NK cells engineered to express a GD2-specific antigen receptor display built-in ADCC-like activity against tumour cells of neuroectodermal origin

Abstract Treatment of high-risk neuroblastoma (NB) represents a major challenge in paediatric oncology. Alternative therapeutic strategies include antibodies targeting the disialoganglioside GD2, which is expressed at high levels on NB cells, and infusion of donor-derived natural killer (NK) cells. To combine specific antibody-mediated recognition of NB cells with the potent cytotoxic activity of NK cells, here we generated clonal derivatives of the clinically applicable human NK cell line NK-92 that stably express a GD2-specific chimeric antigen receptor (CAR) comprising an anti-GD2 ch14.18 single chain Fv antibody fusion protein with CD3-ζ chain as a signalling moiety. CAR expression by gene-modified NK cells facilitated effective recognition and elimination of established GD2 expressing NB cells, which were resistant to parental NK-92. In the case of intrinsically NK-sensitive NB cell lines, we observed markedly increased cell killing activity of retargeted NK-92 cells. Enhanced cell killing was strictly dependent on specific recognition of the target antigen and could be blocked by GD2-specific antibody or anti-idiotypic antibody occupying the CAR’s cell recognition domain. Importantly, strongly enhanced cytotoxicity of the GD2-specific NK cells was also found against primary NB cells and GD2 expressing tumour cells of other origins, demonstrating the potential clinical utility of the retargeted effector cells.


Introduction
Natural killer (NK) cells are part of the innate immune system and the body's first line of defence against virally infected and malig-nant cells. Unlike T cells, which are major histocompatibility complex (MHC) restricted and need to be sensitized and educated about the target, cytotoxicity of NK cells can be activated rapidly and is regulated by a balance of signals from germline-encoded activating and inhibitory cell surface receptors [1,2]. Over the last decade, significant progress has been made towards realizing the potential of NK cells for cancer immunotherapy [3,4]. In addition to autologous or donor-derived primary NK cells, also clinically applicable, continuously growing cytotoxic cell lines such as NK-92 are being developed for adoptive cancer immunotherapy. NK-92 cells exhibit functional characteristics of activated NK cells and express typical NK-cell surface receptors, but lack Fc␥RIII [5]. In preclinical models high intrinsic cytotoxicity of NK-92 against leukaemia, lymphoma, melanoma and other malignancies has been demonstrated [6][7][8]. This high endogenous cytotoxic potential of NK-92 has been attributed to the absence of most of the inhibitory NK cell receptors, although high numbers of activating lectin-like and immunoglobulin-like receptors are being expressed [9]. General safety and tolerability of NK-92 cell infusion has been established in phase I clinical studies, with some patients experiencing clinical responses [10,11].
We extended this approach to target cells that display intrinsic resistance to NK cells by genetic modification of NK-92 cells with chimeric antigen receptors (CARs), which recognize protein antigens differentially expressed on the surface of cancer cells [12][13][14]. Such CAR are transmembrane proteins that consist of an extracellular scFv single-chain Fv antibody as a cell binding domain, genetically linked by a flexible spacer to transmembrane and intracellular parts of signalling molecules such as the CD3chain [15,16]. Here we investigated the suitability of this approach to redirect cytolytic activity of NK-92 to cancer cells that express the non-proteinaceous antigen GD2. The disialoganglioside GD2 is a sialic acid-containing glycosphingolipid, which is thought to play a role in the attachment to components of the extracellular matrix [17]. In normal foetal and adult tissues GD2 expression is mainly restricted to the central nervous system, peripheral nerves and skin melanocytes. As a tumour-associated antigen, GD2 is consistently expressed in malignancies of neuroectodermal origin such as neuroblastoma (NB), melanoma, and to a variable degree in other tumours such as bone and soft-tissue sarcoma, small cell lung cancer and brain tumours [18,19]. Due to its relatively selective expression by tumour cells and its accessibility on the cell surface, GD2 is a suitable antigen for immunotherapy. Murine (3F8 and 14G2a) and chimeric human/mouse (ch14.18) anti-GD2 antibodies have been developed over the past two decades and have been investigated in clinical studies for the treatment of NB and melanoma [18,19].
Neuroblastoma represents the most common extracranial solid tumour during early childhood. Although survival rates of NB patients with locoregional disease exceed 90%, metastatic NB in patients over 18 months of age is often incurable even with advanced multimodality chemotherapy regimens [20][21][22]. For more than one decade, high-risk stage IV NB patients have been treated with high-dose chemotherapy followed by autologous stem cell transplantation (auto-SCT), but outcome still remains poor [23]. Clinical trials using chimeric GD2-specific antibody ch14.18 as a single agent [24], or administration of the antibody together with systemic application of GM-CSF and interleukin (IL)-2 [25] were performed. In the latter phase III study treatment of stage IV NB patients with ch14.18 in combination with IL-2 and GM-CSF following auto-SCT resulted in a 20% increase in survival [25]. To enhance the efficacy of antibody therapy, and to minimize systemic toxicities as well as immunogenicity, an immunocytokine (hu14.18-IL2) was developed that targets IL-2 directly to tumour cells by fusing it to the humanized anti-GD2 antibody hu14.18 [26]. Also immunotherapy with highly purified donor NK cells that exhibit cytotoxic activity against NB cells with low MHC class I expression represents a promising strategy to improve clinical outcome of patients with NB stage IV. In an ongoing phase I/II study haploidentical SCT (haplo-SCT) in combination with infusion of IL-2 stimulated allogeneic donor NK cells is being employed for the treatment of high-risk NB patients after unsuccessful auto-SCT [27].
To combine specific recognition of NB cells with the potent cytotoxic activity of NK cells, here we generated clonal NK-92 cells that stably express a GD2-specific CAR which carries a cell binding domain derived from antibody ch14.18. Such NK-92-scFv(ch14.18)-cells displayed markedly enhanced cytotoxicity against established GD2-expressing NB tumour cells, whereas lysis of GD2targets remained unaffected. Importantly, selective antitumoral activity was also observed against freshly isolated primary NB tumour cells, indicating that this strategy may be suitable for further development as a cell-based immunotherapy for NB.
Primary human NB cells were isolated after surgery from a tumour metastasis of a 4-year-old patient with NB stage IV, and from the bone marrow (BM) of another high-risk patient after relapse. The tumour metastasis from jejunum was minced, treated with papain for digestion and percolated through a 40 m filter (Cell strainer; BD Biosciences, Heidelberg, Germany) to singularize the cells. Cells were resuspended in DMEM/F-12 medium (Invitrogen, Darmstadt, Germany) containing 100 units/ml penicillin and 100 g/ml streptomycin, and cultivated until analysis in Iscove's modified Dulbecco's medium with the same antibiotics. BM cells of healthy donors and NB patients were obtained after informed consent. Mononuclear cells were applied for analysis after centrifugation with Ficoll Hypaque. Research use of anonymized samples of peripheral blood stem cells (PBSC) and CD34 ϩ cells from healthy donors remaining after allogeneic transplantation was approved by the University Hospital Ethics Committee at the University of Frankfurt, Germany.

Production of amphotropic retroviral vectors and transduction of NK-92 cells
FLYA-JET packaging cells [31] were transfected with pL-scFv(ch14.18) HL--SN and pL-scFv(ch14.18)LH--SN constructs by electroporation using the Easyject Optima electroporation system (Equibio, Ashford, UK) with the following parameters: 20 g of plasmid DNA per 1 ϫ 10 6 cells in 0.8 ml of DMEM medium in a 0.4 cm cuvette, and 'standard' settings according to the manufacturer's recommendations. Stable transfectants were selected for 1 week in DMEM growth medium containing 2.4 mg/ml G418. For production of amphotropic retroviral vector, selected packaging cells were grown over night in NK-92 medium. Culture supernatants were passed through a 0.2 m filter and incubated with NK-92 cells in the presence of 8 g/ml polybrene for 5 hrs at 37ЊC. Then NK-92 cells were grown over night in fresh X-VIVO 10 medium, before G418 was added to a final concentration of 0.6 mg/ml for selection of NK-92-scFv(ch14.18)cells.

Surface expression of GD 2
Expression of GD2 on the surface of established cell lines and primary human cells was determined by flow cytometry using custom PE/Cy5 conjugated 14.G2a anti-GD2 antibody and mouse IgG2a isotype control (BD Biosciences). For quantification of GD2 molecules, cellular antigen expression was measured as antibody binding capacity (ABC) units using the Quantum Simply Cellular kit (Bangs Laboratories, Indianapolis, IN, USA) according to the manufacturer's recommendations. The ABC values calculated for GD2 after subtraction of the values of the appropriate isotype control were expressed as molecules per cell. Standard curves were set with the maximum at 250,000 ABC units, and the detection threshold was Ͻ1000 ABC units. The quantitative analyses were performed on an FC 500 cytometer and data were analysed using the CXP v2.2 software (Beckman Coulter, Krefeld, Germany) as described previously [27].
For FACS-based cell killing assays, adherent tumour cells were harvested by treatment with Accutase (PAA Laboratories, Pasching, Austria) for 5 to 10 min. at 37ЊC and singularized. To prevent unwanted cell clumping, the cells were then incubated with 50 g/ml of DNase I (Roche Applied Science, Mannheim, Germany) in 1ϫ reaction buffer prepared from a 10ϫ stock solution (100 mM Tris-HCl, pH 7.6, 25 mM MgCl2, 5 mM CaCl2) (Roche Applied Science) for 15 min. at 37ЊC under constant gentle shaking.
The reaction was blocked by addition of 250 mM ethylenediaminetetraacetic acid to a final concentration of 50 mM. Target cells were incubated with NK-92 or NK-92-scFv(ch14.18)-cells for 2 and 4 hrs in X-VIVO 10 medium at different E/T ratios. Lytic activity was measured by single platform 5-colour flow cytometric analysis for effector cells alone, and for co-cultured effector and target cells on an FC 500 cytometer (Beckman Coulter) as described previously [32,33]. Briefly, calculation of cytotoxicity was based on the loss of living propidium iodide (PI)target cells (CD9 ϩ CD81 ϩ CD45 neg PI neg ). Absolute count-dedicated beads Cytotoxic activity of NK cells upon prolonged co-culture with target cells at low E/T ratios was visualized by microscopy. Microscopic images of cells were taken after 18 hrs of co-culture using an Axiovert 135 microscope (Carl Zeiss, Göttingen, Germany) and a Sony 3CCD camera (Sony, Berlin, Germany).

Microscopic control of the interaction between NK-92 and primary NB cells
Primary NB cells were seeded at low density on 20 mm chamber slides (Nunc, Langenselbold, Germany) and grown for 24 hrs. Then adherent NB cells were co-cultured with gene-modified NK-92-scFv(ch14.18)-or parental NK-92 cells for 4 hrs at an E:T ratio of 5:1. NK-92 cells were identified with CD45-specific antibody followed by FITC-conjugated secondary antibody. CAR expressing NK-92 cells were detected with mAb 9E10 followed by PE-conjugated secondary antibody. All cells were counter-stained with DAPI. Fluorescence microscopy was performed with an Olympus IX71 microscope (Olympus, Hamburg, Germany), and microscopic images were acquired with a charge-coupled device camera.

Statistical analysis
Kruskal-Wallis test with Dunn's multiple comparison and Wilcoxon matched pairs test were applied to assess statistical significance of differences between groups. P values Ͻ0.05 were considered as significant. Data were analysed using GraphPad Prism software (GraphPad Software, San Diego, CA, USA).

Generation of NK cells carrying GD 2 -specific chimeric antigen receptors
GD2-specific scFv(ch14.18) antibody fragments were derived from constructs encoding scFv(ch14.18)-Fc fusion proteins that carry heavy and light chain variable domains of the chimeric mAb ch14.18 [34,35]. To address potential differences in the functionality of scFv(ch14.18) molecules that depend on the orientation of the variable domains, we employed scFv fragments where VH and VL of antibody ch14.18 were either assembled in the orientation VH-linker-VL, or VL-linker-VH, with the synthetic (G4S)4 sequence serving as a flexible linker. Chimeric antigen receptors were constructed by inserting the scFv fragments designated scFv(ch14.18)HL and scFv(ch14.18)LH between a sequence encoding an N-terminal immunoglobulin heavy-chain signal peptide, and sequences encoding a Myc-tag, the CD8␣ hinge region (amino acids 105-165) and the CD3-chain in the retroviral transfer vector pLXSN [30] (Fig. 1A).
Amphotropic retroviral vector particles were produced by stable transfection of FLYA-JET packaging cells [31], and used for transduction of human NK-92 cells. After selection with G418, expression of scFv(ch14.18)HL-and scFv(ch14.18)LH-receptor proteins on the cell surface was analysed by flow cytometry. At this step the majority of cells in the selected cell pools displayed low or undetectable expression of the CARs (Fig. 1B, left panels). To enrich NK-92 cells that express more homogenous receptor levels, cells were sorted with Myc-tag specific mAb 9E10 and immunomagnetic beads (Fig. 1B, middle panels), followed by limiting dilution to obtain single cell clones. This yielded stable NK-92 cell clones consistently expressing high levels of CARs (Fig. 1B, right panels). We did not observe a difference in expression levels between clones carrying scFv(ch14.18)HL-or scFv(ch14.18)LH-( Fig. 1B and data not shown), indicating that the orientation of VH and VL within scFv(ch14.18) had no influence on the overall expression or surface display of the receptors.

Surface expression of GD 2 on NB cells
As a prerequisite for the analysis of CAR functionality and activity of retargeted NK-92 cells, first surface expression of GD2 by established NB cell lines and primary NB cells was investigated by flow cytometry using fluorochrome-labelled GD2-specific murine mAb 14.G2a. Control cells were treated with an irrelevant isotypematched antibody. Established human UKF-NB3, Kelly, BE(2)C and LAN-1 NB cells displayed intermediate to high levels of GD2 on their surface, whereas only a very weak signal was determined with anti-GD2 antibody for SK-N-SH cells ( Fig. 2A). Analysis of primary NB cells from the BM of 12 relapsed NB patients revealed markedly enhanced GD2 expression in these samples when compared to established GD2 ϩ NB cell lines (data not shown).
To illustrate this pronounced difference exemplarily, BM with NB cells from one of these patients was mixed with established UKF-NB-3 cells, and GD2 expression was determined by flow cytometry (Fig. 2B).
For quantification of GD2 molecules, cellular antigen expression on the surface of established and primary NB cells as well as haematopoietic cells from healthy donors was determined in comparison to antibody-binding microbeads as a standard (Fig. 2C). We found very high expression of GD2 in the range of 2 to 4 ϫ 10 5 antibody binding capacity (ABC) molecules/cell for established BE(2)C and UKF-NB3 NB cells, 7 ϫ 10 3 ABC molecules/cell for Kelly cells and 1.5 ϫ 10 3 ABC molecules/cell for SK-N-SH (Fig. 2C), which is near the lower detection limit of 1 ϫ 10 3 epitopes/cell for this assay. Malignant cells of non-NB origin such as K562 (erythroleukaemia) expressed low levels of GD2 (3.5 ϫ 10 3 ABC molecules/cell). Similar results were obtained for A673 peripheral neuroepithelioma tumour (PNET) and rhabdomyosarcoma cells (data not shown). GD2 levels on primary NB cells isolated from a tumour metastasis and the BM of a relapsed NB patient with stage IV disease were above 2 ϫ 10 6 ABC molecules/cells, which was the upper detection limit of the assay (Fig. 2C). Haematopoietic cells from healthy donors expressed very low GD2 levels (sorted CD34 ϩ cells), or yielded ABC signals below the lower detection limit of the assay (lymphocytes in PBSC and BM).
These results demonstrate that expression of scFv(ch14.18)receptors on NK-92 triggers markedly enhanced cell killing towards NB cells, which is dependent on specific recognition of GD2 on the target cell surface. Thereby NK-92-scFv(ch14.18)HLand NK-92-scFv(ch14.18)LH-cells were similarly active, indicating that the orientation of VH and VL in the scFv(ch14.18) antibody fragment does not influence target cell recognition and GD2specific cytotoxicity.

18)-against primary NB cells
To investigate cytotoxic activity of NK-92-scFv(ch14.18)towards primary tumour cells, human NB cells were freshly isolated from a tumour metastasis of a 4-year-old patient with NB stage IV, singularized and cultivated for several days. Quantitative flow cytometric analysis revealed very high levels of GD2 expression on the surface of these cells (Fig. 2C). The primary NB cells were co-cultured at different E/T ratios with NK-92-scFv(ch14.18)HL-or parental NK-92 cells for 2 hrs before analysis of specific lysis in FACS-based assays. Primary tumour cells displayed intermediate sensitivity to parental NK-92 cells (43% specific lysis at an E/T ratio of 10:1), but were highly sensitive to CAR-expressing NK-92-scFv(ch14.18)HL-cells, resulting in 82% specific lysis at an E/T ratio of 10:1 (P Ͻ 0.01) (Fig. 5A). This enhanced cytotoxic activity corresponded well with the high level of GD2 expression on the target cells. In accordance with the results of the cytotoxicity analysis, interaction between genemodified NK-92 and adherent, primary NB cells was substantially enhanced when compared to untargeted parental NK-92 cells (Fig. 5B).

Discussion
High-risk stage IV NB is characterized by disseminated metastasis, and continues to be a therapeutic challenge in paediatric oncology. Despite intense multimodal treatment, prognosis remains poor with long-term survival in only around 30% to 40% of patients [20][21][22], justifying efforts to develop alternative, more efficient treatment strategies. In this study, we have redirected continuously growing human NK cells to GD2 expressing NB cells using CARs that harbour a GD2-specific scFv fragment derived from the ch14.18 antibody. This was linked to the CD3-chain to trigger cytolytic activity upon target cell recognition. Gene-modified NK cell lines were derived by transduction with retroviral vectors followed by selection of single cell clones, which displayed high and stable CAR expression over several months in continuous culture. NK-92 cells carrying CAR that harboured scFv(ch14.18) antibody fragments either in VH-linker-VL or VLlinker-VH orientation both displayed high cytotoxic activity towards GD2 expressing NB and melanoma cells. A moderate cell killing activity was achieved against breast cancer cells with more limited GD2 expression.
Immunotherapeutic approaches that target GD2 have first focused on monoclonal antibodies. To date murine antibodies 3F8 and 14G2a, the chimeric human/mouse antibody ch14.18, and the humanized immunocytokine hu14.18-IL2 have been used in clinical trials for the treatment of NB to enhance antibody-dependent cell-mediated cytotoxicity (ADCC) against GD2 ϩ tumour cells [18,25,26]. Yet there is no consensus about the benefit of these strategies [37]. Promising responses were seen in a recent phase III study combining ch14.18 with GM-CSF and IL-2 [25], and a recent phase II study demonstrating antitumoral activity of the immunocytokine hu14.18-IL2 in patients with relapsed or refractory NB [26]. Other reports described no advantage over conventional therapy [38]. Yu et al. attributed the difference in outcome to the addition of IL-2 and GM-CSF, which augments ch14.18mediated ADCC in vivo [25]. Combination of mAbs with different treatment modalities may be required to improve survival of NB patients, which includes utilization of cellular effector mechanisms [39]. Safety and feasibility of anti-GD2 antibody 3F8 together with haploidentical NK cells for the treatment of high-risk disease is currently being investigated in a phase I study (NCT00877110; clinicaltrials.gov). The lack or very low expression of MHC class I molecules on NB cells make them an ideal target for NK cells. Ex vivo stimulation of the effector cells with IL-2 resulted in increased expression of natural cytotoxicity receptors and NKG2D, and enhanced cytotoxicity towards NB cells [27]. Nevertheless, expression of soluble NKG2D ligands such as MHC class I related protein A by NB cells can have a marked inhibitory effect on NK cells [40,41]. Furthermore, ADCC inducing activity of anti-GD2 antibodies may be affected by Fc receptor polymorphisms, as it has been described for other therapeutic antibodies [42,43].
Our results show that expression of a GD2-specific CAR in NK cells directly couples antibody-mediated recognition of GD2 with the execution of cytotoxicity. This provides the effector cells with built-in ADCC-like activity, which can bypass limitations such as insufficient Fc␥RIII activation by antibodies and overcome the tumour cells' endogenous resistance mechanisms as demonstrated for Kelly and LAN-1 cells. These NB cells were largely resistant to parental NK-92 cells, but were lysed by NK-92-scFv(ch14.18)-with high efficiency. In the case of intrinsically NK-sensitive NB cell lines, we observed markedly increased cell killing activity of NK-92-scFv(ch14.18)-cells. This enhanced activity was strictly dependent on specific recognition of the GD2 target antigen by the NK cells. Blocking of GD2 on the target cell surface with GD2-specific antibody or occupation of the antigen binding site of the scFv(ch14.18) domain of the CAR with an antiidiotypic antibody abrogated cytotoxic activity of NK-92-scFv(ch14.18)-towards GD2-expressing targets. In addition to GD2-dependent cell killing, retargeted NK-92-scFv(ch14.18)cells retained endogenous natural cytotoxicity of NK-92, demonstrated by similar activity of NK-92 and the GD2-specific variant against NK-sensitive SK-N-SH NB cells which express only very low GD2 levels. This may be of advantage for the treatment of tumours that consist of cells with heterogeneous GD2 expression. Importantly, strongly enhanced cytotoxicity of NK-92-scFv(ch14.18)-cells was also found in the case of primary NB cells, indicating that the retargeted effector cells may be of clinical utility. GD2 has previously been established as a relevant cancer antigen for NB, and only minimal intra-or intertumoral heterogeneity of GD2 expression has been described [44,45]. Moreover, persistent GD2 antigen expression after treatment with GD2-specific antibody was demonstrated. Only in 1 out of 62 NB patients the tumour lost GD2 expression after therapy, but in this case underwent phenotypic transformation into a pheochromocytomalike tumour [45].
High level GD2 expression is restricted to NB, melanoma and other tumours of neuroectodermal origin [18,19,39]. We did not observe measurable GD2 expression on lymphocytes from BM and peripheral blood, and only very low GD2 expression on sorted CD34 ϩ progenitor cells. Hence, haematological toxicities induced by treatment with GD2-specific NK cells appear unlikely. Nevertheless, limited antigen expression in other normal tissues can result in unwanted side effects of GD2-targeted therapies. Reported toxicities of treatment with anti-GD2 antibodies with or without GM-CSF included fever, nausea/vomiting, urticara, hypotension, capillary leak syndrome, ocular symptoms, neuropathic pain and neurotoxicity [25,26,38,46,47]. In general, these adverse effects were considerable but manageable, and did not result in discontinuation of the treatment. Recent data from animal models suggest that such toxicities are to a large part due to complement-dependent cytotoxicity (CDC) induced by such antibodies, and may be reduced by limiting CDC [48]. Introduction of a point mutation in GD2-specific ch14.18 antibody which interferes with CDC while retaining ADCC activity, resulted in markedly reduced allodynia in rats when compared to the parental molecule [48]. In the case of retargeted NK-92-scFv(ch14.18)-cells, the antibody fragments within the GD2-specific CAR are restricted to the variable domains of antibody heavy and light chains, and do not contain the Fc portion that would be required to elicit CDC.
Hence, CAR-triggered cytotoxicity of the NK cells is limited to ADCC-like activity, which will likely circumvent CDC-dependent toxicities.
Similar to our results, selective cytotoxicity of GD2-specific primary NK cells and cytotoxic T lymphocytes (CTLs) was found in experimental models [49,50]. Clinical application of GD2targeted CTLs appeared safe, and was associated with tumour regression or necrosis in half of the NB patients treated [51]. In the latter study, Epstein-Barr virus-specific CTLs were employed for genetic modification with a GD2-specific CAR to minimize the risk of autoimmunity caused by reactivity of the endogenous T-cell receptors of the retargeted cells with normal tissue antigens. In studies evaluating untargeted NK-92 cells, the effector cells were irradiated as a safety measure prior to infusion into cancer patients. This prevented permanent engraftment, although cells remained viable and retained cytotoxicity for several days [10,11]. This was also the case for irradiated retargeted NK-92 cells (this study and [12]).
The use of CAR-expressing primary cells for adoptive immunotherapy requires for each individual patient the isolation, expansion and genetic modification of the relevant T-or NK-cell populations. Clinically applicable cytotoxic cell lines such as NK-92 could complement these approaches, especially in cases where autologous effector cells cannot be employed and suitable donors are not available. Methodology for GMP-compliant large scale production of unmodified NK-92 cells is well established [10,11], and could be readily applied for continuous expansion of retargeted derivatives such as GD2-specific NK-92-scFv(ch14.18)-cells. NK-92 cells express HLA class I molecules. Nevertheless, in the clinical trials conducted so far with unmodified, parental NK-92 cells, only few patients developed anti-HLA class I antibodies upon repeated treatment with the allogeneic effector cells [10,11]. This was most likely due to the impaired immune status of the patients following chemo-and radiation therapy. Hence, repeated intravenous NK-92 therapy appears feasible when performed under continuous crossmatch testing with the patients' serum before infusion. Our results demonstrate that clonal NK-92-scFv(ch14.18)-cells selectively and reliably eliminate established and primary NB cells and GD2 expressing tumour cells of other origins. Utilization of NK-92-scFv(ch14.18)-cells could bypass the need for separate genetic modification of effector cells for each individual patient, justifying further development of this approach.