Structural Delineation of the Calcineurin–NFAT Interaction and its Parallels to PP1 Targeting Interactions

doi:10.1016/j.jmb.2004.07.068

J. Mol. Biol. (2004) 342, 1659–1674

Structural Delineation of the Calcineurin–NFAT Interaction and its Parallels to PP1 Targeting Interactions Huiming Li1,2, Anjana Rao1,2 and Patrick G. Hogan2* 1

Department of Pathology Harvard Medical School, 200 Longwood Avenue, Boston MA 02115, USA 2 CBR Institute for Biomedical Research, 200 Longwood Avenue, Boston, MA 02115 USA

Calcineurin is a phosphoprotein phosphatase that channels intracellular Ca signals into multiple biological pathways. Calcineurin is known to interact directly with its substrate nuclear factor of activated T cells (NFAT or NFATc), with other substrates, and with several targeting and scaffold proteins including AKAP79 and Cabin1/cain. The calcineurin–NFAT interaction depends on recognition of a PxIxIT sequence motif present in NFAT-family proteins and in certain other calcineurin-interacting proteins. Here, we define the structural basis for the interaction of calcineurin with NFAT and with other proteins possessing the PxIxIT motif. The calcineurin–PxIxIT contact has a direct parallel in the contact of protein phosphatase 1 with its regulatory proteins, suggesting that the evolution of these related phosphatases involved local remodelling of an ancestral docking site. q 2004 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: calcineurin; NFAT; phosphatase; protein–protein interaction; T cell activation

Introduction Protein serine/threonine phosphatases present an intriguing problem in how specificity is maintained in intracellular signalling networks. In mammalian cells, a few major serine/threonine phosphatases, protein phosphatase 1 (PP1), PP2A, and calcineurin, dephosphorylate an extensive set of phosphoproteins,1–4 seemingly implying that there is little substrate specificity. Paradoxically, however, dephosphorylation of individual substrates, or even of individual phosphorylated sites in the same substrate, can be largely the province of a single phosphatase. The accepted explanation of this paradox is that the enzymatic activity of the protein serine/threonine phosphatases is tightly controlled by their association with targeting proteins, inhibitors, and regulators.5,6 The case of protein phosphatase 1 has been extensively studied.1,6 The functioning PP1 holoenzyme consists of the catalytic subunit associated with one of its regulatory proteins. A particular Abbreviations used: NFAT, nuclear factor of activated T cells; Bpa, p-benzoylphenylalanine; FP, fluorescence polarization. E-mail address of the corresponding author: [email protected]

docking site on PP1, the RVxF recognition site, accounts for binding of many of these regulatory proteins and, in rare cases, for the direct binding of a PP1 substrate.6–8 The structure of PP1 catalytic subunit in complex with an RVxF peptide has been solved by X-ray crystallography, and shows the peptide bound in an extended conformation within a shallow channel on the protein surface, contacting residues that are highly conserved in PP1 from various species.9 The catalytic subunit makes similar canonical contacts in a complex with the targeting subunit MYPT1, and additional contacts that are specific to MYPT1.10 In contrast to our knowledge of PP1 targeting, our understanding of substrate and regulatory protein recognition by other protein serine/threonine phosphatases is much less complete. Calcineurin (PP2B or PP3) is a Ca/calmodulindependent phosphatase that is a principal mediator of cellular responses triggered by intracellular Ca signals.3,4 Calcineurin has a critical role in responses to environmental stress in Saccharomyces cerevisiae and in virulence in pathogenic fungi.11,12 Calcineurin is widely expressed in the tissues of multicellular organisms, where it connects intracellular Ca signals to distinctive outputs characteristic of the cell type or developmental stage. Thus, for example, it modulates sensory reception in Caenorhabditis elegans and engages in crosstalk with

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

1660 the pathways controlling formation of photoreceptor cells and wing veins in Drosophila melanogaster.13,14 In vertebrates, intracellular Ca signals acting through calcineurin participate in the development, physiology, and pathophysiology of the nervous system, the immune system, and striated muscle.3,4,15–17 Among the many phosphoprotein substrates of calcineurin in mammalian cells, nuclear factor of activated T cells (NFAT) has received particular attention. NFAT-family transcription factors were first recognized as downstream effectors of calcineurin in T cell activation.15,17–19 Clinical interest in calcineurin has continued to focus on T cells and other cells of the immune system, since calcineurin– NFAT signalling is the principal target of cyclosporin A and FK506,20 immunosuppressive drugs that are widely prescribed to prevent or treat rejection of transplanted organs and tissues. However, recent work has shown that calcineurin–NFAT signalling also controls gene transcription in a number of other developmentally and physiologically important processes, including osteoclast differentiation, cardiac valve development, differentiation of slow-twitch skeletal muscle fibers, and myocardial hypertrophy.15–17,21 Calcineurin–NFAT signalling depends on transient and reversible recognition by calcineurin of a conserved PxIxIT motif in the N-terminal regulatory domain of NFAT.22 Calcineurin has relatively low affinity (Kdw20 mM) for the native SPRIEIT target sequence in NFAT1, and signalling is severely compromised not only by mutations in the SPRIEIT motif that weaken binding, but also by substitutions that increase the strength of binding.23 Thus, the calcineurin–NFAT interaction poses, in especially sharp focus, a question that is likely to arise for many of the protein–protein interactions that link signalling enzymes to their intracellular substrates and regulatory proteins: how to build high specificity into a readily reversible interaction. Although calcineurin engages in a variety of protein–protein interactions that maintain its signalling specificity (direct binding to substrates, tethering at locations adjacent to substrates, and interaction with regulatory or inhibitory proteins22–38), the structural basis of calcineurin–protein recognition has not been defined for any of its known complexes. Here, we have examined the complex of calcineurin and VIVIT peptide, a ligand with relatively high affinity for the PxIxIT-binding site that was identified through screening a combinatorial peptide library.23 An initial direct approach to determining the structure of the calcineurin– VIVIT complex, cocrystallizing peptide with calcineurin, failed for reasons that became apparent in the course of our studies. Therefore, we have used peptide crosslinking, in silico docking, and experimental analysis of calcineurin mutants to map the PxIxIT docking site on calcineurin and to identify the interactions that underlie specific recognition of NFAT. The results provide a picture of a site that achieves precise recognition in the context of a

The Docking Interaction of Calcineurin with NFAT

modest net free energy of binding, and in addition suggest an unexpected evolutionary parallel between the PxIxIT recognition site of calcineurin and the RVxF recognition site of PP1.

Results Identification of the VIVIT docking site on calcineurin The tight specific binding of VIVIT to the NFAT docking site on calcineurin suggested that the site could be localized by crosslinking the bound peptide to calcineurin. We synthesized the peptides VIVIT*8 and VIVIT*10 (Figure 1a), incorporating the crosslinker p-benzoylphenylalanine (Bpa) at the position indicated. Each peptide was incubated with the calcineurin Aa catalytic domain (here denoted CnAa) and exposed to UV light to activate Bpa, and the reaction mixture was analyzed on an SDS/polyacrylamide gel. Crosslinked product was detected as a band of lower mobility in the gel, due to the w2 kDa increase in its mass, as illustrated for peptide VIVIT*8 in Figure 1b. Only a single retarded band was observed, indicating that the peptide crosslinked to a single site on CnAa. Biotin-labelled VIVIT*8 and VIVIT*10 peptides also reacted efficiently with CnAa, producing a crosslinked product detectable by Western blotting with an antibody to biotin (Figure 1c). CnAa crosslinked to biotinyl-VIVIT*8 was subjected to trypsin digestion, and biotinyl peptides were affinity-purified on immobilized avidin and analyzed by matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) mass spectrometry (Figure 1d). A CnAa sample that was trypsindigested without prior crosslinking served as a control. The peak of mass 2384 Da observed in the cross-linked sample indicated the crosslinking of peptide to a CnAa fragment of 470 Da and identified the likely site of crosslinking as the tripeptide MYR, calcineurin(290-292). Trypsin digestion of CnAa crosslinked to biotinylVIVIT*10 yielded a peak of identical mass, 2384 Da, indicating that the VIVIT*10 peptide labelled the same fragment of CnAa (data not shown). Mutations in the docking site reduce affinity for VIVIT To confirm the location of the VIVIT-binding site, and to map residues participating in peptide binding, we introduced mutations into CnAa. The protein surface surrounding calcineurin(290-292) is largely contributed by residues of b strands 11 and 14, the proximal portions of the b11-b12 loop, and the segment connecting helix 9 and b strand 8.39,40 Initially we replaced the individual residues Y288 and M290 of b strand 11, F299 of the b11-b12 loop, and N330, I331, and R332 of b strand 14 with alanine, and examined the effect on binding of

The Docking Interaction of Calcineurin with NFAT

1661 fivefold by the R332A and M290A substitutions, wtenfold by the Y288A and N330A substitutions, and more than 20-fold by the F299A and I331A substitutions. This mapping of the site by peptide crosslinking and mutagenesis confirms our earlier conclusion22,23 that the VIVIT docking site does not coincide with the active site of CnAa (Figure 2d and e). Proteins that incorporated substitutions at combinations of the sites, either N330A/I331A/ R332A (NIROAAA) or Y288A/M290A/F299A (YMFOAAA), showed complete loss of binding in the range of calcineurin concentrations tested (Figure 3a). Mutation of the docking site selectively impairs dephosphorylation of NFAT We examined whether loss of docking in the NIROAAA mutant had a selective effect on dephosphorylation of NFAT. The phosphatase activity measured in control experiments with standard calcineurin substrates, phospho-RII peptide and p-nitrophenylphosphate (pNPP), was high, although somewhat less than that of wild-type calcineurin (Figure 3b and not shown). However, the NIROAAA triple mutant was ineffective in dephosphorylating NFAT1 (Figure 3c). Further studies will be needed to define calcineurin mutations whose effect is strictly limited to impairment of docking. Crosslinking of VIVIT to mutant calcineurins

Figure 1. UV crosslinking of VIVIT to CnAa and mass spectrometric analysis of the crosslinked product. a, Sequence of VIVIT peptide and the sites of pbenzoylphenylalanine substitution in VIVIT*8 and VIVIT*10. b, SDS-PAGE of crosslinking reactions containing increasing amounts of VIVIT*8 peptide, stained with Coomassie brilliant blue. Crosslinked product (arrowhead) is observed as a band with w2 kDa increase in apparent molecular mass. c, Specific labeling of CnAa by biotinyl-VIVIT*8 and biotinyl-VIVIT*10 peptides. The same blot was stained with Ponceau red (upper panel) and with HRP-labelled anti-biotin (lower panel). d, MALDI-TOF mass spectrometric analysis of CnAa, with and without crosslinking to biotinyl-peptide, after trypsin digestion and affinity purification on avidin resin. The mass of biotinyl-VIVIT*8 is 1915 Da, and the mass of the tryptic fragment with crosslinked biotinyl-VIVIT*8 is 2384 Da.

VIVIT peptide to calcineurin in a fluorescence polarization (FP) assay (Figure 2a–c). Each individual substitution decreased the affinity of calcineurin for VIVIT peptide: the Kd was shifted three- to

In an independent method of assessing the interaction between mutant calcineurins and VIVIT peptide, the proteins were UV-crosslinked to biotinylated VIVIT*8 and VIVIT*10 peptides (Figure 4a). In most cases the degree of crosslinking agreed qualitatively with the ranking derived from FP assays, the labelling with biotinyl-VIVIT*8, for example, being strongest for the R332A mutant; weaker for N330A, Y288A, and F299A; and weakest for I331A. However, two mutant proteins exhibited crosslinking patterns that could not be explained solely by the relative affinities measured in the FP assay. M290A calcineurin was not labelled by either peptide, even though that mutant displayed VIVIT binding similar to R332A calcineurin in the FP assay, indicating that M290 is itself the residue covalently labelled in wild-type CnAa. It has been shown in other instances that the Bpa group reacts readily with the g-methylene or 3-methyl carbon of methionine if it is suitably positioned.41 In addition, the I331A mutant displayed unexpectedly strong crosslinking to the VIVIT*10 peptide (Figure 4a). This suggested that, unlike the interaction with VIVIT (Figure 2c) or with VIVIT*8 (Figure 4a), the interaction between VIVIT*10 peptide and calcineurin was not hindered by the I331A substitution. This conclusion was directly supported by competitive binding assays (Figure 4b and c).

1662

The Docking Interaction of Calcineurin with NFAT

Figure 2. Location of the VIVIT docking site on calcineurin. a, Primary sequence of calcineurin in the region analyzed, with b strands 11–14 indicated and residues that were individually replaced by alanine in the mutated proteins shaded. The locus of crosslinking, calcineurin(290-292), is at the C terminus of b strand 11. b and c, Binding of fluorescent VIVIT peptide (100 nM) to wild-type and mutated calcineurins, detected by the increase in polarization of fluorescence when peptide is bound. Estimated dissociation constants, Kd, were: wild-type (WT), 0.75 mM; R332A, 3 mM; M290A, 5 mM; N330A, 8 mM; Y288A, w10 mM; and F299A and I331A, O20 mM. d and e, The region surrounding the VIVIT docking site is highlighted in two different views of the CnAa catalytic domain (chain A, residues 21–347, from PDB entry 1TCO).

Molecular docking assists in visualizing the interaction between VIVIT and CnAa The location of the docking site explains the difficulty in cocrystallizing calcineurin and VIVIT peptide, because it coincides with a crystal packing contact in several published structures,39,40,42 and

occupancy of the site by peptide would interfere with crystallization. Therefore, we turned to in silico docking simulations using AutoDock 3.0 software43 to complement the experimental analysis of the protein–ligand interaction. The large number of torsional degrees of freedom of VIVIT and PVIVIT requires care to ensure that there is adequate

The Docking Interaction of Calcineurin with NFAT

Figure 3. VIVIT peptide binding and catalytic activity of the CnAa triple mutants. a, Both NIR/AAA calcineurin and YMF/AAA calcineurin showed a profound loss of ability to bind VIVIT peptide in the fluorescence polarization assay. b, Dephosphorylation of RII phosphopeptide by wild-type and mutant calcineurin. In the control sample no calcineurin was added, and in the other samples the CnAa concentration was 200 nM, within the linear range of the enzymatic assay. Phosphate release was monitored, using Malachite green, as absorbance at 650 nm. The catalytic activity was only moderately decreased by the NIR/AAA substitution. c, Dephosphorylation of NFAT1 incubated with wild-type or mutant calcineurin at the concentrations indicated. The phosphorylated and dephosphorylated forms of NFAT were resolved by SDS-PAGE and detected by Western blotting with anti-NFAT1. Dephosphorylation of NFAT1 by the NIR/AAA mutant was inefficient at all concentrations of calcineurin tested.

sampling of peptide conformations, and we addressed this problem by using comparatively large numbers of simulation runs and energy evaluations as noted in Materials and Methods. Because the accuracy of docking results also can be influenced by the length of peptide chosen, we

1663 performed docking simulations with both the 5mer VIVIT peptide and the 6mer PVIVIT peptide. Each docking simulation produced a clearly defined cluster of conformations with lowest predicted energy (Figure 5). The predicted docked complexes were replicated in independent simulations for each peptide, and the results for VIVIT and PVIVIT were in agreement in the region Ile8– Thr11, implying that adequate conformational sampling had been performed. The orientation of the docked 5mer VIVIT peptide with respect to CnAa is shown in Figure 6a. VIVIT assumes a b-strand conformation, juxtaposing itself parallel with b strand 14 of calcineurin so that sheet 2 of the central b sandwich is extended. As expected from the crosslinking data, both Ile8 and Ile10 side-chains project into the vicinity of M290 (Figure 6b). Main-chain hydrogen bonds are formed between Val7 of the peptide and M329 of CnAa, between Val9 of peptide and M329 and I331 of CnAa, and between Thr11 of peptide and I331 and Q333 of CnAa. An additional side-chain hydrogen bond is made between the g-hydroxyl group of Thr11 and the side-chain amide of N330 (Figure 6c). Among van der Waals contacts, the packing interactions between residues Ile8 and Ile10 of the peptide and CnAa are prominent. The Ile8 side-chain rests in a non-polar recess formed by M290, F299, P300, M329, and I331. The side-chain of Ile10 packs against the platform formed by the side-chains of I331 and Y288 (Figure 6d). In contrast, the Val9 side-chain makes only passing contact with N330. In representative structures from docking runs with the 6mer PVIVIT peptide, the backbone of VIVIT aligns well with the model of calcineurin– 5mer VIVIT complex, and the major van der Waals and hydrogen-bonding interactions are preserved (Figure 7). The new feature defined in this complex is a proline pocket in which Pro6 is sandwiched between F299 of the b11-b12 loop and N327 of the b13-b14 turn, fitting into a cavity bounded also by residues Q194, F195, L275, Y324, and M329 (Figure 7b and c). The predicted placement of Pro6 allows good non-polar contacts with calcineurin and typical proline Ca K H /O and Cd K H /O bonds44 to the backbone and side-chain, respectively, of N327. As a result of the calcineurin–Pro6 interaction, Val7 reorients locally without a change in its van der Waals contacts. The predicted protein– ligand contacts are summarized in Table 1. Experimental confirmation that Thr11 in VIVIT interacts with N330 in CnAa The docking simulations consistently indicated a hydrogen bond between Thr11 of VIVIT and N330 of CnAa. To test for this interaction experimentally, we examined binding of a ThrOAla mutant peptide, VIVIA, in more detail. Consistent with the predicted interaction, replacement of Thr11 by alanine in VIVIT peptide caused a reduction in affinity between the ligand and wildtype CnAa (Figure 8a) comparable to that caused

1664

The Docking Interaction of Calcineurin with NFAT

Figure 4. The effect of mutations in CnAa on the efficiency of crosslinking VIVIT*8 and VIVIT*10 peptides. a, Crosslinking of VIVIT*8 or VIVIT*10 to the indicated GST-CnA, wild-type or mutant, was assessed by Western blotting with anti-biotin. Staining for GST (lower panel) served as a loading control. b, Competitive assay for binding of VIVIT peptide or VIVIT*10 peptide to CnAa. Binding was detected through the change in polarization caused by displacement of fluorescent VIVIT from the binding site by the indicated concentrations of competitor peptide. VIVIT*10 showed far lower affinity than unmodified VIVIT for wild-type CnAa. c, A competitive binding experiment similar to that shown in b, except with I331A CnAa. Wild-type VIVIT bound poorly to the mutated protein, confirming the observation of Figure 2c. In contrast, VIVIT*10 binding to I331A CnAa was modestly increased over its binding to wild-type CnAa in b.

by the N330A substitution in calcineurin. More tellingly, the two replacements did not contribute independently to the change in binding energy, since the Thr/Ala substitution in VIVIT did not further impair binding to N330A (Figure 8b). The experimental findings support the model in which Thr11 of VIVIT peptide contacts N330 of calcineurin.

Discussion Docking of NFAT and other protein partners of calcineurin We have mapped the principal calcineurin–NFAT contact by photocrosslinking, MALDI-TOF mass spectrometry, and the introduction of point

The Docking Interaction of Calcineurin with NFAT

Figure 5. Clustering of the 100 docked conformations of PVIVIT from a typical docking trial. The AutoDock 3.0 empirical energy scoring function is plotted on the abscissa. Each bin of the histogram includes all confor˚ all-atom RMSD of its best docked mations within 1 A conformation.

mutations in calcineurin. For this purpose we used a collection of derivatives of VIVIT peptide, an optimized ligand based on the PxIxIT motif of NFAT-family proteins. The core docking site on

1665 calcineurin comprises a proline pocket, bounded by residues in and immediately adjacent to the b13-b14 turn and by a constellation of side-chains projecting from the loops preceding b strands 6, 10, and 12; and a shallow surface trough delimited by b strand 11, b strand 14, and residues of the b11-b12 loop (Figure 7). These findings confirm our earlier conclusion that the docking site is distinct from the calcineurin catalytic site and from the surface region that interacts with immunosuppressive drug/immunophilin complexes.22,23 It is possible that residues flanking the PxIxIT motif in NFAT proteins contact a somewhat more extensive surface on calcineurin, since affinity selection of synthetic peptides has indicated that calcineurin imposes a preference for certain residues at positions flanking the core PxIxIT segment.23 We further refined the map of the contact site using a combination of in silico docking simulations and experimental evidence. The resulting model of contacts between calcineurin and the PxIxIT recognition peptide has the following experimental support. (1) Substitution of alanine for each of six residues that are in close apposition to peptide in

Figure 6. Model of the CnAa–VIVIT complex obtained from docking simulations with 5mer VIVIT peptide. a, The peptide is red in a ball-and-stick representation, and CnAa is grey in a ribbon diagram, with b strands 11–14 highlighted in yellow. b, In the model, the side-chains of Ile8 and Ile10 project toward M290 of CnAa. c, The predicted hydrogen bond between Thr11 of VIVIT and N330 of CnAa is shown. d, van der Waals packing interaction between Ile10 of the VIVIT peptide and I331 and Y288 of CnAa.

1666

The Docking Interaction of Calcineurin with NFAT

Figure 7. Model of the CnAa–VIVIT complex obtained from docking simulations with 6mer PVIVIT peptide. a, Stereo view of docked PVIVIT peptide and CnAa(280-340), showing the alignment of the peptide alongside b strand 14. b Strands 11–14 are marked. b, Linear sequence of calcineurin in the segments contributing to the binding site. c, Stereo view of docked PVIVIT peptide, with relevant calcineurin side-chains labelled. The proline pocket is at the lower right between F299 and N327, and the Ile8 recess to its left between F299 and M290.

the model reduced the affinity of the binding interaction. (2) Modified VIVIT peptides in which the crosslinker p-benzoylphenylalanine was substituted for either Ile8 or Ile10 in the peptide covalently labelled M290 of calcineurin, in agreement with the predicted orientation and position of the Ile8 and Ile10 side-chains. (3) The I331A substitution in calcineurin had opposite effects on binding of wild-type VIVIT and of the VIVIT*10 peptide. These contrasting effects are most readily explained by a loss of favorable protein–peptide interactions involving Ile10 in the first case and by a

reduction of unfavorable protein–peptide interactions involving Bpa10 in the second case, consistent with the placement of Ile10 and I331 in direct contact by the model. (4) Either the N330A substitution in calcineurin or replacement of Thr11 by alanine in the peptide decreased the strength of the interaction, but there was no additive effect when both substitutions were present, strongly suggesting direct contact between these side-chains. Important remaining structural questions are how docking serves to position the multiple phosphoserine residues of NFAT for dephosphorylation, and

The Docking Interaction of Calcineurin with NFAT

1667

Table 1. Summary of calcineurin–PVIVIT contacts

yeasts to nematodes, insects, and mammals, suggesting that they may have an essential role in calcineurin–protein docking in all these organisms. Indeed, although the classical NFAT-family substrates are confined to vertebrates, cognate calcineurin-binding motifs have been observed in two calcineurin substrates in S. cerevisiae, the unrelated transcription factor Crz1p (PIISIQ) and Hph1p (PVIAVN).24,25 On the other hand, some substrates and at least one important signalling pathway downstream of calcineurin, the NF-kB pathway in T cells, have been shown to be insensitive to inhibition by VIVIT peptide.23 Thus, it is likely that additional calcineurin–substrate docking interactions remain to be discovered.

A. Polar contactsa Backbone N327 (C]O) Pro6 Ca K H Val7 (C]O) M329 (NH) Val9 (NH) M329 (C]O) Val9 (C]O) I331 (NH) Thr11 (NH) I331 (C]O) Thr11 (C]O) Q333 (NH) Side-chain N327 (side-chain) Pro6 CdKH Thr11 (g-OH) N330 (side-chain) B. Non-polar contactsa Pro6 Q194, F195, L275, F299, Y324, N327, M329 Val7 V328 Ile8 M290, F299, P300, M329b, I331b Val9 V328, N330 Ile10 Y288, I331 Thr11 R332 a

Criteria for assessing hydrogen bonds and van der Waals contacts are as stated in Materials and Methods. b These contacts are dependent on the Ile8 side-chain conformer.

whether additional recognition sequences45,46 in NFAT are utilized. These questions will be answered ultimately by X-ray crystallography of the protein–protein complex. In silico modeling shows that the site we have identified can accommodate the naturally occurring NFAT sequences PRIEIT, PSIQIT, and PSIRIT without unfavorable steric interactions. Similar recognition motifs have been identified in the endogenous calcineurin inhibitor Cabin1/cain (PEITVT)47,48 and in the viral calcineurin inhibitor A238L (PKIIIT).49 One targeting protein, AKAP79, presents a longer variant sequence, PIAIIIT,50 that can be accommodated in the same binding site with its first isoleucine in the proline pocket and its proline displaced onto the adjacent surface of calcineurin. Notably, the surface residues implicated here in PxIxIT docking are identical in calcineurin from diverse organisms, ranging from

Comparison of docking sites on calcineurin and PP1 Our model of the calcineurin–PVIVIT complex reveals an unexpected parallel between calcineurin and the related protein serine/threonine phosphatase PP1. In the previously determined structure of the PP1 catalytic subunit in complex with a regulatory subunit peptide, the peptide is in an extended configuration as in the calcineurin–PVIVIT complex and is, in part, aligned along b strand 14 of PP1 to extend the b sheet.9 Despite the fact that the two phosphatases prefer different substrates and differ in whether or not they utilize a separate targeting subunit, when the conserved backbones of calcineurin and PP1 are superposed, the two docking sites largely coincide. At a first level of analysis, the sites divide topographically into two sectors. One part interacts with the N-terminal portion of the target peptide, and here the proline pocket of calcineurin stands in sharp contrast to the rolling landscape of PP1 studded with negatively charged side-chains (Figure 9(a)). The other part, contacting the non-polar residues of the peptide, is a hydrophobic trough that

Figure 8. Effect of Thr/Ala replacement in VIVIT peptide. a, Competitive binding assay measuring binding of a 16mer VIVIT peptide or its Thr/Ala mutant, VIVIA, to wild-type CnAa. The rightward shift in the binding curve, and thus the reduction in the free energy of binding, is comparable to that produced by the N330A mutation in CnAa (see Figure 2b and c). Note that the parental VIVIT peptide used for this assay differs in sequence from the 16mer VIVIT peptide described by Aramburu et al.,23 and has a somewhat lower affinity for calcineurin. b, Binding of the same two peptides to N330A CnAa. In this case, the Thr/Ala replacement does not reduce the free energy of binding.

1668

The Docking Interaction of Calcineurin with NFAT

Figure 9. Topography of the docking sites on calcineurin and PP1. a, Views from the end of each site that would be occupied by the N-terminal residues of the corresponding peptide, illustrating the rather different surface topographies. F299 and other labelled residues in the foreground surround the proline pocket. b, Views from the end of each site that would be occupied by the C-terminal residues of the corresponding peptide, illustrating the similar topographies of the hydrophobic troughs in the two proteins. The trough in calcineurin, however, is closed off by the side-chains of F299 and P300, thereby forming the Ile8 recess.

presents a similar topography in the two proteins up to the point where it is truncated in calcineurin by residues F299 and P300 (Figure 9b). More detailed comparison of the two cases shows that only very limited differences in the primary sequences are needed to specify the divergent local architectures. Key differences between calcineurin and PP1 are a lengthened b11-b12 loop in calcineurin, due to the presence of an additional six residues including F299, and a differing configuration of the peptide backbone of the two proteins in the b13-b14 loop (Figure 10a). The relationship between the linear sequences and the three-dimensional structures of the sites is detailed in Figure 10b–d. The insertion in the b11-b12 loop positions F299 against the surface of b sheet 2, forming part of the proline pocket and a protruding ridge that

contacts Ile8 of the peptide. Tightening of the b13-b14 loop to a b turn in calcineurin and its displacement by Y324 permits deployment of the side-chain of M329 into the space between b sheets 1 and 2, with both Y324 and M329 also contributing to the proline pocket. In contrast, the remaining residues delimiting the proline pocket of calcineurin have or could reorient into a similar spatial disposition in the two proteins. The hydrophobic patch across the edge of b strands 11 and 14 is also left unchanged by the shorter b13-b14 turn. Each binding site is consolidated by other sequence specializations. Two residues, in particular, in the b13-b14 loop of calcineurin support its distinctive local architecture: the bulky hydrophobic side-chain of Y324, and N327, which accomodates the b turn and simultaneously traces an edge of the proline pocket. Outside the region of

1669

The Docking Interaction of Calcineurin with NFAT

rearranged backbone, the single notable specialization of the calcineurin peptide-binding site is the presence of N330 in b strand 14, which supplies a hydrogen bond to Thr11 in the peptide. In a similar way, the acidic residues in PP1 that stabilize the binding of RRVSFA peptide by interacting directly with its arginine side-chains or by contributing to the negative electrostatic potential are largely absent from calcineurin and can be considered a specialization of PP1. From an evolutionary perspective, it is intriguing that only very few features of the linear sequence are needed to specify the calcineurin site, suggesting that just a few mutational steps enabled calcineurin and PP1 to engage distinct partners in intracellular signalling networks. Further, both the calcineurin and PP1 sites have rather low affinity for their natural target sequences, indicating that the structural changes in an ancestral site were not constrained by a requirement to maintain highaffinity docking interactions. In light of the central role of F299 and P300 in creating both the proline pocket and the recess occupied by Ile8, a plausible scenario is that a functional and distinct calcineurin docking site was initially produced in a single step by an insertion in the b11-b12 loop, and later refined through evolutionary selection.

Conclusion In conclusion, we have identified the core docking site for NFAT on calcineurin, including features in calcineurin complementary to each of the conserved residues in the PxIxIT motif. These experiments give a structural dimension to the residue preferences determined in the earlier peptide selection experiments, and constitute a necessary step toward a quantitative understanding of specificity in this low-affinity recognition site. Comparison of the calcineurin and PP1 docking sites illuminates a contribution of local protein architecture to their divergent functions in signalling.

Materials and Methods Materials pNPP, reduced glutathione, and goat anti-biotin peroxidase conjugate were purchased from Sigma (St. Louis, MO). RII phosphopeptide and phosphatase activity assay reagents were purchased from either Calbiochem (San Diego, CA) or Biomol (Plymouth Meeting, PA). Sequencing-grade modified trypsin and SoftLinke Soft Release Avidin Resin were from Promega (Madison, WI), RapiGeste SF was from Waters (Milford, MA), and glutathione Sepharose resin and PreScission protease were from Amersham Bioscience (Piscataway, NJ). The peptides used in crosslinking experiments and fluorescence polarization assays were synthesized at Tufts University Core Facility using F-moc chemistry, and purified by reversed phase HPLC. Anti-NFAT151 and fluorescent VIVIT peptide52 were as described.

Plasmid constructs, protein expression, and protein purification The expression construct for the calcineurin Aa catalytic domain, calcineurin(2-347), has been described.23 Mutations were introduced using a PCR-based approach, verified by sequencing, and the cDNAs inserted into pGEX6P-1 vector (Amersham Bioscience, Piscataway, NJ) between the BamHI and XhoI sites. The plasmids were transformed into BL21 (DE3) cells and grown in LB medium with 100 mg/ml of ampicillin at 37 8C. When A600 nm of the culture reached 0.6, protein expression was induced with 0.5 mM IPTG, and incubation was continued at 20 8C overnight. Cells were harvested by centrifugation and lysed by sonication in lysis buffer containing 100 mM NaCl, 50 mM Tris–HCl (pH 8.0), 1 mM EDTA, 1 mM DTT, and 1 mM PMSF. The lysate was clarified by centrifugation and incubated with glutathione Sepharose resin at 4 8C overnight. The resin was extensively washed with lysis buffer, followed by elution with 10 mM reduced glutathione in the same buffer. Calcineurin for UV-crosslinking experiments was cleaved from the fusion protein and purified using PreScission protease as described in the protocol provided by the vendor. UV crosslinking Pilot experiments with synthetic VIVIT peptides, VIVIT*2 to VIVIT*5 and VIVIT*7 to VIVIT*12, each incorporating Bpa at the position indicated, showed that only VIVIT*8 and VIVIT*10 crosslinked efficiently to CnAa. For the experiments presented in Figures 1 and 4, a reaction mixture containing 30 mM CnAa and 15 mM, or other indicated concentration, Bpa-peptide in 150 mM NaCl, 5 mM MgCl2, 5 mM Tris–HCl (pH 8.0), was irradiated with UV light for ten minutes. Efficiency of crosslinking was monitored by SDS-PAGE on a 12% (w/v) polyacrylamide gel stained with Coomassie brilliant blue, or, in the case of biotinyl-VIVIT peptides, by Western blotting using anti-biotin. Trypsin digestion Trypsin digestion was based on a protocol provided by the vendor of RapiGeste SF. Briefly, DTT was added to the CnAa sample to a concentration of 5 mM, RapiGeste SF was suspended as a 0.2% (w/v) stock solution in 50 mM NH4HCO3 and mixed with the CnAa to give a final detergent concentration of 0.1% (w/v), and the sample was heated at 60 8C for 30 minutes. After cooling to room temperature, iodoacetamide was added to a final concentration of 15 mM, and the sample was placed in the dark for 30 minutes to alkylate cysteine sulfhydryl groups. Trypsin was added at a ratio of 1 : 20 (w/w), the sample was incubated at 37 8C for another 60 minutes, and HCl was added to a final concentration of 50 mM to precipitate the detergent. Tryptic peptide purification and mass spectrometry SoftLinke avidin resin (10 ml) was added to 1 ml of 100 mM sodium phosphate (pH 7.0) containing 5 mM biotin and incubated at 4 8C to block irreversible binding sites. The resin was regenerated by washing once with 1 ml of 10% acetic acid, and then three times with 1 ml of 100 mM sodium phosphate (pH 7.0). Digested CnAa was incubated with resin at 4 8C for one hour. The washing procedure was adapted from Girault et al.53 Briefly, after incubation, the supernatant was removed, and the resin

1670

The Docking Interaction of Calcineurin with NFAT

Figure 10. Structural underpinnings of the substrate/regulator recognition sites of calcineurin and PP1. Original coordinates for the two proteins were from PDB entries 1AUI and 1JK7. a, Stereo ribbon diagram of the superposed backbones of calcineurin (red) and PP1 (magenta) in the regions contributing to the peptide binding sites. b Strand 14 at the edge of the upper b sheet and b strand 11 at the edge of the lower b sheet are labelled. The strands depicted in the upper b sheet are b strands 14 and 13, and in the lower b sheet, in sequence, from left to right, b strands 11, 12, 10, 6, and 5. The protein backbones differ appreciably only in the b11-b12 and the b13-b14 loops. b, Linear alignment of calcineurin and PP1 protein sequences in the segments contributing to the peptide binding sites. Vertical lines connect resi˚ in the superposed protein structures. Residues that contribute side-chain dues whose Ca carbon atoms fall within 1 A contacts to the peptide, or both side-chain and backbone contacts, are highlighted in red. Residues in PP1 that provide only backbone carbonyl contacts are highlighted in orange. The contacts on PP1 are as determined by Egloff et al.9 The key role of M290 and C291 in the interaction between PP1 and DARPP-32 has been experimentally confirmed.58 c, Stereo view of the segments of calcineurin corresponding to the linear sequences in b. Only the Ca trace is shown, except for the side-chains of residues that contribute to the PVIVIT-binding site, and the contiguous section of calcineurin backbone in b strand 14 that makes backbone contacts with PVIVIT. The N-terminal residue of each segment is numbered for orientation. For clarity, the upper and lower b sheets have been disassembled and are viewed from slightly different vantage points, indicated by the axes. A view of the site from a single vantage point is available in Supplementary Figure 1. d, Similar stereo view of the segments of PP1 corresponding to the linear sequences in b. The Ca trace is shown, supplemented by side-chains of residues that contribute to the RRVSFA-binding site, and the protein backbone where it contributes to contacts with RRVSFA. Comparison with c makes clear the spatial correspondence between residues constructing the binding site in PP1 and residues constructing the binding site in calcineurin.

Figure 10 a–c (legend opposite)

was washed twice with 100 ml of 50 mM Tris–HCl (pH 7.4), 1 mM DTT, 0.1% BSA (buffer A); sequentially with 200 ml and 100 ml of buffer A supplemented with 0.1% SDS; with 200 ml and 100 ml of buffer A supplemented with 1 M NaCl; and with 200 ml, 100 ml, and 50 ml of water containing 1 mM DTT. The bound tryptic fragments were eluted in 2 ml of the MALDI matrix a-cyano-4-hydroxycinnamic acid and analyzed on a Voyager-DE STR BioSpectrometry Workstation (PE/PerSeptive Biosystem, Foster City, CA). As a control, the trypsin digest of a CnAa sample that had not been crosslinked to VIVIT peptide was prepared by an identical procedure. VIVIT binding assays The calcineurin–peptide interaction was monitored with a fluorescence polarization assay52 using an Analyst

plate reader (LJL Biosystems, Sunnyvale, CA). For direct binding assays, 100 nM fluorescent VIVIT peptide was titrated with increasing amounts of CnAa. In competitive binding assays, increasing amounts of competitor were added to a reaction containing 100 nM fluorescent VIVIT and a fixed concentration of calcineurin. The balance between a robust signal in the fluorescence polarization assay and its sensitivity to competition calls for use of calcineurin at a concentration comparable to the Kd of the calcineurin–fluorescent VIVIT interaction, and hence the following concentrations of protein were used: wild-type CnAa, 1 mM; N330A mutant, 6 mM; and I331A mutant, 60 mM. For the assays with I331A calcineurin, the competing ligands were 14mer VIVIT peptide and the Bpa-substituted VIVIT*10 peptide. For the assays with N330A calcineurin, a 16mer VIVIT peptide (MAGPHP VIVITPGHEE) and its VIVIA mutant (MAGPHPVIVIAP

1671

The Docking Interaction of Calcineurin with NFAT

GHEE) were the competitors. Equilibrium binding data were fitted using two-state or three-state models as described.52

of the incubation, the color was allowed to develop for 30 minutes, and absorption at 650 nm was monitored. In vitro phosphatase assay with NFAT1

In vitro phosphatase assay using non-NFAT substrates In phosphatase assays with standard calcineurin substrates, the concentration of the protein ranged from 200 nM to 5 mM, the reaction was incubated at 30 8C for 30 minutes, and the color was detected by a SpectraMAX 340 PC microplate reader (Molecular Devices, Sunnyvale, CA). Phosphatase assay buffer contained 25 mM Mops (pH 7.0), 5 mM MnCl2, and 0.1 mM CaCl2. When pNPP was used as substrate, it was dissolved in the assay buffer to 20 mM, and 90 ml of pNPP was mixed with 10 ml of CnAa to give the desired final enzyme concentration. The change in color due to the release of p-nitrophenolate anion was monitored at 410 nm. For assays with RII phosphopeptide, the substrate concentration was 0.15 mM. Malachite green (100 ml) was added at the end

NFAT phosphorylated on its physiologically relevant sites was obtained in cell lysates. HEK293 cells were grown in DMEM (Cell-gro) supplemented with 10% (v/v) fetal bovine serum (Omega), 10 mM Hepes (pH 7.4), 100 units/ml of penicillin, 100 mg/ml of streptomycin, 2 mM L-glutamine, and 60 mM b-mercaptoethanol. Cells growing on a 10 cm plate were transfected with NFAT1 expression plasmid using the calcium phosphate method. At 24 hours after transfection, the medium was removed, and the plates were put on ice: 1 ml of lysis buffer (10 mM KCl, 50 mM Tris–HCl (pH 7.4), 0.05% NP-40, 5 mM leupeptin, 10 mg/ml of aprotinin, 1 mM PMSF) was spread evenly on each plate. After five minutes, the lysate was collected and centrifuged in a microcentrifuge for five minutes at 4 8C. The supernatant was frozen immediately in liquid nitrogen and stored at K80 8C.

1672 Stocks of wild-type GST-tagged CnAa and NIR/AAA mutant prepared at 1 mg/ml were diluted to a calcineurin concentration that is within the linear range for both substrates to compare their activities using RII phosphopeptide and pNPP as substrates. In parallel experiments using NFAT as substrate, the same CnAa stocks were serially diluted, and 7 ml of diluted enzyme was combined with 11 ml of HEK293 cell lysate and 2 ml of 10!phosphatase assay buffer, and incubated at 30 8C for 30 minutes. After incubation, SDS sample buffer was added and the samples were boiled for five minutes. The samples were resolved by SDS/8% polyacrylamide gel electrophoresis, transferred to nitrocellulose overnight, and probed with anti-NFAT1 antibody. Docking simulations The coordinates of the calcineurin A chain were taken from the Protein Data Bank (PDB) file of the calcineurin/FKBP/FK506 complex (PDB entry 1TCO). The ligands, 5mer VIVIT peptide and 6mer PVIVIT peptide, were built using the program O. Docking43 was ˚ !60 A ˚ !60 A ˚ performed using AutoDock3.0 in a 60 A ˚ , centered on the 3volume, with grid spacing of 0.375 A methyl group of calcineurin residue M290. The Lamarckian genetic algorithm was utilized, and the number of energy evaluations and population size were set, respectively, at five million and 200, or at ten million and 100. Each simulation carried out a total of 100 runs. The ˚ all-atom docked conformations were clustered with 1 A RMSD (Figure 5), and representatives from the best cluster in each docking run were examined. We adopted the following criteria for assigning van der Waals contacts and hydrogen bonds: participating heavy atoms within ˚ for a hydrogen bond, and within 4.5 A ˚ for van der 3.5 A Waals contact, in the majority of representatives of the cluster. In practice, when the criteria were satisfied, they were met by the vast majority of individual representatives. For example, the N327–Thr11 hydrogen bond was scored as present in 26/27 individual structures in the cluster represented in Figure 7. All torsions except the peptide bonds were unconstrained in the initial simulations with 5mer VIVIT and 6mer PVIVIT. In these simulations, AutoDock consistently preferred an unnatural Cb K Cg torsion in Ile8, presumably because the combined-atom representation of non-polar hydrogen atoms leaves the AutoDock scoring function blind to the eclipsing of H-atom substituents. Since this unnatural Cb K Cg torsion affected the local contacts of Ile8, we repeated the dockings with this single torsion constrained in each of the configurations that are commonly observed in protein structures. AutoDock again gave a well-clustered solution in each case, but the scoring function did not indicate a preference for one conformer over the other. The model presented in Figure 7 shows the conformer with more complete burial of the Ile8 side-chain. In order to examine the relation between the docking sites on calcineurin and PP1, the calcineurin catalytic domain (1AUI) was superposed on the PP1 catalytic subunit (1JK7) using TOP.54 Figures were prepared using RasMol55 and Chimera.56,57

Acknowledgements We thank the Institute of Chemistry and Cell

The Docking Interaction of Calcineurin with NFAT

Biology, Harvard Medical School, for generously providing access to the Analyst plate reader. We are grateful to L. Jin and M. Roehrl for pointing out the crystal contacts on CnAa, to S. Harrison for continuing advice and discussions, and to R. Huey for help with the AutoDock program. This research was supported by grants AI43726 and AI 40127 from the National Institutes of Health and by the Sandler Program for Asthma Research.

Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j. jmb.2004.07.068

References 1. Ceulemans, H. & Bollen, M. (2004). Functional diversity of protein phosphatase-1, a cellular economizer and reset button. Physiol. Rev. 84, 1–39. 2. Millward, T. A., Zolnierowicz, S. & Hemmings, B. A. (1999). Regulation of protein kinase cascades by protein phosphatase 2A. Trends Biochem. Sci. 24, 186–191. 3. Aramburu, J., Rao, A. & Klee, C. B. (2000). Calcineurin: from structure to function. Curr. Top. Cell. Regul. 36, 237–295. 4. Rusnak, F. & Mertz, P. (2000). Calcineurin: form and function. Physiol. Rev. 80, 1483–1521. 5. Janssens, V. & Goris, J. (2001). Protein phosphatase 2A: a highly regulated family of serine/threonine phosphatases implicated in cell growth and signalling. Biochem. J. 353, 417–439. 6. Cohen, P. T. W. (2002). Protein phosphatase 1–targeted in many directions. J. Cell Sci. 115, 241–256. 7. Wakula, P., Beullens, M., Ceulemans, H., Stalmans, W. & Bollen, M. (2003). Degeneracy and function of the ubiquitous RVXF motif that mediates binding to protein phosphatase-1. J. Biol. Chem. 278, 18817–18823. 8. Darman, R. B., Flemmer, A. & Forbush, B. (2001). Modulation of ion transport by direct targeting of protein phosphatase type 1 to the Na-K-Cl cotransporter. J. Biol. Chem. 276, 34359–34362. 9. Egloff, M.-P., Johnson, D. F., Moorhead, G., Cohen, P. T. W., Cohen, P. & Barford, D. (1997). Structural basis for the recognition of regulatory subunits by the catalytic subunit of protein phosphatase 1. EMBO J. 16, 1876–1887. 10. Terrak, M., Kerff, F., Langsetmo, K., Tao, T. & Dominguez, R. (2004). Structural basis of protein phosphatase 1 regulation. Nature, 429, 780–784. 11. Kraus, P. R. & Heitman, J. (2003). Coping with stress: calmodulin and calcineurin in model and pathogenic fungi. Biochem. Biophys. Res. Commun. 311, 1151–1157. 12. Cyert, M. S. (2003). Calcineurin signaling in Saccharomyces cerevisiae: how yeast go crazy in response to stress. Biochem. Biophys. Res. Commun. 311, 1143–1150. 13. Kuhara, A., Inada, H., Katsura, I. & Mori, I. (2002). Negative regulation and gain control of sensory neurons by the C. elegans calcineurin TAX-6. Neuron, 33, 751–763. 14. Sullivan, K. M. C. & Rubin, G. M. (2002). The Ca2Ccalmodulin-activated protein phosphatase calcineurin negatively regulates Egf receptor signaling in Drosophila development. Genetics, 161, 183–193.

1673

The Docking Interaction of Calcineurin with NFAT

15. Crabtree, G. R. & Olson, E. N. (2002). NFAT signaling: choreographing the social lives of cells. Cell, 109, S67–S79. 16. Schiaffino, S. & Serrano, A. (2002). Calcineurin signaling and neural control of skeletal muscle fiber type and size. Trends Pharmacol. Sci. 23, 569–575. 17. Hogan, P. G., Chen, L., Nardone, J. & Rao, A. (2003). Transcriptional regulation by calcium, calcineurin, and NFAT. Genes Dev. 17, 2205–2232. 18. Rao, A., Luo, C. & Hogan, P. G. (1997). Transcription factors of the NFAT family: regulation and function. Annu. Rev. Immunol. 15, 707–747. 19. Serfling, E., Berberich-Siebelt, F., Chuvpilo, S., Jankevics, E., Klein-Hessling, S., Twardzik, T. & Avots, A. (2000). The role of NF-AT transcription factors in T cell activation and differentiation. Biochim. Biophys. Acta, 1498, 1–18. 20. Liu, J., Farmer, J. D., Jr, Lane, W. S., Friedman, J., Weissman, I. & Schreiber, S. L. (1991). Calcineurin is a common target of cyclophilin-cyclosporin A and FKBP-FK506 complexes. Cell, 66, 807–815. 21. Horsley, V. & Pavlath, G. K. (2002). NFAT: ubiquitous regulator of cell differentiation and adaptation. J. Cell Biol. 156, 771–774. 22. Aramburu, J., Garcı´a-Coza´r, F., Raghavan, A., Okamura, H., Rao, A. & Hogan, P. G. (1998). Selective inhibition of NFAT activation by a peptide spanning the calcineurin targeting site of NFAT. Mol. Cell, 1, 627–637. 23. Aramburu, J., Yaffe, M. B., Lo´pez-Rodrı´guez, C., Cantley, L. C., Hogan, P. G. & Rao, A. (1999). Affinity-driven peptide selection of an NFAT inhibitor more selective than cyclosporin A. Science, 285, 2129–2133. 24. Boustany, L. M. & Cyert, M. S. (2002). Calcineurindependent regulation of Crz1p nuclear export requires Msn5p and a conserved calcineurin docking site. Genes Dev. 16, 608–619. 25. Heath, V. L., Shaw, S. L., Roy, S. & Cyert, M. S. (2004). Hph1p and Hph2p, novel components of calcineurinmediated stress responses in Saccharomyces cerevisiae. Eukaryot. Cell, 3, 695–704. 26. Coghlan, V. M., Perrino, B. A., Howard, M., Langeberg, L. K., Hicks, J. B., Gallatin, W. M. & Scott, J. D. (1998). Association of protein kinase A and protein phosphatase 2B with a common anchoring protein. Science, 267, 108–111. 27. Cousin, M. A. & Robinson, P. J. (2001). The dephosphins: dephosphorylation by calcineurin triggers synaptic vesicle endocytosis. Trends Neurosci. 24, 659–665. 28. Frey, N., Richardson, J. A. & Olson, E. N. (2000). Calsarcins, a novel family of sarcomeric calcineurinbinding proteins. Proc. Natl Acad. Sci. USA, 97, 14632–14637. 29. Gorlach, J., Fox, D. S., Cutler, N. S., Cox, G. M., Perfect, J. R. & Heitman, J. (2000). Identification and characterization of a highly conserved calcineurin binding protein, CBP1/calcipressin, in Cryptococcus neoformans. EMBO J. 19, 3618–3629. 30. Hilioti, Z., Gallagher, D. A., Low-Nam, S. T., Ramaswamy, P., Gajer, P. Kingsbury, T. J. et al. (2004). GSK-3 kinases enhance calcineurin signaling by phosphorylation of RCNs. Genes Dev. 18, 35–47. 31. Kingsbury, T. J. & Cunningham, K. W. (2000). A conserved family of calcineurin regulators. Genes Dev. 14, 1595–1604. 32. Lai, M. M., Hong, J. J., Ruggiero, A. M., Burnett, P. E., Slepnev, V. I., De Camilli, P. & Snyder, S. H. (1999). The

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44. 45. 46.

47.

calcineurin-dynamin 1 complex as a calcium sensor for synaptic vesicle endocytosis. J. Biol. Chem. 274, 25963–25966. Liu, J. P., Sim, A. T. & Robinson, P. J. (1994). Calcineurin inhibition of dynamin I GTPase activity coupled to nerve terminal depolarization. Science, 265, 970–973. Rothermel, B., Vega, R. B., Yang, J., Wu, H., BasselDuby, R. & Williams, R. S. (2000). A protein encoded within the Down syndrome critical region is enriched in striated muscles and inhibits calcineurin signaling. J. Biol. Chem. 275, 8719–8725. Ryeom, S., Greenwald, R. J., Sharpe, A. H. & McKeon, F. (2003). The threshold pattern of calcineurindependent gene expression is altered by loss of the endogenous inhibitor calcipressin. Nature Immunol. 4, 874–881. Santana, L. F., Chase, E. G., Votaw, V. S., Nelson, M. T. & Greven, R. (2002). Functional coupling of calcineurin and protein kinase A in mouse ventricular myocytes. J. Physiol. 544, 57–69. Shin, D. W., Pan, Z., Bandyopadhyay, A., Bhat, M. B., Kim, D. H. & Ma, J. (2002). Ca2C-dependent interaction between FKBP12 and calcineurin regulates activity of the Ca2C release channel in skeletal muscle. Biophys. J. 83, 2539–2549. Tavalin, S. J., Colledge, M., Hell, J. W., Langeberg, L. K., Huganir, R. L. & Scott, J. D. (2002). Regulation of GluR1 by the A-kinase anchoring protein 79 (AKAP79) signaling complex shares properties with long-term depression. J. Neurosci. 22, 3044–3051. Kissinger, C. R., Parge, H. E., Knighton, D. R., Lewis, C. T., Pelletier, L. A., Tempczyk, A. et al. (1995). Crystal structures of human calcineurin and the human FKBP12-FK506-calcineurin complex. Nature, 378, 641–644. Griffith, J. P., Kim, J. L., Kim, E. E., Sintchak, M. D., Thomson, J. A., Fitzgibbon, M. J. et al. (1995). X-ray structure of calcineurin inhibited by the immunophilin-immunosuppressant FKBP12-FK506 complex. Cell, 82, 507–522. Rihakova, L., Deraet, M., Auger-Messier, M., Perodin, J., Boucard, A. A., Guillemette, G. et al. (2002). Methionine proximity assay, a novel method for exploring peptide ligand-receptor interaction. J. Recept. Signal Transduct. Res. 22, 297–313. Jin, L. & Harrison, S. C. (2002). Crystal structure of human calcineurin complexed with cyclosporin A and human cyclophilin. Proc. Natl Acad. Sci. USA, 99, 13522–13526. Morris, G. M., Goodsell, D. S., Halliday, R. S., Huey, R., Hart, W. E., Belew, R. K. & Olson, A. J. (1998). Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J. Comput. Chem. 19, 1639–1662. Bhattacharyya, R. & Chakrabarti, P. (2003). Stereospecific interactions of proline residues in protein structures and complexes. J. Mol. Biol. 331, 925–940. Park, S., Uesugi, M. & Verdine, G. L. (2000). A second calcineurin binding site on the NFAT regulatory domain. Proc. Natl Acad. Sci. USA, 97, 7130–7135. Liu, J., Arai, K. & Arai, N. (2001). Inhibition of NFATx activation by an oligopeptide: disrupting the interaction of NFATx with calcineurin. J. Immunol. 167, 2677–2687. Lai, M. M., Burnett, P. E., Wolosker, H., Blackshaw, S.

1674

48.

49.

50.

51.

52.

The Docking Interaction of Calcineurin with NFAT

& Snyder, S. H. (1998). Cain, a novel physiologic protein inhibitor of calcineurin. J. Biol. Chem. 273, 18325–18331. Sun, L., Youn, H. D., Loh, C., Stolow, M., He, W. & Liu, J. O. (1998). Cabin 1, a negative regulator for calcineurin signaling in T lymphocytes. Immunity, 8, 703–711. Miskin, J. E., Abrams, C. C. & Dixon, L. K. (2000). African swine fever virus protein A238L interacts with the cellular phosphatase calcineurin via a binding domain similar to that of NFAT. J. Virol. 74, 9412–9420. Dell’Acqua, M. L., Dodge, K. L., Tavalin, S. J. & Scott, J. D. (2002). Mapping the protein phosphatase-2B anchoring site on AKAP79. Binding and inhibition of phosphatase activity are mediated by residues 315-360. J. Biol. Chem. 277, 48796–48802. Ho, A. M., Jain, J., Rao, A. & Hogan, P. G. (1994). Expression of the transcription factor NFATp in a neuronal cell line and in the murine nervous system. J. Biol. Chem. 269, 28181–28186. Roehrl, M. H. A., Kang, S., Aramburu, J., Wagner, G., Rao, A. & Hogan, P. G. (2004). Selective inhibition of calcineurin-NFAT signaling by blocking protein– protein interaction with small organic molecules. Proc. Natl Acad. Sci. USA, 101, 7554–7559.

53. Girault, S., Chassaing, G., Blais, J. C., Brunot, A. & Bolbach, G. (1996). Coupling of MALDI-TOF mass analysis to the separation of biotinylated peptides by magnetic streptavidin beads. Anal. Chem. 68, 2122–2126. 54. Lu, G. (2000). TOP: a new method for protein structure comparisons and similarity searches. J. Appl. Crystallog. 33, 176–183. 55. Sayle, R. A. & Milner-White, E. J. (1995). RasMol: biomolecular graphics for all. Trends Biochem. Sci. 20, 374. 56. Huang, C. C., Couch, G. S., Pettersen, E. F. & Ferrin, T. E. (1996). Chimera: an extensible molecular modeling application constructed using standard components. Pac. Symp. Biocomput. 1, 724. http://www.cgl.ucsf.edu/chimera 57. Sanner, M. F., Olson, A. J. & Spehner, J. C. (1996). Reduced surface: an efficient way to compute molecular surfaces. Biopolymers, 38, 305–320. 58. Watanabe, T., Huang, H. B., Horiuchi, A., da Cruze Silva, E. F., Hsieh-Wilson, L., Allen, P. B. et al. (2001). Protein phosphatase 1 regulation by inhibitors and targeting subunits. Proc. Natl Acad. Sci. USA, 98, 3080–3085.

Edited by G. von Heijne (Received 7 June 2004; received in revised form 16 July 2004; accepted 20 July 2004)