A recombinant trispecific single‐chain Fv derivative directed against CD123 and CD33 mediates effective elimination of acute myeloid leukaemia cells by dual targeting

Two trivalent constructs consisting of single‐chain Fv antibody fragments (scFvs) specific for the interleukin‐3 receptor α chain (CD123), CD33 and the Fcγ‐receptor III (CD16) were designed and characterized for the elimination of acute myeloid leukaemia (AML) cells. The dual targeting single‐chain Fv triplebody (sctb) [123 × ds16 × 33] and the mono targeting sctb [123 × ds16 × 123] both specifically bound their respective target antigens and were stable in human serum at 37°C for at least 5 d. Both constructs induced potent antibody‐dependent cellular cytotoxicity (ADCC) of two different AML‐derived CD33‐ and CD123 double‐positive cell lines in the low picomolar range using isolated mononuclear cells (MNCs) as effector cells. In these experiments the dual targeting molecule produced significantly stronger lysis than the mono targeting agent. In addition, the sctbs showed a high potency in mediating ADCC of primary leukaemia cells isolated from peripheral blood or bone marrow of seven AML patients. Hence, these novel molecules displayed potent anti‐leukaemic effects against AML cells in vitro and represent attractive candidates for further preclinical development.

Acute myeloid leukaemia (AML) is a haematopoietic malignancy characterized by a block in cell maturation and an accumulation of undifferentiated blasts in the bone marrow and peripheral blood (Lowenberg et al, 2003;Hiddemann et al, 2005). AML is commonly treated with induction chemotherapy followed by consolidation therapy, and sometimes maintenance therapy Robak & Wierzbowska, 2009). As a post-remission therapy, autologous or allogeneic bone marrow transplantation can be performed in selected patients (Shipley & Butera, 2009). The immunoconjugate Gemtuzumab Ozogamicin (GO) is approved by the US Food and Drug Administration (FDA) for the treatment of patients in first relapse older than 60 years and for patients resistant to standard chemotherapy (Bross et al, 2001;Sievers, 2001). GO is a CD33-specific antibody that is chemically coupled to the cytotoxic drug calicheamicin (Sievers et al, 1999;Stasi, 2008). CD33, a 67-kDa transmembrane cell surface glycoprotein specific for the myeloid lineage, is not expressed on non-haematopoietic tissue but is expressed on a subset of normal haematopoietic stem cells (HSCs) and on AML leukaemia stem cells (AML-LSCs;Freeman et al, 1995;Taussig et al, 2005;Hauswirth et al, 2007). In a phase II clinical trial, 30% of relapsed AML patients responded to GO, but severe side-effects were observed (Sievers, 2001;Larson et al, 2002). In a recent European multicentre phase III study, GO was found to provide no benefit when administered as a postremission therapeutic agent to older patients (Lowenberg et al, 2010), and the drug is not approved in Europe. Furthermore, the agent displayed antigen-independent cytotoxicities towards CD33 negative cell lines (Bross et al, 2001;Jedema et al, 2004;Schwemmlein et al, 2006). Despite these problems, GO clearly produces clinical benefits for a subgroup of AML patients, and CD33 is a validated target antigen. However, new therapeutic modalities with fewer toxicities are urgently needed for the treatment of AML (Estey & Dohner, 2006). Another important AML-associated antigen is CD123, the alpha subunit of the interleukin-3 receptor (IL-3Ra), which is predominantly expressed on myeloid cells and on a subpopulation of B-lymphocytes. It is not expressed on platelets, red blood cells, natural killer (NK) cells and peripheral T cells (Moretti et al, 2001), and only in low density on HSCs (Huang et al, 1999;Jin et al, 2009). However, CD123 has been detected in a variety of haematopoietic malignancies and on AML-LSCs (Jordan et al, 2000;Munoz et al, 2001;Testa et al, 2002;Taussig et al, 2005). Strikingly, CD123 shows a fourfold increased expression on AML-LSCs over normal HSCs (Jin et al, 2009). This finding, together with the clinical observations that elevated expression of CD123 in AML is associated with higher blast counts at diagnosis and a lower complete remission rate resulting in poorer prognosis (Testa et al, 2002(Testa et al, , 2004Graf et al, 2004), make CD123 a particularly interesting target for antibody-derived therapeutics (Jin et al, 2009).
AML-LSCs are known to possess several remarkable properties including self-renewal potential and increased resistance against chemotherapeutics and DNA damage (Lapidot et al, 1994;Bonnet & Dick, 1997;Guan & Hogge, 2000;Guzman et al, 2001;Hope et al, 2004;Ishikawa et al, 2007;Dick, 2008). This unique cell population is thought to play a key role in the frequent occurrence of minimal residual disease (MRD) in relapsed AML patients after conventional chemotherapy (van Rhenen et al, 2005;Ravandi & Estrov, 2006). Hence, following the discovery of AML-LSCs, a great amount of effort has been spent to investigate the surface antigen profile of this clinically relevant cell population to identify suitable markers for their targeted elimination (Jordan et al, 2000;Taussig et al, 2005;Jin et al, 2006Jin et al, , 2009Hosen et al, 2007). The fact that both antigens, CD33 and CD123, are expressed on AML-LSCs offers novel prospects for the treatment of AML. So far an immunotoxin (Du et al, 2007), a fusion of IL-3 with a truncated version of diphtheria toxin (Feuring-Buske et al, 2002;Frankel et al, 2008), a neutralizing full-length antibody (7G3; Jin et al, 2009), and a bispecific single-chain Fv (bsscFv; Stein et al, 2010) directed against CD123 have been described, and the monoclonal antibody (mAb) 7G3 has been tested in a phase I clinical study (Roberts et al, 2008; http://clinicaltrials.gov/ ct2/show/NCT00401739?term=CSL360&rank=1). Well-studied antibody-derived agents directed against CD33 include, apart from GO an immunotoxin (Schwemmlein et al, 2006), a fusion protein between a single chain Fv antibody fragment (scFv) and the sTRAIL death receptor ligand (ten Cate et al, 2009), and siRNA-loaded liposomes coated with CD33directed scFvs (Rothdiener et al, 2010).
BsscFvs have advantageous properties over full-length bispecific antibodies and chemically coupled bispecific Fragment antibody binding (Fab) fragments, and have recently produced convincing therapeutic effects in clinical trials (Bargou et al, 2008). To overcome some of the remaining weaknesses of this format, such as an unfavourable plasma retention, the format has been further improved by expanding it to tandem diabodies (Kipriyanov et al, 1999) and singlechain triplebodies (sctbs; Kellner et al, 2008). The incorporation of a second scFv-binding site for tumour-antigens in these extended formats has led to increased avidity for the tumour cell, increased anti-tumour activity in antibody-dependent cellular cytotoxicity (ADCC) reactions, and improved plasma retention in vivo (Kellner et al, 2008). An example of a sctb with improved ADCC activity for AML cells is the sctb [33 · ds16 · 33] (Singer et al, 2010), with two scFv antigen binding sites for CD33 on AML cells and one for CD16, the low affinity Fcc RIII receptor for IgG on NK cells and macrophages (Ravetch & Perussia, 1989;Daeron, 1997). In a direct comparison, this sctb showed increased anti-leukaemic activity for AML cells over the corresponding bsscFv [33 · ds16] with only one binding site each for CD33 and CD16. Before these trivalent single-polypeptide formats, other trispecific antibody derivatives had been explored. One example was a trispecific Fab construct, consisting of three chemically coupled Fabfragments (Somasundaram et al, 1999). A second trispecific construct was called a 'tri-body', a two-chain polypeptide based on a Fab fragment as scaffold (Schoonjans et al, 2001). Two scFvs were fused to the C-termini of the L-and Fd-chains of this Fab, respectively. Finally, a trispecific molecule based on a VH domain has been reported, which carried two scFvs fused to a VH-domain (Song et al, 2003).
The key molecule presented here, the sctb [123 · ds16 · 33], differs significantly in its design from all of these examples. It is a single-chain polypeptide with only one binding site for an effector cell and two for two different tumour antigens, CD123 and CD33. The goal of this design was to achieve improved anti-leukaemic activity by 'dual targeting' of the tumour cell, i.e. by simultaneous binding to two different antigens on the same tumour cell. We expected this molecule to bind with greater avidity to CD123 and CD33 on doublepositive AML cells than the corresponding control sctbs [123 · ds16 · 123] and [33 · ds16 · 33], because it would take advantage of two populations of target antigens rather than only one. Based on theoretical arguments, the avidity of such proteins was expected to be propotional to the combined target antigen density on the tumour cell (Mattes, 1997;Pluckthun & Pack, 1997).
The prediction of added benefits by dual-over monospecific targeting of tumour cells is intuitively appealing, and has been pursued for a number of years (Robinson et al, 2008), but has not yet been conclusively proven for scFv-derived agents recruiting effector cells. A definitive proof of this prediction is experimentally demanding and is still under investigation. However, indirect evidence in favour of the proposition has been obtained for scFv-immunotoxins, and therefore, the concept may also extend to agents recruiting effector cells (Vallera et al, 2005). The molecules presented here provide an opportunity to directly evaluate the benefits of dual over mono-specific targeting for AML cells. Finally, they provide new means to test the prediction that dual targeting a combination of two antigens, which are more highly expressed on AML-LSCs than on HSCs, may permit the preferential elimination of AML-LSCs. It would be important to test this prediction because, if confirmed, it would provide a possibility for a deliberate attack on the LSCs with the intent to eliminate MRD cells and prevent relapse, while preserving some HSCs for haematopoietic reconstitution after the end of treatment, without the need for autologous stem cell transplantation. The molecules presented here provide first steps towards these important goals of rational leukaemia therapy.
Construction of sctbs [123 · ds16 · 33] and [123 · ds16 · 123] Escherichia coli strain XL-1 blue (Stratagene, Amsterdam, The Netherlands) was used as the host for the amplification of plasmids and cloning. For the construction of the sctb [123 · ds16 · 33] expression vector the sequence coding for the CD123-specific scFv was excised from the vector pAK400-CD123scFv (Stein et al, 2010) and cloned as a SfiI-cassette into the existing vector pSecTag2HygroC-STREP-His-CD33 · dsCD16 · CD33 (Singer et al, 2010) replacing the coding sequence for the N-terminal CD33-specific scFv. Thereby the vector pSecTag2HygroC-STREP-His-CD123 · dsCD16 · CD33 was produced. In these constructs, the index ds designates the disulfide-stabilized variants (Bruenke et al, 2004). To generate the vector pSecTag2HygroC-STREP-His-CD123 · dsCD16 · CD123, the sequence coding for the CD123-specific scFv was amplified by polymerase chain reaction (PCR) from the vector pAK400-CD123scFv and ligated into pSecTag2-HygroC-STREP-His-CD123 · dsCD16 · CD33, using XhoI/ EcoRV restriction sites, and replacing the coding sequence for the C-terminal CD33-specific scFv. For construction of the expression vector pSecTag2HygroC-hCD123ex-DsRed, cDNA coding for the extracellular domain of CD123 was amplified by PCR using the existing vector pSecTag2HygroC-hCD123ex-Fc (Stein et al, 2010) and ligated into the vector pSecTag2HygroC-CD33ex-DsRed (Singer et al, 2010), replacing the extracellular domain of CD33. Correct construction of the final constructs was confirmed by DNA sequence analysis on an Applied Biosystems automated DNA sequencer (ABI Prism 310 Genetic Analyser; Perkin-Elmer, Ueberlingen, Germany).

Expression and purification of sctbs [123
For expression of the recombinant sctbs [123 · ds16 · 33] and [123 · ds16 · 123], and the control sctbs ds[19 · 16 · 19] and [7 · ds16 · 7] specific for the cell surface antigens CD 19 and CD7, respectively, 293T cells were stably transfected with the respective expression vectors (Kellner et al, 2008). Cells were cultured under permanent selection with hygromycin C in a mini-PERM bioreactor (Greiner Bio-One, Frickenhausen, Germany) with a dialysis membrane with a 12AE5 kDa cutoff following the manufacturer's instructions. The culture supernatants containing the recombinant protein were collected four times over a period of 2 weeks. For expression of the control sctb [33 · 64 · 33] specific for the cell surface antigens CD33 and CD64 (Stein, unpublished data), 293T cells were transiently transfected with the expression vector using the calcium phosphate technique including 5 mmol/l chloroquine (Sambrook et al, 1989). The transfection medium was replaced by fresh culture medium after 10 h and, for 5 d, supernatants were collected every day and combined. Supernatants were analysed for the presence of antibody fragments by flow cytometry. The recombinant His-tagged proteins were enriched by affinity chromatography using nickel-nitrilotriacetic acid agarose (NTA) beads (Qiagen, Hilden, Germany) and dialysed against phosphate-buffered saline (PBS). Fusion proteins with green or red fluorescence protein were transiently expressed in 293T cells and purified as described above.

Sodium dodecyl sulphate polyacrylamide gel electrophoresis and Western blot analysis
Eluted proteins were analysed by reducing sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) using standard procedures (Laemmli, 1970). Gels were stained with Coomassie brilliant blue R250 (Sigma-Aldrich, Taufkirchen, Germany). In Western blot experiments, recombinant proteins were detected with an unconjugated penta-His antibody (Qiagen) and a secondary horseradish peroxidase-coupled goat anti-mouse IgG antibody (Dianova, Hamburg, Germany). Western Blots were developed using enhanced chemiluminescence reagents (Amersham Pharmacia Biotech, Freiburg, Germany).

Flow cytometry analysis
Immunofluorescence analysis was performed on a FACSCalibur instrument using CellQuest software (Becton Dickinson, Heidelberg, Germany) as described (Schwemmlein et al, 2006). Briefly, 1 · 10 4 events were collected for each sample, and whole cells were analysed using appropriate scatter gates to exclude cellular debris and aggregates. The recombinant sctb proteins were detected using a penta-His antibody and a phycoerythrin-conjugated goat anti-mouse-IgG F(ab') 2 (DAKO Diagnostica GmbH, Hamburg, Germany), unless otherwise stated.

Determination of equilibrium binding constants (K D )
K D values were determined by calibrated flow cytometry as described (Benedict et al, 1997). The highest mean fluorescence value was set to 100%, and all data points were normalized to this value. The experiments were repeated six times. The K D values were calculated using a nonlinear regression curve fit.

Measurement of in vitro serum stability
To determine the in vitro serum stability of sctbs [123 · ds16 · 33] and [123 · ds16 · 123], aliquots of the recombinant proteins were incubated at a sub-saturating concentration of 2AE5 lg/ml at 37°C (day 5) in human serum or stored at )20°C and thawed at defined time points (day 0-4). The residual binding activity was measured by flow cytometry on day 0 (as described above). The experiment was repeated at least four times and results were fitted to a monoexponential decay.

Isolation of mononuclear cells (MNCs) and AML cells from human donors
Citrate buffered or heparinized peripheral blood from both healthy volunteers and AML patients or bone marrow from AML patients were obtained after receiving informed consent and with the approval of the Ethics Committee of the University of Erlangen-Nuremberg. MNCs were enriched by Lymphoflot (Biotest, Dreieich, Germany) Ficoll density centrifugation in Leukosep tubes (Greiner, Frickenhausen, Germany) according to the manufacturer's instructions, and suspended in RPMI 1640 Glutamax-I medium containing 10% FCS, 100 units/ml penicillin and 100 mg/ml streptomycin. Viability was verified by Trypan blue exclusion and exceeded 95%.

ADCC reactions
ADCC assays, using MNCs from healthy donors as effector cells, were performed in triplicates using a 3-h 51 Cr release assay as described (Elsasser et al, 1996). For blocking experiments, the parental antibody 3G8, the IgG1 isotype control, the CD123-and CD33-specific scFvs and the control scFv were added at a 125-and 500-fold molar excess, respectively. Doseresponse curves were recorded using several equimolar fivefold serial dilutions of the respective antibody fragments at a constant effector-to-target-cell (E:T) ratio of 40:1 MNCs to target cells. Background lysis induced by MNCs alone was subtracted from each data point, and EC 50 values (concentration of an antibody fragment producing 50% of maximum specific lysis) and maximal lysis were calculated by using a sigmoidal dose-response curve fit. The experiments were repeated six times and mean values are reported, unless otherwise stated.

ADCC reactions with enriched NK cells
NK cells were enriched from the MNC fraction using a NK cell isolation Kit (Miltenyi, Bergisch Gladbach Germany) and immunomagnetic bead (MACS) technology following the manufacturer's protocols. Purities ranged from 83% to 98% as judged by 2-colour flow cytometry using fluorescein isothiocyanate-coupled CD16-specific antibodies (Beckman Coulter, Brea, CA, USA) and PC7-conjugated CD56-specific antibodies (BD Biosciences, Heidelberg, Germany) according to the manufacturer's instructions. Flow cytometric analysis was performed on a FC 500 flow cytometer (Beckman Coulter, Brea, CA, USA). Isolated NK cells were cultured overnight at a density of 2 · 10 6 cells/ml in RPMI 1640 Glutamax-I medium (Invitrogen) containing 10% FCS (Invitrogen), 100 units/ml penicillin (Invitrogen), and 100 lg/ml streptomycin (Invitrogen). Cytotoxicity reactions were performed at an E:T cell ratio of 10:1 in a standard 4-h 51 Cr assay as described above. After 4 h of incubation, 25 ll of supernatant was mixed with scintillation solution Supermix (Applied Biosystems, Darmstadt, Germany) and incubated for 15 min with agitation. 51 Crrelease from triplicates was measured in counts per minute (cpm) using the scintillation and luminescence counter 1450 Micro Beta TriLux (Perkin Elmer). Maximal 51 Cr release was determined by adding Triton X-100 (1% final concentration) to target cells. USA) and Microsoft EXCEL. Group data were reported as means ± standard error of the mean (SEM). Differences between groups were analysed using paired Student t-test. P-values <0AE05 were considered significant.
To test whether sctb [123 · ds16 · 33] was able to bind to more than one antigen simultaneously, CD33 single-positive U937 cells were incubated with the sctb and stained in parallel with the recombinant proteins CD123ex-RFP and CD16ex-GFP ( Fig 3B). These proteins contained the extracellular domains of the surface proteins CD123 and CD16, genetically fused to the red and green fluorescence proteins, respectively. As a result, the target cells showed green and red fluorescence, indicating that the sctb mediated binding of both CD123ex-RFP and CD16ex-GFP to the same target cell.
To determine the in vitro serum stability of sctbs [123 · ds16 · 33] and [123 · ds16 · 123], the recombinant proteins were incubated in human serum at 37°C for different lengths of time. Residual binding activity was measured by flow cytometry and ranged between 81% and 94% on day 5 for each scFv on respective single-positive cell lines. Therefore, stability for both molecules for at least 5 d was confirmed (Table I).

ADCC of established AML-derived cell lines mediated by sctbs [123 · ds16 · 33] and [123 · ds16 · 123]
To test the ability of both sctbs to mediate ADCC in vitro, ADCC reactions with two different AML-derived cell lines were performed. Both sctbs [123 · ds16 · 33] and [123 · ds16 · 123] mediated potent lysis of the CD123 + / CD33 + double-positive cell lines MOLM-13 and THP-1 in a dose-dependent manner (Fig 4). Samples were incubated for 3 h in a chromium release assay using unstimulated MNCs from unrelated healthy donors at an E:T ratio of 40:1. For MOLM-13 cells the sctbs [123 · ds16 · 33] and [123 · ds16 · 123] gave rise to EC 50 values in the low picomolar range of 21 and 50 pmol/l, respectively. Maximum specific lysis was c. 46% and c. 37%, respectively (Fig 4A). For THP-1 cells, EC 50  (Fig 5). Co-incubation with a simultaneous 500-fold molar excess of both CD33-and CD123-specific scFvs completely blocked lysis of MOLM-13 cells, whereas competition with either scFv alone reduced target cell lysis only to about half-maximum value. Therefore, both scFvs specific for CD33 and CD123 contained within the triplebody must have contributed to the ADCC reaction. Also, co-incubation with a 125-fold molar excess of the parental CD16-specific mAb 3G8 but not with an IgG1 isotype control antibody (4G7), significantly reduced lysis of MOLM-13 cells, highlighting the specificity in triggering CD16 + effector cells.
ADCC of the established AML-derived cell line MOLM-13 mediated by sctbs [123 · ds16 · 33] and [123 · ds16 · 123] and enriched NK cells To investigate whether the sctbs were able to activate CD16 + effector cells not only within the MNC fraction but also to directly activate purified CD16 + NK cells, NK cells sorted with immunomagnetic beads were used in ADCC reactions ( Fig 4C). Purified NK cells and MOLM-13 cells were used at an E:T ratio of 10:1 with triplebody concentrations of 1 nmol/l in three independent experiments. Both sctbs mediated similar significant lysis (c. 60%) of AML cells compared to a control sctb. No specific lysis was observed when samples were incubated in the absence of NK cells. These results show first, that the sctbs were able to directly activate purified CD16 + NK cells in ADCC reactions; second, that the proteins alone without effector cells had no effect on the target cells; and third, that the protein preparations were therefore free of cytotoxic contaminants. Finally, both sctbs were tested for their ability to mediate ADCC of freshly isolated primary leukaemia cells from AML patients. For these experiments, isolated MNCs from either peripheral blood or bone marrow of seven patients were used in ADCC reactions. Overall, six peripheral blood and two bone marrow samples were studied. MNCs from one unrelated healthy donor per AML patient sample were used as effector cells (Fig 6; Table II. For direct comparison of the potencies of the dual targeting and mono targeting sctbs in ADCC of primary leukaemia cells, the previously published sctb [33 · ds16 · 33] (Singer et al, 2010) was also carried along. All three sctbs showed potent lysis of primary AML cells in a concentrationdependent manner (Fig 6). Four representative ADCC reactions of samples from Patients 1 and 2 (peripheral blood; Fig 6A, B) and Patients 6 and 7 (bone marrow; Fig 6C, D) are depicted. At a concentration of 5 nmol/l, all three recombinant proteins produced potent specific lysis (Fig 6E). In 4/8 samples, the dual targeting sctb [123 · ds16 · 33] reached the highest extent of lysis. This finding was confirmed when ADCC data from all six peripheral blood samples were combined (Fig 6F). The EC 50 values derived from this data set were c. 250, c. 130, and c. 250 pmol/l for the sctbs [123 · ds16 · 33], [123 · ds16 · 123] and [33 · ds16 · 33], respectively. Maximum specific lysis was c. 25%, c. 19% and c. 19% of input cells, respectively. Although the differences in maximum specific lysis between the dual targeting sctb [123 · ds16 · 33] and the mono targeting sctbs [123 · ds16 · 123] and [33 · ds16 · 33] did not reach statistical significance, the dual targeting molecule showed a remarkable tendency to mediate higher specific lysis of primary AML cells than its mono targeting relatives. These results clearly demonstrate that the sctbs can induce potent ADCC of primary AML cells of different AML types in vitro, and point to small but distinct advantages for the dual targeting agent over the corresponding mono targeting agents.

Discussion
Here we describe the generation and characterization of two new antibody derivatives designed for use against AML cells.
Although both sctbs are trivalent, the first, [123 · ds16 · 33] represents a 'dual targeting' molecule, whereas the second, [123 · ds16 · 123], is a bispecific molecule with only one specificity for the tumour cell which we refer to as a 'mono specific' or 'mono targeting' agent. To our knowledge, the sctb [123 · ds16 · 33] is the first 'dual targeting' single-chain triplebody simultaneously addressing two different antigens on a tumour cell to be reported. The sctb [123 · ds16 · 123] is an improvement over the existing bsscFv [123 · ds16] (Stein et al, 2010), in which the recently gained knowledge, that addition of a second scFv binding site for a tumour antigen leads to increased anti-tumour activity (Kellner et al, 2008), has been incorporated. Both recombinant proteins were produced in stably transfected eukaryotic cells with expression yields ranging from 1AE2 to 7AE4 mg (sctb [123 · ds 16 · 33]) and 1AE7-2AE8 mg (sctb [123 · ds16 · 123]) per litre of cell culture medium. Both molecules bound specifically to antigen-positive cells. Furthermore, the sctb [123 · ds16 · 33] was able to bind to more than one antigen simultaneously, an essential prerequisite for the recruitment of CD16-positive effector cells. This also highlights the fact that three scFvs in tandem in one polypeptide-chain can produce a functional protein. Equilibrium binding constants on single-positive cells were approximately 22 nmol/l (CD16), 18 nmol/l (CD33), 19 nmol/l (CD123) for the individual components carried in the sctbs, induced potent ADCC of MOLM-13 cells at a concentration of 1 nmol/l. Simultaneous addition of a 500-fold molar excess of CD33-and CD123specific scFvs each completely blocked the ADCC reaction, but not the addition of either CD33-or CD123-specific scFv alone, or a control scFv. Simultaneous incubation with a 125-fold molar excess of 3G8 antibody, but not with a control IgG1 (4G7), significantly reduced ADCC. Data points represent mean percentage of relative specific lysis obtained with isolated MNCs from six different healthy donors at an E:T ratio of 40:1. Specific lysis measured for the sctb [123 · ds16 · 33] was defined as 100%, lysis with no antibody (Ab) was defined as 0%. Specific lysis is total lysis minus spontaneous lysis. *Statistically significant differences (P < 0AE05) in ADCC relative to the control sctb.  (Kellner et al, 2008;Singer et al, 2010). This may be due to differences in antigen density for CD33 and CD123 on double-positive cells, to differences in accessibility of both antigens or their spatial segregation into different subdomains. However, the result clearly indicates that both scFvs against CD33 and CD123 carried by the dual targeting agent participate in binding to the respective antigens on double-positive cells. The K D values of the mono targeting sctb [123 · ds16 · 123] were approximately 25 and 6 nmol/l for the CD16 and CD123 scFvs, respectively. This highlights a threefold gain in affinity for CD123 compared to the dual targeting sctb [123 · ds16 · 33], which must be due to an avidity effect caused by the addition of a second CD123-specific scFv in the mono targeting protein. This result is a further clear indication that both antigens actively participate in binding to the target cell, confirming earlier findings for this molecular format (Kellner et al, 2008;Singer et al, 2010). Another important parameter for antibody-derivatives is stability. When administered into the blood stream as therapeutic agents, such molecules need to be stable ] (open triangle) mediated dose-dependent ADCC of primary AML cells, whereas a control sctb (closed square) failed to induce cellular lysis. (A and B) Induction of ADCC by sctbs of purified primary AML cells isolated from peripheral blood (Patients 1 and 2). (C and D) Induction of ADCC by sctbs of purified primary AML cells isolated from bone marrow (Patients 6 and 7). Data points represent percentage of specific lysis obtained with isolated MNCs from one healthy donor at an E:T ratio of 40:1. (E) Induction of ADCC by sctbs at a concentration of 5 nmol/l for all eight samples. For Patients 1-6 the cells analysed were MNCs isolated from peripheral blood. For Patients 6 and 7, bone marrow-derived cells were studied. White bars: control sctb; black bars: sctb [123 · ds16 · 33]; light grey bars: sctb [123 · ds16 · 123]; dark grey bars: sctb [33 · ds16 · 33]. (F) Induction of ADCC by sctbs of purified AML cells from peripheral blood of six different patients (Patients 1-6), combined data. Data points represent mean percentage of specific lysis averaged over the six patients obtained with isolated MNCs from one healthy donor per patient sample at an E:T ratio of 40:1. Specific lysis is total lysis minus spontaneous lysis. against degradation and unfolding (Carter, 2006). Both proteins have been tested and were stable in human serum at 37°C for at least 5 d. Stability tests in immunocompetent mice similar to those previously reported for the sctb ds[19 · 16 · 19] (Kellner et al, 2008) are planned. Both sctbs induced potent ADCC of the AML-derived, CD123/CD33 double-positive cell lines MOLM-13 and THP-1. For several concentrations the dual targeting sctb [123 · ds16 · 33] revealed a significantly higher lysis than the mono targeting sctb [123 · ds16 · 123]. This interesting result is probably due to the fact that a greater combined antigen density was accessible to the dual targeting agent and hence the molecule showed a prolonged opsonization of the tumour cell. This probably leads to an increased probability to recruit effector cells. Competition experiments revealed that cell lysis by sctb [123 · ds16 · 33] was strictly antigen-specific. Coincubation of the sctb with an excess of either anti-CD33 or anti-CD123 scFvs separately reduced cell lysis to about 50-60%. Remarkably, simultaneous co-incubation with both scFvs completely abolished cell lysis, indicating that both binding moieties played an active role not only in binding to the tumour cell but also in mediating ADCC. The same is true for CD16. Co-incubation with the CD16 mAb 3G8 reduced the measured cell lysis to such an extent that it was no longer statistically significant compared to the lysis obtained with a control sctb. In addition, both sctbs were able to induce potent ADCC with purified CD16 + NK cells as effector cells at an effector-to-target ratio of 10:1. Furthermore, cell lysis due to binding of the proteins to the target cells alone in the absence of effector cells and lysis due to toxic contaminants in the protein preparations were ruled out in these experiments by omission of the effector cells.
Finally, the sctbs also induced potent ADCC of primary cells from peripheral blood or bone marrow of AML patients. For direct comparison of the ability to mediate ADCC of primary cells, the existing mono targeting sctb [33 · ds16 · 33] (Singer et al, 2010) was included in these experiments. Although not statistically significant, the dual targeting sctb [123 · ds16 · 33] induced the highest maximum lysis averaged over all patients. The isolated cells originated from a variety of AML subtypes according to FAB-and WHOclassifications. These samples showed a high variability in the response to the sctbs (Fig 6A-E). This is most likely due to the highly heterogenous nature of the AML, with many subtypes characterized by different genomic alterations, different disease phenotypes, disease progression and responses to treatment . Therefore, a broad variability both in the fraction of total cells expressing CD33 and CD123 and cells susceptible to ADCC lysis mediated by our recombinant proteins was to be expected and had also been previously observed for the susceptibility to the CD33-ETA' immunotoxin (Schwemmlein et al, 2006). It would be interesting in the future to study whether the CD34 + CD38 ) compartment of primary AML-cells, obtained by preparative sorting, which contains the AML-LSCs, shows a stronger extent of ADCC-lysis than other compartments devoid of AML-LSCs. Until now, these experiments could not be performed due to limited availability of primary patient cells.
Recent studies gave rise to the concept that CD16-positive monocytes/macrophages may play a vital role as effector cells in vivo mediating anti-tumour effects of therapeutic antibodies (Uchida et al, 2004;Tedder et al, 2006). In particular, their role in the elimination of AML cells by phagocytosis has been extensively studied Majeti et al, 2009). In our experiments, CD16-positive NK cells represented the main effector cell population in purified MNCs, a leucocyte fraction devoid of monocytes/macrophages. It will be an important question for the future to determine, whether our triplebodies are able to activate not only NK cells by monovalent binding to CD16 but also CD16-positive macrophages. Current studies in our group address this question.
The novel sctbs [123 · ds16 · 33] and [123 · ds16 · 123] described in this study represent promising therapeutic molecules for the treatment of AML in the future. In particular, the dual-targeting sctb [123 · ds16 · 33] might become valuable for the targeted elimination of AML-LSCs. Its novel combination of two scFvs directed against antigens present on AML-LSCs at greater densities than on normal HSCs offers the potential for a preferential targeting of this important cell type in vivo. Future studies will need to answer the question whether preferential targeting indeed is possible in vivo.