Absolute Requirement of C D l l / CD 18 Adhesion Molecules , FcRII , and the Phosphatidylinositol-Linked FcRIII for Monoclonal Antibody-Mediated Neutrophil Antihuman Tumor Cytotoxicity

We have previously shown that 3F8, a murine IgG, monoclonal antibody (MoAb) specific for the ganglioside GD2, mediates tumor cell kill in vitro and in vivo. We now describe receptor requirements of polymorphonuclear leukocytes (PMN) in 3F8-mediated cytotoxicity (ADCC) of human G,(+) melanoma and neuroblastoma cell lines. PMN from a child with leukocyte adhesion deficiency (LAD) were devoid of CD11/CD18 adhesion molecules and mounted no detectable ADCC. MoAb to CDllb, CDllc, and CD18 each efficiently blocked ADCC by normal PMN. In contrast, a panel of different MoAbs to CDl la had no significant inhibitory effect on ADCC, a finding consistent with the low-to-absent expression of the CDl la ligand, intercellular adhesion molecule-1, on the target cells. Granulocyte-macrophage colony-stimulating factor (GM-CSF) significantly increased the expression of CDllb, CDllc, and CD18 on normal PMN, decreased the expression of Fc receptors (FcR), and enhanced ADCC by normal but not by LAD PMN. MoAbs to FcRll and FcRlll each


F8 IS A MURINE IgG, monoclonal antibody (MoAb)
3 that is well-suited for targeted immunotherapy of cancer because (1) its ganglioside G,, target antigen is highly restricted to neuroectodermal tissues and is genetically stable's2; (2) it has excellent tumor localization in patients3; and (3) in vitro, it mediates human tumor cell destruction by human complement and by human lymphocytes, cultured monocytes, and polymorphonuclear leukocytes (PMN).4-7 This constellation of features, which may account for regressions of GD2( +) tumors in patients treated with 3F8,8 has been documented in very few MoAbs. For example, most studies of antibody-dependent cellular cytotoxicity (ADCC) use mononuclear cells as effectors, in part because few MoAbs (clinically applicable or otherwise) are known to mediate ADCC of human tumor cells by PMN.6-9-1' Yet, in vivo, PMN are the predominant circulating leukocyte and are activated and increased in number by cytokines such as granulocyte-macrophage colony-stimulating factor (GM-CSF)."Z'~ Studies showing efficiently blocked ADCC; anti-FcRI MoAb had no effect. Flow cytometry using anti-FcRII MoAb versus anti-FcRIII MoAb did not show cross competition, suggesting that inhibition of ADCC was not a steric effect resulting from FcRll proximity t o FcRIII. PMN deficient in FcRlll (obtained from patients with paroxysmal nocturnal hemoglobinuria) and PMN depleted of FcRlll by treatment with elastase or phosphatidylinositol (PI)-specific phospholipase C produced low ADCC, supporting a role for the PI-linked FcRIII. Thus, optimal ADCC using human PMN, human solid tumor cells, and a clinically active MoAb (conditions that contrast with the heterologous antibodies and nonhuman or nonneoplastic targets used in most models of PMN ADCC) required C D l l b, CDl lc, FcRII, and the PI-linked FcRIII. Furthermore, in this clinically relevant system, GM-CSF enhancement of antitumor PMN ADCC correlated with increased expression of C D l l /CD18 molecules. The cytotoxic mechanisms in these model systems may differ from those operative in PMN ADCC of human solid tumor cells that not only are large in size, and hence may not be readily phagocytosed by PMN, but are also relatively resistant to reactive oxygen The current study focused on receptor requirements of 3F8-mediated PMN ADCC of GDz( +) human melanoma and neuroblastoma cells. Although CD11/CD18 molecules mediate a wide range of adhesion-dependent activities of PMN,= no report to date has described the involvement of this complex in PMN ADCC of human tumor cells. The same holds for Fc receptors (FcR), despite the wellestablished necessity of FcR in ADCC. Of the three reported types of FcR, the low-affinity FcRII and FcRIII predominate on resting PMN.z7 While FcRII has a welldocumented role in many PMN activities, the importance of the more abundant FcRIII, including in ADCC, is uncertain, given its linkage to PMN via a phosphatidylinositol (PI) moiety.% This form of linkage, in contrast to the polypeptide anchor of FcRII, lacks a transmembrane connection to the cytoplasm and may limit the capacity of FcRIII to transmit intracellular signals for PMN function. We now demonstrate a critical role in a clinically relevant model of PMN ADCC, not only for FcRII and all subunits of the CD11/CD18 complex, except CDlla, but also for the PI-linked FcRIII.

MATERIALS AND METHODS
Purified 3F8 (anti-G,,; IgG,) was prepared in our laboratory as described: and was used at a final concentration of 2 p,glmI, unless otherwise stated. AI10 (anti-decay accelerating factor [DAF]) 29   New York, NY). Hybridomas were grown as ascites in BALB/c mice, and partially purified by precipitation in 45% ammonium sulfate at OT, followed by dialysis against phosphate-buffered saline.
PMN were isolated from heparinized peripheral blood by Ficoll-Paque gradients (Pharmacia Fine Chemicals, Piscataway, NJ), followed by 3% dextran separation and hypotonic red blood cell lysis. PMN comprised more than 95% of the resulting cell population by Wright-Giemsa stain. Normal PMN were obtained from healthy laboratory personnel. PMN were also obtained from a 1-year-old girl diagnosed with leukocyte adhesion deficiency (LAD) by Dr Paul J. Leibson of the Mayo Clinic, and referred to Memorial Sloan-Kettering Cancer Center for a bone marrow transplantation procedure to correct life-threatening leukocyte abnormalities. PMN from two patients with paroxysmal nocturnal hemoglobinuria (PNH), diagnosed by Drs Hugo Castro-Malaspina, Lillian Reich, and Wendell F. Rosse, were also used in selected assays.
PI-specific phospholipase C (PI-PLC; from culture supernatants of Bacillus subtilis transfected with the PI-PLC gene from B fhuringiensis) was generously provided by Dr Martin Low (Columbia University, New York, NY). Elastase was from Elastin Products Co (Owenmille, MI). Recombinant human GM-CSF was a gift from Genetics Institute (Cambridge, MA).
Target tumor cells were labeled with sodium 51Cr chromate (Amersham, Arlington Heights, IL) at 100 @/lo6 cells at 37°C for 1 hour. After the cells were washed, loosely bound 51Cr was leaked for 1 hour at 37°C. After further washing, 5 X l@ target cells/well were admixed with effector cells, MoAb, and/or GM-CSF in 96-well polystyrene round-bottom plates (Sarstedt, Germany) to a final volume of 250 pL/well. The

Cell lines.
Effector cells.

Chromium release assay.
plates were incubated at 37°C for 4 hours and then centrifuged at 400g for 5 minutes; the released 51Cr in supernatant was counted in a y-counter (Packard Instrument, Downers Grove, IL). Percentage of specific release was calculated using the formula 100% x (experimental cpmbackground cpm)/(5% sodium dodecyl sulfate [SDS] cpmbackground cpm), where cpm are counts per minute of 51Cr released. Total release was assessed by lysis with 5% SDS (Sigma, St Louis, MO), and background release was measured in the absence of granulocytes. The background was 10% to 25% for LAN-1 and IMR-6 and 7% to 16% for SKMel-1.
Purified PMN were incubated for 20 minutes at 4°C with saturating concentrations of test or control MoAb, washed, reacted with a 1:25 dilution of 1 mg/mL of fluorescein-labeled goat antimouse (IgG and IgM) antibody (Tago, Burlingame, CA) for 20 minutes at 4"C, then washed again, fixed in 0.5% paraformaldehyde, and analyzed on a FACScan (Becton Dickinson).
When assessing changes in PMN phenotypes after treatment with enzymes or with GM-CSF, parallel control studies showed standard deviations (SD) from the mean of 22% to 210% for all antigens, except FcRII, for which the SD from the mean was 212% to 215%. Pvalues were derived using a Student's f-test for paired differences to compare untreated versus treated aliquots of PMN. Examination of the data showed no extreme values that would positively influence this statistic.
Immunofluorescence and flow cytomeoy. Statistical analysis.

RESULTS
Experiments reported elsewherez showing no lysis of "innocent bystander" GD2( -) tumor cell targets suggested a requirement for physical attachment between effector and target cells in 3F8-mediated PMN ADCC, which is a nonphagocytic process (Munn D.H., Kushner B.H., unpublished observations, April 1990). In the current study, results using LAD PMN established the importance of CD11/CD18 molecules in this system. LAD PMN, which were confirmed by immunophenotyping to be devoid of the entire CD11/ CD18 complex, mounted no detectable ADCC ( Table 1). The addition of GM-CSF did not alter this result, even though in parallel assays GM-CSF enhanced ADCC by normal PMN, as previously observed: and increased the expression of CD11/CD18 on normal PMN (see below).
To identify which CD11/CD18 subunits were required for 3F8-mediated PMN ADCC, subunit-specific MoAbs were added to the ADCC assay (Fig 1). MoAbs to CDllb (LM2/1, Mol), CDllc (LeuM5, 3.9), and CD18 (R15.7, IB4, TS1/18) each produced efficient concentration-Requirement for CDll lCD18 adhesion molecules.  dependent inhibition of ADCC, with complete or near complete abrogation of ADCC at 5 to 10 pg/mL. In contrast, three different anti-CDlla MoAbs (TS1/22,2F12, G25.2) in concentrations up to 100 pg/mL had no effect on ADCC. A fourth anti-CDlla MoAb (R3.1) was inhibitory only at concentrations 225 to 50 pg/mL when SKMel-1 was the target, and at concentrations 250 pg/mL when LAN-1 was the target. This lack of effect of anti-CDlla MoAb may be explained by the low expression on melanoma of ICAM-1: which is a ligand for CDlla:6 and the absence of ICAM-1 on LAN-1, which was previously reported4 and which we documented using flow cytometry and the anti-ICAM-1 MoAb R1.l (not shown).
We found variability in the ADCC inhibitory efficiency of different anti-CDll/CD18 MoAb, a phenomenon described with non-ADCC adherence functions of PMN.47 Among the anti-CDll/CD18 MoAbs assayed in our system, the anti-CD18 MoAbs were the most efficient inhibitors (R15.7 more than IB4 and TS1/18 ; Fig l), and the anti-CDllb MoAbs LM2/1 and Mol were more inhibitory than OKM-1 and OKM-10 (not shown). Thus, OKM-1 in concentrations up to 100 pg/mL produced less than 20% inhibition of ADCC against LAN-1, and inhibited ADCC against SKMel-1 by 50% or more only when present at 12.5 pg/mL or greater. In comparison to LM2/1 or Mol, OKM-10 was less inhibitory when LAN-1 was the target, but equally inhibitory when SKMel-1 was the target. The results with OKM-1 and, to a lesser degree, with OKM-10 suggested that nonspecific steric effects did not account for the sensitivity of ADCC to the anti-CDll/CD18 MoAbs. This point was reinforced by the normal ADCC that occurred in the presence of the anti-P-2-microglobulin MoAb BBM.l in concentrations up to 100 pg/mL (not shown).
Blocking MoAbs were used to identify the roles of the different FcR in 3F8-mediated PMN ADCC (Fig 2). The anti-FcRI MoAb 197 in concentrations up to 100 pg/mL had no effect on ADCC. This result was not unexpected because FcRI is present at very low density on PMN.27 In contrast, the anti-FcRII MoAb IV.3 and the anti-FcRIII MoAb 3G8 each completely abrogated ADCC at 1 pg/mL or less. Flow cytometry studies with anti-FcRII MoAb versus anti-FcRIII MoAb did not show cross competition (not shown), confirming prior reports of individual anti-FcR MoAb specifi~ity,".~:'~ and suggesting that steric factors could not account for the respective ADCC blocking effects of IV.3 and 3G8. Thus, both FcRII and FcRIII appeared to be required for successful ADCC.
Because of evidence that the PI-linked FcRIII of PMN may not be critical for PMN function, including tumor cell kill,5052 studies were performed to assess directly the require-   (Table 2). In comparison to normal PMN, the FcRIII-deficient PMN performed poor ADCC in the presence of 3F8 (Table 2). These findings favored an FcRIII requirement for optimal lysis in this system.
Additional support for this conclusion came from studies using PMN from adults with PNH (Table 3). Immunophenotyping confirmed that these PMN, as reported,% had very low expression of PI-linked surface molecules, including FcRIII and DAF. Similar to findings with PMN enzymatically depleted of FcRIII, PNH PMN were ineffective in ADCC as compared with normal PMN (Table 3).
To gain further insight into the roles of FcR and the CDll/CD18 complex in 3F8-mediated PMN ADCC, studies were performed using GM-CSF. This cytokine enhances PMN ADCc8."~' and has obvious potential clinical relevance. GM-CSF significantly increased CDllb (P < .001), CDllc (P = .007), and CD18 (P = .004) expression and, as expected, enhanced 3F8-mediated PMN ADCC ( Table 4). GM-CSF produced a less than significant decrease in both FcRII and FcRIII expression, but did not appreciably change the density of other PMN surface molecules. The inclusion of GM-CSF in the ADCC assay did not alter the relative inhibitory effects of the anti-CDll/ CD18 MoAbs, although higher concentrations of these MoAbs were required to inhibit ADCC (not shown). Thus, while anti-CDlla MoAbs had no effect on ADCC by GM-CSF-stimulated PMN, cytotoxicity curves (such as those in Fig 1) using MoAbs to CDllb, CDllc, and CD18 were shifted to the right in the presence of GM-CSF. These results were consistent with the possibility that GM-CSF enhances 3F8-mediated PMN ADCC at least partly via upregulation of CDllb, CDllc, and CD18.

DISCUSSION
We have defined PMN surface receptors critical for PMN ADCC of human solid tumor cells. In the presence of 3F8, an MoAb that has antitumor activity in patients: ADCC is dependent on a physical interaction between PMN and GDz( +) human neuroectodermal cells that does not involve phagocytosis but requires most subunits of the CDll /CD18 complex, FcRII, and the PI-linked FcRIII. Several of these features differ from those associated with PMN ADCC in model systems using heterologous antibodies and nonhuman or nonneoplastic target cells. PMN were incubated for 2 hours in medium with or without GM-CSF (2 nglmL). washed, and immunophenotyped or tested in ADCC. In this representative study, GM-CSF upregulated C D l l b, CDllc, and CD18 and enhanced 3F8-mediated PMN ADCC.
For personal use only. on October 23, 2017. by guest www.bloodjournal.org From

KUSHNERANDCHEUNG
The availability of LAD PMN allowed us to establish the importance of the CD11/CD18 complex in 3F8-mediated PMN ADCC. Inhibition studies using specific MoAbs showed that all subunits of the complex, except CDlla, were critical for normal PMN cytotoxicity with 3F8. Given the results with LAD PMN, the abrogation of 3F8mediated PMN ADCC produced by MoAbs to CD18, the p subunit common to all three C D l l a subunits, might have been predicted. The same holds for the inhibitory effects of MoAb blockade of CDllb, which is the most abundant of the C D l l a subunits on PMN and has well-established myeloid cell adhesive functions?6 The efficient inhibition produced by MoAbs to CDllc was less expected in view of the relatively low density of CDllc on PMN. Recent studies, however, suggest that CDllc also can function independently as an adhesion molec~le.'~ Our results suggest that either (1) PMN-target cell attachment via CDllb or CDllc is necessary but not sufficient for 3F8-mediated ADCC to proceed, or (2) in our system, when CDllb or CDllc is blocked, the other becomes nonfunctional. A redundancy of function among the CD11/CD18 molecules has previously been noted, and is supported, for example, by the relative structural homology between CDllb and CDllc.= No prior report has described the involvement of CD11/ CD18 molecules in PMN ADCC of human solid tumor cells (as mediated by MoAbs or by heterologous antibodies). The sole previously reported study of ADCC by LAD PMN targeted herpes simplex virus-infected Chang liver cells. 54 Findings in that study differed from our own in that defective ADCC was attributed to the absence of CDlla and CDllb, because anti-CDlla (TS1/22) MoAb alone or the combination of anti-CDllb (OKM-1) plus anti-CD18 (TS1/18) MoAbs blocked ADCC by normal PMN (anti-CDllb or anti-CD18 MoAb alone produced insignificant inhibition). The lack of effect of anti-CDlla MoAb on 3F8-mediated PMN ADCC is consistent with the low-toabsent expression of the CDlla ligand ICAM-1 on the melanoma and neuroblastoma targets. The known counterreceptors for CDllb and CDllc include iC3b, fibrinogen, factor X, as well as ICAh4-lZ; identification of the ligands for CDllb and CDllc in our system awaits further studies.
Our results with GM-CSF are consistent with an important role for CD11/CD18 in 3F8-mediated PMN ADCC, because GM-CSF enhancement of the latter occurred in conjunction with increased expression of CDllb, CDllc, and CD18. Although GM-CSF can affect the respiratory burst of PMN,55,56 two lines of evidence suggested to us that GM-CSF may act via CD11/CD18 in our system: (1) We had previously found that GM-CSF enhances 3F8-mediated ADCC by PMN with defective oxidative metabolism (obtained from patients with chronic granulomatous disease)." (2) The time-course of GM-CSF enhancement of PMN ADCC that we6 and others" had observed was more consistent with its rapid upregulation of PMN CD11/ CD18,",57*58 than with the more delayed onset of its activation of PMN oxidative In addition to its effects on the CD11/CD18 complex, GM-CSF decreased PMN expression of FcRII and FcRIII, a previously reported phen~menon.~'.~~.'~ Our studies cannot rule out the possibility that GM-CSF enhances 3F8mediated PMN ADCC by non-CDll/CDl8 effects, such as increasing the affinity of FcR or mobilizing nonoxidative cytolytic capacities.
Studies of the role of FcR in 3F8-mediated PMN ADCC underscored the importance of physical PMN-target attachment. Thus, anti-FcRII and anti-FcRIII MoAbs each efficiently inhibited ADCC. Furthermore, PMN enzymatically depleted of FcRIII and PNH PMN, which have low FcRIII expression, were ineffective in ADCC as compared with normal PMN. The results suggest that, as with CDllb and CDllc, the separate involvement of each low-affinity FcR is necessary but not sufficient for optimal 3F8-mediated PMN ADCC. The current report is the first to document a requirement for both FcRII and FcRIII in PMN ADCC against human (or nonhuman) tumor or nonneoplastic targets. Our findings are noteworthy because, although Fc-FcR interactions are the well-established basis of ADCC, the issue has been clouded by the discoveg that FcRIII is anchored to the plasma membrane of PMN by a PI moiety and may therefore have a limited capacity for signal transduction (see below).
A requirement for the PI-linked FcRIII in PMN ADCC is evident in some but not all model systems. Thus, PMN lyse murine hybridoma cells bearing antibody directed against FcRII, but not those bearing antibody against FCRIII?"'~ In contrast, PMN lyse chick erythrocytes coated with antibody directed against FcRII or FcRIII"; heteroaggregates of anti-FcRIII and anti-2,4,dinitrophenyl antibodies mediate PMN destruction of 2,4,6-trinitrophenylmodified RDM4 murine leukemia cells"; and salivary PMN, which have very low expression of FcRIII, are ineffective in ADCC of herpes simplex virus-infected Chang liver cells.61 The latter two reports do not include data relating to other FcR, such as FcRIl expression on salivary PMN or the effects on ADCC of anti-FcRII MoAb.
FcR requirements in PMN ADCC of human tumor cells have not been described before the present report. One attempt to do so was unsuccessful when PMN did not lyse JY human B-lymphoblastoid cells that were precoated with a murine IgG, MoAb.= This failure illustrates the difficulty in achieving ADCC of human tumor cells by unstimulated PMN9,1',14; it also supports the hypothesis that PMN ADCC of human tumor cells, which are too large for phagocytosis by PMN and are relatively resistant to reactive oxygen specie^,^^^^ involves cytotoxic mechanisms different from those in model systems using murine targets.
We have established that FcRIII is one of several PMN receptors required for 3F8-mediated PMN ADCC of human solid tumor (neuroctodermal) cells. It remains to be determined whether the PI-linked FcRIII on PMN is generating an intracellular signal to promote PMN cytotoxicity, or is only serving as a receptor for tight binding to MoAb-coated tumor cells, while signal transduction is effected by FcRII.~' Evidence supporting an active role for the PI-linked FcRIII independent of FcRII includes reports that PMN FcRIII can mediate exocytosis of granule