The mechanism of activation of NK-cell IFN-γ production by ligation of CD28

Molecular Immunology 36 (1999) 361±372

The mechanism of activation of NK-cell IFN-g production by ligation of CD28 Jason C. Cheung a, Crystal Y. Koh a, Brian E. Gordon b, Julie A. Wilder a, 1, Dorothy Yuan a,* a

Laboratory of Molecular Pathology, Department of Pathology, University of Texas Southwestern Medical Center, Dallas, TX 75235, USA b Department of Comparative Medicine, Carolina Medical Center, Charlotte, NC 28232, USA Received 16 November 1998; received in revised form 25 January 1999; accepted 2 March 1999

Abstract We have investigated the mechanism by which anti-CD28 antibodies activates IFN-g production by murine NK cells. These studies reveal that engagement of CD28 alone by this antibody is a poor activator of this cytokine response. E€ective stimulation requires simultaneous ligation of the receptor for Fc (FcgRIII, CD16) which on its own is also a poor inducer of murine NK cells. The mechanism by which immobilized anti-CD28 increases IFN-g mRNA abundance involves both upregulation of transcription as well as induction of mRNA stabilization. However, the elevation of transcription is not as evident as that induced by IL-12 which, in contrast, does not induce message stabilization. Thus ligation of CD28 in the presence of IL-12 results in a synergistic increase in production of the cytokine. Using this assay we have also determined that immobilized anti-CD28 cannot induce resting NK cells to produce IFN-g. In contrast, the same cells can be induced by BCL1C11 tumor cells that express high amounts of the CD28 ligand, B7-2. These studies provide important insights into the ability of cells bearing counter-receptor for CD28 to activate NK cell-cytokine production in vivo. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: NK cells; CD28; IFN-g mRNA stability; FcgRIII; IL-12

1. Introduction CD28 has been shown to play an essential role as an accessory molecule for stimulation of T cell-cytokine production. Whether this receptor serves a similar role on NK cells is not clear. Thus, the original documentation of its presence on murine NK cells claimed that anti-CD28, in the presence of PMA can trigger cytotoxicity, growth, as well as cytokine production (Nandi et al., 1994). On the other hand, in the B7-1 or B7-2 mediated triggering of NK cell cytokine production other cell surface interactions were found to * Corresponding author. Tel.: +214-648-4107; fax: +214-6484033. E-mail address: [email protected] (D. Yuan) 1 Present address: Department of Pathology, University of New Mexico Health Science Center, Albuquerque, NM 87131, USA

be necessary suggesting that ligation of CD28 by its counter-receptor alone is not sucient (Hunter et al., 1997). In addition, there is lack of agreement on whether CD28 is expressed on resting NK cells (Nandi et al., 1994; Hunter et al., 1997). Resolution of these questions is important for understanding how NK cells can be activated by targets such as B cells, antigen presenting cells and tumors which express the appropriate counter-receptors. Therefore we have further investigated the ability of anti-CD28 antibodies to activate both resting and IL-2 propagated NK cells. We found that soluble anti-CD28 is a poor inducer of cytokine production by IL-2 propagated NK cells, even in the presence of PMA. However, when immobilized the antibody is a potent stimulator of IFN-g production. This stimulation is blocked by the inclusion of the F(ab)'2 fragment of anti-CD28 as well as antibodies to the FcgRIII on NK cells. Crosslinking of only the

0161-5890/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 1 6 1 - 5 8 9 0 ( 9 9 ) 0 0 0 5 1 - 6

362

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

FCgRIII on NK cells is, however, not sucient because both soluble as well as immobilized antiFcgRIII is a poor stimulator of IFN-g production. Therefore activation by anti-CD28 requires crosslinking of both the FcgRIII and CD28. We have dissected the molecular mechanism by which immobilized anti-CD28 activates IFN-g production and have compared this mode of induction with that mediated by IL-12 or tumor cells. The results of these experiments provide important insights into the manner by which CD28 functions as an accessory molecule for induction of cytokine production in the absence of other co-stimulatory ligands that may be presented on inducer cells. We have also investigated the response of freshly isolated NK cells to stimulation by anti-CD28 and compared it to that by a tumor, BCL1-C11 cells. These results reveal di€erences between mechanism of activation of IFN-g production by resting vs activated NK cells. 2. Experimental procedures 2.1. Animals C.B-17SCID mice were raised in our own colony from a breeding pair originally obtained from Dr Michael Bennett (UT Southwestern Medical Center). C.B-17  C57BL/6 F1 SCID were bred in the Carolina Medical Center, Charlotte, NC. Cells from these strains were used interchangeably since no di€erence was found for the ability of either to produce IFN-g in response to various stimuli. 2.2. Cytokines, antibodies and FACS analysis Recombinant human IL-2 was provided by the Biological Resources Branch of the Biological Response Modi®ers Program, Division of Cancer Treatment/NCI. Recombinant murine IL-12 was obtained from Genetics Institute, Cambridge, MA. Anti-CD28 (Nandi et al., 1994) anti-FcgR (2.4G2, Unkeless, 1979) anti-TCRa/b (Kubo et al., 1989), antiNK1.1 (PK136, Koo and Peppard, 1984), Rat antiIFN-g (R4-6A2, Spitalny and Havell, 1984), and control antibodies were puri®ed from hybridoma culture supernatants using Gamma Bind (Pharmacia, Fine Chemicals, Piscataway, NJ) according to manufacturer's suggestions. The F(ab) '2 fragments of antiCD28 were prepared as described by Yu et al. (1996). CTLA4-Ig was provided by Bristol-Myers Squibb Pharmaceutical Research Institute (Seattle, WA). FITC-conjugated anti-NK1.1 (PK136) and PE-conjugated anti-CD69 was purchased from PharMingen Research (San Diego, CA). Mouse anti-rat antibodies

were purchased from Jackson Immunoresearch, (Westgrove, PA). Cells were stained as previously described (Yuan and Dang, 1997) and analyzed on the FACscan ¯ow cytometer (Becton-Dickinson, San Jose, CA). 2.3. Preparation of non-stimulated and IL-2 propagated NK cells SCID splenocytes were used as a source of nonstimulated NK cells. SCID bone marrow cells were propagated in IL-2 as previously described (Tutt et al., 1986) to produce LAK (lymphokine-activated killer) cells. As mice bearing the SCID mutation are virtually devoid of B and T lymphocytes, after 6 days of culture the remaining cells were routinely found to be >95% NK1.1+ and CD5 ÿ . Cells were used between 6 and 10 days of initiation of culture. 2.4. Stimulation of SCID splenocytes or IL-2 propagated NK cells Cells were cultured in RPMI-1640 supplemented with 10% FCS and 200 units/ml of IL-2 at 1±2  106 per ml with stimulatory antibodies. To immobilize antibodies, 24 or 96 well Falcon plates were incubated with 10±20 mg per ml of antibodies diluted in PBS. After 4 h incubation at 238C, the antibodies were removed and plates were washed with PBS. Alternatively cells were co-cultured with equal numbers of BCL1-C11 cells (Brooks et al., 1983) as previously described (Michael et al., 1989). Where indicated PMA (Sigma Chemical, St Louis, MO) was added at 18±20 ng/ml. The optimal concentration of IL-12 used (5 ng/ml) was established by titrating the e€ect of addition into NK cultures stimulated with plate bound anti-CD28. DRB (5,6-dichloro-1-b-D-ribofuranosylbenzimidazole, Sigma Chemical, St Louis, MO) was added at a ®nal concentration of 0.1 mM. Titration experiments showed, by run-on analysis, that this concentration inhibits transcription of b2M as well as GAPDH and had little e€ect on cell viability over the time period used. 2.5. IFN-g g measurement Levels of IFN-g protein in culture supernatants were measured by sandwich IFN-g ELISA and using recombinant IFN-g in a standard curve as previously described (Wilder and Yuan, 1995). Only values obtained from the average of duplicate or triplicate wells which did not vary by more than 20% are reported.

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

2.6. RNA isolation and probes for S1 and nuclear runon analysis Total cellular RNA was isolated from frozen cell pellets by the method developed by Feramisco et al. (Feramisco et al., 1982). Methods for generation of S1 probes have been previously described (Yuan et al., 1995). Brie¯y, the S1 probes for IFN-g and b2 microglobulin (b2M) were generated by annealing 32P end-labeling speci®c oligos (IFN-g: 5 'AGCTGGTGGACCACTCGGAT, b2M: TGGGT GGCGTGAGTATACTT) to plasmid DNA containing each cDNA followed by primer extension and digestion of the product at a unique enzyme site within the plasmid to yield a fragment of 220 and 166, respectively. The fragments were isolated on a 6% denaturing gel. Protection of appropriately spliced mRNA yields 140 and 98 nt protected species respectively. Plasmids which were immobilized for nuclear run-on analysis have been previously described (Wilder and Yuan, 1995). They contain inserts as follows: (1) b2M: 643 bp cDNA from b2-microglobulin mRNA; (2) GAPDH: 1233 bp complete rat cDNA from glyceraldehyde dehydrogenase mRNA; (3) IFN-g: 643 nucleotides of the murine IFN-g cDNA; (4) 2B4: 1.9 kb complete cDNA cloned from 2B4 mRNA; (5) IL2R-b: 350 bp PCR ampli®ed fragment from the published sequence of IL2-Rb. Control plasmid was pBluescript II KS from Stratagene, (La Jolla, CA). 2.7. Nuclear run-on analysis Run-on analyses were performed as previously described (Yuan and Tucker, 1984). After cell culture with or without stimuli, NI cells were harvested, washed and counted. Equivalent numbers of cells were lysed with 0.01% NP40. Intact nuclei were labeled with 50±150 mCi [32P]a-UTP for 15 min at 308C. Total TCA precipitable radioactivity obtained from each labeling varied from 1±2  106 cpm. The RNA was then extracted and hybridized to nitrocellulose blots slotted with various DNA plasmids containing inserts of interest (5 mg/plasmid). Hybridization proceeded at 428C for 4 days. The ®lters were dried, exposed to phosphor screens, analyzed and quanti®ed by using the PhosphorImager and the Imagequant software package (Molecular Dynamics, Sunnyvale, CA). For each sample, the density derived from hybridization to control plasmid (less than 2% of the GAPDH signal) was ®rst subtracted from the density obtained from hybridization to each plasmid which was subsequently normalized to the GAPDH signal.

363

2.8. S1 analysis S1 analysis was performed as previously described (Yuan et al., 1995). Dried gels were either exposed to a phosphorimager screen and quanti®ed using the Imagequant software package (Molecular Dynamics, Sunnyvale, CA). Data are expressed as relative hybridization which is equal to the density of the protected IFN-g probe/density of the protected b2M probe. 2.9. Semi-quantitative RT-PCR analysis Total RNA was primed with a combination of Oligo (dT)15 and Random Primer (Promega, Madison, WI) and reverse transcribed with M-MLV Reverse transcriptase (GibcoBRL, Grand Island, NY). RTPCR was carried out with a set of oligos speci®c for IFN-g (forward: 5 'GAATGCATCCTTTTTCGCC; reverse: 5 'GTGGCATATAGATGTGGAAGA) resulting in ampli®ed products of 203 bp from cDNA and 298 bp from genomic DNA often present in the RNA samples. GAPDH primers were: forward: 5 'CACCATGGAGAAGGC; reverse: 5 'TGCCAGTGAGCTTCC. Ampli®cation condition was 948C for 3 min followed by 25 cycles of 948C for 45 s, 588C for 1 min, and 728C for 2 min and ended with an extension cycle of 728C for 7 min. By varying the cycle time, 25 cycles were determined to fall within the logarithmic increase phase for both primer sets. In addition, titration of selected cDNA samples at various dilutions were made to ascertain that after 25 cycles of ampli®cation the amount of each product is a linear function of the amount of cDNA. Comparison of the relative abundance of IFN-g mRNA determined by either S1 analysis or semi-quantitative PCR yielded similar values for the same samples. In order to minimize di€erences in RNA or cDNA preparation, equal aliquots of each sample were always analyzed at the same time with the two sets of primers so that each values obtained for the IFN-g mRNA could be normalized to the control mRNA. For easier quanti®cation of products one of each primer set was end labeled to high speci®c activity with [32P]-gATP and added to the reaction mixture. PCR products were run on 6% acrylamide gels of 1.5% agarose gels and appropriate species were quanti®ed after exposure to phosphor screens. 3. Results 3.1. Stimulation by anti-CD28 requires ligation of both CD28 and Fcg gRIII on NK cells In initial experiments conducted to investigate the mechanism by which cytokine production is upregu-

364

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

Table 1 Immobilized anti-CD28 is a more e€ective stimulator of NK cells IFN-g produced (unit/ml)a Experiment No.

Media

Soluble anti-CD28

Soluble anti-CD28+PMA

Immobilized anti-CD28

1 2 3

0.2 0.2 0.1

0.1 2.5 0.1

1.4 2.5 1.7

6.5 6.6 2.9

a 1±2  106 IL-2 propagated NK cells were incubated with either 20 mg/ml of soluble anti-CD28 or with immobilized antibodies as indicated. IFN-g production in supernatants collected after overnight incubation were determined by ELISA. Triplicate determinations of the same samples did not vary by more than 20%.

against FcgRIII which by itself was a poor activator of murine NK cells (Table 3). The stimulation was also not attributable to more e€ective crosslinking of FcgRIII by the immobilization of the antibody because plate bound anti-TCR antibodies of the same isotype yielded only background stimulation (data not shown). Furthermore, Table 3 shows that even plate-bound 2.4G2 was not as e€ective as anti-CD28 in stimulation of IFN-g production. Furthermore, crosslinking of the cell-bound antibodies by a secondary antibody also did not increase the IFN-g secretion level. Despite the use of anity puri®ed antibodies, the extent of stimulation varied considerably between experiments. This is probably attributable to the activation status of the IL-2 propagated cells in each preparation. Because of this variability, conclusions regarding each stimulation condition were made only within the same experiment.

lated when CD28 is triggered we found that stimulation of IL-2 propagated NK cells with soluble antiCD28 resulted in only minimal increases in IFN-g production. On the other hand, when the antibody was immobilized in the culture wells, the extent of the response was in general greatly increased. Table 1 shows three representative experiments including one in which the response to immobilized antibody was lower than usual but nevertheless still exceeded that of stimulation by soluble anti-CD28. The greater ability of immobilized antibodies to stimulate was not attributable to the manner of presentation of the ligand because no response was detected when the F(ab)'2 fragments of anti-CD28 was immobilized in the same manner (Table 2, Experiment 1). Nonetheless, induction by plate-bound anti-CD28 does require interaction with the speci®c ligand because the stimulation was blocked by the inclusion of F(ab)2' fragments of anti-CD28 (Table 2, Experiments 1 and 2). Thus, it appears that the activity of immobilized antibody is mediated by the simultaneous ligation of two di€erent determinants on NK cells. The other ligand is likely to be the Fc portion of anti-CD28 which can interact with the FcgRIII expressed on NK cells. Table 2 (Experiments 3±5) shows that the stimulation was indeed inhibited extensively by the inclusion of 2.4G2, an antibody

3.2. Changes in transcription of the IFN-g g gene upon induction with various stimuli In order to understand the mechanism by which immobilized anti-CD28 induces IFN-g secretion, we performed run-on analysis of IL-2 propagated NK cells stimulated with anti-CD28 to determine if increased transcription can be detected. Four hours after stimu-

Table 2 Stimulation by immobilized anti-CD28 can be blocked by either F(ab)'2 fragments or anti-FcRIII IFN-g produced (unit/ml)a

Stimulus Soluble Ab None None Anti-CD28 F(ab)'2 None Anti-CD28 F(ab)'2 Anti-FcgR Anti-FcgR Control IgG2ab

Immobilized Ab None Anti-CD28 F(ab)'2 None Anti-CD28 Anti-CD28 Anti-CD28 None Anti-CD28

Experiment 1 0.96 0.79 9.0 0.69

Experiment 2 0.04

Experiment 3 0.26

Experiment 4 0.39

Experiment 5

0.05 7.8 0.93

5.9

15.8

11.6

2.8 0.02

3.8 1.4

1.0 0.2 8.6

a IL-2 propagated NK cells were cultured with 20 mg/ml of anti-FcgR or 100 mg/ml of anti-CD28 F(ab)'2 alone or together with immobilized antibodies. IFN-g production in supernatants collected after overnight incubation was determined by ELISA, Triplicate determinations of the same samples did not vary by more than 20%. (Blank spaces denote conditions not included for the particular experiment listed ). b Control IgG2 is a rat monoclonal antibody with no reactivity for NK cells.

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

365

Table 3 Anti-FcRIII is a poor stimulator of murine NK cells IFN-g produced (unit/ml)a Experiment No.

Media

Soluble anti-FcRIII

Soluble anti-FcRIII+MAR

Immobilized anti-FcRIII

Immobilized anti-CD28

1 2 3 1 2

0.08 0.18 0.12 0.08 0.2

0.185 0.16 0.13

0.53 0.34 1.1

3.3 9.5

36.8 22.9

a 1±2  106 IL-2 propagated NK cells were incubated with each of reagents as indicated. Soluble 2.4G2 was added at either 15 mg/ml (Experiments 1 and 2) or 25 mg/ml (Experiment 3). Mouse anti-rat antibodies (MAR) at 10 mg/ml was added after preincubation of the primary antibody at 48C. For immobilized antibodies plates were precoated with antibodies at 20 mg/ml. IFN-g production in supernatants collected after overnight incubation were determined by ELISA. Triplicate determinations of the same samples did not vary by more than 20%.

lation, nascent RNA was labeled by isolating the nuclei and incubating them with a pulse of [32P]aUTP. The relative level of transcription of a number of genes was compared by hybridization of the labeled RNA to a panel of probes. Fig. 1A shows that the extent of hybridization of nascent labeled RNA extracted from anti-CD28 stimulated cells to the plasmid containing IFN-g cDNA was very low, barely above the background level registered from hybridization to the vector (KS) control. Because of the low level, comparison with nascent RNA from non-stimu-

Fig. 1. E€ect of stimulation of NK cells on transcription of the IFNg gene. Run-on analysis was performed as described in Experimental procedures using nuclei (1  107 per sample) obtained from equal numbers of IL-2 propagated NK cells stimulated for 4 h either with immobilized anti-CD28 alone (A, open bars), IL-12 alone (A and B, lightly shaded bars), or IL-12 and immobilized anti-CD28 (A and B, solid bars). Extent of hybridization to each indicated probe was normalized to the extent of hybridization to the GAPDH probe for each sample.

lated control cells did not reveal signi®cant di€erences in a number of experiments (data not shown). We also could not detect signi®cant di€erences when the stimulation time was reduced to 2 h. As a positive control, we performed run-on analysis using IL-12 induced NK cells since the cytokine has been shown to induce transcription upregulation in human NK cells (Chan et al., 1992). Fig. 1A shows that in comparison to cells stimulated with anti-CD28, IL-12 stimulation induced a detectable increase in hybridization to the IFN-g probe. In addition, after stimulation by both antiCD28 and IL-12 the transcription level was further enhanced. Even though the extent of increase was low; however, in another independent experiment (Fig. 1B) the same enhancing e€ect by CD28 was found. In the accompanying bar graphs the levels of hybridization to

Fig. 2. Induction of IFN-g mRNA accumulation and protein secretion from NK cells by anti-CD28 and IL-12. IL-2 propagated NK cells were stimulated with immobilized anti-CD28 either alone or together with IL-12 for 4 h. Subsequently, total RNA was extracted and subjected to RT-PCR analysis as indicated in Experimental procedures. The extent of ampli®cation from IFN-g cDNA (lightly shaded bars) was normalized to the extent of ampli®cation from GAPDH cDNA for each condition. Supernatants removed at the same time were assayed for IFN-g production by ELISA analysis (solid bars).

366

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

Fig. 3. Rate of degradation of IFN-g mRNA in NK cells stimulated under various conditions. (A) Representative S-1 gel analysis of mRNA from IL-2 propagated SCID NK cells stimulated with immobilized anti-CD28 for 16 h, at which time DRB was added to all of the samples and subsequently harvested at the indicated times. M=size markers. P=Probe only. The intensity of species protected by IFN-g mRNA (140 nt) was quanti®ed and normalized to that protected by b2 M (98 nt) mRNA for each sample. This value was then plotted (C) as percent of the 0 time point value (lane 2). Lanes 2±6 represent T=0, 0.5, 1, 2, and 5 h, respectively. (B) Representative semi-quantitative RT-PCR analysis of NK cells treated with DRB after stimulation with immobilized 2.5G2. Lanes 2±6 represent T=0, 0.5, 1, 2, and 5 h, respectively. Lane 7 depicts RTPCR ampli®cation of RNA from cells stimulated with immobilized anti-CD28 as a comparison. The labeled products were quanti®ed by PhosphorImager analysis. The intensity of product from IFN-g mRNA (203 bp) was normalized to the intensity of the product from GAPDH mRNA (350 bp) and plotted (H); g indicates position of genomic DNA ampli®ed by the same primers. (D) The decay rate of mRNA was established for cells treated as in (A) except incubation with DRB was continued for a more extended time period. (E) Cells were incubated with antiCD28 for only 2 h before addition of DRB to inhibit RNA synthesis. Extracted RNA was quanti®ed as in (A). (F) NK cells were stimulated for 16 h with immobilized anti-CD28 and IL-12 before addition of DRB. IFN-g mRNA was determined by semi-quantitative RT-PCR analysis as in (B). (G) The decay curve in (F) was replotted (open circles) together with the decay curve obtained from cells stimulated with only anti-CD28 and analysed in the same manner. Squares indicate level of IFN-g produced in the absence of stimulation (i.e. lane 1 in (A) and (B)). Arrows on X-axis denote estimated half-life of the more labile fraction of RNA. (H) NK cells were stimulated for 16 h with immobilized 2.4G2 before addition of DRB. Samples were processed as in (B).

each probe was normalized to that of the housekeeping gene, GAPDH. This allows the comparison of the extent of increase in IFN-g transcription with possible overall increases in gene transcription that may have been induced by stimulation of the cells. Clearly, a speci®c enhancement in transcription of the IFN-g gene was induced upon stimulation with IL-12 together with anti-CD28. A similar relative increase in IFN-g transcription can be seen if normalization was performed based on hybridization to another non-NK speci®c gene, b2M. The signi®cance of the changes in

2B4 transcription is not clear since the di€erences obtained were not consistent; however, the transcription of another gene, IL-2b receptor, also expressed in NK cells, was not upregulated. Despite the absence of detectable increase in IFN-g gene transcription induced by stimulation with antiCD28 alone, quanti®cation of the accumulated RNA (Fig. 2) shows that anti-CD28 induced a much higher level of mRNA accumulation over a 4-h period than stimulation by IL-12 alone. Furthermore, the two reagents together induced a synergistic increase. The

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

increased mRNA is re¯ected in the amount of protein secreted over the same time period. 3.3. Decay kinetics of IFN-g g mRNA di€ers depending on the stimulus The di€erence in mRNA accumulation in anti-CD28 vs IL-12 stimulated cells suggests that the half-life of mRNA induced under these conditions may di€er. Therefore stability measurements were made. IL-2 propagated NK cells were stimulated with anti-CD28 overnight. Subsequently the stimulated cells were incubated with an inhibitor of RNA polymerase II catalyzed chain elongation, 5,6 dichloro-1b Dribofuranosylbenzimidazole (DRB) for various time periods. At various time points the mRNA was extracted and subjected to S1 analysis. The species representing protection by IFN-g mRNA in the S1 gel shown in Fig. 3A was quanti®ed and plotted after normalization to the b2M mRNA protected species, as a function of time (Fig. 3C). For each independent experiment, the value obtained at 0 time, before addition of DRB but after stimulation was set at 100%. The rate of decrease in IFN-g mRNA levels found for antiCD28 stimulated cells appears to be complex, but it can best be extrapolated to two di€erent slopes representing two populations of RNA, one with a half-life below 2 h and the other with a half-life longer than 6 h. Up to 24 h after inhibition of RNA synthesis, as much as 30% of the mRNA present initially can still be detected (Fig. 3D). By extrapolating the shallower decay slope to the y-axis the fraction of RNA that exhibits a longer half-life can be estimated. In a total of ®ve independent experiments (data not shown), using both S1 analysis and semi-quantitative RT-PCR analysis, we found that this fraction varied between 30±65%, and conversely between 35±70% exhibited a rapid degradation rate. A possible explanation for the presence of two populations of IFN-g mRNA is that at early times after induction, anti-CD28 only activates mRNA transcription and it is only later that mechanisms that activate message stabilization are induced. To test this possibility we tried to determine if the composition of the two populations can be altered by a shorter period of stimulation. NK cells were induced for only 2 h with anti-CD28 before inhibition of RNA synthesis. Fig. 3E indicates that the biphasic decay kinetics was maintained except that the fraction of the mRNA with a longer half-life was signi®cantly decreased suggesting that indeed at early times after induction only the more labile mRNA was present. An alternative explanation that cannot be completely discounted is the possibility that the RNA with the rapid decay slope is derived from uninduced cells. Such RNA has been previously shown to have a similarly short half-life

367

(Wilder and Yuan, 1995). However, it should be noted that the level of IFN-g mRNA present in uninduced cells assayed at the same time was generally less than 5% of the induced level, an amount that is below the 35±70% mRNA estimated to have a rapid degradation rate. If RNA induced initially after transcriptional activation of the IFN-g gene is labile, one would predict that a larger proportion of the RNA induced by the combination of anti-CD28 and IL-12 would have a short half-life since we showed that the combination of these two reagents signi®cantly enhanced transcription. Fig. 3F shows that the decay rate in this case also displayed biphasic properties except that here the proportion of mRNA that exhibited a long half-life represented a much lower fraction of the total RNA (10%). However, because of the increased initial level of mRNA, the amount remaining 5 h after inhibition of RNA synthesis is similar to that remaining after treatment of cells stimulated with anti-CD28 alone. To illustrate this point, in Fig. 3G, the decay curve in Fig. 3F is superimposed on that obtained from cells treated with only anti-CD28 in the same experiment using the 0 time point as the relative amounts present in the two samples at the initiation of DRB addition. Thus it is clear that although only 10% of the mRNA induced in the presence of anti-CD28 and IL-12 has a long half-life but because of the increase in transcriptional level induced by the latter the absolute amount of RNA that persists is much higher, resulting in a synergistic accumulation of total mRNA. Finally, to investigate whether the sequelae of the activation signal induced by anti-CD28 stimulation differs from that by anti-FcgRIII itself, the half life of IFN-g mRNA induced in cells stimulated by immobilized 2.4G2 was also determined in the same manner. As shown in Fig. 3B, lanes 1 and 2, the extent of message induction after overnight incubation is much lower than that of cells stimulated with immobilized antiCD28 (lane 7). Furthermore, the decay kinetics of the induced mRNA in these cells (Fig. 3H) does not di€er from uninduced control cells (data not shown) indicating that ligation by 2.4G2 does not result in message stabilization. In addition, as expected, stimulation of NK cell-IFN-g production by IL-12 was not enhanced by the addition of plate bound anti-FcgRIII (data not shown). Thus the combined engagement of CD28 together with the FcgRIII appears to result in activation of additional signaling pathways that are not triggered by stimulation of the latter alone. 3.4. Responsiveness of non-cultured NK cells to antiCD28 There is some disagreement in the literature as to whether CD28 is expressed on non-cultured NK cells

368

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

Fig. 4. Depletion of the anti-CD28 response in non-cultured SCID splenocytes corresponds to the depletion of CD69 expressing NK cells. IL-2 propagated NK cells (A), freshly isolated SCID splenocytes (B), or nylon wool non-adherent SCID splenocytes (C) were stained with PE-conjugated anti-CD69 and FITC-conjugated anti-NK1.1. PLots indicate staining pro®les of the lymphocyte-sized population. (D) Equal numbers IL-2 propagated NK cells (open bars), fresh SCID splenocytes (lightly shaded bars), or nylon-wool non-adherent SCID splenocytes (solid bars) were cultured in medium alone, with immobilized antibodies, or with equal numbers of BCL1-C11 cells. Supernatants collected after overnight incubation were assayed for levels of IFN-g by ELISA analysis.

(Nandi et al., 1994; Hunter et al., 1997). Part of the di€erence may be attributed to the fact that the fraction of endogenously activated NK cells may di€er in SCID mice kept in di€erent colonies. To address this question from a functional aspect, splenocytes from

SCID mice were passed over nylon wool columns to deplete in vivo activated cells. As shown by a representative experiment (Fig. 4), nylon wool signi®cantly depleted cells that express the CD69 activation marker (Fig. 4A) present on virtually all IL-2 propagated cells

Fig. 5. Mechanism of induction of IFN-g mRNA by BCL1-C11 stimulation. Run-on analysis (A) was performed as described in Experimental procedures using nuclei (2±3  107 per sample) obtained from equal numbers of IL-2 propagated NK cells cultured for 4 h either with (NK+C11 and solid bars in (B)) or without BCL1-C11 cells (NK+(C11) and open bars in (B)). Before lysis, an equivalent number of stimulator cells was added to the non-stimulated NK cells. Extent of hybridization to each indicated probe was normalized to the extent of hybridization to the GAPDH probe for each sample (B). (C) NK cells stimulated overnight with BCL1-C11 cells were treated with DRB and the present residual RNA remaining at the times indicated was determined as in Fig. 3.

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

(Karlhofer and Yokohama, 1991) but enriched for the total fraction of NK cells in the preparation (Fig. 4C). Both non-fractionated and fractionated splenocytes were tested for their ability to produce IFN-g in response to immobilized anti-CD28. Fig. 4D shows that whereas anti-CD28 could e€ectively stimulate freshly isolated NK cells to produce IFN-g, removal of the majority of activated NK cells virtually eliminated all of the response to anti-CD28. The lack of response con®rmed that functional CD28 is not expressed on NK cells until after activation. The same populations were examined for the response to anti-NK1.1 which is expressed on resting as well as activated NK cells. The increase in response to this antibody found for nylon wool-passaged cells re¯ects the slight enrichment of NK1.1 expressing cells in the population. We have previously shown that BCL1-C11 cells can induce IFN-g production by NK cells in vivo (Michael et al., 1989). Attempts to block this interaction with CTLA4-Ig (Linsley et al., 1991) resulted in highly variable results in that e€ective suppression was observed in only some experiments (data not shown). Since resting NK cells do not express functional CD28, we tested the same population for reactivity to BCL-C11 cells. Fig. 4D shows that this population can indeed respond to these tumors albeit at lower levels therefore at least part of the response maybe CD28 independent. 3.5. Mechanism of induction of IFN-g g mRNA by BCL1C11 tumor cells To examine if stimulation of NK cells by BCL1-C11 tumor cells occurs via the same mechanism as that by anti-CD28 or by IL-12, we ®rst tested the e€ect of transcription of the IFN-g after induction of IL-2 propagated NK cells with BCL1-C11 cells. Fig. 5A and B shows that after a 4-h induction period, at which time signi®cant levels of IFN-g mRNA and protein secretion can be detected, transcription of the gene is not signi®cantly increased. On the other hand, the decay kinetics of IFN-g mRNA determined after overnight stimulation (Fig. 5C) shows that none of the newly induced IFN-g mRNA has a short half-life. Therefore the major mechanism of induction by the tumor cells is via mRNA stabilization. 4. Discussion The major conclusion from the experiments presented herein is that, depending on the nature of the stimulus, NK cells utilize two alternative mechanisms as well as combinations thereof to increase accumulation of IFN-g mRNA. Because B7-1 and B7-2 expressed on tumor cells (Djeu et al., 1980; Koh and Yuan, 1997), infected antigen presenting cells (Hunter

369

et al., 1997) and activated B cells (Michael et al., 1989) can possibly activate NK cells via their interaction with CD28 we initiated these experiments to explore the mechanism by which ligation of this counterreceptor initiates the cytokine response. Anti-CD28 was used to analyze this response independent of other factors expressed on stimulatory cell types. However, we ®nd that ligation of CD28 alone results in minimal responses from NK cells. Thus neither soluble antiCD28 nor immobilized F(ab) '2 fragment of the antibody can increase signi®cantly the production of IFNg. On the other hand, CD28 is a very ecient ampli®er of NK responses to ligation via FcgRIII or stimulation by IL-12. Murine NK cells respond poorly to ligation of CD16 because, unlike human NK cells (Anegon et al., 1988), the half-life of IFN-g mRNA induced by this route is not increased and remains as short (less than 1 h) as that in non-induced cells. The reason for the di€erence in response to CD16 ligation is not clear but there are reported di€erences in signaling pathways utilized by this receptor on human vs murine NK cells. For example, in human cells CD16 can be functionally associated with the z chain (Salcedo et al., 1993), whereas FcgRIII signals only through the g chain in murine NK cells (Takai et al., 1994; Arase et al., 1997). Furthermore, CD16 ligation in human cells results in activation of p56lck (Salcedo et al., 1993; Bottino et al., 1994) whereas knock-out of this gene has been shown to not compromise ADCC activity in murine NK cells (Wen et al., 1995). Our experiments show that upon crosslinking of CD28 with FcgRIII much higher levels of IFN-g is produced. Clearly engagement of both receptors are needed because the stimulation can be blocked by either F(ab) '2 fragments of anti-CD28 or by 2.4G2. It is interesting that despite the fact that the Fc portion of soluble anti-CD28 antibodies can interact with the FcgRIII and that the antigen combining site of the same antibodies can interact with CD28 on NK cells, the signal is much less e€ective than when the antibodies are ®xed on a solid surface. Thus appropriate crosslinking of CD28 with CD16 dictated by immobilization may activate additional pathways which cannot be provided by engagement of either receptor alone. We have not ruled out the possibility that other molecules expressed on NK cells may serve a similar role as CD28. A preliminary screen of the ability of immobilized antibodies to a€ect various Ly49 family members and MHC Class I determinants reveal a wide range of responses. Therefore, further experiments are in progress to dissect these e€ects. However, it is clear that immobilized anti-MHC Class I antibodies that are not expressed on NK cells do not activate IFN-g secretion (Cheung, unpublished observations). We have shown that, as in T lymphocytes (Lindsten et al., 1989) the major mechanism by which anti-CD28

370

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

induces NK cell production of IFN-g is via speci®c mRNA stabilization. However, although increased synthesis of the message was not detectable by run-on analysis, anti-CD28 must induce some transcriptional activation because 90% of the newly induced IFN-g mRNA appearing at 2 h after activation has a rapid turn-over rate. Surprisingly, this fraction, while diminished, persists when cells are activated for extended periods and the relative amount is higher than background levels present in cells cultured in only IL-2. Therefore, the basal level of IFN-g transcription must be elevated in stimulated cells. Interestingly, whereas T cell stimulation of lymphokine production by antiCD28 has been attributed to mRNA stabilization, there is also evidence that anti-CD28 can induce transcriptional up-regulation of T cell genes such as IL-2 (Fraser et al., 1991) via a NF-kappa B binding site. Thus it is possible that the IFN-g promoter may also have a binding site for factors that can respond to CD28 ligation. It is interesting that, in contrast, to the biphasic decay slopes found for anti-CD28 stimulation, when NK cells are stimulated with PMA, B cells (Wilder and Yuan, 1995), or as shown herein, BCL1-C11 tumor cells, the biphasic kinetics is not apparent. Therefore, if CD28 plays a role in the activation via these routes, the mechanism of this co-stimulation must be di€erent. It should be noted that this experiment also serves to show that the biphasic decay slopes observed is not likely to be an artifact of the experimental system utilized. Since run-on analysis is a relatively insensitive method of measuring small di€erences in levels of transcription after stimulation of cells that are already partially activated, the measurement of stability of newly induced mRNA may yield more reliable information. In contrast, the extent of elevation of transcription induced by IL-12 is detectable by run-on analysis and treatment with anti-CD28 augments the level although the extent of increase is not easy to quantify. Further analysis is necessary before the mechanistic basis of this enhancement can be fully understood. A testable hypothesis is that anti-CD28 stimulation augments IL12 receptor expression on NK cells. It should be noted, however, that the increased transcription induced by anti-CD28 is apparently sucient to signi®cantly enhance accumulation of total IFN-g mRNA. Thus in spite of the relative short half life of a large fraction of the transcripts synthesized after stimulation by IL-12 and anti-CD28 (Fig. 3G), nevertheless, due to the transcriptional up-regulation that results in greater input into the IFN-g mRNA pool, a net increase in speci®c mRNA is achieved. It is interesting that not all of the RNA induced is stabilized. Whereas it has been shown that stabilization of the IFN-g mRNA is most likely mediated by the cis regu-

latory sequence contained within the AUUUA response element (Shaw and Kamen, 1986), the nature and mode of action of transregulatory factors remain unclear (Sachs, 1993; Myer et al., 1997). The decay pro®le of cells stimulated with IL-12 and CD28 suggests that these putative regulatory proteins are limiting when cells are stimulated in this manner. Thus a substantial fraction of RNA remains relatively unstable. Alternatively, it is possible that the labile fraction is not translated eciently and therefore rapidly degraded (Jack et al., 1989). In spite of the fact that cell surface expressed B7-1 or B7-2 has previously been shown to augment the IL-12 response (Hunter et al., 1997; Kubin et al., 1994; Murphy et al., 1994), whether the mechanism involves transcriptional enhancement has not been investigated. However, the mechanism documented herein for the synergistic induction of IFN-g by anti-CD28 and IL-12 is analogous to the synergistic activity of IL-2 and IL-12 on human NK cells, in that IL-12 increases transcriptional activity of the IFN-g gene, but the addition of a stimulus from IL-2 greatly increases mRNA stability (Chan et al., 1992). Finally, we have found that if the activated fraction of NK cells is removed from freshly isolated splenocytes, the cells are no longer responsive to immobilized anti-CD28 and therefore cannot function as a costimulator for the FcRg-mediated signal. Therefore, regardless of whether CD28 is expressed on resting NK cells, these results show that they are functionally inactive. Thus, unlike the B7-2-CD28 interaction occurring between APCs and resting T lymphocytes (Linsley, 1995; Bluestone, 1995), B7-2, if active for non-stimulated NK cells in vivo, must trigger via alternative counter-receptors. The latter may be related to those that trigger NK mediated cytotoxicity of bone marrow-derived macrophages or dendritic cells. In this case, despite the presence of CD86 on the target cells, the NK response cannot be inhibited by either antiCD28 or anti-CTLA-4 antibodies (Chambers et al., 1996). Thus, the costimulatory activity of CD28 is only operative after NK cells have been activated by other means. We have shown that BCL1-C11 cells also induce IFN-g production from resting NK cells. In vivo, however, despite the ability of these tumor cells to directly induce the cytokine, IL-12 is also necessary for BCL1-C11 to exert detectable functional activity (Koh and Yuan, 1997). These studies provide a mechanistic explanation for this observation. Thus, because stimulation via cellular interactions induces message stabilization, signi®cant ampli®cation of this e€ect can be achieved by transcriptional activation via IL-12. On the other hand, the rapid turnover of mRNA induced by IL-12 alone insures that e€ects of this powerful cytokine are kept in check, so that it can

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

operate e€ectively only in conjunction with other factors that depend on direct cell contact. Acknowledgements We thank Dr Richard O'Hara, Genetics Institute, Cambridge, MA, for providing recombinant IL-12, Dr Peter S. Linsley, Bristol-Myers Squibb Pharmaceutical Research Institute, Seattle, WA, for providing CTLAIg, and Dr Ariizumi Kioshe, Department of Internal Medicine, UT Southwestern, for providing the IL-2Rb plasmid. We are grateful for the expert technical assistance of Denise Broyles, Tam Dang and De Ming Zhou. This work was supported by National Institutes of Health Grant AI38938 and Cancer Immunology Training Grant CA09082. References Anegon, I., Cuturi, M.C., Trinchieri, G., Perussia, B., 1988. Interaction of Fc receptor (CD16) ligands induces transcription of interleukin 2 receptor (CD25) and lymphokine genes and expression of their products in human natural killer cells. J. Exp. Med. 167, 452±472. Arase, N., Arase, H., Park, S.Y., Ohno, H., Ra, C., Saito, T., 1997. Association with FcRg is essential for activation signal through NKR-P1 (CD161) in natural killer (NK) cells and NK1.1+ T cells. J. Exp. Med. 186, 1957±1963. Bluestone, J.A., 1995. New perspectives of CD28:B7-mediated T cell costimulation. Immunity 2, 555±559. Bottino, C., Vitale, M., Olcese, L., Sivori, S., Morelli, L., Augugliaro, R., Ciconne, E., Moretta, L., Moretta, A., 1994. The human natural killer cell receptor for major histocompatibility complex class I molecules. Surface modulation of p58 moles and their linkage to CD3 zeta chain, Fc epsilon RI gamma chain and the p56(lck ) kinase. Eur. J. Immunol. 24, 2527±2534. Brooks, K., Yuan, D., Uhr, J.W., Krammer, P.H., Vitetta, E.S., 1983. Lymphokine-induced IgM secretion by clones of neoplastic B cells. Nature 302, 825±826. Chambers, B.J., Salcedo, M., Ljunggren, H.G., 1996. Triggering of natural killer cells by the costimulatory molecule CD80 (B7-1). Immunity 5, 311±317. Chan, S.H., Kobayashi, M., Santoli, D., Perussia, B., Trinchieri, G., 1992. Mechanisms of IFN-g induction by natural killer cell stimulatory factor (NKSF/IL-12). J. Immunol. 148, 92±98. Djeu, J.Y., Huang, K.Y., Herberman, R.B., 1980. Augmentation of mouse natural killer activity and induction of interferon by tumor cells in vivo. J. Exp. Med. 151, 781±789. Feramisco, J.R., Helfman, D.M., Smart, J.E., Burridge, K., Thomas, G.P., 1982. Coexistence of vinculin and a vinculin-like protein of higher molecular weight in smooth muscle. J. Biol. Chem. 257, 11024±11031. Fraser, J.D., Irving, B.A., Crabtree, G.R., Weiss, A., 1991. Regulation of interleukin-2 gene enhancer activity by the T cell accessory molecule CD28. Science 251, 313±316. Hunter, C.A., Ellis-Neyer, L., Gabriel, K.E., Kennedy, M.K., Grabstein, K.H., Linsley, P.S., Remington, J.S., 1997. The role of the CD28/B7 interaction in the regulation of NK cell responses during infection with toxoplasma gondii. J. Immunol. 158, 2285± 2293. Jack, H.M., Berg, J., Wabl, M., 1989. Translation a€ects immuno-

371

globulin mRNA stability. European Journal of Immunology 19, 843±847. Karlhofer, F.M., Yokohama, W.M., 1991. Stimulation of murine natural killer (NK) cells by a monoclonal antibody speci®c for the NK1.1 antigen. IL-2 activated NK cells possess additional speci®c stimulation pathways. J. Immunol. 146, 3662±3673. Koh, C.Y., Yuan, D., 1997. The e€ect of NK cell activation by tumor cells on antigen-speci®c antibody responses. J. Immunol. 159, 4745±4752. Koo, G.C., Peppard, J.R., 1984. Establishment of monoclonal antiNK-1.1 antibody. Hybridoma 3, 301±303. Kubin, M., Kamoun, M., Trinchieri, G., 1994. Interleukin 12 synergizes with B7/CD28 interaction in inducing ecient proliferation and cytokine production of human T cells. J. Exp. Med. 180, 211±222. Kubo, R.T., Born, W., Kappler, J.W., Marrack, P., Pigeon, M., 1989. Characterization of a monoclonal antibody which detects all murine ab T cell receptors. J. Immunol. 142, 2736±2742. Lindsten, T., June, C.H., Ledbetter, J.A., Stella, G., Thompson, C.B., 1989. Regulation of lymphokine messenger RNA stability by a surface-mediated T cell activation pathway. Science 244, 339±343. Linsley, P.S., Brady, W., Urnes, M., Grosmaire, L.S., Damle, N.K., Ledbetter, J.A., 1991. CTLA-4 is a second receptor for the B cell activation antigen B7. J. Exp. Med. 174, 561±569. Linsley, P.S., 1995. Distinct roles for CD28 and cytotoxic T lymphocyte-associated molecule-4 receptors during T cell activation. J. Exp. Med. 182, 289±292. Michael, A., Hackett, J., Bennett, M., Kumar, V., Yuan, D., 1989. Regulation of B lymphocytes by natural killer cells: Role of interferon-gamma. J. Immunol. 142, 1095±1101. Murphy, E.E., Terres, G., Macatonia, S.E., Hsieh, C.S., Mattson, J., Lanier, L., Wysocka, M., Trinchieri, G., Murphy, K., O'Garra, A., 1994. B7 and interleukin 12 cooperate for proliferation and interferon gamma production by mouse T helper clones that are unresponsive to B7 costimulation. J. Exp. Med. 180, 223±231. Myer, V.E., Fan, X.C., Steitz, J.A., 1997. Identi®cation of HuR as a protein implicated in AUUUA-mediated mRNA decay. EMBO 16, 2130±2134. Nandi, D., Gross, J.A., Allison, J.P. 1994 CD28-mediated costimulation is necessary for optimal proliferation of murine natural killer cells. 1997 152: 3361±3369. Salcedo, T.W., Kurosaki, T., Kanakaraj, P., Ravetch, J.V., Perussia, B., 1993. Physical and functional association of p56lck with FcgRIIIA (CD16) in natural killer cells. J. Exp. Med. 177, 1475± 1480. Sachs, A.B., 1993. Messenger RNA degradation in eukaryotes. Cell 74, 413±421. Shaw, G., Kamen, R., 1986. A conserved AU sequence from the 3' untranslated region of GM-CSF mRNA mediates selective mRNA degradation. Cell 46, 659±667. Spitalny, G., Havell, E., 1984. Monoclonal antibody to murine gamma interferon inhibits lymphokine-induced antiviral and macrophage tumoricidal activities. J. Exp. Med. 159, 1560±1565. Takai, T., Li, M., Sylvestre, D., Clynes, R., Ravetch, J.V., 1994. FcRg chain deletion results in pleiotrophic e€ector cell defects. Cell 76, 531±542. Tutt, M.M., Kuziel, W.A., Hackett Jr, J., Bennett, M., Tucker, P.W., Kumar, V., 1986. Murine natural killer cells do not express functional transcripts of the a-, b-, or d-chain genes of the T cell receptor. J. Immunol. 137, 2998±3001. Unkeless, J.C., 1979. Characterization of a monoclonal antibody directed against mouse macrophage and lymphocyte Fc receptors. J. Exp. Med. 150, 580±596. Wen, T., Zhang, L., Kung, S.K.P., Molina, T.J., Miller, R.G., Mak, T.W., 1995. All-skin graft rejection, tumor rejection and natural

372

J.C. Cheung et al. / Molecular Immunology 36 (1999) 361±372

killer activity in mice lacking p58lck. Eur. J. Immunol. 25 (11), 3155±3159. Wilder, J.A., Yuan, D., 1995. Regulation of interferon-g mRNA production in murine natural killer cells. Int. Immunol. 7, 575±582. Yu, Y.Y.L., George, T., Dorfman, J.R., Roland, J., Kumar, V., Bennett, M., 1996. The role of Ly49A and 5E6 (Ly49C) molecules in hybrid resistance mediated by murine natural killer cells against normal T cell blasts. Immunity 4, 67±76. Yuan, D., Dang, T., 1997. E€ect of the presence of transgenic H and

L chain genes on B cell development and allelic exclusion. Int. Immunol. 9, 1651±1661. Yuan, D., Dang, T., Hawley, J., Jenuwein, T., Grosschedl, R., 1995. Role of the OCTA site in regulation of Ig heavy chain gene transcription during B cell activation. Inter. Immunol. 7, 1162± 1171. Yuan, D., Tucker, P.W., 1984. Transcriptional regulation of the m-d gene in normal murine B lymphocytes. J. Exp. Med. 160, 564± 574.