Designing orthogonal signaling pathways: How to fit in with the surroundings

Protein–protein interactions serve as the primary means for transducing signals within cells and between cells and thus are responsible for controlling cell proliferation, morphology, differentiation, and survival. Accordingly, the ability to rationally manipulate these interactions will represent a key enabling technology for understanding the properties of natural cellular networks, for building new biological responses into cells, and even as a starting point for therapeutic intervention in certain human diseases. Successful integration into cells will require that engineered protein interactions (and later whole engineered pathways) can exist in parallel with endogenous cellular interactions (and pathways) without interference or cross-talk between the two: they must behave in a manner completely orthogonal to one another. A study in PNAS by Kapp et al. (1) reports the computational redesign of a GTPase/guanine nucleotide exchange factor (GEF) protein pair, then goes on to demonstrate the orthogonality of this pair in living cells.

Certain applications, such as introducing an enzyme/inhibitor cognate pair, may not require “wiring” the new interaction into the underlying cellular circuitry. If a functionally similar protein pair exists in the host cell, extensive differences between the artificial pair and the endogenous pair may naturally lead to orthogonality. A new protein pair that is highly dissimilar to endogenous cognate pairs can be transferred from a separate organism (2, 3), produced by extensive rational redesign of the interface (4), or even designed completely from scratch (5). Because of the divergence of the resulting protein pair from existing nodes in the cellular interaction network, however, none of these approaches are likely to allow subsequent incorporation of this artificial pair into existing cellular signaling pathways. This in turn may limit the breadth of phenotypic outputs that can be modulated using the artificial protein pair.

Given that highly connected protein nodes with multiple regulatory interaction partners will provide the most flexibility when reengineering signal transduction pathways, how can we incorporate into cells new orthogonal pairs of “hub” proteins? In addressing this problem, Kapp et al. (1) face several challenging design criteria (Fig. 1), each of which is addressed in this study either explicitly or implicitly.

The first challenge is ensuring GTPase connectivity to existing cellular machinery: this cannot necessarily be expected if a GTPase is borrowed from an unrelated organism. The authors address this by using an endogenous GTPase as a starting point for redesign, then allowing in the redesign process only amino acid substitutions predicted not to disrupt interactions with other known partners. This strong restriction leads to a tricky problem of building orthogonality into the starting GTPase/GEF pair, which the authors solve using a computational second site suppressor approach (4). This entails first identifying a single-point GTPase mutation that fully disrupts interaction with the wild-type GEF, then in a second step altering the GEF to accommodate this GTPase mutation. Remarkably, the authors identify a single GEF mutation that fully shifts interaction selectivity toward the redesigned GTPase. The fact that a single pair of point mutations proved sufficient to produce an orthogonal protein–protein interface implicitly underlies the ability of this new protein pair to address certain additional key design criteria.

A second design challenge is maintaining signal transduction to downstream effectors despite the sequence changes required for orthogonality. Although the approach of Kapp et al. (1) does not consider this design requirement explicitly, their use of structure-based techniques implicitly leads to point mutants predicted to be consistent with the desired protein conformation and not overly destabilizing. The fact that the redesign was limited to a single pair of point mutations provides further reason to expect that signal transduction to downstream effectors will be preserved. Even with these considerations, however, the redesigned GTPase/GEF pair is a weaker catalyst than the starting wild-type pair. This difference is due to unexpected long-range structural consequences arising from this pair of point mutations and represents a particular challenge for the future. Predicting potential structural and functional disruptions arising from variations in sequence remains an active focus of research (6), and improved tools to address this problem will surely prove helpful in selecting mutations that better maintain downstream signaling.

A more subtle design challenge associated with redesign of regulatory “hub” proteins is maintaining the variety of potential interactions with upstream regulators. These upstream regulators may act by a variety of potential mechanisms, including covalently modifying the hub protein, binding of adapter proteins, and altering subcellular localization of the hub protein. Failure to precisely maintain these upstream effects can lead to profound differences not only in activity but also specificity of a hub protein (7). On the basis of this “minimally disruptive” approach taken by Kapp et al. (1), upstream interactions of their redesigned GTPase can be expected to behave exactly as those of the wild-type GTPase.

Returning to the original goal of orthogonality, a final design challenge results from the fact that potential interference may derive from endogenous cellular proteins other than those used as a starting point for design. Given the modularity of many protein interaction domains, a particular concern is inadvertent interaction with other members of the same protein family. Given the finite (and well-established) collection of Rho-like GTPases and their cognate GEFs and their highly conserved modes of interaction (8), it should be possible to use a “multistate” setup to directly design against the possibility of inadvertently introducing new interactions with a competing family member. Indeed, exhaustive consideration of all basic-region leucine zipper (bZIP) families has been used to construct from first principles new sequences selective for a single target bZIP (9). Rather than take this direct approach, however, the fact that

Kapp et al. (1) rely on a single pair of point mutations may have indirectly maintained selectively within their designed pair. Because they collectively evolve in a shared environment, selective pressure prevents unwanted interaction between naturally occurring noncognate pairs (10); the modest sequence changes introduced by Kapp et al. (1) may have indirectly preserved the subtle effects of evolution that eliminate binding to other Rho-like GTPases and their cognate GEFs. Although not explicitly applied in this case, Kapp et al. (1) point out retrospectively that potential mutations that shift the specificity toward a competing family member could in principle be identified at the level of sequence.

In light of these considerations, the “minimally disruptive” approach of Kapp et al. (1) represents a highly efficient strategy to indirectly and simultaneously address a number of important criteria that would otherwise each individually represent a challenging design task. The ability to transfer numerous subtle and nontrivial features from the starting protein pair stands as a major strength of this approach.

Ultimately, the question moving forward will be this: can we build orthogonality into other systems just as easily, or is there something particular about this Cdc42–Intersectin interface? Although interactions critical for signaling are surely evolved for robustness to spontaneous mutations, the demonstration that a single pair of carefully chosen point mutations is sufficient to confer orthogonality with little loss of function (1) leads one to wonder what range of orthogonal protein pairs could be created through similarly minimal changes in sequence. Accordingly, one can certainly envision extending the scale of this approach to build up complete orthogonal pathways that mimic and coexist alongside any number of endogenous cellular pathways in cells.