The human natural killer cytotoxic cell line NK-92, once armed with a murine CD16 receptor, represents a convenient cellular tool for the screening of mouse mAbs according to their ADCC potential

To take advantage of the large number of well-characterized mouse immunoglobulins (IgGs) for the study of antibody-dependent cell-mediated cytotoxicity (ADCC) in human cells, we armed human cytotoxic lymphocytes with a mouse receptor for the Fc portion of IgG antibodies. The human ΝΚ−92 natural killer cell line was transduced with a mouse receptor gene (mCD16), which was stably expressed on the cell surface (referred to as NK-92mCD16). When tested against a B-lymphoblastoid cell line (BLCL) coated with mouse anti-CD20 IgG1, IgG2a or IgG2b monoclonal antibodies (mAbs), the newly expressed mouse Fc receptor enabled the NK-92mCD16 cells to kill the BLCL by ADCC. Next, using the NK-92mCD16 we compared mouse mAbs directed at B lineage specific CD antigens for their ability to induce ADCC against human Epstein-Barr virus- infected B lymphoblastoid (for anti-CD19, -CD20 and -CD21) or against myeloma (for anti-CD38 and –CD138) target cells. Our results demonstrated that the “NK-92mCD16 assay” allows convenient and sensitive discrimination of mouse mAbs for their ability to mediate ADCC in a human cellular system. In addition, our results provide examples of dissociation between opsonization and target cell killing through ADCC. These “murinized” human effector cells thus represent a convenient cellular tool for the study of ADCC.


Introduction
Antibody-dependent cell-mediated cytotoxicity (ADCC) is one of the mechanisms by which therapeutic antibodies achieve clinical efficacy. This mechanism combines humoral immunity, which involves specific antigen (Ag) recognition by an antibody (Ab), with cellular immunity, which involves cellmediated cytolytic destruction of Ab-coated target cells. While the specificity of target cell recognition resides within the Fab portion of the Ab molecule, ADCC occurs upon the interaction between the Fc portion of the target cell-bound Ab and the Fc receptors (FcR) expressed by effector cells, such as FcγRIIIA/ CD16A, which recruit and activate effector cells.
In the context of ADCC-mediated tumor cell lysis, Fabdependent specificity is essential for tumor cell discrimination (and consequently low toxicity), while Fc-dependent effector recruitment is essential for tumor cell killing. An ideal therapeutic Ab would be tumor-specific; however, most of the Ags that are currently targeted in clinical practice are tumor-associated rather than tumor-specific. In addition, because a particular to take advantage of the large number of well-characterized mouse immunoglobulins (IgGs) for the study of antibodydependent cell-mediated cytotoxicity (ADCC) in human cells, we armed human cytotoxic lymphocytes with a mouse receptor for the Fc portion of IgG antibodies. the human NK-92 natural killer cell line was transduced with a mouse receptor gene (mCD16), which was stably expressed on the cell surface (referred to as NK-92 mCD16 ). When tested against a B-lymphoblastoid cell line (BLCL) coated with mouse anti-CD20 IgG1, IgG2a or IgG2b monoclonal antibodies (mAbs), the newly expressed mouse Fc receptor enabled the NK-92 mCD16 cells to kill the BLCL by ADCC. Next, using the NK-92 mCD16 we compared mouse mAbs directed at B lineage specific CD antigens for their ability to induce ADCC against human epstein-Barr virus-infected B lymphoblastoid (for anti-CD19, -CD20 and -CD21) or against myeloma (for anti-CD38 and -CD138) target cells. our results demonstrated that the "NK-92 mCD16 assay" allows convenient and sensitive discrimination of mouse mAbs for their ability to mediate ADCC in a human cellular system. In addition, our results provide examples of dissociation between opsonization and target cell killing through ADCC. these "murinized" human effector cells thus represent a convenient cellular tool for the study of ADCC. Ag may be adequately tumor-associated, but not expressed by the entire tumor cell population, two or more tumor-associated Ags may be considered targets to improve tumor cell killing. ADCC depends not only on the Ag/Ab and the FcR/Fc affinities, but also on the access of the FcR to the Fc once the Ab is associated with the tumor Ag. Thus, at least two levels of Ab screening could be considered a priori: first, to identify an Ag; and second, to identify the best epitope to be targeted on this particular Ag. Indeed, over 25 y ago, ADCC by effector human lymphocytes was suggested to be "apparently sensitive to spatial orientation and organization of target cell-bound Ab." 1 Accordingly, Fc accessibility for the FcR and its consequences on ADCC efficiency may be different for each Ag, depending on the epitope that is recognized. Thus, to optimize tumor cell destruction through ADCC, the monoclonal antibody (mAb) that allows for the best effector cell activation should be chosen. While considerable technological efforts have been made to assess ADCC optimization through Fc modifications, no straightforward technique has been identified to associate epitope specificity and ADCC performance against a particular Ag.
Generation of the hCD16 (NK-92 hCD16 )-and the mCD16 (NK-92 mCD16 )-transduced NK-92 cell line. Amphotropic retroviral vector particles were produced by the transfection of Phoenix-Ampho packaging cells and used for the transduction of the human natural killer (NK) cell line NK-92, which does not express human CD16. 5 Five days after transduction, 35% of the NK-92 hCD16 and 41% of the NK-92 mCD16 cells expressed the hCD16 and mCD16 receptors, respectively, as assessed by flow cytometry (Fig. 2). After immuno-selection, the expression of both transgenes remained stable for the 3 mo of follow-up (data not shown).
ADCC by NK-92 hCD16 and NK-92 mCD16 cells. Before assessing the ADCC activity, we first examined whether the expression of the mCD16 or hCD16 receptors had any effect on the natural (CD16-independent) cytotoxic activity of the NK-92 cell line. As shown in Figure 3A, no significant differences were observed between the NK-92, NK-92 hCD16 and NK-92 mCD16 cells in their ability to kill the NK-sensitive K562 cell line. Next, the ADCC activity of the NK-92, NK-92 hCD16 and NK-92 mCD16 cells was evaluated using a standard 4 h 51 Cr release assay against a human CD20-positive Epstein-Barr virus (EBV)-infected B lymphoblastoid cell line (BLCL) in the presence of increasing concentrations of mouse anti-CD20 IgG1 (clone AT80) (Fig. 3B, left panel) or the chimeric anti-CD20 rituximab (Fig. 3B, right panel).
As shown in Figure 3B (left panel), only the NK-92 mCD16 cells killed the human BLCL incubated with the mouse anti-CD20 Ab. By contrast, both the NK-92 hCD16 and NK-92 mCD16 cells killed the human BLCL that was pre-incubated with the chimeric anti-CD20 Ab (Fig. 3B, right panel). Although these data demonstrate that mouse CD16 can perform ADCC in the presence of rituximab, the complete superposition of the two curves is fortuitous because the expression of human and mouse CD16 by transduced NK-92 cells and the affinity of rituximab for these two Fc receptors are likely dissimilar. All the dose responses reached a plateau at the Ab concentration of 10 -1 μg/ ml.
Mouse isotype and NK-92 mCD16 -mediated ADCC. We next compared the NK-92 mCD16 -mediated ADCC in the presence of mouse IgG1, IgG2a and IgG2b anti-human CD20 mAbs ( Table 1), which recognize closely related epitopes in the large extracellular loop of CD20. 6,7 The three mAbs mediated ADCC against the EBV-LCL in the presence of NK-92 mCD16 , with the following lysis ranking by isotype: IgG2a ≥ IgG2b > IgG1 (Fig. 4). The minimum concentration of Ab able to induce detectable ADCC in our assay was between 10 -4 and 10 -3 μg/ml for IgG2a and IgG2b and between 10 -3 and 10 -2 μg/ml for IgG1. In the presence of the NK-92 hCD16 effectors, only IgG2a anti-CD20 induced detectable ADCC against the EBV-LCL. The cytotoxic score in the latter case, however, was almost negligible compared with that observed in the presence of the NK-92 mCD16 effectors cells, i.e., those armed with the cognate (mCD16) FcγR.
For this purpose it would be advantageous to be able to test in an effector/target human system the currently available mouse mAbs and those that are newly produced by hybridomas, a technology more available than animals that are humanized for the immunoglobulin locus. To this end, we describe here the production and characterization of human cytotoxic lymphocytes armed with a mouse FcγR and show how these "murinized" human effector cells can become useful cellular tools to analyze the ADCC potential of mouse Abs. Moreover, using this approach, we found that the ADCC-mediated lysis of a given target cell opsonized to the same extent by mAbs directed to different Ag can be dramatically different, demonstrating that opsonization is necessary, but not sufficient, to induce ADCC.

Results
Human FcγRIIIA-human FcεRIγ (hCD16A) and mouse FcγRIII-human FcεRIγ chimeric (mCD16) vectors. The chimeric cDNA coding for human FcγRIIIA-V158 linked to human FcεRIγ (hCD16) has been described previously. 2 The chimeric cDNA coding for mouse FcγRIII linked to human FcεRIγ (mCD16) was synthesized, and it comprised the extracellular domain of the C57BL/6 mouse FcγRIII haplotype T 3 linked to the cDNA coding for the human FcεRIγ (nucleotides 83 to 283). The human FcεRIγ comprised a two amino acid (aa) sequence (Pro4-Gln5) of the extracellular domain and the intact transmembrane and intracytoplasmic domains, as previously described. 4 The mCD16 chimeric cDNA was cloned into the HindIII and NotI sites of the pMX retroviral vector (Fig. 1).

Discussion
In the present work, we demonstrate that human lymphocytes can be modified by gene transfer to gain ADCC capacity in the presence of mouse mAbs.
After transduction with a mouse CD16/γ receptor gene (mCD16), the human NK cell line NK-92 mCD16 displayed stable cell surface expression of the mCD16 receptor. Using target B cells and a set of mAbs directed at several B cell Ags, we demonstrated that the mouse receptor was functional in the human NK cell line. These NK-92 mCD16 cells are useful both for analyzing the ADCC potential of mouse mAbs of different isotypes using human (effector and target) cells and consequently for comparing the abilities of different target cell Ags to induce ADCC. Taking advantage of these "murinized" effector cells, we present several examples of the dissociation between the level of opsonization of a target Ag and the ability of effector cells to lyse the Ag-expressing cells by ADCC.
To analyze the ADCC capacity of the NK-92 mCD16 cells, we first focused on the CD20 Ag due to its extensive description as an "ADCC target." When tested against a target BLCL coated with a mouse IgG1 anti-CD20 mAb (clone AT80), the mCD16 receptor enabled the NK-92 mCD16 cells to kill the BLCL by ADCC. Expectedly, mouse IgG2a, IgG2b and IgG1 were all capable of inducing ADCC in the presence of NK92 mCD16 . Of note, the NK-92 mCD16 cells mediated ADCC also in the presence of rituximab (a human IgG1), an observation in agreement with the recent analysis by MB Overdijk et al. 8 These "murinized" NK-92 mCD16 cells present several advantages over human or mouse NK cells for assessing the ADCC activity of mouse mAbs. In addition to the fact that the killing Suitability of NK-92 mCD16 to screen the ADCC potential of mouse mAbs. To compare the ADCC potential of mouse mAbs directed at different B cell target Ags, we used an EBV-LCL that expressed CD19, CD20 and CD21. Figure 5A (left panel) shows staining of the EBV-LCL with the mouse IgG1 mAbs J3-119, AT80 and BL13 (Table 1), which are specific for CD19, CD20 and CD21, respectively. After incubation with the first Ab, cells were washed before incubation with the R-phycoerythrin (PE) goat anti-mouse mAb used at saturating concentration. In addition an enzyme-linked immunosorbent assay was performed to verify that the anti-IgG binds similarly to the three Abs (data not shown). In these conditions, the levels of staining reflect the levels of opsonization of each mAb, and at it is shown in Figure 5A (left panel), at 1 μg/ml, levels of opsonization were very close for CD19, CD20 and CD21. Next, using NK-92 mCD16 and the above mouse mAbs, we compared the ADCC score mediated by each mAb. As shown on Figure  5A, right panel, no correlation was observed between the level of opsonization and the level of ADCC. Of note, in contrast to reports in the literature (see discussion), the mouse anti-CD20 mAb AT80 initiated ADCC against the EBV-LCL. A second example is presented in Figure 5B. First, CD38 and CD138 were tested with mAbs T6 and BA38 for Ag expression and the ability to drive ADCC upon Ag recognition. The target cell was the LP1 myeloma cell line. As in the previous case, when a specific mAb was used, the assay clearly dissociated opsonization from the induction of ADCC. Indeed, although CD38 and CD138 showed close levels of surface staining (Fig. 5B, left  panel), the targeting of CD38, but not CD138, by any of the four mAbs tested (BA38 shown in Fig. 5B, right panel as well as MI15, BB4 and DL101; data not shown) triggered ADCC. The study of ADCC by mouse effector cells is also affected by technical limitations. Mouse NK cells express low levels of FcγRIII 14 and show minimal effector functions when incubated in vitro with target tumor cells. In fact, prior to analyses of NK cell responses, most investigators cultured mouse NK cells in the presence of cytokines or treated mice with IFN-γ to elicit detectable effector functions. [15][16][17][18] Beyond the requirement of animal facilities and the need to treat animals and grow cells prior to experiments, which require time and money, these constraints can also affect the reproducibility of experiments, particularly in the context of a quantitative evaluation.
By contrast, straightforward experiments using NK-92 mCD16 permit reproducible comparisons of ADCC potential with different Ab/Ag pairings, as exemplified by the two sets of data in Figure 5. In the present work, performed to validate the NK-92 mCD16 cells as a cellular tool, ADCC assays were performed with a limited number of murine mAbs. Obviously, the ranking of different target Ag for their interest in inducing ADCC would require further testing. Despite this, our results clearly discriminated CD20 as a good target for ADCC compared with CD19 (as already shown by E. Hooijberg et al. 19 ) and CD21, thus providing an a posteriori validation of the of a human target cell by a murine effector cell may be hampered by the differences between accessory molecules, the major drawback with human NK cells is the FcR-Fc xenogenic interaction because ADCC by human effector cells can be mediated (at least to some extent) by mouse IgG3, IgG2a and IgG2b, but not mouse IgG1. 9,10 For example, using 29 mouse mAbs (21 IgG1s, 4 IgG2as and 4 IgG3s) directed at malignant human B cell lines and human mononuclear cells, Würflein et al. showed that only 3 (anti-HLA-DR) of them consistently induced ADCC. This observation could be explained because the anti-HLA-DR antibodies were of the IgG3 isotype, which in contrast to IgG1, interacts with human CD16. 11 The xenogenic situation also explains why the mouse anti-CD20 (clone 2B8, an IgG1), from which rituximab is derived, was initially thought incapable of mediating ADCC 12 (in the same way as the anti-Tac initially described by Junghans RP et al. 13 ). In these studies, the ADCC potential of mouse IgG1 mAbs was assessed with human PBMC. 12,13 These results were also confirmed by our data showing that no ADCC was observed against the EBV-LCL coated with the IgG1 mouse anti-CD20 (Fig. 3B, left panel). Because most of the murine mAb available are of the IgG1 isotype, human PBMC or NK cells are clearly not suitable for assessing ADCC activity by murine mAbs. One hundred thousand (0.1 × 10 6 ) cells (NK-92, NK-92 mCD16 , NK-92 hCD16 , BLCL or LP1) were incubated for 15 min at room temperature at the indicated mAb concentrations diluted with PBS supplemented with 0.1% human albumin (HA) in a final volume of 30 μl. After staining, plates were centrifuged, the supernatant was discarded by flicking, and wells were washed twice with 200 μl PBS. Negative controls were set up in the presence of a control isotype in case of direct staining or in the absence of first Abs for indirect staining. For indirect staining, cells were washed after the first incubation and the second Ab was used at saturating concentration.
Construction of the human FcγRIII-human FcεRIγ and mouse FcγRIII-human FcεRIγ chimeric cDNAs. The chimeric cDNA coding for human FcγRIIIA-V158 linked to human FcεRIγ was described previously. 2 The chimeric cDNA coding screening. Such discriminations could not have been drawn, as specified above, by using human PBMC.
Comparison of ADCC against a myeloma cell line induced by opsonization with anti-CD38 and anti-CD138 identified only CD38 as a potential target CD for ADCC. The experiments presented in Figure 5 also demonstrate that the level of CD opsonization obtained with a particular mAb (rank order: CD19 = CD21 > CD20) was clearly dissociated from the ADCC that was triggered by the same mAb-coated CD (rank order: CD20 > > CD19, while ADCC against CD21 was not detectable). This dissociation, initially described by JE Christiaansen et al. 1 in a xenogenic FcγR-Fc model (with human effector cells and mouse IGg2a), reveals other levels of complexity regarding the nature of the cell-Ab-cell interactions that drives ADCC. In conclusion, the "murinized" NK-92 mCD16 cells described in the present work represent a sensitive and easy to use cellular tool for the study of ADCC and the screening of mouse mAbs according to their ADCC potential.
Antibodies and staining. The specificity, isotype and source of mAbs used in the study are indicated in Table 1.  The NK-92 cell line was resuspended in RPMI 1640 culture medium supplemented with 10% FBS and 100 UI/ml of recombinant IL-2, seeded at 1 × 10 6 cells in 1 ml per well into 6-well plates and exposed to 2 × 2 ml of retroviral supernatant by spinoculation (2400 g, 1.5 h, 32°C) in the presence of 4 μg/ml polybrene (Sigma). The culture medium was changed 24 h post-infection. Mock (non-transduced) controls were performed in parallel, by which the supernatant of untransfected packaging cells was added to the NK-92 cell line. The transduction efficiencies were assessed 5 d later by staining the for mouse FcγRIII linked to human FcεRIγ was synthesized (GeneCust). This cDNA was composed of the extracellular domain of mouse FcγRIII (nucleotides 1-215, NCBI Reference Sequence: NP-034318.2 from C57BL/6 strain, haplotype T) linked to the cDNA coding for human FcεRIγ (nucleotides 83-283, GenBank Accession number BC033872). The human FcεRIγ included a two aa sequence (Pro4-Gln5) of the extracellular domain and the intact transmembrane and intracytoplasmic domains, as previously described. 4 The chimeric cDNA of mouse FcγRIII-human FcεRIγ (referred to as mCD16) and the chimeric cDNA of human FcγRIII-human FcεRIγ (referred to as hCD16) were cloned into HindIII-NotI and BamHI-NotI sites, respectively, within the Moloney mouse leukemia virus vector pMX under the transcriptional control of the viral longterminal repeat. 20 Retroviral vector production. Transient retroviral supernatants were produced by CaCl 2 precipitation with 15 μg of plasmid. Two million Phoenix-Ampho cells 21 were seeded into 10-cm-diameter dishes 24 h prior to transfection. The transfection was performed with 15 μg pMX/mCD16 or pMX/hCD16 release represents the mean cpm of the target cells incubated with 1% Triton X 100 (Sigma).
Enzyme-linked immunosorbent assay. A 96-well plate (Maxisorp Nunc) was coated for 24 h at 4°C with 5 μg/ml of the mouse antibodies anti-CD19, anti-CD20, anti-CD21 in Bicarbonate/carbonate coating buffer, 100 mM, pH 9.6. The plate was washed three times with PBST (PBS buffer containing 0.05% Tween 20) and then blocked with PBST-1% BSA at room temperature for 1 h, followed by three washes with PBST. The plate was incubated for 1 h at room temperature with the secondary antibody [Goat F(ab') 2 Fragment Anti-Mouse IgG (H + L)-Peroxidase, Beckman Coulter], diluted 1/5000, 1/10,000, 1/25,000, 1/50,000, 1/100,000 in PBST-1% BSA. The wells were washed three times with PBST and one time with PBS. Bound secondary antibody was detected by an incubation of 15 min with the substrate SIGMAFAST™ OPD (o-Phenylenediamine dihydrochloride). The reaction was stopped by adding 0.5 M H 2 SO 4 and the absorbance at 492 and 620 nm was measured with an iEMS reader MF spectrometer (Labsystems).

Disclosure of Potential Conflicts of Interest
No potential conflict of interests were disclosed.

Acknowledgments
This work was supported by institutional grant from INSERM, by the CANCEROPOLE Grand Ouest and the Ligue Régionale contre le Cancer. BC designed research, performed research, analyzed data and wrote the manuscript. RV performed research, CP provided myeloma cell lines and analyzed data, GT performed research, analyzed data and edit the manuscript, HV designed research, analyzed data and wrote the manuscript. mCD16-and hCD16-transduced NK-92 cell lines (referred to as NK-92 mCD16 and NK-92 hCD16 ) with a PE-conjugated rat anti-mouse CD16/32 mAb (clone 93) (Beckman Coulter) or a PE-conjugated mouse anti-human CD16 mAb (clone 3G8) (Beckman Coulter), respectively.
Cytotoxicity and ADCC assay. Cytotoxic activity was assessed using a standard 51 Cr release assay. The target cells were labeled with 100 μCi 51 Cr for 1 h at 37°C, washed four times with culture medium and plated at the indicated effector-totarget cell ratios in 96-well flat-bottom plates. The indicated mAb was incubated with the target cells for 20 min at room temperature before the addition of effector cells. After a 4 h incubation at 37°C, 25 μl of supernatant were removed from each well and mixed with 100 μl scintillation fluid, and 51 Cr activity was counted in a scintillation counter (MicroBeta, Perkin Elmer). Each test was performed in triplicate. The results are expressed as the percentage of lysis, which is calculated according to the following equation: (experimental release − spontaneous release)/(maximal release-spontaneous release) × 100, where the experimental release represents the mean counts per minute (cpm) of the target cells in the presence of effector cells; spontaneous release represents the mean cpm of the target cells incubated without effector cells; and maximal