Natural Killer (NK) Cell–mediated Cytotoxicity: Differential Use of  TRAIL and Fas Ligand by Immature and Mature Primary Human NK Cells

mAbs to CD2 (B67.1, B67.6), CD4 (B66.6), CD5 (B36.1), CD8 (B116.1), CD11b (B43.4), CD14 (B52.1), CD15 (B40.9), CD56 (B159.5 [21, 22]), CD161/NKR-P1A (B199.2) used for cell purification (23), and the TNF-α– neutralizing mAb (B154.2, B154.7 [24]) have been reported previously. mAbs to CD3 (OKT3), CD21 (THB5), CD34, CD32 (IV.3), CD64 (32.2), and the irrelevant P3x63.Ag8.653 Ig were produced from cells from the American Type Culture Collection (Rockville, MD). The Fas-blocking ZB4 mAb and the Fas-triggering CH11 mAb (25) were from Upstate Biotechnology, Inc. (Lake Placid, NY). The goat anti–mouse Ig used for panning was produced in our laboratory, adsorbed on human IgG, and affinity-purified on mouse Ig Sepharose before use (26).

Immature lineage negative (Lin⁻) progenitor cell populations were enriched, as described (23), from umbilical cord blood samples provided by Dr. R. Depp (Department of Obstetrics and Gynecology, Thomas Jefferson University Hospital, Philadelphia, PA). These were >99% CD3⁻/CD161⁻/ CD16⁻/CD56⁻ (indirect immunofluorescence [23]), not cytotoxic against K562 target cells, and did not express CD16 mRNA (reverse transcription [RT]-PCR) or CD161 (23; and data not shown). Their differentiation to NK cells was induced upon coculture with the murine bone marrow stromal cell line Sl/Sl⁴ hSCF²²⁰ (Sl/Sl⁴ [27]; provided by Dr. D. Williams, University of Indiana, Indianapolis, IN) and rIL-2 (50 U/ml; Hoffman-LaRoche, Nutley, NJ, through the Biological Response Modifiers Program, National Cancer Institute, Bethesda, MD) or rIL-15 (10 ng/ml, specific activity 2.95 × 10⁸ U/mg; Immunex Corp., Seattle, WA) during a 20–30-d culture period (23, 28). CD161⁺/ CD56⁻ immature and CD56⁺ mature NK cells were purified (23) from CD3⁻/CD161⁺ cells from 25–30-d cultures with IL-2. Homogeneous populations of CD3⁻/CD56⁺/CD16⁺ NK cells were also obtained by negative selection from short-term, 10-d cultures of umbilical cord blood lymphocytes with 30 Gy–irradiated RPMI-8866 cells (22). Each population contained >98% cells expressing the appropriate phenotype (direct immunofluorescence).

As described (10), Fas, DR4, and DR5 extracellular ligand-binding domain–Fc fusion proteins were produced in human 293 cells transiently transfected with constructs encoding these proteins. Conditioned media were used as source of the soluble proteins. Recombinant His 6–tagged TRAIL was produced in bacteria and purified by affinity chromatography on Ni²⁺ affinity resin.

Cell-mediated cytotoxicity was analyzed in 4- or 6-h ⁵¹Cr-release assays using, respectively, the Fas⁻ K562 and the Fas⁺/DR4⁺ Jurkat cells (J32 clone) as targets (5 × 10³/well) and a series of effector cell numbers; LU were calculated at 45% specific ⁵¹Cr release (21). When indicated, 1 mM EGTA and 2 mM MgCl₂, the Fas-blocking ZB4 or the irrelevant P3x63.Ag8.653 mAb as control (each 1 μg/ml), the Fas-triggering CH11 mAb (0.2 μg/ml), the TNF-α–neutralizing mAb (ascites, 1:500 dilution), His 6-tagged TRAIL (supernatant, 1:400 dilution), and DR4– and DR5–Fc fusion proteins (supernatants, 1:2 dilution) were present throughout the assay. When indicated, effector cells were preincubated (5 × 10⁶ cell/ml, 18 h, 37°C) with rIL-2 (50 U/ml) and rIL-12 (2 ng/ml, specific activity 4.5 × 10⁶ U/mg protein in an IFN-γ–inducing assay; provided by Dr. S. Wolf, Genetics Institute, Andover, MA). J32 cell death was analyzed in a 6-h ⁵¹Cr-release assay with the indicated reagents present throughout the assay. rTNF-α (specific activity 10⁷ U/mg protein on L929 cells, α subline), was provided by Dr. J.S. Price, Cetus Corp., Emeryville, CA.

RT-PCR was performed as described (23) on total RNA extracted from the different cell populations (5 × 10⁴ cells/sample) with RNAzol (Biotecx Laboratories, Houston, TX). FasL and TRAIL primers (produced at the Kimmel Cancer Center Nucleic Acid Facility) used for RT were, respectively, TGGTTGCCTTGGTAGGATTGGGC (5′) and GAGCTTATATAAGCCGAAAAACG (3′) (29), and AACCTCTGAGGAAACCAT (5′) and TTAGCCAACTAAAAAGGC (3′). β-actin primers were published previously (23). RT-PCR products were identified on agarose gels after ethidium bromide staining.

NK cells induced to differentiate from Lin⁻ umbilical cord blood cells in cultures with IL-15 mediated lower levels of cytotoxicity against K562 target cells than those mediated by cells from cultures with IL-2 (Fig. 1,a), but cells from either culture condition, such as mature adult peripheral or umbilical cord blood NK cells (26, 30), mediated similar levels of cytotoxicity after stimulation with IL-2 and IL-12. In both cases cytotoxicity was restricted to the CD3⁻/CD161⁺/CD56⁺ NK cell population, the only one detected in cultures with IL-15 (our unpublished data; Fig. 1, b and c). An additional immature CD161⁺/CD56⁻ NK cell population generated in cultures with IL-2 neither produces IFN-γ (23) nor mediates cytotoxicity against K562 cells (Fig. 1,c) although, like mature NK cells, cells in this population contain perforin and enzymatically active granzyme B (23) and form conjugates with K562 cells (not shown). The inability of this cell population to kill K562 cells likely reflects a defect in either expression or granule exocytosis–triggering function of one or more NK cell receptors activating cytotoxicity or of other accessory molecules involved in it. Interestingly, however, cells in the same population reproducibly mediated significant levels of cytotoxicity against the Jurkat cell line (Fig. 1 c).

Killing of K562 cells by umbilical cord (Fig. 2,a) and adult peripheral blood NK cells (31; data not shown) is completely Ca²⁺ dependent, granule exocytosis mediated; that of the Fas⁺ Jurkat cell line (25) is only partially so (Fig. 2,b, using umbilical cord blood as effector cells; not shown with adult PBLs). In this case, activation of the Ca²⁺- and granule release–independent apoptotic death pathway(s) depends on engagement of Fas by FasL, expressed or induced at significant levels on the effector cells upon target cell interaction (18, 31). Thus, the effector cells function primarily to present target cells with the appropriate configuration of the ligand(s) (FasL) capable of binding and cross-linking membrane receptor(s) (Fas) that initiate the target cell apoptotic signals. The Ca²⁺-independent cytotoxicity mediated against Jurkat cells by NK cells from cultures with IL-15 or IL-2 (Fig. 2, c and d, respectively) was inhibited, as expected, in the presence of Fas-blocking mAb, and no greater inhibition was observed upon further addition of TNF-α–neutralizing mAb. Thus, analogous to the observations with freshly purified mature umbilical cord blood and adult peripheral blood NK cells (31), Ca²⁺-dependent granule exocytosis and Fas/FasL-dependent cytotoxic mechanisms account for most cytotoxic activity of mature NK cells from Lin⁻ cells under the culture conditions analyzed.

Interestingly, the immature CD161⁺/CD56⁻ NK cells from cultures with IL-2 (Fig. 2,e) also lysed Jurkat cells, although less efficiently than their mature CD56⁺ cell counterparts from the same culture. As expected based on lack of detectable granule exocytosis upon target cell recognition (23), Jurkat cell lysis by these cells, unlike that by mature NK cells, was essentially Ca²⁺ independent (Fig. 2,e). Immunofluorescence analysis revealed no detectable FasL, whereas RT-PCR detected constitutive FasL mRNA expression in all of the NK cell populations (not shown), in agreement with findings in mature adult human NK cells (19) and murine NK cell clones (18). However, Fas-blocking mAb did not inhibit killing of Jurkat cells by the immature cells (Fig. 2 e), suggesting that these use soluble or membrane-bound molecules distinct from FasL to mediate this effect.

CD161⁺/CD56⁻ immature NK cells express TNF-α mRNA constitutively (23), and most produce TNF-α upon stimulation (our unpublished observations). Two lines of evidence exclude a significant role of TNF-α, either secreted (32) or in membrane-bound form (3), in Jurkat cell death: (a) no significant ⁵¹Cr release above background was detected from Jurkat cells after a 6-h incubation with up to 10³ U/ml rTNF-α (not shown); and (b) no inhibition of immature NK cell–mediated Jurkat cell killing was detected in the presence of a TNF-α–neutralizing mAb, alone or with added Fas-blocking ZB4 mAb (Fig. 2 e). Similar results with cells from cultures with IL-2 (containing both mature and immature NK cells) confirmed that the TNF-α/TNF-R pathway does not participate significantly in NK cell–mediated Jurkat cell lysis under the experimental conditions used, and ligands other than TNF-α on the immature NK cells bind target cell receptors distinct from TNF-R and Fas but sharing their apoptosis-inducing activity.

Unlike K562 cells (13; not shown), Jurkat cells are sensitive both to Fas-mediated (25; not shown) and to soluble TRAIL-mediated cytotoxicity (13). Engagement of the TRAIL receptors DR4 and DR5 (10, 11) by soluble TRAIL results in apoptosis (33, 34). The J32 Jurkat cell clone used, like most clones of this cell line, undergoes apoptotic cell death upon binding either soluble TRAIL or Fas-triggering mAb (Fig. 3). TRAIL-induced ⁵¹Cr release was inhibited to background levels by a mixture of soluble TRAIL-Rs (DR4 and DR5), but not by soluble Fas or Fas-blocking mAb. Conversely, Fas-induced ⁵¹Cr release using Fas-triggering mAb was inhibited by the same Fas-blocking mAb, but not by the mixture of TRAIL-R–Fc fusion proteins. The lack of inhibition of the anti-Fas mAb–induced cell death by soluble Fas likely reflects inefficient competition of this molecule for the anti-Fas binding site on the mAb. These data establish that functional Fas and TRAIL-R1 and 2 (DR4 and/or DR5) are expressed on this Jurkat cell clone with their expected and exclusive specificity.

TRAIL mRNA expression was detectable (by RT-PCR) in the immature NK cells, although TRAIL protein was undetectable in DR4/DR5 binding analysis (data not shown). To determine whether engagement of functional TRAIL-R by TRAIL, possibly expressed or induced to significant levels on these cells upon target cell interaction, elicits Ca²⁺-independent target cell death, we analyzed the effect of DR4/DR5–Fc fusion proteins on cytotoxicity of mature and immature NK cells against Jurkat cells. Like the Ca²⁺-independent lysis mediated by mature NK cells from short-term (10-d) cultures of primary PBLs (Fig. 4,a), that mediated by CD161⁺/CD56⁺ mature NK cells from cultures with IL-15 (not shown) was reproducibly inhibited by the Fas-blocking mAb, but not by the DR4/DR5–Fc fusion proteins. Addition of these to the anti-Fas mAb in the presence of EGTA did not result in inhibition greater than that mediated by the mAb alone. However, at the same concentrations, these reagents significantly inhibited J32 cell lysis by immature NK cells (Fig. 4 b; >75%, as quantitated based on LU in the cell populations). This indicates that the molecules transducing apoptotic signals in this case are either or both DR4 and DR5, thus implicating TRAIL in the cytotoxicity mediated by this cell subset. No additional inhibition was observed under the same conditions with ZB4 mAb added, excluding a major role, if any, for the Fas/FasL-dependent cytotoxic pathway by immature NK cells. The possibility that FasL on the immature NK cells is induced, and used, at later times after interaction with the appropriate target cells is unlikely, given the relatively fast kinetics of FasL expression both in T and NK cells. However, formal proof awaits results of experiments using additional target cells in longer assays. The possibility that other (yet to be identified) molecule(s) with ligand binding specificity identical to TRAIL and in addition to it are involved in Jurkat cell killing by the immature NK cells is unlikely, based on a recent report showing anti-TRAIL Ab–mediated inhibition of Jurkat and melanoma cell lysis by cytotoxic human CD4⁺ T cell clones (35).

The data reported here provide the first direct evidence for the use of functional TRAIL, or a molecule with identical DR4/DR5 binding specificity, by primary human NK cells, and point to this as a marker of differentiation of these and possibly other cytotoxic cell types. Our results also indicate a switch in the surface molecules used for target cell recognition and/or lysis during NK cell differentiation. TRAIL–TRAIL-R interactions represent the first major (but possibly not exclusive) mechanism of target cell lysis used by NK cells to mediate their functions as innate effectors of defense, and this ability is acquired by NK cells at a specific, phenotypically identifiable stage of differentiation (exclusively CD161⁺) that precedes acquisition of the ability to mediate granule exocytosis– or FasL-dependent cytotoxicity (23; and this paper) and to produce IFN-γ (23).

Interesting and testable questions are raised by these data related to whether (a) the same molecules are used by NK cells at different stages of differentiation to kill other target cells; (b) TRAIL and FasL are also markers of differentiation of other cytotoxic cells, or of specific functional capabilities acquired by activated mature cytotoxic effector cells or lymphocyte subsets, phenotypically similar to the immature CD161⁺/CD56⁻ NK cells and present in peripheral blood (23). This possibility is suggested by preliminary data (RT-PCR, not shown) indicating TRAIL mRNA expression in mature NK cells; and (c) like granule exocytosis (for a review, see reference 36), both FasL- and TRAIL-dependent cytotoxicity are negatively regulated by MHC-binding receptors inhibiting cytotoxicity. TRAIL-Rs, expressed on numerous cell types, have been shown to be functional on transformed, but not normal cell-derived cell lines (37). Definition of the expression and functionality of these receptors on primary tumor cells and hematopoietic cells at distinct stages of differentiation or activation/proliferation is essential to establish whether the use of TRAIL rather than FasL by immature NK cells, or even (a) phenotypically identical subset among mature NK cells (23), may endow these cells with (relative) specificity in the control of tumor and possibly immature hematopoietic cell growth.

We are grateful to Dr. R. Depp and the staff of the Obstetrics and Gynecology Department of Thomas Jefferson University Hospital for providing the umbilical cord blood samples, Dr. C.-C. Liu for critical discussions and suggestions, Mr. B. Abebe for technical assistance, and Ms. M. Hoffman for editing.

Supported in part by United States Public Health Service grants CA37155, CA45284 (to B. Perussia), AG14357 (to E.S. Alnemri), and T32-CA09683 (to I.M. Bennett), and by the Charlotte Greyer Foundation (E.S. Alnemri). L. Zamai was the recipient of a Dottorato di Ricerca in Scienze Morfologiche Umane (University of Bologna, Italy) from the Italian Ministry of Research and University.