The small GTPases Ras and Rheb studied by multidimensional NMR spectroscopy: structure and function

Introduction: the small GTPases Ras and Rheb

Cells are constantly sending messages and are checking nutrient levels and growth rates within the cell as well as with other cells. These messages need to be explicit and one way to amplify signals is to link them to a process that is chemically irreversible, like the cleavage of ATP or GTP. Adenosine triphosphate (ATP) and guanosine triphosphate (GTP) are used in living cells as cofactors for various biochemical transformation reactions.

Whereas ATP is regarded as a ‘molecular unit of currency’ in intracellular energy metabolism, the hydrolysis of GTP to GDP (guanosine diphosphate) mainly plays a regulatory role in biochemical processes, including cell growth, cell differentiation, as well as vesicular and nuclear transport. Proteins capable of binding guanosine nucleotides are called guanosine binding proteins (or guanine nucleotide-binding proteins, GNBPs). The superfamily of monomeric small GTPases includes proteins with a size of 20–25 kDa and a common fold that consists of a central six-stranded mixed β-sheet surrounded by a total of five α-helices.

The regulatory role of the small GTPases is fulfilled through their capability to act as binary switches by toggling between an ‘on’ and ‘off’ state (Figure 1). These states are chemically characterised by the bound nucleotides: GTP (= on) and GDP (= off). Since the intrinsic GTP hydrolysis, performed by the small GTPases, is very slow (10⁻⁶ 1/s), it is accelerated by GTPase activating proteins (GAPs) by several magnitudes (Gibbs et al., 1984). The hydrolysis products GDP and P_i are not automatically released, because the protein exhibits the same affinity for GTP and GDP (Renault et al., 2003; Pasqualato and Cherfils, 2005; Cherfils, 2014). Therefore, the exchange process is also catalysed by a group of different proteins, which are structurally totally unrelated: guanine exchange factors (GEFs). Ultimately, GTPases bind and hydrolyse GTP and thereby switch from an active (GTP-loaded) to an inactive (GDP-loaded) form. Thus they play a crucial role in cellular signalling. The communication between the G-Domain and its various effector proteins is mediated by two regions, switch I and switch II, to transmit external signals from growth factors to intracellular signalling cascades. Both regions undergo dramatic structural changes, depending on which nucleotide is bound. The underlying molecular mechanism is known as the ‘loaded-spring mechanism’, introduced by Vetter and Wittinghofer (2001). This universal mechanism involves two amino acids, a conserved threonine (T35), located in switch I, and a glycine (G60), which adopt distinct conformations when GTP is bound. During the process of GTP-hydrolysis, the ɣ-phosphate is released and the switch regions are allowed to relax into different conformations. Hitherto, 167 proteins have been identified that are part of the human Ras superfamily, of which 39 proteins belong to the Ras family (Rojas et al., 2012).

Figure 1:

The GTPase cycle involves the exchange of GDP by guanine nucleotide exchange factors (GEFs), which control the nucleotide exchange by increasing the dissociation rate.

The GEF protein directly inserts certain amino acids into the nucleotide binding domain (NBD). Thereby, the affinity of the GTPase towards the nucleotide is reduced. Once activated, the GTP-bound state enables GTPases to interact with various effectors. GTPase activating proteins (GAPs) catalyse GTP hydrolysis and return GTPases to their GDP-bound ‘OFF’-state, thereby completing the GTPase cycle.

Our project mainly focused on two small GTPases, Rheb and K-Ras4B. Ras is probably the most famous small GTPase, because of its importance for cell growth, differentiation, and survival. In the human proteome, three Ras isoforms (or splice versions) and one splice variants are found: N-, H-Ras, K-Ras4A, and K-Ras4B. These amino acid sequences share a high similarity (>95%) and differ only in a ‘hyper variable region’ (HVR), that comprises the carboxy-terminal 25 amino acids. The final four residues of the HVR region are known as the CAAX-box, which is the target of posttranslational modification (Figure 2). The Ras proteins undergo four steps of modification to mediate membrane binding: isoprenylation, proteolysis, methylation, and palmitoylation (Aronheim et al., 1994). Additionally, K-Ras4B possesses a polybasic stretch of six lysines that is capable of binding to negatively charged phospholipids of distinct cell membranes. The Ras protein is a known proto-oncogene and approximately up to one-third of all human cancers are caused by a mutation in this particular protein. The highest frequency of Ras mutations are mostly found in pancreatic (90%), lung (40%), and colorectal cancer types (50%) (Schubbert et al., 2007). Three mutations hotspots have been identified, which mainly occur at codons 12, 13, and 61. In total, these three mutation sites occur in 97–99% of all Ras mutations in cancer (Cox and Der, 2010). Interestingly, most mutations are found in K-Ras (85%), followed by N-Ras (12%), and rarely in H-Ras (3%) (Cox et al., 2014).

Figure 2:

Schematic presentation of the four Ras isoforms.

H-Ras, N-Ras, K-Ras4A and K-Ras4B are highly homologous throughout the conserved G domain (amino acids 1–166). The C-terminal hypervariable domain (amino acids 166–188/189) specifies membrane localisation through post-translational modifications that include the farnesylation of each isoform on the C-terminal CAAX motif and palmitoylation of cysteines on H-Ras, N-Ras, and K-Ras4A (highlighted in yellow). Membrane localisation of K-Ras4B is also facilitated by a stretch of lysine residues in support of the farnesyl moiety.

Rheb (Ras homologue enriched in brain) is a small GTPase that is related to Ras, Rap, and Ral (Tee et al., 2003, 2005; Ehrkamp et al., 2013). Rheb, is part of the mTOR pathway that integrates intra- and extracellular signals and thereby regulates cell metabolism and growth proliferation. Unlike Ras, Rheb harbours an Arg and Ser instead of Gly residues in its P-loop that binds the β phosphate of either GDP or GTP (Karassek et al., 2010). As a molecular switch, Rheb regulates cell volume, cell growth, cell cycle progression, neuronal axon regeneration, autophagy, nutritional deprivation, oxygen stress, and cellular energy status (Takei et al., 2004; Karassek et al., 2010). Rheb’s impact on growth and cell cycle progression is mediated by the regulatory influence of growth factors on the insulin/mTORC1/S6K signalling pathway via the mammalian target of rapamycin (mTOR) (Manning and Cantley, 2003). A complex that consist of the tuberous sclerosis gene products, hamartin (Tsc1) and tuberin (Tsc2) serves as a GTPase activating protein (GAP) for Rheb, which implies that Rheb plays an important role in tuberous sclerosis (Manning and Cantley, 2003; Aspuria and Tamanoi, 2004; Karassek et al., 2010). Patients suffering from this genetic disorder characteristically develop benign hamartomatous tumours in the brain, kidneys, heart, lungs, skin or eyes and cannot be medically cured to date. Interestingly, activation of the Rheb-mTORC1 pathway may have opposite cellular effects depending on the cell stress state. While in cancer cells Rheb promotes cell cycle progression, there is an enhancement of apoptosis by Rheb observed after cellular application of various toxic stimuli such as UV stress or ER-stress (Ozcan et al., 2008; Karassek et al., 2010). Furthermore, Rheb synthesis is rapidly up-regulated as an immediate-early response gene after injury (Yamagata et al., 1994; Potheraveedu et al., 2017). Over-expression of Rheb enhances cellular degeneration, which can be prevented by rapamycin or by knock down of apoptosis signal-regulating kinase (ASK-1) (Karassek et al., 2010). Thus, clinical application of rapamycin should take into consideration the cellular stress state which could change its therapeutic effects (Karassek et al., 2010).

Structure in solution of Rheb/GDP and its dynamics on the pico- to nanosecond time scale

Based on the NMR assignments of Rheb bound to both GDP and GppNHp (Berghaus et al., 2007; Schwarten et al., 2007), a non-hydrolysable analogue of GTP, we were able to determine the structure in solution of the Rheb/GDP complex by multidimensional heteronuclear NMR spectroscopy. The complex adopts the typical canonical fold (Figure 3) of Ras-GTPases (Karassek et al., 2010). In the NMR structure, switch I and, in part, switch II of Rheb-GDP are ill-defined in solution because of fluctuations on the pico- to nanosecond time scale, as observed in the heteronuclear NOE experiment (Karassek et al., 2010). This seems to be a common feature of small GTPases, as this has also been observed for the structure of Ras in solution (Kraulis et al., 1994). In sharp contrast to the GDP-loaded inactive state, the ¹H-¹⁵N HSQC spectrum of activated Rheb bound to Gpp(NH)p is characterised by extensive line broadening for numerous resonances of the switch regions. Obviously, switch I adopts multiple conformations in solution that interconvert on an intermediate time scale (Schwarten et al., 2007). This dynamic feature could be of functional relevance as switch I apparently adopts different conformations and some of these conformations might be of catalytical importance for GTP hydrolysis.

Figure 3:

Backbone representation of the structural ensemble of Rheb in solution (PDB data bank code 2L0X).

The unstructured N-terminal residues S1–K5 and C-terminal residues S175–M184 have been omitted for clarity. α-helical secondary structural elements are shown in red, β-sheets in green, and loops in yellow. The dynamic switch I and II regions are shown in orange. The GDP is depicted as a stick model and the magnesium ion is shown in magenta.

The crucial difference between small GTPases affects the switch II region. In the crystal structures of Rheb, switch II is structurally very similar in the GDP- and GTP-bound states (Vetter and Wittinghofer, 2001; Yu et al., 2005). Unlike the crystal structure however, switch II like switch I of Rheb-Gpp(NH)p presumably adopts multiple conformations in solution. These multiple conformations alternate on the micro- to millisecond time scale, which leads to extensive line broadening in the ¹H-¹⁵N HSQC spectrum (Schwarten et al., 2007). This observation suggests that switch I and II can adopt several (two or more) different slowly interconverting conformations (Karassek et al., 2010). It is important to note that a very similar dynamic behaviour also been observed for Ras (Kraulis et al., 1994; Ito et al., 1997; Spoerner et al., 2001; Vetter and Wittinghofer, 2001). This conformational flexibility might be of functional relevance not only for the GTP/GDP cycle but also for the selectivity of small GTPase towards GAPs and GEFs (Kraulis et al., 1994; Ito et al., 1997; Spoerner et al., 2001; Vetter and Wittinghofer, 2001). Based on these cumulative findings, the interaction between Rheb (or Ras) and its GAP and/or GEF is obviously characterised by conformational selection, a feature also observed in numerous other protein ligand complexes (Karassek et al., 2010; Weikl and Paul, 2014).

Rheb’s GAP and GEF

Whereas a protein complex formed by the tuberous sclerosis gene products, the tumour suppressors TSC1 (or hamartin) and TSC2 (or tuberin), has been shown to function as a GTPase activating protein (GAP) for Rheb (Manning and Cantley, 2003; Aspuria and Tamanoi, 2004; Karassek et al., 2010), the GEF for Rheb remains elusive to date. In fact, Rheb may not possess a genuine GEF but instead basal nucleotide exchange rates may suffice to load Rheb with GTP given the fact that hormones down-regulate its GAP, TSC (Wang and Proud, 2011). Amino acids mediate mTOR/raptor signalling through activation of class 3 phosphatidylinositol 3OH-kinase (Nobukuni et al., 2005). It could be shown that TCTP is essential for growth and proliferation through regulation of Rheb GTPase in Drosophila (Hsu et al., 2007) as Rheb acts on the PI3 kinase/mTOR pathway and it was suggested that TCTP could act as a GEF for Rheb.

TCTP regulates apoptosis and is involved in malignant transformation (Hsu et al., 2007). However, several studies suggest that TCTP is very unlikely to act as a GEF for Rheb, because neither in vitro nor in vivo experiments could show any nucleotide exchange activity of TCTP towards Rheb bound to GDP (Rehmann et al., 2008; Wang et al., 2008). In particular, ¹H-¹⁵N HSQC-NMR-based binding studies did not reveal any significant NMR chemical shift perturbation for Rheb backbone amide resonances up to a threefold molar excess of TCTP. This is in sharp contrast to ¹H-¹⁵N HSQC-NMR spectra of Ras in complex with SOS, a well-established GEF, which are characterised by substantial line broadening. This suggests that the interaction between Rheb and TCTP is – if at all – extremely weak and not of any physiological relevance (Rehmann et al., 2008). TCTP is structurally related to Mss4, a known GEF for the Rab family (Thaw et al., 2001). However, a detailed comparison of TCTP proteins and the MSS4Rab3 complex clearly shows that while the core of TCTP and Mss4 can be superimposed, a sequence insertion in TCTP sterically clashes with Rab3 (Rehmann et al., 2008). This excludes a mode of interaction for TCTP and Rheb similar to that of Mss4 and Rab. Obviously, more studies in the future are still required in order to fully understand the relationship between TCTP and the TSC/Rheb/mTOR pathway.

Low molecular weight compounds as functional modulators of small GTPases

Antagonising biomolecular interactions by means of small molecules is pivotal to medicinal chemistry and is regarded a crucial strategy in cancer therapy. Yet, despite some progress in recent years, the inhibition of protein-protein interactions by low molecular weight compounds still represents a major challenge in numerous research endeavours. Biophysical interactions between biomolecules are usually mediated by rather large surfaces, which is why each residue of the biomolecular interface only minimally contributes to the overall binding free energy.

Initially, we set out to search for low molecular weight compounds that target Rheb and might provide novel candidate scaffolds that could, once chemically optimised, be used for treatment of tuberous sclerosis-mediated tumour growth. Inspired by previous studies, which, for example, performed an NMR-based screening of a large 11000-member library (Sun et al., 2012), we designed a small fragment library based mainly on the privileged structure concept and the ‘rule of five’ (Choy and Prausnitz, 2011). This focused library that contained not more than 100 compounds was screened by multidimensional NMR spectroscopy to identify structures that selectively bind to Rheb. In detail, we have applied multidimensional heteronuclear NMR spectroscopy and chemical shift perturbation analysis in order to characterise the interactions between low molecular weight compounds and Rheb. It is interesting to note that most of the molecular hits contained structural elements that had either a linear shape with a biphenyl scaffold or were bent, induced by an sp²- or sp³-hybridised carbon atom between the two phenyl rings (Schoepel et al., 2013). In the case of Rheb, the most significant chemical shifts were found for Bisphenol A, 4,4′-biphenol, and 4,4′-dihydroxybenzophenone. The identified binding sites were then used as experimental restraints in a molecular docking procedure, using the HADDOCK programme suite (High Ambiguity Driven biomolecular DOCKing) (Dominguez et al., 2003). Structures obtained from these docking calculations were refined using OPLS_2005 force field minimisations in explicit water as solvent (Schoepel et al., 2013). In the refined docking model, the complex between Rheb and 4,4′-biphenol is stabilised by three hydrogen bridges between I69 and I78 to one 4,4′-biphenol hydroxyl group and by another hydrogen bond between the second 4,4′-biphenol hydroxyl group and K109, which correlate well with the observed chemical shift perturbation. The hydrogen bonds stabilise the 4,4′-biphenol horizontally at the top of a deep lipophilic binding pocket and the two aromatic rings are twisted by approximately 30°. As the switch II region of Rheb has been shown to exhibit an increased flexibility on the pico- to nanosecond time scale, conformational selection might indeed play a role in ligand binding to Rheb (Karassek et al., 2010). Based on the multidimensional NMR data on biphenol and Rheb, the affinity is only approximately 1500±200 μm (Schoepel et al., 2013). Such affinities are typical for fragment-based screening approaches and imply further optimisation through chemical synthesis in order to increase binding affinity (Maurer et al., 2012; Sun et al., 2012). If the two phenol moieties of 4,4′-biphenol are separated by introducing an sp² carbon, like in benzophenone, or by a quaternary sp³ carbon with limited conformational freedom, like in 4-[2-(4-Hydroxyphenyl)propan-2-yl]phenol known as Bisphenol A, the bent structure directs the aromatic rings deeply into the binding pocket. The K_D value was extracted from multidimensional NMR spectra for a Bisphenol A/Rheb complex and is approximately only 1800±500 μm (Schoepel et al., 2013). In the refined docking model, Bisphenol A is stabilised in the lipophilic pocket by three hydrogen bridges, similar to 4,4′-biphenol (Figures 4 and 5). In Rheb, S68 and Q72 form hydrogen bonds with one hydroxyl group of Bisphenol A. Y67, which is located at the bottom of the pocket, participates in another hydrogen bond with the second hydroxyl group of Bisphenol A. In addition, π-cation interaction between the side-chain amino group of K109 and the phenol group of Bisphenol A located deep inside the pocket contribute to the overall binding energy. A similar binding geometry is found for 4,4′-dihydroxybenzophenone. The highly flexible 4,4′-methylenediphenol, however, only induces minor chemical shift changes in the NMR spectra of Rheb (Schoepel et al., 2013).

Figure 4:

The observed weighted chemical shift differences projected onto the molecular surface of K-Ras and Rheb, respectively.

On the left, weighted chemical shift perturbation of K-Ras4B bound to GDP upon titration with Bisphenol A are projected onto the molecular surface of K-Ras4B. On the right, weighted chemical shift perturbation of Rheb bound to GDP upon titration with Bisphenol A are projected onto the molecular surface of Rheb. The colour code is yellow for small, orange for medium, and red for large NMR chemical shift perturbations upon ligand binding. Grey colour indicates no significant changes in the 2D ¹H-¹⁵N NMR spectra.

Figure 5:

Examples of HADDOCK-derived structures of K-Ras4B in complex with BPA and BPS.

BPA (left) exhibits a higher degree of flexibility due to an sp³ hybridisation of its central carbon atom. BPS (right) lacks this flexibility and remains more rigid. Observed chemical shift perturbations are coloured in orange (medium) and red (large), respectively.

Finally, we could also show that 4,4′-biphenol selectively inhibits Rheb signalling and induces cell death, which suggests that this compound might be a novel candidate for treatment of tuberous sclerosis-mediated tumour growth (Schoepel et al., 2013). In order to investigate as to whether there is a cellular effect of 4,4′-biphenol by Rheb we determined dose-response curves for cellular degeneration and signalling. Microscopic assessment of cells indicated maximal cellular degeneration and blockade of phospho-S6 at around 100 μm thus correlating with Rheb’s binding affinities to 4,4′-biphenol while there was no such effect of bisphenol A. Obviously, 4,4′-biphenol inhibits the mTOR downstream signalling without affecting Ras signalling (Schoepel et al., 2013).

The switch-dependent heterogeneity of the Ras conformations, especially in the GDP-bound form, is only poorly portrayed by structures that were obtained by crystallography. Moreover, on the basis of these measurements only, the Ras protein was believed to be undruggable, since it lacks hydrophobic cavities on its molecular surface (Cox et al., 2014). Nevertheless, during the past years several small molecules and zinc-chelating compounds have been identified and characterised as Sos antagonists of K-Ras (Palmioli et al., 2009; Rosnizeck et al., 2012). In 2012, co-crystal structures of several ligands bound to the full-length and truncated Ras were published (Maurer et al., 2012; Sun et al., 2012). With the help of a NMR-based fragment screening, Genentech identified DCAI (4,6-dichloro-2-methyl-3-aminoethyl-indole) as a Ras ligand (Maurer et al., 2012), which binds in a hydrophobic cavity located between helix α2 and the core β-sheet, β1–β3. The dimensions of this binding pocket are of approximately 7×7 Å at the opening with a depth of 5 Å. The interacting amino acids include K5, L6, V7, I55, L56, and T74. The DCAI compound has a K_D value of 1.1±0.5 mm, as derived from NMR experiments. The same binding pocket was identified in another study (Sun et al., 2012). There, a indolic compound (S)-N-(2-((1H-indol-3-yl)methyl)-1H-benzo[d]imidazol-5-yl)pyrrolidine-2-carboxamide) (Sun et al., 2012) was discovered as an interaction partner for Ras. This compound was found to bind to mutant (G12V and G12D, respectively) versions of the Ras protein. This compound binds to K-Ras(G12D) with an affinity of K_D=190 μm. Since the binding pocket of both compounds is located in the Ras-GDP/SOS interaction interface, it is not surprising that the interaction between both proteins is disturbed after ligand binding. The prolonged indole compound showed an inhibition of the SOS-meditated nucleotide exchange of 58±8% and the DCAI molecule blocked both nucleotide exchange and release reactions with an IC₅₀ of 342±22 μm and 155±36 μm, respectively (Maurer et al., 2012; Sun et al., 2012). Interestingly, both authors highlight tyrosine 71, which is located in switch II. This residue plays an important role as it occupies the hydrophobic pocket in the apo-crystal structure. Upon ligand binding this residue is replaced by the ligand and it is this replacement that exposes the hydrophobic cavity.

Although several (fragment-based) screenings for small molecular antagonists of protein-protein interactions between Ras and its effectors have been proven to be difficult as affinities remain often rather low (Ganguly et al., 1998; Maurer et al., 2012; Rosnizeck et al., 2012; Cox et al., 2014; Schoepel et al., 2016), GTP analogues that bind tightly or even irreversibly to Ras could in principle lead to severe side effects because of unselective cross-reactivities towards other GTPases. Nonetheless, a promising study recently reported on selectively targeting a Cys mutant of Ras (Ostrem et al., 2013; Hunter et al., 2014; Min Lim et al., 2014; Wiegandt et al., 2015; Ostrem and Shokat, 2016). These ligands bind irreversibly to a mutant version of the Ras protein (G12C) (Ostrem et al., 2013). The synthesised compounds rely on the mutant cysteine to be present in the GTPase amino acid sequence for binding due to acrylamide and sulphonamide coupling groups and should therefore not affect the wild-type protein. This binding pocket for these compounds differs from the site previously described and is located between the central β-sheet of Ras, and the α2-(switch II), and α3-helices (Maurer et al., 2012; Rosnizeck et al., 2012; Sun et al., 2012; Ostrem et al., 2013; Schoepel et al., 2013). Like the aforementioned other binding pockets, this ligand binding cavity does not exist in the crystal structures of GDP-bound wildtype Ras and hence is probably highly dynamic in solution. Inspired by this work and the fact that we recently had found a Rheb ligand (as described above) we also tested those ligands on the K-Ras4B protein. We did not observe any chemical shift perturbation when we tested 4,4′- biphenol for its binding to K-Ras4B as judged from ¹H-¹⁵N HSQC spectra. Obviously, this ligand is quite selective for Rheb. Bisphenol A however, binds to K-Ras4B with an affinity of K_D=600±200 μm, which is therefore significantly higher than for Rheb (Schoepel et al., 2013). The binding site in K-Ras identified in our study for Bisphenol A is identical to the one described by Genentech and Sun et al. in 2012 and is located between α-helix and the core β-sheet, β1–β3. Accordingly, we could show that amino acids L6, V7, L56, T74, and G75 contact Bisphenol A. In addition, our HADDOCK-based docking models show that the aromatic rings must adopt an almost right angle for optimal binding. Our structural analyses reveal that the binding site in Rheb is considerably bigger than in K-Ras so that stretched, linear molecules such as 4,4′-biphenol cannot bind into the latter pocket (see Figures 4 and 5) (Schoepel et al., 2013).

We also addressed the question as to whether our ligands could interfere with the SOS-meditated nucleotide exchange, since the SOS protein mainly binds both switch-regions of the GTPase, which is close to the binding pocket of the afforested ligands. Indeed, we could detect a reduced of the Sos-mediated nucleotide exchange of both H- and K-Ras by a factor of 2.5-fold and 1.6-fold, respectively. This is again in accordance with recent studies, in which a comparable stoichiometric excess of small molecular compounds was used (Maurer et al., 2012; Sun et al., 2012). Judged from our HADDOCK docking models and the available co-crystal structures, it seems highly likely, that these different ligands, including Bisphenol A, bind to GDP-Ras and interfere with SOS complex formation due to steric reasons. Interestingly, the observed affinity of Bisphenol A is in the same micromolar range as the affinity between Ras-GDP and the SOS molecule itself (Palmioli et al., 2009).

The HADDOCK-based docking studies suggest that the Bisphenol A binding site on K-Ras is smaller than the corresponding Rheb site. It seems like that rigid, linear molecules such as 4,4′-biphenol cannot bind in this pocket, whereas the more flexible bisphenolic compound does. Bisphenol A (BPA) is one of the most-produced chemicals worldwide with approximately 10 billion tonnes in 2011 (Vom Saal et al., 2012). BPA, in which central carbon is accompanied by two methyl and two phenolic groups, is used in food can linings, thermal papers, and other daily life plastic products. The main source for human exposal to bisphenols is based on the uptake through food and drinks (Vandenberg et al., 2007). Therefore, bisphenols are found in human serum (Ribeiro-Varandas et al., 2013), urine (Matsumoto et al., 2003), adipose (Ben-Jonathan et al., 2009), and placental tissues (Arbuckle et al., 2015). Outside the human body, significant amounts of bisphenols can be found in drinking and wastewater, air, as well as dust (Vandenberg et al., 2007). The second main source is thermal paper, used, e.g. cashier receipts, in which the BPA is not polymerised. This renders it more available for exposure than polymerised BPA in resins or plastics (Pivnenko et al., 2015).

Lately, BPS has been adopted as a BPA substitute because of public pressure and new governmental restrictions (Glausiusz, 2014). For example, BPS can be found in ‘BPA-free’-labelled thermal paper (Liao et al., 2012). Also, BPS is a known endocrine disruptor and is believed to have a comparable physiological effect (Hashimoto et al., 2001; Grignard et al., 2012). Lately, we turned our attention to this bisphenolic analogue of BPA. While the two ring systems in BPA are connected via a sp³-hybridised carbon atom, these rings are connected via a sulphonyl group (SO₂) in Bisphenol S, which leads to reduced flexibility. Judged from the chemical shift perturbations, BPS also binds to K-Ras4B, although with a lower affinity of 5.8±0.7 mm (Schoepel et al., 2016). In addition to those induced by BPA we observed a shift (twice the standard deviation) for the amino acid S39, which is part of the switch I region. Obviously, BPS adopts a different, twisted orientation in concordance with our docking results (Dominguez et al., 2003), which showed a different bending angle for BPA and BPS, respectively.

As judged from the NMR titration data, BPS and BPA bind K-Ras4B. However, we did not observe any reduction in the exchange rate for Bisphenol S in the SOS-nucleotide exchange assay. We therefore conclude, that only ligands with a K_D value in the micromolar range, such as BPA, are able to interfere with the nucleotide exchange because the affinity between the Ras and SOS is in a similar affinity regime (Palmioli et al., 2009). Taken together, BPS is a ligand for K- Ras4B, whose affinity is ten times lower than that of BPA. Thereby, BPS apparently loses its antagonising function in the Sos-meditated nucleotide exchange of K-Ras4B.

Overall, numerous groups were able to show that low molecular weight compounds can selectively bind to Ras and interfere with the Sos-mediated nucleotide exchange in H- and K-Ras (Taveras et al., 1997; Spoerner et al., 2001; Gorfe et al., 2008; Araki et al., 2011; Patgiri et al., 2011; Maurer et al., 2012; Sun et al., 2012; Hocker et al., 2013; Ostrem et al., 2013; Shima et al., 2013; Min Lim et al., 2014; Leshchiner et al., 2015; Winter et al., 2015). Apparently, the activity of small GTPases can be indirectly modulated by small molecule through disrupting the Ras/SOS guanine nucleotide exchange complex. After decades in search of low molecular weight ligands for small GTPases, recent progress now paves the way for future design of much-needed GTPase-selective antagonists with higher affinity to benefit the treatment of Rheb-mediated brain disease and Ras-driven tumours.

For membrane-associated proteins such as Rheb and Ras, a completely different strategy from the ones described above aims at masking the crucial C-terminal CaaX-box with peptidomimetic receptors. These will interfere with or even prevent the membrane insertion and localisation of Ras-proteins that is a prerequisite for their normal biological function. Ras-proteins are post-translationally lipidated by prenyl transferases at the C-terminal CaaX-box of Ras-proteins with farnesyl and/or palmitoyl moieties (Figure 2) (Brunsveld et al., 2006; Triola et al., 2012). For Rheb and Ras to be physiologically active, proper membrane attachment and localisation is essential even though Rheb is only monofarnesylated (Buerger et al., 2006). During the past decade, several inhibitors of these farnesyl transferases (FTIs) could be developed, some of which even entered clinical trials (Appels et al., 2005). Although only one farnesyl transferase farnesylates all GTPases that accept the farnesyl moiety, FTIs are likely to have side effects on related GTPases (Basso et al., 2006) K-Ras4B, one most important oncogenic Ras-proteins, can even be alternatively prenylated by geranyl-geranyl transferases (Rowell et al., 1997). It was also shown that antagonising the prenyl-binding protein PDEδ with small molecules efficiently disturbs the spatial organisation and signalling of farnesylated K-Ras4B (Zimmermann et al., 2013). To interfere with the Ras-prenylation via sequence-selective recognition of the CaaX-box of a specific Ras-protein by a synthetic small receptor molecule has been largely undervalued. This complementary approach to develop low molecular weight antagonists holds the promise of GTPase selectivity as CaaX-boxes sequences of Ras-proteins are different. Like other CaaX-box regions present in small GTPases, the C-terminus of Rheb is highly flexible and completely solvent-exposed (Karassek et al., 2010). The latter feature is an important prerequisite for a suitable receptor molecule to recognise its target sequence (Schmuck et al., 1999). Such potential CaaX-box antagonists might therefore be less toxic compared to known farnesyl transferase inhibitors. We were able to develop for the first time more drug-like CaaX-box receptors that contain d-, β-, and other non-proteinogenic amino acids. These receptors bind to the C-terminal CSVM sequence of Rheb in a sequence-selective manner. Indeed, saturation transfer difference (STD) NMR spectra, a ligand-based approach, confirm that these receptors also bind to the full-length Rheb protein in an aqueous buffer (Düppe et al., 2014). These pincer-like receptors might thus constitute excellent molecular scaffolds for future peptidomimetic sequence-selective receptors that exhibit higher affinities and may provide another promising new approach to selectively inactivate Ras proteins.

Conclusion and future perspectives

Today – with ever increasing computer power for molecular modelling of (protein) receptor complexes (using programmes such as HADDOCK) and the now widespread application of fragment-based drug discovery and development – medicinal chemistry is progressing at an impressive pace. Together with additional biophysical techniques, such as surface plasmon resonance (SPR) spectroscopy, isothermal titration calorimetry (ITC), microscale thermophoresis technology (MST), the analysis of small molecular weight fragments can complement classical high throughput screening endeavours, which can be described as a fragment-assisted rather than fragment-based technology (Renaud et al., 2016). NMR spectroscopy uniquely contributes to this process in that in not only reports on the affinity of a ligand for a selected protein target but also reveals vital structural information on this interaction at atomic resolution. This is (often) regarded as the basis for subsequent structure-activity relationship (SAR) studies in order to improve binding affinities of initial fragment hits in transit to lead structures to ultimately achieve a significant therapeutic effect. This process further benefits from improved methods and protocols in structural biology, including both X-ray crystallography and biomolecular NMR spectroscopy.

Among all strategies to target small GTPase like Ras (Ostrem and Shokat, 2016), the GEF reaction is regarded to be a prime target as it currently holds the greatest promise for success within the (near) future. In addition, it now even seems possible to selectively target tumour-specific Ras mutants, such as G12C and G12D, or even Ras isoforms (e.g. K- or H-Ras), which seemed an unachievable goal not so long ago (Ostrem and Shokat, 2016). Recently, numerous small molecular compounds have emerged as low molecular weight GEF/Ras antagonists (Thaw et al., 2001; Dominguez et al., 2003; Palmioli et al., 2009; Maurer et al., 2012; Schoepel et al., 2013; Min Lim et al., 2014). All of these studies have focused on the GDP-loaded form as Ras as it is commonly accepted that this target will probably lead to an efficient (and therefore efficacious) antagonistic lead structure much faster. Allosteric low molecular weight compounds with high affinity and selectivity for certain GTPase or even isoforms thereof have not yet been described. But such compounds promise to perform much better in modulating the interaction of GTPases with GEF and/or their effectors, which are characterised by rather large molecular contact interfaces.

Thus, the low molecular compounds that have recently emerged as ligands of the hitherto ‘undruggable’ Ras proteins will be instrumental in modulating the GEF/GTPase-mediated signalling cascade that plays a fundamental role in conveying and amplifying biological signals. In the light of this progress, a Ras-targeted cancer therapy now does not appear an unreachable goal anymore and the supremacy of NMR spectroscopy combined with fragment-based drug design shall substantially contribute to pharmacological and biochemical studies of Rheb- and/or Ras-related malignancies in the (not too distant) future.