Characterization of High Molecular Weight FK-506 Binding Activities Reveals a Novel FK-506-binding Protein as Well as a Protein Complex*

The immunoregulant FK-506 potently inhibits par- ticular calcium-associated signal transduction events that occur early during T-lymphocyte activation and during IgE receptor-mediated exocytosis in mast cells. FK-506 binds to a growing family of receptors termed FK-506-binding proteins (FKBPs), the most abundant being a 12-kDa cytosolic receptor, FKBP12. To date, there is no formal evidence proving that FKBPlP is the sole receptor mediating the immunosuppressive effects or toxic side effects of FK-506. Using gel filtration chromatography as an assay for novel FK-506- binding proteins, we identified FK-506 binding activities in extracts prepared from calf brain and from JURKAT cells. Both of these new activities comigrated with apparent molecular masses of 110 kDa. However, further characterization of both binding activities re- vealed that the two are not identical. The 110-kDa activity observed in brain extracts appears to be the FKBP12*FK-506*calcineurin (CaN) complex previ- ously J., Weissman, I., and Schreiber, S . (1991) Cell 66, 807-815) while the 110 kDa activity observed in JURKAT cells is a novel FK-506-binding protein. Our characterization of the FKBPlZ-FK-506-CaN complex reveals a dependence upon calmodulin (CaM) for formation of the complex and demonstrates that the peptidyl-prolyl cis-trans isomerase (PPIase) activity of FKBPlP is not required for binding of FKBP12-FK-506 to CaN or for inhibition of CaN phosphatase activ- ity. The novel FK-506-binding protein in JURKAT cells has been purified to homogeneity, migrates with an apparent mass of 51 kDa on denaturing gels, and has been termed FKBP51. Like FKBP12, FKBP51 has PPIase activity, but, unlike FKBP12-FK-506, FKBP51-FK-506 does not complex with or inhibit the phosphatase activity of, CaN. These results indicate that complex formation with CaN may not be a general property of the FKBPs. Peptide sequencing reveals that FKBP5l may be similar, if not identical, to hsp56, a component of non-transformed steroid receptors.

The immunoregulant FK-506 potently inhibits particular calcium-associated signal transduction events that occur early during T-lymphocyte activation and during IgE receptor-mediated exocytosis in mast cells. FK-506 binds to a growing family of receptors termed FK-506-binding proteins (FKBPs), the most abundant being a 12-kDa cytosolic receptor, FKBP12. To date, there is no formal evidence proving that FKBPlP is the sole receptor mediating the immunosuppressive effects or toxic side effects of FK-506. Using gel filtration chromatography as an assay for novel FK-506binding proteins, we identified FK-506 binding activities in extracts prepared from calf brain and from JURKAT cells. Both of these new activities comigrated with apparent molecular masses of 110 kDa. However, further characterization of both binding activities revealed that the two are not identical. The 110-kDa activity observed in brain extracts appears to be the FKBP12*FK-506*calcineurin (CaN) complex previously reported (Liu, J., Farmer, J., Lane, W., Friedman, J., Weissman, I., and Schreiber, S . (1991) Cell 66, 807-815) while the 110 kDa activity observed in JURKAT cells is a novel FK-506-binding protein. Our characterization of the FKBPlZ-FK-506-CaN complex reveals a dependence upon calmodulin (CaM) for formation of the complex and demonstrates that the peptidyl-prolyl cis-trans isomerase (PPIase) activity of FKBPlP is not required for binding of FKBP12-FK-506 to CaN or for inhibition of CaN phosphatase activity. The novel FK-506-binding protein in JURKAT cells has been purified to homogeneity, migrates with an apparent mass of 51 kDa on denaturing gels, and has been termed FKBP51. Like FKBP12, FKBP51 has PPIase activity, but, unlike FKBP12-FK-506, FKBP51-FK-506 does not complex with or inhibit the phosphatase activity of, CaN. These results indicate that complex formation with CaN may not be a general property of the FKBPs. Peptide sequencing reveals that FKBP5l may be similar, if not identical, to hsp56, a component of non-transformed steroid receptors. The macrolide FK-506 is a powerful immunosuppressant that, like the cyclic undecapeptide drug, cyclosporin A (CsA),' inhibits specific calcium-dependent signal transduction events leading to T-lymphocyte activation, selectively blocking transcription of a set of coordinately expressed, earlyphase genes crucial for lymphocyte growth and differentiation (2). Like CsA, FK-506 also selectively blocks calcium-dependent intracellular signaling events in other, possibly related, signal transduction pathways (3-5). FK-506 binds to cytosolic receptors (6) which are distinct from the major CsA receptor, cyclophilin A (7). A 12-kDa cytosolic FK-506-bindingprotein, FKBP12, was purified and characterized first (8,9) and, like cyclophilin A, has since been shown to be a member of a growing family of receptors termed FKBPs (10-12). In addition to binding immunosuppressive ligands, another property shared by the FKBPs and the cyclophilins is that they are both peptidyl-prolyl isomerases (PPIases), enzymes that catalyze isomerization about peptidyl-prolyl bonds (8,9,(12)(13)(14).
Like cyclophilin A, FKBPl2 is one of the most abundant cytosolic proteins in eukaryotes, is found in most tissues and cell types (15), and is extraordinarily well-conserved throughout phylogeny. These observations suggest that FKBPl2 has a critical and central role in cellular physiology and may explain why FK-506 has a number of toxic side effects in animals and man (16). A structural relative of FK-506, rapamycin, is another immunomodulator which also binds to the FKBPs but whose immunosuppressive activity is a consequence of a block in the proliferative response of T-cells to growth-promoting lymphokines (17). The observations that FK-506 and rapamycin bind to a family of receptors and affect multiple signaling pathways are indicative that FKBPs may have multiple and diverse roles within cells.
Recently, it has been demonstrated that FKBPl2. FK-506 binds specifically to the calcium and calmodulin (CaM)-dependent serine/threonine phosphatase, calcineurin (CaN), inhibiting its phosphatase activity in vitro (1). FKBP12, without drug present, will not bind to CaN. Furthermore, two members of the human cyclophilin family, cyclophilins A and C, will also bind to and inhibit CaN only when they are complexed with CsA. These results implicate CaN as a common downstream target of both FK-506 and CsA and help to explain the parallel effects of the two drugs. That one of the steps 21753 involved in the early T-cell activation events is a calciumregulated dephosphorylation event is also consistent with these observations. CaNs have been implicated in controlling signal transduction pathways emanating from the second messenger, calcium (18), but prior to the discovery that immunophilin-drug complexes bind to CaN, it had not been possible to assign a particular function to the CaNs in any one signal transduction process. One hypothesis explaining CaN's role in calcium-dependent T-cell activation is that it directly or indirectly (via a phosphatase cascade) dephosphorylates a transcription factor that is necessary for early lymphokine gene expression (19).
An important objective is to identify and characterize all of the FKBPs that are relevant t o the immunosuppressive and toxic effects of FK-506 and rapamycin in lymphoid and nonlymphoid cells. In crude extracts prepared both from calf brain and from the T-lymphocytic line, JURKAT, we discovered two FK-506 binding activities that both migrate with apparent molecular masses of about 110 kDa on gel filtration columns. Subsequent to the report that FKBP12 binds to CaN in the presence of FK-506 (l), we performed reconstitution experiments using purified proteins and found that the FKBP12 .FK-506. CaN complex comigrates with the 110 kDa activities we observe in calf brain extracts. X n this report we confirm and extend the characterization of the FKBPl2. FK-506. CaN complex. We also characterize the 110-kDa FK-506 binding activity in JURKAT cells and show that it is a protein similar in sequence to the hsp56 component of rabbit progestin receptor complexes (20).

MATERIALS AND METHODS
Purification of FKBP51 from JURKAT Cell.-A cytosolic extract prepared from 3 X 10" cells was prepared as described (5), dialyzed in buffer containing 10 mM potassium phosphate (pH 7.2), 5 mM pmercaptoethanol, 1 mM EDTA, and 0.5 mM PMSF, and applied to a 5 X 20 cm Affi-Gel Blue column equilibrated with the same buffer. After washing the column with 2 column volumes of equilibration buffer, the column was washed with a linear gradient (total volume, 2 liters) of 100-500 mM potassium phosphate (in the equilibration buffer). Fractions (12.5 ml each) were collected at a flow rate of 125 ml/h. FK-506 binding activity eluted between 200 and 325 mM potassium phosphate. Active fractions were combined (total volume 124 ml) and applied to a column (5 X 15 cm) of DEAE-Sepharose equilibrated with buffer containing 25 mM HEPES (pH 7.51, 5 mM 0-mercaptoethanol, 0.5 mM PMSF, and 50 mM KC1. The column was washed with the equilibration buffer containing 100 mM KC1 and then with equilibration buffer containing 250 mM KCl, and fractions (6.5 ml each) were collected at a flow rate of 125 ml/h. Active fractions were concentrated by ultrafiltration to 8.6 ml and dialyzed against 4 liters of 25 mM Bis-Tris (pH 6.3). The concentrated material was applied to a MonoP HR 5/20 fast protein liquid chromatography chromatofocusing column equilibrated in the dialysis buffer. After washing the column with 10 column volumes of the dialysis buffer, the column was developed with 80 ml of polybuffer 74 (pH 4.0), and fractions (1 ml) were collected at a flow rate of 1 ml/min. The active binding fractions, which eluted between pH 4.6 and 4.1, were combined and dialyzed overnight against buffer containing 25 mM HEPES (pH 7.5), 50 mM KC1, and 5 mM 0-mercaptoethanol. The dialyzed material was dialyzed again for 2 h against buffer containing 50 mM sodium phosphate (pH 7.0), 0.6 M (NH&SO4, and 5 mM pmercaptoethanol and applied to a phenyl-Superose HR 5/5 fast protein liquid chromatography column equilibrated in the same buffer. The column was developed with a linear gradient (total volume, 5 ml) of 0.6-0.0 M (NH4)~SO4 in 50 mM sodium phosphate (pH 7.0) and 5 mM p-mercaptoethanol. Fractions (0.5 ml) were collected at a flow rate of 0.5 ml/min. Active fractions were combined, dialyzed in 4 liters of the DEAE-Sepharose equilibration buffer, and concentrated to 1.6 ml by ultrafiltration. Protein was assayed using the Bio-Rad protein assay (Bio-Rad). Purified FKBP51 was stored at -80 "C.
To obtain peptides suitable for amino acid sequence analysis, FKBP51 was treated with trypsin in a 50-pl reaction containing 50 mM ammonium bicarbonate, pH 9.0, and a trypsin/FKBP51 ratio of 1:lOO (w/w). FKBP51 was digested a t 37 "C for 16 h after which time the digest was neutralized with 5 pl of 10% trifluoroacetic acid. Tryptic peptides were separated by reverse-phase HPLC using a 1.0 X 100-mm C18 column (ABI, Foster City, CA) and a buffer gradient of 2% &0/0.060% trifluoroacetic acid to 75% acetonitrile/HzO (1090) containing 0.055% trifluoroacetic acid. Peptides were sequenced directly from PVDF membranes using an AB1 477 protein sequencer (ABI).
Peptidyl-prolyl cis-trans Isomerase Assay-Peptidyl-prolyl isomerization was assayed as previously described (5) with the following changes: the peptide substrate used was N-succinyl-Ala-Leu-Pro-Phe-p-nitroaniiide (BACHEM, Switzerland) at a final concentration of 73 p~ and chymotrypsin (Sigma) was present in the assay at a concentration of 66 p~. The release ofp-nitroanilide by chymotrypsin was quantitated by measuring the increase in absorbance at 405 nm using a Beckman DU68 spectrophotometer. After an initial rapid increase in absorbance due to hydrolysis of the trans peptide, the slow secondary increase in absorbance, which reflects the conversion of cis peptide to trans peptide, was measured at 3-s intervals. The data were fit to a simple first-order rate law and the first-order rate constant, k (s-'), calculated. K, values for the inhibition of FKBP51 PPIase activity by FK-506 and rapamycin were determined from the dependence of the first-order rate constant, 12, on inhibitor concentration using a computer program written by Nancy Thornberry of the Department of Enzymology, Merck Research Laboratories (28).
Calcineurin Binding Assay-Incubations (total volume, 500 pl) were performed for 15 min at 30 "C and contained various combinations of the following components: 1.5 pg of purified bovine calcineurin, 1 pg of recombinant human FKBPIS, 1 pg of bovine calmodulin (Sigma), and 100 nM [3H]dihydroFK-506. The incubation buffer contained 20 mM Tris (pH 7.5), 100 m M NaCl, 6 mM MgCl,, 0.1 mM CaC12, 0.1 mg/ml BSA, and 0.5 mM dithiothreitol. The incubation reaction was chromatographed on a Bio-Si1 SEC 250 HPLC column (Bio-Rad) at a flow rate of 1 ml/min, and fractions (0.4 ml) were assayed for radioactivity. The chromatography buffer was the same as the incubation buffer except that BSA was omitted. In some incubations calmidazolium (25 pglml, Calbiochem) was added 10 min prior to the addition of [3H]dihydroFK-506 in order to inhibit calmodulin binding.
Isolation of the cDNA Encoding Human FKBP12"DNA probes and primers were synthesized on a Milligen Cyclone model 8400 DNA synthesizer. Based upon amino acid sequencing of human FKBPl2 purified from JURKAT cells (5), one degenerate, deoxyinosine-containing 38-mer oligonucleotide probe was synthesized in order to isolate the cDNA encoding human FKBPl2 from a JURKAT cDNA library cloned in the vector hgtl0. Deoxyinosines were used to probe T, C, A, G ambiguities as well as some T, C ambiguities, while thymidines were used to probe certain A, G ambiguities according to the recommendations outlined by Martin and Castro (23). The oligonucleotide synthesized and the corresponding human FKBP amino acid sequence from which it was derived is shown below.

HIS TYR THR GLY MET LEU GLU ASP GLY LYS LYS PHE ASP CAT TAT ACI GGI ATG TTI GAI GAT GGI AAI AAI
TTT GA c c C The JURKAT cDNA library was plated on Escherichia coli strain C600 hflA. To isolate the human FKBPl2 cDNA, ten 150-mm plates containing approximately 30,000 plaques/plate of the JURKAT cDNA library were screened. Plate replicas were made on Nytran membranes (Schleicher & Schuell). Denaturation, renaturation, and baking of the phage DNA to the filters were performed according to the manufacturer's instructions. The filters were prehybridized and then hybridized to the 32P-labeled oligonucleotide probe according to the methods described by Itoh et al. (24). Briefly, the filters were prehybridized for 4 h at 45 "C in a solution containing 5 X SSC, 10 X Denhardt's, 50 mM sodium phosphate, and 0.1% SDS. The 32Plabeled oligonucleotide probe (1 ng/ml, 4 x IO7 cpm/pmol) was added to the prehybridization solution and the filters hybridized at 45 "C for 16 h. The filters were washed four times at 20 min/wash at room temperature in a washing solution containing 5 X SSC, 0.1% SDS, and 50 mM sodium phosphate. This was followed by an additional wash at 45 "C for 1 min. The wet filters were wrapped in Saran Wrap and exposed to x-ray film (XAR-5, Eastman Kodak) for 36 h at -80 "C using intensifying screens. Positive phage were plaque-purified, and the phage insert was subcloned into the EcoRI site of pUC19. Sequencing of both DNA strands by the dideoxy method was performed directly from denatured plasmid miniprep DNA. The cDNA was partial in that the open reading frame was missing the nucleotides encoding the NHz-terminal 11 amino acids. The nucleotide sequence of this partial cDNA clone was in agreement with the sequence of the human FKBP (from nucleotides 34 through 1454) that was published The resuspended cells were frozen by dripping into liquid nitrogen, stored at -80 "C, and FKBP released by subsequent thawing at room temperature. Unbroken cells and debris were removed by centrifugation, the supernatant collected, and protamine sulfate was added to a final concentration of 0.04%. After centrifugation (15,000 X g, 30 min), the supernatant was concentrated by using an Amicon filtration device and a YM-5 membrane. FKBPl2 was purified from the concentrated protein by chromatography on a 600 X 21.5-mm Bio-Si1 TSK125 HPLC column. FKBPl2 was eluted at 5 ml/min in buffer containing 20 mM sodium phosphate (pH 6.8), 50 mM NaZSO4, 5 mM (3-mercaptoethanol, 1 mM EDTA, and 0.5 mM PMSF. Yields varied between 50 and 200 mg of FKBP12/liter of bacterial culture. The purified protein appeared homogeneous by Coomassie staining of SDS-polyacrylamide gels and had nearly identical FK-506 binding and PPIase activities compared to FKBPl2 isolated from JURKAT cells. Amino acid sequencing of the recombinant protein demonstrated that the first 38 amino acids matched those encoded by the cDNA except that the NHz-terminal methionine was missing. Expression and Purification of Radiolabeled FKBP12"An overnight culture of the FKBPl2-producing bacterial strain described above was used to innoculate 50 ml of M9 media containing 50 pg/ ml ampicillin such that the ODsm was 0. concentration) were added, and the culture was incubated overnight at 37 "C. The cells were lysed and FKBPl2 purified to radiochemical homogeneity as described above. The specific activity of the radiolabeled protein was 1.3 Ci/mol. Construction of the F36Y FKBPl2 Mutant-The TTT (Phe) codon at position 36 in human FKBP12 was changed to TAC (Tyr) by a polymerase chain reaction with overlapping mutant primers using a procedure described by Ho et al. (27). Oligomers flanking the ATG and TGA codons were used in a subsequent polymerase chain reaction to incorporate the F36Y alteration into the entire open reading frame. The resulting fragment was digested with NcoI and BamHI and ligated to NcoIand BamHI-digested pET3d vector DNA. After transformation into BL21(DE3) cells, the 327-base pair DNA fragment encoding the mutant FKBPl2 of one recombinant was sequenced to verify the presence of the mutation. The F36Y isolate used in these studies has an insertion of 3 residues (TAC) at the NcoI site, and as a result cannot be cut with this enzyme. The presence of these additional residues did not significantly alter expression levels of this protein in E. coli compared with the wild-type recombinant FKBP12 gene that has an intact NcoI site.
Cells expressing FKBPl2 F36Y were grown in LB media containing 100 pg/ml ampicillin supplemented with 0.4% glucose until the OD, reached 1.0-1.5 OD units. Then the cells were induced with 0.1 mM (final concentration) isopropyl-1-thio-(3-D-galactopyranoside and grown for an additional 15 h. The induced cells were subjected to centrifugation for 10 min at 5000 X g, resuspended in 1/50 of the original culture volume with 20 mM Tris (pH 7.4) and stored at -70 "C until purified as described above.

Purification of a Novel FK-506
Binding Activity from JUR-KAT Cells-We prepared extracts from JURKAT cells and from a non-lymphoid tissue, calf brain, incubated the extracts with [3H]dihydroFK-506 and fractionated the incubation mixture by HPLC gel filtration. In addition to the major peak of radioactivity associated with FKBP12 ( Fig. 1, peak B), we observed in both extracts, a smaller peak of activity corresponding to a protein with a relative mass of 110 kDa (Fig. 1,  peak A ) . Attempts to purify the 110-kDa FK-506 binding activity from extracts of calf brain proved unsuccessful. In contrast, the 110 kDa binding activity from JURKAT cells was more amenable to purification. Table I (Table I), respectively. Panel B, reverse-phase HPLC of step V-purified FKBP51.50 pl(11 pg of protein) of step V-purified FKBP51 was applied to a 1.0 X lOO-mm, AB1 C4 reverse-phase column. FKBP51 was eluted using a buffer gradient of 15% H20, 0.6% trifluoroacetic acid to 75% 9010 acetonitrile, HZ0 containing 0.55% trifluoroacetic acid. Chromatography was performed using an AB1 130 separation system. proteins purified to date, this new FKBP does not bind CsA and the FK-506 binding is competitively inhibited by rapamycin (data not shown). We also found that binding activity copurified with PPIase activity that was inhibitable by both FK-506 and rapamycin, another property shared with several other FK-506-binding proteins. However, the PPIase activity could not be determined until after the DEAE step due to the presence of cyclophilin A, also a PPIase, in earlier purification steps. Furthermore, the standard LH-20 assay used to measure CsA and FK-506 binding to cyclophilin and to FKBP12, respectively, could not be used for assaying the 110 kDa activity because the activity was completely retained by the LH-20 resin. Therefore, a binding assay using Bio-Gel P-6DG was devised. Because crude extracts contained large quantities of FKBP12, assays of the 110 kDa activity in crude extracts were performed by HPLC gel filtration and, in subsequent purification steps, by the P-6DG assay. When the purified material was subjected to electrophoresis using an SDS-polyacrylamide gel, one major protein band with a relative molecular mass of 51 kDa was evident ( Fig. 2A). We therefore refer to this protein as FKBP51. Purity of the novel protein was confirmed by HPLC reverse-phase chromatography on a C18 column (Fig. 2B). Because 51 kDa is approximately half that of the value obtained from HPLC gel filtration of the same protein bound to FK-506 (Fig. 3), it suggests that FKBP51 may dimerize in its native state or that FK-506 promotes dimerization. In the absence of FK-506, we failed to recover functional protein (as assayed in both the Bio-Gel P6-DG and PPIase assays) after fractionation by HPLC gel filtration and, therefore, could not distinguish between FK-506-induced dimerization or a normal homodimeric structure for FKBP51.
Purified FKBP51 has a specific activity of 426-510 ng of dihydroFK-506 bound/mg of protein (Table I). This calculates to a 473-fold purification and would seem to suggest that FKBP51 is as abundant as FKBPl2 which it clearly is not (Fig. 1). Furthermore, assuming a molecular mass of 51 kDa for the active FK-506-binding subunit and a 1:l ratio of protein to ligand, one would expect a theoretical specific activity of 15,700 ng of [3H]dihydroFK-506 bound/mg of FKBP51. These discrepancies may be due to an underestimation of FK-506 binding using Bio-Gel P-6DG due to the lower affinity of FK-506 for FKBP51 as compared to FKBP12 (see below). In contrast, when specific activities are measured using the PPIase assay, FKBP51 appears to be fully active when compared to both FKBPl2 and FKBP25. We have determined that kaJKm for the PPIase activity of FKBP51 is 0.60 X lo6 M". s-', which is comparable to the PPIase activities of both FKBP12 (kat/Km = 0.64 x lo6 M".s") (281, and FK-506 and rapamycin are potent inhibitors of FKBP51 PPIase activity (Fig. 4) allowing kinetic data to be used to assess the affinity of these ligands for FKBP51. We have  (28)). Peptide sequencing of the protein revealed a single NH2 terminus, a further indication that FKBP51 was homogenous (Table 11). Shown below six of the seven peptide sequences derived from FKBP51 is an alignment with related sequences  (Table I)) along with increasing concentrations of FK-506 (c".) or rapamycin (W). After incubation for 10 min at 25 "C, remaining assay components were added and PPIase activity determined as described under "Materials and Methods.'' The data shown were fitted using a four parameter logistic function (Sigmaplot, Jandel Scientific, Corte Madera, CA).

Alignment between amino acid sequences derived from sequencing human FKBP51 and the corresponding sequences from the open reading frame of cloned rabbit uterine hsp56
The first alignment (A) is between the NH2-terminal sequence of FKBP51 with an internal sequence of rabbit hsp56. All other alignments are between sequences of tryptic peptides from human FKB56 with internal sequences of the cloned rabbit gene. FKBP51 peptide fragments were generated by trypsin treatment as described under "Materials and Methods." found in the open reading frame predicted by the DNA sequence of the cDNA encoding rabbit uterine hsp56 (20). Hsp56 is a heat shock protein found in nontransformed steroid receptor complexes and contains a domain having 55% identity to FKBP12. It is not known if rabbit uterine hsp56 binds FK-506 or if it has PPIase activity. The similarity of six of the seven FKBP51 peptide sequences to sequences in rabbit hsp56 as well as the similarity in size of the two proteins suggests that FKBP51 is either the human homolog of rabbit uterine hsp56 or another closely related protein.
FK-506. FKBP Complex Formation with Calcineurin-As specified previously, we were unable to purify the 110 kDa activity from bovine brain even though an FK-506 binding activity, similar to that observed in JURKAT extracts, was clearly present in crude extracts. Recently, it has been shown that FKBP12.FK-506 complexes bind to the Ca2+ and CaMdependent protein phosphatase, CaN (1). In view of the fact that CaN is extraordinarily abundant in brain, we reasoned that the 110-kDa [3H]dihydroFK-506 complex observed in brain extracts might be such a complex. In order to test this hypothesis, we first determined the requirements for complex formation between FKBPl2 and CaN.
We find that incubation of bovine CaN, bovine CaM, recombinant human FKBP12, [3H]dihydroFK-506, Ca2+, and M e results in the formation of a protein complex which has an apparent mass of about 110 kDa as determined by HPLC gel filtration (Fig. 5A). (The larger peak of radioactivity migrating with an apparent mass of 12 kDa is due to free FKBP12-[3H]dihydroFK-506 complexes.) In the absence of CaN, no complex is detected and all of the [3H]dihydroFK-506 is bound to FKBPl2 alone (Fig. 5A). Furthermore, we find that in the absence of CaM, the 110-kDa complex is virtually undetectable (Fig. 5A). A residual amount of complex is occasionally observed in the absence of CaM but may be the result of CaM contamination of CaN. CaM binds with very high affinity to CaN (29) and is difficult to remove completely. The potent CaM antagonist, calmidazolium (30), was used to verify our observation that formation of the FKBP12. FK-506. CaN complex is dependent upon CaM. Addition of calmidazolium to an incubation mixture containing all of the components required for complex formation prevented formation of the complex (Fig. 5B). Calmidazolium had no inhibitory effect on FK-506 binding by FKBP12. These results differ from those reported by Liu et al. (1) who did not observe a strict CaM requirement for FKBP12.CaN complex formation. Using labeled FKBP12, we also confirmed that FKBP12-calcineurin complex formation is ligand-specific. Rapamycin, which binds to FKBP12 with an affinity similar to that of FK-506, but which is inactive as an inhibitor of interleukin-2 synthesis (2), fails to promote complex formation with CaN (Fig. 5C) Nature of the High Molecular Weight FK-506 Complex in Bovine Bruin-Our finding that complex formation between FKBPl2. FK-506 and CaN is dependent upon CaM provided us with the means to directly determine the makeup of the high molecular weight FK-506 complex observed in crude bovine brain extracts. Addition of calmidazolium to brain extracts almost totally abolishes formation of the 110-kDa peak (Fig. 6A). Addition of calmidazolium to the same amount of protein from JURKAT extracts only partially reduces the 110-kDa peak (Fig. 6B). These results suggest that virtually all of the 110 kDa binding activity in brain is due to the FKBP12. FK-506. CaN . CaM complex while, in JURKAT extracts, a majority of the 110-kDa FK-506 binding activity is due to FKBP51. Interestingly, when FKBP51 was substituted for FKBPl2 in either the CaN binding assay or the CaN phosphatase assay, no complex formation or inhibition of phosphatase activity was observed (data not shown). These observations suggest that CaN binding, in the presence of FK-506, is not a general property of FK-506-binding proteins.
PPIase Activity of FKBPl2 Is Not Required for CaN Complex Formation-The conformation of FK-506 when bound to FKBP12 is known to be dramatically different from the conformations existing in solution. The amide bond of FK-506 exists in both cis and trans conformations in solution while only the trans form is associated with FKBP12. In addition, the pyranose ring is on the outside of the macrocycle when FK-506 is in solution but is on the inside of the macrocycle when it is bound to FKBPl2 (31)(32)(33). It is not known if the enzymatic activity of FKBPl2 is responsible for the observed alterations in the conformation of FK-506 upon binding. By site-directed mutagenesis, we have constructed a number of single amino acid changes in residues that are most conserved throughout phylogeny in FKBP12. One mutant, termed F36Y (a Tyr for Phe substitution at position 36), has 85-90% of the FK-506 binding activity relative to wild-type FKBPl2 (data not shown) but has less than 0.1% of the wildtype's PPIase activity (Fig. 7A). The requirements for functional PPIase activity in the FK-506-dependent inhibition of CaN phosphatase activity were examined using the PPIase- deficient human FKBPlP mutant. Increasing amounts of wild-type recombinant FKBPl2 or PPIase-deficient FKBPlP were added to a cocktail containing all of the components necessary for complex formation and for measurement of CaN phosphatase activity (Fig. 7 B ) . The wild-type and the PPIasedeficient FKBP12 proteins exhibited nearly identical IC60 values in the CaN phosphatase assay. These results indicate that the PPIase activity of FKBPl2 is not required for the inhibition of CaN phosphatase activity by FKBP12. FK-506 complexes.

DISCUSSION
Using gel filtration as a method to identify novel FK-506 binding activities and/or high molecular weight complexes, we observed peaks of activity in crude extracts prepared from lymphoid and non-lymphoid sources that were well-separated from the previously identified peak corresponding to FKBP12. Although both of the new activities migrated with identical apparent molecular masses of approximately 110 kDa, further characterization of the activities revealed important differences. We were unable to further purify the activity from calf brain, while purification of the activity from JURKAT cells yielded a homogenous protein with an apparent molecular mass on SDS-polyacrylamide gel electrophoresis gels of 51 kDa. This suggests that the new protein, which we have labeled FKBP51, either forms a dimer in its native state or, due to an unusual tertiary shape, migrates anomalously on sizing columns.
Peptide sequence analysis of FKBP51 suggests that it will be similar, if not identical, to hsp56, a heat shock protein found complexed with hsp70 and hsp90 in nontransformed progestin, glucocorticoid, androgen, and estrogen receptor complexes isolated from the human lymphocytic line, IM-9. Purified human hsp56 has been subjected to Edman degradation and the last six amino acids of the 20 obtained from NHp-terminal sequencing (34) match the first six amino acids that we obtained from NHz-terminal sequencing of FKBP51. Although the cDNA encoding human hsp56 has not been isolated, the cDNA encoding rabbit hsp56 has been cloned and sequenced. Five of six sequenced tryptic peptides from FKBP51 are similar to amino acid sequences found in the translation product of the rabbit cDNA sequence. The open reading frame of rabbit hsp56 contains regions highly homologous to known FK-506-binding proteins indicating that it is a member of the FKBP family. However, experiments demonstrating that human or rabbit hsp56 bind to FK-506 in solution or that they have PPIase activity have not been reported. Recently, a 60-kDa human protein purified by affinity to immobilized FK-506 has been described (35). It is not known if this 60-kDa protein binds FK-506 in solution or if it has PPIase activity. NHz-terminal sequencing of this protein revealed that it has the same NHp-terminal sequence as human hsp56 (34). Presently, there is not enough sequence information from FKBP51, human hsp56, or from the 60-kDa protein isolated on the FK-506 affinity column to determine whether or not all three proteins are identical. Although it is possible that the difference we observe at the NHz terminus is due to proteolysis of FKBP51 during purification, isolation and sequencing of the corresponding cDNA must be performed in order to firmly establish the degree of relatedness between FKBP51 and hsp56.
If FKBP51 is, in fact, the human homolog of rabbit hsp56, it will be interesting to determine what effect FK-506 binding to FKBP51 has upon integrity of protein-protein interactions in glucocorticoid receptor complexes or if FK-506 can compete with glucocorticoids for binding. The latter result might suggest that mammalian cells have an endogenous small molecule ligand that is mimicked by FK-506. Like FKBP12, we have found that FKBP5l isomerizes peptidyl-prolyl bonds in peptide substrates; however, no cellular substrates for FKBPl2 or FKBP51 have been identified to date. It is not known if hsp56 has PPIase activity but if FKBP51 and hsp56 are identical, then hsp70, hsp90, or the glucocorticoid receptor itself would be obvious candidates for an endogenous substrate. One could imagine that subsequent to steroid binding by the glucocorticoid receptor, FKBP51-catalyzed isomerization about a particular peptidyl-prolyl bond helps promote disassociation of the proteins contained in the complex.
Two of our results demonstrated that the 110 kDa binding activity discovered in calf brain was different from the activity we purified from JURKAT cells. First, the activity in the brain extract was labile and could not be purified. Once the cDNA encoding FKBP51 is obtained, multiple tissue Northern blots will be performed in order to determine if the low level of FKBP51 activity we have observed in brain is due to weak expression of the corresponding mRNA in that tissue. Second, subsequent to the report that FKBP12. FK-506 binds to CaN (l), we added the CaM inhibitor calmidazolium to crude extracts prepared from both sources and were unable to detect any 110 kDa binding activity in the brain extract. In contrast, activity was retained in the crude extract prepared from JURKAT cells when calmidazolium was added. We have confirmed the observation that FKBP12.FK-506 binds specifically to the phosphoprotein phosphatase, CaN, inhibiting its phosphatase activity and the observation that FKBP12rapamycin does not complex with CaN. The latter observation correlates well with the experimental evidence that, in uiuo, FK-506 and rapamycin are potent reciprocal antagonists of one another (36, 37). Another potent competitive antagonist of FK-506, L-685,818, also binds with equivalent affinity to FKBPl2 but fails to promote CaN complex formation: In contrast to previous findings (l), we observe a dependence upon CaM for the binding of FKBPl2 .FK-506 to CaN. In a reconstituted system using purified components, almost no complex was detected unless CaM was present in the incubation mixture. Further corroboration of the CaM requirement was made using the CaM antagonist, calmidazolium. Addition of calmidazolium either to the complete, reconstituted system or to the crude brain extract completely abolished FKBP12 .FK-506. CaN complex formation. The residual complex formation observed in our reconstituted system when CaM was omitted from the reaction mixture might be due to the presence of small amounts of CaM in the CaN preparation. CaN has been identified as the major binding protein for the calcium receptor, CaM (39). CaM binding may evoke a conformational change in CaN which neutralizes a COOH-terminal CaN inhibitory domain (40). Our results are compatible with a model which proposes that in uiuo, native CaN (in the absence of CaM) exists in a conformational state unsuitable for FKBP12 .FK-506 binding. Interaction with CaM would reveal a site for FKBP12 FK-506 binding. Previous reports of FKBP12 .FK-506 interaction in the absence of CaM (1) might be explained by the investigators use of partially purified CaN which may be contaminated with CaM or by the presence of proteolytically fragmented CaN. The reported interaction of FKBP12.FK-506 complex with the 43-kDa CaN A fragment lacking the CaM binding and autoinhibitory domains (41) would be consistent with our model since the catalytic and FKBPl2.FK-506 binding domains would be expected to be exposed. Further clarification of whether there is a strict CaM requirement for FKBPl2 .FK-506. CaN complex formation or whether CaM enhances complex formation might be made using recombinant bacterially produced CaN once it becomes available.
CaN has a heterogeneous tissue distribution and is extremely abundant in the brain, comprising nearly 1% of total brain protein (42). CaN is much less abundant in peripheral tissues although it has also been identified as the major CaM binding protein in T-lymphocytes (43). The FKBP12.FK-506 interaction with CaN was originally discovered in extracts prepared from brain and CaN's tissue distribution explains why we observed so much of the complex in the extracts prepared from brain. The FKBPl2. FK-506. CaN interaction has not yet been reported using extracts prepared from Tlymphocytes although the partial calmidazolium sensitivity of our 110-kDa peak in JURKAT extracts is strongly suggestive that the same interaction does, in fact, occur in drugtreated T-cells.
We have identified an FKBP12 mutant, F36Y, that has near-normal affinity for FK-506 but which has less than 1/ 1000th the PPIase activity of FKBP12. Furthermore, the wild-type and F36Y mutant proteins are equivalent in their abilities to inhibit CaN phosphatase activity in the presence of FK-506. Our observations indicate that the PPIase activity of FKBPl2 is unrelated to its affinity for FK-506 and that the conformation of FK-506 required for interaction with CaN is not induced by the PPIase activity of FKBP12. These results also correlate well with the observation that inhibition of PPIase activity is not germane to the immunosuppressive activity of FK-506 (36, 37). CsA, like FK-506, is markedly hydrophobic and, again like FK-506, has strikingly different conformations when free in organic solvents and when bound to its cognate immunophilin (44,45). However, recent results strongly suggest that cyclophilin does not induce the bound conformation of CsA but binds a conformation preexisting in aqueous solution (38). Collectively, all of these results suggest that FK-506 and CsA are not substrates for the PPIase activities of their respective binding proteins but that the bound conformations are the preferred ones in the hydrophobic binding pockets of FKBPl2 and cyclophilin, respectively.