WO2023052462A1

WO2023052462A1 - Identification of a cryptic and druggable pocket in the active site of ret with therapeutic potential: the post-lysine pocket

Info

Publication number: WO2023052462A1
Application number: PCT/EP2022/077036
Authority: WO
Inventors: Iván PLAZA MENACHO
Original assignee: Fundación Del Sector Público Estatal Centro Nacional De Investigaciones Oncológicas Carlos III (F.S.P. CNIO)
Priority date: 2021-09-28
Filing date: 2022-09-28
Publication date: 2023-04-06

Abstract

The present invention relates to in vitro or in silico screening methods for compounds which inhibit RET kinase activity which bind to at least one of amino acids K758, L760, E768 and L772 of RET polypeptide.

Description

Identification of a cryptic and druggable pocket in the active site of RET with therapeutic potential: the post-lysine pocket

FIELD OF THE INVENTION

BACKGROUND

Protein kinases play a causative role in human disease and cancer when deregulated by oncogenic mutations or overexpression. A current hallmark for precision and personalized medicine is the development of highly specific protein kinase inhibitors that can be translated into the clinic for the successful treatment of cancer patients (1 ). To date, the U.S. Food and Drug Administration (FDA) has approved more than 30 kinase inhibitors that are used in the clinic to treat cancer and other human disorders.

However, these drugs target only a small percentage of the entire human kinome (5%) and they usually display non-specific crosstalk and lack of activity against drug-resistant secondary mutations.

The RET (REarranged-during Transfection) proto-oncogene encodes a receptor tyrosine kinase for members of the glial cell line-derived neurotrophic factor (GDNF) family of extracellular ligands (2). Oncogenic RET mutations and rearrangements result in constitutive ligand-independent activation of RET catalytic activity and signalling (2).

Both fusions and mutated RET are actionable drivers in non-small cell lung cancer (NSCLC) and thyroid cancers (3,4). In particular, RET rearrangements have been identified in ~2% of lung adenocarcinomas (NSCLC), 20% of papillary thyroid carcinoma (PTC), and less frequently in other types of cancer including breast, salivary gland intraductal carcinoma, pancreatic and colon cancers (3).

Rational and precise targeting of oncogenic drivers is a crucial hallmark in the cancer research field. Over the last years, the FDA approved the repurposing of several multityrosine kinase inhibitors (MKIs) with pharmacological activity against RET for the treatment of thyroid, lung adenocarcinoma and other RET-positive cancers. These inhibitors tested in the clinic had significant limitations due to non-selective activity against multiple kinases, dose-limiting toxicities as well as suboptimal target inhibition in the presence of resistance- associated secondary mutations, resulting altogether in modest survival outcomes in a limited subset of patients compared with other targeted therapies clinically successful. Overall, the clinical outcomes in response to RET-directed therapies were limited and modest compared with those achieved with other drugs targeting other oncogenes in solid tumors including EGFR, B-RAF, ALK, and ROS1.

Recently developed RET inhibitors LOXO-292 (selpercatinib) and BLU-667 (pralsetinib) exhibited >100-fold selectivity against a wide variety of RET oncogenic mutations in preclinical and clinical models (5-7). Data from the phase I clinical trial (ARROW) revealed that BLU-667 treatment resulted in a substantial response in RET-rearranged NSCLC patients with overall response rate (ORR) of 60% and disease control rate (DCR) of 100% (6). Additionally, clinical global phase l/ll trial of LOXO-292 (LIBRETTO-001 ) for RET rearranged-positive NSCLC showed substantial response as a potent inhibitor with ORR of 68%, median progression-free survival (PFS) of 18.5 months, and median duration of response of 20.3 months (5,7). LOXO-292 also demonstrated a high ORR of 91 % in NSCLC patients with central nervous system metastasis (8,9). In the registration dataset of RET- mutant medullary thyroid carcinoma (MTC), the acquired clinical trial data showed ORR of 56% and 53%, in multi-kinase inhibitors (MKI)-treated and MKI-naive patients, respectively. Another clinical study including 26 patients with RET-fusion positive PTC showed 62% ORR (10). As a consequence, the FDA recently approved LOXO-292 and BLU-667 for the treatment of cancer patients presenting oncogenic RET rearrangements or mutations (8,1 1 ). Despite such promising scenario, recurrent mechanisms of resistance to these selective inhibitors have already been described (12). In particular, refractory mutations located at the solvent front pocket.

There is a need for understanding these mechanisms of resistance as well as the structural dynamics and pharmacophoric features required for highly specific and effective RET kinase inhibition. Also, there is a need for the design and development of clinically successful compounds able to overcome refractory RET mutations.

DESCRIPTION OF THE INVENTION

The inventors have found a cryptic sub-pocket adjacent to the catalytic lysine of RET, defined by K758, L760, E768, and L772, that they called the post-lysine pocket, as a key structural determinant for efficient and selective RET kinase inhibition. This cryptic pocket has important consequences on RET tyrosine kinase activity. The identification of the postlysine pocket as a new druggable vulnerability in the RET kinase exploited by second generation RET inhibitors allows the screening of compounds and also drug design and the development of personalized therapies for patients with RET-driven cancers.

In a first aspect, the present invention relates to a method for testing the ability of a compound to inhibit RET tyrosine kinase activity comprising assessing the ability of said compound to bind to at least one of amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration. In a preferred embodiment, the method tests the ability of a compound to inhibit RET tyrosine kinase activity comprising assessing the ability of said compound to bind to amino acids K758, L760, E768 of RET.

As used herein, “RET” refers to proto-oncogene tyrosine-protein kinase receptor Ret, preferably to human proto-oncogene tyrosine-protein kinase receptor Ret, with reference number in Ensembl Database ENSG00000165731 and UNIPROT database P07949-2.

As used herein, “open glycine-rich loop and aC helix-in configuration” refers to the conformation or structural arrangement in which: i) GRL loop adopts an open position compatible with ATP binding, ii) aC helix adopts a catalytically competent inner position able to engage the catalytic salt-bridge between lysine (P3 K758) and (E775, aC). The skilled person can monitor these conformational modes by x-ray crystallography, as well as enzymatic and binding assays (DSF, ITC) measuring ATP binding and/or y-phospho- transfer.

In a preferred embodiment of the first aspect, the method comprises assessing the ability of said compound to bind to at least two of amino acids K758, L760, E768 and L772 of RET. In a preferred embodiment of the first aspect, the method comprises assessing the ability of said compound to bind to at least three of amino acids K758, L760, E768 and L772 of RET. In a most preferred embodiment of the first aspect, the method comprises assessing the ability of said compound to bind to amino acids K758, L760, E768 and L772 of RET. A preferred embodiment relates to a method for testing the ability of a compound to inhibit RET tyrosine kinase activity comprising assessing the ability of said compound to bind to at least one of amino acids K758, L760, E768 of RET, optionally in combination with L772 of RET, or any combinations thereof.

In a preferred embodiment of the first aspect, the method further comprises assessing the ability of the compound to bind to amino acid G810 and/or amino acid V738 of RET. In a preferred embodiment, the method further comprises assessing the ability of the compound to bind to at least one amino acid selected from L730, V738, A756, I788, V804, A807, L881 of RET, or any combination thereof.

In a preferred embodiment of the first aspect, the ability of the compound to bind said amino acids is tested by differential scanning fluorimetry (such as with the technology called Tycho of Nanotemper and SYPRO Orange), isothermal titration calorimetry or by Western Blotting of lysates from cells treated with increasing concentrations of the compounds or any combination thereof.

A second aspect of the present invention relates to a screening method for compounds able to inhibit RET tyrosine kinase activity comprising virtually screening if the compounds interact with at least one of amino acids K758, L760, E768 and L772 of RET, using RET crystal structure, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration. The screening method of the second aspect is an in-silico method. In a preferred embodiment, the method of the second aspect comprises virtually screening if the compounds interact with at least two of amino acids K758, L760, E768 and L772 of RET, using RET crystal structure. In a preferred embodiment, the method of the second aspect comprises virtually screening if the compounds interact with at least three of amino acids K758, L760, E768 and L772 of RET, using RET crystal structure. In a preferred embodiment, the method of the second aspect comprises virtually screening if the compounds interact with the four amino acids K758, L760, E768 and L772 of RET, using RET crystal structure. A preferred embodiment relates to a screening method for compounds able to inhibit RET tyrosine kinase activity comprising virtually screening if the compounds interact with at least one of amino acids at least one of amino acids K758, L760, E768 of RET, optionally in combination with L772 of RET, or any combinations thereof, using RET crystal structure. In a preferred embodiment, the method further comprises virtually screening if the compounds interact with at least one amino acid selected from L730, V738, A756, I788, V804, A807, L881 of RET, or any combination thereof.

RET crystal structure (Protein Data Bank (PDB): 2IVS GRL closed and PDB: 5AMN open, PDB: 4CKJ has both conformations) can be in the open or closed GRL configurations. In a preferred embodiment, both structures are used in the screening method.

A third aspect of the present invention relates to a method of screening for a compound capable of inhibiting the tyrosine kinase activity of RET, comprising assessing the ability of said compound to bind to at least one, at least two, at least three or four of amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration. The screening method of the second aspect is an in vitro method. In a preferred embodiment, the method of the third aspect further comprises comparing the binding of said compound with the binding of LOXO-292 and/or BLU-667 to amino acids K758, L760, E768 and/or L772 of RET. Preferably, the method of the third aspect comprises selecting the compound if its binding to amino acids K758, L760, E768 and/or L772 of RET is tighter than the binding of LOXO-292 and/or BLU-667. A preferred embodiment relates to a method of screening for a compound capable of inhibiting the tyrosine kinase activity of RET, comprising assessing the ability of said compound to bind to at least one of amino acids K758, L760, E768 of RET, optionally in combination with L772 of RET, or any combinations thereof. In a preferred embodiment, the method further comprises assessing the ability of said compound to bind to at least one amino acid selected from L730, V738, A756, I788, V804, A807, L881 of RET, or any combination thereof.

A fourth aspect of the present invention relates to a compound that binds to at least one of RET’s amino acids K758, L760, E768, and L772, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration, other than LOXO-292 or BLU-667. In a preferred embodiment, the compound binds to at least two, at least three or even more preferably the four of RET’s amino acids K758, L760, E768, and L772. Preferably, wherein said compound does not contact amino acids G810 and/or V738 of RET.

A fifth aspect of the present invention relates to the use of a compound of the fourth aspect to inhibit RET. A preferred embodiment is the compound of the fourth aspect for use in the treatment of RET-driven cancers, preferably of lung cancer, breast cancer or thyroid cancer. This aspect of the invention relates also to the use of the compounds of the fourth aspect for the manufacture of a medicament for the treatment of RET-driven cancers, preferably of lung, breast or thyroid cancers. This aspect also relates to a method of treating a subject in need thereof by administering a therapeutic effective amount of a compound of the fourth aspect of a pharmaceutically acceptable salt thereof.

A sixth aspect of the present invention is a polypeptide with an amino acid sequence comprising RET polypeptide sequence comprising mutation L760A or mutations L760/772A or at least two mutations selected from K758M, L760A and L772A, preferably comprising mutation L760A or mutations L760/772A. In a preferred embodiment said polypeptide has an amino acid sequence comprising RET polypeptide sequence comprising mutation L760A or mutations L760/772A. As explained above, RET polypeptide sequence is the one in Ensembl Database ENSG00000165731 and UNIPROT database P07949-2. A seventh aspect of the present invention is a nucleic acid with a nucleotide sequence comprising a sequence codifying for the amino acid sequence of the sixth aspect.

The methods of the first, second and third aspect may use the polypeptide of the sixth aspect of the invention or the nucleic acid of the seventh aspect of the invention. An eighth aspect of the present invention relates to a method for modifying compounds to inhibit RET, comprising the following steps: (a) Testing the binding of the compound to amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration; (b) Modifying the chemical structure of the compound; (c) Repeating steps (a) and (b) for 1 to 100 times; and (d) selecting the modified compound when is binds to amino acids K758, L760, E768 and L772 of RET or when its binding to amino acids K758, L760, E768 and L772 of RET is more specific and stronger than before the modification. The binding specificity and/or strength is measured by methods known by the skilled person, such as measuring binding capacity by DSF (Ti and Tm), affinity by isothermal titration calorimetry (ITC, (K_D) and IC50), in WBs of cells expressing wild type (WT) or the indicated mutants in dose-dependent experiments with the inhibitors. The modifications of step (b) are aimed at having more interactions, such as hydrogen bonds or electrostatic interactions, between the compound and the 4 amino acids of the post-lysine pocket.

As used herein, the term “comprises” also includes “consists”. All embodiments of all aspects can be combined and all embodiments of all aspects are embodiments of the other aspects.

DESCRIPTION OF THE DRAWINGS

Figure 1 . RET active site druggability landscape is determined by the dynamics of the GRL (A) Cartoon representation of RET KD crystal structure with two discrete conformations of the GRL (PDB 4CKJ). Hydrogen bonds and salt bridge interactions defining both open (upper) and close (lower) conformations. (B) Superimposition of the indicated RET KD crystal structures depicting side chains of F735 from the GRL and catalytic K758 (j33) and their corresponding distances (A) and defined volumes of the active site (A³) in each configuration. FT-site mapping of druggable pockets within the active site (color coded) of RET under different GRL-configurations: (C) Open (PDB 5AMN), (D) intermediate (PDB 2IVT and 2IVV) and (E) closed (PDB 2IVS) depicting residues contributing to each pocket, ABP (adenine-binding pocket), FP-II (front pocket-ll) and catalytic loop-HRD motif pocket (CP). Surface representation (C-D insets) with side-chain of residues defining a front subpocket- 11 defined by K758, L760, L772 and E775. Please note that a fully competent (available) pocket appears in the open crystal structure, being mutually exclusive with presence (occupancy) by the side chain of F775 from GRL. Figure 2. Extensive perturbation of the active site disrupts the closed tether and induces an open conformation of the GRL Surface representations of the L-RIP identified pockets (versus reference, closed structure PDB 2IVS) combining both trajectories.

Figure 3. Structural identification and dynamical characterization of a druggable post-lysine pocket (A) Surface representation of RET post-lysine pocket composites under different GRL-configurations with amino acid side chains represented in sticks: open (PDB 5AMN), intermediate (PDB 2 IVT) and closed (PDB 2 IVS). In the case of RET a fully competent post-lysine pocket is available when the side chain of residue 735 (GRL) is pointing away from the active site (i.e. GRL open) and E768 is pointing inwards. (B) Evolutionary conservation of post-lysine pocket composites using Consurf, closer view of the pocket (inset, indicates high evolutionary rate) (C) Protein sequence alignment with secondary structural elements of a set of protein tyrosine kinases which are targets of known RET TKIs for which structural information is available showing conservation of post-lysine pocket residues (*)The sequences correspond in consecutive order to SEQ ID NO: 9 TO SEQ ID NO: 26. (D) Cartoon representation of a superimposition of crystal structures from (C) with RET post-lysine pocket residues depicted (RET, VEGFR1 , VEGFR2, FGFR1 , FGFR2, FGFR3, FGFR4 and PDGFRa). (E) Schematic representation of a competent post-lysine pocket: left panel, array of different GRL-configurations (from versus RET aC-in. Right panel, array of different aC helices configurations versus RET open GRL.

Figure 4. LOXO-292 and BLU-667 target the post-lysine (A) Cartoon representation of RET KD crystal structure in complex with BLU-667 (PDB 7JU5) with secondary structural elements. (B) Close-up view of the active site in an open (PDB 7JU5) and superimposed closed GRL (PDB 2IVS) configuration showing post-lysine pocket residues. (C) 2D- pharmacophore representation of BLU-667 interactions: hydrophobic, hydrogen bond and electrostatic interactions (pi-cation). (D) Lateral view of B, without depicting the GRL. (E) Upper view from B, without depicting the GRL. (F) Cartoon representation of RET kinase domain crystal structure in complex with LOXO-292 (PDB 7JU6) with secondary structural elements. (G) Close-up of the active site in an open and closed GRL configuration showing post-lysine pocket residues. (H) 2D-pharmacophore representation of LOXO-292 interactions: hydrophobic, hydrogen bond and electrostatic interactions (pi-cation). (I) Lateral view of G, without depicting the GRL. (J) Upper view from G, without depicting the GRL. Superimposition of the RET active site (upper view) with gate-keeper and post-lysine pocket residues, in complex with: (K) ATP based on a crystal structure of FGFR2 (PDB 2PVF), LOXO (PDB 7JU6) and BLU-667 (PDB 7JU5). (L) ATP (PDB 2PVF), ZD7464 (PDB 2IVU) and BIBF1120 (PDB 6NEC). Figure 5. Functional evaluation of mutants targeting the post-lysine pocket WBs data of lysates from HEK293 cells ectopically expressing RET WT kinase or the indicated mutants treated with increasing concentrations of the inhibitors LOXO-292 and BLU-667 using the indicated antibodies. ZD6474 was used as an internal control for the RET KD V804M mutant.

Figure 6. Graphic summary First-generation RET inhibitors were multityrosine kinase inhibitors (TKIs) derived from secondary pharmacology targeting the adenine-binding pocket that resulted in poor clinical outputs. Recently developed second-generation RET inhibitors (derived from primary pharmacology), exploit in addition further vulnerabilities within the active site e.g. the post-lysine pocket. We define the structural and dynamical determinants conferring high selectivity to these inhibitors towards RET by targeting the post-lysine pocket, making them clinically successful.

Figure 7. Cartoon representation of crystals structures: RET KD in complex with (A) BLU- 667 (PDB 7JU5), (B) LOXO-292 (PDB 7JU6) and (C) ZD6474 (PDB 2IVU) and (D) KIT KD in complex with Ponatinib (PDB 4U01 ) with representations of R- and C-spine composing residues (side-chain in sticks and surface representations), colour coded secondary structural elements and functional motif together with depicted druggable pockets within de active site. ABP (Adenine binding pocket), GP (gate-keeper pocket), SP (solvent pocket), FP (front pocket), PKP (post-lysine pocket).

Figure 8. FT-site mapping of druggable pockets within the active site of (A) FGFR2 (PDB 2PVF) and (B) FGFR3 (PDB 6LVM). Inset, depiction of post-lysine pocket residue composite of FGFR2 and protein sequence alignment of RET closely related kinases FGFR1 -4 and PDGFRa showing conservation of post-lysine pocket residues (the indicated sequences correspond consecutively to SEQ ID NO: 9, SEQ ID NO: 18 - SEQ ID NO: 21 and SEQ ID NO: 23.

EXAMPLES

The following examples illustrate the present invention:

1. The druggability landscape of the RET active site is determined by the dynamics of the glycine-rich loop (GRL)

We explored the druggability landscape of the RET active site under different GRL configurations. Our rationale was driven by a previously solved high-resolution crystal structure of a RET Kinase domain (KD) displaying two discrete conformations of the ATP- binding loop. In the closed structure the F735 (GRL) side chain is clamped over the active site and the side chain of aC E768 points away from the cleft. This conformer was further stabilized by a triad of tethered residues between E734 (GRL), R912 (activation loop) and D771 (aC) (Fig. 1 A). In contrast, in the open structure, the F735 side chain was solvent- exposed through a large displacement of the loop from the active site, whereas E768 points inward into the cleft. The two different conformations are defined by the mutually exclusive configurations of E768 and F735 side chains, which restrict nucleotide binding and accessibility. The open structure is further stabilized by salt-bridge and hydrogen bonding interactions between residues in the GRL and the p3-aC loop (Fig. 1 A). First, we measured the conformational space within the active site of RET using several crystal structures capturing different GRL conformational states: 2IVS (closed), 2IVT (intermediate), 5AMN (open) and 4CKJ (open and closed), by calculating the center of mass distance between the P3 catalytic K758 side chain (NZ atom) and GRL E734 main chain (Co) and their corresponding active site volumes (figure 1 B, and table 1 ).

While distances (and volumes) in the closed and intermediate states were 7.6 A (646.623 A³) and 8.5 A (902.283 A³), respectively, the open state showed significant larger values of 15.0 and 16.6 A (1870.6 and 2311.7 A³). Transition of GRL to an open state was also associated with an increase in the distance between the gatekeeper residue V804 and the catalytic K758 (table 1 ). These data indicated that a larger druggability space within the active site of RET is available in the opened structure. Next, we mapped druggable regions within the active site of RET susceptible to conformational changes by the GRL using the FTSite web server (Fig. 1 C-E and table 2).

Table 1 . Distances between indicated residues and active site volumes from different RET kinase domain crystal structures with their corresponding ligand and GRL configurations.

Table 2 shows the residue composition of druggable pockets identified in RET KD using the FT-site mapping: Closed (PDB 2IVS)

ntermediate

ntermediate (PDB 2IVU)

Opened (PDB 5AMN)

Three pockets were identified and ranked within the active site of RET in the open structure (Fig. 1 C). The adenine-binding pocket (ABP) linked to the front pocket-l (FP-I), and gatekeeper pocket displayed average druggable scores. A small pocket adjacent to the catalytic loop-HRD (His-Arg-Asp) motif pocket (catalytic pocket (CP)) made up of D874, R878, N879 and P914 side chains together with R912 main chain was detected with low druggability score (see table 2 for details). Interestingly, a small pocket with maximum druggability potential was detected in the front-pocket-ll adjacent to the catalytic lysine defined by K758 (|33), L760 (P3-aC loop), E768 (aC), and L772 (aC). In the case of the intermediate structure (Fig. 1 D, right panel) the ABP and FP-I ranked as the top druggable pocket. The CP was also detected with averaged druggable potential. We noticed that the FP-I I was not detected as it was partially occupied by the GRL and F735 side chain (Fig. 1 D, right panel). This is in contrast with another intermediate RET crystal structure in which F735 side chain electron density is not defined (PDB 2IVV). In this situation the FP-II is partially accessible and appears as the top ranked druggable region (Fig. 1 D, right and central panels). In the case of the closed structure (Fig. 1 F) only the ABP and FP-I region appear as druggable. This is attributed to the degree of occupancy of the active site by F735 side chain and the GRL itself impeding the access to the front solvent pocket-ll. These data highlight: i) the existence of additional druggable vulnerabilities within the active site that were not exploited by first generation RET inhibitors, and ii) that optimal and selective RET kinase inhibition will depend on interactions other than those taking place at the adenine- binding and gatekeeper pockets.

2. Extensive perturbation of the active site disrupts the closed tether and expands its druggability landscape by inducing an open GRL

To simulate the perturbation of the active site of RET by an inhibitor able to exploit an extensive druggable space and its impact on the transition from a closed to an open GRL, we applied RIP (rotamerically induced perturbations) MD simulations methods. First, we used the RIPlig (RIP by a pseudo-ligand) method to identify large conformational changes within the active site. In the case of RET, this approach resulted in a significant 10 A displacement of the GRL from a closed tether to a fully open solvent accessible configuration of the GRL as indicated by a large displacement of E734. These data were in good agreement with the crystal structure data and the rationale of our study. Next, we applied the Langevin-RIP (L-RIP) approach by perturbing the dihedral angles of residues E734 and F735, both key determinants from our MD and X-ray data, to evaluate their specific contribution in GRL transition. Perturbation of the dihedral angles of these two residues resulted in intermediate GRL configurations associated with significant changes in the druggability potential and modification of active site physicochemical properties. Whereas perturbation of E734 resulted in lower and less consistent conformational changes in the GRL and druggability, as indicated by volume and exposure values. Other physicochemical properties e.g. hydrophobicity, hydrogen acceptor, and negative ion dependencies were increased in some particular frames. On the other hand, F735 perturbation resulted in higher druggability potential and the induction of complex conformational changes, as evidenced by higher volume and exposure values, which in turn resulted in increased hydrophobicity, hydrogen acceptor, and negative charges of the active site. Next, we found that residues from the GRL (aa 730-739), aC helix (aa 766-773), P3- aC loop (aa 756-760) and hinge (aa 805-810) regions showed highly dynamic behavior indicative of significant GRL motions by the and its coordination with other secondary structural elements such the aC helix, P3- aC loop and hinge. Furthermore, the two trajectories revealed several druggable pockets within the active site with high occurrence throughout the simulation (Fig. 2) that matched the druggable space observed in the open and intermediate crystal structures by the FTSite method comprising the solvent front pocket (ABP), front pocket-l (FP-I) and front pocket-ll (FP-II). These results were in good agreement with the FTSite analyses and showed that perturbation of the closed autoinhibitory tether causes a large conformational motion of the GRL that expands the druggability landscape of the RET kinase active site.

3. Structural identification and dynamic characterization of a druggable post-lysine pocket

The FTSite analyses revealed a sub-pocket adjacent to the catalytic lysine consisting of K758 (P3), L760 (P3-aC loop), E768 (aC), and L772 (aC) that we name post-lysine pocket (Fig. 1 C). The pocket is defined in a central axis by hydrophobic L760 and L772, which are flanked on one side by catalytic K758 and E768 on the other (Fig. 1 C-D). While in the closed and intermediate configurations F735 points to the center of the pocket, in the open structure F735 points away from the cleft and the side chain of E768 adopts and inner position together with K758, defining a fully accessible post-lysine pocket (Fig. 3A). We consistently observed an invariant aC helix-in (active) in the different configurations with a proper alignment of the regulatory and catalytic spine residues and DFG-in conformation (DFG, i.e.: Asp-Phe-Gly) (Fig. 7). This newly identify druggable pocket appears to be evolutionary conserved (Fig. 3B). We examined the presence and conservation of the FP- II and post-lysine pocket in other protein kinases that are pharmacologically inhibited by RET tyrosine kinase inhibitors for which crystal structure data are available (Fig. 3C), including ALK (PDB 4TT7), VEGFR1 (3HNG), VEGFR2 (3VHE), MET (3DKC), Tie-2 (2OSC), TrKB (4ASZ), Axl (5U6B), ROS1 (3ZBF), FGFR1 (4V05), FGFR2 (1 GJO), FGFR3 (6LVM), FGFR4 (4XCU), c-Src (4U5J), PDGFRa (6J0L), Abl (3IK3), EGFR (5Y9T) and B- RAF (3C4C). From this set of structures, we looked at the conservation of the residues defining the post-lysine pocket. RET F735 (GRL) and L760 (P3-aC loop) were highly conserved residues within the dataset, with exceptions only in c-Abl (Y253) at the equivalent position to F735 and Tie-2 (M857) and Axl (M569) in the equivalent position to L760. An acidic residue equivalent to RET E768 (aC) lacked conservation in TrKB (A597), c-Src (S303), EGFR (A755) and B-RAF (Q494). RET L772 was substituted by phenylalanine in the case of ALK, Tie-2, T rKB, Axl, Ros, c-Src, Abl, B-RAF or isoleucine in the case of EGFR. These data indicate that the conservation of residues defining the post-lysine pocket occurred within the closest RET phylogenetic group of receptor tyrosine kinases including FGFR1 , FGFR2, FGFR3, FGFR4, VEGFR1 , VEGFR2, and PDGFRa (Fig. 3C and 8). Interestingly, the majority of the indicated structures have a GRL-closed and/or an aC-out configuration relative to RET, which results in a non-accessible post-lysine pocket (Fig. 3D- E). Using the FTSite server, a druggable post-lysine pocket was only detectable in structures with aC-in and open GRL conformers including RET and FGFR2 (PDB 2PVF). In the latter, however, despite displaying an accessible post-lysine pocket the front solvent pocket-ll did not appear as druggable, probably as a consequence of the aC being slightly shifted toward an intermediate position (Fig. 8A). To further corroborate the dynamic crosstalk between the GRL and aC helix to restrict or allow access to the post-lysine pocket, we analyzed the druggable pockets in the FGFR-3 active site using the FTSite server. The catalytic domain of FGFR3 (PDB 6LVM) displayed a very similar conformation to FGFR2 with RMSD value of 0.6 A, with aC-in but closed GRL-configuration. As predicted, neither the front solvent pocket-ll nor post-lysine pocket were identified as druggable sites in the FGFR3 structure (Fig. 8B). These data demonstrated that conservation of the post-lysine pocket is a common feature of RET closest RTKs phylogenetic group, but this fact is not sufficient for occupancy as dynamic inputs from GRL and aC restrict access to the pocket. A fully accessible and druggable post-lysine pocket requires a GRL-open and aC-in configuration that are only seen in RET kinase crystal structures, as illustrated in figure 3E.

4. LOXO-292 and BLU-667 target the post-lysine pocket

We explored the binding mode of highly specific RET inhibitors LOXO-292 and BLU-667 under different GRL configurations. The two compounds were able to accommodate properly into the active site of RET only in an open state. Superimposition with intermediate (2IVV) or closed (2IVS) RET structures (Fig. 4A and B) demonstrated that both compounds would clash with the loop in those settings in which compounds are restricted to accommodate only onto the adenine-binding pocket, as seen in the crystal structures of RET KD in complex with vandetanib and PP1 . Opening of the GRL results in suitable occupancy of the FP-II and the post-lysine pocket by LOXO-292 and BLU-667, respectively (Fig. 4A and B). Despite their atypical binding modes, both inhibitors do bind into a DFG-in (active) configuration with proper alignment of the regulatory and catalytic spines contrary to type II inhibitors, which perturb such alignment of the spines, see figure S1 .

BLU-667 (pralsetinib) targets the post-lysine pocket by accommodating the 4-fluoropyrazole ring into the patch forming a pi-cation interaction with the catalytic K758 and the fluorine group forming hydrophobic contacts with post-lysine pocket residues L760 and L772. Three hydrogen bonds were formed with E805 and A807 from the hinge region by the 5-methyl- pyrazol group with additional hydrophobic interactions with A756, V804, L881 , I788 and V738. The methyl-pyrimidine group forms hydrophobic contacts with L730 in addition to coordinating with one water molecule via the N1. Of note there are two other water molecules coordinated with the compound: one with N2 of the 4-fluoropyrazole group, and the other with the methoxy oxygen bounded to the cyclohexane ring (Fig. 3A).

LOXO-292 (selpercatinib) targeted the post-lysine pocket by accommodating the 2- methoxypyridine ring and forming also a pi-cation interaction with K758. The pyrazolo[1 ,5- a] pyridine group forms an additional hydrogen bond with the main chain nitrogen atom of A807 at the hinge and hydrophobic interactions with L730 and L881 .

Further hydrophobic contacts were made by the central pyridine group and V738. Two water molecules are coordinated with the oxygen atoms from the 2-methoxypyridine group and the pyrazolo[1 ,5-a] pyridine ring, respectively.

Both compounds appeared to exploit extensive ATP binding mimicry when compared to first-generation RET inhibitors whose prototypical interactions are mostly limited to the hinge and adenine-binding site. In fact, both compounds could only accommodate into the active site in an open GRL-state (Fig. 4 A-B) as the GRL would clash with the compounds in the closed or intermediate configurations, in a manner reminiscent to what it does with ATP. First generation type I RET inhibitors like vandetanib (ZD6474) and nindetanib (BIBF1120) get accommodated parallel to the hinge and perpendicular to the inner part of the aC helix (Fig. 4 L and K). On the other hand, second-generation RET inhibitors accommodate in the path towards the GRL below and in diagonal from the hinge across the proximal aC helix exploiting other druggable hotspots reaching to the solvent pocket, FP-I and FP-II (Fig. 4 L and K). Taking as a reference the crystal structure of FGFR2 in complex with ATP (PDB 2PVF), we observed the adenine group of ATP forming hydrogen bonds with E565 and A567 at the hinge mirroring the interactions of RET A807 and Y806 hinge residues with the pyrazolo-pyridine and methyl-pyrazol groups from LOXO-292 and BLU-667, respectively. Furthermore, hydrogen bonds between FGFR2 main chain N571 and R630 atoms and the ribose of the adenosine moiety were formed. In the case of RET equivalent residues e.g. L881 made hydrophobic contacts with the inhibitors and in the case of S81 1 it was shown to form a water bridge with the quinazoline group of vandetanib. In FGFR2 the catalytic K515 (as part of the post-lysine pocket) form a salt bridge and hydrogen bonds with a and p phosphate groups, whereas in the case of RET the equivalent K758 formed pi-cation interactions with fluoropyrazole and methoxypyridine rings from BLU-667 and LOXO-292 respectively. FGFR2 F492 (equivalent to RET F735) and A491 (RETG733) main chain atoms formed hydrogen bonds with the y-phosphate group of the ATP, and this was mirrored in the case of LOXO-292 crystal structure by two coordinated water molecules interacting with main chains G733 and E734 atoms from the GRL. Altogether, these data demonstrate that LOXO-292 and BLU-667 target the post-lysine pocket by promoting and open GRL conformer and exploiting extensive ATP mimicry a feature not observed before with inhibitors of the first generation.

5. Molecular dynamics and stability of apo and complexed structures

The stability and dynamics of RET kinase in the apo (PDB 2IVS) and complexed with LOXO- 292 (PDB ID 7JU6), BLU-667 (PDB ID 7JU5) and ZD6474 (PDB ID 2IVU) were investigated using a 100-nsconventional MD trajectory.

The root mean-square deviation (RMSD) of protein backbone atoms (C-N-Ca) was computed to assess the stability of each protein-ligand complex system with respect to their initial frame. The initial inspection of the computed RMSD profile showed that all systems were equilibrated during the simulation. The RMSD of the protein backbone was fluctuating just below 2 A for the complexed systems with LOXO-292, BLU-667 and ZD6474 which shows that the systems were stable throughout the simulation process.

In contrast to the complexed ensembles, the RET apo system displayed significant fluctuations during the simulation process with a maximum RMSD value of 2.3 A. In the case of LOXO-292 and BLU-667, and contrary to ZD-6474, the molecule adopted an opened GRL conformation, which is required for full access and accommodation into the post-lysine pocket.

T o test the influence of ligand binding on protein flexibility, the root mean-square fluctuations (RMSF) parameter of the protein backbone atoms was computed throughout the simulation. The apo system displayed sharp peaks in the pi - P2 (residues 726-738), and p3-aC loop (residues 760-767) regions and a relative high flexibility in the residues 907-914 of the activation loop. However, RET LOXO-292 and BLU-667 complexed systems showed a remarkable flexibility reduction in the indicated regions, especially those defining the GRL. These data together prove that LOXO-292 and BLU-667 have a direct impact on reducing the flexibility of the ATP-binding loop by forcing it to an open conformation that is required for post-lysine pocket occupancy.

Next, the MM/GBSA (molecular mechanics Boltzmann surface area) method was implemented to estimate the stability and binding free energies of the inhibitors (Table 3). The computed binding free energy (AG total) of RET-LOXO-292 and BLU-667 complexed systems were -55.3 ± 3.3 and 51 .5 ± 3.0, respectively, which were significantly lower than ZD6474 (-43.6 ± 2.3) indicative of higher affinity binding by the second-generation RET inhibitors. This was further corroborated by the per-residue energy decomposition analysis identifying the residues that contributed significantly to the ligand binding through intermolecular interactions (table 3).

Table 3. Molecular mechanics Boltzmann surface area (MM/GBSA) method va ues for the stability and binding free energies of the indicated inhibitors. The computed binding free energy of RET LOXO-292 and BLU-667 complexed systems were respectively significantly lower than ZD6474, indicative of higher affinity binding by second generation RET inhibitors

For LOXO-292 and BLU-667, the major favourable energetic contributions in the post-lysine pocket originated from the catalytic K758 with energy values of -6.4 Kcal/mol and -3.5 Kcal/mol, respectively. These substantial energetic values were derived from the intense TT- cation interactions with the K758 side chain. Both inhibitors displayed significant hydrophobic contacts with post-lysine pocket residue L760. In contrast to BLU-6676 4- fluoropirazole group, LOXO-292 displayed relatively higher interaction with L772 by the 2- methoxypyridine occupying the post-lysine pocket with an energy value of -1.88 Kcal/mol. Strong hydrophobic interactions of both inhibitors with L730 (pi ), V738 (|32), Y806 (hinge), and G810 (front solvent pocket) further stabilize the inhibitors in the adenine binding pocket and the solvent front pocket. Interestingly, refractory mutations at these sites that reduce the affinity of these inhibitors and confer resistance have been already identified in patients.

6. Functional evaluation of mutants targeting the post-lysine pocket reveals an impact on both inhibitor sensitivity and RET tyrosine kinase activity In order to investigate the impact of post-lysine pocket residues on inhibitor binding and cellular response, we applied a multidisciplinary approach combining protein biochemistry, biophysics and cell-based assays. First, we generated recombinant RET KD WT and post- lysine pocket variants K758M, L760A, L772A and L760/772A.

Table 4. Differential Scanning Fluorimetry (DSF) data measured by i) a direct method based on intrinsic fluorescence (IF) providing inflection temperature (Ti,) and SYPROTM Orange providing melting temperatures (T m). Data shown mean ± standard error of the mean (SEM) are from 4 to 8 independent experiments.

In addition, we also generated a KD V804M mutant as a negative control for the binding of a prototypical type-l inhibitor e.g. ZD6474. Next, we applied two independent differential scanning fluorimetry (DSF) methods: i) a direct method measuring intrinsic fluorescence upon a fast temperature gradient providing an inflection temperature (Ti), and ii) an indirect method based on SYPRO™ orange dye that provides melting temperatures (Tm,). We measured the effect of LOXO-292 and BLU-667 binding on the thermal stability of the apocontrol versus complexed proteins (table 4), ZD-6474 was also tested as a negative control for the gatekeeper mutant V804M. While both LOXO-292 and BLU-667 binding to RET KD WT conferred a significant increment in the protein thermal (deltaTi/m) of about 10 ^eC. Mutations of some of the post-lysine pocket composites resulted in a significant effect on the binding of the inhibitors, in particular we observed a detrimental LOXO-292 binding effect to L772A with a lower Tm of 7.0 ^eC (vs 9,5 ^eC control WT), whereas the double mutant L760/772A had a larger BLU-667 Tm of 14.0 ^eC (vs 1 1 ,6 ^eC control WT). Furthermore, the K758M mutant displayed a significant thermal stability increment by LOXO-292 and BLU- 667 of 1 1 ,9 and 14,2 ^eC (vs 9,5 and 11 ,6 ^eC control WT, respectively). The results from the K758M and double L760/772A mutants were surprising to some extend as we anticipated that perturbation of the post-lysine pocket composites would result in a detrimental effect on the binding on these inhibitors. Next, we tested the effect of post-lysine pocket mutants on the phospho-tyrosine kinase activity of the RET KD in vitro in time course autophosphorylation assays. We found that, contrary to the L760A mutant, which displayed comparable levels to the WT, RET L772A had a loss of function effect on the tyrosine kinase activity as indicated by both total phospho-tyrosine and phospho-specific RET Y905 antibodies. The same detrimental effect was observed with the double L760/772A mutant. Finally, in order to recapitulate these results in cell-based assays we used HEK293 cells ectopically expressing a full-length RET receptor with a C634R mutation in the extracellular domain, which bypasses the GDNF ligand and co-receptor requirements for activation (with intact intracellular domain) subjected to a dose-dependent treatment with LOXO-292 and BLU-667 for 90 min (Fig. 5). While treatment of RET WT KD with increasing concentrations of LOXO-292 and BLU-667 resulted in a consistent inhibition of RET auto-phosphorylation and ERK1/2 downstream activity at 10 nM, no significant differences were observed with the L760A mutant. In the case of the L772A mutant, as anticipated from the kinase assays we found a loss of function effect. Strikingly the double L760/772A mutant had a rescue effect on the tyrosine kinase activity by the L760A mutation, displaying also increased sensitivity to LOXO-293 and BLU-667 as 1 nM concentrations of the inhibitors resulted in a significant reduction on RET auto- and ERK1/2- phosphorylation levels. As a control we used RET KD V804M, which was sensitive to both LOXO-292 and BLU-667 at similar concentrations than the WT, but was resistant to ZD6474 treatment. These data were consistent with the DSF results, and showed the important implications of the post-lysine pocket composites on both inhibitor sensitivity and RET tyrosine kinase activity.

7. Virtual Screening

Virtual screening (VS) using the in house 50K CNIO virtual library was performed with no filters applied (No Lipinski'2 Rule of 5) using Rigid docking approach. The screening was performed using RET crystal structures with different P-loop configurations: open (PDB 5AMN) and closed (PDB 2IVS) in order to segregate between compounds able to target the cryptic pocket (i.e. post-lysine pocket) and compounds that target other un-restricted areas in the active site.

With this VS, we aimed at identifying in silico hits targeting previously identified druggable and cryptic-pockets within the RET active site, at performing molecular dynamics (MD) simulations to evaluate stability of the binders, at characterizing the initial binding by DSF and at chemically designing enhanced binders to the post-lysine pocket. We have collected a series of potential hits (60-70) from our initial VS strategy for the open P-loop configuration of the kinase domain. We have ranked them based on docking, XP (extra precision) glide scores and inspection of their capacity to bind to areas other than the adenine and gatekeeper binding areas e.g. post-lysine pocket among others using structure based VS, then selection of compounds using Docketing score, XP glide score, then studying protein-ligand interactions, performing 100 ns MD simulations and then doing free energy calculations using MM/GBSA (Molecular Mechanics/Generalized Boltzmann Surface Area).

We performed 100 ns MD simulations with top-ranked compounds and calculated their binding free energies by the MM/GBSA method. LOXO-292 and BLU-667 were used as internal controls.

Table 5. Molecular Mechanics Generalized Boltzmann Surface Area (MM/GBSA) method values for the stability and binding free energies of the indicated inhibitors. Data energy unit: Kcal/mol.

The root mean-square deviation (RMSD) of the protein Ccr atom with respect to the initial frame was computed to evaluate the stability of each protein-ligand complex system compared with the apo state. Initial examination of the computed RMSD profile showed that despite initial fluctuations all the complexed systems equilibrated with average fluctuations values of 2 A or below. The per-residue energy decomposition for RET apo, LOXO-292, BLU-667 and ZD6474 complexes throughout the simulation showed highly energetic interactions for some residues, if we take e.g. LOXO-292 and BLU-667 as control or references.

We performed initial evaluation of binding of these inhibitors by DSF using both direct intrinsic fluorescence (direct) and an indirect method using SYPRO™ Orange with a WT RET KD and an oncogenic deltaCCDC6-RET construct with a consistent increment in the thermal stability of 2.3 and 3.19 °C.

Table 6. DSF data by SYPRO™ Orange in thermal stability assays depicting Tm (°C) for RET Kinase Domain. ETP4 has a tighter binding to the RET kinase domain (wild-type and mutant variant) as indicated by higher deltaTm and Ti. Binding affinities are determined by isothermal titration calorimetry (ITC). Also, tumour cells lines driven by RET mutations (MZ-CRC-1 and MTTC- TT) are treated with the hit compounds to determine inhibition of auto-phosphorylation and cellular effect on the proliferation.

Discussion

Extensive analysis of the active site showed that potent and selective RET kinase inhibition requires the exploitation of vulnerabilities beyond the occupation of the adenine- binding pocket and gate-keeper vicinity. The inventors have found that intrinsic flexibility of the glycine-rich loop (GRL) and the aC helix reshapes the druggability landscape in the RET active site (Fig. 1 ). While in the closed GRL structure only the adenine-binding, gate-keeper and solvent pockets were potentially druggable, in the intermediate and open GRL structures the FP-II appeared with a significant higher druggability potential relative to the adenine-binding site. This pocket is distinguished by the existence of a small sub-pocket adjacent to the catalytic lysine defined by K758, L760, E768, and L772 that we defined as the post-lysine pocket. This new druggable pocket was fully accessible in the case of the open RET structures. The assembly of these residues and the accessibility to the post- lysine pocket was regulated by the position of F735 side chain and coordinated by the dynamics of the GRL and the aC helix. An interesting observation was that, as part of the adenine-binding site, the gatekeeper sub-pocket was found only in the structures with an intermediate and opened GRL conformer. This is related to the F735 transition away from the post-lysine pocket, which creates more space for the K758 side chain rotamers.

The recently discovered S904F acquired mutation in the activation loop of RET resulted in ZD6474 resistance. The crystal structure of the RET KD S904F mutant (PDB ID 6FEK) revealed a closed GRL conformer with an unusual K758 side chain rotamer with NZ atom pointing towards the gatekeeper pocket. A superimposition of that structure with the RET KD complexed to ZD6474 (PDB ID 2IVU) revealed steric clashes between the inhibitor and K758 side chain. This is further supported by a shorter V804-K758 distance when compared to other closed GRL structures (table 1 ). In addition, long-unbiased MD simulation of RET KD WT and mutant S904F complexed with ZD6474 revealed higher energy state of the mutant kinase with intermediate GRL conformer with F735 pointing towards the active site. As a result, we anticipate that K758 regulates the accessibility to the gatekeeper pocket as a result of the crosstalk between the gatekeeper and the post-lysine pockets being regulated by the position of F735 side chain. RIP MD simulations to show that extensive (RIPlig) and site-directed (L-RIP) perturbation of the GRL disrupts the close tether and induces an opening of the GRL conformation and active site that consequently expands the druggability landscape, being F735 the key determinant for such transition (Fig. 2). Furthermore, we examined the presence and conservation of the FP-II and post-lysine pocket in other protein kinases that are pharmacologically inhibited by RET tyrosine kinase inhibitors for which structural information was available (Fig. 3). We found that conservation of the post-lysine pocket is also a common feature of the phylogenetically closer group of RTKs (i.e. FGFR1 -4, VEGFR1 -2), but is not sufficient for occupancy and competency as dynamic inputs from GRL and aC restrict access to the pocket. A fully accessible and druggable post-lysine pocket requires of a synchronous GRL-open and aC-in configuration only seen in RET crystal structures.

Superimposition of crystals structures of the RET kinase in complex with LOXO-292 and BLU-667 with previous RET structures in complex with type-l inhibitors showed that both compounds could only fit into the active site in an open GRL-state (Fig. 4) as the ATP-loop would clash with the compounds in the closed or intermediate configurations, in a reminiscent manner to ATP. The free energy calculations by the MM/GBSA method shows that LOXO-292 and BLU-667 forms important interactions with L730 (b1) and V738 (b2), both from the GLR, and also with hinge residues Y806, K808 and G810. Interestingly, refractory mutations in those sites have been found in patients and resistant cell lines and clones. This is due to steric clashes with the inhibitors caused by the replacement of bulky amino acid chains in those regions, as well as the disruption of strong intermolecular hydrophobic contacts as shown by MM/GBSA analysis.

Development of next generation RET inhibitors will require optimal design of chemical moieties with less contacts with the solvent pocket and P3.

Functional characterization of mutants targeting the post-lysine pocket revealed a dual role on drug specificity and tyrosine kinase activity. The results from the double mutant (L760/772) and K758M were surprising as we anticipated that perturbation of the post-lysine pocket composites would result in a priory detrimental effect on the binding on these inhibitors. However, a tighter binding was observed in these cases as indicated by increased Tm and Ti (table 4). This can be potentially explained by the fact that substitutions by shorter side chains may allow a better accommodation of the compounds into the post- lysine pocket, resulting in a sensitizing effect as seen with the double mutant (L760/772A) (Fig. 5). In auto-phosphorylation assays using recombinant isolated RET kinase, the L760A mutant appeared to have higher background phosphorylation levels which were slightly enhanced over the time course compared with WT. These results suggest the L760A mutant could be a better substrate for Sf9 endogenous kinases, as seen before by both oncogenic mutations targeting the kinase domain (in particular M918T) and mutants disrupting the closed tether. On the contrary, L772A and L760/772A both had a significant detrimental effect on RET phospho-tyrosine kinase activity. When the same mutants were evaluated in a RET full-length context in dose-dependent cell-based assays (Fig. 5) we obtained results in line with the recombinant protein data e.g. a loss of function effect by the L772A mutation. Strikingly we also found a rescue effect by the L760A mutation and increased sensitivity by the double mutant L760/772A. These data were supported by the binding and DSF data where both L760/772A and the K758M mutants displayed a significant increment in their Tm and Ti upon inhibitor binding (Table 4). These results suggested that mutating specific post-lysine pocket components to alanine favored the binding potentially by allowing a better accommodation of the 2-methoxypyridine (LOXO- 292) and 4-fluoropyrazole (BLU-667). Furthermore, the striking effect of the double mutant in terms of rescuing the null effect of the L772A mutation indicates a potential crosstalk between thej33-aC loop and the aC helix, together with L772, which forms part of a recently described PIF-like pocket in RET. In both cases, there are clear signs of allosteric inputs to the catalytic site by regulating also R- and C-spines assemblies.

MATERIAL AND METHODS

Mapping of ligand binding site and hotspot residues

The FTSite server was used to explore the druggable pockets within the active site of RET in crystal structures with different GRL-conformations including closed- (PDB 2IVS), intermediate- (PDB 2IVT), and open- conformers (PDB 5AMN).

Transient pocket analyses

Mapping of transient pockets within the RET active site was performed by T ransient Pockets in Protein (TRAPP) webserver. The overall workflow of the TRAPP webserver consists of three stages: i) ensemble of the generated structures, ii) superimposition and clustering and iii) detection and characterization. The TRAPP structure module contains several simulation methods for the generation of protein ensembles. The pseudo-ligand (RIPlig) and Langevin rotamerically induced perturbations (L-RIP) MD-based methods were used for the generation of protein ensembles. Following each perturbation, the structures were relaxed for 0.6 ps in an implicit solvent MD simulation coupled to a Langevin thermostat. The TRAPP analysis module was used to align and superimpose the generated structures using the backbone of the previously chosen binding pocket residues using the RMSD metric, and clustered using a hierarchal algorithm with an RMSD threshold value of 3 A. The TRAPP pocket module was employed to identify transient regions within the active site. The protein cavities near the binding pocket are calculated and saved on the grid. Furthermore, the physicochemical properties of the side chain residues in the detected cavities, surface area, and pocket are computed by this module.

Conventional MD simulation

MD simulation was performed using the Amber 16 software package with GPU acceleration. In particular, the ff 14SB and GAFF (generalized Amber force field) for proteins and ligands, respectively, and the TIP3P model was chosen for water molecules. The partial charges of each ligand were calculated implemented in the AM1 - BCC charge method using the Antechamber module from AmberTools 16. Proteins were protonated at pH 7.4 and each molecular ensembled was immersed in a cubic box with a separation margin from the surface of the solute of 10 A. The system was electroneutralized by addition of sufficient Cl- counterions. The long-range electrostatic interactions were treated by the Particle Mesh Ewald (PME) method, while the short- range interactions; electrostatic and van der Waal, were calculated with a distance cut- off of 8 A. The SHAKE algorithm was applied to constrain the bond lengths involving hydrogen atoms to their equal volumes. The integration time step was 2fs. The systems were subjected to two minimization stages (a total 5000 steps). The system coordinates were saved every 10 ps. The generated trajectories were analyzed using CPPTRAJ from the AmberTools 17. Finally, hydrogen bond analysis was carried out by VMD hydrogen bonds tools with distance and angle cut-offs of 3.0 A and 135 degrees, respectively. All the plots were generated by Gnuplot and Python matplotlib.

Binding free energy calculations using MM-GBSA method

The Molecular Mechanics Generalized Boltzmann Surface Area (MM-GBSA) method was employed for binding free energy calculations using an implicit solvent model. This method additionally allows the energy decomposition analysis, which provides detailed information about the residual energetic contributions. The free energy of ligand binding to the receptor to form a complex is estimated using the following equations:

AG(bind) = G(RL) - G(R) - G(L) t is further decomposed into several contributed interactions:

AG(bind) = AH - TAS = AE(MM) + AG(_Soi) - TAS

In which: AE(MM) =AE(int) +AE(ele) +AE(vdw)

AG(soi) = AG(PB/GB> + AG(SA)

AG (SA) = y. SAS A + b

AG bind is the interaction energy between the receptor and ligand in a vacuum, equivalent to the sum of polar (AG PB/GB) and non-polar (AG SA) interactions between the solute and the continuum solvent model. The GB model used for the calculation of polar interactions provides an analytical expression of the polar interactions, which is faster than the Poisson Boltzman (PB) method. The Generalized Boltzman (GB) model (ig=2) was used to measure the electrostatic solvation energy. The per residue energy decomposition method implemented in the MM/GBSA was used to analyze the residual energy interaction contribution to the total binding free energy. This method considers the intermolecular and solvation energies without the inclusion of the entropy.

Site directed mutagenesis

Site directed mutagenesis was performed on a pBac-PAK-RET kinase domain (KD, aa 705- 1013) codon optimized construct (ref) as template using a modified Q5-polymerase- based protocol in which complementary primers were employed (see below) followed by Dpnl treatment at 37 degrees for at least 120 min before transformation in Q5-DH5a bacterial strain. Primers used (codon optimized sequences) were:

L760A_forward 5'-TGTGAAGATGgcgAAGGAGAACGCTTCCC-3' (SEQ ID NO: 1 ) L760A_reverse 5'-GGGAAGCGTTCTCCTTcgcCATCTTCACA-3' (SEQ ID NO: 2) L772A_forward 5'-GCTGCGTGACgcgCTGTCCGAGTTC-3' (SEQ ID NO: 3) L772A_reverse 5'-GAACTCGGACAGcgcGTCACGCAGC -3' (SEQ ID NO: 4) V804M forward 5'-GCTGCTGATCatgGAGTACGCTAAG-3' (SEQ ID NO: 5) V804M_forward 5'-CTTAGCGTACTCcatGATCAGCAGC-3' (SEQ ID NO: 6) K758M_forward 5'- CCGTGGCTGTGATGATGCTGAAGG (SEQ ID NO: 7) K758M reverse 5'-CCTTCAGCATCATCACAGCCACGG (SEQ ID NO: 8)

Expression and purification of recombinant RET KD

Expression of (codon optimized) RET KD (WT and indicated mutants) was performed in Sf9 insect cells using a baculovirus system following already established and published protocols. Protein purification was performed by tandem IMAC (Ni+2) and Glutathione- beads gravity flow chromatography and in-gel 3C-protease digestion followed by sizeexclusion chromatography when indicated. Differential Scanning Fluorimetry (DSF)

To evaluate the thermal stability of RET kinase WT and indicated mutants in the absence of (apo) and in complex with inhibitors LOXO-292, BLU-667 and ZD6474 (vandetanib) as control, we applied two different scanning fluorimetry methods. First, an indirect SYPRO Orange-based method. For this assay the total reaction volume was adjusted to 40 pL at 1 - 2 pM protein, 10 pM inhibitor, and 2xSYPRO Orange concentrations subjected to a gradient of temperature from 20 to 95 °C. Fluorescence was measured on an Applied Biosystem 7300 Real-Time PCR system. Second, a direct method based on changes in intrinsic fluorescence upon a quick gradient of temperature was measured using a tycho nanotemper instrument following manufacturer's instructions.

Western-blotting

For auto-phosphorylation assays with recombinant proteins, 2.5-5 mM of isolated RET KD (WT and indicated mutants) was incubated with 2 mM MgCI2 and 1 mM ATP for the indicated time points, after which samples were mixed with 5X sample loading buffer and boiled for 5 min. For cell lysates, transfected HEK293 cells subjected to the indicated concentrations of drug treatment were lysed in ice-cold 50 mM Tris pH 7.5, 150 mM NaCI, 1 mM DTT, 5% glycerol, supplemented with a cocktail of protease and phosphatase inhibitors and the total cell extract was centrifuged for 15 min at 5000-6000 rpm, after which soluble sample was mixed with 5x sample buffer and boiled for 5 min prior protein quantification with Bradford. Equal amounts of samples were run in SDS-PAGE gels and subjected to immunoblotting using the indicated antibodies.

REFERENCES

1 . Zhang, J., Yang, P. L., and Gray, N. S. (2009) Nat Rev Cancer 9, 28-39.

2. Plaza-Menacho, I. (2018) Endocr Relat Cancer 25, T79-T90.

3. Kohno, T. et al. (2012) Nat Med 18, 375-377.

4. Lipson, D. et al. (2012) Nat Med 18, 382-384.

5. Drilon, A. et al. (2020) N Engl J Med 383, 813-824.

6. Subbiah, V. et al. (2018) Cancer Discov 8, 836- 849.

7. Wirth, L. J. et al. (2020) N Engl J Med 383, 825-835.

8. Wright, K. M. (2020) Oncology (Williston Park) 34, 406- 406;431.

9. Guo, R. et al. (2019). JCO Precis Oncol 3

10. Ackermann, C. J. et al. (2019) Onco Targets Ther 12, 7857-7864.

1 1 . Bradford, D. et al. (2020) Clin Cancer Res

12. Lin, J. J. et al. (2020) Ann Oncol

Claims

29 CLAIMS

1 . A method for testing the ability of a compound to inhibit RET tyrosine kinase activity comprising assessing the ability of said compound to bind to at least two of amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop (GRL) and aC helix-in configuration.

2. The method according to claim 1 , comprising assessing the ability of said compound to bind to at least three of amino acids K758, L760, E768 and L772 of RET.

3. The method according to any one of claims 1 or 2, comprising assessing the ability of said compound to bind to amino acids K758, L760, E768 and L772 of RET.

4. The method according to any one of claims 1 to 3, further comprising assessing the ability of the compound to bind to amino acid G810 and/or amino acid V738 of RET.

5. The method according to any one of claims 1 to 4, wherein the ability of the compound to bind said amino acids is tested by differential scanning fluorimetry, isothermal titration calorimetry or Western Blotting of lysates from cells treated with increasing concentrations of compounds or a combination thereof.

6. A screening method for compounds able to inhibit RET tyrosine kinase activity comprising virtually screening if the compounds interact with at least two of amino acids K758, L760, E768 and L772 of RET, using RET crystal structure, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration.

7. A method of screening for a compound capable of inhibiting the tyrosine kinase activity of RET, comprising assessing the ability of said compound to bind to at least two, at least three or four of amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration.

8. The method of screening according to claim 7, further comprising comparing the binding of said compound with the binding of LOXO-292 and/or BLU-667 to amino acids K758, L760, E768 and/or L772 of RET.

9. The method according to claim 8, comprising selecting the compound if its binding to amino acids K758, L760, E768 and/or L772 of RET is tighter than the binding of LOXO- 292 and/or BLU-667.

10. A compound that binds to at least one of RET’s amino acids K758, L760, E768, and L772, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration, other than LOXO-292 or BLU-667. 30

11. The compound according to claim 10, wherein said compound does not contact amino acids G810 and/or V738 of RET.

12. Use of a compound according to any one of claims 10 or 11 to inhibit RET.

13. The compound according to any one of claims 10 or 1 1 for use in the treatment of RET-driven cancers, preferably of lung cancer, breast cancer or thyroid cancer.

14. A polypeptide having an amino acid sequence comprising RET polypeptide sequence comprising mutation L760A or mutations L760/772A.

15. A nucleic acid having a nucleotide sequence comprising a sequence codifying for the amino acid sequence of claim 14.

16. A method for modifying compounds to inhibit RET, comprising the following steps: a. Testing the binding of the compound to amino acids K758, L760, E768 and L772 of RET, preferably wherein RET has an open glycine-rich loop and aC helix-in configuration; b. Modifying the chemical structure of the compound; c. Repeating steps (a) and (b) for 1 to 100 times; and d. selecting the modified compound when it binds to amino acids K758, L760, E768 and L772 of RET or when its binding to amino acids K758, L760, E768 and L772 of RET is more specific and/or stronger than before the modification.