Structure-based design, synthesis and crystallization of 2-arylquinazolines as lipid pocket ligands of p38α MAPK

In protein kinase research, identifying and addressing small molecule binding sites other than the highly conserved ATP-pocket are of intense interest because this line of investigation extends our understanding of kinase function beyond the catalytic phosphotransfer. Such alternative binding sites may be involved in altering the activation state through subtle conformational changes, control cellular enzyme localization, or in mediating and disrupting protein-protein interactions. Small organic molecules that target these less conserved regions might serve as tools for chemical biology research and to probe alternative strategies in targeting protein kinases in disease settings. Here, we present the structure-based design and synthesis of a focused library of 2-arylquinazoline derivatives to target the lipophilic C-terminal binding pocket in p38α MAPK, for which a clear biological function has yet to be identified. The interactions of the ligands with p38α MAPK was analyzed by SPR measurements and validated by protein X-ray crystallography.


Introduction
Protein kinases have been a frequent topic in medicinal chemistry and drug development due to their key function as mediating components in signal transduction, regulating cellular pathways on a molecular level, thereby playing a crucial role in the emergence of several diseases. The conventional approach towards the treatment of kinase-related diseases has involved the administration of ATP-competitive inhibitors which potently occupy and thereby block the enzyme's active site where the phosphotransfer from ATP to target substrates takes place [1,2]. However, development of specifically selective inhibitors for a certain targeted kinase within the related members of this enzyme family remains a major hurdle in drug research [3,4].
Successful strategies to gain improved selectivity within the kinome have revolved around employing unique structural features of individual kinases, such as covalent modification of cysteines [5,6] or identifying and targeting alternative binding pockets distant from the active site [7]. Alternative bindings sites far from the ATP-pocket can directly regulate kinase affinity PLOS  and can potentially be addressed by small molecules which alter the kinase activity in a dual manner, via both inhibition and activation [8,9]. In addition to advantages in the development of selective kinase modulators, these binding sites can aid in distinguishing so called non-catalytic functions, those processes triggered by protein-protein (or protein-target) interactions, where kinases serve as scaffolds, e.g., for the formation of multi-enzyme-complexes [10]. In this way, these remote sites can serve as (allosteric) effectors of target molecules, directly affecting the location or the activity state of interaction partners, or can influence cell proliferation, differentiation, and apoptosis. An increasing number of those scaffolding functions of kinases are gradually being discovered and their functions may even exceed the significance of the solely catalytic properties [10][11][12].
Accordingly, the identification and exploration of non-conserved regions may provide insights into the putative unknown functions of protein kinases beyond catalysis and allosteric regulation. Thus, elucidation of unique structural features that modulate protein kinases through employing alternative binding pockets and investigations of the design of corresponding small molecules as alternatives to ATP-competitive ligands has moved to the forefront of kinase inhibitor research and kinase biology. Against this background, we recently identified a novel class of p38α MAPK (mitogen-activated protein kinase) binders addressing a C-terminal lipophilic binding pocket (LP) accessible for small molecules located at several angstroms distance from the enzymes active site. The discovered 2-arylquinazolines bind to the LP and their corresponding co-crystal structures revealed a very distinct binding mode of these lipid pocket ligands (LiPoLis) to p38α (Fig 1B). We considered that these ligands may serve as interesting starting points to study the yet unexplored functions of this binding site in p38α [13]. Several small molecules have been described to address this pocket ( Fig 1A) and they can be classified into detergent-like molecules, β-octyl-D-glucopyranoside (BOG) [14] and phosphatidylinositol ether lipid analogues [15], and small organic compounds, 1 (4-[3-(4-fluorophenyl)-1H-pyrazol-4-yl]pyridine) [16] and 2 (4-(trifluoromethyl)-3-(3-(trifluoromethyl)phenyl)-1Hpyrazolo [3,4-b]pyridine-6(7H)-one) (Fig 1A) [17]. Overall, the majority of the published compounds that address the LP have exhibited a rather low affinity towards p38α [17,18].
The LP consists of the two α-helices 1L14 and 2L14, the αEF/αF loop and a deep lipophilic sub-pocket that is mostly decorated with hydrophobic amino acid side chains (Fig 1B). Based on published co-crystal structures and biochemical assay data, there are indications that some of the already described LiPoLis alter the kinase conformation in a way that the enzymatic activity could be directly influenced by ligand binding, although the corresponding data was measured at high concentrations [15][16][17]. Previous publications also speculated that there might be a regulatory fine-tuning as a result of conformational changes within the LP when addressed by ligands [15,18]. Thus, the biological role of the LP is still not fully understood and may well serve some yet unknown function [19].
To gain a more detailed insight into the role of the p38α LP, we undertook SAR studies based on our previously found lead structure 3 and the known binding mode. Here, we present a structure-based design, synthesis, and validation by surface plasmon resonance (SPR) analysis and protein X-ray crystallography of novel LiPoLis that target the LP in p38α MAPK.
Merck 60 F245 aluminium-backed silica gel plates. Compounds were purified by column chromatography using silica gel from Baker (40-70 μm particle size) or VWR Prolabo (Normasil 60, 40-63 μm particle size). Flash column chromatography was done using a Biotage Isolera One system with Biotage SNAP and SNAP Ultra columns, respectively, and monitored by UV at 254 and 360 nm. Preparative HPLC was conducted on a Varian HPLC system (Pro Star 215) with a VP 250/21 Nucleosil C18 PPN column from Macherey-Nagel and monitored by UV at 254 nm. All presented compounds were analyzed by HPLC to determine and ensure a purity of ! 95%.

Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-8-yl)benzenesulfonamide (10e).
To a stirred solution of 3 (60 mg, 0.18 mmol) in pyridine (2 drops) and DCM (5 mL) at 0˚C benzenesulfonyl chloride (26.7 μL in 1 mL DCM, 0.19 mmol) was added dropwise. The reaction mixture was then heated to 50˚C and was stirred for another 6 h. After quenching with 6 N NaOH and extraction, the combined organic layers were concentrated in vacuo.  (14). A solution of 13 (200 mg, 0.97 mmol) in concentrated HCl (5 mL) and stirred overnight at rt. The reduced crude product (591 mg, 3.58 mmol) was used without any further purification. After dissolving in TFA (11 mL), TES (1.374 mL, 8.6 mmol) was added dropwise and the resulting mixture was stirred at rt overnight. H 2 O and cold saturated NaHCO 3 were added to the reaction that was subsequently thoroughly extracted with EtOAc. The combined organic layers were concentrated in vacuo and silica gel column chromatography (1% ! 5% MeOH/DCM) yielded the target compound 14 as a hydrochloride salt (620 mg, 92%).

Protein expression, purification and crystallization
The expression and purification of inactive, non-phosphorylated p38α wt MAPK was done as previously reported [20]. Briefly, an N-terminal His 6 -p38α wt construct was transformed in to E. coli BL21(DE3) and expressed overnight at 18˚C. The protein was purified by Ni 2+ -NTAaffinity chromatography, followed by anion exchange and size exclusion chromatography after removal of the His-tag by proteolytic cleavage. For SPR experiments the corresponding protein batch did not undergo His-tag cleavage. The pure protein was subsequently concentrated to 10-30 mg/mL, aliquoted, flash frozen in liquid N 2 and stored at -80˚C.
Various LiPoLis were co-crystallized with p38α wt using conditions similar to those as described previously [14]. Briefly, protein-ligand complexes were prepared by mixing 40 μL p38α wt (10 mg/mL) with 1 μL compound (50 mM in DMSO) and incubated for 60 min on ice. The samples were centrifuged at 13,000 rpm for 10 min to remove excess ligand. Crystals were grown in 24-well crystallization plates (EasyXtal Tool, Quagen, Hilden, Germany) using the hanging drop vapor diffusion method and by mixing 1.5 μL protein-ligand solution with 0.5 μL reservoir (100 mM MES pH 5.6-6.2, 20-30% PEG4000 and 50 mM BOG). In some cases, crystals were obtained when BOG was absent in the reservoir solution and when 125 μM BIRB-769 were used instead of BOG, respectively. The crystals were protected using 25% PEG400 before they were flash frozen in liquid N 2 . Diffraction data of the p38α-ligand complexes were collected at the PX10SA beam line of the Swiss Light Source (PSI, Villigen, Switzerland) using wavelengths close to 1 Å. The datasets were integrated with XDS [21] and scaled using XSCALE [21]. The complex structures were solved by molecular replacement with PHASER [22] using the published p38α structure of 4 (PDB: 4DLI) [13] as template. Molecules in the asymmetric unit were manually modified using the program COOT [23]. Inhibitor topology files were generated using the Dundee PRODRG server [24]. Final refinement was done employing the PDB-redo server [25] and PHENIX [26]. Refined structures were validated by Ramachandran plot analysis with RAMPAGE [27]. Data collection, structure refinement statistics, PDB-entry codes and the Ramachandran plot results are shown in S2 Table. Electron densities used for the omit maps were generated via a simulated annealing refinement with ligand and water molecule occupancy set to"0". PyMOL [28] was used to produce

Thermal shift assay
Melting curves of p38α wt were measured at two compound concentrations (50, 10 μM). Compounds in 10 mM DMSO stock solutions were diluted to 2.5 mM and 0.5 mM with DMSO, respectively. From each of the dilution 0.8 μL were added to a well of a 96-well plate (TW-MT, Biozym), except water and reference (DMSO) wells. A protein stock solution (30 mg/mL p38α wt) was diluted to 1 μM in sample buffer (10 mM HEPES, 150 mM NaCl, 5x dye, pH 7.5), already containing the dye SYPRO 1 orange (1000x diluted 5000x stock in DMSO, Sigma) and kept protected from light. To each well of the measurement plate were added 39 μL of this protein solution and the samples were appropriately mixed. After the plates were sealed with optical foil and spun down (200 g, rt, 1 min), they were subsequently placed in a LightCycler 1 480 II (Roche). Experiment was carried out at 492 nm excitation and 610 nm emission wavelength and the temperature ramping from 25˚C to 95˚C with a rate of 1˚C/min. ΔT was determined from the difference to DMSO control samples. Structure-based design of lipid pocket ligands of p38α MAPK

Structure-based design and synthesis of a focused 4-amino-2-arylquinazoline compound library
Starting from the crystal structures of 3 and 4 in complex with p38α (PDB-codes: 4DLI and 4DLJ) and some knowledge of the underlying binding mode, we employed a rational approach for the design and synthesis of new 4-amino-2-arylquinazolines. As a major design aspect, we made no changes to the main interactions contributing to ligand binding, namely the paralleldisplaced π-π stacking between the quinazoline core and the aromatic side chain of Trp197 and a direct hydrogen bond between the anilinic amine and the Asp294 carboxylic acid function. The deep sub-pocket decorated with lipophilic amino acid residues is occupied by a hydrophobic moiety, in this case a phenethyl substituent in the 4-position. This site was thought to be suitable for derivatization to potentially increase compound affinity towards the kinase, since it has been shown that small molecules with differently sized moieties at this position could bind to the LP, with 3 being the most demanding LiPoLi observed thus far [13].
Since previous studies showed that 3 was poorly soluble in aqueous media, we selected the 2-phenyl ring pointing outside the LP towards the solvent for modification with solubilizing groups (Fig 1B). Furthermore, the 7-position of the scaffold should be easily derivatized while maintaining the NH-group as a hydrogen bond donor. Taken together, these observations led us to choose the lipophilic moiety in the 4-position, the solvent-exposed 2-phenyl group as well as the 7-amine as putative sites of modification for the preparation of a diverse quinazoline-based compound library (Fig 1C).
The alignment of LiPoLis 1 and 3 (PDB-codes: 3HVC and 4DLJ) followed by structural analysis prompted us to our initial concepts of derivatization of the quinazoline core scaffold (Fig 2A). In our first design approach, we increased the size of the lipophilic group in the 4-position by introducing fluoro-and difluorophenyl substituents (9b-f) (Fig 2B). These hydrophobic elements might fill out the sub-pocket and thereby form additional favorable interactions, e.g., halogen bonds to backbone carbonyls [30]. Furthermore, the binding of fluorinated compounds to receptors may be entropically favored due to the liberation of water by desolvation, particularly at hydrophobic moieties [31]. Substitution of the phenyl ring with a thiophene (Tph) moiety and variation of the linker length to the scaffold (9g,h) emerged as interesting alternatives since thiophene is a bioisostere with respect to the phenyl group and features different chemical and electronic properties [32]. Using a shorter methylene linker may provide sufficient space for expansion of the five-membered ring, which was the basis of the design of 9j (Fig 2C). Aliphatic moieties like the present cyclopropyl group in building block 14 that was used for the synthesis of 9j can make a greater contribution to the hydrophobic interactions than aromatic substituents, and thus these aliphatic structures may be advantageous for improving ligand affinity [33]. As an alternative strategy, we also designed a hybrid compound (9i), which retained the scaffold of 3 while the 4-position was derivatized with a fragment of the known LiPoLi 1 (Fig 2D).
Furthermore, the solvent exposed 2-phenyl ring and the amine in the 7-position were chosen as derivatization sites for the introduction of solubilizing groups, such as acetylpiperazine (9l), morpholine (9m), methylsulfinyl (19), amides and sulfonamides (10a-e). Also, 8-amino derivative 9k was designed to investigate any impact on the binding affinity, since new hydrogen bonds to the amino acid chains could be formed. The structural basis for this assumption is the relatively high density of polar amino acid side chains in this region of the LP, such as Asp294 being directly involved in the binding of 3.
We developed a common synthetic route which successfully led to the proposed target compounds (Fig 3). The quinazoline cores 7a-d were built up in the reaction by use of anthranilic acids 5a and 5b with benzamide and benzoic anhydride moieties, respectively [34,35], or by using anthranilic amide 6 and aldehyde building blocks [36,37]. Substitution in the 4-position with an amine took place after activation of the hydroxyl group with hexachlorocyclotriphosphazene (HCCP) [38], yielding the nitro compound precursors 8a-n, which were subsequently reduced to the corresponding amines 9a-m. These were feasible substrates to undergo nucleophilic substitution with either carboxylic acids or sulfonyl chlorides to yield compounds 10a-e.
Some LiPoLis required the synthesis of building blocks to be used for coupling to the scaffold (14,16) or as components for condensation to the quinazoline core (17a,b). Starting with the protection of 11 via acetylation and subsequent Friedel-Crafts acylation of the furan ring 12, the final amine 14 was generated by reduction and simultaneous deprotection of intermediate compound 13 under acidic conditions (Fig 4). Dihydroimidazole 16 was prepared by condensation of 4-fluorobenzaldehyde 15 with ethylene diamine (Fig 5).
To decorate the quinazolines 9l,m with corresponding solubilizing groups, the aldehydes 17a and 17b were synthesized by substituting 4-fluorobenzaldehyde 15 with acetylpiperazine and morpholine, respectively (Fig 6). For the generation of 19 starting from 7a, the intermediate 8n was oxidized to the sulfinyl compound 18 that was subsequently reduced under mild conditions to give the final amine (Fig 7). Following these procedures, we synthesized a library of over 30 compounds.
As a potential control ligand, we also synthesized 2 [17], starting with the condensation of nitrile component 20 and hydrazine [39]. The resulting pyrazolamine 21 was converted with 4,4,4-trifluoro-3-oxobutanoate to yield the final product 2. Analogues 22 and 23 were designed to explore possible chemical space with a putative alternative LiPoLi scaffold and were easily generated by using the corresponding oxobutanoates (Fig 8). To subsequently validate our design concept and to confirm that the newly developed LiPoLis would address the LP of the p38α MAPK, we undertook SPR experiments.

Surface plasmon resonance
For SPR measurements, the His 6 -p38α wt was immobilized on a trisNTA [29] sensor. The system was tested with the active site inhibitor SB203580 that showed reproducible results (k on = 1.43 Á 10 6 (M Á s) -1 , k off = 2.43 Á 10 −3 s -1 , K D = 1.7 nM) in good agreement with the literature [34,40] (S1A Fig). Thus, the chosen assay was subsequently used for the characterisation of the synthesized library of quinazoline-based LiPoLis.
Injection of LiPoLi samples gave a concentration dependent response for some of the synthesized derivatives, indicating a specific binding event to the p38α kinase domain (S1B-S1D Fig) since other LiPoLis did not show any increasing signal upon compound injection (S1F Fig). Hence, this renders the separation between actual ligands and non-binding molecules to the immobilized protein and the identification of tolerated structural modifications to the LiPoLis by SPR possible. Concerning the ligands that didn't show any response, all nitro derivatives 8a-m as well as 7-and 8-substituted LiPoLis 10a-e can be considered not to bind to p38α under the chosen experimental conditions. Notably, also reference compound 2 and its derivatives 22 and 23 did not show any response in the SPR experiments with increasing analyte concentrations.
Those ligands that were identified to positively bind to p38α MAPK in these studies all exhibited a similar shape of the detected sensorgram as the initial lead compound 3. Starting at baseline level, sensorgrams showed an increasing response during analyte injection in a concentration-dependent manner, which indicates a positive binding event to the immobilized protein. Unspecific binding to the sensor surface of the reference channel could be Structure-based design of lipid pocket ligands of p38α MAPK essentially excluded, since corresponding sensorgrams did not represent any signal indicating unwanted interactions with the reference surface blocked with the His 6 -peptide. At higher concentrations the tested analytes did steadily bind to the active sensor surface without reaching an equilibrium, indicating accumulating effects to the kinase. This behaviour complicates the determination of kinetic parameters in terms of association as a 1:1-Langmuir fit model did not properly represent the binding event. However, the observed effect was reversible, as baseline level was recovered after switching from compound injection to running buffer flow.
Compounds that showed a characteristic response in the SPR measurements allow conclusions regarding the tolerated chemical space concerning compound modifications that can be made to the 2-arylquinazoline scaffold without impairing the ability to address the LP. The exemplified results for the characterisation of 3, 9c and 9m showcase the commonly observed shape of sensorgrams of the tested LiPoLis, that were reflected by slow association and significantly fast dissociation (k off = 0.04-0.06 s -1 ) (S1B-S1D Fig), typically leading to weak binding affinities. Substituting the phenethyl moiety in 4-position with bioisosteres in form of thiophenes 9g and 9h or replacement by fluorophenyl residues (9b-d) gave comparable and reproducible sensorgrams. On note, derivatives 9e,f carrying the difluorophenyl moiety appeared to lose any affinity to the protein compared to the other 4-aminoquinazolines. In direct comparison to 3, the solubilizing group-bearing derivatives 9l,m showed a significantly faster association before reaching the equilibrium state at lower ligand concentrations (S1D Fig). Thus, these results demonstrate, that even sterically more demanding moieties are tolerated at the phenyl ring in the 2-postion. In one case, the binding signal of 9j detected from the active channel was apparently masked by non-specific binding occurring at both, the active surface and the blocked reference surface, reflected in pseudo-irreversible binding after doublereferencing (S1E Fig). In summary, we could establish a robust SPR assay system that was set up using the known inhibitor SB203580 and generated reliable and reproducible data for the characterisation of the presented ligands. Weak affinity of the analytes towards the target kinase was generally observed in SPR and orthogonal assays, e.g., thermal shift assays (S3 Fig). To further characterise the designed LiPolis regarding the exact binding mode when bound to p38α MAPK cocrystallization experiments were conducted.

Crystallography
For the crystallization, we pursued different strategies that resulted in different co-crystallized structures of p38α MAPK in complex with LiPoLi quinazolines. Crystal growth varied depending on the experimental protocol followed and compounds used. When previously published conditions were used in the presence of BOG [14], crystals and the corresponding protein structures were usually obtained that accommodated the detergent in the LP. BOG is known to bind to the LP and the concentration used under these conditions may compete with the LiPoLi quinazolines for the occupation of the LP. Therefore, we also conducted crystallization trials in the absence of BOG. A third crystallization method focused on using the active site inhibitor BIRB-769 to stabilize the kinase and thereby facilitate crystal growth and eventually binding of the LiPoLis. Generally, crystal growth took place spontaneously or within 21 days, yielding needle-shaped crystals when BOG or BIRB-796 were present. More spherical-shaped crystals were obtained when the protein was exclusively incubated with the LiPoLis (S4 Fig). Protein crystals showing the spherical morphology usually were occupied with bound LiPoLis to the LP and could thereby serve as an indicator for successful co-crystallization (a summary of these findings can be found in S1 Table).
When crystallization trials were set up with ligands bearing an alternative substituent other than an amine in the 7-position, immediate precipitation of the protein was observed. Thus, no crystals could be obtained for nitro compounds 8a-n and the derivatives 9k, 10a-e as well as 2 and its analogues. This outcome indicated that sterically demanding substituents in the 7-position might interfere with protein binding, resulting in the loss of the previously observed H bond between the amine and Asp294.
We started our trials of the 4-amino derivatives following the direct co-crystallization protocol and the complex of 9c was one of the first structures that was successfully solved ( Fig  9A). The generally proposed binding mode and key interactions were still maintained, including the phenethyl moiety binding deeply into the lipophilic sub-pocket, the π-π-interaction between Trp197 and the core quinazoline scaffold, and a direct hydrogen bond between the amine in the 7-position and Asp294 as well as a hydrogen bonding network with surrounding water molecules and amino acid side chains and the protein backbone, e.g., including Ser251 and Lys249, and partly stabilized by contacts involving crystal symmetry mates, particularly Glu336 (S5 Fig). Furthermore, we obtained crystals for 9g and 9h that were conceived as bioisosteres of 3 by substituting the 4-residue with the corresponding thiophene moieties (Fig 9B  and 9C). Thus, we demonstrated that the hydrophobic sub-pocket of the LP was capable of harboring that five-membered heterocycle while maintaining the general binding mode described above. These findings were the starting point for the design of 9j as an "extended derivative" of 9g, which was achieved by adding another methylene cyclopropyl to the furan Structure-based design of lipid pocket ligands of p38α MAPK ring. We were gratified to observe that the co-crystallization of this LiPoLi succeeded in presence of BIRB-796. It is noteworthy that the binding mode of 9j within the LP was not significantly different from the previously mentioned key interactions (Fig 9D). The cyclopropyl moiety of 9j was able to access the deeply buried regions of the LP. Furthermore, the cyclopropyl group did not lead to any perturbation of the protein structure based on the arrangement of adjacent amino acid side chains, as compared to the other LiPoLi bound structures. The difference electron density map was not defined at the position where the cyclopropyl substituent was situated, indicating a significant flexibility and the ability to twist within the lipophilic sub-pocket of the LP.
Interestingly, some compounds only crystallized in the absence of BOG in the reservoir solution, and these samples also exhibited a new crystal morphology as observed for 9l and 9m (S4 Fig). The co-crystal structures for both compounds were successfully solved and confirmed our design approach as they revealed the same binding geometry as the initial lead compound 3 (Fig 9E and 9F). Compounds 9l and particularly 9m were designed to improve the compound's solubility by introducing a solubilizing group at the phenyl ring in the para-position. The 2-phenyl substituent was orientated towards the solvent, similar to 3, although no interactions with the surrounding water molecules were found. Notably, the network of water molecules surrounding the LP was best resolved in the p38α-9m complex crystal structure (Fig 9F). Performing a simulated annealing refinement, mFo-DFc omit maps were generated for all cocrystallized LiPoLis and show defined electron density that could be clearly assigned to the corresponding ligands and demonstrated a consistent mode of binding (S6 Fig). Only for 9j, no defined density of the cyclopropyl moiety within the LP was observed, likely due to conformational flexibility, and 9h only showed partial density of the quinazoline scaffold, potentially caused by lower occupancy at the chosen experimental conditions. When we compared our new complex crystal structures with p38α-1 and p38α-3, we commonly observed subtle conformational changes of secondary structure elements shaping the LP. Interestingly, a more significant displacement of helix 2L14 was found in the p38α-9h complex structure, resulting in an almost complete opening of the lower section of the LP (Fig 10). This underlines a pronounced flexibility of the entire binding site rendering it putatively addressable by even more complex synthetic ligands or yet to be identified biological binding partners.

Conclusions
Due to their central role in cell signaling pathways, protein kinases are prominent targets in drug research and development. Several diseases are caused by direct dysregulation of the corresponding kinase or their mediators and interaction partners, implying that not only the integrity of enzymatic mechanisms, but also of non-catalytic functions, often referred to as scaffolding functions [10], are mandatory to preserve the sensitive and well-regulated processes within the cell. The successful targeting of scaffolding functions requires detailed knowledge of the participating interaction partners as well as the mechanisms of communication, which include the important structural elements such as binding epitopes and their ligands as well as their biological roles. Here, we set out to target the previously identified LP in p38α MAPK. By applying structure-guided derivatization of 3, we introduced structural variations in the hydrophobic moiety in the 4-position and the functionalization of the 2-phenyl ring and the 7-/8-positions that led to a focused library of over 30 compounds. An SPR assay system was set up to get insight into the kinetics of LiPoLi binding towards p38α MAPK. The interactions of the tested ligands and the immobilized kinase was typically described by a slow association phase and saturation followed by fast dissociation. Most of the LiPoLis that showed a characteristic response in the SPR experiments were also successfully co-crystallized with p38α MAPK. A series of six complex crystal structures showed that the designed LiPoLis indeed target the LP of p38α MAPK and validated our design approaches. The complex structure of 9j could be solved although in the SPR experiments only unspecific binding to the sensor surface was detected.
In summary, we identified substitution patterns of the LiPoLi scaffold to be crucial for the opening of the LP underlining the flexible nature of this binding site. The characterization of the ligand-binding event by SPR indicated, however, that these LiPoLis are most likely not suitable to serve as functional probes given their weak binding affinities. Anyhow, the results presented here will encourage further compound modifications to focus particularly on the 2-phenyl ring and alternative substitutions to generate more potent LiPoLis to finally dissect the functional role of the lipid pocket in p38α.  Performing a simulated annealing refinement, mFo-DFc omit maps (green, contoured at 2.5σ) were calculated for A) 9c, B) 9g, D) 9h, E) 9j, E) 9l and F) 9m, as well as for surrounding water molecules. 2Fo-Fc maps for the ligands, waters and key residues Trp197 and Asp294 were contoured at 1.0σ (blue). Maps indicate partial occupancy for 9h due to multiple molecules bound to the protein and conformational flexibility of the cyclopropyl moiety in 9j. (TIF) S1 Table. Crystallographic statistics of p38α MAPK in complex with LiPoLi derivatives. Statistics for co-crystals with LiPoLis 9c, 9g, 9h, 9j, 9l and 9m (PDBs: 5N63, 5N64, 5N65, 5N66, 5N67 and 5N68). Values in parenthesis refer to the highest resolution shell. (TIF) S1 File. NMR spectra.