Intimate Cell Conjugate Formation and Exchange of Membrane Lipids Precede Apoptosis Induction in Target Cells during Antibody-Dependent, Granulocyte-Mediated Cytotoxicity1

Ab-dependent polymorphonuclear granulocyte (PMN)-mediated cytotoxicity may play an important role in the control of malignant diseases. However, little is known as to which particular pathways are used for the killing of malignant cells by PMN. The production of reactive oxygen intermediates (ROI) has been observed to occur during Ab-dependent, cell-mediated cytotoxicity (ADCC). However, PMN from a patient with chronic granulomatous disease demonstrated strong ADCC against malignant lymphoma cells. Furthermore, the inhibition of ROI production in PMN from healthy donors had no significant effect on ADCC. Therefore, ROI production by the NADPH oxidase of PMN does not appear to be mandatory for PMN-mediated ADCC. Recent data suggest a role for perforins in PMN-mediated cytotoxicity. However, in our assays concanamycin A, an inhibitor of perforin-mediated ADCC by mononuclear cells, had no inhibitory effect on PMN-mediated ADCC. Using electron microscopy we observed that PMN and their target cells intimately interact with the formation of interdigitating membrane protrusions. During PMN and target cell contact there was a mutual exchange of fluorescent membrane lipid dyes that was strongly increased in the presence of tumor-targeting Abs. This observation may be closely related to the recently described process of trogocytosis by lymphocytes. The presence of transient PMN-tumor cell aggregates and the accumulation of PMN with tumor cell-derived membrane lipids and vice versa were associated with effective ADCC as measured by chromium-release or apoptosis induction.

T he capacity to mediate Ab-dependent cell-mediated cytotoxicity (ADCC) 4 has been demonstrated in vitro for monocytes/macrophages and NK cells as well as for polymorphonuclear granulocytes (PMN). PMN are potent effector cells against a wide range of malignancies in vitro (1-3) and are considered important for the rejection of malignant tumors in vivo (4). In contrast to T cells, the cytotoxicity of PMN is dependent on Ab binding to the target cells. PMN from G-CSF-treated patients were the predominant effectors for the killing of breast cancer cells in vitro in the presence of bispecific Ab (bsAb), recognizing both Fc␥RI and tumor target Ag HER-2/neu (2). Recent therapeutic advances in Ab-mediated cancer therapy (5,6) have renewed the interest in PMN-mediated ADCC. However, the mechanism of target cell death remains elusive (7). Usually, PMN-mediated ADCC has been measured by 3 h of 51 Cr-release assays (8,9) with high E:T cell ratios (10). Previous experiments with PMN effectors demonstrated unexplained differences in tumor cell-related Ag serving as targets for ADCC. Thus, mAb against HLA class II or against members of the epidermal growth factor receptor family such as HER-2/neu were highly effective in recruiting PMN as effector cells, whereas CD10, CD19, CD21, CD37, and CD38 Ab were ineffective in this regard (1-3, 11, 12).
To recruit cell-mediated effector mechanisms, Ab must interact with Ig Fc receptors, which are classified as Fc␥, Fc␣, or Fc receptors depending on their specificity for IgG, IgA, or IgE, respectively (13). The pivotal role of Fc receptors for the antitumor activities of mAb in vivo has been demonstrated in knockout mice in which the signaling machinery of the Fc receptors had been genetically disrupted (14). Furthermore, recent studies of non-Hodgkin's lymphoma showed a correlation between the Fc␥RIIIa (CD16a) and Fc␥RIIa (CD32) genotype and the clinical and molecular responses to a CD20-directed therapy with the mAb rituximab (15)(16)(17). However, conflicting results have been reported during the investigation of other malignancies treated with mAb (18).
PMN constitutively express Fc␣RI (CD89), the low affinity Fc␥RIIa (CD32a), and Fc␥RIIIb (CD16b) (19). The expression of the high-affinity Fc␥RI (CD64) can be induced in vivo by stimulation with G-CSF (20). We have recently demonstrated that Fc␣RI-directed bsAb proved more effective in triggering tumor cell lysis than the respective Fc␥-directed constructs, particularly with unstimulated PMN (19,21). In contrast to some members of the Fc␥R family, Fc␣RI is exclusively expressed on cytotoxic cells (22).
A breakdown of the membrane lipid asymmetry, the loss of matrix adhesion and mitochondrial membrane potential, the "boiling" of cytoplasm, the condensation of chromatin, and, finally, the internucleosomal cleavage of the nuclear DNA characterize apoptotic cell death (23). At an early stage of apoptosis the exposure of phosphatidylserine on the outer leaflet of the cytoplasmic membrane is detected by the binding of annexin V. DNA fragmentation and the loss of chromatin resulting in a sub-G 1 DNA content of the dying cells is a characteristic feature of the later stages of apoptosis. Under physiological conditions when high E:T cell ratios are unlikely to occur, apoptosis induction is supposed to be the major pathway of cell death (7). After activation, PMN produce reactive oxygen intermediates (ROI) that are necessary for their fungicidal and bactericidal effects, but their role in ADCC remains to be defined. Recently, PMN have been shown to contain granzyme B and perforin (24,25). However, contradictory observations exist (26,27). Presuming that PMN contain granzyme B and perforin, we tried to inhibit PMN-mediated ADCC with concanamycin A (CMA), which is a potent inhibitor of perforin-mediated NK cell cytotoxicity (28). Morphological, immunological, and functional characterizations of PMN-mediated ADCC were performed by electron microscopy, confocal laser-scanning microscopy, and FACS.

Isolation of effector cells
The experiments reported here were conducted in accordance with the Declaration of Helsinki. After informed consent, 10 -20 ml of citrate anticoagulated peripheral blood was drawn from normal healthy donors (NHD) or a patient with chronic granulomatous disease (CGD) for effector cell preparation. The male CGD patient had an X-linked deletion of C1475 in exon 12 of the CYBB gene encoding gp91 phox , predicting a frame shift and resulting in a stop codon in the NADPH binding domain (31). PMN were isolated from peripheral blood as described (11). Briefly, citrate anticoagulated blood was layered over a discontinuous Percoll gradient (62 and 70%; Seromed). After centrifugation, PMN were collected at the interphase between the two Percoll layers. The remaining erythrocytes were removed by hypotonic lysis. The purity of PMN as determined by cytospin preparations exceeded 95%, and cell viability was Ͼ95% as tested by trypan blue exclusion. Mononuclear cells (MNC) containing NK cells, T cells, B cells, and monocytes/macrophages were collected between the plasma and the 62% Percoll layer.

ADCC assays
Chromium 51 release assays were performed as described elsewhere (11). Briefly, target cells were labeled with 200 Ci of 51 Cr for 2 h. To investigate the influence of caspases on PMN-mediated ADCC, Raji H cells were preincubated with 100 M z-Val-Ala-DL-Asp-fluoromethylketone (z-VAD-fmk; Alexis), a pan-caspase inhibitor, during the labeling process with chromium. After extensive washing with the RF10 ϩ medium, cells were adjusted to 10 5 /ml. PMN or MNC at an E:T ratio of 40:1 and sensitizing Ab at final concentrations of 2 g/ml were added to round-bottom microtiter plates (Nunc) unless otherwise noted. To address the role of ROI production in ADCC of Raji H , PMN were preincubated with 25 M diphenyleneiodonium chloride (DPI; Sigma-Aldrich), an inhibitor of neutrophil NADPH oxidase (32), for 5 min at room temperature and added to assays without washing, giving a final concentration of DPI of 6.3 M. Assays were performed in RF10 ϩ medium in the presence of 50 U/ml GM-CSF and started by adding 50 l of chromium-labeled target cells, resulting in a final volume of 200 l. After 3 h of incubation (37°C with 5% CO 2 ), assays were stopped by centrifugation, and 51 Cr release from triplicate samples was measured. The percentage of cellular cytotoxicity was calculated using the following formula: specific lysis ϭ [(experimental cpm Ϫ basal cpm) Ϭ (maximal cpm Ϫ basal cpm)] ϫ 100%.
Maximal 51 Cr-release was determined by adding perchloric acid to target cells at a final concentration of 3%. Basal release was measured in the absence of sensitizing Ab and effector cells. No Ab-independent cytotoxicity was observed with PMN as effector cells.

Oxidative burst assays
The luminol-ECL method was used for the detection of total (intracellular and extracellular) ROI production. PMN (4 ϫ 10 5 ) from NHD were gently centrifuged (40 ϫ g for 3 min at 4°C) on targets (1 ϫ 10 5 Raji H cells) opsonized with or without Ab (2 g/ml) in PBS supplemented with 10% FCS and 8.8 mM glucose. To inhibit ROI production, PMN from NHD were preincubated with 25 M DPI for 5 min at room temperature and added to assays without washing, giving a final concentration of DPI of 5.7 M. GM-CSF (50 U/ml) and luminol (50 M) were added, giving a final volume of 220 l. Luminol-ECL was measured in triplicates as counts per

Electron microscopy
Transmission electron microscopy was performed as described elsewhere (33). Briefly, 187.1 rat hybridoma cells sorted for high membrane expression of a mAb against a mouse L chain and freshly isolated PMN from a patient treated with G-CSF were incubated at an E:T ratio of 5:1 with or without 2 g/ml mAb 22 (mIgG1), retargeting Fc␥RI on effector cells. Incubation was stopped by adding cold Ito's fixative. The cell suspension was transferred on 6.5-mm diameter, 0.4-m pore-size Transwell filters (Costar) and fixed for 4 h at 4°C. Subsequently, filters were postfixed in reduced osmium, encapsulated in 2% agar, stained with uranyl acetate, dehydrated in a graded series of ethanolic solutions completed by pure acetone, and finally embedded in Epon 812. Ultrathin sections were contrasted with uranyl acetate and lead citrate and examined with a Zeiss EM T109 transmission electron microscope.

E:T cell aggregate formation assays
The Vybrant lipophilic carbocyanine dyes 2- After washing three times with RF10 ϩ medium, effector and target cells were incubated for 0, 30, and 90 min at 37°C at an E:T ratio of 2.5:1 or Ͼ200 min. Conjugate formation was analyzed on 150,000 cells in the presence or absence of sensitizing Ab constructs using an EPICS Profile flow cytometer. Fluorescence of DiO and DiI was measured in FL1 and FL2 channels, respectively. After the exclusion of double negative events, the double fluorescent cells presented in Table I were calculated for each population by multiplying the percentage of each population determined by FSC/SSC with the respective percentage of double fluorescent cells.

Apoptosis assays
For the measurement of programmed cell death, a previously described method was adapted (35). Raji H target cells at 10 6 per ml were stained for

The target Ag and the Ab contribute to effective PMN-mediated ADCC
Ab directed against the target Ags HER-2/neu, HLA class II, or CD19 were analyzed for their ability to induce ADCC of Raji H cells by PMN. As shown in Fig. 1A, PMN-mediated lysis is restricted to distinct target Ags. Significant cytotoxicity occurred with Ab against the target Ag HLA class II and HER-2/neu. Consistent with our previously published data, PMN failed to induce ADCC in the absence of Ab or in the presence of the mAb 4G7c against CD19 (11,30), although the Ag density of HER-2/neu (9.9 Ϯ 1.7, mean fluorescence intensity Ϯ SEM) and CD19 (5.3 Ϯ 1.3, mean fluorescence intensity Ϯ SEM) was in a similar range (n ϭ 10, p ϭ 0.2). The HLA class II-directed mAb F3.3 demonstrated significant tumor cell killing, starting from 0.08 g/ml to 10 g/ml (Fig. 1B).  With the bsAb [A77 ϫ trastuzumab] recognizing Fc␣RI and HER-2/neu, cytotoxicity was observed starting from 0.4 g/ml to 10 g/ml, reaching an optimum at 2 g/ml (Fig. 1B). In both cases specific lysis was found starting from an E:T ratio as low as 10 to 1 (Fig. 1C). In summary, the target Ag, the Ab concentration, and the E:T ratio all play important roles in determining the efficiency of the PMN-mediated ADCC of Raji H cells.

ROI production is not mandatory for PMN-mediated ADCC
The generation of ROI during ADCC against Raji H was monitored by luminol-amplified chemiluminescence (36). We observed Abdependent ROI production of PMN from NHD with Abs targeting HER-2/neu ([A77 ϫ trastuzumab]) or HLA class II (F3.3), but not with the mAb against CD19 (4G7c), which did not mediate ADCC (Fig. 1D). This suggests an involvement of ROI in PMN-mediated ADCC. However, the Ab that mediated the most efficient cytotoxicity (the classical mAb against HLA class II) was not the Ab that induced the highest ROI production (the bsAb using the Fc␣RI of PMN and HER-2/neu on target cells). Therefore, ROI production did not strictly correlate with ADCC activity.
To further investigate the role of ROI production, we compared the PMN from a patient suffering from CGD with the PMN from a NHD in regard to ADCC with Raji H cells. CGD is characterized by the complete lack of ROI generation and, consequently, of oxidative burst activity. As shown in Fig. 2A, the PMN of both the CGD patient and the NHD efficiently killed tumor targets in the presence of the HLA class II-directed mAb or the HER-2/neudirected bsAb. Efficient ROI production was observed with PMN from the NHD stimulated by F3.3-or [A77 ϫ trastuzumab]-coated targets in contrast to the PMN from the CGD patient, which were unable to produce ROI (Fig. 2B). These results were further supported by the inhibition of the neutrophil NADPH oxidase of PMN from the NHD with DPI, which resulted in a Ն95% reduction of ROI production. This had no significant effect on the PMN-mediated specific lysis of Raji H compared with untreated PMN (n ϭ 6) by targeting HLA class II (47 Ϯ 9 and 57 Ϯ 6%, respectively) or HER-2/neu (24 Ϯ 4 and 26 Ϯ 6%, respectively). Therefore, ROI production by the NADPH oxidase of PMN does not appear to be mandatory for the PMN-mediated lysis of Raji H cells.

The perforin inhibitor CMA does not inhibit PMN-mediated ADCC
As previously reported, CMA inhibits the perforin-based pathway of NK cell-and cytotoxic T cell-mediated cytolysis (37). Furthermore, recent data suggest a role for perforins and granzymes in PMN-mediated cytotoxicity (25). On the basis of these findings, we tried to elucidate the role of perforins in PMN-mediated ADCC by inhibition of the perforin-based pathway with CMA (Fig. 3). We used PMN and MNC as effectors targeting the B cell lymphoma ARH-77 via HLA class II, the breast cancer cell line SK-BR-3 via HER-2/neu, and Raji H cells via both HLA class II and HER-2/neu. In the presence of CMA, PMN-mediated lysis of ARH-77 via HLA class II was significantly enhanced using the mAb F3.3 (Fig. 3A). No inhibition by CMA of PMN-mediated ADCC was observed for SK-BR-3 opsonized with the bsAb [A77 ϫ trastuzumab] (Fig. 3C). The PMN-mediated ADCC of Raji H cells remained unaffected by CMA by the targeting of HER-2/neu or HLA class II with bsAb [A77 ϫ trastuzumab] and mAb F3.3, respectively (Fig. 3E). In contrast, perforin-mediated ADCC by MNC was significantly inhibited in the presence of CMA for both target Ags (Fig. 3, B, D, and F). Therefore, we conclude that PMN-mediated ADCC is not dependent on the perforin pathway.

Aggregate formation and membrane dye exchange is induced by Ab
To initiate ADCC, PMN have to adhere to their target cells. As shown by electron microscopy, PMN made intimate contacts with Ab-opsonized tumor cells (Fig. 4, A-E). Importantly, PMN did not adhere to target cells in the absence of an opsonizing Ab (Fig. 4F), indicating that an Ab is required for aggregate formation. These data were further supported by flow cytometry showing clusters between PMN and tumor cells. We adapted a commonly used assay for the quantification of aggregates (38). The plasma membranes of PMN effector and Raji H cells were labeled with the fluorescent carbocyanines DiO or DiI, respectively. Labeled cells were cocultured in the presence or absence of a sensitizing Ab and analyzed for cluster formation at the time points 0, 30, and 90 min (Table I and Fig. 5). In FSC/SSC dot plots three populations were to be observed. Population 1 with high SSC and low FSC and population 2 with low SSC and FSC represent PMN and Raji H cells, respectively. Clustered cells were found in population 3 characterized by their increased FSC/SSC. After coculture for 30 or 90 min, an ϳ10-fold increase in the number of aggregates was found in the presence of bsAb, whereas no significant cluster formation was to be observed without a bsAb (Table I). In addition to the aggregates, after 30 min significantly more double fluorescent PMN (population 1) or target cells (population 2) appeared in cocultures with bsAb compared with assays without bsAb. The percentage of double fluorescent PMN further significantly increased from 30 to 90 min, indicating an ongoing recruitment of PMN for ADCC. A small increase of double fluorescent cells was also observed for PMN and Raji H independently of bsAb (Table I and Fig. 5).
In Fig. 5 we show a more detailed evaluation of the cocultures. High percentages of double fluorescent PMN-target cell clusters (92-96%) were to be observed in population 3 of the aggregated cells in the presence of bsAb [A77 ϫ trastuzumab] (Fig. 5, C and  E). Cocultures without bsAb displayed only low percentages of double fluorescent cell clusters. Furthermore, both Raji H and PMN showed a tendency for homotypic aggregation in the absence of bsAb (Fig. 5, A, B, and D). Interestingly, the single effector cells (population 1) and the target cells (population 2) contained more double fluorescent cells after coculture in the presence of bsAb (Fig. 5, C and E) than without bsAb (Fig. 5, B and D).
The time course of effector-target cell clustering analyzed by flow cytometry is shown in Fig. 6. For this purpose, labeled cells were cocultured in the presence of the bsAb [A77 ϫ trastuzumab] at 37°C and the double fluorescent cells in each population were repeatedly measured at short time intervals. Directly after mixing, double fluorescent cell clusters were detected. After reaching a maximum between 30 and 60 min, the number of double fluorescent clusters slowly decreased. In contrast, double fluorescent single effector or target cells were Ͻ1% for the first 10 min of coculture. Subsequently, the number of double fluorescent PMN and target cells increased. The percentage of double fluorescent target cells (population 2) remained stable and lower than that observed for  PMN. Coculture at 4°C revealed bsAb-dependent formation of cell clusters, but no membrane dye exchange occurred (not shown). We conclude that heterotypic clustering precedes mutual membrane lipid exchange between effector and target cells during ADCC.
To visualize aggregate formation and membrane dye exchange, confocal laser scanning microscopy was performed (Fig. 7). DiIlabeled Raji H and DiO-labeled PMN were detected as red and green cells, respectively. Labeled PMN and Raji H cells that had mutually exchanged fluorescent membrane lipids appeared yellow in the merged image. Morphologically, effector and target cells could be distinguished by their characteristic nuclear features. An increased number of yellow cells were found in the presence of the bsAb [A77 ϫ trastuzumab] compared with control experiments without Ab. Therefore, membrane lipid exchange increased dramatically in the presence of an Ab.

Apoptosis is induced in PMN-mediated ADCC
To elucidate the mechanism of Ab-mediated Raji H cell death by PMN, we adopted a previously described method for the detection of apoptosis (35). Raji H cells were stained with CFSE, washed, and mixed with isolated PMN. CFSE remains stably associated with the nuclei of cells even after hypotonic lysis. In this way, we were able to differentiate the CFSE-stained Raji H nuclei from unlabeled PMN nuclei during FACS analysis. After 6, 24, and 40 h of coculture in the presence or absence of the bsAb [A77 ϫ trastuzumab], cells were stained with hypotonic PI-Triton X-100. The PI fluorescence of CFSE-tagged nuclei was measured and CFSE-positive nuclei with sub-G 1 DNA content were considered being apoptotic target cell nuclei. In ADCC assays with an E:T ratio as low as 2.5:1 significant apoptosis was detected in the presence of the bsAb [A77 ϫ trastuzumab] 24 and 40 h after coculture. Apoptosis was even more pronounced in assays with an E:T ratio of 20:1 (Fig. 8). Simultaneously, the number of intact target cell nuclei (G 0 -G 1 ) decreased. As expected for GM-CSF-exposed PMN, apoptosis ranged from 30 to 50% from 24 to 40 h irrespective of the presence of bsAb. We found no indication for the aggregation of nuclear fragments from osmotically lysed apoptotic Raji H cells or PMN. If we had not been able to differentiate apoptotic effector and target cell nuclei we would expect to observe much higher levels of apoptotic Raji H nuclei. In long term culture for 6 days, medium consumption of Raji H cells ceased in the presence of bsAb at an E:T ratio of 20:1, indicating almost complete target cell death.

Discussion
Several factors affect the efficiency of PMN in ADCC. Thus, the specificity, concentration, and format of Ab, as well as E:T ratio, density, and domains of target Ag, contribute substantially to effective Ab-mediated tumor cell killing by PMN (3,7,11,12,19,21,30). As previously shown, HER-2/neu and HLA class II proved to be appropriate target Ag of PMN-mediated lysis of Raji H cells. In contrast, no ADCC was mediated via the CD19 Ag despite a similar surface density to HER-2/neu (30). To get insight into the mechanisms of ADCC, we used Raji H cells as targets and freshly isolated PMN as effector cells. We demonstrated efficient recruitment of effectors by HER-2/neu-directed F(abЈ) 2 bsAb or HLA class II-directed IgG mAb at low E:T ratios and Ab concentrations. Activated PMN produce ROI; however, there are contradictory reports about the role of ROI in PMN-mediated ADCC (39,40). ROI production of PMN from NHD was triggered by mAb to HLA class II or bsAb to HER-2/neu, but not by CD19 mAb. This observation may implicate an Ag restriction in oxidative burst activity as previously described for ADCC (11). However, ROI production did not strictly correlate with ADCC activity. Thus, the oxidative burst targeting of HER-2/neu with bsAb was clearly stronger than the targeting of HLA class II with IgG mAb, whereas the relative effectiveness of ADCC was vice versa. More important, PMN from a CGD patient completely lacking ROI production mediated equally efficient ADCC like the PMN from NHD targeting HLA class II and HER-2/neu. Furthermore, the inhibition of ROI production in PMN from NHD had no effect on ADCC against Raji H cells. These data are in accordance with data from Katz et al. (39) that demonstrated a role for ROI in ADCC exclusively against erythrocytes but not against lymphoid target cells. From our experiments we conclude that NADPH-dependent ROI production is not mandatory for PMNmediated ADCC against Raji H .
To elucidate the contribution of perforin to PMN-mediated ADCC we used CMA, an inhibitor of the vacuolar type H ϩ ATPase of perforin-based cytotoxicity. CMA inhibits the acidification of lytic granules and the activation of perforin in NK cells, resulting in reduced NK cell-mediated cytotoxicity (28). However, in experiments with PMN from NHD as effector cells, we found that CMA had no or even a stimulatory effect on PMN-mediated ADCC (Fig.  3). CMA induces a rise in intracellular Ca 2ϩ via inhibition of the vacuolar type H ϩ ATPase (41). Increasing [Ca] i activates PMN and may thus enhance PMN-mediated ADCC. This indicates that either PMN-mediated ADCC is not perforin-mediated or that perforin-release is regulated differently in PMN and NK cells.
To get insight into the interactions of PMN with their target cells during ADCC, we used transmission electron microscopy, confocal laser-scanning microscopy, and flow cytometry. Recently, cellcell fusion between cancer cells and myeloid cells has been described (42), and cell-cell fusion was detected as double fluorescent events in flow cytometric studies (38). Interestingly, during intimate contact formation we observed cytoplasmic communication between effector and target cells (Fig. 4, D and E) in the presence of mAb. This cytoplasmic communication may specifically characterize the "immunological synapse" formed between PMN and target cells during ADCC, which will be referred to as the "cytotoxic synapse." ␤ 2 integrins are described to be necessary for the formation of the immunological synapse and to redistribute into membrane rafts. In this way they may isolate the small cytoplasmic communications we have observed from the remaining cell membrane, preventing the progression to cell fusion. Additionally, in our experiments video-assisted microscopy of living cells (not shown) and confocal laser-scanning microscopy revealed the formation of target-effector cell conjugates and enhanced membrane dye transfer in the presence of bsAb (Fig. 7) without evidence of cell-cell fusion.
Transient aggregate formation between effector and target cells was detected by flow cytometry immediately after mixing in the absence of bsAb. This observation suggests that cells make rapid contacts by accidental collision. In the presence of bsAb persistent aggregate formation was observed (Fig. 6). Thus, the bsAb targeting Fc␣RI and HER-2/neu establishes a long-lasting tethering between effector and target cells by interaction with the Fc receptor and the tumor cell Ag, respectively. Adhesion molecules such as integrins on the PMN and intercellular adhesion molecules on the surface of tumor cells may further contribute to the stable formation of the "cytotoxic synapse" (36,43). In the absence of bsAb we detected homotypically aggregated Raji H cells (population 3, upper left quadrants in Fig. 5, A-C) that have been described as adhering to each other and exchanging membrane fragments spontaneously (44). After bsAb-mediated adhesion of PMN to Raji H , heterotypic aggregates formed and the number of exclusively DiIpositive aggregated Raji H cells decreased. From our data we cannot differentiate whether PMN disrupted the homotypic aggregates of Raji H cells to form heterotypic aggregates or whether PMN were recruited to the homotypic aggregates of Raji H .
Our data demonstrate a small increase in the number of double fluorescent solitary PMN and Raji H over time that is independent of bsAb, implying membrane dye exchange during some random collisions between effector and target cells (Figs. 5 and 6 and Table  I). However, in the presence of bsAb significantly more double fluorescent single PMN or Raji H cells were to be detected. This may be explained by: 1) prolonged contacts between effector and target cells as shown for retargeted T cells (45); 2) the generation of chemoattractants for PMN by tumor cells after opsonization with bsAb; or 3) the activation of the mobility or adhesiveness of the PMN by triggering Fc␣RI with bsAb. For Raji H , a significant bsAb-dependent uptake of membrane dye from PMN was observed at 30 min but not at 90 min (Table I). This most likely reflects [A77 ϫ trastuzumab]-mediated early target cell death, which was previously shown to require the interaction between the Fc receptors on PMN and target cells for the induction of apoptosis (7). As shown in Fig. 8, we were also able to detect apoptosis induction in Raji H cells in the presence of both bsAb and PMN. According to our previously described results in breast cancer cells (7) and our current experiments with lymphoma cells (Fig. 8), we suggest that ADCC induced by PMN involves the induction of apoptosis in the target cells, which is supported by our current finding that the inhibition of caspases in target cells decreases HER-2/neu-directed cytotoxicity. However, cytotoxicity targeting HER-2/neu was not completely blocked by a broad spectrum caspase inhibitor, which suggests more than one pathway leading to PMN-mediated cytotoxicity against Raji H . The observed caspase-independent lymphoma cell death targeting HLA class II is in accordance with previous data (46) and suggests an important role for alternative apoptosis pathways. Caspase-independent induction of apoptosis involving NF-B inhibition or the mitochondrial flavoprotein apoptosis-inducing factor has recently been described (47,48). Therefore, we suppose the involvement of caspase-independent death pathways, although we cannot completely exclude the possibility that the induction of necrosis might also play a role in 51 Cr release assays.
Chemical structure and lipophilic character enable DiI and DiO to insert into the lipid bilayers of the plasma membrane of intact cells where they display a free lateral diffusion. In general, the membrane dyes do not leak and move from cell to cell (34). Thus, the observed membrane dye exchange may represent a mutual transfer of membrane fragments between PMN and tumor cells, similar to trogocytosis. Trogocytosis was initially described in T cells, NK cells, and APC. It is induced after the activation of ITAM-containing surface molecules like the BCR or TCR (49). The Ab dependence of membrane lipid exchange in our experiments is in accordance with the binding of the bsAb to Fc␣RI and subsequent ITAM-mediated activation of the effector cells. Fc receptors as well as BCR or TCR are members of the multichain immune recognition receptor family signaling via ITAM. Furthermore, the temperature dependence of dye exchange indicates that membrane exchange is an active and energy-dependent process. Thus, our observation of membrane lipid exchange during neutrophil-mediated, Ab-dependent tumor cell lysis may be closely related to trogocytosis.
Because successful ADCC has always been associated with an intimate membrane interaction of PMN and target cells and with a mutual exchange of membrane lipids, both processes are supposed to mechanistically contribute to PMN-mediated ADCC and apoptosis induction. In conclusion, we suggest that the mutual membrane transfer between PMN and tumor cell targets that occurs during the formation of the "cytotoxic synapse" is involved in the induction of programmed target cell death.