Report

Affinity-Driven Peptide Selection of an NFAT Inhibitor More Selective Than Cyclosporin A

See allHide authors and affiliations

Science  24 Sep 1999:
Vol. 285, Issue 5436, pp. 2129-2133
DOI: 10.1126/science.285.5436.2129

Abstract

The flow of information from calcium-mobilizing receptors to nuclear factor of activated T cells (NFAT)–dependent genes is critically dependent on interaction between the phosphatase calcineurin and the transcription factor NFAT. A high-affinity calcineurin-binding peptide was selected from combinatorial peptide libraries based on the calcineurin docking motif of NFAT. This peptide potently inhibited NFAT activation and NFAT-dependent expression of endogenous cytokine genes in T cells, without affecting the expression of other cytokines that require calcineurin but not NFAT. Substitution of the optimized peptide sequence into the natural calcineurin docking site increased the calcineurin responsiveness of NFAT. Compounds that interfere selectively with the calcineurin-NFAT interaction without affecting calcineurin phosphatase activity may be useful as therapeutic agents that are less toxic than current drugs.

Transcription factors of the NFAT family regulate immune responses as well as adaptive responses in heart and skeletal muscle (1–3). Four of the five NFAT proteins (NFAT1/p, NFAT2/c, NFAT3, and NFAT4/x) are cytoplasmic and are activated by stimulation of cell surface receptors coupled to Ca2+ mobilization (1). The Ca2+-activated phosphatase calcineurin dephosphorylates these NFAT proteins, promoting their nuclear translocation and activation (1, 4). Calcineurin docks at a site in the conserved NFAT regulatory domain that has the consensus sequence PxIxIT (5, 6) (Fig. 1A). Interfering with docking of calcineurin at the PxIxIT sequence impairs NFAT activation and NFAT-dependent reporter gene expression (5).

Figure 1

Evolution of an optimized peptide that inhibits the NFAT-calcineurin interaction. (A) Calcineurin docking sequences present in NFAT family proteins. (B) First round of selection. A combinatorial peptide library anchored by the sequence PxIxIT from NFAT was selected by binding to GST-calcineurin (residues 2 through 347) (8). Positions fixed in the first (B) and second (C) degenerate peptide libraries are shown in the single-letter amino acid code, and randomized positions are indicated by X. Within the general library sequence, each X position contains roughly equimolar amounts of all amino acids except cysteine. Boxed residues are those conserved in all NFAT proteins. After extensive washing, the bound peptides were eluted and sequenced, and amino acids within each sequencing cycle were normalized to their abundance in the original library mixture. Particular amino acids selected in the degenerate positions are shown with preference values indicated in parentheses. Residues showing strong selection are shown bold and underlined. (C) Second round of selection. An alternative set of residues was chosen based on the initial screen (B) to orient a secondary library, and the second library was selected on GST-calcineurin (residues 2 through 347) to derive high-affinity peptides. Residues that were locked in based on the screening in (B) are boxed. Z indicates a member of a set of nonnatural amino acids (8). This screen revealed extremely strong selection for particular amino acids within the sequence, resulting in the optimal peptide VIVIT. (D) Sequence of the VIVIT peptide used in subsequent experiments. mer, oligomer.

To develop high-affinity NFAT inhibitors based on the PxIxIT sequence, we constructed combinatorial peptide libraries (7,8) (Fig. 1). The first library, with the sequence MAxxxPxIxITxxHKK (where x represents a mixture of natural amino acid residues) was randomized in seven residues not fully conserved within the NFAT family (Fig. 1B). Peptides were selected for their ability to bind a glutathione S-transferase (GST) fusion protein containing the calcineurin catalytic domain (8). The peptide pool eluted from the calcineurin column showed moderate selection for glycine, serine, and lysine at position 3; no preferred residues at position 4; histidine or aliphatic residues at position 5; and moderate selection for polar residues (threonine, lysine, glutamine, and glutamic acid) at position 7 (Fig. 1B). Position 9 showed weak selection for aliphatic residues, notably valine. Glycine and proline were selected at positions 12 and 13, which suggests that the NFAT binding site in calcineurin imposes a turn at the COOH-terminal end of the PxIxIT motif.

To refine the peptide selection further, we used an affinity-driven peptide evolution technique (7). A second degenerate peptide library, MAGxHP[T/x][z/x]xIxGPHEE (where z represents a set of nonnatural residues) was synthesized, locking in some of the residues selected in the first screening while randomizing other positions that had been previously fixed (8) (Fig. 1C). Positions 3 and 5 were fixed as glycine and histidine; positions 12 and 13 were fixed as glycine and proline to favor the putative turn. Position 7 was biased to contain 50% threonine and 50% all other residues. Position 8 contained natural amino acid residues and five additional nonnatural amino acid residues with large aromatic or cyclic groups (8), to determine whether binding affinity could be improved by substituting a large hydrophobic side chain for the isoleucine side chain naturally conserved at this position in NFAT proteins. Position 10 was fixed as isoleucine, positions 9 and 11 were randomized, and the two COOH-terminal lysines were replaced by glutamates to prevent any bias produced by their positive charge.

Screening with the second library yielded strongly preferred residues at most of the randomized positions (Fig. 1C). The polar residues that occur naturally at the variable residues of the PxIxIT sequence in NFAT1-4 (R/S at position 7; E/R/Q at position 9) were not highly selected; rather, bulky or β-branched hydrophobic residues (valine, isoleucine, and leucine) were preferred. Proline was preferred at position 4, echoing its occurrence at this position of NFAT1 (but not NFAT2-4) (Fig. 1A). Isoleucine was stringently preferred at position 8, with lesser selection for other hydrophobic amino acids and no selection for the nonnatural amino acids, which is consistent with the invariance of isoleucine at this position of the PxIxIT sequence in all four NFAT proteins (Fig. 1A). Finally, there was strong selection for the conserved threonine at position 11 of the PxIxIT sequence, with a weaker preference for serine.

We synthesized the predicted optimal peptide MAGPHPVIVITGPHEE (6) (Fig. 1D) and examined its effect on the interaction of calcineurin with NFAT (9) (Fig. 2). This peptide (referred to hereafter as VIVIT) was about 25 times more effective than the original SPRIEIT peptide (5) at inhibiting the binding of activated calcineurin to GST-NFAT1 (Fig. 2A). The VIVIT peptide was also superior at inhibiting calcineurin-mediated dephosphorylation of NFAT1, NFAT2, and NFAT4 in cell extracts (Fig. 2, B and C). When expressed as a fusion protein with green fluorescent protein (GFP), the VIVIT peptide efficiently inhibited calcineurin-dependent nuclear translocation of NFAT1 (10) as well as activation of an NFAT-AP-1 reporter by endogenous or overexpressed NFAT1, NFAT2, and NFAT4 (11) (Fig. 2D). Thus, iterative peptide selection based on calcineurin binding yielded a highly inhibitory peptide, capable of disrupting all aspects of NFAT activation by calcineurin much more effectively than peptides spanning the natural calcineurin docking sequences of NFAT.

Figure 2

The VIVIT peptide is a potent inhibitor of the NFAT-calcineurin interaction, and its substitution into the calcineurin docking site enhances the calcineurin responsiveness of NFAT1. (A) Inhibition of the NFAT-calcineurin interaction (9). Calcineurin (Cn) was activated with calmodulin (CaM) and CaCl2 (Ca2+), and its binding to GST (lane 1) and GST-NFAT1 (residues 1 through 415) (lanes 2 through 11) was evaluated by protein immunoblotting. Cn A, calcineurin A chain. (B and C) Inhibition of the calcineurin-mediated dephosphorylation of NFAT proteins (9). Lysates of HeLa cells expressing HA-NFAT1 (B) or lysates from HEK 293T cells expressing HA-NFAT1, HA-NFAT2, or HA-NFAT4 (C) were incubated with the phosphatase inhibitor sodium pyrophosphate (NaPPi, lane 1) or with activated calcineurin (Cn+CaM+Ca2+) in the absence or presence of peptides at the indicated micromolar concentrations. The phosphorylation status of NFAT proteins was evaluated by protein immunoblotting with anti-HA. The positions of phospho- and dephospho-NFAT are indicated by arrows in (B). (D) Inhibition of NFAT-dependent gene expression. (Left panel) Jurkat cells were cotransfected with a 3xNFAT-Luc reporter plasmid and with expression plasmids encoding murine NFAT1, GFP, GFP-SPRIEIT, or GFP-VIVIT as indicated (13). (Right panel) Jurkat cells were cotransfected with a 3xNFAT-Luc reporter plasmid and with expression plasmids encoding GFP, GFP-VIVIT, and murine NFAT1, human NFAT2, and human NFAT4 as indicated (11). Twenty-four hours after transfection, luciferase activity induced by endogenous NFAT (Endog.) or by overexpressed NFAT proteins was measured in unstimulated cells and in cells stimulated for 6 hours with PMA and ionomycin. (E) Substitution of the VIVIT sequence into NFAT1 (12). Cl.7W2 murine T cells were transfected with wild-type HA-NFAT1-GFP or with the mutant HA-NFAT1[VIVIT]-GFP, in which HPVIVITGP replaces SPRIEITPS. Cells were stimulated with ionomycin (Iono) at the concentrations indicated in the absence or presence of 1 μM CsA, and the phosphorylation status of NFAT1 was assessed by protein immunoblotting with anti-HA. (F) Localization of NFAT1 and NFAT1[VIVIT] in cells (12). HeLa cells transiently expressing wild-type HA-NFAT1-GFP or HA-NFAT1[VIVIT]-GFP were left untreated or were treated with 10 μM CsA (for 16 hours). NFAT1 proteins were visualized in fixed cells by GFP fluorescence.

We asked whether substituting the high-affinity VIVIT sequence into a wild-type NFAT protein would increase its responsiveness to calcineurin. Compared to wild-type NFAT1, a mutant NFAT1[VIVIT] protein with the sequence SPRIEITPS replaced by HPVIVITGP (12) was significantly dephosphorylated even in resting cells and required lower concentrations of ionomycin to be fully dephosphorylated (Fig. 2E). Dephosphorylation of NFAT1[VIVIT] was calcineurin-dependent because it was blocked by cyclosporin A (CsA). Further, whereas wild-type NFAT1 was cytoplasmic in resting cells (Fig. 2F, left), NFAT1[VIVIT] consistently showed partial nuclear accumulation, which was prevented by CsA (Fig. 2F, right).

Inhibition of NFAT activation by the VIVIT peptide did not reflect inhibition at the calcineurin active site, because the peptide, at a concentration (100 μM) that effectively inhibited NFAT1-calcineurin binding and NFAT1 dephosphorylation (Fig. 2, A through C), did not inhibit calcineurin phosphatase activity toward the RII phosphopeptide (Fig. 3A). In the same experiment, CsA–cyclophilin A complexes (10 μM) inhibited calcineurin phosphatase activity by ∼95% (Fig. 3A). Consistent with this observation, expression of the GFP-VIVIT fusion protein inhibited activation of an NFAT reporter but not of an NF-κB reporter (Fig. 3B), although both reporters were equivalently sensitive to inhibition of calcineurin with CsA (Fig. 3C). Thus, the VIVIT peptide selectively inhibits NFAT activation without disrupting other calcineurin-dependent pathways. GFP-VIVIT, but not GFP, inhibited reporter gene expression driven by the interleukin-2 (IL-2) and tumor necrosis factor α (TNF-α) promoters in T cells stimulated with phorbol ester (PMA) plus ionomycin or with antibody to CD3 (anti-CD3) plus anti-CD28 (Fig. 3D).

Figure 3

The VIVIT peptide selectively inhibits NFAT activation but not calcineurin activity. (A) The VIVIT peptide did not inhibit calcineurin phosphatase activity, assayed as radiolabel released (counts per minute × 10−3) from32P-phospho-RII peptide (9). The numbers next to each peptide label indicate peptide concentrations (micromolar). CsA/CypA complexes were used at 10 μM. (B) Selective inhibition of NFAT reporter activity (11). Jurkat cells were cotransfected with 3xNFAT-Luc (left panel) or 2xNF-κB-Luc (right panel) reporter plasmid, and with GFP and GFP-VIVIT expression plasmids as indicated (measured in micrograms of plasmid per 106cells). Twenty-four hours after transfection, cells were left untreated (open bars) or were stimulated for 6 hours with PMA and ionomycin (solid bars). (C) Calcineurin dependence of NFAT and NF-κB reporter activity in T cells (11). Jurkat cells were transfected with 3xNFAT-Luc (left panel) or 2xNF-κB-Luc (right panel) reporter plasmid. Twenty-four hours after transfection, cells were left unstimulated or were stimulated for 6 hours with PMA and ionomycin (P+I) in the absence or presence of CsA. (D) Inhibition of NFAT-dependent activation of the IL-2 and TNF-α promoters (11). Jurkat cells were cotransfected with GFP or GFP-VIVIT expression plasmids and with luciferase reporter plasmids driven either by the human IL-2 promoter (left panel) or by the human TNF-α promoter (right panel). Twenty-four hours after transfection, cells were left unstimulated (open bars) or were stimulated for 6 hours with PMA and ionomycin (solid bars) or with anti-CD3 and anti-CD28 (hatched bars).

We tested the ability of the VIVIT peptide to inhibit expression of endogenous NFAT-dependent genes (Fig. 4). Jurkat cells highly enriched for expression of GFP-VIVIT or GFP (13) were stimulated and analyzed for cytokine expression by ribonuclease (RNase) protection assay. GFP-VIVIT inhibited the inducible expression of IL-2, IL-13, IL-3, TNF-α, granulocyte-macrophage colony-stimulating factor (GM-CSF) and macrophage inflammatory protein 1α (MIP-1α) (Fig. 4, A and B), thus establishing these genes as NFAT-dependent genes and confirming earlier reporter assays indicating the presence of functional NFAT sites in their promoter-enhancer regions (1). In contrast, GFP-VIVIT did not affect CsA-sensitive expression of TNF-β and lymphotoxin-β (LT-β) (Fig. 4C), establishing that there are CsA-sensitive (presumably calcineurin-dependent) genes that are not controlled by NFAT.

Figure 4

The VIVIT peptide distinguishes NFAT- dependent and NFAT-independent categories of cyclosporin-sensitive genes. The expression of cytokine mRNAs by Jurkat T cells expressing mCD4 and GFP or GFP-VIVIT as indicated is shown (13). Cells were left unstimulated (dashes) or were stimulated for 3 hours with PMA and ionomycin (P+I) in the presence or absence of CsA, and levels of cytokine mRNAs were analyzed by RNase protection assay. RNA loading is indicated by the intensity of housekeeping transcripts L32 and GAPDH. (A and B) NFAT-dependent expression of IL-2, IL-13, IL-3, TNF-α, GM-CSF, and MIP-1α mRNAs. Autoradiogram exposure times were 24 hours (A) and 12 hours (B) for the upper panel and 4 hours (A) and 2 hours (B) for the lower panel. (C) NFAT-independent but CsA-sensitive expression of TNF-β and LT-β mRNAs. Exposure times were 36 hours for the upper panel and 12 hours for the lower panel.

Our studies extend the range of experimental approaches for probing NFAT function in vivo. In addition to their established role in the immune response (1), NFAT and calcineurin have been implicated in cardiac and skeletal muscle hypertrophy (2,3), in slow fiber differentiation in skeletal muscle (3), in cardiac valve development (14), and in differentiation of a preadipocyte cell line to adipocytes in culture (15). These conclusions have relied on analysis of mouse models, identification of plausible NFAT sites in gene regulatory regions, expression of modified NFATs and calcineurins, and the use of CsA and FK506. The VIVIT peptide constitutes a highly selective inhibitor of NFAT, which can now be used for direct identification of NFAT target genes in these various cell types.

Substitution of the VIVIT sequence, a high-affinity calcineurin docking site, into wild-type NFAT1 causes it to be dephosphorylated and activated even in resting cells, which have low basal calcineurin activity. Evidently evolution has selected for an NFAT-calcineurin interaction of low to moderate affinity that precludes NFAT activation in resting cells. Our results call attention to the general point that many protein-protein interactions, especially those involving enzyme-substrate interactions or transient docking interactions (16), may be constrained to a range of low or moderate affinities in order to facilitate information transfer from one intracellular location to another, to ensure reversibility, and to prevent inappropriate activation at subthreshold levels of stimulus. The protein-protein interfaces involved in these reversible interactions would lend themselves to the design of small peptide or nonpeptide inhibitors.

The immunosuppressants CsA and FK506, used clinically to prevent transplant rejection, inhibit the phosphatase activity of calcineurin toward all its protein substrates, including NFAT (1). Although these drugs have revolutionized transplant therapy, their use is associated with progressive loss of renal function, hypertension, neurotoxicity, and increased risk of malignancy (17). It is not yet clear to what extent these toxicities are due to inhibition of NFAT, to interference with dephosphorylation of other calcineurin substrates, or to a potentially non–calcineurin-dependent up-regulation of TGF-β. Selective NFAT inhibitors will allow us to address these questions directly. NFAT inhibitors with less toxicity than CsA and FK506 could be useful in treating chronic ailments such as myocardial hypertrophy, allergy, arthritis, and autoimmune disease.

  • * To whom correspondence should be addressed. E-mail: arao{at}cbr.med.harvard.edu (A.R.); hogan{at}cbr.med.harvard.edu(P.G.H.)

REFERENCES AND NOTES

View Abstract

Navigate This Article