Report

Reprogramming Control of an Allosteric Signaling Switch Through Modular Recombination

See allHide authors and affiliations

Science  26 Sep 2003:
Vol. 301, Issue 5641, pp. 1904-1908
DOI: 10.1126/science.1085945

Abstract

Many eukaryotic signaling proteins are composed of simple modular binding domains, yet they can display sophisticated behaviors such as allosteric gating and multi-input signal integration, properties essential for complex cellular circuits. To understand how such behavior can emerge from combinations of simple domains, we engineered variants of the actin regulatory protein N-WASP (neuronal Wiskott-Aldrich syndrome protein) in which the “output” domain of N-WASP was recombined with heterologous autoinhibitory “input” domains. Synthetic switch proteins were created with diverse gating behaviors in response to nonphysiological inputs. Thus, this type of modular framework can facilitate the evolution or engineering of cellular signaling circuits.

Cellular behavior is mediated by circuits of interconnected signal transduction proteins. Many of these proteins are allosteric—their catalytic output activity is gated by specific upstream inputs such as ligand binding or covalent modification. Most eukaryotic signaling proteins are composed of modular domains with binding or catalytic functions (1, 2). It has been proposed that domain recombination could facilitate the evolution of proteins with novel signaling functions (14).

Consistent with such a model, complex allosteric gating in some signaling switches is mediated by modular, autoinhibitory interactions (4, 5). For example, the actin regulatory switch N-WASP (6, 7) (Fig. 1A), which displays sophisticated signal integration, contains an output region (“VCA” domain) that in isolation is constitutively active—it stimulates actin polymerization by binding and activating the actin-related protein (Arp) 2/3 complex. However, two modular domains, a highly basic (B) motif and a guanosine 5′-triphosphatase (GTPase)–binding domain (GBD) repress activity through autoinhibitory interactions (8, 9). Two activating stimuli, the phosphoinositide PIP2 and the activated GTPase Cdc42, bind the B and GBD motifs, respectively, and disrupt autoinhibition (9, 10). Because the two inputs act cooperatively, N-WASP approximates an AND gate in which strong activation is only observed in the presence of both inputs (9, 11). Such multi-input regulation is thought to yield precise spatial and temporal control over actin polymerization.

Fig. 1.

Design of synthetic switch gated by heterologous ligand. (A) N-WASP is a modular allosteric switch: its output domain constitutively stimulates Arp2/3-mediated actin polymerization but is repressed by autoinhibitory interactions involving two domains, the GTPase-binding domain (GBD) and a basic (B) motif. Input ligands activate by disrupting autoinhibitory interactions: GTP-loaded Cdc42 binds GBD; PIP2 binds B motif. These two inputs act synergistically (9, 11), thus, N-WASP resembles an AND gate. (B) Design strategy for a synthetic single-input switch using N-WASP's output domain and a PDZ domain-ligand pair as heterologous autoinhibitory module (α-syntrophin PDZ; ligand comprising NH2-GVKESLV-COOH; Kd = 8 μM). (C) Synthetic switch protein is basally repressed but can be activated by addition of exogenous PDZ ligand. We tested switches with an in vitro pyrene-actin polymerization assay (fig. S1), using the time required to reach 50% polymerization (t1/2) as the (relative) activity metric. Basal repression is observed only in constructs containing the intramolecular PDZ domain–ligand pair (fig. S2). Peptide concentration required for half-maximal activation (Kact) is 50 μM. Studies with variant switches show that degree of repression is correlated with affinity of the intramolecular ligand (KPDZ), whereas sensitivity to external PDZ ligand shows an inverse correlation (fig. S3). Assays in this and all other figures were performed with 50 nM switch protein, 5 nM Arp2/3, and 1.3 μM actin (10% pyrene-actin).

We explored the flexibility of such modular regulation by attempting to use domain recombination to reprogram input control of N-WASP. As a simple test of whether modular autoinhibition is interchangeable, we engineered a synthetic signaling switch gated by a single heterologous ligand (Fig. 1B). The design involved tethering an unrelated modular domain-ligand pair—in this case a PDZ domain and its cognate C-terminal peptide ligand—to the termini of the N-WASP output domain. This design would create a potential autoinhibitory interaction that could be relieved by competitive binding of an external PDZ ligand.

Under basal conditions, this synthetic switch was repressed in an in vitro actin polymerization assay (12) (Methods, fig. S1). Repression required an intact, intramolecular autoinhibitory interaction: constructs containing only one interaction partner were not repressed, and addition of saturating free PDZ domain (∼10-fold > Kd) in trans to a construct bearing only the PDZ ligand did not yield repression (fig. S2). The intramolecular PDZ interaction likely locks the output domain in an inactive conformation or restricts dynamic properties required for activity.

The switch was activated by increasing concentrations of free PDZ ligand (Fig. 1C), with maximal activity close to that of the isolated output domain. Half-maximal activation (Kact) required 50 μM input. Precise gating behavior was dependent on the affinity of the autoinhibitory interaction (fig. S3); reducing affinity of the internal ligand resulted in lower basal repression but increased input sensitivity (reduced Kact), as would be expected if the intramolecular PDZ interaction was required for repression.

As in electronic circuits, complex cellular regulation often requires multi-input integrating gates (AND, OR, XOR, etc.) used in combinatorial control or feedback and feed-forward loops (4). We attempted the design of synthetic AND-gate switches by covalently tethering two modular domain-ligand pairs to N-WASP's output domain such that the intramolecular interactions might cooperatively repress activity. Such a switch would respond cooperatively to the combination of both competing external ligands (Fig. 2A). Because of increased complexity of two-input switches, we created a combinatorial library in which switch design parameters including domain type, domain-ligand affinity, linker length, and domain architecture were varied (Fig. 2B). To further increase variability, we used two forms of the N-WASP output domain, long and short; both display constitutive activity (13).

Fig. 2.

Design of synthetic dual-input switch library. (A) Two-input switch design strategy. (B) Switch library constructed by domain recombination. Components used (table S1) are two output domains of N-WASP (output A and B), which differ in length; three different input domains (PDZ, SH3, GBD); from one to three intramolecular ligands of differing affinities for each of the input domains (ligand for GBD is contained within the output domain); and four different-length interdomain linkers (Gly-Ser repeats). Switch architecture and design parameters are listed at left. Component affinities are given in micromolar units. Observed gating behavior is listed at right. Activity of library members was screened in the presence of no inputs, each individual input, and both inputs simultaneously by using a standard set of input concentrations (Cdc42GTP-γ-S: 10 μM; PDZ ligand: 200 μM; SH3 ligand: 10 μM; all of these concentrations are 20 to 100 times the Kd for input binding to its isolated recognition domain). Relative activity (measured as in fig. S1) under these conditions is indicated by a color code (from low to high: black, green, yellow, white). (C) Classes of gating behavior observed in the library (see fig. S4 for class definitions). For linkers n and m, see the scheme at the left; lig., ligand; polym., polymerization.

Two classes of switches were designed. For the first class—“chimeric” switches—the target behavior was dual regulation by PDZ ligand and Cdc42, a nonnative and a native N-WASP regulator, respectively. We constructed these switches using a PDZ domain and the native N-WASP GBD as regulatory modules. The GBD binds a peptide within the N-WASP output region (residues 461 to 479), an interaction that is competitively disrupted by activated Cdc42 (8). Although the intramolecular GBD interaction is required for autoinhibition in native N-WASP, it is not sufficient: the interaction does not repress N-WASP activity unless combined with the autoinhibitory interaction of the B module (the PIP2 responsive element). (9). For the second class—“heterologous” switches—the target behavior was dual regulation by PDZ and SH3 domain ligands, two nonnative inputs. We constructed these switches using the PDZ domain from α-syntrophin and the SH3 domain from Crk. SH3 domains recognize short proline-rich motifs (14, 15).

A library of 34 such switches (Fig. 2B) was tested for gating by the appropriate high-affinity intermolecular ligands. Activity was tested in the presence of no inputs, each individual input, and both inputs together. Like most signaling proteins, these modular allosteric switches did not give simple binary responses; the precise response observed depended on the input concentrations used. We therefore performed activation screens under a standard set of input concentrations: 10 μM Cdc42–GTP-γ-S [guanosine 5′-O-(3′-thiotriphosphate)], 200 μM PDZ ligand, and 10 μM SH3 ligand. Each of these concentrations is 20 to 100 times the Kd observed for input ligand binding to its isolated recognition domain.

Switches could be divided into diverse behavioral classes (Fig. 2C). At the extremes, five switches showed little or no basal repression, and nine were extremely well-repressed, but could not be activated under any of the tested conditions. Most constructs, however, showed some type of gating behavior. Of the remaining 20 switches, 16 showed positive gating (both inputs activate). Two of the proteins displayed antagonistic gating: one input activates, whereas the other represses (mechanism discussed later). The positively gated dual-input switches could be further subdivided. Two proteins showed OR gate–like behavior (roughly equivalent activation in the presence of either individual input or both together), five proteins showed clear AND gate–like behavior, whereas the remaining constructs showed intermediate behavior. Thus, this relatively small library yielded a diversity of switch behaviors, including several with the targeted AND-gate behavior.

Several design principles were revealed by examining how switch parameters alter behavior. Basal repression and input sensitivity were directly linked to the affinity of autoinhibitory interactions. For example, the chimeric switch C11, which has an intramolecular PDZ ligand with Kd = 8 μM, was well repressed under basal conditions but insensitive: It could not be activated by the standard concentrations of PDZ ligand or Cdc42, even in combination (Fig. 3A). However, if the intramolecular PDZ ligand-affinity was reduced (KPDZ =100 μM), the protein then resembled an AND gate (switch C12).

Fig. 3.

Synthetic switches that resemble AND gates. (A) Chimeric switch C12 (right) resembles an AND gate; it shows strong actin polymerization only in the presence of both PDZ ligand and Cdc42. Adjacent bar graph shows maximal polymerization rates under each condition normalized to the basal rate (no input). Related switch (C11) with identical architecture but a higher affinity intramolecular PDZ ligand (left) is insensitive or overrepressed. (B) Heterologous switch H14 (right) resembles an AND gate that responds to SH3 and PDZ ligand. A related switch (H15) with identical architecture but a weaker affinity intramolecular PDZ ligand (left) resembles an OR gate; individual ligands yield relatively strong activation. (C) Switch H14, which resembles an AND gate, can spatially target actin polymerization in a Xenopus oocyte extract. Polystyrene beads were coated with GST fusions to no ligand, SH3 ligand, PDZ ligand, or a tandem SH3-PDZ ligand (see Methods, Supporting Online Material). The tandem ligand was used at half concentration relative to monovalent ligands. Only beads coated with the tandem ligand and incubated with switch H14 (100 μM) nucleated polymerization of rhodamine-labeled actin (red). Merge of bright-field and fluorescence images are shown. Fraction (fract.) of beads displaying actin polymerization (polym.) is given.

Heterologous switch behavior was also dependent on affinity of the autoinhibitory interactions. For example, switch H15, which has internal SH3 and PDZ ligands with KSH3 = 10 μM and KPDZ = 100 μM, resembled an OR gate (Fig. 3B). However, increasing the affinity of the internal PDZ ligand by ∼10-fold (KPDZ = 8 μM) within the same architecture yielded a well-behaved AND gate (switch H14). Interestingly, in one architectural context, the 8 μM PDZ affinity was too high to yield AND-gate behavior (switch C11), whereas in another context this affinity was ideal (switch H14). This difference may be due to differences in the affinity of the partner domain; in C11 the partner domain is the GBD, which binds its internal ligand with Kd = 1 μM (9, 16), whereas in H14 the partner domain is an SH3 domain with KSH3 = ∼10 μM. Maintaining a balance between switch repression and sensitivity may require balancing the affinities of the highly coupled autoinhibitory interactions.

Linker length also affected switch behavior. For example, if the linker length between the PDZ and SH3 domains in H14 was increased from 5 to 20 residues, the switch became more sensitive to the isolated inputs (switch H16), indicative of reduced domain coupling. This finding is consistent with observations that coupling between regulatory domains of Src family kinases depends strongly on conformational and energetic features of the interdomain linker (17). Within this library, however, increasing interdomain linker length did not uniformly reduce coupling, which suggests that these effects are context-dependent.

Synthetic AND-gate switches were tested for targeted activation of actin polymerization in Xenopus oocyte extracts (Fig. 3C). Carboxylated polystyrene beads were coated with glutathione S-transferase (GST) fusions to the relevant input ligands: no ligand (GST alone), SH3 ligand, PDZ ligand, or SH3 and PDZ ligands connected in tandem. When beads were incubated with soluble H14 switch and oocyte extract, actin filament nucleation was observed only on beads coated with the tandem SH3-PDZ ligand, consistent with multi-input targeting.

The combinatorial switch library also yielded switches with the unexpected behavior of antagonistic or negative input control (H1, H2) in which PDZ ligand acted as an activator, but SH3 ligand acted as a repressor (Fig. 4A). Detailed examination of the gating properties of switch H2 in various input concentration regimes revealed that PDZ ligand always acts as an activator; SH3 ligand, however, increased the basal level of repression (Fig. 4B). Antagonistic regulation is consistent with a model in which the intramolecular PDZ interaction is solely responsible for autoinhibition, and the intramolecular SH3 interaction destabilizes the intramolecular PDZ interaction, but, by itself, has no direct effect on output activity (Fig. 4C). We modeled this scheme by assuming that the state in which both intramolecular interactions take place is unfavorable and unpopulated (fig. S5). Such a scheme predicted an activation surface (Fig. 4C) resembling the observed behavior of switch H2 (Fig. 4B). For related switches (H1 to H3), the maximum level of repression observed (in the presence of SH3 ligand), directly correlated with PDZ affinity, a trend consistent with repression driven solely by the intramolecular PDZ interaction.

Fig. 4.

Mechanism of antagonistic switch. (A) PDZ and SH3 ligands have opposing effects on the activity of switch H2; PDZ ligand is an activator, SH3 ligand is a repressor. Fraction (fract.) of beads displaying actin polymerization (polym.) is shown; lig. ligand; rel., relative. (B) Effect of PDZ ligand on switch H2 activity in the presence of different, constant concentrations of SH3 ligand. (C) Antagonistic behavior of switch H2 can be explained by a model in which the SH3 and PDZ intramolecular interactions are anti-cooperative (i.e., the state with both intramolecular interactions is unfavorable and unpopulated), and the intramolecular PDZ interaction mediates autoinhibition. Although the intramolecular SH3 interaction is neutral, it indirectly relieves repression by opposing the intramolecular PDZ interaction. A simple circuit diagram shows how this nested series of regulatory interactions yields antagonistic input control. Positive and negative net effects of inputs are indicated. Modeling of such a switch predicts an activation surface (green) consistent with experimental behavior (fig. S5). (D) Model of switches that resemble AND gates. Both intramolecular domain interactions contribute to autoinhibition. Thus, both ligands are positive regulators. Modeling yields an activation surface (green) consistent with more potent activation in the presence of both ligands simultaneously (fig. S5).

In this type of antagonistic switch, the two domains appear to act in a nested manner: The SH3 intramolecular interaction regulates the PDZ intramolecular interaction negatively, which in turn negatively regulates the output activity (Fig. 4C). Addition of exogenous SH3 ligand, therefore, stabilizes the autoinhibitory PDZ interaction, leading to the observed inhibitory effect. In contrast, in positive integrating switches that resemble AND gates, the two domains work in concert to negatively regulate output function (Fig. 4D). Consequently, disruption of both intramolecular interactions yields activation.

This unanticipated class of switches highlights a striking feature of the library: Subtle changes in switch parameters can lead to dramatic changes in gating behavior. The architecture of antagonistic switches (H1, H2) is identical to a set of positive switches (H7 to H12) except for the size of the output domain (long output in the antagonistic switches; short in the positive switches). The geometry of the output domain must have significant impact on the coupling between regulatory domains, presumably by altering stability of the various conformational states of the switch.

These results demonstrate that multidomain signaling switches like N-WASP are functionally modular; diverse and complex gating behaviors can be generated through relatively simple recombination events between input and output domains, even among domains with no known evolutionary relation. By allowing the establishment of novel regulatory connections between molecules with no previous physiological relation, such recombination events would be a powerful force driving evolution of novel cellular circuitry (18). This interchangeability exists because, in modular allosteric switches, regions that mediate input control are physically separable from output regions. Facile interchange of gating properties is unlikely to occur in conventional allosteric proteins in which input and output activities are centralized in a single folded structure, and gating is mediated by subtle conformational shifts.

Domain recombination space sampled in these experiments proved functionally rich: Although constructs showed a range of different gating behaviors (negative-positive, integrating-nonintegrating, etc.), nearly all of them show some form of gating. Gating as an emergent property, therefore, does not appear to be extremely rare, as might be expected if only very precise domain arrangements yielded regulation. This modular framework, in addition to promoting switch protein evolution, could be used to engineer proteins with novel regulatory control and, in principle, novel cellular circuits.

Supporting Online Material

www.sciencemag.org/cgi/content/full/301/5641/1904/DC1

Materials and Methods

Figs. S1 to S5

Table S1

References and Notes

References and Notes

Stay Connected to Science

Navigate This Article