Characterization of the Nuclear and Cytoplasmic Components of the Lymphoid-specific Nuclear Factor of Activated T Cells (NF-AT) Complex*

The lymphoid-specific transcription complex, NF-AT, is involved in early gene activation in T cells and is assembled from a pre-existing, T cell restricted cytoplasmic factor and an inducible ubiquitous nuclear component within 30 min after activation through the antigen receptor. Recent studies have implicated the family of AP1 factors as components of the murine NF-AT complex. Evidence is provided here that the nuclear component of human NF-AT contains the phorbol es-ter-inducible transcription factor AP1 (Jun/Fos). We further characterize which AP1 family members can assume this role. Antisera to Fos inhibits NF-AT DNA binding as does an oligonucleotide containing a binding site for AP1. Constitutive expression in vivo of Fos, and to a lesser extent Fra-1, eliminates the requirement for phorbol12-myristate 13-acetate (PMA) stimulation, leaving NF-AT-directed transcription responsive to calcium ionophore alone. Overexpression of cJun or JunD, but not JunB, also eliminates the requirement for PMA, indicating that many but not all Jun- and Fos-related proteins functionally activate NF-AT-dependent transcription in the presence of the cytoplasmic component. NF-AT DNA binding can be reconstituted in vitro using semi-purified AP1 proteins mixed with cytosol from T lymphocytes. Fos proteins are

AT can be reconstituted from a ubiquitous nuclear component that requires protein synthesis for induction and a T cellspecific constitutive cytoplasmic component (NF-AT,). This cytoplasmic component associates with the nucleus in response to calcium signaling in a manner that is inhibited by the immunosupressive drugs cyclosporin A (CsA) and FK506. Like the nuclear component of NF-AT, AP1 activity can be induced in many different cell types. AP1 can contain any one of a group of related proto-oncogenes, the Jun family (8)(9)(10)(11)(12), which are capable of binding to specific responsive elements as homodimers (8)(9)(10)(11)(12)(13) or as heterodimers with the product of the c-fos gene (13)(14)(15)(16)(17) or with other Fos-related proteins (18-20). The nuclear component of NF-AT can be induced with PMA, is not sensitive to CsA or FK506, and can be seen in cells of non-T cell origin such as HeLa (7) and Cos cells.* Transcription of the proto-oncogene c-fos occurs within 30 min following treatment of T cells with PMA (21,22) and is insensitive to CsA and FK506 (22). In light of these similarities, we have investigated whether Fos or a Fos-related antigen (Fra) might participate in the formation of the NF-AT complex.
Recent studies (23) have implicated Fos as a component of the murine NF-AT complex; however, the antisera used in this work recognizes the Fos M peptide which is essentially conserved in every Fos-related protein thus far identified including FosB, FosB2, Fra-1, and Fra-2 (18-20). The nuclear portion of the murine NF-AT complex binds to an oligonucleotide derived from the IL-2 enhancer containing a 6 of 7 base pair match with a consensus AP1 binding site and may contain one of the known Jun proteins (23).
Using a T cell line expressing the SV40 T antigen, we have characterized the nuclear and cytoplasmic components of human NF-AT. We find that any of three Jun family members can complement T cell cytosol to form NF-AT binding activity in vitro but only cJun and JunD, not JunB, are able to functionally replace PMA induction of this component in uiuo. Both Fos and Fra-1 can functionally replace PMA induction, and although Fos protein can participate in NF-AT DNA binding activity, it is not necessary; Jun alone can form ' The abbreviations used are: NF-AT, nuclear factor of activated T cells; HNF-la, hepatocyte nuclear factor-la; PMA, phorbol 12myristate 13-acetate; IL-2, interleukin-2; Oct-1, octamer factor-1; Fra, Fos-related antigen; CsA, cyclosporin A; Tag, T antigen; PAGE, polyacrylamide gel electrophoresis; PMSF, phenylmethylsulfonyl fluoride; DTT, dithiothreitol; PAGE, polyacrylamide gel electrophoresis; CAT, chloramphenicol acetyltransferase.
J. Northrop and G. Crabtree Gel Mobility Shifts-A nuclear extract from Jurkat T cells stimulated for 3 h. with 20 ng/ml PMA (Sigma) and 2 p~ ionomycin (Calbiochem) was prepared as described (26). Rat liver nuclear extract was a generous gift of D. Mendel and was prepared as described (27). Nuclear extracts were preincubated for 30 min a t room temperature in a buffer containing 10 mM Tris-HCI, pH 7.5, 50 mM NaCI, 0.5 mM EDTA, S% glycerol, and 1.7 pg of poly(d1-dC) with or without antisera (0.5 pl) or competitor oligonucleotides (25 ng, 200-fold excess). Endlabeled oligonucleotide probes were then added to each mixture and incubations carried out for another 45 min a t room temperature. DNA-protein complexes were resolved on 4% polyacrylamide, 0. and HNF (828 probe as defined, Ref. 29). The sequence of the immunoglobulin K-chain enhancer E-box containing oligonucleotide is as follows: (5'-GATCAAGGCAGGTGGCCCAGATC-3'). Antisera to HNF-la has been described (27) as has the antisera to Fos (30). Antisera specific for JunD was raised in Balb/c mice using a protein containing a region of human JunD (gift of N. Nomura) spanning from amino acid 44 to 193 fused to the carboxyl terminus of glutathione S-transferase (Pharmacia). The fusion protein was purified FIG. 1. NF-AT contains an A P l binding activity and is immunologically cross-reactive with Fos. A, inhibition of NF-AT and AP1 DNA binding activities using anti-Fos antibodies. Jurkat nuclear extract (lanes 1-9) or rat liver nuclear extract (lanes [10][11][12] was preincubated with no additions (lanes I , 4, 7, and IO) or with 0.5 pl of anti" peptide antisera (lanes 2,5. 8, and 1 I ) or 0.5 pl of anti-HNF-la antisera (lanes 3, 6, 9, and 12) before adding the indicated end labeled oligonucleotide probes. Specific complexes are indicated by arrows.
B, cross-competition between NF-AT and AP1. Gel mobility shift assays were done using Jurkat nuclear extract as in A except that preincubations were done with either no competitor or with the specified competitor. Lanes 1-4, using the NF-AT probe; lanes 5-8, using the AP1 probe. using glutathione-agarose and mixed with trehalose dimycolate emulsion (RIB1 Immunochem Research, Inc.) prior to intraperitoneal injection.
Tra~~~ections-Luciferase reporter constructs were as follows. NF-ATLuc was provided hy Peter Kao and contains three copies of the NF-AT binding site (-286 to -257 of human IL-2) linked to the human IL-2 promoter (-72 to +47) driving luciferase (31). NFKRLuc contains three copies of the human immunoglobulin r-chain enhancer B site (22) in place of the three NF-AT binding sites, and II,"LLuc contains the human IL-2 enhancer/promoter sequence -326 to +47. The expression constructs SVFos, RSVcJun, RSVJunH, and RSVJunD encoding murine cFos, c.Jun, JunR, and .JunD, respectively.
were a gift of M. Yaniv. An expression vector for murine JunD tagged with 6 histidine residues at the amino terminus was constructed by placing the coding region of murine junD into the hacterial expression vector QE-9 (Quiagen). An EcoRI fragment containing the sequences encoding the histidine t a i l as well as the entire open reading frame of junll was then cloned into the EcoRI site of pR.JS, placing it under the control of the SRn promoter. These manipulations added the sequence Met-Arg-Gly-Ser-His-His-His-His-His.His~~lv-Ser.l'ro to the amino end of dunD. Transfection of this construct into .Jurkat cells indicates that it remains functional.' An expression vector for murine cFos tagged with 6 histidine residues at the amino terminus was constructed by cloning genomic sequences hetween the unique RglI site and the NaeI site immediately downstream from the stop codon into the vector QE-9. A fragment containing the histidine tag and these genomic sequences was cloned into the I.:coRI site of pRJ5 as above. This procedure removed the first 21 amino acids of Fos and replaced them with the sequence Met-Arg-Cly-Ser-His-His-His-his^ His-His-Gly-Ser-Val-Asp. This construct also remains functional as assayed by transfection into Jurkat cells.' Expression vectors for rat cFos and rat Fra-1 were constructed by cloning I,'coRI fragments containing the entire coding sequence for each of these proteins into the EcoRI site of pRJS. Jurkat cells expressing the SV40 T antigen were transiently transfected hy electroporation (240 V, 960 p F ) with luciferase-based reporter constructs for NF-AT, NFxR, and the 11,-2 enhancer/promoter and with expression vectors for cFos, Fra-1. cJun. dunR JunD, or pRdS control plasmid. Each transfection contained 2 pg of reporter plasmid and 4 pg of expression construct. Stimulations with 2 p M ionomycin plus or minus 20 ng/ml PMA were started 24 h after transfection and were allowed to proceed for 7 h before harvesting cells for luciferase assays (31).
Reconstitution of NF-AT--Jurkat cells expressing the SV40 T antigen (approximately 1.2 X 10" cells) were transfected hy electroporation as above using either 10 pg of HislunD. HisFos, or SVFos expression vectors or 20 pg of .JunI) or JunR expression vectors as indicated. After 36 h, whole cell extracts were prepared by suspending cells in 300 pI of buffer containing 20 mM HEPES, pH 7.8. 0.4 M KCI, 20% glycerol, 2 mM DTT, and 0.2 mM I'MSF on ice. This cell suspension was frozen and thawed on ice two times and then spun at 100,000 X g for 12 min a t 4 "C. The supernatant was diluted with an and then bound proteins were eluted with three 150-pl aliquots of buffer B containing 80 mM KC1 and 100 mM imidazole. Aliquots of the column load and flow-through (0.7% of total, 5 pl) and elutions (3.3% of total, 5 pl) were run on SDS-PAGE and blotted with the JunD-specific antisera described above. Binding of the primary antisera was detected with a horseradish peroxidase-conjugated rabbit anti-mouse antibody (Zymed) and the enhanced chemiluminescence system (Amersham Corp.). Ni2+ column fractions (elution 2, 5 pl total), either alone or mixed with cytoplasmic extract (5 rg, prepared as described (7)), plus or minus competitor oligonucleotide (25 ng) were allowed to associate for 15 min in a buffer containing 10 mM Tris-HC1, pH 7.5, 40 mM NaCl, 0.5 mM EDTA, 5% glycerol, and 300 ng of poly(d1-dC) before addition of labeled oligonucleotide. Reactions were further processed as described above under gel mobility shifts.
Samples were dialyzed against buffer B above containing 50 mM KC1 and 0.1 mM EDTA in a system 100 microdialyzer (Pierce Chemical Co.). Aliquots of these samples (1 pl) and 5 pg of enriched NF-ATc starting material were analyzed by SDS-PAGE. NF-ATc was assayed in 10 p1 of these fractions by mixing with 4 pg of nuclear extract prepared from HeLa cells stimulated for 2 h with 20 ng/ml PMA as described (7). Gel shift conditions were as described above under "Gel Mobility Shifts."

RESULTS
As shown previously (23), an antisera against the conserved M peptide region of Fos (30) inhibited NF-AT binding activity from nuclear extracts of Jurkat T cells; however, there is no inhibition by an antisera which recognizes hepatocyte nuclear factor-la (HNF-la), a liver-enriched transcription factor not found in T cells (27) (Fig. L4, lanes 1-3). As expected, similar results with the anti" peptide antisera were found using an oligonucleotide to detect AP1 binding proteins (Fig. lA, lunes  4-6). In contrast, no inhibition is seen of Oct-1 DNA binding (lanes 7-9) or of HNF-la DNA binding (lanes 10-12). The H N F -l a antisera induces a supershift of HNF-la of slower mobility (lune 12) as observed previously (27). To investigate whether an entire AP1 binding activity forms a component of NF-AT, we performed competitions with the cognate binding sites for these two factors (Fig. 1B). As expected, AP1 binding activity was effectively competed by excess AP1 oligonucleotide but not by the binding site for NF-AT or the K E~ site of the immunoglobulin K enhancer (lanes 5-8). This implies that the NF-AT binding site does not contain a cryptic AP1 binding site. Surprisingly, while the K E~ site does not compete for the NF-AT binding activity, the AP1 site competes effec- tively (lanes 2-4). These results suggest that NF-AT contains an AP1 binding activity and that Fos or a Fra can participate in the formation of the nuclear component of NF-AT.
T o determine if AP1 binding proteins could replace the requirement for PMA stimulation in NF-AT-directed transcription, we cotransfected either murine or rat c-fos with a reporter gene under the direction of NF-AT. Complete independence from PMA stimulation is observed, whereas the requirement for a calcium signal remains intact (Fig. 2 A ) . Murine Fos appears more effective than rat Fos in this assay. Cotransfection of an expression vector for rat Fra-1, which was isolated by virtue of the fact that it is immunologicaIly cross-reactive with Fos (32), also replaces the PMA signal, although less efficiently than rat Fos. There was no effect of c-fos cotransfection on NFKRor IL-2-directed transcription ( Fig. 28), indicating that Fos cannot substitute for all PMA generated signals necessary for IL-2 expression. Since Fos does not bind to an APl site as a monomer or homodimer (14)(15)(16)(17), it seemed likely that the competition results in Fig. 1B could best be explained if Fos or a Fra were complexed with a Jun protein capable of binding to an AP1 site. Cotransfection of expression vectors for either c-jun orjunD resulted in partial independence from the PMA signal for induction of NF-AT transcriptional activity, whereas expression of junR had no effect ( Fig. 2A ).
To show directly that Jun proteins are capable of participating in NF-AT DNA binding activity, we overexpressed J u n D tagged with histidine and nontagged Fos in Jurkat cells which constitutively express SV40 T antigen. Extracts from these cells were passed over Ni"-nitrilotriacetic acid-agarose affinity columns and eluted with imidazole. A Western blot using a JunD-specific antisera shows that the His-tagged J u n D is found predominantly in the second fraction eluted from the column, whereas the column flow-through is effectively depleted (Fig. 3A, upper panel). In contrast, wild-type JunD remains in the flow-through and is not eluted off the Ni'+ column (Fig. 3A, lower panel). To control for nontagged proteins in the extract which elute with the His-tagged JunD, the second elution fraction from each column was used in reconstitution experiments (Fig. 3 R ) . Although neither Jurkat cytosol nor the HisJunD elution alone contain NF-AT binding activity (lanes I and 2 ) , mixing these two fractions (lane 3 ) resulted in regeneration of NF-AT binding activity which is specifically competed by excess NF-AT oligonucleotide hut not by a nonspecific oligonucleotide (lanes 8 and 9). T h e wildtype JunD does not appear in the column elution and, as expected, this elution is incapable of reconstituting an NF-A T gel shift (lanes 4  overexpressed Fos copurifies with the HisJunD in these experiments by virtue of their strong interaction through the leucine zipper motif (17, 33-35). The presence of JunD/Fos heterodimers is shown in Fig. 3, C and D (see below). Finally, Cos cell cytosol alone or when combined with the HisJunD elution ( lanes 6 and 7) fails to regenerate NF-AT binding activity. This confirms the T cell-specific nature of the cytosolic component (7). We tested the same elution fractions alone or when mixed with Jurkat cytosol for AP1 binding (Fig. 3C). The column elution from the His-tagged JunD transfection contains AP1 binding activity even in the absence of cytosol, as expected (lane 2 ) , whereas the corresponding elution for the nontagged JunD does not (lane 4 ) . Two specific AP1 DNA binding activities appear in Fig. 3C, an upper sharp band and a lower more diffuse band. Expression and purification of His-tagged JunD alone gives rise to the upper band (Fig. 30, lane 1  are coexpressed (lune 4 ) . We interpret these results as indicating the presence of Jun/Jun homodimers (upper bund) and Jun/Fos heterodimers (lower bund).
Since expression of JunD alone replaces the PMA requirement for NF-AT-directed transcription, we tested whether JunD alone could reconstitute NF-AT DNA binding in the presence of the cytoplasmic component. Addition of cytosol to JunD homodimers (Fig. 4, lunes 7-9) results in reconstituted NF-AT DNA binding with the correct specificity (lunes [10][11][12]. This result is consistent with the functional data presented above. For comparison, reconstitution using coexpressed JunD and Fos is shown in Fig. 4A, lunes 1-6. Jurkat cells do not express Fos in the absence of PMA stimulation (22); however, JunD is expressed a t low levels in nonstimulated Jurkat cells:' and therefore is available to pair with exogenously expressed Fos. This may explain why expression of Fos or Fra-1 alone can functionally replace the PMA signal. T o investigate whether Fos alone could reconstitute NF-AT DNA binding, cytosol was added to the nickel column elution shown in Fig. 30, lane 2, and in Fig. 4A, lunes 13-15. A very small amount of NF-AT can be detected (compare lunes 16 and 25). These results indicate that the functional assay with the luciferase reporter is far more sensitive than the gel shift assay. Coexpression of JunD with HisFos results in a large amount of Jun/Fos heterodimer (Fig. 4A, lunes 19-21) and consequently a large increase in the amount of reconstituted NF-AT (lanes 22 -24). As shown in Fig. 3, the reconstituted NF-AT DNA binding activity correlates with the presence of immunoreactive JunD. As expected, JunD is easily detected in fractions able to reconstitute appreciable NF-AT DNA binding (Fig. 4B) with the exception of HisFos/JunD (Fig.  4B, lune 12, E2). A longer exposure of this blot shows some JunD copurifying with the HisFos (data not shown), indicating that the Western blot is yet less sensitive than the gel shift assay. Since JunB is unable to transactivate the NF-ATLuc reporter (Fig. %I), we determined whether it could participate in the NF-AT complex. Coexpression of JunB with HisFos enhances reconstituted NF-AT DNA binding as K. Ullman and G. Crabtree, unpublished observations. did JunD (Fig. 4A). This indicates that JunB can participate in a nonfunctional NF-AT complex.
T o begin characterization of the cytoplasmic component of NF-AT, we fractionated Jurkat cytosol enriched in NF-ATc by SDS-PAGE and renatured proteins eluted form gel slices. These fractions are shown in Fig 5A. Using HeLa nuclear extract as a source of the nuclear component (7), we were able to reconstitute NF-AT DNA binding activity specifically from three fractions (fractions 4-6, Fig. 5B) but mostly from fraction 5 (94-116 kDa). The reconstituted complexes migrate slower in the gel with higher molecular mass fractions, and each forms a subset of the migration of the NF-AT reconstituted from the starting material ( L ) . This may indicate the presence of multiple forms of the cytoplasmic component.

DISCUSSION
Based on our results, we propose that human NF-AT contains both a PMA-inducible and relatively ubiquitous AP1 binding activity, forming the nuclear component, as well as a T cell-restricted cytosolic component. Recent work (23) has implicated AP1 in the formation of the inducible murine NF-AT complex. Our results confirm this notion for the human NF-AT complex. We further show that expression of Fos, Fra, and Jun proteins can eliminate the need for PMA induction of the nuclear component and characterize which Jun proteins are functional and which are able to reconstitute NF-AT binding in the presence of the cytosolic component.
The nuclear component of human NF-AT appears to contain both a Jun family member and Fos or a Fra. Functionally, both cJun and JunD operate to replace the PMA requirement as does Fos and Fra-1. Although the effect of cJun or JunD is not nearly as pronounced as that of Fos, this may reflect known differences between the transcriptional activation potential of cJun homodimers as compared with cJun/Fos heterodimers at a 12-0-tetradecanoylphorbol-13-acetate-responsive element (15, 36). Although JunB is able to participate in the NF-AT complex, it is not able to activate an NF-ATdependent reporter. This finding is provocative given previous studies (37,38) showing differences in the transactivation potential and mutual antagonism between cJun and JunB on certain 12-0-tetradecanoylphorbol-13-acetate-responsive element constructs and in malignant transformation assays.
There is some evidence (23) that NF-ATc can bind a specific DNA sequence independently of the nuclear component. Thus, given the similarities between NF-AT and AP1 transactivation with regard to the activities of various Jun and Fos family members, NF-ATc may serve to couple APl to a new DNA binding site where Jun and Fos contribute all transcriptional activation functions. Antibody inhibition and functional studies implicate Fos or a Fra as a participant in the NF-AT complex. Although the native NF-AT complex includes Fos or a Fra, we have shown both functionally and by reconstitution of DNA binding activity that Fos is not necessary and that cJun and JunD are sufficient to replace the nuclear component. Expression of Fos of Fra-1 functionally activate NF-AT-dependent transcription in the presence of NF-ATc, likely because of constituitive low level expression of Jun proteins in Jurkat cells. 3 The Fos gene family contains a number of members already and is likely to grow; thus it remains a formal possibility that the native NF-AT complex in Jurkat cells contains an as yet unidentified Fra, pairing with Jun.
NF-ATc not only serves to couple AP1 to the NF-AT DNA binding site but also imparts dependence on a calcium flux and, thus, sensitivity to CsA and FK506. Therefore, with regard to calcium-mediated signal transduction in T cells and immunosupressant action, NF-ATc is a key component. We have partially purified this protein and shown it to have a molecular mass between 94 and 116 kDa. NF-ATc is present in more than one molecular mass range, indicating some heterogeneity in this factor. This heterogeneity is unlikely due to proteolysis during the denaturation-renaturation steps as the mobility of the reconstituted complexes all fall within the range of the native NF-AT complex isolated from the nuclei of stimulated Jurkat cells. Purification and cloning of the gene for NF-ATc will be necessary to fully answer these questions. Although the full nature of the cytosolic component of NF-AT is not known, the complex formed with AP1 acquires the specificity of NF-AT, thereby explaining how a ubiquitous transcription factor like AP1 can contribute to the biologically specific pattern of early gene activation in T cells.