Antigen-Specific Cytolysis by Neutrophils and NK Cells Expressing Chimeric Immune Receptors Bearing ζ or γ Signaling Domains

TCR and FcγR are structurally and functionally related immune receptors expressed on T lymphocytes and myeloid/NK cells, respectively, which activate a variety of critical biologic activities including cytolysis, cytokine/inflammatory mediator release, and phagocytosis. One of the primary immune defense mechanisms in the control of disease is killing of malignant and virally infected cells by both TCR and FcγR-bearing effectors. TCR and FcγR are comprised of two functional components: the first for target cell/Ag recognition via binding to MHC-presented peptides (TCR) or Abs (FcγR); the second for activation of signal transduction pathways that culminate in effector function. The latter function is mediated primarily by two structurally related subunits, TCR-ζ and FcεRIγ. Whereas the ζ-chain is primarily associated with the TCR complex, the FcεRI γ subunit (γ) is primarily associated with FcγRI expressed on neutrophils and macrophages (1, 2, 3) and the TCR of γδ T cells (4). In addition, both ζ and γ subunits have been found associated with FcγRIIIA expressed on NK cells and macrophages (5, 6). The γ and ζ subunits contain one and three copies, respectively, of the conserved 18-amino acid Ig tyrosine activation motif (ITAM).⁶ Although the tyrosine and leucine residues present within the ITAM sequence [YXXL-X(7)-YXXL] are conserved between the two subunits, ζ and γ, the XX and flanking amino acids differ (7). The tyrosine residues of the conserved YXXL are required for signal transduction (8). TCR and FcγR cross-linking initiates a signal transduction cascade which begins with activation of Src and Syk family protein tyrosine kinases (PTKs) and phosphorylation of ITAM tyrosine residues and culminates in cellular activation (7, 8, 9, 10, 11, 12, 13, 14).

Chimeric immune receptors (CIRs) have been described in which the ligand-binding domain of a heterologous receptor or single-chain Ab is fused directly to the cytoplasmic domain of a member of the ζ family (15, 16, 17, 18). In this way, MHC-unrestricted CIRs directed against a variety of tumor-associated or viral-derived ligands may be generated. We have previously described an HIV-specific CIR composed of the extracellular domain of CD4 fused to ζ (18). CD4 binds the gp120 moiety of the HIV envelope protein (HIVenv) expressed on the surface of virally infected cells. When CD4ζ is introduced into mature human CD8⁺ T cells (18) or NK cells (19) via retroviral-mediated transduction, cytotoxic function of the TCR- or FcγR-bearing effector cell populations can be efficiently and specifically directed to kill gp120-expressing cells in vitro. More recently, we have described the generation of T cell-independent systemic immunity in SCID mice reconstituted with CD4ζ-expressing myeloid and NK cells following bone marrow transplantation (20). Such Ag-specific CIRs provide an ideal model system with which to investigate immune cell function and develop novel approaches for the treatment of disease. Although much is known about the role of ζ in TCR-bearing cells, the relationship between ζ/γ structure and FcγR-mediated effector function is not well understood. In this report, we compare ζ- and γ-based CIRs for their ability to activate the cytolytic function of two distinct classes of FcγR-bearing effector cells, neutrophils and NK cells. Since fully differentiated NK cells may be propagated in vitro, ζ/γ-CIRs were introduced into mature human NK cells. In contrast to NK cells, neutrophils are short-lived with a half-life of ∼18 h in vitro. Mature neutrophils were therefore generated by 1) in vitro differentiation of retrovirally transduced human CD34⁺ cells or 2) in vivo differentiation of retrovirally transduced murine hemopoietic stem cells posttransplantation. Long term expression of both ζ- and γ-CIRs on reconstituted myeloid and NK cells is observed following murine bone marrow transplantation. The ζ-CIR is consistently more efficient than the γ-based receptor at mediating target-specific cytolysis by both NK cells and neutrophils, even though the latter primarily utilize the γ-chain for FcγR signaling.

The CD4ζ chimeric receptor has been previously described (18), and bears the extracellular and transmembrane (TM) domains of the human CD4 receptor (residues 1–372 and 372–395 of the mature CD4 chain, respectively), fused to the cytoplasmic domain of the human TCR ζ-chain (residues 31 to 142 of the mature ζ-chain). CD4γ has the same extracellular and TM domains as CD4ζ, but the cytoplasmic domain is derived from the human FcεRI γ-chain (residues 27–68 of the mature γ-chain). The CD4^del gene is a truncated form of the CD4 receptor in which the cytoplasmic residues have been deleted (residues 403–433 of the mature CD4 receptor deleted; Gln at position 403 is mutated to a stop codon). Oligonucleotide-directed deletional mutagenesis was used to form specific junctions between gene sequences.

A high efficiency retroviral transduction system, kat, was used to generate high titer retroviral supernatants containing the CD4ζ, CD4γ, or CD4^del gene, from 293 cells transiently transfected with packaging (pkat) and retroviral vector (rkat) plasmids as previously described (21). rkat43.2 is a variant of rkat4 (21) in which Moloney murine leukemia virus (MMLV) Psi sequences (up to the ATG of Gag which has been changed to TAG) have been replaced by the corresponding sequences from Moloney Murine Sarcoma Virus (MMSV). Viral Env sequences between the CIR gene region and the reverse strand binding site have been deleted. SV40 poly(A) and ori sequences have been inserted 3′ of the retroviral sequences on a pSK11 (Stratagene, La Jolla, CA) plasmid backbone. Following reverse transcription and integration into target cells, transcription of both unspliced and spliced mRNAs initiates from the viral long terminal repeat (LTR). rkat43.3pgk was generated from rkat43.2 by 1) replacing the XhoI-EcoRI splice acceptor fragment with a PCR-generated 520-bp XhoI-EcoRI fragment encoding the human phosphoglycerate kinase (pgk) promoter/enhancer (nucleotides 5–516, GenBank accession number M11958) and 2) replacing the 3′ MMLV LTR with an LTR in which the sequences from PvuII-XbaI (nucleotides 7935–8113, GenBank accession number MLMCG) have been deleted. Following reverse transcription and integration into target cells, transcription of CD4ζ, CD4γ, or CD4^del is initiated only from the pgk promoter of rkat43.3pgkF3. Previous work in this laboratory has shown that this vector yields stable levels of gene expression over 6 mo in transplanted mice, whereas viral LTR-driven expression diminishes rapidly over 1 to 2 mo (our unpublished data).

pkat2ampacUTd was derived from pkat2ampac (21), by deleting untranslated sequences 3′ of the envelope gene: the 172-bp ClaI-NheI fragment of pkat2ampac was replaced with a 100-bp PCR-generated ClaI-NheI fragment containing only the 3′ coding regions of MMLV 4070a to give pkat2ampacUTd. pkat2ecopac was derived from pkat2ampacUTd by replacing the 4-kb SalI-ClaI pol-env-encoding fragment of pkat2ampacUTd with the analogous fragment from ecotropic MMLV (GenBank accession number MLMCG).

Retroviral supernatants were generated essentially as previously described (21). Briefly, TSA54 cells were plated at 6.5 × 10⁵/10-cm plate and 48 h later cotransfected with 10 μg of rkat43.2 or rkat43.3pgk retroviral vector encoding the relevant CIR and 5 mg of packaging vector: pkat2ampacUTd for transduction of human CD34⁺ or NK cells; and pkat2ecopac for transduction of murine bone marrow. The medium was changed after 15 to 24 h and replaced with either NK, CD34⁺, or bone marrow culture medium as described below. After an additional 24 h, the culture supernatant was harvested, filtered through a 0.45-μm pore size filter, frozen on dry ice, and stored at −70°C. Retroviral titers on National Institutes of Health 3T3 cells ranged from 6 × 10⁶ to 1 × 10⁷ viral particles/ml.

Retroviral transduction of NK3.3 cells was conducted with the kat retroviral producer system as previously described (19) except that supernatant instead of coculture transduction was used. NK cells were plated at 1 × 10⁶ cells/well of a 24-well plate in 1 ml of NK medium (19) and exposed to 1 ml of retroviral supernatant and 2 μg/ml Polybrene. After 18 to 24 h, 1 ml of medium was removed and replaced with 1 ml of fresh supernatant and Polybrene. The transduction procedure was then repeated twice more at 18- to 24-h intervals, and the NK3.3 cells were allowed to recover for an additional 48 h in NK medium. Stable expression of CD4ζ and CD4γ chimeric receptors was then analyzed 2 days posttransduction by flow cytometry with FITC-conjugated anti-CD4 mAbs as described below. CD4ζ/γ-expressing NK cells were subsequently purified by immunoaffinity anti-CD4 mAb-coated flasks (Applied Immune Sciences, Santa Clara, CA).

Human bone marrow was obtained from consenting healthy adult volunteers by aspiration from the posterior iliac crest. Light density mononuclear cells were isolated using Ficoll-Hypaque (Pharmacia, Piscataway, NJ) density centrifugation within 24 h of harvest. Isolation of CD34⁺ progenitors was performed using the CellPro Ceprate System (CellPro, Bothell, WA). After purification, CD34⁺ cells were prestimulated for 2 days in a CD34⁺ basal medium of Myelocult Long Term Culture Medium (Stem Cell Technologies, Vancouver, British Columbia, Canada) with the addition of 10⁻⁶ M hydrocortisone, human stem cell factor (SCF) (100 ng/ml), human IL-3 (50 ng/ml), and human IL-6 (10 ng/ml). Cells were then cultured at 8 × 10⁵ cells/ml in the presence of freshly collected retroviral supernatant with the addition of 8 μg/ml Polybrene, 100 ng/ml human SCF, 50 ng/ml human IL-3, and 10 ng/ml human IL-6. Transduction was conducted in 10-cm dishes precoated with 5 μg/ml anti-human CD44 and 5 μg/ml anti-human CD49d mAb (Coulter, Westbrook, ME) to increase transduction efficiency. After 4 h, cells were collected, washed, and resuspended in viral supernatant for a second exposure. After 8 h, cells were collected, washed, and resuspended in Myeloid Long Term Culture Medium with the addition of growth factors for liquid expansion into neutrophils. Two days after retroviral infection, relative transduction efficiency was determined by flow cytometric analysis for CD4 expression.

CD34⁺ cells were expanded and differentiated into neutrophils utilizing the following schedule of cytokine treatment: during days 1 to 6, the cells were cultured in human SCF (100 ng/ml), human IL-3 (50 ng/ml), and human IL-6 (10 ng/ml). On days 7, 9, and 11, fresh human SCF (10 ng/ml) and human granulocyte CSF (G-CSF) (2 ng/ml) were added to the cultures. On days 14 and 16, cells were fed with G-CSF (10 ng/ml) alone. On days 18 to 22, cells were harvested, analyzed via cytospin preparations and flow cytometry, and utilized in chromium release cytotoxicity assays as described below.

C.B-17 scid/scid (SCID) mice (Charles River Laboratories, Wilmington, MA) were utilized for bone marrow transplant studies. Bone marrow donors were 8- to 16-wk-old males and females, and recipients were 8- to 12-wk-old males. Animals were housed in sterile laminar airflow hoods and fed ad libitum with sterile food and water. All animal procedures conformed to institutional guidelines. Donor SCID mice were injected via the tail vein with 5-fluorouracil (100 mg/kg) (Roche Laboratories, Nutley, NJ) to enrich for immature hemopoietic progenitors with long term repopulating ability. After 6 days, mice were euthanized by CO₂ asphyxiation. Femurs were harvested and flushed with bone marrow culture medium (DME, 4.5 g/L glucose, 15% FCS, glutamine, penicillin, and streptomycin) and 5 mM EDTA. Low density cells (LDC) were isolated by density gradient separation using Lympholyte-M (Cedar Lane, Hornby, Ontario, Canada). Briefly, bone marrow cells were layered over an equal volume of gradient and spun at 2200 rpm for 20 min at 20°C in a table top centrifuge. Interphase cells were collected, washed, and resuspended in culture medium. LDC were plated at 3 × 10⁶ cells/well of a 6-well plate (Corning Glass, Corning, NY) and exposed to retroviral supernatant in the presence of 6 mg/ml Polybrene. Virus was packaged in ecotropic envelope for efficient transduction of murine cells. After 2 h, the medium was hemidepleted, and fresh viral supernatant was added. Transduced cells were harvested from the plates after 4 h, washed, and resuspended in 0.9% normal saline and 0.1% BSA for injection. Via tail vein injection, 10⁶ transduced LDC/mouse were infused into sublethally irradiated (350 rads) SCID mice.

Heparinized blood (300 μl) was attained by retroorbital bleeds of CIR transplanted and control mice at various time points posttransplant. RBC were depleted by ammonium chloride lysis. Approximately 2 × 10⁵ cells were incubated with the following murine-specific mAbs conjugated to FITC: anti-GR-1; anti-Mac-3; anti-5E6 (PharMingen, San Diego, CA), in addition to anti-human CD4⁻ phycoerythrin (PE) (Becton Dickinson, San Jose, CA) according to manufacturer’s instructions. FITC- and PE-conjugated isotype-matched mAbs served as negative controls. Stained cells were analyzed on a FACScan cytometer (Becton Dickinson).

At 3 wk posttransplant, 4 CIR-transduced and 4 control SCID mice were treated with 7 daily s.c. injections of human G-CSF (Amgen, Thousand Oaks, CA) (100 μg/kg/day). Mice were then euthanized, and heparinized blood was recovered by cardiac puncture. Neutrophils were isolated using a modification of the standard “1-step Polymorph” procedure (Accurate Chemical, Westbury, NY) in which 0.8 ml of 1.5% NaCl was added to 10 ml of stock gradient. Murine PB (4 ml) was layered over 4 ml of modified gradient, and tubes were spun for 30 min at 450 × g at 20°C in a table top centrifuge. The neutrophil fraction was removed and washed twice. An aliquot was removed for cytospin analysis and was >95% neutrophils as judged by standard Wright-Giemsa staining techniques. Additional cells were removed for measurement of CIR transduction efficiency by flow cytometric analysis. The remaining cells were used in a chromium release cytotoxicity assay as described below.

Cytotoxicity of transduced neutrophil and NK effectors was evaluated by a standard ⁵¹Cr release method. Raji cells derived from Burkitt’s lymphoma were used as targets, either as parental (Raji-p) or those transfected with HIVenv, Raji-env (19). Raji-p and Raji-env cells were labeled with sodium [⁵¹Cr]chromate (100 mCi/10⁶ cells) for 3 h at 37°C and washed three times. Labeled target cells were then incubated with effector cells as described below. The percentage of specific lysis was calculated from triplicate samples using the following formula: [(CMP-SR)/(MR-SR)] × 100, where cpm is the counts per min released by targets incubated with effector cells, MR is the counts per min released by targets lysed with 100 μl of 1% Triton X-100, and SR is the counts per min released by targets incubated with medium alone. In vitro-differentiated human neutrophils were harvested at day 21 post cytokine treatment and cultured overnight in culture medium with the addition of 10 ng/ml human G-CSF and 500 U/ml human γ-IFN. The next day, cells were analyzed via cytospin preparations and flow cytometry, and utilized in a ⁵¹Cr release cytotoxicity assay as described above. Briefly, human neutrophils (control and CIR expressing) were resuspended in 100 μl of assay medium (RPMI, 10% FCS, 20 ng/ml human granulocyte-macrophage-CSF) and plated in triplicate in 96-well plates. To each well 10^{5 51}Cr-labeled targets were added in 100 μl of assay medium to establish a range of E:T ratios. To measure Ab-dependent cellular cytotoxicity (ADCC), polyclonal rabbit anti-human lymphocyte serum (Cedar Lane) (5 mg/ml) was added to control cells. The assay was initiated by centrifugation (50 × g, 2 min, 20°C) and allowed to incubate at 37°C for 5 h. From each well, 100 μl of supernatant were removed and counted in a gamma counter for the assessment of ⁵¹Cr release. Freshly isolated murine neutrophils were resuspended in 100 μl of assay medium (RPMI, 10% FCS, 10 ng/ml murine granulocyte-macrophage-CSF) at various concentrations and utilized in a ⁵¹Cr release cytotoxicity assay as described for human neutrophils.

The structure of the primary immune receptors utilized by T cells and myeloid/NK cells, the TCR and FcγR, respectively, is depicted in Figure 1. HIV gp120-specific CIRs were constructed by fusing the cytoplasmic domains of either the TCR-ζ or FcRγ-signaling chains to the extracellular and TM domains of the human CD4 receptor. Both CD4-CIRs are expressed as monomers since they do not possess the cysteine-bearing TM domain of the ζ- or γ-chain. Furthermore, by virtue of the CD4 TM domain, the ζ/γ chimeras are expressed independently of endogenous ζ/γ-chains (Refs. 15 and 16 and data not shown). Although we used HIV-specific CIRs as a model system in this study, CIRs can be generated against a variety of cell surface Ag on virally infected or malignant cells by utilizing single-chain derivatives of mAbs with the appropriate specificity (17, 18).

CD4ζ and CD4γ CIRs were introduced into NK cells, murine bone marrow, or human CD34⁺ progenitors using the kat retroviral vector transduction system previously described for high efficiency gene transfer into CD8⁺ T lymphocytes (21). High titer retroviral supernatants were generated by transient cotransfection of the retroviral vectors rkat 43.2 or rkat 43.3pgk with amphotropic or ecotropic packaging plasmids as described in Materials and Methods. Following reverse transcription and integration into target cells, transcription initiates from the viral LTR of rkat 43.2 (Fig. 2,A) and the internal phosphoglycerate kinase (pgk) promoter of rkat43.3pgk (which has a disabled 3′-LTR) (Fig. 2 B).

The human NK clone NK3.3 has the phenotypic and functional characteristics of primary NK cells (22), exhibiting both natural and FcγR-mediated ADCC. Such NK clones can be readily expanded ex vivo and are therefore suitable for evaluating the expression and activity of ζ- and γ-CIRs. NK3.3 cells were transduced with rkat43.2 bearing either CD4ζ or CD4γ as described in Materials and Methods. Although both CIRs were efficiently expressed on the cell surface following transduction, the intensity of expression for CD4ζ was routinely twofold higher than CD4γ despite equivalent levels of gene transfer (data not shown). Subsequent flow cytometric sorting resulted in two CD4γ populations: the first with CD4 levels equivalent to the CD4ζ NK population (CD4γ^low); the second with CD4 levels approximately twofold higher than the CD4ζ population (CD4γ^high) (Fig. 3 A). Expression of endogenous FcγRIIIA in the transduced populations was unaffected by coexpression of either ζ- or γ-CIRs (data not shown).

We have previously described the ability of CD4ζ-modified NK cells to kill HIVenv-expressing target cells as efficiently as CD4ζ T cells (19). In this study, we compared the CD4ζ, CD4γ^low, and CD4γ^high NK populations shown in Figure 3,A for their ability to mediate cytolysis in standard killing assays. Target cells were either unmodified or modified Raji cells expressing low levels of the CIR-specific ligand, HIVgp120. CIR-mediated cytolysis was compared with FcγR-mediated ADCC in each case (Fig. 3,B). As shown previously (19), the level of CD4ζ killing exceeded that of ADCC, with 55% cytolysis reached at E:T ratios of 10:1 for CD4ζ (Fig. 3 B) and over 100:1 for ADCC (data not shown). In contrast, the level of cytolytic activity mediated by CD4γ^low NK cells was similar to ADCC. CD4γ^low NK cells were also markedly less efficient than CD4ζ NK cells expressing similar CIR levels, with 20% lysis attained at E:T ratios of 10:1 for CD4γ^low compared with only 0.07:1 for CD4ζ. Despite the higher CIR levels expressed on the CD4γ^high NK population, cytolysis was still less efficient than for CD4ζ effectors, with 20% lysis requiring an E:T ratio of 12.5:1. In summary, the ζ-CIR is markedly more efficient than either endogenous FcγRIIIA or the γ-CIR at mediating NK cytolysis, even when expressed at lower levels than the latter receptor. These data show that the γ-chain is inherently less active than the ζ-chain at triggering cytolytic function of this NK effector population.

Neutrophils are the most numerous immune cell type, constituting approximately one-half to two-thirds of all circulating human white blood cells. The bone marrow is responsible for maintaining a constant supply of neutrophils at the remarkable rate of ∼1 × 10¹¹ per day (23), each of which is terminally differentiated and programmed to die within 1 to 2 days. To evaluate the ability of ζ- and γ-bearing immune receptors to initiate and specifically direct human neutrophil cytolytic activity, CD4ζ or CD4γ CIRs were introduced into human CD34⁺ cells in vitro by retroviral transduction with rkat43.2. Transduced progenitors were analyzed by two-color flow cytometry with mAbs specific for CD4 (to detect CIR expression) and the pan-leukocyte marker, CD45. As shown in Figure 4,A, ∼70 and 49% of CD45⁺ cells expressed CD4ζ or CD4γ, respectively. The transduced populations were subsequently expanded and differentiated into neutrophils in liquid culture containing appropriate cytokines. After 16 days of exposure to cytokines, ∼25 and 8% of CD15⁺ neutrophils derived from the transduced progenitor population expressed CD4ζ or CD4γ, respectively (Fig. 4 B). In addition to the decrease in the percentage of CD4⁺ cells detected, the intensity of CIR expression per cell decreased during differentiation. Southern analysis of the CD4ζ- and CD4γ-transduced neutrophil populations revealed that the frequency of CIR-marked cells was similar in both cases (approximately three copies per cell; data not shown), suggesting that the differential change in CD4ζ and CD4γ surface expression observed during neutrophil maturation from progenitors is most likely posttranscriptional in origin.

The transduced and unmodified neutrophils shown in Figure 4,B were subsequently tested in cytotoxicity assays against parental Raji cells and Raji cells expressing gp120. Figure 5 shows the relative cytolytic activity of the CD4ζ- and CD4γ-expressing populations, with the E:T ratio corrected for differences in the absolute percentage of CIR-expressing neutrophils in each case. The data show that the CD4ζ receptor in particular was able to efficiently activate cytolysis by the human neutrophil effector population against the Raji-env targets. The level of CD4ζ-mediated lysis was very impressive (specific lysis above background was seen at E:T ratios as low as 1.5:1) and was even more efficient than ADCC at E:T ratios of <25:1. ζ/γ-CIR-mediated killing was highly ligand specific as shown by the absence of detectable Raji-p killing. Unmodified neutrophils exhibited no significant activity above background against Raji-env or Raji-p cells. CD4ζ neutrophils exhibited higher levels of target-specific lysis than the CD4γ population, with 25% maximal lysis by CD4ζ neutrophils at E:T ratios of 25:1 as compared with only 5% maximal lysis by CD4γ neutrophils at E:T ratios of 5:1. In summary, the ζ-CIR is able to direct the cytolytic activity of human neutrophils in a highly target-specific manner and with impressive efficiency and is more efficient than the γ-CIR. However, correction of E:T ratios for differences in the absolute percentage of CD4ζ and CD4γ expression may still underestimate CD4γ activity. We therefore went on to compare both CD4ζ and CD4γ expression and function in the murine transplant system described below.

The data described above demonstrate that CIRs are expressed and functionally active in neutrophils differentiated in vitro from retrovirally transduced human CD34⁺ progenitors. To confirm that such observations are also relevant to mature neutrophils derived in vivo, we went on to evaluate CIR expression and function following transplantation of SCID mice with CIR-transduced syngeneic bone marrow. As in humans, SCID mice rapidly reconstitute donor-derived myeloid and NK lineages posttransplant. We have previously shown that SCID bone marrow/hemopoietic stem cells (HSC) transduced with ζ-CIR successfully differentiate into CIR-expressing myeloid and NK cells in vivo, giving rise to murine neutrophils with ζ-CIR-directed cytolytic function (20). In the current study, we have extended this observation to include γ-bearing CIRs, thereby enabling subsequent comparison of CD4γ and CD4ζ expression and neutrophil-mediated cytotoxicity.

High titer retroviral supernatants bearing CD4ζ, CD4γ, or a “silent” CD4 receptor in which the cytoplasmic domain is deleted, CD4^del, were generated using the kat retroviral system as described in Materials and Methods. Transduction of mouse bone marrow stem cells was accomplished as described in Materials and Methods using chemotherapy plus cytokine incubation to induce stem cell cycling before transduction. Surface expression of each CD4- CIR on transduced cells preinfusion is shown in Figure 6 A. Although the total number of cells expressing CD4ζ and CD4γ was similar (78 and 62%, respectively), the mean intensity of expression was considerably higher for the ζ-CIR and CD4^del receptor. Quantitative-competitive PCR analysis confirmed the immunofluorescence data and showed similar levels of integrated provirus for the CIR-transduced populations (data not shown). The transduced bone marrow was subsequently transplanted into sublethally irradiated syngeneic mice via tail vein injection.

Reconstitution of the myeloid and NK compartment is generally complete within 15 to 25 days (24). Mice were therefore tested for successful transduction and expression of CD4-CIR as early as 3 wk posttransplantation. After 3 wk, three mice from each of the three transplant groups, CD4ζ, CD4γ, and CD4^del, were treated with G-CSF for 7 days and then killed. In vivo administration of G-CSF results in two- to threefold increases in the neutrophil count of mice as well as humans, thereby facilitating neutrophil analysis. Neutrophils were then purified from peripheral blood by density gradient centrifugation as described in Materials and Methods and subjected to flow cytometric analysis to confirm surface receptor expression. In contrast to the marked difference in CD4ζ and CD4γ expression levels on CD34-derived human neutrophils (Fig. 4,B), the absolute percentage of murine neutrophils expressing CD4ζ and CD4γ is very similar (Fig. 6 B). Similar levels of CD4ζγ expression were also seen in the absence of G-CSF treatment (data not shown). Despite the similar percentage of murine neutrophils expressing either receptor, however, the intensity of CD4γ expression was six- and eightfold lower than that of CD4^del and CD4ζ, respectively.

The CD4ζ, CD4γ, and CD4^del neutrophil populations (Fig. 6,B) were subsequently tested in cytotoxicity assays for their ability to kill Raji-p and Raji-env cells over a range of E:T ratios. Although the absolute level of CD4ζ-mediated cytotoxicity is relatively low in the murine as compared with the human system, it is absolutely specific for Raji cells expressing gp120 (Fig. 7,A). Efficient killing via ADCC was observed for each of the three CD4-CIR neutrophils, with specific lysis of >50% target cells at E:T ratios of >25:1 (data not shown). Although the murine and human ζ-chains exhibit a high degree of homology (25), the human ζ-chain may be less active in murine neutrophils. Furthermore, human cells may be less efficient targets for murine than for human effectors. Alternatively, differences in the absolute levels of CD4ζ/γ activity between the human and murine systems may not be species specific but may correlate with ex vivo vs in vivo modes of neutrophil generation, respectively. A previous study with CD4ζ-transplanted C3H mice yielded even higher levels of Raji-env specific lysis, however, with 17% lysis attained at an E:T ratio of 40:1 (Fig. 7 B). Although some specific CD4γ-mediated cytolysis against Raji-env cells was observed over background, the efficiency was much lower than for CD4ζ and rapidly decreased with E:T ratio. The lower level of CD4γ-mediated cytolytic activity was confirmed in two subsequent experiments (data not shown). As expected, CD4^del neutrophils exhibited no specific cytolysis in these assays. In summary, these data confirm and extend the results described above for CD34-derived neutrophils.

SCID mice transplanted with each of the 3 CIRs were subjected to more extensive immunophenotyping to determine the relative efficiency of γ- vs ζ- expression in both myeloid and NK cell compartments over time. SCID mice possess a higher percentage of NK cells than their immunocompetent counterparts, thereby facilitating phenotypic analysis of this relatively small effector population. Peripheral blood was isolated from reconstituted animals at various times posttransplant and analyzed by two-color flow cytometry using mAbs against human CD4 and the following mouse blood cell lineage markers: GR-1 (expressed on all granulocytes including neutrophils and basophils); Mac-1 (expressed on monocyte/macrophages and neutrophils); or NK-1.1 (expressed on NK cells). Expression of CD4ζ, CD4γ, and CD4^del was detected in granulocytes, monocytes, and NK cells in the peripheral blood of transplanted animals at 16 weeks posttransplant (Fig. 8). The percentage of myeloid and NK cells expressing CD4ζ and CD4^del was similarly high, with averages of 30% for CD4ζ and 40% for CD4^del. In contrast, surface expression of CD4γ was significantly lower, with an average of 12% of myeloid and NK cells expressing this receptor. In addition to the lower number of cells expressing detectable CD4γ, the lower intensity of CD4γ receptors parallels that observed for the transduced human NK- and CD34-derived neutrophil populations described above (Figs. 3 and 4). It is striking that the percentages of GR-1-, Mac-1-, and NK-1.1-positive populations expressing a particular CD4-CIR are almost identical. Since the “neutral” receptor, CD4^del, gives a similar expression profile to CD4ζ/γ, it appears that expression of γ- or ζ-bearing receptors in primitive progenitors early in development does not negatively impact subsequent maturation of myeloid and NK cells. Furthermore, although the overall level of expression is lower for CD4γ than for CD4ζ, no major qualitative differences in expression are seen for γ- as compared with ζ-bearing receptors. Finally, expression of all three CD4-CIRs was stable over at least a 7-month period (data not shown).

FcγRIIIA is the primary FcγR utilized for by NK cells for ADCC (26) and requires associated homo- and heterodimers of ζ- and γ-chains for both surface expression and effector function (5, 6). We have previously shown that CIRs bearing the ζ cytoplasmic domain can efficiently activate NK cells (19). In this report, we have extended these studies to compare the cytolytic capacity of ζ- and γ-CIRs in NK cells, and we show that CIRs possessing the γ cytoplasmic domain also redirect the cytolytic activity of NK cells in an Ag-dependent manner. Presumably, both ζ- and γ-receptors mediate NK cytolytic activity by coupling to the endogenous FcγRIIIA signaling pathway. However, our results reveal that ζ-CIR-mediated cytotoxicity is far more efficient than that triggered by either γ-CIR or FcγRIIIA (ADCC), even when expressed at lower levels than γ-CIR. These results suggest that in NK cells, the ζ-chain has intrinsically more cytolytic activity than γ.

The role of the γ signaling chain in facilitating FcγR-mediated ADCC in neutrophils formed the basis for the studies described in this report in which we compare ζ/γ-CIRs for their ability to activate neutrophil cytolytic function. Of the three classes of FcγRs that neutrophils use to trigger the lytic machinery, FcγRI appears to be the only one that utilizes a member of the ζ/γ family for signal transduction. Specifically, FcγRI requires the γ-chain for assembly and effector functions such as phagocytosis and presumably ADCC (27, 28). In contrast, FcγRIIA possesses an ITAM-like motif within the cytoplasmic domain that mediates cytolysis in the absence of the γ-chain (29), and does not require γ for assembly. The third FcγR expressed on human neutrophils, FcγRIIIB, is anchored to the plasma membrane through a glycosylphosphatidylinositol linkage and is not associated with ζ/γ subunits. Although FcγRIIIB cannot mediate ADCC independently, it may function synergistically with other FcγR (30, 31). Since there is no homologous glycosylphosphatidylinositol-anchored FcγRIIIB in mice, however, murine neutrophils express FcγRIIIA. We show in this report that CIRs bearing the signal transduction domain of the ζ subunit efficiently and specifically direct the cytolytic activity of neutrophils. This finding is intriguing given the fact that human neutrophils primarily utilize the γ subunit for FcγR-mediated cytolysis. To understand the relationship between signal transduction pathways initiated by FcγRs and ζ-chimeras to induce neutrophil cytotoxicity, a number of issues require further clarification including the impact of ζ and γ ITAM structure on PTK binding and activation, the role of individual classes of FcγR in ADCC, and the influence of ligand identity on function.

Although CD4ζ-mediated killing was marginally more efficient than FcγR-driven ADCC by human neutrophils (e.g., twofold difference based on E:T ratio required to achieve 20% lysis), CD4ζ-mediated cytolysis by human NK cells routinely exceeded that of FcγRIIIA-mediated ADCC (e.g., 14-fold difference based on E:T ratio required to achieve 20% lysis). The consistently smaller CIR/FcγR “cytolytic ratio” for human neutrophils as compared with NK cells does not appear to result from major differences in ADCC efficiency between the two effector cell types. Despite the potential for the direct and synergistic contribution of all three classes of neutrophil FcγR in harnessing both γ/ζ-dependent and -independent pathways, ADCC by human neutrophils was equal to or somewhat less efficient than FcγRIIIA-mediated ADCC by human NK cells in our system. The observation that the ζ-CIR functions less efficiently relative to endogenous FcγR for human neutrophils than do NK cells is intriguing. An interesting report potentially relevant to this observation describes how neutrophil-ADCC but not NK-ADCC against Raji cells shows a striking dependency on target ligand identity rather than expression level per se (32).

CD4ζ mediates significantly higher levels of cytolytic activity than CD4γ when expressed in NK cells, even when expressed at lower levels than the latter receptor. Therefore, the ζ signaling chain is inherently more active than γ at harnessing the cytolytic pathway in NK effectors. For neutrophils, however, the significantly lower intensity of CD4γ as compared with CD4ζ expression (e.g. eightfold for murine neutrophils) may also contribute to the lower cytolytic function of CD4γ observed. The structure of the ζ and γ subunits differs in two major ways: 1) quantitatively: ζ and γ contain three and one copies, respectively, of the ITAM; and 2) qualitatively: specific internal and flanking amino acids differ between ζ and γ ITAMs, although the tyrosine and leucine residues are conserved. Therefore, ζ-CIR may be more effective than γ-CIR at NK-mediated cytolysis for quantitative and/or qualitative reasons. Quantitative differences between ζ- and γ-based chimeras are supported by studies showing a direct relationship between the number of ζ-chain ITAMs and the efficiency of cytokine release (33), apoptosis (34), and thymocyte selection (35). However, microheterogeneity in the amino acid sequence between ζ- and γ-chain ITAMs can also result in qualitative differences in signal transduction and effector function. For example, differential binding and activation of Src and Syk family PTKs have been reported for ζ- and γ-chain ITAMs (7, 36, 37, 38, 39), and the γ subunit is approximately six times more efficient than the ζ subunit at mediating phagocytosis (7). Indeed, recent data from our laboratory reveal that CD4γ is far more efficient than CD4ζ at mediating phagocytosis despite the lower level of CD4γ surface expression (M. Wang and M. R. Roberts, unpublished data). Further studies will be necessary to dissect the relationship between ζ- and γ-ITAM structure, Src/Syk family PTK interaction, and cytolytic function.

We have previously shown that ζ-CIR are efficiently expressed on myeloid and NK cells following transplantation of CIR-modified bone marrow (20). In this report, we show that bone marrow modified with a CIR bearing the signaling domain of the FcεRI γ-chain can also give rise to mature FcR⁺ effector cells expressing γ-CIR in vivo. However, ζ-CIR are always expressed more efficiently on the cell surface than their γ-bearing counterparts. This observation is consistent with a study of ζ knockout mice expressing the γ-chain that showed that TCR complexes associated with γ rather than ζ are expressed less well on T cells (40). The reasons for this interesting phenomenon are unclear, but they may involve differences in the half-life of ζ- and γ-derived receptors and/or differences in PTKs associated with them. For example, activation of Lck maintains low TCR expression in immature thymocytes by causing retention and degradation of TCR components (41). Future experiments will test this hypothesis for ζγ-CIR expression in NK cells and neutrophils.

In a previous study, we showed that CD4ζ-transplanted SCID mice were protected from challenge with a lethal dose of a human B cell leukemia (Raji) expressing HIVenv (20). Cytolysis must clearly be an essential component of this in vivo antitumor effect, although the relative contribution of different effector functions (phagocytosis, for example) and classes of myeloid and NK cells remains to be determined. In the current report, we show that ζ-CIR is more effective than γ-CIR at triggering NK- and neutrophil-mediated cytolysis in vitro. However, γ-CIR may be more effective at mediating other therapeutically relevant effector functions. For example, the γ subunit is more efficient than ζ in mediating phagocytosis (7), a finding consistent with preliminary data from our laboratory comparing phagocytic signaling by γ- and ζ-CIR (M. Wang and M. R. Roberts, unpublished data). It will therefore be of interest in future studies to compare ζ- and γ-CIRs for their ability to generate T cell-independent systemic immunity in vivo.

The pleiotropic biologic responses triggered via FcγR make them potentially attractive candidates for directed immunotherapy. Indeed, FcγR-directed bispecific Abs are currently being clinically evaluated for treating cancer and infectious disease (42, 43). Introduction of disease-specific chimeric immune receptors into bone marrow progenitor cells may have significant therapeutic potential for malignant and infectious diseases. Such an approach would provide a constant supply of Ag-specific myeloid and NK cells from gene-marked HSC in vivo and therefore may have advantages over ex vivo modification of terminally differentiated effector cells such as T lymphocytes or systemic delivery of FcγR-directed bispecific Abs. Finally, the ability to readily harness myeloid and NK cell function by genetically modifying HSC and progenitors with Ag-specific chimeric receptors provides a powerful experimental approach with which to study the role of nonlymphoid effector cells in the host immune system.

We thank Drs. Ronald Germain, Harold Malech, and Arthur Weiss for a critical reading of the manuscript and Jacqueline Rothwell for preparation of the manuscript.