R, Fc receptor for IgG, mAb, monoclonal antibody.
redirected
Demanet et al., 1991, 1994; Brissinck et al., 1991; Renner et al., 1994) and preliminary clinical studies using bsAbs in cancer patients have also been promising (Nitta et al., 1990; Bolhuis
et al., 1992; Kroesen
et al., 1993).
If bsAbs are to be used in uivo, they must be pure, homogeneous proteins, preferably lacking Fc portions which could, by interacting with FcyR on various cell types, induce premature triggering, lysis, or sequestration of effector cells. In the past, homogeneous bispecific F(ab’), fragments have been produced by digestion of hybrid hybridoma antibodies (Staerz and Bevan, 1986; Lanzavecchia and Scheidegger, 1987; Segal et al., 1992, 1993) or by crosslinking two different F(ab’)s through their hinge cysteines (Glennie et al., 1987; Segal et al., 1993). More recently, several types of genetically engineered bsAbs have been designed to facilitate their production (Holliger and Winter, 1993). For example, Fab’ molecules produced in bacteria (Shalaby et al., 1992; Carter et al., 1992) and in myeloma cells (Kostelny et al., 1992) were directly crosslinked by their hinge sulfhydryls, thus eliminating digestion steps. Genetic engineering has also led to the production of bsAbs consisting primarily of the variable domains of antibodies. These constructs are smaller than those containing whole F(ab’) fragments and would
211
lysis.
therefore
be expected
to have improved
tumor
212
C. R. JOST et al.
penetration. One construct uses two different polypeptide chains, Vu,-V,, and VL,-VH2, that associate laterally and non-covalently to form a bsAb known as a “diabody” (Holliger et al., 1993). Other constructs are composed of single-chain Fvs (sFvs), molecules in which V, and V, from the same mAb are connected by a polypeptide linker, and which have the Ag binding properties of the parental mAb (Huston et al., 1988, 1993; Bird et al., 1988). Single-chain bsAbs have been produced by joining two different sFvs with a polypeptide linker. Mallender and VOSS(1994) first described a single-chain bispecific sFv [bs(sFv),] having specificity for two different haptens, and Gruber et al. (1994) subsequently showed that a bs(sFv), having specificity for a TCR idiotope and for a hapten could redirect lysis against a haptenmodified cell. Both groups produced their bs(sFv),s in bacteria, and the proteins required refolding and affinity purification before they could be used. In another study, Hayden et al. (1994) produced a bs(sFv)> in a COS cell transient expression system as a fusion protein that contained the Fc region of human IgGl. The COS cells secreted active bsAb that redirected lysis but was heterogeneous in size due to aggregation of the Fc portions. In order to study the properties of a single, undimerized bs(sFv),, we have produced a construct consisting of 2Cl I-sFv (anti-murine CD3 t-chain) linked to OKT9sFv (anti-human transferrin receptor [TfR]). This was expressed in COS-7 cells, thus taking advantage of endoplasmic reticulum-based protein folding machinery and quality control mechanisms that allow only fully folded proteins to be secreted. We show that the bs(sFv), is produced and secreted as rapidly as each of the parental sFvs, that most of the bs(sFv), is active, and that the secreted material specifically redirects lysis mediated by activated mouse CTL at rig/ml levels.
MATERIALS
AND METHODS
Cell lines and antibodies
The following cell lines and Abs were used: COS-7 monkey kidney fibroblasts, K562 human erythroleukemia cells (both from American Type Culture Collection, Rockville, MD), 2B4 murine T hybridoma cells (Hedrick et al., 1982), HUT 102 human T lymphoma cells (Gootenberg et al., 1981), 9ElO mAb against the cmyc peptide (Evan et al., 1985), 2.4G2 mAb against the mouse FcyRII and III (Unkeless, 1979), 145-2C11 mAb (termed 2Cll) against mouse CD3 E chain (Leo et al., 1987), OKT9 mAb against human transferrin receptor (TfR) (Goding and Burns, 1981; Schneider et al., 1982) W6/32 mAb against human MHC class I molecules (Barnstable et al., 1978), and 36-7-5 mAb against H-2 Kk (Sachs et al., 1981). Fab fragments of W6/32 and 36-75 were prepared as described (Perez et al., 1986). Fab fragments from 2C11 and OKT9 were prepared using an ImmunoPure Fab Preparation Kit (Pierce, Rockford, IL) according to the manufacturers’ instructions. OKT9 Fab was labelled with FITC as described (Segal et al., 1987). Polyclonal rabbit anti-mouse IgG was from Cappel
(Organon Technika, Durham, NC). Heteroconjugates were prepared by crosslinking 2Cll mAb to OKT9 mAb using succinimidyl-3-(2-pyridyidithiol)-propionateas described (Segal, 1993). sFv and bs(sFu), constructs
The 2Cl I-sFv and OKT9-sFv (Nicholls et al., 1993) containing plasmids (pcDNA/AMP-2Cll and pSVLOKT9) have been described (Jost et al., 1994). Both sFvs have a VL-linker-V, design, a 5’ light-chain leader sequence and a 3’ c-myc peptide tag. The bs(sFv), was made by inserting a modified version of the OKT9-sFv into pcDNA/AMP-2Cll at the 3’ end of the Vu domain of 2Cl I-sFv. To this end we introduced a unique EcoR I site in pcDNA/AMP2Cll between Vu and the c-nzyc peptide using a Transformer Mutagenesis Kit (Clonetech Laboratories Inc, Palo Alto, CA). The mutating and selecting primers were, CC GTC TCC TCA TCA GAA TTC AAA CTC ATC TCA GAA GAG G and GGG AGA CCC ACG CGT GGT ACC GAG CTC GG, respectively. The OKT9-sFv insert was generated by PCR using pSVLOKT9 as a template and a 5’-primer, ATATA AAG CTT GAA TTC GCC AAA ACA ACA GCC CCA TCG GTC TAT CCA CTG GCC CCT GTG CTC GAG TCT AGT GGT AGC GGA GAC ATC AAG ATG ACC CAG TC and a 3’-primer TATAT TCT AGA CTC TTA TTA ATT CAG ATC CTC. The PCR product consisted of an EcoR I site followed sequentially by the inter-sFv linker, OKT9-sFv, the c-myc peptide tag, stop codons and a Xba I site. The PCR product was directly ligated into pCR 11 (Invitrogen, San Diego, CA), which was then digested with EcoR I and Xba I. The resulting fragment was ligated into the pcDNA/AMP-2Cll (MUT) that had been digested with the same restriction enzymes yielding the bs(sFv), construct shown in Fig. 1. Transfection and metabolic labeling
Procedures used for transfection and metabolic labeling have been previously described (Jost et al., 1994). In short, COS-7 cells were cultured in 1O-cm culture dishes in DMEM medium (Dulbecco’s modified Eagle’s medium supplemented with penicillin (100 U/ml), streptomycin (100 U/ml), L-glutamine (2 mM) and 10% fetal calf serum) and transfected with 20 pg of plasmid DNA using the calcium phosphate precipitation method (Davis et al., 1986). When the transfected cells were used to produce non-radioactive bs(sFv),, the cells were allowed to grow for an additional 48 hr, after which the supernatant was harvested, diluted and used in cytotoxicity experiments. To produce “S-methionine labeled bs(sFv), [j5Sbs(sFv),], cells were trypsinized 16 hr after transfection, replated in three IO-cm culture dishes, and allowed to grow for an additional 24 hr prior to labeling. Cells (90% confluent) were pulsed for 1.5 hr at 37°C with ‘5Smethionine (Transj5S-label [ICN Radiochemicals, Irvine, CA]) at a concentration of 0.15 mCi/ml in a total volume of 4 ml and chased for 2.5 hr in DMEM. The radioactive supernatants from identical transformations were pooled
Recombinant
single-chain
213
bispecific Fv, redirects lysis by CTL OKT9 SFV
2C11 SFV N
C MYC PEPTIDE
VL
inter-&G
ala lys thr thr ala pro ser val tyr pro leu ala pro vai
2C1 1 VH - glu phe
I
LINKER
I
2Cll
ser ser gly ser giy - OKT9
leu glu
I
CHl
I
VL
flexible
Fig. 1. Diagram of the 2Cll-OKT9 bs(sFv), protein. The bs(sFv), protein has a 2Cll-sFv-linkerOKT9-sFv design. The asterisk indicates the Ag binding sites of the bs(sFv),. N and C refer to the N- and C-termini of the molecule. The glu-phe and leu-glu residues in the linker result from EcoR I and Xba I sites, respectively.
and supplemented with 10 mM HEPES and 0.01% sodium azide prior to use. In the experiment shown in Fig. 2, cells were pulsed for 1 hr and chased for varying times in DMEM containing 15 mg/ml of L-methionine.
Immunoprecipitation
Immunoprecipitation and the analysis of the precipitated proteins was performed at 4°C as described (Jost
OKTS-sFv -4)---Cells
__)
%.
P
2Cll -sFv
4-Supematant
--b
d----Cells
__)
.x
2h
r)-
Supematant
--W
-
6h
P
2h
6h
2h
P
6h
P
2h
Bispecific-sFv2 +Cells
69kD-
---W
M-Supematant
+
-Qmlb :
P
..
2h
6h
P
2h
6h
Fig. 2. Production and secretion of sFvs by COS-7 cells. COS-7 cells transfected with plasmids encoding the indicated protein were pulsed for 1 hr with “S-methionine and chased for 0 (“P” in Fig.), 2 or 6 hr, as indicated. sFvs were immunoprecipitated from either cell lysates (“Cells”) or from culture media (“Supernatant”), subjected to SDS-PAGE under reducing conditions, and visualized by autoradiography. The OKT9-sFv and the bs(sFv), contain an N-linked carbohydrate group that contributes to their observed molecular weights.
6h
214
C. R. JOST et al.
et al., 1994). Briefly, cells were lysed and proteins were precipitated using 9ElO (anti-c-m-rc peptide) and protein A-Sepharose precoated with rabbit anti-mouse Ig. In some experiments, protein was eluted from the beads by incubation for 5 min at 95°C in elution buffer (20 mM Tris, I mM EDTA, 2% SDS 0.1% Bromphenol Blue and 5% 2-ME pH 7.8) and eluates were analysed by reducing 12.5% SDS-PAGE and autoradiography. In other experiments, beads were divided into four equal portions and treated as above with elution buffer (without Bromphenol Blue). Eluates were transferred directly to 2.5 ml of Ecolume (ICN, Cleveland, OH) scintillation solution and counted.
,ug OKT9 mAb, incubated for various times at 4 C and spun for 10 sec. Cell pellets were lysed, bs(sFv)z and sFv were immunoprecipitated, and the immunoprecipitates analysed by scintillation counting as described above. The amount of sFv and bs(sFv), that bound non-specifically to K562 cells was determined from the radioactivity bound in the presence of excess OKT9 mAb. Quadruplicate samples were averaged, and the non-specifically associated radioactivity (less than 10% of that bound to K562 cells at 0 time) was subtracted. In both radioactive and fluorescence measurements, background-corrected data were divided by background corrected 0 time values and multiplied by 100 (per cent of maximal bound).
bs(sFv), binding
Cytotoxicit),
K562 cells (9 x 10”) and 2B4 cells (2 x 10’) were preincubated at 4°C with or without OKT9 Fab (30 pg) or 2C 11 Fab (50 pg), respectively, in 200 ~1 wash medium (HBBS, 0.1% BSA, 0.01% sodium azide). After 30 min, 2 ml of ‘5S-bs(sFv),-containing medium was added and cells were incubated for 2 hr at 4°C. Cell-associated bs(sFvs)s was immunoprecipitated and analysed by SDS PAGE as described above. In other studies, cells were incubated with varying amounts of parental or control Fab before addition of 1 ml “S-bs(sFv)+,. Immunoprecipitated bs(sFv)z was quantified by scintillation counting as described above. Non-specific binding [the amount of radioactivity bound in the presence of excess (50 pg) parental Ab] was subtracted and % inhibition was calculated using the formula:
Cytotoxicity was studied by using a standard “Crrelease assay (Wunderlich and Shearer, 1991). “Cr-labeled HUT 102 cells were used as target cells. CTL effector cells were generated in 5-6 day primary allogeneic MLR cultures by incubating splenic effector cells with irradiated allogeneic stimulator cells. Strain combinations used to generate CTLs were Balb/c antiC57Bl/lO, C57Bl/lO anti-Balb/c and C3H/Hen antiBalb/c. A varying amount of effector cells was added to each well of a 96-well microtiter plate. Varying dilutions of COS-7 supernatant containing either bs(sFv)z, OKT9sFv, 2Cl I-sFv. medium alone or medium containing graded concentration of the 2Cl l-OKT9 heteroconjugate were then added followed by either 5 x 10’ or 1 x lo4 target cells. When using the 2Cll x OKT9 heteroconjugate, 2 pg of 2.4G2 anti-FcyR mAb was added to each well to prevent interactions with Fc receptors. In inhibition experiments, effecters and targets were added directly to the wells (no wash step) after a brief preincubation with inhibiting or control antibodies. Percentage inhibition was calculated as described above for binding but in this case I and T refer to the per cent lysis in the presence (I) or absence (T) of inhibitor.
% Inhibition
= lOO(1 - Z/T),
where Z and T equal the number of counts bound specifically in the presence and absence of inhibitor, respectively. Dissociation
qf OKT9
Fab
K562 cells (10’) were incubated for 1 hr at 4°C with 2 pg of FITC-OKT9 Fab, with (background) or without (samples) 50 pg OKT9 mAb in a total volume of 1 ml of wash medium. Cells were centrifuged for 5 min at 2OOg, resuspended in 1 ml wash medium containing 50 rug OKT9 mAb, and incubated at 4’C. At various times, samples were analysed by flow cytometry. Mean fluorescence values were determined and background values (12% of initial value, 22% of final value) were subtracted at each time point. Zero time was taken as the first point measured.
RESULTS sFv and bs(sFv)2 constructs
The bs(sFv), and its parental sFvs were produced in COS-7 cells transfected with vectors encoding the following molecules: 2Cl I-sFv, against the murine CD3-c chain; OKT9-sFv, against the human TfR; and the bs(sFv), against both Ags. All constructs contained light chain leaders to direct newly-synthesized proteins to the endoplasmic reticulum and a C-terminal c-m~‘c peptide to facilitate detection. Both parental sFvs contained an Dissociation qf OKT9 sFv and the bs(sFv), intra-sFv (Gly,-Ser), linker connecting the C-terminal V, with the N-terminal V,,. The bs(sFv), is shown diaMultiple samples of K562 cells (6 x lo6 per sample) in 1 ml of COS-7 medium containing “5S-bs(sFv), or 35S- grammatically in Fig. 1. An inter-sFv linker connects the OKT9-sFv were incubated 1 hr at 4°C in the presence or C-terminus of 2Cll -sFv to the N-terminus of the OKT9sFv. This inter-sFv linker consists of two residues arising absence of excess OKT9 mAb. Cells were centrifuged for 5 min at 2OOg, and one sample (0 time point) was from an EcoR I site, the first 13 amino acids of the 2Cll resuspended in cold wash medium and immediately spun C,l domain followed by two residues from a Xho I site for 10 set in an Eppendorf microcentrifuge. Other sam- and a flexible hydrophilic region. The bs(sFv),-encoding a vector that ples were suspended in 1 ml wash medium containing 50 construct was cloned into pcDNA/AMP,
21.5
Recombinant single-chain bispecific FvZ redirects lysis by CTL
(a)
expresses large amounts of protein in COS-7 cells under control of the CMV promoter.
2B4 cells
K562 cells
Production and secretion oj’sFv molecules Transfected COS-7 cells were labeled with “S-methionine for 60 min and then chased for either 0, 2 or 6 hr. The sFvs and bs(sFv), were immunoprecipitated from both the cell lysates and the supernatants and analysed by SDS-PAGE. Figure 2 shows that the bs(sFv)z and both parental sFvs were secreted from the cells at similar rates; by 6 hr, most of the radiolabeled bs(sFv), and sFvs had left the cells and were found in the supernatant. Thus, the increased size and complexity of the bs(sFv), resulting from the concatenation of two sFv molecules does not affect their ability to be processed and secreted by COS7 cells.
inhibitor
+
-
+ 2Cll
OKT9 (b) 2C11 Fab
36-7-5
Fab
0.2 E 0.6 2 6 20 _
I+ b
20 >
Secreted bs(sFv)> spec$cally binds both Ags To determine if secreted sFvs were active, we tested the ability of radiolabeled material from the medium of transfected COS-7 cells to specifically bind Ag. Figure 3a shows that the bs(sFv), bound to 2B4 (CD3+) and to K562 (TfR+) cells and that binding was inhibited by the parental antibodies. These antibodies inhibited %bs(sFv), binding in a dose-dependent fashion (b) but inhibition did not occur when antibodies binding to MHC class I molecules on these cells were used as controls. Thus, the bs(sFv), specifically binds both Ags. To determine the fraction of bs(sFv), that had binding activity, we depleted 35S-bs(sFv), from the COS-7 medium by four sequential incubations with either 2B4 or OKT9 cells, and then immunoprecipitated the remaining bs(sFv), using the 9ElO mAb which recognizes the c-my peptide tag. In Fig. 4 the top panel shows that most of the bs(sFv), was removed by absorption to K562 cells and very little could subsequently be immunoprecipitated with the 9ElO mAb, indicating that a large fraction of the bs(sFv), present in the supernatant had the TfR binding activity of OKT9. The lower panel shows that a small and nearly equal amount of bs(sFv)? was bound to 2B4 cells at each absorption step but that a large amount of bs(sFv), remained in the medium even after four absorptions. Because the amount of bs(sFv), removed at each absorption step was a small fraction of the total, we were unable to determine the fraction of bs(sFv)2 having antiCD3 binding activity. The low binding of the bs(sFv), to 2B4 cells was not due to the 2Cll portion being in the bispecific configuration, since similar results were obtained with the parental 2Cll -sFv (Jost et al., 1994).
OKT9Fab
W6/32
‘,l.k-
Fab
,
10 h 0
f 20
40 Percent
60 inhibition
60
100
of binding
Fig. 3. Secreted bs(sFv), specifically binds its Ags.(a) K562 (TfR+) and 2B4 (CD3+) cells were incubated for 2 hr at 4°C with medium containing “S-bs(sFvs), in the presence or absence of the indicated parental Ab. Bound bs(sFv)2 was immunoprecipitated and analysed by SDS-PAGE under reducing conditions. (b) Dose dependence of the inhibition was studied by incubating K562 cells (shaded bars) and 2B4 cells (open bars) with 35S-bs(sFv)1 in the presence of the indicated concentrations (in pg/ml) of parental Fab or with a control Fab (anti-Kk Fab for 2B4 cells and anti-HLA I Fab for K562 cells). After the incubation, bound bs(sFv), was immunoprecipitated, counted, and the percentage inhibition was calculated as described in Materials and Methods. The asterisk indicates no inhibition.
that the bs(sFv), dissociates from K562 cells at a rate intermediate to those of the sFv and Fab, but that there is not a great difference in dissociation rates of the three different molecules. Dissociation rates are directly related to binding affinity (Berzofsky et al., 1993) implying that the affinities of free OKT9-sFv and the OKT9-sFv in the bs(sFv), are similar. Thus, the presence of the N-terminal inter-sFv linker has little detrimental effect on the binding affinity of the bs(sFv)2 for the TfR. The bs(sFv)2 is stable at 37’ C
The bs(sFv), binds TfR with a similar af$nity to OKT9 sFr and Fab The inter-sFv linker attaches to the N-terminus of the OKT9-sFv and might therefore interfere with Ag binding (Fig. 1). To determine whether the presence of this linker had an effect on the binding of the bs(sFv), to the TfR, we compared the dissociation rate of the bs(sFv), with those of the OKT9-sFv and OKT9 Fab. Fig. 5 shows
It has been reported that incubation in serum at 37°C can lead to rapid inactivation of some sFvs (Glockshuber et al., 1990; Reiter et al., 1994). However, immunoprecipitation and sequential absorption experiments showed that most of the bs(sFv), remained in solution and still had binding activity after incubation for 16 hr at 37°C in medium containing 10% serum (data not shown).
216
C. R. JOST et al. T
1
2
4
3
R
K562
Sequential Precipitations T
2
1
___) 4
3
R
Sequential Precipitations ---W Fig. 4. Depletion of bs(sFv)2from COS-7 supernatant. Medium (4 ml) containing 95S-bs(sFv),was sequentially absorbed with either K562 cells (TfR+, 2 x IO’ cells/absorption) or 2B4 cells (CD3+, 6 x lo7 cells/absorption). T indicates the total amount of bs(sFv), present in 4 ml culture medium before absorptions, numbers above lanes indicate the absorption step (numbered lanes contain cell-bound bs(sFv),), and R indicates bs(sFv)z remaining in the medium after the final absorption. Shown are the SDS-PAGE analyses of immunoprecipitates.
allogeneic MLR were used as effector cells. Figure 6a shows that the bs(sFv), efficiently redirected lysis at E : T ratios of 5 : 1 and 10 : 1, while supernatants containing the parental sFvs did not. Figure 6b shows that the bs(sFv), mediated lysis in a dose-dependent fashion, and still gave significant lysis when diluted 128-fold. Lysis mediated by a chemically crosslinked heteroconjugate between 2Cll and OKT9 was much lower than that mediated by the bs(sFv),. Heteroconjugates made in the same way from other antibodies have been more potent at redirecting lysis (Perez et al., 1986), and it may be that either 2Cll or OKT9 is especially susceptible to inactivation by the crosslinking reagent. Both parental Fabs inhibited lysis, while irrelevant antibodies that bound to either effector or target cells did not (Fig. 7). Thus, the simultaneous binding of the bs(sFv), to both effector and target cells was required for lysis.
2Cll-sFv 30
20
10 , :gtgy 2
‘.
-.
- - - - - - - __
4
6
6
10
12
14
16
Bs(sFv),
Effector
OKTS-sFv OKT9
Fab
60
20
to target
ratio
4o (W 30
i
04 0
20
40
Time
60
I 80
--o--
2CllxOKT9
d-
bs(sfv),
---m-
OKT9-sFv
-+-
2Cll-SFV
:_I_
(Min)
Fig. 5. Dissociation of OKT9-sFvs, bs(sFv)z and OKT9 Fab from K562 cells. 35S-bs(sFv)2(H), OKT9-sFv (0) and FITC labeled OKT9 Fab (V) were bound to K562 cells. Cells were washed and resuspended in excess unlabeled OKT9 mAb. Radioactive protein bound at the indicated times was precipitated and measured by scintillation counting. Bound FITC labelled Fab was analysed by FACS.
Redirected cytotoxicity Because the bs(sFv), has dual specificity for murine CD3 and human TfR we asked whether it could redirect mouse T cells to lyse the TfR+ HUT-102 cells. We used COS-7 cell supernatants that had been harvested 48 hr after transfection with vectors encoding either the bs(sFv), or parental sFvs. CTLs generated in a primary
0
1
2
Log,
3
4
Dilution
Fig. 6. Targeting of lysis using bs(sFv),. Diluted COS-7 supernatants containing bs(sFv), (V), OKT9-sFv (m), 2Cl I-sFv (A), chemically crosslinked 2C I 1 x OKT9 heteroconjugate (0) or medium alone (-•-) were used in a standard 3.5 hr “Cr releaseassaywith HUT-102 cellsas targets and activated mouse spleen cells as effecters. (a) Lysis by titrated doses of Balb/c allo-stimulated effector cells. COS-7 cell supernatants were diluted 1 : 16 and the heteroconjugate was used at 0.38 pg/ml. (b) Two-fold serial dilutions of the transfected COS-7 supernatants with a 1 : 16 dilution as starting point, and 2Cll x OKT9 heteroconjugate (0.5 pg/ml starting point) were used to redirect lysis mediated by C57B1/6 allo-stimulated effector cells at an E:Tratio of 10: I.
Recombinant single-chain bispecific Fvz redirects lysis by CTL
217
fore, the rate at which proteins are secreted is an indication of the rate at which they are folded and assembled. We had previously found that COS-7 cells fold and secrete genetically engineered sFvs at different rates depending on the sFv in question (Jost et al., 1994). We were therefore surprised to find that the bs(sFv), is secreted at the same rate as its constitutive sFvs since the bispecific construct is a more complex molecule conOKT9 Fab 1 taining four disulfide bonds and two sets of VL-V, domain pairs. Apparently, the additional complexity of the bs(sFv), poses no problem for the protein folding W6/32 Fab IO machinery in COS-7 cells. 100 20 40 60 80 0 We were concerned that the presence of two different inhibition of lysis sFvs within one molecule could lead to mispairing of V Fig. 7. The specificity of redirected lysis. The specificity of lysis domains as has been reported in the production of bispecof HUT 102 cellswas studied in the presenceor absenceof the ific antibodies from hybrid hybridomas (De Lau et al., indicated amounts (in pg/ml) of parental or irrelevant Fabs. Open bars, C3H allo-stimulated effecters were preincubated 1991). However, we found, by absorption, that almost with 2Cll or 36-7-5 (anti-Kk) Fab and used at an E : T ratio of all bs(sFv), had OKT9 binding activity suggesting that 8.5 : 1 in the presence of a 1 : 8 dilution of bs(sFv),-containing the variable domains are correctly matched in most culture medium. Shaded bars, HUT 102 target cells were pre- secreted molecules. Absorption experiments could not be incubated with OKT9 or W6/32 (anti-HLA 1) Fab. Balb/c allo- used to test whether all bs(sFv), molecules have anti-CD3 stimulated effecters were used at an E : T ratio of 5 : 1 in the activity because too few bs(sFv)? molecules bound to 2B4 presence of a 1 : 8 dilution COS-7 supernatant containing cells at each absorption step. This could be due to the bs(sFv),. The asterisk indicates no inhibition. low number of CD3 molecules on the 2B4 cells and/or to the low affinity of 2Cll for its antigen (K, of 2Cll Fab for CD3 is approximately 0.7 x lo-’ MP’, and 2B4 cells We tried to measure the amount of bs(sFv)z used in express about 2 x IO4 CD3 molecules/cell, T. Bakacs, the “Cr-release assay by inhibiting the binding of 35S- unpublished result). Thus, although we cannot formally prove that all bs(sFv), molecules have CD3 binding bs(sFv), with the medium used to redirect lysis [which contained cold bs(sFv),]. No inhibition was observed activity, we have no reason to believe that this is not the even when the medium was used undiluted. As seen in case. Fig. 3B, OKT9 Fab inhibited the binding at 0.3 pg/ml Inter-sFv linkers that have been used in bs(sFv), conbut not at 0.1 pg/ml. Therefore, assuming that the structs include: a 24-amino acid flexible linker modelled bs(sFv), and the OKT9-Fab bind with similar affinities, after a secreted fungal cellulase domain (Mallender and there is probably less than 0.3 pg/ml of cold bs(sFv), in Voss, 1994); a 25-amino acid linker rich in ala, asp and the culture supernatant. On the other hand, the same lys residues (Gruber et al., 1994); a 27-amino acid helical OKTbs(sFv), produced in bacteria can redirect lysis at peptide linker containing a recurring EEAKK motif concentrations as low as 0.2 rig/ml (Kurucz et al., 1995). (Hayden et al., 1994); and the 23-residue linker we used Therefore, the concentration of bs(sFv), secreted by the consisting of the first 14 amino acids of the 2Cll C,, transfected COS-7 cells is likely to be in the range of 0.3domain followed by a five-amino acid flexible hydrophilic 0.02 pg/ml. sequence. All of these linkers resulted in active bispecific molecules, suggesting that the compositions of the intersFv linkers are relatively unimportant. Nevertheless one DISCUSSION would expect that a minimal linker length might be In this paper we describe a genetically engineered required to prevent the V domains of one sFv from intersingle-chain bsAb consisting of two sFvs linked together by acting with those of the other sFv. We were curious a polypeptide spacer, one sFv reactive against the human whether the presence of the inter-sFv linker, which TfR and the other against murine CD3. When transfected attaches to the N-terminus of the OKT9-sFv, would into COS-7 cells, the bs(sFv), is secreted at the same rate decrease the affinity of the OKT9 portion of the bispecific as the parental sFvs and has a molecular weight of 69 molecule for Ag. We found, however, that the affinity of kDa, as expected for a glycosylated bs(sFv),. The secreted the bs(sFv&, as measured by its rate of dissociation from bs(sFv), recognizes both Ags and specifically redirects TfR+ cells, is similar to that of the OKT9-sFv and Fab. mouse CTL to kill human TfR+ target cells. The bs(sFv), In addition, we have subsequently produced this protein protein is stable for at least 16 hr at 37”C, and was still in bacteria in large amounts and have found its affinity active after several months of storage at 4°C. for CD3 to be indistinguishable from that of the parental It has been established that mammalian cells fold pro2Cll Fab (Kurucz et al., 1995). This suggests that at least teins in the endoplasmic reticulum and that a quality with this bsAb, the inter-sFv linker did not interfere with control mechanism is in place to prevent incorrectly Ag binding. folded or unassembled proteins from leaving the endoThe 2C11 x OKT9 bs(sFv), present in culture superplasmic reticulum (Gething and Sambrook, 1992). Therenatants was still able to redirect lysis at high dilutions 2Cll
Fab
0.1
C. R. JOST et al.
218
after at least 3 months storage at 4 ‘C, and the bs(sFv), and its parental sFvs were stable to an overnight incubation in culture medium at 37’C. It has been reported (Glockshuber et al., 1990; Reiter et al., 1994) and we also have found (unpublished observation) that some sFvs are unstable at 37°C. Thus, it is to be expected that not all bs(sFv),s will be as stable as the 2Cll x OKT9 bs(sFv),. Since instability would present problems both when producing the proteins in mammalian cells and when using them in biological systems, sFvs should first be screened for stability before using them in bs(sFv), constructs. Because the bs(sFv), was active at low protein concentrations and because mammalian cells secrete primarily active proteins, the retargeting activity of the bs(sFv), could be studied using supernatants directly. Therefore, the main advantage of the COS-7 cell secretion system is that it allows rapid testing of newly produced bs(sFv), constructs. Once a bs(sFv), construct has proven to be stable, to recognize both antigens, and to redirect lysis, it could be produced in large amounts either in bacteria or in stably transfected mammalian cells. The next step would be to determine the fate and efficacy of the bs(sFv)z in vim. Acknowledgements-The authors thank Apurva Sarin, Mark Williams and Belen Moreno for providing allo-stimulated effector cells, David Winkler for producing the OKT9 mAb, also Tibor Bakacs and Giuseppe Sconocchia for thoughtful suggestions during the preparation of the manuscript.
REFERENCES Barnstable C. J., Bodmer W. F., Brown G., Galfre G., Milstein C., Williams A. F. and Ziegler A. (1978) Production of monoclonal antibodies to group A erythrocytes, HLA and other human cell surface antigens-new tools for genetic analysis. Cell14,
9-20.
Berzofsky J. A., Berkower I. J. and Epstein S. L. (1993) Antigen-antibody interactions and monoclonal antibodies. In Fundumental Immunology (Edited by Paul W. E.), p. 421. Raven Press Ltd, New York. Bird R. E., Hardman K. D., Jacobson J. W., Johnson S., Kaufman B. M., Lee S. M., Lee T., Pope S. H., Riordan G. S. and Whitlow M. (1988) Single-chain antigen-binding proteins. Science 242,423-426. Bolhuis R. L., Lamers C. H., Goey S. H., Eggermont A. M., Trimbos J. B., Stoter G., Lanzavecchia A., Di Re E., Miotti S., Raspagliesi F., Rivoltini L. and Colnaghi M. I. (1992) Adoptive immunotherapy of ovarian carcinoma with bsMAb-targeted lymphocytes: a multicenter study. ht. J. Cancer 7(suppl.), 78-81. Brissinck J., Demanet C., Moser M., Leo 0. and Thielemans K. (1991) Treatment of mice bearing BCLl lymphoma with bispecific antibodies. J. Immun. 147,4019-4026. Carter P., Kelley R. F., Rodrigues M. L., Snedecor B., Covarrubias M., Velligan M. D., Wong W. L. T., Rowland A. M., Kotts C. E., Carver M. E., Yang M., Bourell J. H., Shepard H. M. and Henner D. (1992) High level Escherichia coli expression and production of a bivalent humanized antibody fragment. Bio/Technology 10, 163-167. Davis L. G., Dibner M. D. and Battey J. F. (1986) Basic Methods in Molecular Biology. Elsevier, New York. De Lau W. B. M., Heije K., Neeges J. J., Oosterwegel M.,
Rozemuller E. and Bast B. J. E. G. (1991) Absence of preferential homologous H/L chain association in hybrid hybridomas. J. Immun. 146,90&914. Demanet C., Brissinck J., Van Mechelen M., Leo 0. and Thielemans K. (1991) Treatment of murine B cell lymphoma with bispecific monoclonal antibodies (anti-idiotype x anti-CD3). J. Immun. 147, 1091~1097. Demanet C., Brissinck J., Leo O., Moser M. and Thielemans K. (1994) Role of T-cell subsets in the bispecific antibody (anti-idiotype x anti-CD3) treatment of the BCL, lymphoma. Cancer Res. 54,2973-2978. Evan G. I., Lewis G. K., Ramsay G. and Bishop J. M. (1985) Isolation of monoclonal antibodies specific for human c-myc proto-oncogene product. Molec. Cell Biol. 5, 361&3616. Garrido M. A., Valdayo M. J., Winkler D. F., Titus J. A., Hecht T. T.. Perez P., Segal D. M. and Wunderlich J. R. (1990) Targeting human T lymphocytes with bispecific antibodies to react against human ovarian carcinoma cells in nu/nu mice. Cancer Res. 50,42274232. Gething M. J. and Sambrook J. (1992) Protein folding in the cell. Nature 355, 3345. Glennie M. J., McBride H. M., Worth A. T. and Stevenson G. T. (1987) Preparation and performance of bispecific F(ab)‘? antibody containing thioether-linked Fab’z fragments. J. Immun. 139,2367-2375. Glockshuber R., Malia M., Pfitzinger I. and Pluckthun A. (1990) A comparison of strategies to stabilize immunoglobulin Fv-fragments. Biochemistry. 29, 136221367. Goding J. W. and Burns G. F. (1981) Monoclonal antibody OKT-9 recognizes the receptor for transferrin on human acute lymphocytic leukemia cells. J. Immun. 127, 12561258. Gootenberg J. E., Ruscetti F. W., Mier J. W., Gazdar A. and Gallo R. C. (1981) Human cutaneous T cell lymphoma and leukemia cell lines produce and respond to T cell growth factor. J. P.Y~. Med. 154, 140331418. Gruber M., Schodin B. A., Wilson E. R. and Kranz D. M. (1994) Efficient tumor cell lysis mediated by a bispecific single chain antibody expressed in Escherichia coli. J. Immun. 152, 5368-5374.
Hayden M. S., Linsley P. S., Gayle M. A., Bajorath J., Brady W. A., Norris N. A., Fell H. P.. Ledbetter J. A. and Gilliland L. K. (1994) Single-chain mono- and bispecific antibody derivatives with novel biological properties and antitumour activity from a COS cell transient expression system. Therapeut. Immun. 1, 3315. Hedrick S. M., Matis L. A.. Hecht T. T., Samelson L. E., Longo D. L., Heber-Katz E. and Schwartz R. H. (1982) The fine specificity of antigen and Ia determinant recognition by T cell hybridoma clones specific for pigeon cytochrome c. Cell 30, 141-I 52. Holliger P.. Prosper0 T. and Winter G. (1993) “Diabodies”: small bivalent and bispecific antibody fragments. Proc. natn. Acad. Sci. U.S.A. 90, 64446448. Holliger P. and Winter G. (1993) Engineering bispecific antibodies. Curr. Opin. Biotech. 4, 44G-449. Huston J. S., Levinson D., Mudgett-Hunter M., Tai M. S., Novotny J., Margolies M. N., Ridge R. J., Bruccoleri R. E., Haber E., Crea R. and Oppermann H. (1988) Protein engineering of antibody binding sites: recovery of specific activity in an anti-digoxin single-chain Fv analogue produced in Escherichia coli. Proc. natn. Acad. Sci. U.S.A. 85, 58795883.
Huston J. S., McCartney J., Tai M.-S., Mottola-Hartshorn C., Jin D., Warren F., Keck P. and Oppermann H. (1993) Medical applications of single-chain antibodies. Int. Rev. Immun. 10, 195-217.
Recombinant
single-chain
bispecific
Jost C. R., Kurucz I., Jacobus C. M., Titus J. A., George A. J. T. and Segal D. M. (1994) Mammalian expression and secretion of functional single-chain Fv molecules. J. biol. Chem. 269,26261-26213. Kostelny S. A., Cole M. S. and Tso J. Y. (1992) Formation of a bispecific antibody by the use of leucine zippers. J. I~n~lmzm. 148, 1541-l 553. Kroesen B. J., Ter Haar A.: Spakman H., Willemse P., Sleijfer D. T., De Vries E. G. E., Mulder N. H., Berendsen H. H., Limburg P. C.. The T. H. and De Leij L. (1993) Local antitumour treatment in carcinoma patients with bispecificmonoclonal-antibody-redirected T cells. Crtnceu Immun. Immunother. 37, 4OWOl. Kurucz I., Titus J. A., Jost C. R., Jacobus C. M. and Segal D. M. (1995) Retargeting of C’TL by an efficiently refolded bispecilic single-chain Fv dimer produced in bacteria. J. Immun. 154,4576-4582. Lanzavecchia A. and Scheidegger D. (1981) The use of hybrid hybridomas to target human cytotoxic T lymphocytes. Eur. J. Immuti. 17, 105-I II. Leo O., Foo M., Sachs D. H., Samelson L. E. and Bluestone J. A. (1981) Identification of a monoclonal antibody specific for a murine T3 polypeptide. Proc. natn. Acad. Sci. C’.S.A. 84, 1314-l 318. Mallender W. D. and Voss E. W. Jr (1994) Construction, expression, and activity of a bivalent bispecific single-chain antibody. J. hiol. C’hem. 269, 199-206. Mezzanzanica D., Garrido M. A., Neblock D. S., Daddona P. E., Andrew S. M.. Zurawski V. R., Segal D. M. and Wunderlich J. R. (1991) Human T-lymphocytes targeted against an established ovarian carcinoma with bispecific F(ab’), antibody prolong host survival in a murine xenograft model. C~mc.er.Res. 51, 5116~5121, Nicholls P. J.. Johnson V. G., Blanford M. D. and Andrew S. M. (1993) An improved method for generating single-chain antibodies from hybridomas. .I. immunol. Meth. 165, 8 I-91. Nitta T., Sato K., Yagita H.. Okumura K. and Ishii S. (1990) Preliminary trial of specific targeting therapy against malignant glioma. Lutwr 335, 368~~37 1. Perez P.. HolTman R. W.. Shaw S., Bluestone J. A. and Segal D. M. (1985) Specific targeting of cytotoxic T cells by antiT3 linked to anti-target cell antibody. Nature 316, 354-356. Perez P.. Hoffman R. W., Titus J. A. and Segal D. M. (1986) Specific targeting of human peripheral blood T cells by heteroaggregates containing anti-T3 crosslinked to antitarget cell antibodies. J. exp. Med. 163, 166-118. Reiter Y.. Brinkmann U., Kreitman R. J., Jung S. H., Lee B. and Pastan I. (1994) Stabilization of the Fv fragments in recombinant immunotoxins by disulfide bonds engineered into conserved framework regions. Biochemistr~~ 33, 545 I5459.
Renner C., Jung W.. Sahin U., Denfeld R., Pohl C., Trumper L.. Hartmann F., Diehl V., Van Lier R. and Pfreundschuh M. (1994) Cure of xenografted human tumors by bispecific monoclonal antibodies and human T cells. Science 264,833835.
Sachs D. H., Mayer N. and Ozato K. (1981) Hybridoma antibodies directed toward murine H-2 and Ia antigens. In Monorlonul Antibodies and T-cell Hybridomus (Edited by
Fvz redirects lysis by CTL
219
Hammerling J., Hammerling U. and Kearney J. F.), p.95. Elsevier/North-Holland, New York, Schneider C.. Sutherland R., Newman R. and Greaves M. (1982) Structural features of the cell surface receptor for transferrin that is recognized by the monoclonal antibody OKT9. J. hiol. Chetn. 251, 8516-8522. Segal D. M., Titus J. A. and Stephany D. A. (1987) The use of fluorescence flow cytometry for the study of lymphoid cell receptors. In Methods in Etqtnolog~~: Immunological Technigues, Parr K: in vitro Models qf’B and T Cell Fun&on and Lymphoid Cell Receptors (Edited by Di Sabato G.). p. 478. Academic Press, New York. Segal D. M., Urch C. E., George A. J. T. and Jost C. R. (1992) Bispecific antibodies in cancer treatment. In Biologic Therap) o/‘Cuncer Updutes, Vol. 2 (Edited by DeVita V. T., Hellman S. and Rosenberg S. A.), p. 1. J. B. Lippincott, Philadelphia. Segal D. M. (1993) SPDP crosslinking of antibodies to form heteroconjugates mediating redirected cytotoxicity. In CJtotosic. Cells; Recognition, &ffkctor Function, Generution, and Methods (Edited by Sitkovsky M. V. and Henkart P. A.), p. 485. Birkhauser, Boston. Segal D. M., Jost C. R. and George A. J. T. (1993) Targeted cellular cytotoxicity. In Cvtoto.uic Cells: Generation, Rrcognition, IY#kror Funcfions. Method~y (Edited by Sitkovsky M. V. and Henkart P. A.). p. 96. Springer/Verlag, Berlin. Segal D. M. and Snider D. P. (1989) Targeting and activation of cytotoxic lymphocytes. Chem. Immun. 47, 179-2 13. Shalaby M. R.. Shepard H. M., Presta L., Rodrigues M. L., Beverley P. C. L., Feldmann M. and Carter P. (1992) Development of humanized bispecific antibodies reactive with cytotoxic lymphocytes and tumor cells overexpressing the HER2 protooncogene. J. exp. Med. 175,217-225. Staerz U. D., Kanagawa 0. and Bevan M. J. (1985) Hybrid antibodies can target sites for attack by T cells. Nafure 314, 628-631.
Staerz U. D. and Bevan M. J. (1986) Hybrid hybridoma producing a bispecific monoclonal antibody that can focus effector T-cell activity. Proc. ncrtr2. Acad. Sc~i. U.S.A. 83, 14531457.
Unkeless J. C. (1979) Characterization of a monoclonal antibody directed against mouse macrophage and lymphocyte Fc receptors. J. r.\-p. Med. 150, 580-596. Van Dijk J.. Zegveld S. T., Fleuren G. J. and Warnaar S. 0. (1991) Localization of monoclonal antibody G250 and bispecific monoclonal antibody CD3iG250 in human renalcell carcinoma xenografts: relative effects of size and affinity. Inr. J. Cancer 48, 138-143. Weiner G. J., Kostelny S. A., Hillstrom J. R., Cole M. S., Link B. K., Wang S. L. and Tso J. Y. (1994) The role of T cell activation in anti-CD3 X antitumor bispecific antibody therapy. J. Immun. 152, 2385-2392. Weiner G. J. and Hillstrom J. R. (1991) Bispecific anti-idiotype/anti-CD3 antibody therapy of murine B cell lymphoma. J. Immun. 147, 40354044. Wunderlich J. R. and Shearer G. M. (1991) Induction and measurement of cytotoxic T lymphocyte activity, Unit 3. I 1. In Currenf Protocols in Immunology. Vol. 1 (Edited by Coligan J. E.. Kruisbeek A. M., Margulies D. H., Shevach E. M. and Strober W.), p. 3.11.1. John Wiley and Sons, New York.
Pergamon 0161~5890(95)00087-9
A SINGLE-CHAIN MAMMALIAN CAROLINA Experimental
BISPECIFIC Fv, MOLECULE PRODUCED CELLS REDIRECTS LYSIS BY ACTIVATED CTL R. JOST, JULIE DAVID
Immunology
A. TITUS, ISTVAN M. SEGAL*
Branch, National
(First received 3 Murch
Cancer Institute,
KURUCZ
IN
and
Bethesda, MD 20892, U.S.A.
1995; accepted in recised,form
15 MUJ 1995)
Abstract-Single-chain Fv (sFv) molecules consist of the two variable domains of an antibody (Ab) connected by a polypeptide spacer and contain the binding activities of their parental antibodies (Abs). In this paper we have attached the C-terminus of 2Cl I-sFv (anti-mouse CD3 c-chain) to the N-terminus of OKT9-sFv (anti-human transferrin receptor [TfR]) through a 23 amino acid inter-sFv linker consisting primarily of C,, region residues from 2Cl1, to form a single-chain bispecific Fv, [bs(sFv)J molecule. The bs(sFv), was expressed in COS-7 cells, and was secreted at the same rate as the two parental sFvs. The secreted protein had both anti-CD3 and anti-TfR binding activities. Essentially all of the secreted bs(sFv)z molecules bound TfR and the binding affinity of the bs(sFv), was comparable to that of OKT9 sFv and Fab. Thus, the attachment of the inter-sFv linker to the N-terminus of OKT9-sFv did not impair its binding function. The bs(sFv), retained both binding specificities after long-term storage at 4 C or overnight incubation at 37’C. It redirected activated mouse CTL to specifically lyse human TfR+ target cells at low (rig/ml) concentrations and was much more active than a chemically cross-linked heteroconjugate prepared from the same parental mAbs. Because bs(sFv), molecules secreted by mammalian cells are homogeneous proteins containing two binding sites in a single polypeptide chain, they hold great promise as an easily obtainable, economic source of a bispecific molecule suitable for in cico use. Kql words: Bispecific antibody,
single-chain
antibody,
INTRODUCTION Bispecific antibodies (bsAbs) are proteins that have two different binding specificities within the same molecule. They can be used in any application requiring the crosslinking of two different molecules, an important one being the redirection of lysis by CTL (Perez et al., 1985; Staerz et al., 1985; Segal and Snider, 1989). By linking CD3/TCR complexes on CTL with cell surface components on target cells, bsAbs induce CTL to lyse target cells regardless of their initial specificity. In particular, bsAbs with specificity for the TCR and a tumor Ag redirect CTL to specifically lyse tumor cells. In vivo treatment of tumor-bearing mice with such bsAbs results in the eradication of the tumor or slows its growth (Garrido et al., 1990; Mezzanzanica et al., 1991; Van Dijk et al., 1991; Weiner er a/., 1994; Weiner and Hillstrom, 1991;
*Author to whom correspondence should be adressed at: Building 10, Room 4Bl7, National Institutes of Health, Bethesda, MD 20892, U.S.A. Abbreviations: sFv, single-chain Fv, TfR, transferrin receptor, bs(sFv),, single-chain bispecific Fv,, bsAb, bispecific antibody, Fc,.R, Fc receptor for IgG, mAb, monoclonal antibody.
redirected
Demanet et al., 1991, 1994; Brissinck et al., 1991; Renner et al., 1994) and preliminary clinical studies using bsAbs in cancer patients have also been promising (Nitta et al., 1990; Bolhuis
et al., 1992; Kroesen
et al., 1993).
If bsAbs are to be used in uivo, they must be pure, homogeneous proteins, preferably lacking Fc portions which could, by interacting with FcyR on various cell types, induce premature triggering, lysis, or sequestration of effector cells. In the past, homogeneous bispecific F(ab’), fragments have been produced by digestion of hybrid hybridoma antibodies (Staerz and Bevan, 1986; Lanzavecchia and Scheidegger, 1987; Segal et al., 1992, 1993) or by crosslinking two different F(ab’)s through their hinge cysteines (Glennie et al., 1987; Segal et al., 1993). More recently, several types of genetically engineered bsAbs have been designed to facilitate their production (Holliger and Winter, 1993). For example, Fab’ molecules produced in bacteria (Shalaby et al., 1992; Carter et al., 1992) and in myeloma cells (Kostelny et al., 1992) were directly crosslinked by their hinge sulfhydryls, thus eliminating digestion steps. Genetic engineering has also led to the production of bsAbs consisting primarily of the variable domains of antibodies. These constructs are smaller than those containing whole F(ab’) fragments and would
211
lysis.
therefore
be expected
to have improved
tumor
212
C. R. JOST et al.
penetration. One construct uses two different polypeptide chains, Vu,-V,, and VL,-VH2, that associate laterally and non-covalently to form a bsAb known as a “diabody” (Holliger et al., 1993). Other constructs are composed of single-chain Fvs (sFvs), molecules in which V, and V, from the same mAb are connected by a polypeptide linker, and which have the Ag binding properties of the parental mAb (Huston et al., 1988, 1993; Bird et al., 1988). Single-chain bsAbs have been produced by joining two different sFvs with a polypeptide linker. Mallender and VOSS(1994) first described a single-chain bispecific sFv [bs(sFv),] having specificity for two different haptens, and Gruber et al. (1994) subsequently showed that a bs(sFv), having specificity for a TCR idiotope and for a hapten could redirect lysis against a haptenmodified cell. Both groups produced their bs(sFv),s in bacteria, and the proteins required refolding and affinity purification before they could be used. In another study, Hayden et al. (1994) produced a bs(sFv)> in a COS cell transient expression system as a fusion protein that contained the Fc region of human IgGl. The COS cells secreted active bsAb that redirected lysis but was heterogeneous in size due to aggregation of the Fc portions. In order to study the properties of a single, undimerized bs(sFv),, we have produced a construct consisting of 2Cl I-sFv (anti-murine CD3 t-chain) linked to OKT9sFv (anti-human transferrin receptor [TfR]). This was expressed in COS-7 cells, thus taking advantage of endoplasmic reticulum-based protein folding machinery and quality control mechanisms that allow only fully folded proteins to be secreted. We show that the bs(sFv), is produced and secreted as rapidly as each of the parental sFvs, that most of the bs(sFv), is active, and that the secreted material specifically redirects lysis mediated by activated mouse CTL at rig/ml levels.
MATERIALS
AND METHODS
Cell lines and antibodies
The following cell lines and Abs were used: COS-7 monkey kidney fibroblasts, K562 human erythroleukemia cells (both from American Type Culture Collection, Rockville, MD), 2B4 murine T hybridoma cells (Hedrick et al., 1982), HUT 102 human T lymphoma cells (Gootenberg et al., 1981), 9ElO mAb against the cmyc peptide (Evan et al., 1985), 2.4G2 mAb against the mouse FcyRII and III (Unkeless, 1979), 145-2C11 mAb (termed 2Cll) against mouse CD3 E chain (Leo et al., 1987), OKT9 mAb against human transferrin receptor (TfR) (Goding and Burns, 1981; Schneider et al., 1982) W6/32 mAb against human MHC class I molecules (Barnstable et al., 1978), and 36-7-5 mAb against H-2 Kk (Sachs et al., 1981). Fab fragments of W6/32 and 36-75 were prepared as described (Perez et al., 1986). Fab fragments from 2C11 and OKT9 were prepared using an ImmunoPure Fab Preparation Kit (Pierce, Rockford, IL) according to the manufacturers’ instructions. OKT9 Fab was labelled with FITC as described (Segal et al., 1987). Polyclonal rabbit anti-mouse IgG was from Cappel
(Organon Technika, Durham, NC). Heteroconjugates were prepared by crosslinking 2Cll mAb to OKT9 mAb using succinimidyl-3-(2-pyridyidithiol)-propionateas described (Segal, 1993). sFv and bs(sFu), constructs
The 2Cl I-sFv and OKT9-sFv (Nicholls et al., 1993) containing plasmids (pcDNA/AMP-2Cll and pSVLOKT9) have been described (Jost et al., 1994). Both sFvs have a VL-linker-V, design, a 5’ light-chain leader sequence and a 3’ c-myc peptide tag. The bs(sFv), was made by inserting a modified version of the OKT9-sFv into pcDNA/AMP-2Cll at the 3’ end of the Vu domain of 2Cl I-sFv. To this end we introduced a unique EcoR I site in pcDNA/AMP2Cll between Vu and the c-nzyc peptide using a Transformer Mutagenesis Kit (Clonetech Laboratories Inc, Palo Alto, CA). The mutating and selecting primers were, CC GTC TCC TCA TCA GAA TTC AAA CTC ATC TCA GAA GAG G and GGG AGA CCC ACG CGT GGT ACC GAG CTC GG, respectively. The OKT9-sFv insert was generated by PCR using pSVLOKT9 as a template and a 5’-primer, ATATA AAG CTT GAA TTC GCC AAA ACA ACA GCC CCA TCG GTC TAT CCA CTG GCC CCT GTG CTC GAG TCT AGT GGT AGC GGA GAC ATC AAG ATG ACC CAG TC and a 3’-primer TATAT TCT AGA CTC TTA TTA ATT CAG ATC CTC. The PCR product consisted of an EcoR I site followed sequentially by the inter-sFv linker, OKT9-sFv, the c-myc peptide tag, stop codons and a Xba I site. The PCR product was directly ligated into pCR 11 (Invitrogen, San Diego, CA), which was then digested with EcoR I and Xba I. The resulting fragment was ligated into the pcDNA/AMP-2Cll (MUT) that had been digested with the same restriction enzymes yielding the bs(sFv), construct shown in Fig. 1. Transfection and metabolic labeling
Procedures used for transfection and metabolic labeling have been previously described (Jost et al., 1994). In short, COS-7 cells were cultured in 1O-cm culture dishes in DMEM medium (Dulbecco’s modified Eagle’s medium supplemented with penicillin (100 U/ml), streptomycin (100 U/ml), L-glutamine (2 mM) and 10% fetal calf serum) and transfected with 20 pg of plasmid DNA using the calcium phosphate precipitation method (Davis et al., 1986). When the transfected cells were used to produce non-radioactive bs(sFv),, the cells were allowed to grow for an additional 48 hr, after which the supernatant was harvested, diluted and used in cytotoxicity experiments. To produce “S-methionine labeled bs(sFv), [j5Sbs(sFv),], cells were trypsinized 16 hr after transfection, replated in three IO-cm culture dishes, and allowed to grow for an additional 24 hr prior to labeling. Cells (90% confluent) were pulsed for 1.5 hr at 37°C with ‘5Smethionine (Transj5S-label [ICN Radiochemicals, Irvine, CA]) at a concentration of 0.15 mCi/ml in a total volume of 4 ml and chased for 2.5 hr in DMEM. The radioactive supernatants from identical transformations were pooled
Recombinant
single-chain
213
bispecific Fv, redirects lysis by CTL OKT9 SFV
2C11 SFV N
C MYC PEPTIDE
VL
inter-&G
ala lys thr thr ala pro ser val tyr pro leu ala pro vai
2C1 1 VH - glu phe
I
LINKER
I
2Cll
ser ser gly ser giy - OKT9
leu glu
I
CHl
I
VL
flexible
Fig. 1. Diagram of the 2Cll-OKT9 bs(sFv), protein. The bs(sFv), protein has a 2Cll-sFv-linkerOKT9-sFv design. The asterisk indicates the Ag binding sites of the bs(sFv),. N and C refer to the N- and C-termini of the molecule. The glu-phe and leu-glu residues in the linker result from EcoR I and Xba I sites, respectively.
and supplemented with 10 mM HEPES and 0.01% sodium azide prior to use. In the experiment shown in Fig. 2, cells were pulsed for 1 hr and chased for varying times in DMEM containing 15 mg/ml of L-methionine.
Immunoprecipitation
Immunoprecipitation and the analysis of the precipitated proteins was performed at 4°C as described (Jost
OKTS-sFv -4)---Cells
__)
%.
P
2Cll -sFv
4-Supematant
--b
d----Cells
__)
.x
2h
r)-
Supematant
--W
-
6h
P
2h
6h
2h
P
6h
P
2h
Bispecific-sFv2 +Cells
69kD-
---W
M-Supematant
+
-Qmlb :
P
..
2h
6h
P
2h
6h
Fig. 2. Production and secretion of sFvs by COS-7 cells. COS-7 cells transfected with plasmids encoding the indicated protein were pulsed for 1 hr with “S-methionine and chased for 0 (“P” in Fig.), 2 or 6 hr, as indicated. sFvs were immunoprecipitated from either cell lysates (“Cells”) or from culture media (“Supernatant”), subjected to SDS-PAGE under reducing conditions, and visualized by autoradiography. The OKT9-sFv and the bs(sFv), contain an N-linked carbohydrate group that contributes to their observed molecular weights.
6h
214
C. R. JOST et al.
et al., 1994). Briefly, cells were lysed and proteins were precipitated using 9ElO (anti-c-m-rc peptide) and protein A-Sepharose precoated with rabbit anti-mouse Ig. In some experiments, protein was eluted from the beads by incubation for 5 min at 95°C in elution buffer (20 mM Tris, I mM EDTA, 2% SDS 0.1% Bromphenol Blue and 5% 2-ME pH 7.8) and eluates were analysed by reducing 12.5% SDS-PAGE and autoradiography. In other experiments, beads were divided into four equal portions and treated as above with elution buffer (without Bromphenol Blue). Eluates were transferred directly to 2.5 ml of Ecolume (ICN, Cleveland, OH) scintillation solution and counted.
,ug OKT9 mAb, incubated for various times at 4 C and spun for 10 sec. Cell pellets were lysed, bs(sFv)z and sFv were immunoprecipitated, and the immunoprecipitates analysed by scintillation counting as described above. The amount of sFv and bs(sFv), that bound non-specifically to K562 cells was determined from the radioactivity bound in the presence of excess OKT9 mAb. Quadruplicate samples were averaged, and the non-specifically associated radioactivity (less than 10% of that bound to K562 cells at 0 time) was subtracted. In both radioactive and fluorescence measurements, background-corrected data were divided by background corrected 0 time values and multiplied by 100 (per cent of maximal bound).
bs(sFv), binding
Cytotoxicit),
K562 cells (9 x 10”) and 2B4 cells (2 x 10’) were preincubated at 4°C with or without OKT9 Fab (30 pg) or 2C 11 Fab (50 pg), respectively, in 200 ~1 wash medium (HBBS, 0.1% BSA, 0.01% sodium azide). After 30 min, 2 ml of ‘5S-bs(sFv),-containing medium was added and cells were incubated for 2 hr at 4°C. Cell-associated bs(sFvs)s was immunoprecipitated and analysed by SDS PAGE as described above. In other studies, cells were incubated with varying amounts of parental or control Fab before addition of 1 ml “S-bs(sFv)+,. Immunoprecipitated bs(sFv)z was quantified by scintillation counting as described above. Non-specific binding [the amount of radioactivity bound in the presence of excess (50 pg) parental Ab] was subtracted and % inhibition was calculated using the formula:
Cytotoxicity was studied by using a standard “Crrelease assay (Wunderlich and Shearer, 1991). “Cr-labeled HUT 102 cells were used as target cells. CTL effector cells were generated in 5-6 day primary allogeneic MLR cultures by incubating splenic effector cells with irradiated allogeneic stimulator cells. Strain combinations used to generate CTLs were Balb/c antiC57Bl/lO, C57Bl/lO anti-Balb/c and C3H/Hen antiBalb/c. A varying amount of effector cells was added to each well of a 96-well microtiter plate. Varying dilutions of COS-7 supernatant containing either bs(sFv)z, OKT9sFv, 2Cl I-sFv. medium alone or medium containing graded concentration of the 2Cl l-OKT9 heteroconjugate were then added followed by either 5 x 10’ or 1 x lo4 target cells. When using the 2Cll x OKT9 heteroconjugate, 2 pg of 2.4G2 anti-FcyR mAb was added to each well to prevent interactions with Fc receptors. In inhibition experiments, effecters and targets were added directly to the wells (no wash step) after a brief preincubation with inhibiting or control antibodies. Percentage inhibition was calculated as described above for binding but in this case I and T refer to the per cent lysis in the presence (I) or absence (T) of inhibitor.
% Inhibition
= lOO(1 - Z/T),
where Z and T equal the number of counts bound specifically in the presence and absence of inhibitor, respectively. Dissociation
qf OKT9
Fab
K562 cells (10’) were incubated for 1 hr at 4°C with 2 pg of FITC-OKT9 Fab, with (background) or without (samples) 50 pg OKT9 mAb in a total volume of 1 ml of wash medium. Cells were centrifuged for 5 min at 2OOg, resuspended in 1 ml wash medium containing 50 rug OKT9 mAb, and incubated at 4’C. At various times, samples were analysed by flow cytometry. Mean fluorescence values were determined and background values (12% of initial value, 22% of final value) were subtracted at each time point. Zero time was taken as the first point measured.
RESULTS sFv and bs(sFv)2 constructs
The bs(sFv), and its parental sFvs were produced in COS-7 cells transfected with vectors encoding the following molecules: 2Cl I-sFv, against the murine CD3-c chain; OKT9-sFv, against the human TfR; and the bs(sFv), against both Ags. All constructs contained light chain leaders to direct newly-synthesized proteins to the endoplasmic reticulum and a C-terminal c-m~‘c peptide to facilitate detection. Both parental sFvs contained an Dissociation qf OKT9 sFv and the bs(sFv), intra-sFv (Gly,-Ser), linker connecting the C-terminal V, with the N-terminal V,,. The bs(sFv), is shown diaMultiple samples of K562 cells (6 x lo6 per sample) in 1 ml of COS-7 medium containing “5S-bs(sFv), or 35S- grammatically in Fig. 1. An inter-sFv linker connects the OKT9-sFv were incubated 1 hr at 4°C in the presence or C-terminus of 2Cll -sFv to the N-terminus of the OKT9sFv. This inter-sFv linker consists of two residues arising absence of excess OKT9 mAb. Cells were centrifuged for 5 min at 2OOg, and one sample (0 time point) was from an EcoR I site, the first 13 amino acids of the 2Cll resuspended in cold wash medium and immediately spun C,l domain followed by two residues from a Xho I site for 10 set in an Eppendorf microcentrifuge. Other sam- and a flexible hydrophilic region. The bs(sFv),-encoding a vector that ples were suspended in 1 ml wash medium containing 50 construct was cloned into pcDNA/AMP,
21.5
Recombinant single-chain bispecific FvZ redirects lysis by CTL
(a)
expresses large amounts of protein in COS-7 cells under control of the CMV promoter.
2B4 cells
K562 cells
Production and secretion oj’sFv molecules Transfected COS-7 cells were labeled with “S-methionine for 60 min and then chased for either 0, 2 or 6 hr. The sFvs and bs(sFv), were immunoprecipitated from both the cell lysates and the supernatants and analysed by SDS-PAGE. Figure 2 shows that the bs(sFv)z and both parental sFvs were secreted from the cells at similar rates; by 6 hr, most of the radiolabeled bs(sFv), and sFvs had left the cells and were found in the supernatant. Thus, the increased size and complexity of the bs(sFv), resulting from the concatenation of two sFv molecules does not affect their ability to be processed and secreted by COS7 cells.
inhibitor
+
-
+ 2Cll
OKT9 (b) 2C11 Fab
36-7-5
Fab
0.2 E 0.6 2 6 20 _
I+ b
20 >
Secreted bs(sFv)> spec$cally binds both Ags To determine if secreted sFvs were active, we tested the ability of radiolabeled material from the medium of transfected COS-7 cells to specifically bind Ag. Figure 3a shows that the bs(sFv), bound to 2B4 (CD3+) and to K562 (TfR+) cells and that binding was inhibited by the parental antibodies. These antibodies inhibited %bs(sFv), binding in a dose-dependent fashion (b) but inhibition did not occur when antibodies binding to MHC class I molecules on these cells were used as controls. Thus, the bs(sFv), specifically binds both Ags. To determine the fraction of bs(sFv), that had binding activity, we depleted 35S-bs(sFv), from the COS-7 medium by four sequential incubations with either 2B4 or OKT9 cells, and then immunoprecipitated the remaining bs(sFv), using the 9ElO mAb which recognizes the c-my peptide tag. In Fig. 4 the top panel shows that most of the bs(sFv), was removed by absorption to K562 cells and very little could subsequently be immunoprecipitated with the 9ElO mAb, indicating that a large fraction of the bs(sFv), present in the supernatant had the TfR binding activity of OKT9. The lower panel shows that a small and nearly equal amount of bs(sFv)? was bound to 2B4 cells at each absorption step but that a large amount of bs(sFv), remained in the medium even after four absorptions. Because the amount of bs(sFv), removed at each absorption step was a small fraction of the total, we were unable to determine the fraction of bs(sFv)2 having antiCD3 binding activity. The low binding of the bs(sFv), to 2B4 cells was not due to the 2Cll portion being in the bispecific configuration, since similar results were obtained with the parental 2Cll -sFv (Jost et al., 1994).
OKT9Fab
W6/32
‘,l.k-
Fab
,
10 h 0
f 20
40 Percent
60 inhibition
60
100
of binding
Fig. 3. Secreted bs(sFv), specifically binds its Ags.(a) K562 (TfR+) and 2B4 (CD3+) cells were incubated for 2 hr at 4°C with medium containing “S-bs(sFvs), in the presence or absence of the indicated parental Ab. Bound bs(sFv)2 was immunoprecipitated and analysed by SDS-PAGE under reducing conditions. (b) Dose dependence of the inhibition was studied by incubating K562 cells (shaded bars) and 2B4 cells (open bars) with 35S-bs(sFv)1 in the presence of the indicated concentrations (in pg/ml) of parental Fab or with a control Fab (anti-Kk Fab for 2B4 cells and anti-HLA I Fab for K562 cells). After the incubation, bound bs(sFv), was immunoprecipitated, counted, and the percentage inhibition was calculated as described in Materials and Methods. The asterisk indicates no inhibition.
that the bs(sFv), dissociates from K562 cells at a rate intermediate to those of the sFv and Fab, but that there is not a great difference in dissociation rates of the three different molecules. Dissociation rates are directly related to binding affinity (Berzofsky et al., 1993) implying that the affinities of free OKT9-sFv and the OKT9-sFv in the bs(sFv), are similar. Thus, the presence of the N-terminal inter-sFv linker has little detrimental effect on the binding affinity of the bs(sFv)2 for the TfR. The bs(sFv)2 is stable at 37’ C
The bs(sFv), binds TfR with a similar af$nity to OKT9 sFr and Fab The inter-sFv linker attaches to the N-terminus of the OKT9-sFv and might therefore interfere with Ag binding (Fig. 1). To determine whether the presence of this linker had an effect on the binding of the bs(sFv), to the TfR, we compared the dissociation rate of the bs(sFv), with those of the OKT9-sFv and OKT9 Fab. Fig. 5 shows
It has been reported that incubation in serum at 37°C can lead to rapid inactivation of some sFvs (Glockshuber et al., 1990; Reiter et al., 1994). However, immunoprecipitation and sequential absorption experiments showed that most of the bs(sFv), remained in solution and still had binding activity after incubation for 16 hr at 37°C in medium containing 10% serum (data not shown).
216
C. R. JOST et al. T
1
2
4
3
R
K562
Sequential Precipitations T
2
1
___) 4
3
R
Sequential Precipitations ---W Fig. 4. Depletion of bs(sFv)2from COS-7 supernatant. Medium (4 ml) containing 95S-bs(sFv),was sequentially absorbed with either K562 cells (TfR+, 2 x IO’ cells/absorption) or 2B4 cells (CD3+, 6 x lo7 cells/absorption). T indicates the total amount of bs(sFv), present in 4 ml culture medium before absorptions, numbers above lanes indicate the absorption step (numbered lanes contain cell-bound bs(sFv),), and R indicates bs(sFv)z remaining in the medium after the final absorption. Shown are the SDS-PAGE analyses of immunoprecipitates.
allogeneic MLR were used as effector cells. Figure 6a shows that the bs(sFv), efficiently redirected lysis at E : T ratios of 5 : 1 and 10 : 1, while supernatants containing the parental sFvs did not. Figure 6b shows that the bs(sFv), mediated lysis in a dose-dependent fashion, and still gave significant lysis when diluted 128-fold. Lysis mediated by a chemically crosslinked heteroconjugate between 2Cll and OKT9 was much lower than that mediated by the bs(sFv),. Heteroconjugates made in the same way from other antibodies have been more potent at redirecting lysis (Perez et al., 1986), and it may be that either 2Cll or OKT9 is especially susceptible to inactivation by the crosslinking reagent. Both parental Fabs inhibited lysis, while irrelevant antibodies that bound to either effector or target cells did not (Fig. 7). Thus, the simultaneous binding of the bs(sFv), to both effector and target cells was required for lysis.
2Cll-sFv 30
20
10 , :gtgy 2
‘.
-.
- - - - - - - __
4
6
6
10
12
14
16
Bs(sFv),
Effector
OKTS-sFv OKT9
Fab
60
20
to target
ratio
4o (W 30
i
04 0
20
40
Time
60
I 80
--o--
2CllxOKT9
d-
bs(sfv),
---m-
OKT9-sFv
-+-
2Cll-SFV
:_I_
(Min)
Fig. 5. Dissociation of OKT9-sFvs, bs(sFv)z and OKT9 Fab from K562 cells. 35S-bs(sFv)2(H), OKT9-sFv (0) and FITC labeled OKT9 Fab (V) were bound to K562 cells. Cells were washed and resuspended in excess unlabeled OKT9 mAb. Radioactive protein bound at the indicated times was precipitated and measured by scintillation counting. Bound FITC labelled Fab was analysed by FACS.
Redirected cytotoxicity Because the bs(sFv), has dual specificity for murine CD3 and human TfR we asked whether it could redirect mouse T cells to lyse the TfR+ HUT-102 cells. We used COS-7 cell supernatants that had been harvested 48 hr after transfection with vectors encoding either the bs(sFv), or parental sFvs. CTLs generated in a primary
0
1
2
Log,
3
4
Dilution
Fig. 6. Targeting of lysis using bs(sFv),. Diluted COS-7 supernatants containing bs(sFv), (V), OKT9-sFv (m), 2Cl I-sFv (A), chemically crosslinked 2C I 1 x OKT9 heteroconjugate (0) or medium alone (-•-) were used in a standard 3.5 hr “Cr releaseassaywith HUT-102 cellsas targets and activated mouse spleen cells as effecters. (a) Lysis by titrated doses of Balb/c allo-stimulated effector cells. COS-7 cell supernatants were diluted 1 : 16 and the heteroconjugate was used at 0.38 pg/ml. (b) Two-fold serial dilutions of the transfected COS-7 supernatants with a 1 : 16 dilution as starting point, and 2Cll x OKT9 heteroconjugate (0.5 pg/ml starting point) were used to redirect lysis mediated by C57B1/6 allo-stimulated effector cells at an E:Tratio of 10: I.
Recombinant single-chain bispecific Fvz redirects lysis by CTL
217
fore, the rate at which proteins are secreted is an indication of the rate at which they are folded and assembled. We had previously found that COS-7 cells fold and secrete genetically engineered sFvs at different rates depending on the sFv in question (Jost et al., 1994). We were therefore surprised to find that the bs(sFv), is secreted at the same rate as its constitutive sFvs since the bispecific construct is a more complex molecule conOKT9 Fab 1 taining four disulfide bonds and two sets of VL-V, domain pairs. Apparently, the additional complexity of the bs(sFv), poses no problem for the protein folding W6/32 Fab IO machinery in COS-7 cells. 100 20 40 60 80 0 We were concerned that the presence of two different inhibition of lysis sFvs within one molecule could lead to mispairing of V Fig. 7. The specificity of redirected lysis. The specificity of lysis domains as has been reported in the production of bispecof HUT 102 cellswas studied in the presenceor absenceof the ific antibodies from hybrid hybridomas (De Lau et al., indicated amounts (in pg/ml) of parental or irrelevant Fabs. Open bars, C3H allo-stimulated effecters were preincubated 1991). However, we found, by absorption, that almost with 2Cll or 36-7-5 (anti-Kk) Fab and used at an E : T ratio of all bs(sFv), had OKT9 binding activity suggesting that 8.5 : 1 in the presence of a 1 : 8 dilution of bs(sFv),-containing the variable domains are correctly matched in most culture medium. Shaded bars, HUT 102 target cells were pre- secreted molecules. Absorption experiments could not be incubated with OKT9 or W6/32 (anti-HLA 1) Fab. Balb/c allo- used to test whether all bs(sFv), molecules have anti-CD3 stimulated effecters were used at an E : T ratio of 5 : 1 in the activity because too few bs(sFv)? molecules bound to 2B4 presence of a 1 : 8 dilution COS-7 supernatant containing cells at each absorption step. This could be due to the bs(sFv),. The asterisk indicates no inhibition. low number of CD3 molecules on the 2B4 cells and/or to the low affinity of 2Cll for its antigen (K, of 2Cll Fab for CD3 is approximately 0.7 x lo-’ MP’, and 2B4 cells We tried to measure the amount of bs(sFv)z used in express about 2 x IO4 CD3 molecules/cell, T. Bakacs, the “Cr-release assay by inhibiting the binding of 35S- unpublished result). Thus, although we cannot formally prove that all bs(sFv), molecules have CD3 binding bs(sFv), with the medium used to redirect lysis [which contained cold bs(sFv),]. No inhibition was observed activity, we have no reason to believe that this is not the even when the medium was used undiluted. As seen in case. Fig. 3B, OKT9 Fab inhibited the binding at 0.3 pg/ml Inter-sFv linkers that have been used in bs(sFv), conbut not at 0.1 pg/ml. Therefore, assuming that the structs include: a 24-amino acid flexible linker modelled bs(sFv), and the OKT9-Fab bind with similar affinities, after a secreted fungal cellulase domain (Mallender and there is probably less than 0.3 pg/ml of cold bs(sFv), in Voss, 1994); a 25-amino acid linker rich in ala, asp and the culture supernatant. On the other hand, the same lys residues (Gruber et al., 1994); a 27-amino acid helical OKTbs(sFv), produced in bacteria can redirect lysis at peptide linker containing a recurring EEAKK motif concentrations as low as 0.2 rig/ml (Kurucz et al., 1995). (Hayden et al., 1994); and the 23-residue linker we used Therefore, the concentration of bs(sFv), secreted by the consisting of the first 14 amino acids of the 2Cll C,, transfected COS-7 cells is likely to be in the range of 0.3domain followed by a five-amino acid flexible hydrophilic 0.02 pg/ml. sequence. All of these linkers resulted in active bispecific molecules, suggesting that the compositions of the intersFv linkers are relatively unimportant. Nevertheless one DISCUSSION would expect that a minimal linker length might be In this paper we describe a genetically engineered required to prevent the V domains of one sFv from intersingle-chain bsAb consisting of two sFvs linked together by acting with those of the other sFv. We were curious a polypeptide spacer, one sFv reactive against the human whether the presence of the inter-sFv linker, which TfR and the other against murine CD3. When transfected attaches to the N-terminus of the OKT9-sFv, would into COS-7 cells, the bs(sFv), is secreted at the same rate decrease the affinity of the OKT9 portion of the bispecific as the parental sFvs and has a molecular weight of 69 molecule for Ag. We found, however, that the affinity of kDa, as expected for a glycosylated bs(sFv),. The secreted the bs(sFv&, as measured by its rate of dissociation from bs(sFv), recognizes both Ags and specifically redirects TfR+ cells, is similar to that of the OKT9-sFv and Fab. mouse CTL to kill human TfR+ target cells. The bs(sFv), In addition, we have subsequently produced this protein protein is stable for at least 16 hr at 37”C, and was still in bacteria in large amounts and have found its affinity active after several months of storage at 4°C. for CD3 to be indistinguishable from that of the parental It has been established that mammalian cells fold pro2Cll Fab (Kurucz et al., 1995). This suggests that at least teins in the endoplasmic reticulum and that a quality with this bsAb, the inter-sFv linker did not interfere with control mechanism is in place to prevent incorrectly Ag binding. folded or unassembled proteins from leaving the endoThe 2C11 x OKT9 bs(sFv), present in culture superplasmic reticulum (Gething and Sambrook, 1992). Therenatants was still able to redirect lysis at high dilutions 2Cll
Fab
0.1
C. R. JOST et al.
218
after at least 3 months storage at 4 ‘C, and the bs(sFv), and its parental sFvs were stable to an overnight incubation in culture medium at 37’C. It has been reported (Glockshuber et al., 1990; Reiter et al., 1994) and we also have found (unpublished observation) that some sFvs are unstable at 37°C. Thus, it is to be expected that not all bs(sFv),s will be as stable as the 2Cll x OKT9 bs(sFv),. Since instability would present problems both when producing the proteins in mammalian cells and when using them in biological systems, sFvs should first be screened for stability before using them in bs(sFv), constructs. Because the bs(sFv), was active at low protein concentrations and because mammalian cells secrete primarily active proteins, the retargeting activity of the bs(sFv), could be studied using supernatants directly. Therefore, the main advantage of the COS-7 cell secretion system is that it allows rapid testing of newly produced bs(sFv), constructs. Once a bs(sFv), construct has proven to be stable, to recognize both antigens, and to redirect lysis, it could be produced in large amounts either in bacteria or in stably transfected mammalian cells. The next step would be to determine the fate and efficacy of the bs(sFv)z in vim. Acknowledgements-The authors thank Apurva Sarin, Mark Williams and Belen Moreno for providing allo-stimulated effector cells, David Winkler for producing the OKT9 mAb, also Tibor Bakacs and Giuseppe Sconocchia for thoughtful suggestions during the preparation of the manuscript.
REFERENCES Barnstable C. J., Bodmer W. F., Brown G., Galfre G., Milstein C., Williams A. F. and Ziegler A. (1978) Production of monoclonal antibodies to group A erythrocytes, HLA and other human cell surface antigens-new tools for genetic analysis. Cell14,
9-20.
Berzofsky J. A., Berkower I. J. and Epstein S. L. (1993) Antigen-antibody interactions and monoclonal antibodies. In Fundumental Immunology (Edited by Paul W. E.), p. 421. Raven Press Ltd, New York. Bird R. E., Hardman K. D., Jacobson J. W., Johnson S., Kaufman B. M., Lee S. M., Lee T., Pope S. H., Riordan G. S. and Whitlow M. (1988) Single-chain antigen-binding proteins. Science 242,423-426. Bolhuis R. L., Lamers C. H., Goey S. H., Eggermont A. M., Trimbos J. B., Stoter G., Lanzavecchia A., Di Re E., Miotti S., Raspagliesi F., Rivoltini L. and Colnaghi M. I. (1992) Adoptive immunotherapy of ovarian carcinoma with bsMAb-targeted lymphocytes: a multicenter study. ht. J. Cancer 7(suppl.), 78-81. Brissinck J., Demanet C., Moser M., Leo 0. and Thielemans K. (1991) Treatment of mice bearing BCLl lymphoma with bispecific antibodies. J. Immun. 147,4019-4026. Carter P., Kelley R. F., Rodrigues M. L., Snedecor B., Covarrubias M., Velligan M. D., Wong W. L. T., Rowland A. M., Kotts C. E., Carver M. E., Yang M., Bourell J. H., Shepard H. M. and Henner D. (1992) High level Escherichia coli expression and production of a bivalent humanized antibody fragment. Bio/Technology 10, 163-167. Davis L. G., Dibner M. D. and Battey J. F. (1986) Basic Methods in Molecular Biology. Elsevier, New York. De Lau W. B. M., Heije K., Neeges J. J., Oosterwegel M.,
Rozemuller E. and Bast B. J. E. G. (1991) Absence of preferential homologous H/L chain association in hybrid hybridomas. J. Immun. 146,90&914. Demanet C., Brissinck J., Van Mechelen M., Leo 0. and Thielemans K. (1991) Treatment of murine B cell lymphoma with bispecific monoclonal antibodies (anti-idiotype x anti-CD3). J. Immun. 147, 1091~1097. Demanet C., Brissinck J., Leo O., Moser M. and Thielemans K. (1994) Role of T-cell subsets in the bispecific antibody (anti-idiotype x anti-CD3) treatment of the BCL, lymphoma. Cancer Res. 54,2973-2978. Evan G. I., Lewis G. K., Ramsay G. and Bishop J. M. (1985) Isolation of monoclonal antibodies specific for human c-myc proto-oncogene product. Molec. Cell Biol. 5, 361&3616. Garrido M. A., Valdayo M. J., Winkler D. F., Titus J. A., Hecht T. T.. Perez P., Segal D. M. and Wunderlich J. R. (1990) Targeting human T lymphocytes with bispecific antibodies to react against human ovarian carcinoma cells in nu/nu mice. Cancer Res. 50,42274232. Gething M. J. and Sambrook J. (1992) Protein folding in the cell. Nature 355, 3345. Glennie M. J., McBride H. M., Worth A. T. and Stevenson G. T. (1987) Preparation and performance of bispecific F(ab)‘? antibody containing thioether-linked Fab’z fragments. J. Immun. 139,2367-2375. Glockshuber R., Malia M., Pfitzinger I. and Pluckthun A. (1990) A comparison of strategies to stabilize immunoglobulin Fv-fragments. Biochemistry. 29, 136221367. Goding J. W. and Burns G. F. (1981) Monoclonal antibody OKT-9 recognizes the receptor for transferrin on human acute lymphocytic leukemia cells. J. Immun. 127, 12561258. Gootenberg J. E., Ruscetti F. W., Mier J. W., Gazdar A. and Gallo R. C. (1981) Human cutaneous T cell lymphoma and leukemia cell lines produce and respond to T cell growth factor. J. P.Y~. Med. 154, 140331418. Gruber M., Schodin B. A., Wilson E. R. and Kranz D. M. (1994) Efficient tumor cell lysis mediated by a bispecific single chain antibody expressed in Escherichia coli. J. Immun. 152, 5368-5374.
Hayden M. S., Linsley P. S., Gayle M. A., Bajorath J., Brady W. A., Norris N. A., Fell H. P.. Ledbetter J. A. and Gilliland L. K. (1994) Single-chain mono- and bispecific antibody derivatives with novel biological properties and antitumour activity from a COS cell transient expression system. Therapeut. Immun. 1, 3315. Hedrick S. M., Matis L. A.. Hecht T. T., Samelson L. E., Longo D. L., Heber-Katz E. and Schwartz R. H. (1982) The fine specificity of antigen and Ia determinant recognition by T cell hybridoma clones specific for pigeon cytochrome c. Cell 30, 141-I 52. Holliger P.. Prosper0 T. and Winter G. (1993) “Diabodies”: small bivalent and bispecific antibody fragments. Proc. natn. Acad. Sci. U.S.A. 90, 64446448. Holliger P. and Winter G. (1993) Engineering bispecific antibodies. Curr. Opin. Biotech. 4, 44G-449. Huston J. S., Levinson D., Mudgett-Hunter M., Tai M. S., Novotny J., Margolies M. N., Ridge R. J., Bruccoleri R. E., Haber E., Crea R. and Oppermann H. (1988) Protein engineering of antibody binding sites: recovery of specific activity in an anti-digoxin single-chain Fv analogue produced in Escherichia coli. Proc. natn. Acad. Sci. U.S.A. 85, 58795883.
Huston J. S., McCartney J., Tai M.-S., Mottola-Hartshorn C., Jin D., Warren F., Keck P. and Oppermann H. (1993) Medical applications of single-chain antibodies. Int. Rev. Immun. 10, 195-217.
Recombinant
single-chain
bispecific
Jost C. R., Kurucz I., Jacobus C. M., Titus J. A., George A. J. T. and Segal D. M. (1994) Mammalian expression and secretion of functional single-chain Fv molecules. J. biol. Chem. 269,26261-26213. Kostelny S. A., Cole M. S. and Tso J. Y. (1992) Formation of a bispecific antibody by the use of leucine zippers. J. I~n~lmzm. 148, 1541-l 553. Kroesen B. J., Ter Haar A.: Spakman H., Willemse P., Sleijfer D. T., De Vries E. G. E., Mulder N. H., Berendsen H. H., Limburg P. C.. The T. H. and De Leij L. (1993) Local antitumour treatment in carcinoma patients with bispecificmonoclonal-antibody-redirected T cells. Crtnceu Immun. Immunother. 37, 4OWOl. Kurucz I., Titus J. A., Jost C. R., Jacobus C. M. and Segal D. M. (1995) Retargeting of C’TL by an efficiently refolded bispecilic single-chain Fv dimer produced in bacteria. J. Immun. 154,4576-4582. Lanzavecchia A. and Scheidegger D. (1981) The use of hybrid hybridomas to target human cytotoxic T lymphocytes. Eur. J. Immuti. 17, 105-I II. Leo O., Foo M., Sachs D. H., Samelson L. E. and Bluestone J. A. (1981) Identification of a monoclonal antibody specific for a murine T3 polypeptide. Proc. natn. Acad. Sci. C’.S.A. 84, 1314-l 318. Mallender W. D. and Voss E. W. Jr (1994) Construction, expression, and activity of a bivalent bispecific single-chain antibody. J. hiol. C’hem. 269, 199-206. Mezzanzanica D., Garrido M. A., Neblock D. S., Daddona P. E., Andrew S. M.. Zurawski V. R., Segal D. M. and Wunderlich J. R. (1991) Human T-lymphocytes targeted against an established ovarian carcinoma with bispecific F(ab’), antibody prolong host survival in a murine xenograft model. C~mc.er.Res. 51, 5116~5121, Nicholls P. J.. Johnson V. G., Blanford M. D. and Andrew S. M. (1993) An improved method for generating single-chain antibodies from hybridomas. .I. immunol. Meth. 165, 8 I-91. Nitta T., Sato K., Yagita H.. Okumura K. and Ishii S. (1990) Preliminary trial of specific targeting therapy against malignant glioma. Lutwr 335, 368~~37 1. Perez P.. HolTman R. W.. Shaw S., Bluestone J. A. and Segal D. M. (1985) Specific targeting of cytotoxic T cells by antiT3 linked to anti-target cell antibody. Nature 316, 354-356. Perez P.. Hoffman R. W., Titus J. A. and Segal D. M. (1986) Specific targeting of human peripheral blood T cells by heteroaggregates containing anti-T3 crosslinked to antitarget cell antibodies. J. exp. Med. 163, 166-118. Reiter Y.. Brinkmann U., Kreitman R. J., Jung S. H., Lee B. and Pastan I. (1994) Stabilization of the Fv fragments in recombinant immunotoxins by disulfide bonds engineered into conserved framework regions. Biochemistr~~ 33, 545 I5459.
Renner C., Jung W.. Sahin U., Denfeld R., Pohl C., Trumper L.. Hartmann F., Diehl V., Van Lier R. and Pfreundschuh M. (1994) Cure of xenografted human tumors by bispecific monoclonal antibodies and human T cells. Science 264,833835.
Sachs D. H., Mayer N. and Ozato K. (1981) Hybridoma antibodies directed toward murine H-2 and Ia antigens. In Monorlonul Antibodies and T-cell Hybridomus (Edited by
Fvz redirects lysis by CTL
219
Hammerling J., Hammerling U. and Kearney J. F.), p.95. Elsevier/North-Holland, New York, Schneider C.. Sutherland R., Newman R. and Greaves M. (1982) Structural features of the cell surface receptor for transferrin that is recognized by the monoclonal antibody OKT9. J. hiol. Chetn. 251, 8516-8522. Segal D. M., Titus J. A. and Stephany D. A. (1987) The use of fluorescence flow cytometry for the study of lymphoid cell receptors. In Methods in Etqtnolog~~: Immunological Technigues, Parr K: in vitro Models qf’B and T Cell Fun&on and Lymphoid Cell Receptors (Edited by Di Sabato G.). p. 478. Academic Press, New York. Segal D. M., Urch C. E., George A. J. T. and Jost C. R. (1992) Bispecific antibodies in cancer treatment. In Biologic Therap) o/‘Cuncer Updutes, Vol. 2 (Edited by DeVita V. T., Hellman S. and Rosenberg S. A.), p. 1. J. B. Lippincott, Philadelphia. Segal D. M. (1993) SPDP crosslinking of antibodies to form heteroconjugates mediating redirected cytotoxicity. In CJtotosic. Cells; Recognition, &ffkctor Function, Generution, and Methods (Edited by Sitkovsky M. V. and Henkart P. A.), p. 485. Birkhauser, Boston. Segal D. M., Jost C. R. and George A. J. T. (1993) Targeted cellular cytotoxicity. In Cvtoto.uic Cells: Generation, Rrcognition, IY#kror Funcfions. Method~y (Edited by Sitkovsky M. V. and Henkart P. A.). p. 96. Springer/Verlag, Berlin. Segal D. M. and Snider D. P. (1989) Targeting and activation of cytotoxic lymphocytes. Chem. Immun. 47, 179-2 13. Shalaby M. R.. Shepard H. M., Presta L., Rodrigues M. L., Beverley P. C. L., Feldmann M. and Carter P. (1992) Development of humanized bispecific antibodies reactive with cytotoxic lymphocytes and tumor cells overexpressing the HER2 protooncogene. J. exp. Med. 175,217-225. Staerz U. D., Kanagawa 0. and Bevan M. J. (1985) Hybrid antibodies can target sites for attack by T cells. Nafure 314, 628-631.
Staerz U. D. and Bevan M. J. (1986) Hybrid hybridoma producing a bispecific monoclonal antibody that can focus effector T-cell activity. Proc. ncrtr2. Acad. Sc~i. U.S.A. 83, 14531457.
Unkeless J. C. (1979) Characterization of a monoclonal antibody directed against mouse macrophage and lymphocyte Fc receptors. J. r.\-p. Med. 150, 580-596. Van Dijk J.. Zegveld S. T., Fleuren G. J. and Warnaar S. 0. (1991) Localization of monoclonal antibody G250 and bispecific monoclonal antibody CD3iG250 in human renalcell carcinoma xenografts: relative effects of size and affinity. Inr. J. Cancer 48, 138-143. Weiner G. J., Kostelny S. A., Hillstrom J. R., Cole M. S., Link B. K., Wang S. L. and Tso J. Y. (1994) The role of T cell activation in anti-CD3 X antitumor bispecific antibody therapy. J. Immun. 152, 2385-2392. Weiner G. J. and Hillstrom J. R. (1991) Bispecific anti-idiotype/anti-CD3 antibody therapy of murine B cell lymphoma. J. Immun. 147, 40354044. Wunderlich J. R. and Shearer G. M. (1991) Induction and measurement of cytotoxic T lymphocyte activity, Unit 3. I 1. In Currenf Protocols in Immunology. Vol. 1 (Edited by Coligan J. E.. Kruisbeek A. M., Margulies D. H., Shevach E. M. and Strober W.), p. 3.11.1. John Wiley and Sons, New York.