Recognizing and Defining True Ras Binding Domains I: Biochemical Analysis

Common domain databases contain sequence motifs which belong to the ubiquitin fold family and are called Ras binding (RB) and Ras association (RalGDS/AF6 Ras associating) (RA) domains. The name implies that they bind to Ras (or Ras-like) GTP-binding proteins, and a few of them have been documented to qualify as true Ras effectors, defined as binding only to the activated GTP-bound form of Ras. Here we have expressed a large number of these domains and investigated their interaction with Ras, Rap and M-Ras. While their (albeit weak) sequence homology suggest that the domains adopt a common fold, not all of them bind to Ras proteins, irrespective of whether they are called RB or RA domains. We used fluorescence spectroscopy and isothermal titration calorimetry to show that the binding affinities vary over a large range, and are usually specific for either Ras or Rap. Moreover, the specificity is dictated by a set of key residues in the interface. Stopped-flow kinetic analysis showed that the association rate constants determine the different affinities of effector binding, while the dissociation rate constants are in a similar range. Manual sequence analysis allowed us to define positively charged sequence epitopes in certain secondary structure elements of the ubiquitin fold (β1, β2 and α1) which are located at similar positions and comprise the hot spots of the binding interface. These residues are important to qualify an RA/RB domain as a true candidate Ras or Rap effector.

Introduction

Ras is a GTP binding protein involved in many signal transduction processes such as control of growth, differentiation and apoptosis. It acts as a molecular switch that cycles between an inactive GDP- and an active GTP-bound state.1, 2 In the resting state it contains tightly bound GDP, whose dissociation is increased by many orders of magnitude by the action of guanine nucleotide exchange factors. This leads to the activation and GTP loading of Ras. In the GTP-bound form it interacts with effector proteins that are defined as proteins having a high affinity to the GTP- but not the GDP-bound form of Ras. Signal transduction to the effectors is terminated by the GTPase reaction that is intrinsically very slow, and is stimulated by GTPase-activating proteins (GAPs) which increase the reaction by many orders of magnitude.³

The first effector of Ras to be identified was the Ser/Thr protein kinase c-Raf1 which initiates a cascade of protein kinases, the MAP kinase module, to activate MEK1, which in turn activates the MAPKs ERK1 and ERK2.4, 5 c-Raf-1 contains a Ras binding domain (RBD): a stable 81-residue protein domain, sufficient to bind to Ras in a GTP-dependent manner. The structural analysis showed that it has an ubiquitin fold which interacts with Ras (and Ras-like proteins) by forming an inter-protein β-sheet, involving the outer strands of the two proteins.6, 7, 8 The interface of that complex was shown to be mostly determined by complementary charge interactions, the surface of Ras being negatively charged and the surface of Raf-RBD being positive. The structure of the RBD of the Schizosaccharomyces pombe protein Byr2, genetically and hierarchically a homologue of the mammalian MEK1 or Raf, was also solved both alone9, 10 and in complex with Ras.¹¹ While the overall fold is similar, the details in the intermolecular interactions forming the interface with Ras are different.

RalGDS (Ral guanine dissociation stimulator) and the isoforms Rgl, Rgr Rlf, and the α,β,γ PI(3) kinases have also been identified as effectors of Ras.12, 13, 14, 15 They contain an RA domain (RalGDS/AF6 or Ras association domain), which in spite of low sequence homology has a similar structure. The structures of the RA/RB domains, and of their complexes with Ras proteins, show that they share the same fold and also bind to the effector switch I region of Ras via a similar binding mode.16, 17, 18, 19, 20 This interaction, however, involves a different set of residues.21, 22, 23 In the case of the complex between Ras and PI(3)Kγ, which is the first structure with a complete effector, rather than the RA/RB domain alone, the interaction also involves switch II, and this interaction is thought to regulate the activity of the enzyme allosterically.

Using yeast two-hybrid analysis and pull-down techniques, a large number of additional putative effectors have been identified, such as adenylate cyclase in Saccharomyces cerevisiae,²⁴ AF6,25, 26 phospholipase Cε,²⁷ Nore1 (novel Ras effector 1)28, 29 and other isoforms of the tumour suppressor RASSF1.³⁰ In a recent publication, Rodriguez-Viciana and co-workers have performed an extensive study on the interaction of RA and RB domains, in complex with various members of the Ras sub-family, using pull-down experiments.³¹

The RA (RalGDS/AF6) domain has been defined by Ponting & Benjamin, based on database searches and sequence analysis.³² This domain is present in members of the RalGDS family, RIN1 and many proteins not previously recognized as downstream effectors of Ras such as AF6 and the Drosophila homologue canoe, Myr-5, DAG kinase, Ste50 and Ste4. Surprisingly, Raf-RBD scored very low in the sequence comparison and was not considered to belong to this domain family although the authors assumed, as was later verified by structural analysis, that the RA and RB domain have the same fold. Since ubiquitin also has the same fold, we will define from now the overall ubiquitin-fold family, comprising RA/RBD and ubiquitin, as UB domains.

A further complication in the search for effectors for Ras arises from the fact that the superfamily of Ras-like proteins includes a Ras subfamily containing, among others, the proteins Rap (five isoforms) and R-Ras (three isoforms). These are believed to have functions at least partially overlapping with Ras and also seem to interact with RA and RB domain containing proteins.33, 34 Thus, R-Ras and Rap1 have been shown to bind to various RA or RB domains in vitro, in a GTP-dependent manner, and in the case of RalGDS the binding is tighter than to Ras.³⁵ Furthermore, numerous functional studies have shown interactions between Ras-like proteins and effectors containing RA or RB domains to be functionally meaningful.

At present, 108 RA and 20 RB domains, from human proteins, are listed in the SMART database.36, 37 The role of many of these in signal transduction, via Ras/Rap/R-Ras, is not obvious. Not counting possible redundancies, it seems unlikely that all of them are true Ras effectors. Using model building, it has in fact been shown that Myr-5 is unlikely to form a productive interface with Ras.³⁸ In order to define the requirements for productive binding of predicted RA and RB domains to Ras-GTP, or possibly other Ras-like proteins, we have measured the interaction of a large number or RA and RB domains, using different methodologies. Based on these experiments, we have derived common denominators for efficient complex formation. Using this information in the accompanying paper (Kiel et al.)³⁹ and through the use of homology modelling and prediction of binding energy, we develop a methodology to do genome-wide prediction for this type of interaction.