In vivo regulation of murine granzyme B gene transcription in activated primary T cells.

A murine granzyme B promoter fragment that extends 243 base pairs upstream of the transcription start site confers high levels of luciferase reporter gene activity in transient transfection assays into T cells and mouse L cell fibroblasts. This promoter fragment contains canonical binding sites for the transcription factors AP-1, core binding factor (CBF), Ikaros, and the cyclic AMP responsive element binding protein (CREB). Oligonucleotides containing the granzyme B AP-1 or CBF elements form specific complexes with proteins present in nuclear extracts from activated CD8+ splenocytes, MTL cells, EL4 T cells, and L cells. A strong DNase1 hypersensitive site that coincides with the closely associated AP-1, CBF, Ikaros, and CRE elements is present in activated CD8+ T cells but not in resting T cells or L cells. Both in vitro and in vivo footprints are observed at these sequence elements in activated cytotoxic T cells (CTL) but not in resting T cells. The endogenous granzyme B gene is CTL-specific as no mRNA is detectable in EL4 or L cells. We propose that a condensed chromatin structure at the granzyme B promoter is responsible for transcription factor inaccessibility and repression of transcription in non-T cells.

A murine granzyme B promoter fragment that extends 243 base pairs upstream of the transcription start site confers high levels of luciferase reporter gene activity in transient transfection assays into T cells and mouse L cell fibroblasts. This promoter fragment contains canonical binding sites for the transcription factors AP-1, core binding factor (CBF), Ikaros, and the cyclic AMP responsive element binding protein (CREB). Oligonucleotides containing the granzyme B AP-1 or CBF elements form specific complexes with proteins present in nuclear extracts from activated CD8 ؉ splenocytes, MTL cells, EL4 T cells, and L cells. A strong DNase1 hypersensitive site that coincides with the closely associated AP-1, CBF,

Ikaros, and CRE elements is present in activated CD8 ؉ T cells but not in resting T cells or L cells. Both in vitro and in vivo footprints are observed at these sequence elements in activated cytotoxic T cells (CTL) but not in resting T cells. The endogenous granzyme B gene is CTLspecific as no mRNA is detectable in EL4 or L cells. We propose that a condensed chromatin structure at the granzyme B promoter is responsible for transcription factor inaccessibility and repression of transcription in non-T cells.
The body's major defense against viral infections is mediated by cytotoxic T lymphocytes. These cells can sometimes mount an attack on healthy tissue and cause autoimmune diseases, and they are responsible for organ and tissue transplant rejection. Whether the response is appropriate or not, resting T lymphocytes are activated upon recognition of antigen, in the context of major histocompatibility complex molecules (Chien and Davis, 1993). One approach to understanding the events that occur is to study the specialized set of genes that are induced during the acquisition of killing potential. These include perforin and cytotoxic serine proteinases (granzymes) that are major components of the killing machinery.
Granzymes are implicated in granule-mediated target cell death because their expression is closely correlated with killing activity (Prendergast et al., 1992) and by their localization in cytolytic granules (Henkart, 1985;Redmond et al., 1987). Granzyme B is a member of the cytotoxic proteinase gene family that is clustered on mouse chromosome 14 (Crosby et al., 1990). The exact function of granzymes in granule-mediated killing is unclear, but CPP32, a protease believed to be important in apoptosis, has recently been identified as a substrate for granzyme B (Darmon et al., 1995).
The granzyme genes provide an excellent system in which to study cell-specific gene induction as their expression is restricted to activated T lymphocytes and NK cells. Studies of the human and murine granzyme B proximal promoters reveal that they share several conserved sequences. These include T cell-specific transcription factor binding sites such as Ikaros and core binding factor (CBF 1 /PEBP2) (Haddad et al., 1993;Kamachi et al., 1990;Wang and Speck, 1992) as well as recognition sequences for the ubiquitous transcription factors AP-1 and the cyclic AMP response element binding factor (CREB). These sequences have been shown to be sufficient to induce reporter gene expression in immortalized T cell lines in which many of these transcription factors are constitutively active (Frégeau and Bleackley, 1991;Hanson et al., 1993).
We are primarily interested in the events that take place at the endogenous granzyme B locus as resting lymphocytes make the transition to activated killers. We have developed a method for the transfection of reporter gene plasmids into primary mouse splenocytes and show that the minimal granzyme B promoter is able to induce significant levels of luciferase activity in activated CD8 ϩ T cells. Electrophoretic mobility shift analysis was used to examine the DNA binding activities of transcription factors before and after CD8 ϩ T cell activation. We have precisely established the sequences involved in transcription factor binding by in vitro footprinting studies using nuclear extracts derived from a cytotoxic T cell clone. DNase1 hypersensitivity analysis identified potentially important regulatory regions in the granzyme B promoter in CD8 ϩ cells. Finally, we were able to observe these protein/DNA interactions in the endogenous promoter by in vivo footprinting analysis in resting and activated CD8 ϩ T cells using the dimethyl sulfate (DMS)/LMPCR genomic footprinting method. Together these results have enabled us to probe the status of the endogenous gene before and after T cell activation and in a physiologically relevant system. and L cells are a mouse fibroblast line. All cells were cultured in RHFM (RPMI supplemented with 20 mM HEPES (pH 7.5), 100 M ␤-mercaptoethanol, and 10% fetal bovine serum (5% for L cells)). Type 1 cells, MTL 2.8.2 cells, and primary splenocytes were cultured in the presence of 60 units/ml human recombinant IL-2. Primary splenocytes were stimulated with 5-10 g/ml concanavalin A (Sigma), 1:300 to 1:500 dilution of hamster anti-mouse ␣CD3 monoclonal antibody supernatant (Leo et al., 1987), and 1:2000 dilution of ␣CD28 (Pharmingen), alone or in combination. EL4 cells were stimulated with 10 ng/ml phorbol 12myristate 13-acetate (PMA) (Sigma) and 4 M ionomycin (Sigma).
Luciferase and ␤-Galactosidase Assays-Three 10 -20-l aliquots of cell lysates were measured for 20 s following the injection of luciferase reagent (Luciferase Assay, Promega) by a LUMAT LB9501 luminometer (Berthold Systems Inc.). ␤-Galactosidase assays were performed as described (Sambrook et al., 1989). Final activities are given as luciferase/␤-galactosidase values. Because the incubation periods for the ␤-galactosidase assay varied between the various cell types, the luciferase/ ␤-galactosidase values are relative only within each transfection experiment but are not relative between the different cell types.
Northern Blot Analysis-RNA was prepared by acid guanidinium phenol extraction (Chomczynski and Sacchi, 1987). Total RNA was separated on denaturing formaldehyde agarose gels and transferred onto Hybond-N nylon membranes (Amersham) by capillary transfer.
Plasmids-The promoterless luciferase reporter gene plasmid p19LUC was obtained from J. R. De Wet (De Wet et al., 1987). Granzyme B promoter fragments (from our C11 gene clone), obtained by restriction enzyme digestion or PCR amplification, were inserted upstream of the luciferase gene, and orientation was confirmed by sequencing. Two ␤-galactosidase control plasmids were used. SV␤-gal has the bacterial ␤-galactosidase gene under the control of the SV2 viral promoter (Promega), and 906 (from A. Puschel) has the bacterial ␤-galactosidase gene under the control of the human actin promoter. Plasmids were grown in DH5␣ Escherichia coli, and high quality supercoiled DNA was purified for transfection by CsCl 2 density gradient centrifugation.
Electrophoretic Mobility Shift Assays-Nuclear extracts were performed as in Schreiber et al. (1989). Mobility shift assays were performed as in Lin et al. (1993) with minor modifications. Briefly, 3-5 g of nuclear extract was incubated with 2 g of poly(dI-dC) and approximately 0.1 to 0.5 ng (10 -20,000 cpm) of 32 P-labeled oligonucleotide in a 15-20-l reaction containing 12 mM HEPES (pH 7.9), 50 mM KCl, 0.5 mM EDTA, 2.5 mM dithiothreitol, and 10% glycerol. Oligonucleotides were end-labeled with T4 kinase and [␥-32 P]ATP or annealed, and the ends were filled in with Klenow polymerase and [␣-32 P]dCTP. Reactions containing antibody or antiserum were preincubated for 30 min at 4°C. The reactions were run on 5% nondenaturing polyacrylamide gels.
Antisera and Purified Antibodies-Antisera to the PEBP2 ␣ and ␤ subunits was kindly provided by Yoshiaki Ito. Anti-␣A1N35 serum reacts with the ␣ subunit, and rabbit anti-␤2 serum reacts with the ␤ subunit of the PEBP2 complex. Purified c-Fos (K-25) and c-Jun/AP-1 (D) antibodies were obtained from Santa Cruz Biotechnology.
Oligonucleotides and Probes-The sequences of each oligonucleotide used in EMSAs are as follows: granzyme B AP-1 agctTCTCTGAGTCA-TTCa, granzyme B CBF agctTCTGCTGTGGTTACTTCATa, granzyme B Ikaros agctTACAACACCCCCACCCCTA, and granzyme B mutant Ikaros GGCTACAACACTCTCACTCCTATGCCCTT (nucleotides not present in granzyme B are shown in lowercase). Northern blots were probed with a murine granzyme B cDNA and a human ␥-actin cDNA [␣-32 P]dCTP labeled by random priming. Southern blots obtained in the DNase1 hypersensitivity experiments were probed with a TaqI/PstI restriction fragment that extends from Ϫ828 to Ϫ546 of the granzyme B promoter. A human c-Fos gene fragment that extends 80 bp upstream of the transcription start site to 360 bp into intron A was used as a positive control. All hybridizations were performed in 50% formamide hybridization solution at 42°C.
Cell Separation-CD8 ϩ primary splenocytes were isolated from whole splenocyte populations by incubation with an ␣CD8 primary antibody (Serotec) followed by immunomagnetic separation with magnetic Dynabeads ® (Dynal ® ) or by passage over Cellect TM immunocolumns (Biotex Laboratories Inc.). The purity of the selected cell populations were routinely confirmed by flow cytometry analysis.
In Vitro DNase1 Footprinting-Nuclear extracts were prepared as in Ohlsson and Edlund (1986). Extracts were routinely made from 1 ϫ 10 9 cells. The footprint reaction consisted of the following components in a 50-l volume: 25 mM HEPES (pH 7.8), 50 mM KCl, 0.05 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 5% glycerol. The reactions contained approximately 1 ng of end-labeled DNA fragment, 1 g of poly(dI-dC), and up to 100 g of extract. The extract was preincubated for 20 -30 min at 4°C, after which the end-labeled fragment was added and incubated for an additional 10 -15 min at room temperature. DNase1 digestion was initiated by the addition of MgCl 2 and CaCl 2 to final concentrations of 5 mM and 1 mM, respectively. The amount of DNase1 added was empirically determined to give an even pattern of cleavage products. Usually, 500 ng was used per reaction (2.5 ng for controls) and allowed to digest for 45 s. Digestion was stopped by the addition of 100 l of stop solution (100 mM Tris-HCl (pH 8.0), 100 mM NaCl, 1% Sarkosyl, 10 mM EDTA, 100 g/ml proteinase K, and 25 g/ml calf thymus DNA). The reactions were phenol-extracted, and the DNA was ethanol-precipitated. The fragments were visualized on a 7% polyacrylamide, 7 M urea sequencing gel.
In Vivo DMS-Piperidine Treatment of DNA-Dimethyl sulfate (DMS)-piperidine treatment of cells was performed as described in Mueller et al. (1992). Cells were pelleted at 300 ϫ g for 5 min. Approximately 1 ml of medium was left behind to resuspend cells. They were then transferred to a microcentrifuge tube and incubated in a 37°C water bath. 10 l of a 10% DMS/ethanol solution was added to the cells and incubated for 1 min. The methylation reaction was stopped by transferring cells to 49 ml of ice-cold PBS followed by centrifugation at 300 ϫ g for 5 min at 4°C. The cell pellet was resuspended in 1-2 ml of cold PBS, 49 ml of ice-cold PBS was added, and cells were pelleted at 300 ϫ g for 5 min at 4°C. The cell pellet was resuspended in 0.3 ml of cold PBS and added to 2.7 ml of lysis buffer (300 mM NaCl, 50 mM Tris⅐HCl (pH 8.0), 25 mM EDTA (pH 8.0), 200 g/ml proteinase K, 0.2% SDS). The DNA was purified, and piperidine cleavage was performed. A naked DNA control was prepared at the same time.
MTL 2.8.2s were grown on 24 ϫ 24-cm plates to a final cell density of 2.5-4 ϫ 10 7 cells/plate. The medium was removed, and 100 ml of prewarmed RHFM containing 0.1% DMS was added and incubated for 2 min. The cells were washed three times with prewarmed PBS. Approximately 8 ml of lysis buffer was layered over the cells, and the plates were swirled gently for 5 min. The lysed cells were scraped off the plates into 50-ml tubes and incubated at 37°C for 5 h. DNA was prepared from the lysates, and a naked DNA control was processed at the same time. In vitro DMS treatment of naked DNA and subsequent piperidine cleavage of both in vitro and in vivo DMS-treated DNA was performed.
Ligation-mediated PCR Genomic Footprinting-LMPCR genomic footprinting was performed as detailed in Mueller et al. (1992). Oligonucleotides used to detect interactions in the noncoding strand of the Ϫ243 C11 promoter were as follows. Primer 1, 5Ј-cctaggtcccagcgtcaagagt-3Ј (T m 61.8°C); primer 2, 5Ј-gagaggaagaaggcagagggggctct-3Ј (T m 66.3°C), primer 3, 5Ј-gaggaagaaggcagagggggctctgtgacc-3Ј (T m 69.6°C). The staggered linker (LMPCR.1) was changed by a single base to raise the T m . The A residue at position 7 was changed to a G residue. The new T m was 66°C. LMPCR hybridization temperatures were 63°C, 68°C, and 71°C (for primers 1, 2, and 3, respectively). For the end-labeling reaction, 3 cycles of PCR were performed, and the final products were precipitated in duplicate. The PCR products were run on 7% polyacrylamide, 7 M urea sequencing gels. Fixed and dried gels were exposed on Kodak XAR-5 film without an intensifying screen.

RESULTS
Granzyme B RNA Is CTL-specific-To determine the expression pattern of the granzyme B RNA in the various cell types utilized in our experiments, we prepared total RNA and probed with a granzyme B cDNA. Fig. 1 depicts the relative message levels that were produced from two murine cytotoxic T cell lines,theIL-2-dependentMTL2.8.2line,theIL-2-andalloantigendependent CTL21.9 (type 1) line, and stimulated CD8 ϩ splenocytes. Also included were EL4, a murine thymoma T cell line, L cell fibroblasts, and a granzyme-independent cytolytic hybridoma PMM-1 (Kaufmann et al., 1981). The granzyme B mRNA was expressed at high levels in CD8 ϩ splenocytes and our cytotoxic T cell clones but was completely absent in EL4, PMM-1, or L cells. We were not able to further induce the mRNA levels in type 1 T cells or induce transcription in EL4 cells by stimulation with PMA/ionomycin, ␣CD3, or concanavalin A (data not shown).
The granzyme B mRNA was inducible, however, in primary splenocytes stimulated in the presence of IL-2, ␣CD3, and concanavalin A. We examined this induction in a population of primary lymphocytes by Northern blot analysis after 1, 2, and 3 days following stimulation. On day 3, lymphocytes were iso-lated by immunomagnetic separation, and total RNA was prepared for analysis. Fig. 2 depicts the induction profile of the granzyme B message in the whole splenocyte population and of the mRNA in the CD8 ϩ -and CD4 ϩ /CD8 ϩ -depleted cells. We believe that the lower band is mature mRNA while the upper corresponds to a processing intermediate. Granzyme B mRNA appeared on day 1, was very high by day 3, and was not observed in the CD4 ϩ /CD8 ϩ -depleted cell fraction. Even though the control actin band is very low at day 1, the intensities of the 18 and 28 S RNAs were very similar in all lanes. Together, these Northern blot data indicate that this gene is efficiently transcribed, or the mRNA is sufficiently stable, only in cytotoxic T cells. A

243-bp Promoter Fragment Confers High Levels of Reporter Gene Activity in T Cells and L Cells in Transient Transfection
Analysis-To define important transcriptional regulatory elements in the granzyme B promoter, a series of deletion fragments was constructed and inserted upstream of a promoterless luciferase gene. These constructs were then transfected into a variety of cell types, and the relative levels of reporter gene expression were examined. Two promoter fragments, one that extends 243 bp and another that extends 828 bp upstream from the transcription start site consistently produced the highest levels of luciferase activity in Type 1 CTL cells (Fig.  3A). The Rous sarcoma viral promoter was typically as active as the 828-bp granzyme B (C11 gene clone) promoter in T cells, and no luciferase expression was ever observed from the parental p19LUC plasmid. Larger 5Ј-flanking sequences were examined (up to 5 kb) and were much less effective in activating luciferase gene expression than the 828-bp or the 243-bp fragments. These larger fragments were active in both T cells FIG. 1. Northern blot analysis of the granzyme B mRNA in various cell types. Total RNA was collected from L cells, EL4, PMM1, CD8 ϩ (stimulated for 3 days with IL-2, ␣CD3, and concanavalin A, and isolated by passage over a CD8 ϩ immunocolumn), Type 1, and MTL 2.8.2 cells, and 5 g of each was separated on a 0.9% denaturing formaldehyde/agarose gel. The resulting blot was probed with a murine granzyme B cDNA and a human ␥-actin cDNA.
FIG. 2. Northern blot analysis of granzyme B mRNA accumulation in stimulated whole and CD8 ؉ (day 3) primary splenocytes. Total cellular RNA was collected from whole splenocytes following 1, 2, or 3 days of stimulation with IL-2, ␣CD3, and concanavalin A and from the day 3 CD8 ϩ -(isolated by positive immunomagnetic separation) and CD4 ϩ /8 ϩ -depleted subpopulations. RNA (5 g) was separated on a 0.9% denaturing formaldehyde/agarose gel, and the resultant Northern blot was probed with a murine granzyme B cDNA and a human ␥-actin cDNA. and non-T cells (data not shown). We decided to focus our studies on the smaller, but highly active, 243-bp fragment.
Fragments extending 108 bp, 169 bp, and 243 bp upstream of the transcription start site were transfected into MTL 2.8.2, EL4, L cells, and whole splenocytes. Splenocytes were activated by stimulation with IL-2, ␣CD3, and concanavalin A for approximately 20 h prior to transfection and were then re-exposed to stimulus for another 2 days. The immunomagnetic separation of the CD8 ϩ fraction was performed prior to harvest on day 3 post-initial activation. Significant increases in luciferase activity were observed in all cell types, except EL4, as the 3Ј-CBF and AP-1 binding sites (contained within Ϫ169) and the 5Ј-CBF and Ikaros binding sites (contained within Ϫ243, see below) were included in the constructs (Fig. 3B). We did ob-serve, however, a low but significant level of luciferase expression in EL4 cells upon stimulation with PMA/ionomycin. Apparently, two major elements, one between Ϫ108 and Ϫ169 and another between Ϫ169 and Ϫ243, seem to be very important for the high levels of reporter gene expression observed from the Ϫ243 fragment and for the inducibility by PMA/ionomycin in EL4 cells. These transfection studies indicate that the minimal granzyme B promoter confers high levels of expression in transient assays but is not necessarily restricted to T cells.
The Granzyme B Ϫ243 Promoter Contains Binding Sites for Four Known Transcription Factors-Within the Ϫ243 promoter fragment there exist consensus sequence binding sites for the transcription factors AP-1, core binding factor (CBF), Ikaros, and the cyclic AMP responsive element (CRE). These sites are located at approximately Ϫ200 (Ikaros), Ϫ180/Ϫ126 (CBF), Ϫ150 (AP-1) and at Ϫ90 (CRE) nucleotides relative to the transcription start site (Fig. 4).
Nuclear extracts were prepared from MTL 2.8.2 T cells, L cells, and PMA/ionomycin-stimulated and unstimulated EL4 cells and incubated with oligonucleotides containing the granzyme B AP-1 and CBF sequence elements. Electrophoretic mobility shift assays (EMSAs) showed that both binding sites formed specific complexes with nuclear proteins present in all of these cells (Fig. 5A). These results show that the AP-1 and CBF regulatory factors are present in the nuclei of cells that both express and do not express granzyme B.
We then performed mobility shift assays using nuclear extracts from purified CD8 ϩ murine splenocytes to compare complexes in resting and activated cells. Fig. 5B shows that the granzyme B AP-1 oligonucleotide formed a complex with nuclear extracts from stimulated CD8 ϩ cells whereas this complex was absent in resting splenocytes. It has been previously observed that c-Fos mRNA is absent in resting T cells (Jain et al., 1992). A supershift was observed in activated cells with a c-Fos antibody although the c-Jun antibody used in this assay had a negligible effect on the complex in activated CD8, EL4, and L cells. These results indicate that in resting splenocytes the AP-1 complex is either absent or does not bind to DNA, and activation through the T cell receptor is required for effective DNA binding activity.
The granzyme B CBF oligonucleotide formed two complexes in CD8 ϩ lymphocytes (Fig. 5C). In nuclear extracts obtained from resting splenocytes, a weak, indistinct complex was

FIG. 3. Transient transfection analysis of granzyme B (C11) promoter deletion fragments in T cells and L cells.
A, the granzyme B promoter/luciferase deletion series was transfected into a Type 1 (CTL21.9) cytotoxic T cell line and relative expression values for one experiment are shown as a luciferase/␤-galactosidase activity ratio. B, various promoter/luciferase constructs were transfected into MTL 2.8.2, stimulated CD8 ϩ splenocytes (immunomagnetically separated), L cells, and EL4 helper cells (resting or activated for 20 h with 10 ng/ml PMA and 4 M ionomycin, black bars).
FIG. 4. The nucleotide sequence of the granzyme B proximal promoter. Major potential transcription factor binding sites are in boldface, and the transcription factor designations are shown above the sequences (AP-1, activator protein-1; CBF, core binding factor; and CRE, cyclic AMP responsive element). In vitro footprints are shown in boxes. Upward arrows denote protected G residues of in vivo footprints, and an asterisk denotes the hypersensitive residue. The open arrows indicate hypersensitive residues found only in MTL 2.8.2. The transcription start site is indicated with an arrow, and the translation initiation codon (ATG) is indicated in boldface.
formed with the CBF oligonucleotide. Upon stimulation of the cells for 45 h with ␣CD3, a slower mobility complex was observed. The complex was inhibited when the granzyme B CBF oligonucleotide was incubated with nuclear extracts in the presence of a 50 or 200 molar excess of unlabeled granzyme B CBF oligonucleotide. An interesting observation was made when antisera to either PEBP2 ␣A (CBF ␣-subunit) or PEBP2␤ (CBF ␤-subunit) was added to the reaction. The anti-␣A serum did not appear to affect the major complex; however, the anti-␤ serum was capable of disrupting the complex. The ␣A antiserum cross-reacts with all three of the known ␣ subunits, 2 including ␣B and ␣C, which are expressed at high levels in T cells. We may be observing an as yet unknown variation of the ␣ subunit or an entirely different protein that is capable of binding to the CBF/PEBP2 binding site and interacting with the ␤ subunit.
It has been determined previously that the Ikaros gene gives rise to a lymphoid-restricted family of functionally distinct transcription factor proteins which are involved throughout lymphocyte development Molar and Georgopoulos, 1994). The Ikaros element in granzyme B formed a specific complex that was present in both unstimulated and ␣CD3-stimulated CD8 ϩ splenocytes (Fig. 5D). This complex was competed off with an excess of the granzyme B binding site, whereas competition with a mutant Ikaros binding site did not affect the complex. To confirm the binding of regulatory factors along the granzyme B promoter, we performed in vitro DNase1 footprinting. When the granzyme B Ϫ243 promoter fragment was incubated with MTL 2.8.2 nuclear extracts and treated with DNase1, four areas of protection from DNase1 digestion were evident (Fig. 6). Two distinct footprints were visible over the AP-1 and the CRE sequence elements. Less distinct footprints were detected over the two CBF elements, the Ikaros element and a TATA-like element at Ϫ30. Together, this series of assays reveal that there are both activation-dependent and ubiquitous transcription factors binding to the granzyme B promoter in CD8 ϩ splenocytes.
Activated CD8 ϩ Splenocytes Possess a Strong DNase1 Hypersensitive Site That Is Absent in Resting CD8 ϩ Cells and L Cells-Permeabilization of cell membranes with lysolecithin 2 Y. Ito, personal communication.

FIG. 5. Nuclear extracts from T cells and non-T cells form complexes with oligonucleotide probes containing the granzyme B AP-1 and CBF binding sites.
A, nuclear extracts prepared from MTL 2.8.2, EL4 (unstimulated or stimulated with 10 ng/ml PMA and 4 M ionomycin), and L cells were incubated with the granzyme B AP-1 or CBF oligonucleotides and subject to electrophoresis on a nondenaturing polyacrylamide gel. DNA-protein complexes are marked. B, radiolabeled granzyme B AP-1 oligonucleotide incubated with 2 g of nuclear extracts prepared from unstimulated and ␣CD3-stimulated CD8 ϩ splenocytes. For supershift binding assays, 0.5 g of anti-c-Fos or anti-c-Jun antibody was added. C, the same nuclear extracts as in B were incubated with the radiolabeled granzyme B CBF oligonucleotide. 3 l of the ␣A or ␤ antisera was added as indicated. D, the same extracts used in B and C were incubated with the granzyme B Ikaros oligonucleotide. A 200 molar excess of the mutant Ikaros oligonucleotide was added as indicated. A 50 or 200 molar excess of competitor binding sites was added for all shifts. allows DNase1 to penetrate into living cells and cleave exposed regions of DNA within intact nuclei. Regions of chromosomal DNA that are accessible to, or are bound by, transcription factors tend to be hypersensitive to DNase1 digestion. DNase1 hypersensitive sites can be visualized on Southern blots with probes designed to specifically end-label a restriction fragment that contains the site of interest. The genomic granzyme B gene is contained within a 4.4-kb EcoRI fragment whose 5Ј end is 961 bp upstream of the transcription start site. A restriction fragment probe that extends from Ϫ828 to Ϫ546 relative to the transcription start site and does not contain repetitive sequence elements was used to end label this fragment.
We treated permeabilized, resting CD8 ϩ splenocytes with increasing amounts of DNase1 and observed that the 4.4-kb EcoRI fragment gradually decreased in intensity with higher concentrations of DNase1 (Fig. 7A). No specific sub-bands appeared that would indicate a hypersensitive region. In activated T cells, however, an area just proximal to the transcription start site was highly sensitive to DNase1 cleavage. Following activation of CD8 ϩ splenocytes for 3 days (IL-2, ␣CD3, concanavalin A, and ␣CD28), the 4.4-kb EcoRI fragment diminished with higher concentrations of DNase1 and a 750-to 900-bp sub-band appeared (Fig. 7B). The boundaries of this roughly 150-bp hypersensitive region directly correspond to the sequences which contain the AP-1, CBF, Ikaros, and CRE transcription factor binding sites. We were only able to detect a very faint hypersensitive site in L cell nuclei (Fig. 7C). Thus, even though there are nuclear factors in L cells that are capable of binding to granzyme B promoter sequences in vitro, they are not binding to the endogenous, chromosomal DNA. As a positive control, the constitutively active, c-Fos gene promoter was probed in resting splenocytes and L cells. Together, this series of experiments indicates that the granzyme B promoter undergoes a structural modification upon T cell activation that allows transcription factors access to the locus, and this phenomenon is cell type-specific and activation-dependent.
Granzyme B Activation Correlates with Binding of Regulatory Proteins in Vivo-In vivo footprinting is a powerful assay that permits direct detection of protein/DNA interactions within the intact appropriate cell type. Moreover, this method can discriminate between accessible and inaccessible protein binding sites in the chromatin of living cells. In vivo footprint analysis in MTL 2.8.2 cells showed that in intact cells there exist several regions in the granzyme B promoter that were protected from dimethyl sulfate (DMS) methylation (Fig. 8A). One such footprint corresponds to the AP-1 binding site, and two others correspond to the CBF binding sites. The footprint at the CRE is indicated by a reduction of the band correlating to the G residue midpoint in the binding site. Two hypersensitive bands were observed corresponding to the A residues that flank the Ikaros sequence element (Georgopoulos et al., 1992). CD8 ϩ splenocytes isolated by passage over a CD8 immunocolumn were treated directly with DMS or stimulated for 48 h with ␣CD3 and then subject to DMS treatment. A comparison of the resting and activated lanes in Fig. 8B shows that only in activated CD8 ϩ cells did footprints appear at the AP-1, CRE, and both CBF sites. A hypersensitive site was detected at the 5Ј boundary of the AP-1 recognition sequence. Bands corresponding to the internal G residues of both the AP-1 and CBF sites were reduced. The hypersensitive A residues of the Ikaros element observed in MTL 2.8.2 were not apparent, probably due to the resolution limits of the gel. The in vivo footprints correlate well with the in vitro data and present a snapshot of the endogenous promoter as it is activating transcription. The absence of proteins bound in resting cells suggests that the transcription factors that are present do not have access to the DNA, and, moreover, it suggests a lack of repressor interactions. The resting cell profile was invariably identical with the naked DNA control in all experiments performed. In nonexpressing L cells, no protein-DNA interactions were observed (data not shown). These data indicate that the AP-1, CBF, and CRE binding sites are not occupied by their respective transcription factors in resting splenocytes and that T cell activation is required for DNA binding.

DISCUSSION
Each peripheral cytotoxic lymphocyte spends its life in a continuous search for the foreign antigen that can turn it into a potent killer. The acquisition of cytotoxic function requires de novo synthesis and assembly of the killing machinery. The granzyme B gene encodes one of the components of this machinery and is fully activated within 3 days. The major thrust of our experiments has been directed at understanding granzyme B gene induction in physiologically relevant primary lymphocytes.
We have used purified CD8 ϩ splenocytes in reporter gene transfections, DNase1 hypersensitivity analysis, mobility shift assays, and in vivo footprinting. There was a recapitulation of general motifs that have been noted in regulation studies of other inducible or developmentally regulated genes. cis-Acting transcriptional enhancers of tissue-specific genes tend to consist of a cluster of ubiquitous and tissue-specific transcription factor binding sites. The granzyme B promoter region contains FIG. 6. In vitro DNase1 footprint analysis with MTL nuclear extracts shows clear footprints at the AP-1, CBF, and CRE elements. A less prominent footprint over the Ikaros element is also observed. 50 or 75 g of MTL 2.8.2 nuclear extracts were incubated with an end-labeled granzyme B promoter fragment and subject to digestion with DNase1. The cleavage products were separated on a 7% 7 M urea-polyacrylamide gel. A control reaction containing the cleavage products of naked DNA and a Maxam-Gilbert GϩA cleavage ladder were included in the two left lanes.
binding sites for the widely expressed AP-1 and CRE transcription factors as well as for the T cell-specific core binding factor and Ikaros. These sequences can bind to nuclear proteins present in activated T cells and are sufficient to activate high levels of reporter gene expression in transient transfections. These sequences are important for granzyme B regulation in vivo as every binding site is fully accessible to, and bound by, transcription factors in activated CD8 ϩ T cells. The spacing of these sequences is interesting in that the close proximity may be necessary for the concerted enhancer effect of several differentially regulated factors at this specific locus.
These trans-acting factors must be activated at the appropriate time by developmentally or externally derived signals. The Ikaros protein is expressed throughout T cell development. We have shown that it is present in resting CD8 ϩ splenocytes and capable of binding DNA. Our data also show Ikaros to be bound in vivo in MTL 2.8.2 cells. We showed by EMSA and in vivo footprinting that the AP-1 transcription factor does not bind to its sequence element in the granzyme B promoter without prior activation through the T cell receptor. The transcription factor AP-1 is comprised of Fos and Jun heterodimers whose activity is regulated by de novo synthesis as well as FIG. 7. DNase1 hypersensitivity analysis in resting and activated CD8 ؉ splenocytes and L cells shows that the granzyme B promoter is hypersensitive to DNase1 digestion in activated T lymphocytes. Lysolecithin-permeabilized cells were treated with between 0 and 10.0 g/ml DNase1. Genomic DNA was cut with EcoRI, separated on 1.2% agarose gels (15 g/lane), and transferred to nylon membranes. The blots were probed with a 279-bp 5Ј-granzyme B probe that indirectly end-labels a 4.4-kb EcoRI fragment whose 5Ј end is 961 bp upstream of the first transcribed nucleotide. A, the Southern blot of CD8 ϩ unstimulated splenocytes (6-day exposure). Longer exposures of up to 2 weeks failed to reveal any hypersensitive sites. B, hypersensitive site formation in CD8 ϩ splenocytes stimulated for 3 days with IL-2, ␣CD3, concanavalin A, and ␣CD28 (2-week exposure). C, the granzyme B locus in L cells (6-day exposure).
post-translational modification. Upon activation through the T cell receptor, they are phosphorylated through a protein kinase C/Ca ϩ -mediated signal transduction cascade (reviewed in Crabtree and Clipstone (1994), Karin and Smeal (1992), and Rincon and Flavell (1994)). The phosphorylated AP-1 complex can bind efficiently to its cognate DNA sequence element and act as a potent transactivator of transcription. Core binding factor has been implicated in the regulation of many T cellspecific genes such as the T cell receptor ␣, ␤, and ␦ genes, the CD3 ⑀ and ␦ genes (Hallberg et al., 1992;Hsiang et al., 1993;Prosser et al., 1992). CBF consists of two heterologous subunits, a DNA binding ␣ subunit and a non-DNA binding ␤ subunit Zaiman et al., 1995). Through heterodimerization, the ␤ subunit augments the DNA binding affinity of the ␣ subunit. The ␣ subunit has been shown to localize to the nucleus while the ␤ subunit is found in the cytoplasm (Lu et al., 1995). Little is known about the signaling events that lead to the translocation of the ␣/␤ heterodimer into the nucleus. The hypothesis put forth suggests that the ␣ subunit requires modification to make it more amenable to association with the ␤ subunit. This modification would occur after the cell has been activated through the appropriate cell surface receptors. Results of mobility shifts show a weak com-plex in resting cells that could indicate a low affinity, partially dissociated ␣ subunit complex. The complex in activated cells was much more intense and well defined.
Both CBF and AP-1 are able to act in combination with other transcription factors, most notably ets and NF-AT (nuclear factor of activated T cells) (Jain et al., 1993;Wotton et al., 1994). In the regulation of the IL-2 gene, NF-AT translocates to the nucleus in response to increases in intracellular levels of calcium. The nuclear NF-AT then unites with activated Fos/ Jun to form a complex with high affinity DNA binding and transactivation properties. Recently, it has been determined that there is more than one form of NF- AT (Northrop et al., 1994). Core binding factor is encoded by members of a multigene family, and one member of the ␣ subunit family has been shown to be T cell-specific ). An attractive model, similar to IL-2 gene induction, is that upon T cell activation CBF and AP-1 are modified by separate signaling pathways and then unite at the level of DNA binding and transactivation.
Chromatin is no longer viewed as a passive participant in eukaryotic gene regulation. Activated trans-acting factors must be allowed access to their target cis-enhancer elements only in the appropriate tissues and at the appropriate times. The formation of nuclease hypersensitive sites in chromatin has been correlated with such important regulatory elements as enhancers, silencers, and locus control regions (reviewed in Felsenfeld (1993) and Gross and Garrard (1988)). For example, the human ␤-globin locus is composed of five developmentally regulated genes that are induced and expressed sequentially during embryonic, fetal, and adult development. The timely expression of these genes is controlled by a series of four stage-specific DNase1 hypersensitive sites that exist many kilobases upstream of the 5Ј-most gene in the cluster (Fraser et al., 1993). These hypersensitive sites consist of binding sites for ubiquitous and erythroid-specific transcription factors. Their stagespecific appearance has been shown to be dependent upon interaction with individual gene promoters of the globin genes (Reitman et al., 1993). It appears in this case that a cluster of binding sites alone is not sufficient to create a hypersensitive site, but an interaction with a distant element is necessary to induce structural changes in the locus.
The granzyme gene locus is potentially very interesting in this respect. We do not know whether the granzyme B proximal promoter sequences alone are sufficient to create the observed hypersensitive site in activated T cells or if another sequence element located elsewhere is required. The possibility may exist that all or a subset of the granzyme genes could be coordinately regulated by a higher level of control that involves the interaction of the individual promoters with a distant locus control region that would make the genes amenable to transcription upon the reception of the appropriate induction signal. We have looked for additional hypersensitive sites up to 3.5 kb 5Ј of the transcription start site and have found none in this region.
There is increasing evidence that in vitro footprinting assays do not always reflect the true DNA/protein interactions occurring in the chromatin of intact cells. The state of chromatin condensation may sequester cis-elements and prevent binding of available transcription factors. In addition, the nonexpressing cells showed little evidence of a DNase1 hypersensitive site in the vicinity of the granzyme B gene. Interestingly, a hypersensitive site was apparent in activated but not resting CD8 ϩ lymphocytes. An inducible hypersensitive site has been observed at the human granzyme B promoter in a PEER T cell line when activated with TPA and dibutyryl-cAMP . This site extends from approximately 30 to 400 FIG. 8. In vivo footprint analysis in MTL 2.8.2, stimulated and unstimulated CD8 ؉ T cells show proteins bound at the CRE, AP-1, and core binding factor sites. Genomic DNA was isolated following DMS treatment of cells and subject to piperidine cleavage and ligation-mediated PCR as described under "Experimental Procedures." The PCR products were separated on a 7% polyacrylamide, 7 M urea gel.
The MTL in vitro lane shows the DMS-LMPCR cleavage pattern of naked DNA. A footprint is observed as bands that are reduced in intensity relative to the control DNA or as hypersensitive sites. These bands are indicated with arrows. Bands that are reduced in intensity appear at the 5Ј-and 3Ј-CBF elements and at the AP-1 and CRE elements in MTL and stimulated CD8 ϩ relative to control DNA or to unstimulated CD8 ϩ cells. A hypersensitive site is detectable at the boundary of the AP-1 element in both the MTL 2.8.2 and stimulated CD8 ϩ cells. Hypersensitive sites are also observed within the Ikaros binding site in MTL 2.8.2. These sites are indicated with asterisks. nucleotides upstream of the transcription start site and roughly corresponds to the promoter sequences that are highly conserved between mice and humans. We were able to define more precisely the limits of the murine hypersensitive site in activated splenocytes and show that transcription factors form sequence-specific footprints at this site in vivo.
This series of experiments, together with the Northern analysis, transfection, and EMSA data, leads us to infer that the granzyme B locus may not be accessible to transcription factors in resting T cells or non-T cells. Ikaros and CREB proteins are present in resting cells but only upon activation of the T cells are the AP-1 and CBF capable of binding to their cognate sequence elements and, at this time, the promoter becomes accessible to them. There is presumably a highly coordinated mobilization of factors once the decondensation signal takes place. The promoter must be sensitive to multiple signaling events, and it is possible that the assembly of factors occurs only after all of the signaling events have been achieved. The IL-2 enhancer displays an all or nothing chromosomal binding phenomenon in EL4 cells stimulated with TPA/A23187 . When cyclosporin A treatment was used to block the activation of the calcium-dependent NF-AT and NF-B, transcription factors that normally bind to the IL-2 enhancer, no binding of even cyclosporin-insensitive factors was observed in vivo and the enhancer was unoccupied.
It has yet to be determined whether the transcription factor binding sites become accessible because of a shift in the nucleosome structure at the promoter or if the transcription factors themselves mediate the formation of the hypersensitive site. Alternatively, the granzyme B locus may be differentially methylated in different cell types, thus influencing DNA protein binding access. The granzyme B locus is relatively insensitive to DNase1 digestion in L cell nuclei. This is not surprising because at no time in a fibroblast's existence does it produce granzyme B. Histone modifications, such as hyperacetylation of histone tails, or nucleosomal binding proteins such as nucleoplasmin and SWI/SNF Kwon et al., 1994;Wolffe, 1994;Workman and Buchman, 1993) are potentially important players in gene induction. These proteins have been shown to facilitate transcription factor binding and nucleosome disruption and may be essential components of the chromatin rearrangement and transcriptional induction processes. Whether these molecules or others play a role in granzyme B regulation in vivo awaits further investigation.
In this investigation we have confirmed that more than one level of regulation is required to permit the expression of granzyme B. Our proposed model of granzyme B regulation involves decondensation at the chromosomal locus in response to T cell activation. The integration of multiple signals would culminate in the synthesis or activation of the necessary transcription factors and result in a staged assembly of factors at the newly accessible granzyme B promoter. The subsequent three dimensional structure would then activate the basal transcription machinery to initiate transcription and elongate the nascent RNA. This is similar to the model proposed for the regulation of IL-2 transcription . Both systems require timely activation in response to lymphocyte stimulation and must be readily reversible in order to terminate the immune response.