A functional genetic screen identifies regions at the C-terminal tail and death-domain of death-associated protein kinase that are critical for its proapoptotic activity

Abstract

Death-associated protein kinase (DAP-kinase) is a Ca⁺²/calmodulin-regulated serine/threonine kinase with a multidomain structure that participates in apoptosis induced by a variety of signals. To identify regions in this protein that are critical for its proapoptotic activity, we performed a genetic screen on the basis of functional selection of short DAP-kinase-derived fragments that could protect cells from apoptosis by acting in a dominant-negative manner. We expressed a library of randomly fragmented DAP-kinase cDNA in HeLa cells and treated these cells with IFN-γ to induce apoptosis. Functional cDNA fragments were recovered from cells that survived the selection, and those in the sense orientation were examined further in a secondary screen for their ability to protect cells from DAP-kinase-dependent tumor necrosis factor-α-induced apoptosis. We isolated four biologically active peptides that mapped to the ankyrin repeats, the “linker” region, the death domain, and the C-terminal tail of DAP-kinase. Molecular modeling of the complete death domain provided a structural basis for the function of the death-domain-derived fragment by suggesting that the protective fragment constitutes a distinct substructure. The last fragment, spanning the C-terminal serine-rich tail, defined a new regulatory region. Ectopic expression of the tail peptide (17 amino acids) inhibited the function of DAP-kinase, whereas removal of this region from the complete protein caused enhancement of the killing activity, indicating that the C-terminal tail normally plays a negative regulatory role. Altogether, this unbiased screen highlighted functionally important regions in the protein and revealed an additional level of regulation of DAP-kinase apoptotic function that does not affect the catalytic activity.

Death-associated protein kinase (DAP-kinase) is a positive mediator of apoptosis. It was isolated by a function-based gene-cloning methodology named technical knockout (TKO) selection, which involved expression of an anti-sense cDNA library in cells, followed by selection of clones that survived in the continuous presence of an apoptotic stimulus (1). In this system, specific inhibition of DAP-kinase protein expression by anti-sense mRNA protected HeLa cells from apoptosis induced by IFN-γ (2). Furthermore, DAP-kinase was shown to modulate cell death triggered by Fas, tumor necrosis factor α (TNF-α) (3), and detachment from extracellular matrix (4), indicating its general relevance to apoptosis.

DAP-kinase is a Ca⁺²/calmodulin (CaM)-regulated serine/threonine kinase that is localized to the cytoskeleton, where it associates with the actin microfilament system (5). In addition to the kinase domain, which is essential for the death-promoting effects, the protein carries eight ankyrin repeats, a cytoskeleton-binding region, and a death domain. The multidomain structure of DAP-kinase and its participation in a wide range of apoptotic systems imply that this protein may interact with various intracellular components to exert its action. It is therefore conceivable that different regions of the protein may protect from apoptosis when expressed ectopically, potentially through sequestering those putative interacting proteins. In agreement with this view, transfections with the death-domain module by itself protected cells from various death-inducing signals by specifically neutralizing the function of the endogenous DAP-kinase (3).

In this work, we devised a genetic screen aimed at identifying minimal regions in DAP-kinase protein that are critical for its biological function as a mediator of apoptosis and that can act in a dominant-negative manner when expressed ectopically. We utilized an approach that was based on the previously described genetic-suppressor element methodology, in which single or multiple cDNA clones are used to generate libraries of random short cDNA fragments, cloned in both orientations in a retroviral vector (6–8). Biologically active cDNA fragments are then isolated from cells expressing the library by a positive functional selection for a specific phenotype. This concept has been used previously to identify functional regions in several genes including topoisomerase II (9), kinesin heavy chain (10, 11), and p53 (12, 13). Here we adjusted the screen to the conditions and principles of the TKO selection procedure (1). We generated, in an Epstein—Barr virus (EBV)-based vector, an expression library of random fragments derived from DAP-kinase cDNA, expressed it in HeLa cells, and selected for biologically active peptides that could render cells resistant to apoptosis induced by IFN-γ.

Using this unbiased method, we have isolated four biologically active peptides corresponding to different regions in DAP-kinase. Fragment no. 1 covered a part of the ankyrin repeats, and no. 2 mapped to a “linker” region. Fragment no. 3 mapped to a central core of the death-domain module defining a distinct substructure with functional implications in this domain. Fragment no. 4 spanned the C-terminal tail of the protein and revealed the existence of negative autoregulation exerted by this region. The last two fragments were studied in more detail in this work, because they provided structural and functional information on DAP-kinase protein.

Materials and Methods

Construction of DAP-Kinase cDNA Library.

Full-length DAP-kinase cDNA, a 5-Kb fragment, was excised from a pBluescript II (Stratagene) plasmid and purified. Five micrograms of this cDNA was subjected to partial DNase I (Sigma) digestion as described (8). The reaction was stopped by addition of EDTA to a final concentration of 25 mM at different time points to obtain fragment preparations of various average lengths (between 50 bp and 2 Kb). The vector in which the library was constructed was based on pTKO1 that was previously developed in our laboratory (1). First, a synthetic adaptor containing a Flag epitope, a blunt cloning site, and three stop codons was inserted into pTKO1 under a simian virus-40 promoter, as detailed in Fig. 1. cDNA fragments were blunted with T₄ DNA polymerase and Klenow (NEB, Beverly, MA) and directly ligated with an excess of vector. The cassette was flanked by unique restriction sites on both sides to allow directional resubcloning of the inserts. The amount of DNA used for ligation was adjusted so that more than 100,000 colonies were obtained, a complexity that maximized the potential representation of fragments from different regions of the molecule and at a large variety of sizes. Plasmid DNA was prepared directly from pooled bacterial colonies to avoid an amplification step, and PCR analysis of the resulting library confirmed that the insert size range remained unchanged.

Figure 1.

Isolation of DAP-kinase fragments that confer resistance to IFN-γ-induced apoptosis. (A) Screening strategy. Purified human DAP-kinase cDNA underwent partial DNaseI digestion, and fragments were subcloned into an EBV-derived expression vector to generate a cDNA library of random fragments. This library was introduced into HeLa cells, and elements that conferred resistance to apoptosis were isolated and further analyzed. (B) Vector design for the DAP-kinase-fragmented cDNA library. pTKO1, an EBV derived vector, was modified to accommodate the cDNA library by insertion of the indicated adaptor (for details, see Materials and Methods).

Cell Culture and Transfection Procedures.

293, MCF7, and HeLa cells were maintained in DMEM (GIBCO) with 10% FCS (Bio-Lab, Jerusalem)/2 mM glutamine/100 units/ml penicillin/streptomycin (GIBCO). Human IFN-γ (Rephrogen) was added to the culture media at a 1,000 units/ml concentration. Transfections were performed by the standard CaPO₄ precipitation method. For transient apoptosis assays, 293 cells were plated a day before transfection (6 × 10⁵ cells/9-cm plate). Transfection mix for each plate included 1 μg of enhanced green fluorescent protein (EGFP) plasmid (CLONTECH), 3 μg of the “killing” plasmid (pcDNA3 carrying either p55 TNF receptor or DAP-kinase ΔCaM), and 9 μg of either empty vector or a plasmid carrying the library fragments. To compare killing potency of different forms of DAP-kinase, we used 1 μg pEGFP and 10 μg of pcDNA3 carrying each of the indicated forms. Apoptosis was assessed 24 hr after transfection under a fluorescent microscope, according to morphological features.

Selection of Functional Elements in Cells.

Six plates of HeLa cells (5 × 10⁵/9 cm plate) were each transfected with 10 μg of the library DNA. Forty-eight hours after transfection, cells were removed and plated on 20 plates (6 × 10⁵ cells/15 cm plate) in medium containing 200 μg/ml hygromycin B (Calbiochem) and 1,000 units/ml IFN-γ. After 3 wk of selection, cells from surviving colonies were collected, and episomes were extracted by Hirt's method (1).

Protein Analysis.

Cell pellets were lysed in PLB buffer, separated by SDS/PAGE, and blotted onto nitrocellulose membranes as described (5). Filters were incubated with anti DAP-kinase monoclonal antibodies (Sigma) diluted 1:1,000 and then with horseradish peroxidase-conjugated goat anti mouse antibody (Jackson ImmunoResearch). Antibodies were visualized by enhanced chemiluminescence according to manufacturer's instructions (Supersignal, Pierce).

Molecular Modeling.

The sequence alignment presented in Fig. 4A was obtained by using the Smith–Waterman algorithm (14) and is almost identical to our previously published alignment (15) with the difference of one amino acid shift in the position of helix 3. The three-dimensional (3D) model was constructed by using the homology module of InsightII (MSI/Biosym Technologies, San Diego), on the basis of the NMR structure of the closely related p75 neurotrophin receptor (16), which was recently deposited in the Protein Data Bank (17). The initial model was energy minimized by the encad program (18). In this minimization, the Cα atoms were restrained to their initial positions so that the overall structure was not disrupted. The energy minimized structure was verified by calculating the 3D-one-dimensional compatibility scores (19). These identified only a single potentially misfolded region, confined to the 12-aa loop between helices 3 and 4, which is much shorter in p75, and for which a reliable structure could not be constructed.

Figure 4.

Model of the death domain of DAP-kinase. (A) Sequence alignment of the death domains of DAP-kinase (amino acids 1300–1398) and p75 neurotrophin receptor (amino acids 334–420). Amino acids that are included in the protective fragment (1320–1371) are marked by blue letters. Amino acids that form the six α-helical structures are emphasized in bold letters and brackets. (B) Model structure of the death domain of DAP-kinase (colored ribbons) constructed by comparative modeling and overlaid on the NMR-based structure of the p75 neurotrophin receptor (white ribbon). The six helices (α1 to α6) are accentuated by yellow cylinders. In the DAP-kinase death-domain model, regions that are inside and outside the protective fragment are marked with blue and purple, respectively. Note the extended loops between helices α1 and α2, and α3 and α4 in DAP-kinase compared with p75. (C) Model structure of the death domain of DAP-kinase presented as space-filling spheres. As in B, the protective fragment is colored blue.

In Vitro Kinase Assay.

293 cells were transfected by the CaPo₄ procedure with 1 μg pEGFP and 10 μg of pcDNA3 carrying each of the indicated forms of DAP-kinase, all tagged with a hemagglutinin (HA) epitope at their N terminus. Cells were lysed in PLB as described (5), and immunoprecipitation of recombinant DAP-kinase protein from 1 mg total extract was performed with 3 μl of anti-HA monoclonal antibodies in 500 μl of PLB with protease and phosphatase inhibitors for 2 hr at 4°C. The immunoprecipitates were washed three times with PLB and once with reaction buffer (50 mM Hepes, pH 7.5/8 mM MgCl₂/0.1 mg/ml BSA). The proteins bound to the beads were incubated for 10 min at 30°C in 50 μl of reaction buffer containing 15 μCi [γ-³²P] ATP (3 pmol), 50 μM ATP, 2 μg myosin light chain (MLC) (Sigma), CaCl₂ (0.1 mM) and bovine CaM (1 μM, Sigma). Proteins were analyzed on 12% SDS/PAGE. The gel was blotted onto a nitrocellulose membrane, ³²P-labeled proteins were visualized by autoradiography, and the rate of relative MLC phosphorylation was measured by using a PhosphorImager (Fuji). DAP-kinase levels were determined by immunoblotting.

Results

Functional Dissection of DAP-Kinase by a Genetic Screen.

To isolate biologically active peptides of DAP-kinase, we generated an expression library of its randomly fragmented cDNA and carried out a positive functional selection in HeLa cells treated with IFN-γ (Fig. 1A), the system from which DAP-kinase was originally isolated (2). The library was constructed in pTKO1 (1), an EBV-based episomal vector that carries an IFN-responsive enhancer element that stimulates the expression of the library inserts during selection and was previously shown to be very effective in this procedure (1). Into this vector we first introduced a suitable expression cassette, which provided an initiator methionine in a favorable translation initiation context within a Flag epitope, followed by a cloning site and stop codons in all three reading frames (Fig. 1B). DAP-kinase cDNA fragments were generated by incomplete DNase I digestion and ligated into the vector. Because the fragmentation and subcloning direction were both random, we assumed that half of the fragments were inserted in a sense orientation, and that one-third of these (i.e., about 16% of the total inserts) would be expressed in the correct reading frame. The choice of an episomal shuttle vector provided three major advantages: (i) relatively high efficiency of stable transfection not requiring chromosomal integration; (ii) reduced appearance of false positive clones that could result from insertional mutagenesis; and (iii) direct recovery of plasmids from cells that survived the selection and immediate propagation of these plasmids in either bacterial or mammalian cells.

The DAP-kinase cDNA library was introduced into HeLa cells by transfection, and the cells were then subjected to double selection with hygromycin B and IFN-γ for 3 wk. Cell colonies that survived this prolonged selection were pooled, and episomes were isolated by Hirt's extraction and used to transform bacteria. The cDNA inserts of plasmids from 70 randomly chosen bacterial colonies were amplified by PCR and sequenced. Thirty-one fragments turned out to be inserted in a sense orientation and of the sense fragments, 18 clones encoded peptides in the authentic reading frame of DAP-kinase. Of the 18 clones, 4 fragments appeared only once, and the rest appeared multiple times corresponding altogether to 9 different fragments. Because our aim was to study the function of different structural motifs in the protein, we concentrated on these sense fragments in the correct frame. To distinguish between functionally active and false positive peptides, we tested the biological function of individual fragments as detailed below.

A Secondary Screen Identified Functional Peptides That Inhibit DAP-Kinase Function in TNF-Induced Apoptosis.

The HeLa/IFN-γ system, from which DAP-kinase was isolated, was suitable for the first round of functional selection, because in this system a single genetic change was sufficient to yield a weak yet selectable phenotype of increased survival in a population of cells subjected to a long-term selection (1). However, because this characteristic of the system may also lead to a significant level of nonspecific background, fragments obtained in this selection had to be individually tested by a secondary screen.

For the second assay, we chose apoptosis induced by high levels of p55 TNF receptor, a system in which an essential role for DAP-kinase has been previously established (3). In this assay system, we could rapidly examine the ability of the isolated fragments to inhibit DAP-kinase function and thus protect cells from apoptosis. The p55 TNF receptor was transfected into human kidney epithelial 293 cells together with either an empty pTKO1 vector or a vector carrying individual fragments from the first selection, and the level of apoptosis was assessed 24 hr later. Transfected cells were identified by GFP coexpression, and the rate of apoptosis was scored microscopically according to typical morphological features. A peptide was defined positive if its cointroduction into cells reduced the extent of TNF receptor-induced apoptosis by more than 50%.

By using this criterion, the ability of four of the nine different DAP-kinase-derived peptides to confer resistance to apoptosis was established (Fig. 2). One of them (fragment no. 3) appeared six times, and each of the other three appeared twice among the 18 sense-oriented clones. The remaining fragments were scored as false positive in these assays. The extent of protection by the positive clones ranged between 60 and 70% (Fig. 3A) and resembled the extent of protection obtained by the already established inhibitory fragment of DAP-kinase—the 99-aa-long death domain (Fig. 3A; see also ref. 3). As previously discussed, this partial yet significant extent of protection is characteristic of targets that function at a downstream position along apoptotic pathways. In contrast, and as expected, the dominant-negative mutant of FADD, which functions much earlier in the signal transduction, i.e., at the receptor proximal level (20), completely blocked the p55 TNF receptor-induced apoptosis (Fig. 3A). This screen was first performed with the rescued fragments in the original pTKO1 vector (Fig. 3A) and then repeated after subcloning the fragments into pcDNA3 vector, producing similar results (not shown).

Figure 2.

Protective fragments of DAP-kinase isolated by the genetic screen. Biologically active fragments that passed the two successive screens are listed in the table (Upper). (Lower) Schematic representation of DAP-kinase full-length protein with the position of the library-derived cell death-protective fragments shown underneath.

Figure 3.

DAP-kinase-derived fragments protect cells from apoptosis. (A) Apoptosis was induced in 293 cells by transient overexpression of p55 TNF-receptor. The receptor was expressed together with an empty vector, a dominant-negative mutant of FADD, or the different DAP-kinase fragments, as indicated. Transfected cells were identified by GFP expression, and the rate of cell death was scored according to typical morphological features. The graph represents average values obtained from three independent experiments, each of which included at least 300 GFP-positive cells. (B) 293 cells were induced to undergo apoptosis by transient overexpression of activated DAP-kinase (Δ-CaM). The cells were transfected with a plasmid encoding this mutant, together with either an empty vector or different fragments, as indicated. Transfectants were identified by GFP expression, and apoptosis was scored as in A. Below is an immunoblot containing equal amounts of total cell extracts, which were reacted with anti-DAP-kinase antibodies to compare the levels of exogenous DAP-kinase in the different transfections (the endogenous levels are below detection levels under these exposure conditions). (C) MCF7 cells were transfected with p55 TNF-receptor together with an empty vector, the same vector containing the FADD death domain or different DAP-kinase derived fragments, as indicated. Apoptosis was scored as in A.

The detailed mapping of the functionally active fragments to several regions in the protein is shown in Fig. 2. Three of these fragments span short regions (48–55 amino acids in size) and map to the ankyrin repeats, the “linker region” (which carries no known protein motives), and the death domain. The fourth fragment is a stretch of 17 amino acids at the very C terminus of the protein. The isolation of an element residing within the death domain was notable, because it validated our selection strategy.

Two lines of experiments were carried out to confirm that the fragments acted by specifically inhibiting DAP-kinase function. First, we showed that the different fragments could protect 293 cells from death induced by overexpression of DAP-kinase itself. In this assay, we utilized an activated mutant of DAP-kinase, in which a deletion of the CaM binding and regulatory regions rendered the kinase domain constitutively active, thereby generating a potent death-inducing protein (5). The activated DAP-kinase (designated Δ-CaM) was transfected into 293 cells together with either an empty vector or the same vector carrying the different selected cDNA fragments. As shown in Fig. 3B, expression of the various peptides reduced the extent of apoptosis induced by the activated DAP-kinase, without affecting its protein levels. Because the levels of the ectopically expressed activated DAP-kinase exceeded by far those of the endogenous kinase, it was not surprising that the extent of protection in this experimental setting was lower than in the TNF-based system. In other words, the high levels of the transfected DAP-kinase could not be neutralized by the peptides as efficiently as in the previous experiment (Fig. 3A), where only the endogenous DAP-kinase was present.

To examine further the specificity of these fragments, we tested their function in MCF7 breast carcinoma cells that do not contain DAP-kinase and are sensitive to TNF (3). In this system, cotransfection of DAP-kinase-derived fragments failed to protect the cells from apoptosis induced by transfection with p55 TNF receptors (Fig. 3C). These results indicate that the function of these fragments in protecting cells from apoptosis is DAP-kinase dependent, because in the absence of DAP-kinase these elements lack any antiapoptotic activity.

Obviously, the extent of the screen described here has not been exhaustive, and it is likely that more peptides would be isolated should a more comprehensive screen be performed. Yet, from the four protective fragments that were already isolated, the two that map to the death domain and the tail seemed particularly interesting, and we focused on further analysis of these.

Molecular Modeling of the Death Domain Reveals a Distinct Three-Dimensional Substructure for Fragment No. 3.

Fragment no. 3 was derived from the death domain, a conserved module whose critical role in DAP-kinase-mediated apoptosis has been established (3). To elucidate the structural basis for the activity of the rescued fragment, we constructed a 3D model structure of the death domain. The molecular modeling was based on sequence similarity between DAP-kinase death domain and the closely related intracellular domain of the p75 neurotrophin receptor, for which an NMR structure was recently published (16).

The predicted structure of DAP-kinase death domain consists of 99-aa residues that fold to form six α-helices (Fig. 4B). As in p75, helices 1, 5, and 6 lie parallel to each other and are perpendicular to helices 2, 3, and 4. One notable difference between the two death domains is the presence of longer loops extending between helices 1 and 2 and helices 3 and 4 of DAP-kinase death domain. Because these loops are unique to DAP-kinase, they may contribute to the specificity of interaction with other proteins. The region marked in blue (Fig. 4 A–C) represents fragment no. 3, which was isolated as an effective inhibitor of DAP-kinase. This region, whose exact composition is marked in the sequence alignment (Fig. 4A), completely spans helices 2, 3, and 4, which, as revealed by the molecular model and emphasized in Fig. 4C, are clustered to form a structurally distinct element. This substructure is stabilized by hydrophobic interactions, particularly between helices 2 and 4, which are likely to direct correct folding of the peptide even when expressed separately and hence allow binding of this distinct module to its putative target protein(s). Of note, the addition of an N-terminal Flag epitope (see Fig. 1) is not expected to interfere with the folding of the peptide, first because the Flag tag is composed mostly of hydrophilic residues, and second because it is connected to a 9-aa-long stretch of the loop between helices 1 and 2, which physically separates the tag from the core of the module.

Thus, according to this model structure, fragment no. 3 constitutes a defined substructure in the death domain, a feature that may underlie its effective biological function.

An Autoinhibitory Function of DAP-Kinase C-Terminal Tail.

Fragment no. 4 was particularly interesting. It mapped to the very last 17 amino acids in the short C-terminal region that immediately follows the death domain. The amino acid sequence of this peptide (SCNSGTSYNSISSVVSR) is rich in serines, a feature that is typical for many death-domain-containing proteins (15). In one case, the Fas receptor, it has been shown that the C-terminal tail negatively regulates signal transduction (21). The fact that ectopic expression of the C-terminal tail could inhibit DAP-kinase function indicated that the tail could have a defined regulatory role in the context of the complete molecule as well. However, we could not determine a priori whether the tail is normally essential for activation of DAP-kinase, execution of its function, or negative autoregulation.

To explore further the nature of regulation that is mediated by the C-terminal tail, we deleted it from the full-length DAP-kinase and tested the effect of this deletion on the rate of apoptosis induced by the protein. In the case of a positive regulation by the tail or of a direct role in execution of the protein's function, a deletion is expected to cause a reduction in apoptotic activity. Conversely, if the tail is involved in negative regulation of DAP-kinase, then a deletion should result in potentiation of the death-promoting function of the protein. We generated a tail-truncated (Δ-tail) version of DAP-kinase by introducing a stop codon at position 1415, thus removing the last 17 amino acids of the protein. Quantitative apoptosis assays were then used to compare the potency of this deletion mutant to that of the wild-type (wt) protein. The constitutively active kinase mutant (Δ-CaM) was used in these assays as a control for gain of function. It was found that truncation of the last 17 amino acids greatly potentiated the ability of DAP-kinase to induce apoptosis in 293 cells (Fig. 5A). The extent of apoptosis induced by the truncated mutant was significantly higher than that induced by the wt protein (58% vs. 25%) and comparable to the ΔCaM mutant (63%), whereas expression levels were equal for all forms (Fig. 5B). To test whether this enhancement of killing potency was because of an increase in the catalytic activity of the protein, we performed an in vitro kinase assay, in which we immunoprecipitated recombinant proteins from transfected cells and incubated them with MLC as an external substrate (5). In this assay, the truncated Δ-tail mutant displayed kinase activity that was indistinguishable from that of the wt protein, whereas the ΔCaM mutant, carrying an activated kinase domain, clearly showed an increased kinase activity (Fig. 5C). Thus, the enhancement of killing potency could not be attributed to a higher kinase activity. These results imply that the tail does not directly affect the kinase domain but rather acts at another level of controlling the proapoptotic activity of DAP-kinase, conceivably through modulating interactions of other domains in the protein with critical targets.

Figure 5.

Deletion of the last 17 amino acids of DAP-kinase potentiates its activity. (A) 293 cells were transfected with either an empty vector or the same vector carrying the indicated forms of DAP-kinase. WT, wild-type protein; Δ-tail, a truncated mutant lacking the 17 C-terminal amino acids; ΔCaM, a deletion mutant lacking the CaM regulatory region. Transfected cells were identified by GFP coexpression, and the rate of cell death was assessed 24 hr posttransfection, according to typical apoptotic morphology. The graph represents values from three independent experiments, each including at least 300 GFP-positive cells. (B) Protein extracts were prepared from the transfected 293 cells (see A), and DAP-kinase protein levels were analyzed by Western blotting by using specific monoclonal antibodies. The endogenous DAP-kinase was below detection levels under these exposure conditions. (C) An in vitro kinase assay was performed with proteins immunoprecipitated from 293 cells transfected with the indicated plasmids. The assay included an exogenous substrate, MLC, and the relative activity was determined according to the rate of MLC phosphorylation, as quantified by using a PhosphorImager.

Discussion

Identification of functional regions in complex proteins is key to understanding their mechanism of action and is commonly directed by conserved structural motifs. We have analyzed the effect of expressing distinct domains of DAP-kinase and found that overexpression of the entire death domain could protect cells from apoptosis induced by the complete DAP-kinase (3) and that the linker region could protect as well, whereas overexpressing the kinase domain had no effect (unpublished data). Here, we have utilized a random unbiased approach to identify minimal regions in DAP-kinase that are critical for its ability to participate in apoptotic processes. Functional domains of DAP-kinase were identified by expression and selection of dominant-negative peptides that inhibited the activity of the protein and thus prevented cells from undergoing apoptosis. Altogether, four functional fragments were identified, two of which (fragments nos. 3 and 4) were studied here in more detail. The possible unique function of the selected segments from the region encompassing the ankyrin repeats (no. 1) and the unidentified yet “linker” (no. 2) should be studied further in the future. The mere rescue of these biologically active fragments highlighted critical functional domains of DAP-kinase and defined minimal portions of the molecule that are capable of interfering with its death-inducing activities when artificially expressed in cells. When tested in MCF7 cells that lack endogenous DAP-kinase, these fragments did not inhibit apoptosis, indicating that their function is DAP-kinase dependent. Ectopic expression of each of these elements could protect 293 cells from TNF-receptor-induced apoptosis to approximately the same extent, suggesting that each of these domains is necessary for mediating the apoptotic effects. These data are consistent with a model according to which several independent regions are simultaneously involved in executing the proapoptotic function of DAP-kinase.

Interestingly, fragment no. 3, which appeared as several independent copies in the selection, resides at the core of the death-domain module. We have previously shown that the death domain was essential for DAP-kinase function, because deletion of this entire domain from the full-length protein reduced its ability to kill cells without affecting the kinase activity (3). Death domains in general either homodimerize or heterodimerize with other death-domain-containing proteins. The death domain of DAP-kinase did not homodimerize, nor was it capable of interacting with some of the known death domains, which suggested that it may interact with yet unidentified proteins (3). The isolation of fragment no. 3 in this screen attributed a functional role to a precise substructure in this domain. The molecular modeling of the death domain of DAP-kinase shown here suggested that it consists of six α-helical structures, and that fragment no. 3 encompasses the central three, implying a critical role for these helices in mediating protein–protein interactions. According to the model structure, the position of helices 2, 3, and 4 with respect to each other is stabilized by hydrophobic interactions among them. This observation lends support to the possibility that this 3D structure may also form when this fragment is expressed separately, enabling it to recognize and physically associate with proteins that normally bind DAP-kinase death domain. This hypothesis is reinforced by the specific biological activity of this fragment and is consistent with mutational analysis and recent molecular modeling of the death domains of Fas and FADD, which proposed that helices 2 and 3 from the two proteins are directly involved in forming the contact between them (22, 23).

Besides the death domain, the other three elements that were isolated by this screen map to regions in DAP-kinase that had not been shown previously to be necessary for the induction of cell death. Among those, the major emphasis in this work was put on the C-terminal tail. Because DAP-kinase is expressed in growing cells, its ability to induce cell death should be tightly regulated and activated only in response to apoptotic triggers. One level of regulation relates to the catalytic activity that is enhanced by binding of Ca⁺²/CaM (5). A second regulatory mechanism was revealed in this study and engages the C-terminal amino acid tail that immediately follows the death domain. Fragment no. 4, consisting of the very last 17 amino acids of DAP-kinase, can inhibit in trans the function of the complete protein. The biological activity of this tail could hypothetically be modeled in one of two ways: (i) The tail is normally required for execution of DAP-kinase function by interaction with another protein, hence excess of the peptide may sequester a cellular downstream effector, or (ii) the tail is involved in negative regulation of DAP-kinase, probably through intramolecular folding that prevents interaction of other regions in DAP-kinase with downstream effectors, and therefore excess of the peptide can inhibit the effect of DAP-kinase on its targets. The actual mechanism was revealed when we deleted the last 17 amino acids of the protein and found that this truncation greatly potentiated the death-inducing potency of DAP-kinase when overexpressed in cells without affecting its catalytic activity. These results are consistent with the second proposed model, according to which the tail normally serves an autoinhibitory function.

Remarkably, the presence of such serine-rich tails, although not their exact sequence, is conserved among other death domain-containing proteins (15). A similar regulatory function was previously described for Fas, where deletion of 15 amino acids from the C terminus enhanced the killing activity of the receptor (21). More recent studies proposed at least two mechanisms by which the C-terminal tail of Fas may regulate its signal transduction. One observation was that deletion of the tail enhanced the interaction between the death domains of Fas and FADD (24). In addition, a protein phosphatase capable of binding the Fas C terminus has been isolated (25). Although the exact role that Fas-associated phosphatase plays in signal transduction is unclear (26, 27), its association with Fas inhibited death induction in certain cell systems. In this latter case, excess of peptide, which titrated out the inhibitor, stimulated rather than inhibited apoptosis, a scenario that does not apply for DAP-kinase tail. Notably, the apparent difference in electrophoretic mobility between the wt and the tail-deleted proteins is larger than predicted by size differences alone, suggesting that the tail may undergo a posttranslational modification, perhaps phosphorylation, to regulate its activity.

The study described here identified functional regions in DAP-kinase and established the tools to perform a more comprehensive screen. In addition to defining key regions in DAP-kinase, the peptides that have already been identified can serve as a basis for isolation of proteins that interact with these regions in DAP-kinase and for generating specific low molecular-mass inhibitors for this protein, which can be applied to modulate apoptotic processes.

Abbreviations

DAP-kinase

death-associated protein kinase

TNF-α

tumor necrosis factor α

GFP

green fluorescent protein

TKO

technical knockout

EBV

Epstein–Barr virus

CaM

calmodulin

MLC

myosin light chain

3D

three-dimensional

wt

wild type