Inhibition of the Calcineurin-NFAT Interaction by Small Organic Molecules Reflects Binding at an Allosteric Site*

Transcriptional signaling from the Ca2+-calmodulin-activated phosphatase calcineurin to its substrate NFAT (nuclear factor of activated T cells, also termed NFATc) is critically dependent on a protein-protein docking interaction between calcineurin and the PXIXIT motif in NFAT. Several inhibitors of NFAT-calcineurin association (INCA compounds) prevent binding of NFAT or the peptide ligand PVIVIT to calcineurin. Here we show that the binding site on calcineurin for INCA1, INCA2, and INCA6 is centered on cysteine 266 of calcineurin Aα and does not coincide with the core PXIXIT-binding site. Although ample evidence indicates that INCA1 and INCA2 react covalently with cysteine 266, covalent derivatization alone is not sufficient for maximal inhibition of the calcineurin-PVIVIT interaction, because the maleimide INCA12 reacts with the same site and produces only very modest inhibition. Thus, inhibition arises through an allosteric change affecting the PXIXIT docking site, which may be assisted by covalent binding but depends on other specific features of the ligand. The spatial arrangement of the binding sites for PVIVIT and INCA makes it probable that the change in conformation involves the β11-β12 loop of calcineurin. The finding that an allosteric site controls NFAT binding opens new alternatives for inhibition of calcineurin-NFAT signaling.

Transcriptional signaling from the Ca 2؉ -calmodulin-activated phosphatase calcineurin to its substrate NFAT (nuclear factor of activated T cells, also termed NFATc) is critically dependent on a protein-protein docking interaction between calcineurin and the PXIXIT motif in NFAT. Several inhibitors of NFAT-calcineurin association (INCA compounds) prevent binding of NFAT or the peptide ligand PVIVIT to calcineurin. Here we show that the binding site on calcineurin for INCA1, INCA2, and INCA6 is centered on cysteine 266 of calcineurin A␣ and does not coincide with the core PXIXIT-binding site. Although ample evidence indicates that INCA1 and INCA2 react covalently with cysteine 266, covalent derivatization alone is not sufficient for maximal inhibition of the calcineurin-PVIVIT interaction, because the maleimide INCA12 reacts with the same site and produces only very modest inhibition. Thus, inhibition arises through an allosteric change affecting the PXIXIT docking site, which may be assisted by covalent binding but depends on other specific features of the ligand. The spatial arrangement of the binding sites for PVIVIT and INCA makes it probable that the change in conformation involves the ␤11-␤12 loop of calcineurin. The finding that an allosteric site controls NFAT binding opens new alternatives for inhibition of calcineurin-NFAT signaling.
Ca 2ϩ -calcineurin signaling serves vital purposes in mammalian cells, among them the provision of a direct link between the mobilization of cytoplasmic Ca 2ϩ and gene expression (1,2). In the physiological activation of T cells, calcineurin acts through NFAT 4 family transcription factors and other transcriptional effectors (3)(4)(5)(6). Levels of calcineurin in T cells are limiting, and upward or downward modulation of calcineurin enzymatic activity is directly reflected in increased or decreased transcription from cytokine gene promoters (7)(8)(9)(10)(11)(12)(13)(14). The immunosuppressive drugs cyclosporin A (CsA) and FK506 owe their clinical effectiveness to the ability of CsA-cyclophilin or FK506-FKBP12 complexes to inhibit calcineurin in cells of the immune system (15,16). However, the beneficial actions of CsA and FK506 are counterbalanced by serious toxicities attributed at least in part to their interference with calcineurin signaling in other cells and tissues (17,18).
The centrality of calcineurin and the calcineurin-NFAT pathway to immune responses has suggested that interrupting signaling at any of several points could provide a therapeutic alternative to current immunosuppressive drugs. Promising target points are the Ca 2ϩ signal that activates calcineurin (19,20), the protein-protein interaction of calcineurin and NFAT (21)(22)(23)(24)(25), and the cooperative binding of NFAT and AP1 on DNA (26 -30). Each strategy has the potential to be more selective and, hence, less toxic than treatment with CsA or FK506, but none has yet advanced to the stage of small nonpeptide inhibitors that can be tested in vivo.
There is considerable evidence that targeting the calcineurin-NFAT protein-protein interaction will produce a selective inhibition. We have demonstrated that the calcineurin-NFAT interaction is based on recognition of a PXIXIT motif in NFAT and that this recognition is essential for efficient signaling (23). Peptides that compete for binding at the PXIXIT recognition site both inhibit the calcineurin-NFAT interaction in vitro and selectively inhibit calcineurin-NFAT signaling in cells (23,24,31). A recently published study (32) shows that a competitor peptide modified to promote its uptake into cells can prevent heterologous graft rejection in mice. Further, high throughput screening of a library of organic compounds has led to identification of nonpeptide inhibitors of NFAT-calcineurin association (INCA compounds) (25). These compounds interfere with calcineurin-NFAT signaling in cells, motivating a continued search for inhibitors that have higher affinity and reduced nonspecific toxicity and that are suitable for in vivo administration to animals.
The further development of inhibitors can be guided by structural information about the sites of protein-protein interaction and inhibitor binding. To this end, we have determined the structure of the NFAT docking site on calcineurin, in which the PXIXIT recognition peptide of NFAT binds in an extended configuration and each conserved residue of the peptide directly contacts calcineurin (33). Here we extend the structural studies by identifying a distinct binding site for INCA compounds at a cysteine residue adjacent to the PXIXIT peptide docking site. It is covalent binding or tight noncovalent binding of the bulky INCA compounds at this second site that allosterically inhibits recognition of PXIXIT peptide and NFAT. Our findings serve to identify the approaches that are most likely to be fruitful in developing improved inhibitors. In addition, the findings suggest that calcineurin-substrate recognition, like calcineurin catalytic activity (34 -37), may be modulated physiologically by redox reactions.

EXPERIMENTAL PROCEDURES
cDNA Constructs, Protein Expression, and Synthetic Peptides-The expression construct for GST-calcineurin-(2-347), encoding glutathione S-transferase and the catalytic domain of human calcineurin A␣ with the substitutions Y341S, L343A, and M347D, has been described (24). Constructs encoding mutant proteins with individual Cys 3 Ala or Cys 3 Val replacements or with the combined Cys 3 Ala/Cys-266 (all surface Cys residues changed to Ala except for Cys-266) and Cys 3 Ala/C266V (all surface Cys residues changed to Ala except at position 266, where Cys is changed to Val) replacements were made by PCR mutagenesis with appropriate primers and subcloning. In each case, the sequence of the cDNA insert was verified.
GST-calcineurin fusion proteins were expressed in Escherichia coli strain BL21-CodonPlus-RP (Stratagene) by overnight growth at 18°C after the addition of 1 mM isopropyl-␤-D-1-thiogalactopyranoside. Calcineurin was purified from the bacterial lysate by affinity chromatography on glutathione-Sepharose (Amersham Biosciences) and cleavage in 50 mM Tris⅐HCl, pH 7.0, 150 mM NaCl, 1 mM EDTA, and 1 mM DTT with PreScission protease (Amersham Biosciences). Purified calcineurin-(2-347) was concentrated by Centricon-30 filtration, aliquoted, flash-frozen, and stored at Ϫ80°C. In most cases, the concentration of DTT in the buffer was reduced to 0.1 mM during concentration of the protein. This step and the further substantial dilution into fluorescence polarization assays minimized any interference of residual DTT with the INCA compounds. In some cases, protein was concentrated in buffer free of DTT immediately before freezing.
Inhibitors-INCA1, INCA2, INCA6, and INCA12 ( Fig. 1) were obtained from ChemBridge. Inhibitor stocks were prepared at 10 mM in anhydrous dimethyl sulfoxide (Me 2 SO) and stored desiccated at Ϫ20°C. Corresponding aliquots of anhydrous Me 2 SO were stored desiccated at Ϫ20°C for addition to control incubations. Me 2 SO at concentrations up to 1% had no effect on the parameters studied.
Fluorescence Measurements-Fluorescence measurements were made on 10-l samples in a black 384-well plate (Molecular Devices) using the fluorescein filter set (excitation at 485 nm, emission at 530 nm) in an Analyst plate reader (Molecular Devices). Each well contained 100 mM NaCl, 2 mM magnesium acetate, 20 mM HEPES, pH 7.4, 0.1% (w/v) bovine IgG, calcineurin, 100 nM fluorescent PVIVIT, and other additions as specified. Calcineurin was omitted for measurements of the fluorescence emitted by unbound peptide. In competitive binding assays, calcineurin was typically present at 1 or 1.5 M; in direct binding titrations, calcineurin was used at 0.15-10 M. Except in time course experiments, adequate time was allowed for the signal to reach a stable value as verified by repeated readings of the same samples. Pretreatment with iodoacetamide (IAM), N-ethylmaleimide (NEM), or INCA12 was for 30 min to 2 h.
Data from direct titrations with calcineurin have been fitted to a model for equilibrium binding to a single class of sites (38). Competition and time course data for INCA compounds have been plotted without attempting to fit a theoretical curve, because the data are not sufficient to specify parameters for covalent binding, noncovalent binding, and side reactions of the INCA compounds.  To assess short-term stability of the complex, samples were extracted by vortex mixing with 2 volumes of buffer-saturated ether for 2 min and briefly centrifuged to separate the phases. Extraction of samples containing only calcineurin and labeled PVIVIT had no effect on peptide binding and did not prevent subsequent displacement of peptide by INCA2. Extraction of samples containing only the INCA compound prior to incubation with calcineurin and PVIVIT demonstrated that removal of unbound INCA2 from the aqueous phase was complete within 10 s. To assess long-term stability of the complex, samples containing PVIVIT and blocked calcineurin were supplemented with 5 mM NEM or 5 mM DTT, and PVIVIT binding was monitored during the following 4 h. Because the calcineurin concentration in fluorescence polarization assays is greater than the K d of the calcineurin-PVIVIT interaction, each of these procedures is a sensitive test for the unmasking of blocked PVIVIT-binding sites. Mass Spectrometry-For MALDI-TOF mass spectrometry, samples were desalted and concentrated by binding the peptides to C18 resin in a ZipTip (Millipore) and eluting with 70% (v/v) acetonitrile and 0.1% (v/v) trifluoroacetic acid. Eluted peptides were diluted 1:10 with ␣-cyano-4-hydroxycinnamic acid matrix (10 mg/ml in 50% (v/v) acetonitrile and 0.1% (v/v) trifluoroacetic acid) and dried onto the target plate. Mass spectra were obtained using an Applied Biosystems Voyager-DE STR instrument operated in reflector mode with accelerating voltage of 20 kV, grid voltage at 72%, guide wire at 0.1%, and extraction delay time of 175 ns. Calibration was performed using the MSCAL2 mass standard set (Sigma).
A minor peak frequently observed in the peptide sample having a mass of ϳ1776 Da is apparently the peptide fragment RGSSY-FYSYPAVCEF. Corresponding peaks in samples that had been incubated with INCA1, INCA2, or NEM or in samples subjected to prolonged incubation with INCA6 to allow the reaction to approach  completion were shifted by the same mass as the main peak in those incubations. A peak of mass ϳ1397 Da, matching the calculated mass of the fragment RGSSYFYSYPAV, was present occasionally both in untreated peptide samples and in samples that had been incubated with NEM or INCA compounds. The reaction or lack of reaction observed with these peptide fragments reinforces the conclusion that all of the compounds investigated reacted with the cysteine -SH group.
Cellular Assays-Dephosphorylation of NFAT, nuclear import of NFAT, and induction of cytokine mRNAs in D5 T cells were assessed as described in previous publications (23)(24)(25).
Calcineurin activity in cell lysates was measured using a standard assay (39) with phosphorylated RII peptide as substrate. D5 T cells were chilled on ice, collected by centrifugation at 4°C, and resuspended in 50 mM Tris⅐HCl, pH 7.5, 1 mM EDTA, 0.1 mM EGTA, 0.2% (v/v) Nonidet P-40, 50 g/ml phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, and, where specified, 30 mM sodium pyrophosphate. After 5 min on ice, lysates were centrifuged, and supernatants were transferred to fresh tubes containing an equal volume of 50 mM Tris⅐HCl, pH 7.5, 100 mM NaCl, 2.5 mM CaCl 2 , 100 g/ml bovine serum albumin, and 1.5 M okadaic acid. Blank samples were prepared with resuspension buffer in place of cell lysate. The reaction was initiated by the addition of 32 P-labeled RII peptide, each tube was incubated for 15 min at 30°C, and the reaction was terminated by the addition of 100 mM potassium phosphate buffer containing 5% (w/v) trichloroacetic acid. Each sample was applied to a Dowex AG 50W-X8 cation exchange column (Bio-Rad), and the effluent was collected and its content of radiolabel determined by liquid scintillation counting.
Structural Modeling-Structural modeling was based on the coordinates of calcineurin A␣ from Protein Data Bank entries 1AUI and 1TCO (40,41). The coordinates of docked PVIVIT peptide were from Li et al. (33). Fig. 9 was prepared using RasMol.

RESULTS
The compounds INCA1, INCA2, INCA6, and INCA12 ( Fig. 1) have been shown to inhibit recognition of NFAT by calcineurin as assessed quantitatively by the binding of the fluorescent PVIVIT peptide to calcineurin. Several observations have raised the possibility that a covalent INCA-protein complex is involved in the inhibition of PVIVIT peptide binding. First, the most effective inhibitors were quinones or quinoneimines, compounds that are chemically reactive and that are known to form protein adducts. Second, the kinetics of binding were slow at intermediate concentrations of an INCA compound, as illustrated for INCA1 ( Fig. 2A), with a plateau level of inhibition reached only after tens of minutes at room temperature. This behavior is often indicative of the formation of a covalent ligand-protein complex. Finally, INCA1, INCA2, and INCA6 were inactivated as inhibitors in the competitive binding assay by preincubation with DTT ( Fig. 2B and not shown). Inactivation could reflect either reduction of the compounds by DTT to a less reactive form or covalent reaction with DTT.
Effect of Sulfhydryl-modifying Reagents-Protein sulfhydryl groups are likely sites of covalent modification by quinones. Consistent with this possibility, preincubation of calcineurin with the sulfhydryl-modifying reagent IAM impaired INCA2 competition, whereas preincubation of calcineurin by itself had no effect (Fig. 3A). In more extensive studies, preincubation of calcineurin with the maleimide INCA12 impaired competition by INCA1, INCA2, and INCA6 ( Fig. 3B and not  shown).
Pretreatment of calcineurin with the sulfhydryl-modifying reagents IAM, NEM, and INCA12 reduced the calcineurin-PVIVIT polarization signal even in the absence of an INCA competitor ( Fig. 3 and not shown). Very low concentrations of INCA12 were sufficient to reduce the calcineurin-PVIVIT polarization signal, but even high concentra-  tions did not fully eliminate the signal from bound PVIVIT (Fig. 4A). Similarly, for high concentrations of IAM and NEM, the calcineurin-PVIVIT signal settled to a plateau that was well above the signal of free peptide (not shown). The explanation for this behavior, documented below, is that pretreated calcineurin retains the ability to bind PVIVIT with reduced affinity.
In principle, the effect of the nonspecific alkylating reagents IAM and NEM could be due either to true impairment of calcineurin-peptide binding or to impaired immobilization of the fluorescent label when the PVIVIT peptide is bound. The second case would be evident in a reduction in the maximal polarization signal obtained by titrating fluorescent PVIVIT with IAM-pretreated or NEM-pretreated calcineurin and a corresponding reduction in the slope of the central portion of the binding curve. In fact, the curves for pretreated calcineurin showed no change in slope but instead were shifted rightward to an extent compatible with a 1.5-2-fold loss in affinity of the calcineurin-peptide interaction (Fig. 4B). Thus IAM and NEM are themselves partial inhibitors of the calcineurin-PVIVIT interaction, and experiments described below indicate that they act at the same site as the INCA compounds.
Identification of a Target Cysteine Residue-The inhibitory activity common to IAM, maleimides, and quinones/quinoneimines strongly supported the notion that these compounds derivatize a cysteine sulfhydryl of calcineurin. Pilot experiments examining the effect of several single Cys 3 Ala substitutions in calcineurin on the inhibitory activity of INCA2 pointed to Cys-266 as the likely reactive residue, and a more detailed study confirmed that the point mutant C266A was insensitive to the inhibitory effect of INCA2 (Fig. 5). Similar experiments demonstrated that this substitution also compromised the effectiveness of INCA1, INCA6, IAM, and maleimides (not shown).
Conversely, the other surface-exposed cysteine sulfhydryl groups were not required for the action of INCA compounds. The crystal structure of calcineurin shows six exposed cysteine residues in the catalytic domain, Cys-166, Cys-184, Cys-228, Cys-256, Cys-266, and Cys-336. We compared calcineurin in which all exposed cysteine side chains except Cys-266 were changed to alanine (Cys 3 Ala/Cys-266) to calcineurin with the same substitutions plus a C266V substitution (Cys 3 Ala/C266V). Binding of fluorescent PVIVIT to the two proteins was comparable (Fig. 6A). However, whereas replacement of the cysteines other than Cys-266 did not compromise the effectiveness of INCA1, INCA2, or INCA6, the C266V substitution in the context of these other Cys 3 Ala replacements completely blocked the effects of the inhibitors ( Fig. 6B and not shown). The C266V substitution in this context also FIGURE 7. MALDI-TOF mass spectrometric analysis of the reactions of NEM and INCA compounds with the synthetic calcineurin peptide. Data in each spectrum are plotted as intensity relative to the principal peak, and the abscissae have been labeled as mass (rather than as m/z) because all of the species detected are singly charged. Values given are for the monoisotopic mass. A, the unmodified peptide (upper spectrum) is detected at 2382.02 Da (calculated mass, 2382.06 Da). There is neither disulfide formation nor oxidation of cysteine to a sulfenic acid, sulfenyl-amide, sulfinic acid, or sulfonic acid. A minor peak at 1775.62 Da is provisionally identified as the N-terminal fragment RGSSYFYSYPAVCEF (calculated mass, 1775.77 Da) as discussed under "Experimental Procedures." After reaction with 1 mM NEM for 60 min (lower spectrum, displaced vertically for clarity), both the peptide and its N-terminal fragment have been derivatized at a single site, which can be confidently identified as the cysteine sulfhydryl given the reaction pH, 7.35, and the presence of a free sulfhydryl group in the peptide sample. B-D, reactions with INCA1, INCA2, and INCA6. The INCA compounds were present at 50 M initially, and reaction was allowed to proceed for 60 min. The reaction step was followed in panels B and C by a further 60-min incubation with NEM; however, omission of the NEM incubation did not alter the products observed. The peptide-INCA adducts were not further characterized, but chemically reasonable product structures based on the masses are shown as insets. B, the INCA1 reaction yields a principal peak at ϳ2732 Da. Examination of this peak at higher resolution reveals a minor overlapping peak at ϳ2730 Da that we interpret as the oxidized form of the adduct (supplemental Fig. 1, available in the on-line version of this article). The peak at ϳ2125 Da is the product of reaction with the ϳ1775 Da peptide fragment. C, the INCA2 reaction gives a peak at ϳ2536 Da, consistent with hydrolysis of the labile quinoneimine and loss of the chlorine substituents. An overlapping peak at ϳ2538 Da is resolvable when the data are plotted on an expanded m/z-axis (not shown). D, in the INCA6 reaction, at 60 min the main peak is unreacted peptide and the minor peak at ϳ2666 Da is the adduct. blocked the effect of NEM (not shown). These observations showed that the presence of Cys-266 is necessary and sufficient for a complete block of PVIVIT binding by INCA compounds.
Formation of a Peptide-INCA Adduct-We tested the ability of the three INCA compounds to react with the synthetic calcineurin peptide RGSSYFYSYPAVCEFLQHNN. The peptide is calcineurin-(254 -273) with a C256S substitution to eliminate the possibility of a covalent reaction at that position.
MALDI-TOF mass spectrometry demonstrated that INCA1 and INCA2, at micromolar concentrations, react covalently with the synthetic calcineurin peptide (Fig. 7, A-C). Peptide-INCA adduct formation was blocked in each case by prior incubation with excess NEM under conditions that derivatized the -SH group (Fig. 7A and not  shown). Conversely, NEM failed to react with the peptide-INCA adducts present after a first incubation with INCA1 and INCA2 (Fig. 7,  B and C), confirming that the INCA compounds block the sulfhydryl group. Because the peptide is unlikely to have a preferred conformation in solution, the experiments show that INCA1 and INCA2 are sufficiently reactive to modify accessible cysteine sulfhydryl groups without necessarily forming an initial noncovalent complex.
INCA6 reacted more slowly (Fig. 7D), with full labeling of the peptide requiring several hours. Reaction was again blocked by pretreatment with excess NEM. The sluggish reaction suggests that covalent reaction with Cys-266 in calcineurin would require higher concentrations of INCA6, a local environment that increases the nucleophilicity of the Cys-266 thiol or a noncovalent interaction that assists in targeting INCA6 to the site.
Direct examination of tryptic digests of INCA6-treated calcineurin did not provide evidence of an INCA-protein adduct. Rather, the same tryptic peptide containing Cys-266, identified by comparison with digests of C266V calcineurin, was detected in the digests of untreated and treated calcineurin. However, it has been difficult to demonstrate adducts in tryptic digests of other proteins that are known to be covalently modified by quinones, probably because the sulfur-quinone bond is labile (42,43). Lability of the adduct would not prevent its detection in the experiments with the synthetic peptide, because in those experiments the INCA compound was present in excess and the initial reaction mixture was directly examined by mass spectrometry.
Reversibility of Modification by INCA2-Physical association of INCA compounds with calcineurin has been demonstrated by NMR spectroscopy and by cochromatography (25). Two further results indicate that spontaneous dissociation of INCA2, if it occurs, is extremely slow. First, extraction with ether for 2 min did not restore PVIVIT binding after blockade with the INCA compound was complete, even though the nonpolar INCA2, when not bound to calcineurin, rapidly partitions into ether. Second, incubation of INCA2-blocked calcineurin with excess NEM, at an NEM concentration that rapidly blocks the inhibitory site on unmodified calcineurin, did not lead to any recovery of PVIVIT binding during the subsequent 4 h. The modification by INCA2 is nevertheless chemically labile, because treatment with excess DTT largely reversed the inhibitory effect.
Effects of INCA2 in Cells-The chemical reactivity of INCA1 and INCA2 has raised the issue of their suitability for cellular studies. In previous work, INCA6 inhibited the dephosphorylation of NFAT and calcineurin-NFAT signaling without inhibiting calcineurin enzymatic activity (25). INCA1 and INCA2 were not studied at that time because of their cytotoxicity, and even the less reactive INCA6 exhibited toxicity in some types of cells.
It proved possible to examine the effects of INCA2 in D5 T cells, for which INCA2 is not cytotoxic when used at low micromolar concentra-tions. At first glance, INCA2 had effects similar to those of INCA6, preventing the dephosphorylation of NFAT, the nuclear import of NFAT, and the induction of mRNAs encoding tumor necrosis factor-␣, interferon-␥, and macrophage inflammatory proteins MIP-1␣ and MIP-1␤ (Fig. 8, A-C, and not shown). However, closer examination revealed that the physiological effects were associated with a general inhibition of calcineurin catalytic activity (Fig. 8D). A likely mechanism

DISCUSSION
The principal result of this study has been to locate the binding site on calcineurin for the calcineurin-NFAT signaling inhibitors INCA1, INCA2, and INCA6. The position of the INCA-binding site in the vicinity of Cys-266 of calcineurin and its spatial relationship to the previously mapped NFAT docking site on calcineurin are illustrated in Fig. 9, A and B. The distance from S ␥ of Cys-266 to the nearest proline ring atom of the bound PVIVIT peptide is Ͼ15 Å. Thus, contrary to expectation, the most efficacious INCA inhibitors do not displace PVIVIT peptide and NFAT by binding competitively in the core PXIXIT recognition site but rather act by inducing an allosteric change in the NFAT-binding site.
The probable locus of the allosteric change is the ␤11-␤12 loop of calcineurin, which positions residues Phe-299 and Pro-300 to form the proline and isoleucine pockets of the peptide-binding site (33). Cys-266 and other residues in or just preceding helix 10 are in intimate contact with the residues that anchor both ends of the ␤11-␤12 loop (Fig. 9C). Tyr-262 makes extensive contacts with the side chains of Arg-292 and Ser-301 and forms a template for the peptide backbone from residue 292 to residue 295. Val-265 and Leu-269 are in contact with Ser-301, and Cys-266 itself is exposed in a canal on the surface of calcineurin, nestled against Tyr-262 and in loose contact with the side chains of Ser-294 and Ser-301. Binding of a ligand in direct contact with the Cys-266 sulfhydryl, whether binding is covalent or noncovalent, is likely to require a local structural rearrangement and could alter the conformation of the ␤11-␤12 loop. In addition to any changes in the immediate neighborhood of Cys-266, movement of helix 10 could also reposition Leu-275 in the short segment connecting helix 10 and ␤-strand 10, thereby altering its packing against Phe-299 and its contribution to formation of the proline pocket.
A concise explanation of the inhibitory effect of INCA compounds is that introduction of these substituents at Cys-266 induces a structural rearrangement that alters the PVIVIT docking site and that formation of a covalent sulfur-INCA bond provides part of the energy for the rearrangement. In fact, all of the efficacious INCA compounds have the potential for covalent binding. INCA1 and INCA2, in particular, are highly reactive with the cysteine sulfhydryl group. Their specific action on PVIVIT binding, via derivatization of Cys-266, could be simply a byproduct of their general reactivity, or there could be a noncovalent interaction that targets the compounds to the vicinity of Cys-266 prior to covalent binding. INCA6 is less intrinsically reactive under our experimental conditions. Efficient formation of a calcineurin-INCA6 covalent adduct would require assistance through targeting of INCA6 to the site or through heightened reactivity of the Cys-266 sulfhydryl. The possibility also remains that INCA6 is not primarily a covalent inhibitor.
A more refined analysis of the reaction with calcineurin will entail characterization of the calcineurin-INCA adducts that arise under physiological conditions. Steric and other constraints in the protein might favor products that differ from those of the reaction with synthetic peptide discussed in the legend to Fig. 7. For example, the INCA2 linkage to protein could involve a thioether bond, as observed in quinone cofactors of some amine dehydrogenases (44,45) and as inferred for many other protein-quinone adducts (42, 43, 46 -49), or an ipso adduct at the imine carbon, as observed in the complex of N-acetyl-pquinoneimine with papain (50). Cysteinyl-quinone thioether adducts are themselves highly reactive when in the oxidized form (51) and could undergo further reactions following initial adduct formation. An alternative mechanism of quinone-initiated modification of proteins is the production of reactive oxygen species (52), which in their turn react with cysteine side chains to give oxides of sulfur or sulfenyl amides (52,53). However, NMR measurements support a physical interaction between calcineurin and INCA compounds (25), and the experiments with synthetic peptide yielded no evidence that the INCA compounds oxidize cysteine sulfhydryl groups under our conditions. A sidelight to our experiments is that the covalent inhibitors IAM, NEM, and INCA12 react with Cys-266 and reduce the affinity of calcineurin for PVIVIT measurably, but much less dramatically than INCA1, INCA2, and INCA6. This finding highlights the role of an induced conformational change rather than simple covalent derivatization of Cys-266 in causing inhibition. It further serves as an indicator that other partial inhibitors of calcineurin-PVIVIT binding identified by high throughput screening (25) may bind covalently or noncovalently at this allosteric site.
The current experiments were undertaken as a step toward the identification of better inhibitors of calcineurin-NFAT signaling. Our earlier mapping of the PXIXIT-binding site (33) has provided a structural template for the design of inhibitors that compete directly with the NFAT docking peptide. Inhibitors binding at the Cys-266 site will present a less tractable task for structure-based design until the structure of the calcineurin-INCA complexes has been determined. Two immediately workable approaches suggested by our experiments are further high throughput screening with modifications of the screening assay, such as use of C266V calcineurin or preincubation of library compounds with DTT or another thiol, that will tip the balance in the direction of noncovalent inhibitors and examination of tethered ligands (54) focused on the site we have defined. Less conventionally, it may prove possible to develop covalent ligands that are targeted to the site by a noncovalent interaction and are relatively unreactive with nonspecific sites (55,56). The pursuit of covalent inhibitors is an uncertain exercise in most cases because of the difficulty of discriminating between a desired increase in the strength of the noncovalent interaction and an unwanted increase in chemical reactivity. Here, though, the combination of a single physiologically relevant site of interaction, judicious use of cysteine mutants, and careful control experiments may provide an avenue around this technical obstacle.
The sensitivity of protein-ligand affinity at the PXIXIT-binding site to minor modifications at Cys-266, including carboxymethylation and the C266A substitution, raises an intriguing possibility for regulation of calcineurin signaling in cells. Previous work has focused attention on the possibility that calcineurin catalytic activity is regulated by the physiological production of reactive oxygen species and oxidation of Fe 2ϩ at the active site (34 -37, 57-59). If redox reactions or endogenous regulators in cells also act at the INCAbinding site, they could complement redox modulation of overall calcineurin catalytic activity with a layer of regulation having selective effects on different substrates.