Mapping of dynamic allostery within p38 alpha kinase via network analyses and NMR spectroscopy

Mapping of dynamic allostery within p38 alpha kinase via network analyses and NMR spectroscopy | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Mapping of dynamic allostery within p38 alpha kinase via network analyses and NMR spectroscopy Rasmus Linser, Suchandra Acharyya, Jörn Weisner, Rafael Bernardi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8958159/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Kinases are major drug targets especially in cancer therapy. However, the high degree of conservation of their active sites hinders the development of selective inhibitors, motivating a deeper understanding of kinase conformational ensembles and allosteric communication pathways. Here, we use dynamical network analysis to identify key residues involved in a dynamic allostery between the N- and C-lobes that connects the major functional units of the MAP kinase p38α. By combining NMR spectroscopy, activity assays, and in silico analysis of wildtype protein and mutants in the presence or absence of an active-site inhibitor, we experimentally validate the obtained architecture with respect to global protein motion and long-range allosteric modulation. Notably, the identified network highlights communication pathways across several functional sites, prominently involving the allosteric site, the activation loop, and the lipid-binding domain with its embedded cryptic pocket in the C-lobe. These findings provide mechanistic insight into p38α allostery and suggest new opportunities for the rational design of allosteric modulators of MAP kinases. Biological sciences/Biophysics/Molecular biophysics/Molecular conformation Physical sciences/Physics/Techniques and instrumentation/NMR spectroscopy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction The family of p38 mitogen-activated protein kinases (MAPK) represents one of the principal elements within the MAPK signalling cascade, critically involved in mediating cellular responses to stress and inflammatory stimuli. A robust cluster of experimental evidence suggests that p38 can exert pro-oncogenic functions in various types of cancer 1 , which has made p38 one of the most important drug targets across a wide range of cancers. p38 kinases are prototypical kinases comprising a bi-lobal structural core that is divided into a smaller N-terminal and a larger C-terminal lobe. The catalytic active site lies between the two domains, forming a hinge bearing high flexibility, which is considered to be an important parameter for nucleotide exchange 2 . The regulation of p38α activity involves several highly dynamic structural elements, including the catalytic loop, the N-terminal segment of the activation loop containing the conserved “DFG-motif”, the αC helix, and the hinge region 3 . Drug design efforts for this kinase have been focused predominantly on small-molecule inhibitors of the ATP binding site, prohibiting the activating phosphorylation of residues T180 and Y182. Addressing this critical functional element, however, a component widely conserved across the entire kinome, entails dose-limiting adverse effects owing to non-selective inhibition. Consequently, the identification of other potential locales that bind substrates, inhibitors, or allosteric effectors is of great interest 1 . Modulators that bind in the “DFG-out” conformation, exploiting an additional site opening up close to the active site, have also been designed 4 . Sorafenib (Nexavar®) is such a (“type II") inhibitor, binding to this site but potently inhibiting nine other kinases 5 ( Fig. 1a ). Its strategic use in combination with the drug SB202190 at the ATP-binding site, e.g., entails a synergistic effect that increases the apoptotic response in colorectal cancer cells 5 . However, the resemblance of this site to the conserved phosphate-binding site has resulted in little or no increase in the specificity of drug binding, toxicity, and dosage constraints. A rather distant, cryptic pocket is associated to the lipid-binding domain, hosting the “MAP kinase insert” 6 , which – in conjunction with a significant conformational change – has been shown to accommodate a range of lipophilic molecules like n -octyl-β-glucopyranoside (β-OG) as well as suitable covalent binders addressing a conserved Cys closeby 7 . This hidden pocket in the C-terminal region of the protein, the “lipid pocket” 6 ( Fig. 1a ), has been shown to bear a primary sequence conserved over the p38 family and potentially allows for binding of more drug-like small molecules specifically generated in the quest for new allosteric p38 inhibitors 8 . Allosteric communication refers to an intramolecular mechanistic crosstalk between spatially distant sites that participates in modulating functionality as a function of external events 9 . Allostery, in a wider sense, is the event in which one site somehow “feels” changes in a different site 10 , 11 . As such, it can either be based on significant structural changes or rather – even in the complete absence of the latter – hinge on dynamic features that are mutually dependent across the protein 12 . This latter mechanism, commonly referred to as “dynamic allostery” 9 , involves the redistribution of correlated motions across the protein, which can be modulated or induced by regulatory events such as ligand binding, protein–protein interactions, or phosphorylation. In p38α complexes, evidence for long-range dynamic coupling is provided by both structural and dynamical observations: Structural superimposition of the “DFG-out” p38α complexes bound to Sorafenib alone (PDB 3HEG 13 ) and to both Sorafenib and β-OG (PDB 3GCS 14 ) reveals a flip of the pyridyl nitrogen and methyl substituent of Sorafenib, despite a separation of more than 30 Å from β-OG bound in the lipid pocket ( Fig. 1b) . This slight but noteworthy difference in ground-state structures suggests that ligand binding at the distal lipid pocket alters the conformational ensemble of the kinase in solution. In addition, activation of p38a incurs dynamics differentially modulated across timescales, with activation-loop phosphorylation having been observed to quench ps–ns motions without altering the average conformation. Instead, uniform µs–ms backbone dynamics are induced by substrate binding, flattening the energy landscape and rendering key allosteric sites accessible 15 . An cross-lobe interdependency in the realm of dynamic allostery also seems to involve the lipid pocket, occupation of which was observed to lead to widespread chemical-shift perturbations and changes in dynamics 16 . Similaly, for p38g kinase, another one of the four MAPK isoforms and a close relative to p38a with ~60 % sequence identity, motional changes in the kinase due to distant effectors have been assessed from various angles 17, 18, 19 . These observations align with the identification of long-range community networks in protein kinase A (PKA), fundamentally orchestrating overall dynamics and hence functionality 20 , as well as findings in the Scr kinases 21 , epidermal growth factor receptor (EGFR) 22 , and Abl kinase 23 , where catalytic activity was seen to be modulated by very distant sites. 19 Corroborating Cooper and Dryden’s original theory, activation of PKA has recently been ascribed to a redistribution of fast and intermediate-timescale thermal fluctuations (the “violin model”), which explains the lack of apparent structural changes. Thereby, the formation of “hydrophobic spines”, specifically the C-spine assembled upon ATP binding, and the R-spine, linked to positioning of the αC-helix, an assembly regulated by upstream processes (also shown in Fig. 1a) , is thought to aid in orchestrating suitable dynamic networks. 24 In MAP kinases, insights into such motional dependencies are of general interest from a biological perspective. However, they are of particular importance in the light of a motional connectivity between the different regulatory sites, the active site, or the lipid binding domain. A residue-specific mapping of such a dynamic network and identification of its key participants may aid understanding the molecular underpinnings of p38 regulation and even launch future pharmacological avenues based on the underlying pathways ( Fig. 1c ). An in-depth interrogation of dynamic networks has become possible using molecular-dynamics simulations in combination with dynamic network analyses, where the correlations of motions assessed as generalised correlation coefficients based on mutual information are exploited to determine the strength of dynamic connectivity between different residues of the system. 25 Whereas these in-silico findings are extremely rich in information, they do require verification by experimental means. Providing site-specific access to a large range of chemical and motional parameters, NMR spectroscopy is particularly well suited as a complement to such simulations 26 . Here, we use state-of-the-art computational approaches based on a dynamical network analysis software (Dynetan) 25 to characterize p38 dynamic networks from an NAMD 27 -based molecular-dynamics simulation. Perturbing those residues that turn out to be pivotal for the motional network through site-directed mutagenesis, we then use NMR in solution to monitor the consequences for the various mutants experimentally. Together, these data reveal an extended dynamic network spanning both kinase lobes and comprising several distinct allosteric “hotspots” in p38α. In addition to features common to other kinases, our analysis identifies the MAP kinase-specific lipid-binding domain and its associated cryptic lipid pocket as integral components of the dynamic network, highlighting their potential relevance for allosteric therapeutic intervention. Results Following the workflow summarized in Fig. S1, and the system setup summarized in Table. S1, we performed molecular dynamics (MD) simulations of apo p38α, sorafenib-bound p38α, and p38α bound to both sorafenib and β-OG using NAMD 3 27 with the CHARMM36 force field 29 . (See methodological details in the SI text. For the apo protein, Figs. S2 and S3 show root-mean-square deviations as a function of time as well as root-mean-square fluctuations as a function of residue.) For each system, five independent replica simulations of 1 µs each were carried out. For subsequent network and correlation analyses, the final 200 ns of each replica were concatenated and analyzed using Dynetan 25 (described in more detail in the Methods and graphically summarized in Fig. S4). This analysis included community detection and quantification of betweenness centrality, which identifies residues that are central to allosteric communication. Community detection, the first output of network analyses, partitions the network into clusters that share common motion. This enables probing of consistency between the rich, atomic-resolution MD data here with coarser experimental studies on this or similar systems in the past. For the wild type, apo protein, this analysis yields 14 clusters, shown by different colors in Fig. 2a . Here, each community denotes a group of residues that exhibit strongly correlated motions with each other but weaker correlations with residues outside the group, reflecting coherent dynamical behavior within the protein. Despite the ~60 % sequence identity between the isoforms only as well the stark differences in methodology, the clusters obtained from the in-silico assessment here have a remarkable congruency to those (15) clusters found via purely experimental analyses (methyl scanning and chemical-shift perturbations) for p38g 17 : Like in this previous study, our clusters divide the N-lobe into different active-site (magenta) and allosteric-pocket (brown) regions and take accountability of the R-spine (gray) and C-spine regions (pink). The C-lobe is divided into three clusters, where the MAP kinase insert (lipid binding domain) is the cluster shown in green in Fig. 2a and deviates from the cluster containing the activation loop (shown in purple). Notably, several of these dynamic features also align with community structures identified previously for a different kinase (protein kinase A, PKA) using Girvan–Newman-based network analysis 20 . Here, communities were segregated into 13 regions: In the N-lobe, the ATP binding pocket (magenta) is reminiscent of the PKA “Com A1”, the C-helix; the extended allosteric pocket of the p38α (brown) corresponds to “Com B”, and the R-spine of PKA (“Com C”) matches the gray cluster in p38α here 20 . In the C-lobe of PKA, “Com D” is a cluster congruent to the activation loop in p38α (purple), “Com E”, the C-spine cluster in PKA, matches the pink cluster in p38α, and “Com F” in the PKA case is the substrate binding region, which bears at least a slight similarity to the lipid pocket in p38α (green), which is part of the lipid binding domain only applicable for MAP kinases 20 . More interestingly, we interrogated betweenness-centrality, which property identifies allosteric-network hotspots on the basis of their participation extent within different communication pathways. These results are shown in Fig. 2b and more specifically in Table S2. Whereas naturally, a one-to-one comparison with other kinases is compromised due to the different primary structures and differences in the techniques applied, these pathways again seem qualitatively consistent with the experiment-based pathways connecting different regions of p38γ, where methyl mutations of several conserved sites were explored. 17 For example, the memory node L170 in p38γ that takes additional input from the DFG-motif, appears as a motional hotspot in our study as well. Likewise, the hinge cluster of Abl kinase, constituting the D400 of the DFG-motif, had similar dominance in the long-range communication, suggesting the aspartate of the kinase to be a significant player in dynamic allostery 30 . In case of PKA, the F185 of the DFG motif coordinated together was found to be part of the allosteric communication instead of the vicinal aspartate, evocating the hypothesis whether the DFG motif itself is a necessary allosteric modulator 31 . D150, another node identified by our betweenness centrality study, belongs to the catalytic HRD-motif. In PKA, the homologous aspartate D166 was found to be the most crucial node that preorganizes the substrate’s phospho-acceptor site for efficient phosphotransfer 32 . (For Abl kinase and EGFR, the residues of this central catalytic machinery are the homologous D363 and D813, respectively. 33, 34, 35 ) R49, another mediator that we identified as a residue with a high betweenness centrality, is a part of the extended allosteric pocket in p38α. While it reinstates its allosteric role in p38a in our study, a homologous residue in similar kinases that pinpoints its governance in their allosteric mechanism has not been found 36 . Most interestingly, however, W207, identified here as another key residue of the network, is positioned near the substrate binding groove. In PKA, the homologous W222 has been found to be an allosteric hub in the αF/FC region and a structural keystone anchoring the C-lobe. 37, 38 (Abl and EGFR kinases have similar hydrophobes but there is no exact analogue to pinpoint in this region 39 .) In p38a, this residue plays a particular role as it directly faces the lipid pocket, for which the possibility of designing allosteric effectors has been speculated. 7 Its high betweenness centrality hence directly rationalizes a participation of the lipid pocket as part of the dynamic network. Beyond the overall inter-lobe correlations (shown in Fig. 2c and Table S3), which confirm both short-range connectivities dictated by the three-dimensional fold of the protein (e.g., R70/F169 ) and longer-range cross-talk such as between residues L104 and L167 or N102 and K165, we specifically examined the communication pathways linking the allosteric pocket, the activation loop, and the lipid pocket. For this reason, we picked one representative residue of the network for each of these sites/clusters and computed the cumulative pairwise correlation (the sum of degrees of correlation between any two neighbors) within the shortest path of communication. For the allosteric site, represented by residue K66, towards the activation loop (Y188), the information flux traverses the D168 and G170 of the DFG motif, similar to p38γ ( Fig. 3b , compare Table S5) 17, 18, 19 . Accordingly, the shortest path between the activation loop and the lipid pocket, with residues such as K233 and I235, is shown in Fig. 3c . Pathways between the two pockets (between K66 in the allosteric pocket and K233, and I235 in the lipid pocket) were also calculated without specific consideration of the activation loop. This path involves the DFG motif, but bypasses the activation loop to channelise the communication to the lipid binding domain ( Fig. 3e ). Importantly, all of the above analyses identify a network of allosteric communication that interconnects not only the N- and the C-lobe generally but also involves the lipid-binding-domain and specifically the lipid pocket. These in silico results rationalizes the existence of functional communication pathways between the active site and the lipid-binding domain, supporting potential avenues for targeting MAP kinases through allosteric modulation. To assess whether those networks are inherently apparent already from crystallography, the nodes identified were further examined in terms of B-factors derived from the crystal structures 29 . Crystallographic B-factors provide an experimental proxy for local rigidity and structural order. Fig. S5 shows a depiction of B-factors. However, the B-factors of the hotspots of the dynamic network are mostly in the intermediate range (see a list of B-factors for all major nodes of the network in Table S4), denoting that these residues are not particularly disordered in the crystal. (For example, residues R49, D150, and D168 have values of 0.509, 0.342, and 0.497, respectively. W207 shows an even lower value of 0.056, making it a very well-defined residue.) Even though a flexible character/flexibility of specific residues at physiological temperatures can sometimes be forecasted from the cryogenic crystallographic viewpoint, it seems that the allosteric hotspots are not recognizable through a specific degree of disorder. For further elucidation of changes to the networks, we hence turned to room temperature dynamics data, assessed by NMR spectroscopy (see below). To validate the existence and characteristics of the dynamic allosteric network identified by dynamical network analysis, we pursued a two-level approach involving additional in silico analyses and experimental validation. In both cases, the network was probed by “mutating” residues identified as key nodes based on high betweenness centrality. For this downstream analysis, in particular correlation analyses, ligand (sorafenib) binding studies, solution NMR chemical-shift perturbations and NMR relaxation, as well as biochemical assays (see below), four (R49, D150, D168, and W207) were selected and replaced by alanine. W207 is of particular interest as it forms part of the lipid pocket in the C-lobe. MD data (five 1 µs trajectories for each of the four mutants) were obtained using similar strategies as for the wild type (Fig. S6) and re-analyzed with the mentioned network framework. These analyses confirmed both, differences in betweenness-derived pathways and altered community structure (Fig. S7) compared to the wild type. The effects of mutation on long-range communication were also quantified by evaluating inter-lobe (N–C) correlations and correlations along the wild-type-defined shortest communication paths (Tables S2, S3, and S5). Total correlation maps for all ensembles are provided in Fig. S8 and S9. Whereas the tertiary structure and hence many correlations are necessarily maintained with respect to Fig. 2c , changes with respect to long-range correlations can be witnessed in each of the cases. (Difference maps (mutant – wild type) are shown in Fig. S10.) To focus on pathway-level effects, we again selected one central residue per structural element (K66 for the allosteric pocket, Y188 for the activation loop, and I235 for the lipid pocket as before) and computed the shortest paths between them also for the mutants. The wild-type path served as the reference, and the cumulative amounts of pairwise correlations between any two consecutive residues along this path were evaluated for wild type and each mutant. Uncertainty was estimated from five replicas, each partitioned into five windows (Table S6). Across all paths (allosteric-to-activation loop, Fig. 3b ; lipid-pocket-to-activation loop, Fig. 3d ; and allosteric-to-lipid pocket paths, Fig. 3f ), mutant correlations were consistently lower than in the wild type protein. Interestingly, W207A, despite its residence in the C-lobe, has a very strong effect on the allosteric site-to-activation loop path (~20 %, Fig. 3b ). Conversely, for the lipid-pocket-to-activation loop, D168A, the DFG-mutant lying in the hinge of the N-lobe has a strong impact. (~10 %, Fig. 3d ); both residues lie distal to the respective pathways, consistent with a dense allosteric network. ( Fig. 3d ). Intriguingly, with respect to the direct pathway between the allosteric pocket and the lipid pocket, all the nodes identified by the network mutated out (R49A with ~10 %, D150A, D170A, both with ~20 %, and W207A, ~25 %) show a substantial, statistically significant drop ( Fig. 3f ). In addition to the apo proteins, we also examined how the presence of an active/allosteric-site binder would influence the network. For this purpose, we employed the in-silico interrogation of wild type and mutants outlined above, however, starting from the sorafenib-bound X-ray structure 3HEG. Fig. S11 and Table S8 show the embedding of this ligand as interrogated using Discovery Studio 40 . MD simulations again afforded trajectories of 5 replicas each (Fig. S12). At first, we assessed motional correlations between the protein and the drug molecules. Fig. S13 shows these correlations, where for simplicity the ligand sites were coarse-grained into a Western and an Eastern part analysed individually. Interestingly, we observed that the total amount of cumulated correlation drops slightly upon mutation, most prominently for mutant R49A (Fig. S13). More importantly, however, we then pursued the above examination of interdomain communication within the ligand-bound protein. For the wild-type complex, shortest communication paths connecting the allosteric pocket, activation loop, and lipid pocket were identified using the same representative residues as in the apo analyses. Structural representations (Fig. S14A, C, E) reveal that inhibitor binding now reroutes the preferred communication pathways linking the allosteric pocket, the activation loop, and the lipid pocket with each other. For each complex, we quantified the cumulative pairwise correlations along these paths, using the ligand-bound WT as the reference. Now, taking the new, ligand-bound routes of the wild type protein as the reference and mutating the above-mentioned network hospots, no significant reduction in correlation is observed anymore (Fig. S14B, D, F). This indicates that ligand binding rigidifies the protein and largely suppresses its intrinsic dynamic network. Consistently, no systematic gains or losses in cumulative correlations are detected along the ligand-bound paths. Together, these results demonstrate that (i) mutations at network sites attenuate communication in a pathway- and residue-dependent manner, (ii) sorafenib reshapes the communication routes relative to the apo protein, and (iii) ligand binding to the active site globally rigidifies the protein, effectively quenching dynamic network propagation. Assessing a single shortest communication path neglects the fact that allosteric signaling typically arises from the collective contribution of multiple pathways. Moreover, the in silico identification of long-range communication paths spanning many network edges is intrinsically less sensitive than the analysis of shorter network fragments, such as those shown in Fig. 2b . Even with the extensive sampling afforded by five independent microsecond-long MD replicas, these pathway-level correlations retain a certain statistical error, underscoring the need for experimental validation based on vastly larger ensembles of molecules. To experimentally (with a much bigger ensemble) probe the functional consequences of hijacking the dynamic network, we therefore first assessed the enzymatic integrity of the individual mutants relative to the wild-type protein using Homogeneous Time-Resolved Fluorescence (HTRF) assays 41 ( Fig. 3g , see details in the Methods). HTRF is a robust, high-throughput, low-background FRET-based technique that combines the sensitivity of time-resolved fluorescence with a homogeneous assay format. In this study, HTRF was employed to quantitatively compare protein substrate binding efficacy (using GST-tagged activating transcription factor 2 as a substrate) as well as the integrity of binding of ATP between wild-type and mutant constructs 41 . Adding to the wild-type protein, whose expression and purification has been described in the literature 42 , all of the above mentioned mutants were generated via site-directed mutagenesis and the protein expressed recombinantly in E. coli and purified according to the existing protocols (see the Methods and Fig. S15, for biochemistry details). The dependence of reaction speed on either ATP or substrate concentration serves as a quantitative proxy for binding of either interaction partner (see details in the Methods). Kinetic characterisation was pursued via Michaelis-Menten fits ( Fig. 3h and Fig. 3i ), confirming highest binding competencies, with the lowest Michaelis-Menten constants ( K M ) for the wild type enzyme (Table S7). Strongly impaired affinity occurs in mutants D168A, R49A, with a twofold decrease in ATP binding each. Similarly, for the mutant D150A, the binding to the substrate is diminished, and ATP binding competency is slightly lower. Also for W207A, even though the mutation is deep in the C-lobe, the ATP-driven regulation and catalysis appear to be lost. The most severe phenotypes were observed in D150 and W207 , even though neither of these sites is part of the actual enzymatic process, consistent with the possibility of disrupted enzymatic potency due to allosteric modulation from distant sites of mutation. Even though the consistent decrease for both interaction partners even for mutation sites far away from binding interfaces and in particular for W207 as a lipid pocket residue is noteworthy, more specific motional analyses were sought to corroborate the findings. If protein motions were purely local and independent, a point mutation would be expected to affect primarily the substituted residue and its immediate environment. By contrast, NMR studies have shown that perturbations within dynamic allosteric networks can induce long-range changes in both, site-specific chemical shifts and relaxation properties across multiple timescales 17 43 . To assess such putative long-range changes in protein dynamics, we used solution-state NMR on 13 C/ 15 N/ 2 H labelled protein samples expressed in triple labelled media and purified using successive rounds of Ni-affinity, anion exchange, and size exclusion chromatography according to published procedures 42 . In spite of multiple attempts and in contrast to all other proteins mentioned here, the plasmid of mutant W207A could not be successfully expressed as a stable triple-labeled construct in deuterated minimal media, precluding its structural and dynamic interrogation by NMR. Peak assignments were achieved via 3D HNCA spectra for both wild-type and mutant constructs, complemented by TROSY-HSQC used as fingerprint spectra (Fig. S16), upon which existing assignments (BMRB entry: 17471) could be transferred in a stepwise manner. To complement the missing fourth mutant, we newly introduced V38A, in this case, however, specifically as an experimental negative control. V38 lies in a folded element of the tertiary structure, but outside the predicted dynamic network, allowing us to compare any “regular” alterations induced locally to global changes entailed by a tight dynamic network. We first employed chemical-shift perturbation (CSP) analysis to probe changes in the average chemical environment of backbone amides across the protein. In the absence of allosteric communication, the chemical shifts of very distant residues would not be expected to change even in the case of slight structural changes around the mutation site. Instead, the analysis revealed widespread deviations across the backbone for all functional mutants involved in the allosteric coupling. Notably, D168A, the "DFG-loop mutant", exhibited the most dramatic global extent of CSPs across the structure ( Fig. 4a, c, and d ), including long-range effects such as the complete disappearance of the W207 peak in the C-lobe. As mentioned above, apart from its identification as a spatially distant key residue involved in the interdomain communication, W207 also represents a marker for the lipid pocket. This corroborates both, D168A’s sensitive influence on allosteric interactions and domain coupling as well as the participation of W207 in the allosteric network. The distributed nature of dynamic perturbations throughout the protein can be recognized from Fig. 4b , exemplarily depicting chemical-shift perturbations for this variant compared to the wildtype. (CSPs and spectral overlays for the other mutants are shown in Fig. S17.) In contrast to the mutants focused on hotspots found for the dynamic network, the negative control V38A showed indeed very limited chemical-shift changes (see Fig. 4d and Fig. S17). The stark differences in chemical-shift perturbations between mutations within the dynamic network and the negative control is also seen from the violin plots shown in Fig. 4c , supporting V38A’s role as a mutation benign to the conformational ensemble and underscoring the global conformational remodeling induced by the actual network-disruptive mutations on the contrary. These findings provide experimental support for mutations within hotspots identified by dynamical network analysis to indeed induce substantial allosteric rewiring of the kinase conformational ensemble. To determine whether the observed chemical-shift perturbations really reflect changes in global protein dynamics, we employed solution-state 15 N relaxation, more specifically, [ 15 N, 1 H] heteronuclear NOE (hetNOE) and 15 N transverse relaxation rates ( R 2 ). These partly complementary relaxation parameters probe site-specific motion on the ps-ns timescale, with R 2 being additional influenced by motion on the µs timescale. Among all mutants tested, D150A exhibited the most pronounced deviations from wild-type behavior across both, hetNOE and R₂ datasets ( Figs. 5 and S18). The further reduction in hetNOE values compared to the wildtype at several positions throughout the protein suggests a further increase in fast-timescale local flexibility. In parallel, increased R ₂ values, especially in the C-lobe (residues around 200 – 250), indicate an additional increase of conformational-exchange contributions, stemming from increased µs timescale fluctuations and suggesting a widely altered energy landscape introduced by the mutation. These observations also align with significant exchange broadening, which is usually incurred by enhanced µs–ms conformational dynamics. The differences to the wildtype are also visualized through structural mapping of per-residue deviations ( Fig. 5a-b ), witnessing dynamic perturbations extending well beyond the respective mutation sites. Intriguingly, D168A and R49A also show vivid alterations in the C-lobe regarding the R 2 rates, and more slightly for hetNOE values, again supporting that these sites represent key nodes in long-range allosteric coupling. Also the extent of dynamic reorganization in D150A appears to be globally distributed. In disparity, the control variant V38A, with its mutation site outside the predicted dynamic network, exhibited again only minimal, local changes. Both its hetNOE and R₂ profiles remain largely aligned with the wild-type except for subtle shifts near the site of mutation (Fig. S18), confirming its neutrality regarding the kinase’s conformational ensemble. Boxplot comparisons across the variants ( Fig. 5c-d ) support a general enhancement in both, fast-timescale flexibility (decreased hetNOE) and a concurrent rise in µs timescale dynamics (higher R ₂ values), suggesting that mutations in dynamic network positions generally increase protein plasticity and conformational exchange. These shifts correlate with a loosening of the dynamic network upon mutation, upon which motion of the individual residues becomes less coherent and less restricted overall. To probe conformational-exchange processes on the microsecond-to-millisecond timescale, we finally performed 15 N CPMG relaxation dispersion (RD) measurements on wildtype protein and the mutants (Fig. S19). Data quality was insufficient to perform a more comprehensive analysis, but exchange contributions R ex to the effective transverse relaxation values could be obtained with reasonable accuracy, offering residue-specific identification of transient conformational substates. In RD experiments, generally, the exchange contribution R ex is derived from the difference in R 2eff measured at varying CPMG refocusing efficiency. High R ex values are indicative of residues undergoing chemical exchange (allosteric transitions, functional loop dynamics, or domain breathing motions) between conformers in particular on the fast ms timescale, being stronger both, for elevated excited-state populations and increased chemical-shift differences for those transient states. Globally, wild-type samples exhibited the largest number of residues with significant R ex contributions throughout the N-lobe and the C-lobe, indicating widespread conformational plasticity (Fig. S19A-B). As expected, a highly similar pattern as for the wild type is observed for V38A, both with respect to the residue level (Fig. S19A) and regarding the distribution of R ex contributions depicted in the form of a histogram (Fig. S19B), consistent with its position outside the dynamic network and confirming its dynamically silent nature in terms of allosteric communication. The mutant construct R49A shows a slight loss of ms timescale exchange in the N-lobe and a slight increase in the C-lobe, confirming the above reorganization of motional behavior (Fig. S19C). For mutants D150A and D168A, by contrast, a slight decrease of R ex in the C-lobe and a slight increase in the N-lobe was observed (Fig. S18C). All of these changes are, however, rather sporadic and any trends are difficult to discern. The apparent lack of strong differences between the samples results on the one hand from a rather noticeable error associated with the individual R 2eff rates. Fig. S18D exemplifies this for A172, a residue lying in between the DFG motif and the activation loop, via comparison of dispersion curves for the different samples. A number of other dispersion curves are shown in Fig. S19. However, the higher overall similarities across the R ex histograms in Fig. S18B suggests that the differences induced by the perturbations to the dynamic networks, as consistently seen upon CSPs, hetNOEs, and R 2 analyses, are rather limited to motions faster than the ms regime. Discussion The results presented here demonstrate that p38α kinase harbors a dynamic allosteric network that connects the N- and C-lobes and explicitly incorporates the lipid-binding domain and its associated lipid pocket. Mutations at residues identified as central nodes within this network induce substantial perturbations in both local and global dynamics, as evidenced by a consistent set of computational and experimental observations, including molecular dynamics simulations, dynamical network analysis, pathway-level correlation changes, chemical-shift perturbations, heteronuclear NOE measurements, and transverse relaxation rates. While ligand binding partially modulates these effects, the dominant impact of network disruption is observed in the apo ensembles. Collectively, these findings provide experimental validation for the simulation-derived prediction that the p38α scaffold supports long-range dynamic communication between distinct functional regions of the protein. Whereas the firm connectivity between dynamic-network residues tethers the plastic kinase architecture tightly together, thereby conducting concerted dynamics across long distances, mutations at key network sites abrogate long-range communication, entailing increased fast-timescale flexibility and conformational exchange. This conclusion can rationalize prior observations in a more biological context in a residue-specific manner. For example, local perturbations, such as activation-loop phosphorylation or binding of upstream regulators, can induce dynamical changes at distal sites within the kinase, albeit on slower timescales than those primarily interrogated here 15 . With the revelation of a widespread dynamic network prominently involving the activation loop and encompassing in particular the active site, the top part of the C-lobe, and the lipidic pocket, the various prior observations hinting to dynamic allostery across the kinase architecture and regulation come to no surprise. Several residues identified as key network nodes coincide with sites previously implicated in pathological or regulatory phenotypes, despite not being directly involved in catalysis. D150, located in proximity to the catalytic site 44 of the protein, forms an intramolecular H-bond with T185 that is required for TAB1-induced autoactivation 45 . R49K/A is used as a convenient tool mutation for studying PRMT1-mediated control of p38α 46 . Also D168 mutation is known to render the enzyme dysfunctional and has been used as a valuable mechanistic probe 47, 48, 49 . Its prominent functional role, being key for the in/out conformational switch of the DFG motif, required for exchanging ATP/ADP at the active site, clearly exceeds the role as a mere dynamics-transducing element of the network. It is still important to realize that this site can also be ascribed a prominent role in transducing information across the protein scaffold, which would mean an impact on the motional properties and allosteric sensing associated with the conformational changes at the DFG motif. Importantly, our data establish the lipid-binding domain as an integral component of the p38α dynamic network, which provides further prospects for pharmacological avenues based on lipid-pocket-binding allosteric modulators. 7 Taken together, these observations demonstrate how wide-spread dynamic networks, rather than the local mobility associated with the individual structural elements, underly the integration of regulatory signals in p38α and likely in other kinases as well. Beyond the shear identification/mapping of communication pathways between the p38a N- and C-lobes, in particular including the so-far largely unexploited lipid pocket, in the presence and absence of the inhibitor, the above results may also offer a conceptual framework for the in-silico identification of possible allosteric sites in a kinase drug discovery context. The knowledge about, in particular, the existence and whereabouts of surface sites associated to dynamic networks could facilitate the design of next-generation allosteric modulators that do not target the known pockets, such as the ATP binding site or the “allosteric pocket”. It is so far unclear whether the lipid pocket will ever be addressed pharmacologically. The confirmation that a dynamic network extends to this site and in particular the identification of W207 at the rim of the lipid pocket suggests the site may serve as a useful starting point for new allosteric p38α inhibitors. With first-generation, low-affinity binders (lipid pocket ligands), lipid pocket occupation (at the low populations obtained) has not been found to bear a strong impact on kinase activity. 7 However, given that for a desired systemic effect even a weak molecular impact can suffice, the above data encourage further work, involving bulkier or sterically optimized ligands, which might result in new perspectives to overcome the off-site effects usually associated with targeting kinases. More generally, awareness and spatial mapping of dynamic networks in enzymes may guide the identification of regulatory sites amenable to therapeutic intervention or rational engineering. It remains to be seen whether dynamic allostery will prove valuable not only for pharmacological inhibition but also for fine-tuning of enzymatic activity in biotechnological applications through targeted allosteric mutations 50, 51, 52 . In summary, we identify and map a dynamic allosteric network within p38α MAP kinase that spans both lobes, prominently involves inter-lobe segments of the architecture and the activation loop, and explicitly includes the lipid-binding domain and its associated lipid pocket. By integrating computational and experimental structural biology approaches, we reveal specific residues that act as key information relays within a coordinated motional framework, transmitting dynamics across spatially distant regions of the protein. Mutations at these hotspots largely corrupt the integrity of the network, thereby significantly reducing inter-lobe communication and inducing global changes in residue-specific dynamics, underscoring their central role in long-range allostery. Together, these findings provide new insights into the dynamic basis of kinase regulation and pathology. With the lipid pocket surface, in particular residue W207, found to form an integral part of the dynamic network, the study also highlights opportunities for identifying and targeting allosteric sites in future pharmacological strategies. Methods Dynamical Network Analysis All atomic molecular dynamics (MD) simulations were performed using NAMD 3.0 27 in conjunction with the QwikMD plugin 53 . To quantify ensemble-level differences associated with the reported ligand-dependent conformational changes, simulations were initiated from two crystallographic structures of p38α in which the inhibitor ring flip was observed (PDB IDs: 3HEG 13 and 3GCS 14 ). For apo p38α (1WFC 29 ), missing segments were rebuilt by comparative modeling. NOESY-derived restraints were applied during energy minimization (20,000 steps), slow heating to 300 K, and a ~1 μs restrained equilibration; restraints were then released and five independent 1 μs production replicas were generated. Dynamical network analysis identified betweenness-centrality hotspots, which were alanine-scanned using QwikMD; each mutant underwent the same minimize–anneal–equilibrate protocol, followed by five 1 μs replicas for correlation and network metrics. Sorafenib was docked into the allosteric pocket (Table S6), prepared identically, and simulated in five 1μs replicas. Correlations and communication pathways were compared across apo, mutant, and ligand-bound ensembles. A total of ~50 μs simulations are used for this study. All simulations employed the CHARMM36 29 force field for proteins and ligands. Covalent bonds involving hydrogen atoms were constrained using SHAKE, and equations of motion were integrated using a 1 fs time step. Long-range electrostatics were treated using the particle-mesh Ewald method under periodic boundary conditions, with a real-space cutoff of 12 Å and a pair-list distance of 14 Å. Systems were equilibrated under NVT conditions for thermalization and NPT conditions for pressure coupling prior to production. Temperature was maintained at 300 K using Langevin dynamics, and pressure was controlled at 1 atm using the Langevin piston method under periodic boundary conditions. The topology and trajectory files from the MD simulations were incorporated into the Python notebooks of Dynetan 25 . The software uses the incorporation of MDAnalysis to analyze the trajectory simulation files 54 . Contact detection optimization of the code was already performed via Numba 55 and Cython 56 . Network statistics and the determination of optimal paths were carried out using the Floyd–Warshall algorithm, provided by the NetworkX package. For Floyd–Warshall calculations, the “distance” between nodes was defined as d = −log(r MI ), which uses mutual information as a method for contact detection consistent with previous applications of this method which can be seen as: r MI [ i,j ] = (1- e -2/3 I [ i,j ] ) 1/2 where i and j are the position of the alpha carbon of each residue and I [ i,j ] are the mutual information between them which is computed by the density estimator described as: I [ i,j ] = ψ (k) - 1/k - ‹ ψ ( n i ) + ψ ( n j )› + ψ ( N ) where here N is the total number of simulation frames, ψ (x) is the digamma function, n i is the number of frames in which residue i is close to the one in the reference, and stands for the average of the trajectory, using a neighboring parameter k of 6. Each amino acid residue is regarded as a node. The network is structured by these nodes and nodes lie within a cutoff distance (4.5 Å) for at least 75% of an MD trajectory. The residues which lie farther than the cutoff value are excluded which mainly includes some solvent and ion molecules. The nodes are connected via links known as “edges”. Betweenness centrality was computed as : where σ (s, t) is the number of shortest paths between nodes s and t, V is the ensemble of graph nodes, and C is a normalization factor. The simulations were analyzed using in-house Python and TCL scripts along with VMD 57 . Contact maps for residue–residue interactions were generated with VMD and PyContact. For the statistical testing, we normalized all measurements by the cumulative of the WT (or LWT) row. For each variant we report the per-variant mean and plot error bars as the sample standard deviation across replicates. Group comparisons to WT used a non-parametric permutation bootstrap: we pooled the two groups, permuted labels, recomputed the mean difference, and repeated this 10,000 times (two-sided). The p-value is the proportion of permuted differences whose absolute value exceeded the observed absolute difference. No parametric assumptions (normality or equal variances) were required and plotted using python notebooks. All the structural figures for simulations are rendered through VMD. The figures showing residue-wise properties are extracted through Chimera attributes. Sample Preparation The plasmid map of the kinase was elucidated by nanopore sequencing to entail the specific design of mutagenesis primers. The primers for the mutants were designed such that the mutation residue lies at the center for the forward and reverse primers. Phusion™ High-Fidelity DNA Polymerase , a fusion polymerase containing a Mastermix of nucleotides, enzymes, and buffers were used to carry out the polymerase chain reaction. The PCR product was digested using the DpnI enzyme at 37 °C for 1 hr and was immediately transformed overnight at 37 °C to E.coli XL 10 gold cells. The transformed colonies were grown in an overnight culture at 37 °C and the plasmid was extracted using a PEG gold miniprep kit and sent to Eurofins for Sanger Sequencing. The wildtype as well as the mutants was expressed into BL21(DE3) E. coli cells using 13 C and 15 N labelling for assignment and relaxation experiments. D 2 O adaptation was carried out and the cultures were grown in 1 L at 37 °C to reach the OD 600 of 0.6-0.7. They were then cooled for 30 min to room temperature, and induced with 1 mM IPTG overnight (∼20 h) at 18 °C while shaking at 180 rpm. For the mutant W207A, expression in deuterated minimal media could not be achieved. The cells were lysed using a 50 mM Tris Buffer, benzonase, and lysozyme in the presence of a protease inhibitor, centrifuged, and the supernatant was exposed to affinity chromatography (Binding Buffer: 50 mM Tris, 500 mM NaCl, 25 mM Imidazole and 5 % Glycerol; Elution Buffer: 50 mM Tris, 500 mM NaCl, 500 mM Imidazole and 5 % Glycerol) using a nickel column. The eluted fractions were dialyzed to a pH of 7 and were subjected to an overnight cleavage in the presence of thrombin. Anion exchange (Binding Buffer: 25 mM Hepes and 5 % Glycerol; Elution Buffer: 25 mM Hepes, 1 M NaCL, 5 % Glycerol) and Size Exclusion Chromatography (SEC Buffer: 20 mM Hepes, 50 mM NaCl, 100 mg/L methionine and 5 % Glycerol) was then used to obtain pure protein. HTRF Analysis Wild-type and mutant p38α constructs (unlabelled) were activated with constitutively active MKK6 S207E/T211E (Thermo Scientific, Lot 877061F) in activation buffer (50 mM Tris, 10 mM MgCl₂, 1 mM ATP, 1 mM DTT, 0.001% Tween-20, pH 7.4) at 37 °C for 90 min with shaking at 400 rpm. Reactions were dialyzed overnight at 4 °C into storage buffer (20 mM HEPES, 50 mM NaCl, 5% glycerol, pH 7.1), concentrated to ~0.2 mg/mL, and stored at −80 °C. ATP K M was determined using the HTRF® KinEASE™ kit (Cisbio) following the manufacturer’s instructions. Activated p38α was added per well at 0.04 ng (WT), 3.5 ng (mutant/unlabelled), or 20 ng (mutant/labelled) and incubated for 10/10/20 min, respectively, with 1 µM GST-ATF2 substrate across 0.4–900 µM ATP in reaction buffer (50 mM HEPES, 0.1 mM Na₃VO₄, 0.02% NaN₃, 0.01% w/v BSA, 10 mM MgCl₂, 1 mM MnCl₂, 1 mM DTT, 0.01% Triton X-100, pH 7.0) in 384-well black flat-bottom plates (Greiner Bio-One). Reactions were stopped with detection solution (50 mM HEPES, 0.1% w/v BSA, 800 mM KF, 20 mM EDTA, 0.666 nM anti-phospho-ATF2-Eu(K) antibody, 100 nM anti-GST-d2 antibody, pH 7.0) and incubated 60 min at room temperature. Time-resolved fluorescence was read on an EnVision 2104 (PerkinElmer) at 620 nm and 665 nm, 60 µs after excitation at 317 nm. The 665/620 signal ratio was plotted versus ATP concentration and fitted to the Michaelis–Menten equation in Origin (OriginLab). Solution NMR The protein from the SEC Buffer was exchanged to “NMR Buffer” containing 50 mM of HEPES and 150 mM NaCl (pH ~6.8), and the sample was concentrated to approximately 450 µM. NMR experiments were recorded on an 800 MHz Bruker NEO spectrometer. 2D 15 N- 1 H TROSY HSQC and 3D TROSY HNCA were recorded to gain the assignments for the mutants via assignment transfer from the BMRB (entry: 17471). CCPN version 3.1 was used for assigning the mutant peaks 58 . 15 N- 1 H TROSY hetNOE was performed to assess fast-timescale dynamics and was analysed using a CCPN macro for relaxation. 15 N TROSY-CPMG experiments were acquired in an interleaved manner as follows for the CPMG frequencies: 0, 2000, 25, 1500, 1000, 100, 750, 200, 1250, 500, 200 Hz with a CPMG delay of a total of 0.022 s and a recycling delay of 1.5 seconds. The experiments were analyzed with NESSY software 59 . Graphs and analysis were produced using in house python scripts. All the raw data has been deposited in the public repository of Technical University Dortmund (https://data.tu-dortmund.de/previewurl.xhtml?token=ff748a60-a45c-4c14-8f09-b7e397880aca) and can be freely accessed 60 . Declarations Acknowledgements Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy - EXC 2033 – 390677874 – RESOLV, and EXC-114 – 24286268 – CiPS-M. Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 27112786, 325871075 and the Emmy Noether program. Funded/co-funded by the European Union (ERC, 101082494 bypassNMR). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. The authors gratefully acknowledge the computing time provided on the Delta-AI clusters, the clusters of the Bernardi group in Auburn University and the Linux HPC cluster at Technical University Dortmund (LiDO3), which is partially funded in the course of the Large-Scale Equipment Initiative by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) as project 271512359. R.C.B. was supported by the National Science Foundation under Grant MCB-2143787. Computational resources were provided in part through NSF ACCESS (Allocation BIO250251). This research made use of the DeltaAI advanced computing and data resource, supported by the National Science Foundation (Award OAC-2320345) and the State of Illinois, as well as the Delta advanced computing and data resource, supported by the National Science Foundation (Award OAC-2005572) and the State of Illinois. Delta and DeltaAI are joint efforts of the University of Illinois Urbana-Champaign and the National Center for Supercomputing Applications. References Perry JJ, Harris RM, Moiani D, Olson AJ, Tainer JA. p38alpha MAP kinase C-terminal domain binding pocket characterized by crystallographic and computational analyses. J. Mol. Biol. 391 , 1-11 (2009). Ung PM, Thompson AD, Chang L, Gestwicki JE, Carlson HA. Identification of key hinge residues important for nucleotide-dependent allostery in E. coli Hsp70/DnaK. PLoS Comput. Biol. 9 , e1003279 (2013). Wang Z, Harkins PC, Ulevitch RJ, Han J, Cobb MH, Goldsmith EJ. The structure of mitogen-activated protein kinase p38 at 2.1-A resolution. Proc. Natl. Acad. Sci. U. S. A. 94 , 2327-2332 (1997). Wrobleski ST, Doweyko AM. Structural comparison of p38 inhibitor-protein complexes: a review of recent p38 inhibitors having unique binding interactions. Curr. Top. Med. Chem. 5 , 1005-1016 (2005). Grossi V , et al. Sorafenib inhibits p38α activity in colorectal cancer cells and synergizes with the DFG-in inhibitor SB202190 to increase apoptotic response. Cancer Biol. Ther. 13 , 1471-1481 (2012). Diskin R, Engelberg D, Livnah O. A novel lipid binding site formed by the MAP kinase insert in p38 alpha. J. Mol. Biol. 375 , 70-79 (2008). Bührmann M, Hardick J, Weisner J, Quambusch L, Rauh D. Covalent Lipid Pocket Ligands Targeting p38α MAPK Mutants. Angew. Chem., Int. Ed. 56 , 13232-13236 (2017). Zhang XH , et al. Targeting the non-ATP-binding pocket of the MAP kinase p38γ mediates a novel mechanism of cytotoxicity in cutaneous T-cell lymphoma (CTCL). FEBS Lett. 595 , 2570-2592 (2021). Cooper A, Dryden DT. Allostery without conformational change. A plausible model. Eur. Biophys. J. 11 , 103-109 (1984). Monod J, Wyman J, Changeux JP. ON THE NATURE OF ALLOSTERIC TRANSITIONS: A PLAUSIBLE MODEL. J. Mol. Biol. 12 , 88-118 (1965). Kuriyan J, Eisenberg D. The origin of protein interactions and allostery in colocalization. Nature 450 , 983-990 (2007). Koshland DE, Jr., Némethy G, Filmer D. Comparison of experimental binding data and theoretical models in proteins containing subunits. Biochemistry 5 , 365-385 (1966). Namboodiri HV, Bukhtiyarova M, Ramcharan J, Karpusas M, Lee Y, Springman EB. Analysis of imatinib and sorafenib binding to p38alpha compared with c-Abl and b-Raf provides structural insights for understanding the selectivity of inhibitors targeting the DFG-out form of protein kinases. Biochemistry 49 , 3611-3618 (2010). Simard JR , et al. Development of a fluorescent-tagged kinase assay system for the detection and characterization of allosteric kinase inhibitors. J. Am. Chem. Soc. 131 , 13286-13296 (2009). Kumar GS , et al. Dynamic activation and regulation of the mitogen-activated protein kinase p38. Proc. Natl. Acad. Sci. U. S. A. 115 , 4655-4660 (2018). Medina Gómez S, Homberg LT, Bührmann M, Rauh D, Linser R. Lipid Pocket Binders Impose Allosteric Changes of Protein Dynamics Around the Active Site of the Protein Kinase p38α Angew. Chem., Int. Ed. , DOI: 10.1002/anie.202522665 (2026). Aoto PC, Martin BT, Wright PE. NMR Characterization of Information Flow and Allosteric Communities in the MAP Kinase p38γ. Sci. Rep. 6 , 28655 (2016). Aoto PC, Stanfield RL, Wilson IA, Dyson HJ, Wright PE. A Dynamic Switch in Inactive p38γ Leads to an Excited State on the Pathway to an Active Kinase. Biochemistry 58 , 5160-5172 (2019). Avitzour M, Diskin R, Raboy B, Askari N, Engelberg D, Livnah O. Intrinsically active variants of all human p38 isoforms. FEBS J. 274 , 963-975 (2007). Ahuja LG, Kornev AP, McClendon CL, Veglia G, Taylor SS. Mutation of a kinase allosteric node uncouples dynamics linked to phosphotransfer. Proc. Natl. Acad. Sci. U. S. A. 114 , E931-E940 (2017). Foda ZH, Shan Y, Kim ET, Shaw DE, Seeliger MA. A dynamically coupled allosteric network underlies binding cooperativity in Src kinase. Nat. Commun. 6 , 5939 (2015). Zhang X, Gureasko J, Shen K, Cole PA, Kuriyan J. An Allosteric Mechanism for Activation of the Kinase Domain of Epidermal Growth Factor Receptor. Cell 125 , 1137-1149 (2006). Skora L, Mestan J, Fabbro D, Jahnke W, Grzesiek S. NMR reveals the allosteric opening and closing of Abelson tyrosine kinase by ATP-site and myristoyl pocket inhibitors. Proc. Natl. Acad. Sci. U. S. A. 110 , E4437-E4445 (2013). Hu J , et al. Kinase regulation by hydrophobic spine assembly in cancer. Mol. Cell. Biol. 35 , 264-276 (2015). Melo MCR, Bernardi RC, de la Fuente-Nunez C, Luthey-Schulten Z. Generalized correlation-based dynamical network analysis: a new high-performance approach for identifying allosteric communications in molecular dynamics trajectories. J. Chem. Phys. 153 , 134104 (2020). Erin S, George PL. Analysis of coordinated NMR chemical shifts to map allosteric regulatory networks in proteins. Methods 209 , 40-47 (2023). Phillips JC , et al. Scalable molecular dynamics on CPU and GPU architectures with NAMD. J. Chem. Phys. 153 , 044130 (2020). Haque A , et al. Interaction Analysis of MRP1 with Anticancer Drugs Used in Ovarian Cancer: In Silico Approach. Life 12 , 383 (2022). Wilson KP , et al. Crystal structure of p38 mitogen-activated protein kinase. J. Biol. Chem. 271 , 27696-27700 (1996). Krishnan K, Tian H, Tao P, Verkhivker GM. Probing conformational landscapes and mechanisms of allosteric communication in the functional states of the ABL kinase domain using multiscale simulations and network-based mutational profiling of allosteric residue potentials. J. Chem. Phys. 157 , 245101 (2022). Taylor SS, Zhang P, Steichen JM, Keshwani MM, Kornev AP. PKA: lessons learned after twenty years. Biochim. Biophys. Acta 1834 , 1271-1278 (2013). Ahuja LG, Aoto PC, Kornev AP, Veglia G, Taylor SS. Dynamic allostery-based molecular workings of kinase:peptide complexes. Proc. Natl. Acad. Sci. U. S. A. 116 , 15052-15061 (2019). Panjarian S, Iacob RE, Chen S, Engen JR, Smithgall TE. Structure and dynamic regulation of Abl kinases. J. Biol. Chem. 288 , 5443-5450 (2013). Griswold IJ , et al. Kinase domain mutants of Bcr-Abl exhibit altered transformation potency, kinase activity, and substrate utilization, irrespective of sensitivity to imatinib. Mol. Cell. Biol. 26 , 6082-6093 (2006). Shan Y , et al. Oncogenic mutations counteract intrinsic disorder in the EGFR kinase and promote receptor dimerization. Cell 149 , 860-870 (2012). Shah NG , et al. Novel Noncatalytic Substrate-Selective p38α-Specific MAPK Inhibitors with Endothelial-Stabilizing and Anti-Inflammatory Activity. J. Immunol. 198 , 3296-3306 (2017). Taylor SS, Kornev AP. Protein kinases: evolution of dynamic regulatory proteins. Trends Biochem. Sci. 36 , 65-77 (2011). Kornev AP, Taylor SS. Dynamics-Driven Allostery in Protein Kinases. Trends Biochem. Sci. 40 , 628-647 (2015). Huse M, Kuriyan J. The conformational plasticity of protein kinases. Cell 109 , 275-282 (2002). BIOVIA. BIOVIA Discovery Studio Visualizer v21.1. Dassault Systèmes, San Diego, SA, USA , (2023). Degorce F, Card A, Soh S, Trinquet E, Knapik GP, Xie B. HTRF: A technology tailored for drug discovery - a review of theoretical aspects and recent applications. Curr. Chem. Genomics 3 , 22-32 (2009). Keesler GA , et al. Purification and activation of recombinant p38 isoforms alpha, beta, gamma, and delta. Protein Expr. Purif. 14 , 221-228 (1998). Holliday MJ, Camilloni C, Armstrong GS, Vendruscolo M, Eisenmesser EZ. Networks of Dynamic Allostery Regulate Enzyme Function. Structure 25 , 276-286 (2017). Sok P , et al. MAP Kinase-Mediated Activation of RSK1 and MK2 Substrate Kinases. Structure 28 , 1101-1113.e1105 (2020). Thapa D , et al. TAB1-Induced Autoactivation of p38α Mitogen-Activated Protein Kinase Is Crucially Dependent on Threonine 185. Mol. Cell. Biol. 38 , (2018). Liu MY, Hua WK, Chen CJ, Lin WJ. The MKK-Dependent Phosphorylation of p38α Is Augmented by Arginine Methylation on Arg49/Arg149 during Erythroid Differentiation. Int. J. Mol. Sci. 21 , (2020). Ambrosino C , et al. Negative feedback regulation of MKK6 mRNA stability by p38alpha mitogen-activated protein kinase. Mol. Cell. Biol. 23 , 370-381 (2003). Huang S, New L, Pan Z, Han J, Nemerow GR. Urokinase plasminogen activator/urokinase-specific surface receptor expression and matrix invasion by breast cancer cells requires constitutive p38alpha mitogen-activated protein kinase activity. J. Biol. Chem. 275 , 12266-12272 (2000). McLaughlin MM , et al. Identification of mitogen-activated protein (MAP) kinase-activated protein kinase-3, a novel substrate of CSBP p38 MAP kinase. J. Biol. Chem. 271 , 8488-8492 (1996). Bhattacharya S , et al. NMR-guided directed evolution. Nature 610 , 389-393 (2022). Saavedra HG, Wrabl JO, Anderson JA, Li J, Hilser VJ. Dynamic allostery can drive cold adaptation in enzymes. Nature 558 , 324-328 (2018). Singh H, Das CK, Buchmuller BC, Schäfer LV, Summerer D, Linser R. Epigenetic CpG duplex marks probed by an evolved DNA reader via a well-tempered conformational plasticity. Nucleic Acids Res. 51 , 6495-6506 (2023). Ribeiro JV , et al. QwikMD - Integrative Molecular Dynamics Toolkit for Novices and Experts. Sci. Rep. 6 , 26536 (2016). Michaud-Agrawal N, Denning EJ, Woolf TB, Beckstein O. MDAnalysis: a toolkit for the analysis of molecular dynamics simulations. J. Comput. Chem. 32 , 2319-2327 (2011). Lam SK, Pitrou A, Seibert S. Numba: a LLVM-based Python JIT compiler. In: Proceedings of the Second Workshop on the LLVM Compiler Infrastructure in HPC ). Association for Computing Machinery (2015). Behnel S, Bradshaw R, Citro C, Dalcin L, Seljebotn DS, Smith K. Cython: The Best of Both Worlds. Comput. Sci. Eng. 13 , 31-39 (2011). William H, Andrew D, Klaus S. VMD: Visual molecular dynamics. J. Mol. Graph. Model. 14 , 33-38 (1996). Skinner SP, Fogh RH, Boucher W, Ragan TJ, Mureddu LG, Vuister GW. CcpNmr AnalysisAssign: a flexible platform for integrated NMR analysis. J. Biomol. NMR 66 , 111-124 (2016). Bieri M, Gooley PR. Automated NMR relaxation dispersion data analysis using NESSY. BMC Bioinformatics 12 , 421 (2011). Roy Acharyya S. Identification of allosteric communication pathways within p38 alpha kinase from dynamical network analysis and NMR spectroscopy.). DRAFT VERSION edn. TUDOdata (2026). Additional Declarations There is NO Competing Interest. Supplementary Files checklist.pdf Checklist for the simulations SIclean.pdf Supporting Information pdf Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8958159","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":602033016,"identity":"aff2e36a-568a-449f-96dc-e95388752a6a","order_by":0,"name":"Rasmus Linser","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEElEQVRIiWNgGAWjYBADGTYwVQEiEqBiBxgM8GnhgWg5YwDRcoAYLWCSsY0ILebtx59u5vnDwMMn3f7wceW8P4kN7MnHPn/MOZzHd4B54wNsfjiTY3abtw3oMJkzxoZntxkkNvA8S55xcNvhYskDbMXYrJFgyGG7zdsA1CKRwybZCNSy/0aOMQNQS+KGAzxmEti08D9/dhvkMDaJ9Oc/G+cAbZFAaDH/gU2LRILZbWBwAbUkmDE2NqBqMcPmfQmJN2Y357ZJgBxmLNlwzNgY5BeGs9vSE2ceZivG7rD0Zzfe/LGRk5+R/vBjQ42cLDDEDjNUbrNO7DvevPEDNmtgoYAFMONWPwpGwSgYBaMAPwAARzJkV+wMNL0AAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-8983-2935","institution":"TU Dortmund University","correspondingAuthor":true,"prefix":"","firstName":"Rasmus","middleName":"","lastName":"Linser","suffix":""},{"id":602033017,"identity":"dc995860-c590-465f-a2d5-e31f9e50ebbd","order_by":1,"name":"Suchandra Acharyya","email":"","orcid":"","institution":"TU Dortmund University","correspondingAuthor":false,"prefix":"","firstName":"Suchandra","middleName":"","lastName":"Acharyya","suffix":""},{"id":602033018,"identity":"56676967-46ef-4afc-8430-35e0254a2c9a","order_by":2,"name":"Jörn Weisner","email":"","orcid":"https://orcid.org/0000-0002-6103-7371","institution":"TU Dortmund","correspondingAuthor":false,"prefix":"","firstName":"Jörn","middleName":"","lastName":"Weisner","suffix":""},{"id":602033019,"identity":"e5e15b0e-be6c-422a-80f7-ea611dfebadc","order_by":3,"name":"Rafael Bernardi","email":"","orcid":"https://orcid.org/0000-0003-0758-2026","institution":"Auburn University","correspondingAuthor":false,"prefix":"","firstName":"Rafael","middleName":"","lastName":"Bernardi","suffix":""}],"badges":[],"createdAt":"2026-02-24 13:46:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8958159/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8958159/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104404930,"identity":"93082291-34de-452d-9d5a-7799021b664a","added_by":"auto","created_at":"2026-03-11 12:21:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1306194,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe interdomain communication to the lipid binding domain in p38\u003c/strong\u003eα\u003cstrong\u003e in question.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e Structural representation of p38α kinase, highlighting the interactions of ligands binding to the allosteric pocket (exemplified for sorafenib) and the lipid pocket (shown for β-OG), as produced using Discovery studio\u003csup\u003e28\u003c/sup\u003e. Key structural features such as the activation loop, the catalytic C- spine, and the regulatory R-spine are highlighted. \u003cstrong\u003eb \u003c/strong\u003eObserved ring flip of the type-II inhibitor sorafenib between structures in the presence (semi-transparent yellow, PDB 3GCS) or absence (semi-transparent cyan, PDB 3HEG) of a lipid pocket binder (β-OG). \u003cstrong\u003ec\u003c/strong\u003e Sketch of the allosteric pathways between the active site, the adjacent, so-called allosteric pocket, and the lipid pocket in question.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/382eb4af39955d2d07e45744.png"},{"id":104404509,"identity":"9bc89d5b-d792-41e2-9394-2cd00472764a","added_by":"auto","created_at":"2026-03-11 12:20:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":721329,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDynamic network analysis of p38α kinase\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea \u003c/strong\u003eClustering of the protein into 14 communities, largely reminiscent of the different structural elements of the kinase (right). \u003cstrong\u003eb\u003c/strong\u003e Residues with top betweenness centrality, i.e., the highest degree of information flow traversing them, also highlighting those network sites chosen for mutagenesis studies for network validation (red highlights). \u003cstrong\u003ec\u003c/strong\u003e Map of inter-residue dynamic-network correlations, specifically focusing on connections between N- and C-lobe residues. Pairs in black writing denote long-range correlations with a correlation coefficient above 0.3. Note that part of the C-terminal (320-360) lies in the N-Lobe, while the first transition from N- to C-lobe occurs at around residue 107. Correlations involving these covalent linkages are grayed out.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/97e3b2b1f9e211dbfaecaf11.png"},{"id":104200477,"identity":"30a03417-c5bd-476f-bcf4-77996fa47fa6","added_by":"auto","created_at":"2026-03-09 05:17:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":588562,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparison of shortest-path correlations and kinase activity upon dynamic-network mutation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e Strategy to verify the results of dynamic-network analyses, using in-silico mutagenesis, functional assays of mutants, and NMR chemical-shift perturbation and relaxation. \u003cstrong\u003eb\u003c/strong\u003e Residue-wise depiction of the shortest communication path from the allosteric pocket to the activation loop. \u003cstrong\u003ec\u003c/strong\u003e Total correlation of residue pairs along the allosteric-pocket-to-activation-loop pathway, normalized to wild-type (WT = 1) and with statistical annotations ** for p \u0026lt; 0.05 and *** for p \u0026lt; 0.01. Statistical significance was assessed by non-parametric bootstrap tests, see the Methods for details. A marked reduction is observed in all mutants, with W207A exhibiting the most significant drop (p = 0.039). \u003cstrong\u003ed\u003c/strong\u003e Shortest path from the lipid pocket to the activation loop with key residues highlighted. \u003cstrong\u003ee\u003c/strong\u003e Correlation sums along the lipid-pocket-to-activation-loop pathway, normalized to WT, D168A highlighting a significant decrease (p = 0.043). \u003cstrong\u003ef\u003c/strong\u003e Residue pathway showing the shortest communication route from the allosteric pocket to the lipid pocket. \u003cstrong\u003eg\u003c/strong\u003e Total correlation for the allosteric-pocket-to-lipid-pocket pathway, normalized to WT, with all mutants showing statistically significant reduced correlations (p = 0.017 for R49A; p = 0.0026 for D150A; p = 0.0022 for D168A; p = 0.005 for W207A). \u003cstrong\u003eh\u003c/strong\u003e HTRF-based ATP titration assay comparing the integrity of ATP binding between wild-type and mutant p38α kinase constructs across a concentration range of 0–1000 μM ATP. The result of a Michaelis-Menten fit of the fluorescence ratio \u003cem\u003ef\u003c/em\u003e\u003csub\u003e\u003cem\u003e665\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e/f\u003c/em\u003e\u003csub\u003e\u003cem\u003e620\u003cbr\u003e\n\u003c/em\u003e\u003c/sub\u003eobtained after a given time in the initial-rate regime serves as a quantitative proxy for ATP binding (best-fit K\u003csub\u003eM\u003c/sub\u003e shown for each curve). \u003cstrong\u003ei\u003c/strong\u003e HTRF substrate titration assay quantifying substrate binding affinity of wild-type and mutant kinases using increasing concentrations of a GST-tagged activating transcription factor 2 (ATF2) substrate. K\u003csub\u003eM\u003c/sub\u003e fit values are again shown for each curve. While wild-type shows high fidelity of substrate association, the data for the dynamic-network mutants consistently indicate slightly impaired ATP and substrate binding competency.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/76aef7e966ff29f8806a8537.png"},{"id":104403798,"identity":"cd40657d-cb39-40f6-9cb3-f2902b5d3999","added_by":"auto","created_at":"2026-03-11 12:19:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":541145,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eChemical-shift perturbations (CSPs) induced upon mutation of key dynamic-network residues.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e Overlay of [\u003csup\u003e15\u003c/sup\u003eN, \u003csup\u003e1\u003c/sup\u003eH] TROSY-HSQC spectra for wild-type (black) and mutant D168A (red). The mutation results in significant spectral reorganization, including disappearance of the W207 resonance in the lipid pocket. (Note that, in contrast to 2D spectrum shown here, the actual CSPs were read out from 3D spectra.) \u003cstrong\u003eb\u003c/strong\u003e CSPs between wild-type and D168A mapped on the protein scaffold. Perturbations extend beyond the local environment of the mutation, indicating widespread conformational remodeling throughout the protein. CSPs for the other mutants are shown on the protein structure in Fig. S17.\u003cstrong\u003e c\u003c/strong\u003e Violin plots summarizing CSP distributions across network mutants (R49A, D150A, D168A) and the control variant (V38A). Blue and red dots mark residues from the lipid pocket and active site, respectively. \u003cstrong\u003ed\u003c/strong\u003e Quantitative comparison of CSPs as a function of sequence. Dynamic network mutants display various elevated perturbation even at structurally remote positions, whereas the control mutant V38A induces minimal CSPs outside its immediate vicinity.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/d1058ac27ac51a5e524fd775.png"},{"id":104404839,"identity":"f84b2b92-2ec2-44c8-87c1-06b8c43dd2fb","added_by":"auto","created_at":"2026-03-11 12:21:11","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":608990,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eChanges in p38\u003c/strong\u003eα\u003cstrong\u003e fast-timescale dynamics, as seen by heteronuclear NOEs and \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eR\u003c/strong\u003e\u003c/em\u003e\u003csub\u003e\u003cstrong\u003e2 \u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003erelaxation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003e Structural mapping of Δ\u003cem\u003eR\u003c/em\u003e₂ (mutant – WT) values onto the three-dimensional structure of p38α kinase. Red regions indicate residues with \u003cem\u003eR\u003c/em\u003e₂ higher in the mutants, predominantly localized to the catalytic core and distal C-lobe, supporting widespread conformational exchange. \u003cstrong\u003eb\u003c/strong\u003e Structural visualization of ΔhetNOE values. Blue regions highlight residues with reduced NOE values in the mutants, indicative of enhanced backbone flexibility. \u003cstrong\u003ec\u003c/strong\u003e Box plot distribution of \u003cem\u003eR\u003c/em\u003e₂ values for wild-type, dynamic network mutants (R49A, D168A, D150A), and the negative control (V38A). Active-site residues (red) and lipid pocket residues (blue) are specifically highlighted by colored spheres. All dynamic-network mutants exhibit significantly elevated \u003cem\u003eR\u003c/em\u003e₂ values compared to the wild-type and control, consistent with a global enhancement in slow conformational exchange. \u003cstrong\u003ed\u003c/strong\u003e Box plots of hetNOE values across the same variants. Dynamic network mutants show a marked reduction in NOE values, reflecting increased local flexibility, including at functionally critical sites. By contrast, the control mutant V38A shows limited perturbation, confirming an impact on the p38 dynamic network to be absent.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/0667dbf0293803b9d3eef24c.png"},{"id":104409120,"identity":"89c5d436-29e1-47d5-9820-9fc63b857399","added_by":"auto","created_at":"2026-03-11 12:44:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4138089,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/dd98f9a9-918f-4c73-822b-9f9581247858.pdf"},{"id":104200479,"identity":"03a86367-6736-46ad-8d65-3835a754ed2d","added_by":"auto","created_at":"2026-03-09 05:17:06","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":63830,"visible":true,"origin":"","legend":"Checklist for the simulations","description":"","filename":"checklist.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/797fa592c2d3477bf1d4b4fe.pdf"},{"id":104200483,"identity":"f4b4befe-4f72-4390-ba6f-5dce73320682","added_by":"auto","created_at":"2026-03-09 05:17:06","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":9276147,"visible":true,"origin":"","legend":"Supporting Information pdf","description":"","filename":"SIclean.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8958159/v1/678cfeb95cdce7a01e92118b.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Mapping of dynamic allostery within p38 alpha kinase via network analyses and NMR spectroscopy","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe family of p38 mitogen-activated protein kinases (MAPK) represents one of the principal elements within the MAPK signalling cascade, critically involved in mediating cellular responses to stress and inflammatory stimuli. A robust cluster of experimental evidence suggests that p38 can exert pro-oncogenic functions in various types of cancer\u003csup\u003e1\u003c/sup\u003e, which has made p38 one of the most important drug targets across a wide range of cancers. p38 kinases are prototypical kinases comprising a bi-lobal structural core that is divided into a smaller N-terminal and a larger C-terminal lobe. The catalytic active site lies between the two domains, forming a hinge bearing high flexibility, which is considered to be an important parameter for nucleotide exchange\u003csup\u003e2\u003c/sup\u003e. The regulation of p38\u0026alpha; activity involves several highly dynamic structural elements, including the catalytic loop, the N-terminal segment of the activation loop containing the conserved \u0026ldquo;DFG-motif\u0026rdquo;, the \u0026alpha;C helix, and the hinge region\u003csup\u003e3\u003c/sup\u003e. Drug design efforts for this kinase have been focused predominantly on small-molecule inhibitors of the ATP binding site, prohibiting the activating phosphorylation of residues T180 and Y182. Addressing this critical functional element, however, a component widely conserved across the entire kinome, entails dose-limiting adverse effects owing to non-selective inhibition.\u003c/p\u003e\n\u003cp\u003eConsequently, the identification of other potential locales that bind substrates, inhibitors, or allosteric effectors is of great interest\u003csup\u003e1\u003c/sup\u003e. Modulators that bind in the \u0026ldquo;DFG-out\u0026rdquo; conformation, exploiting an additional site opening up close to the active site, have also been designed\u003csup\u003e4\u003c/sup\u003e. Sorafenib (Nexavar\u0026reg;) is such a (\u0026ldquo;type II\u0026quot;) inhibitor, binding to this site but potently inhibiting nine other kinases\u003csup\u003e5\u003c/sup\u003e (\u003cstrong\u003eFig. 1a\u003c/strong\u003e). Its strategic use in combination with the drug SB202190 at the ATP-binding site, e.g., entails a synergistic effect that increases the apoptotic response in colorectal cancer cells\u003csup\u003e5\u003c/sup\u003e. However, the resemblance of this site to the conserved phosphate-binding site has resulted in little or no increase in the specificity of drug binding, toxicity, and dosage constraints. A rather distant, cryptic pocket is associated to the lipid-binding domain, hosting the \u0026ldquo;MAP kinase insert\u0026rdquo;\u003csup\u003e6\u003c/sup\u003e, which \u0026ndash; in conjunction with a significant conformational change \u0026ndash; has been shown to accommodate a range of lipophilic molecules like \u003cem\u003en\u003c/em\u003e-octyl-\u0026beta;-glucopyranoside (\u0026beta;-OG) as well as suitable covalent binders addressing a conserved Cys closeby\u003csup\u003e7\u003c/sup\u003e. This hidden pocket in the C-terminal region of the protein, the \u0026ldquo;lipid pocket\u0026rdquo;\u003csup\u003e6\u003c/sup\u003e (\u003cstrong\u003eFig. 1a\u003c/strong\u003e), has been shown to bear a primary sequence conserved over the p38 family and potentially allows for binding of more drug-like small molecules specifically generated in the quest for new allosteric p38 inhibitors\u003csup\u003e8\u003c/sup\u003e .\u003c/p\u003e\n\u003cp\u003eAllosteric communication refers to an intramolecular mechanistic crosstalk between spatially distant sites that participates in modulating functionality as a function of external events\u003csup\u003e9\u003c/sup\u003e. Allostery, in a wider sense, is the event in which one site somehow \u0026ldquo;feels\u0026rdquo; changes in a different site\u003csup\u003e10\u003c/sup\u003e\u003csup\u003e,\u003c/sup\u003e\u003csup\u003e11\u003c/sup\u003e. As such, it can either be based on significant structural changes or rather \u0026ndash; even in the complete absence of the latter \u0026ndash; hinge on dynamic features that are mutually dependent across the protein\u003csup\u003e12\u003c/sup\u003e. This latter mechanism, commonly referred to as \u0026ldquo;dynamic allostery\u0026rdquo; \u003csup\u003e9\u003c/sup\u003e, involves the redistribution of correlated motions across the protein, which can be modulated or induced by regulatory events such as ligand binding, protein\u0026ndash;protein interactions, or phosphorylation.\u003c/p\u003e\n\u003cp\u003eIn p38\u0026alpha; complexes, evidence for long-range dynamic coupling is provided by both structural and dynamical observations: Structural superimposition of the \u0026ldquo;DFG-out\u0026rdquo; p38\u0026alpha; complexes bound to Sorafenib alone (PDB 3HEG\u003csup\u003e13\u003c/sup\u003e) and to both Sorafenib and \u0026beta;-OG (PDB 3GCS\u003csup\u003e14\u003c/sup\u003e) reveals a flip of the pyridyl nitrogen and methyl substituent of Sorafenib, despite a separation of more than 30 \u0026Aring; from \u0026beta;-OG bound in the lipid pocket (\u003cstrong\u003eFig. 1b)\u003c/strong\u003e. This slight but noteworthy difference in ground-state structures suggests that ligand binding at the distal lipid pocket alters the conformational ensemble of the kinase in solution. In addition, activation of p38a incurs dynamics differentially modulated across timescales, with activation-loop phosphorylation having been observed to quench ps\u0026ndash;ns motions without altering the average conformation. Instead, uniform \u0026micro;s\u0026ndash;ms backbone dynamics are induced by substrate binding, flattening the energy landscape and rendering key allosteric sites accessible\u003csup\u003e15\u003c/sup\u003e. An cross-lobe interdependency in the realm of dynamic allostery also seems to involve the lipid pocket, occupation of which was observed to lead to widespread chemical-shift perturbations and changes in dynamics\u003csup\u003e16\u003c/sup\u003e. Similaly, for p38g kinase, another one of the four MAPK isoforms and a close relative to p38a with ~60 % sequence identity, motional changes in the kinase due to distant effectors have been assessed from various angles\u003csup\u003e17, 18, 19\u003c/sup\u003e. These observations align with the identification of long-range community networks in protein kinase A (PKA), fundamentally orchestrating overall dynamics and hence functionality\u003csup\u003e20\u003c/sup\u003e, as well as findings in the Scr kinases\u003csup\u003e21\u003c/sup\u003e, epidermal growth factor receptor (EGFR)\u003csup\u003e22\u003c/sup\u003e, and Abl kinase\u003csup\u003e23\u003c/sup\u003e, where catalytic activity was seen to be modulated by very distant sites.\u003csup\u003e19\u003c/sup\u003e Corroborating Cooper and Dryden\u0026rsquo;s original theory, activation of PKA has recently been ascribed to a redistribution of fast and intermediate-timescale thermal fluctuations (the \u0026ldquo;violin model\u0026rdquo;), which explains the lack of apparent structural changes. Thereby, the formation of \u0026ldquo;hydrophobic spines\u0026rdquo;, specifically the C-spine assembled upon ATP binding, and the R-spine, linked to positioning of the \u0026alpha;C-helix, an assembly regulated by upstream processes (also shown in \u003cstrong\u003eFig. 1a)\u003c/strong\u003e, is thought to aid in orchestrating suitable dynamic networks.\u003csup\u003e24\u003c/sup\u003e In MAP kinases, insights into such motional dependencies are of general interest from a biological perspective. However, they are of particular importance in the light of a motional connectivity between the different regulatory sites, the active site, or the lipid binding domain. A residue-specific mapping of such a dynamic network and identification of its key participants may aid understanding the molecular underpinnings of p38 regulation and even launch future pharmacological avenues based on the underlying pathways (\u003cstrong\u003eFig. 1c\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eAn in-depth interrogation of dynamic networks has become possible using molecular-dynamics simulations in combination with dynamic network analyses, where the correlations of motions assessed as generalised correlation coefficients based on mutual information are exploited to determine the strength of dynamic connectivity between different residues of the system.\u003csup\u003e25\u003c/sup\u003e Whereas these in-silico findings are extremely rich in information, they do require verification by experimental means. Providing site-specific access to a large range of chemical and motional parameters, NMR spectroscopy is particularly well suited as a complement to such simulations\u003csup\u003e26\u003c/sup\u003e. Here, we use state-of-the-art computational approaches based on a dynamical network analysis software (Dynetan)\u003csup\u003e25\u003c/sup\u003e to characterize p38 dynamic networks from an NAMD\u003csup\u003e27\u003c/sup\u003e-based molecular-dynamics simulation. Perturbing those residues that turn out to be pivotal for the motional network through site-directed mutagenesis, we then use NMR in solution to monitor the consequences for the various mutants experimentally. Together, these data reveal an extended dynamic network spanning both kinase lobes and comprising several distinct allosteric \u0026ldquo;hotspots\u0026rdquo; in p38\u0026alpha;. In addition to features common to other kinases, our analysis identifies the MAP kinase-specific lipid-binding domain and its associated cryptic lipid pocket as integral components of the dynamic network, highlighting their potential relevance for allosteric therapeutic intervention.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eFollowing the workflow summarized in Fig. S1, and the system setup summarized in Table. S1, we performed molecular dynamics (MD) simulations of apo p38\u0026alpha;, sorafenib-bound p38\u0026alpha;, and p38\u0026alpha; bound to both sorafenib and \u0026beta;-OG using NAMD 3\u003csup\u003e27\u003c/sup\u003e with the CHARMM36 force field\u003csup\u003e29\u003c/sup\u003e. (See methodological details in the SI text. For the apo protein, Figs. S2 and S3 show root-mean-square deviations as a function of time as well as root-mean-square fluctuations as a function of residue.) For each system, five independent replica simulations of 1 \u0026micro;s each were carried out. For subsequent network and correlation analyses, the final 200 ns of each replica were concatenated and analyzed using Dynetan\u003csup\u003e25\u003c/sup\u003e (described in more detail in the Methods and graphically summarized in Fig. S4). This analysis included community detection and quantification of betweenness centrality, which identifies residues that are central to allosteric communication.\u003c/p\u003e\n\u003cp\u003eCommunity detection, the first output of network analyses, partitions the network into clusters that share common motion. This enables probing of consistency between the rich, atomic-resolution MD data here with coarser experimental studies on this or similar systems in the past. For the wild type, apo protein, this analysis yields 14 clusters, shown by different colors in \u003cstrong\u003eFig. 2a\u003c/strong\u003e. Here, each community denotes a group of residues that exhibit strongly correlated motions with each other but weaker correlations with residues outside the group, reflecting coherent dynamical behavior within the protein. Despite the ~60\u0026nbsp;% sequence identity between the isoforms only as well the stark differences in methodology, the clusters obtained from the in-silico assessment here have a remarkable congruency to those (15) clusters found via purely experimental analyses (methyl scanning and chemical-shift perturbations) for p38g\u003csup\u003e17\u003c/sup\u003e: Like in this previous study, our clusters divide the N-lobe into different active-site (magenta) and allosteric-pocket (brown) regions and take accountability of the R-spine (gray) and C-spine regions (pink). The C-lobe is divided into three clusters, where the MAP kinase insert (lipid binding domain) is the cluster shown in green in Fig. 2a and deviates from the cluster containing the activation loop (shown in purple). Notably, several of these dynamic features also align with community structures identified previously for a different kinase (protein kinase A, PKA) using Girvan\u0026ndash;Newman-based network analysis\u003csup\u003e20\u003c/sup\u003e. Here, communities were segregated into 13 regions: In the N-lobe, the ATP binding pocket (magenta) is reminiscent of the PKA \u0026ldquo;Com A1\u0026rdquo;, the C-helix; the extended allosteric pocket of the p38\u0026alpha; (brown) corresponds to \u0026ldquo;Com B\u0026rdquo;, and the R-spine of PKA (\u0026ldquo;Com C\u0026rdquo;) matches the gray cluster in p38\u0026alpha; here\u003csup\u003e20\u003c/sup\u003e. In the C-lobe of PKA, \u0026ldquo;Com D\u0026rdquo; is a cluster congruent to the activation loop in p38\u0026alpha; (purple), \u0026ldquo;Com E\u0026rdquo;, the C-spine cluster in PKA, matches the pink cluster in p38\u0026alpha;, and \u0026ldquo;Com F\u0026rdquo; in the PKA case is the substrate binding region, which bears at least a slight similarity to the lipid pocket in p38\u0026alpha; (green), which is part of the lipid binding domain only applicable for MAP kinases\u003csup\u003e20\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eMore interestingly, we interrogated betweenness-centrality, which property identifies allosteric-network hotspots on the basis of their participation extent within different communication pathways. These results are shown in \u003cstrong\u003eFig. 2b\u0026nbsp;\u003c/strong\u003eand more specifically in Table S2. Whereas naturally, a one-to-one comparison with other kinases is compromised due to the different primary structures and differences in the techniques applied, these pathways again seem qualitatively consistent with the experiment-based pathways connecting different regions of p38\u0026gamma;, where methyl mutations of several conserved sites were explored.\u003csup\u003e17\u003c/sup\u003e For example, the memory node L170 in p38\u0026gamma; that takes additional input from the DFG-motif, appears as a motional hotspot in our study as well. Likewise, the hinge cluster of Abl kinase, constituting the D400 of the DFG-motif, had similar dominance in the long-range communication, suggesting the aspartate of the kinase to be a significant player in dynamic allostery\u003csup\u003e30\u003c/sup\u003e. In case of PKA, the F185 of the DFG motif coordinated together was found to be part of the allosteric communication instead of the vicinal aspartate, evocating the hypothesis whether the DFG motif itself is a necessary allosteric modulator\u003csup\u003e31\u003c/sup\u003e. D150, another node identified by our betweenness centrality study, belongs to the catalytic HRD-motif. In PKA, the homologous aspartate D166 was found to be the most crucial node that preorganizes the substrate\u0026rsquo;s phospho-acceptor site for efficient phosphotransfer\u003csup\u003e32\u003c/sup\u003e. (For Abl kinase and EGFR, the residues of this central catalytic machinery are the homologous D363 and D813, respectively.\u003csup\u003e33, 34, 35\u003c/sup\u003e) R49, another mediator that we identified as a residue with a high betweenness centrality, is a part of the extended allosteric pocket in p38\u0026alpha;. While it reinstates its allosteric role in p38a in our study, a homologous residue in similar kinases that pinpoints its governance in their allosteric mechanism has not been found\u003csup\u003e36\u003c/sup\u003e. Most interestingly, however, W207, identified here as another key residue of the network, is positioned near the substrate binding groove. In PKA, the homologous W222 has been found to be an allosteric hub in the \u0026alpha;F/FC region and a structural keystone anchoring the C-lobe.\u003csup\u003e37, 38\u003c/sup\u003e (Abl and EGFR kinases have similar hydrophobes but there is no exact analogue to pinpoint in this region\u003csup\u003e39\u003c/sup\u003e.) In p38a, this residue plays a particular role as it directly faces the lipid pocket, for which the possibility of designing allosteric effectors has been speculated.\u003csup\u003e7\u003c/sup\u003e Its high betweenness centrality hence directly rationalizes a participation of the lipid pocket as part of the dynamic network.\u003c/p\u003e\n\u003cp\u003eBeyond the overall inter-lobe correlations (shown in \u003cstrong\u003eFig. 2c\u0026nbsp;\u003c/strong\u003eand Table S3), which confirm both short-range connectivities dictated by the three-dimensional fold of the protein (e.g., R70/F169 ) and longer-range cross-talk such as between residues L104 and L167 or N102 and K165, we specifically examined the communication pathways linking the allosteric pocket, the activation loop, and the lipid pocket. For this reason, we picked one representative residue of the network for each of these sites/clusters and computed the cumulative pairwise correlation (the sum of degrees of correlation between any two neighbors) within the shortest path of communication. For the allosteric site, represented by residue K66, towards the activation loop (Y188), the information flux traverses the D168 and G170 of the DFG motif, similar to p38\u0026gamma; (\u003cstrong\u003eFig. 3b\u003c/strong\u003e, compare Table S5)\u003csup\u003e17, 18, 19\u003c/sup\u003e. Accordingly, the shortest path between the activation loop and the lipid pocket, with residues such as K233 and I235, is shown in \u003cstrong\u003eFig. 3c\u003c/strong\u003e. Pathways between the two pockets (between K66 in the allosteric pocket and K233, and I235 in the lipid pocket) were also calculated without specific consideration of the activation loop. This path involves the DFG motif, but bypasses the activation loop to channelise the communication to the lipid binding domain (\u003cstrong\u003eFig. 3e\u003c/strong\u003e). Importantly, all of the above analyses identify a network of allosteric communication that interconnects not only the N- and the C-lobe generally but also involves the lipid-binding-domain and specifically the lipid pocket. These in silico results rationalizes the existence of functional communication pathways between the active site and the lipid-binding domain, supporting potential avenues for targeting MAP kinases through allosteric modulation.\u003c/p\u003e\n\u003cp\u003eTo assess whether those networks are inherently apparent already from crystallography, the nodes identified were further examined in terms of B-factors derived from the crystal structures\u003csup\u003e29\u003c/sup\u003e. Crystallographic B-factors provide an experimental proxy for local rigidity and structural order. Fig. S5 shows a depiction of B-factors. However, the B-factors of the hotspots of the dynamic network are mostly in the intermediate range (see a list of B-factors for all major nodes of the network in Table S4), denoting that these residues are not particularly disordered in the crystal. (For example, residues R49, D150, and D168 have values of 0.509, 0.342, and 0.497, respectively. W207 shows an even lower value of 0.056, making it a very well-defined residue.) Even though a flexible character/flexibility of specific residues at physiological temperatures can sometimes be forecasted from the cryogenic crystallographic viewpoint, it seems that the allosteric hotspots are not recognizable through a specific degree of disorder. For further elucidation of changes to the networks, we hence turned to room temperature dynamics data, assessed by NMR spectroscopy (see below).\u003c/p\u003e\n\u003cp\u003eTo validate the existence and characteristics of the dynamic allosteric network identified by dynamical network analysis, we pursued a two-level approach involving additional in silico analyses and experimental validation. In both cases, the network was probed by \u0026ldquo;mutating\u0026rdquo; residues identified as key nodes based on high betweenness centrality. For this downstream analysis, in particular correlation analyses, ligand (sorafenib) binding studies, solution NMR chemical-shift perturbations and NMR relaxation, as well as biochemical assays (see below), four (R49, D150, D168, and W207) were selected and replaced by alanine. W207 is of particular interest as it forms part of the lipid pocket in the C-lobe. MD data (five 1 \u0026micro;s trajectories for each of the four mutants) were obtained using similar strategies as for the wild type (Fig. S6) and re-analyzed with the mentioned network framework. These analyses confirmed both, differences in betweenness-derived pathways and altered community structure (Fig. S7) compared to the wild type. The effects of mutation on long-range communication were also quantified by evaluating inter-lobe (N\u0026ndash;C) correlations and correlations along the wild-type-defined shortest communication paths (Tables S2, S3, and S5). Total correlation maps for all ensembles are provided in Fig. S8 and S9. Whereas the tertiary structure and hence many correlations are necessarily maintained with respect to \u003cstrong\u003eFig. 2c\u003c/strong\u003e, changes with respect to long-range correlations can be witnessed in each of the cases. (Difference maps (mutant \u0026ndash; wild type) are shown in Fig. S10.) To focus on pathway-level effects, we again selected one central residue per structural element (K66 for the allosteric pocket, Y188 for the activation loop, and I235 for the lipid pocket as before) and computed the shortest paths between them also for the mutants. The wild-type path served as the reference, and the cumulative amounts of pairwise correlations between any two consecutive residues along this path were evaluated for wild type and each mutant. Uncertainty was estimated from five replicas, each partitioned into five windows (Table S6). Across all paths (allosteric-to-activation loop, \u003cstrong\u003eFig. 3b\u003c/strong\u003e; lipid-pocket-to-activation loop, \u003cstrong\u003eFig. 3d\u003c/strong\u003e; and allosteric-to-lipid pocket paths, \u003cstrong\u003eFig. 3f\u003c/strong\u003e), mutant correlations were consistently lower than in the wild type protein. Interestingly, W207A, despite its residence in the C-lobe, has a very strong effect on the allosteric site-to-activation loop path (~20 %, \u003cstrong\u003eFig. 3b\u003c/strong\u003e). Conversely, for the lipid-pocket-to-activation loop, D168A, the DFG-mutant lying in the hinge of the N-lobe has a strong impact. (~10 %, \u003cstrong\u003eFig. 3d\u003c/strong\u003e); both residues lie distal to the respective pathways, consistent with a dense allosteric network. (\u003cstrong\u003eFig. 3d\u003c/strong\u003e). Intriguingly, with respect to the direct pathway between the allosteric pocket and the lipid pocket, all the nodes identified by the network mutated out (R49A with ~10\u0026nbsp;%, D150A, D170A, both with ~20\u0026nbsp;%, and W207A, ~25\u0026nbsp;%) show a substantial, statistically significant drop (\u003cstrong\u003eFig. 3f\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn addition to the apo proteins, we also examined how the presence of an active/allosteric-site binder would influence the network. For this purpose, we employed the in-silico interrogation of wild type and mutants outlined above, however, starting from the sorafenib-bound X-ray structure 3HEG. Fig. S11 and Table S8\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eshow the embedding of this ligand as interrogated using Discovery Studio\u003csup\u003e40\u003c/sup\u003e. MD simulations again afforded trajectories of 5 replicas each (Fig. S12). At first, we assessed motional correlations between the protein and the drug molecules. Fig. S13 shows these correlations, where for simplicity the ligand sites were coarse-grained into a Western and an Eastern part analysed individually. Interestingly, we observed that the total amount of cumulated correlation drops slightly upon mutation, most prominently for mutant R49A (Fig. S13). More importantly, however, we then pursued the above examination of interdomain communication within the ligand-bound protein. For the wild-type complex, shortest communication paths connecting the allosteric pocket, activation loop, and lipid pocket were identified using the same representative residues as in the apo analyses. Structural representations (Fig. S14A, C, E) reveal that inhibitor binding now reroutes the preferred communication pathways linking the allosteric pocket, the activation loop, and the lipid pocket with each other. For each complex, we quantified the cumulative pairwise correlations along these paths, using the ligand-bound WT as the reference. Now, taking the new, ligand-bound routes of the wild type protein as the reference and mutating the above-mentioned network hospots, no significant reduction in correlation is observed anymore (Fig. S14B, D, F). This indicates that ligand binding rigidifies the protein and largely suppresses its intrinsic dynamic network. Consistently, no systematic gains or losses in cumulative correlations are detected along the ligand-bound paths.\u003c/p\u003e\n\u003cp\u003eTogether, these results demonstrate that (i) mutations at network sites attenuate communication in a pathway- and residue-dependent manner, (ii) sorafenib reshapes the communication routes relative to the apo protein, and (iii) ligand binding to the active site globally rigidifies the protein, effectively quenching dynamic network propagation. Assessing a single shortest communication path neglects the fact that allosteric signaling typically arises from the collective contribution of multiple pathways. Moreover, the in silico identification of long-range communication paths spanning many network edges is intrinsically less sensitive than the analysis of shorter network fragments, such as those shown in \u003cstrong\u003eFig. 2b\u003c/strong\u003e. Even with the extensive sampling afforded by five independent microsecond-long MD replicas, these pathway-level correlations retain a certain statistical error, underscoring the need for experimental validation based on vastly larger ensembles of molecules. To experimentally (with a much bigger ensemble) probe the functional consequences of hijacking the dynamic network, we therefore first assessed the enzymatic integrity of the individual mutants relative to the wild-type protein using Homogeneous Time-Resolved Fluorescence (HTRF) assays\u003csup\u003e41\u003c/sup\u003e (\u003cstrong\u003eFig. 3g\u003c/strong\u003e, see details in the Methods).\u003c/p\u003e\n\u003cp\u003eHTRF is a robust, high-throughput, low-background FRET-based technique that combines the sensitivity of time-resolved fluorescence with a homogeneous assay format. In this study, HTRF was employed to quantitatively compare protein substrate binding efficacy (using GST-tagged activating transcription factor 2 as a substrate) as well as the integrity of binding of ATP between wild-type and mutant constructs\u003csup\u003e41\u003c/sup\u003e. Adding to the wild-type protein, whose expression and purification has been described in the literature\u003csup\u003e42\u003c/sup\u003e, all of the above mentioned mutants were generated via site-directed mutagenesis and the protein expressed recombinantly in \u003cem\u003eE. coli\u003c/em\u003e and purified according to the existing protocols (see the Methods and Fig. S15, for biochemistry details). The dependence of reaction speed on either ATP or substrate concentration serves as a quantitative proxy for binding of either interaction partner (see details in the Methods). Kinetic characterisation was pursued via Michaelis-Menten fits (\u003cstrong\u003eFig. 3h\u003c/strong\u003e and \u003cstrong\u003eFig. 3i\u003c/strong\u003e), confirming highest binding competencies, with the lowest Michaelis-Menten constants (\u003cem\u003eK\u003c/em\u003e\u003csub\u003eM\u003c/sub\u003e) for the wild type enzyme (Table S7). Strongly impaired affinity occurs in mutants D168A, R49A, with a twofold decrease in ATP binding each. Similarly, for the mutant D150A, the binding to the substrate is diminished, and ATP binding competency is slightly lower. Also for W207A, even though the mutation is deep in the C-lobe, the ATP-driven regulation and catalysis appear to be lost. The most severe phenotypes were observed in \u003cstrong\u003eD150\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eW207\u003c/strong\u003e, even though neither of these sites is part of the actual enzymatic process, consistent with the possibility of disrupted enzymatic potency due to allosteric modulation from distant sites of mutation. Even though the consistent decrease for both interaction partners even for mutation sites far away from binding interfaces and in particular for W207 as a lipid pocket residue is noteworthy, more specific motional analyses were sought to corroborate the findings.\u003c/p\u003e\n\u003cp\u003eIf protein motions were purely local and independent, a point mutation would be expected to affect primarily the substituted residue and its immediate environment. By contrast, NMR studies have shown that perturbations within dynamic allosteric networks can induce long-range changes in both, site-specific chemical shifts and relaxation properties across multiple timescales\u003csup\u003e17\u003c/sup\u003e \u003csup\u003e43\u003c/sup\u003e. To assess such putative long-range changes in protein dynamics, we used solution-state NMR on \u003csup\u003e13\u003c/sup\u003eC/\u003csup\u003e15\u003c/sup\u003eN/\u003csup\u003e2\u003c/sup\u003eH labelled protein samples expressed in triple labelled media and purified using successive rounds of Ni-affinity, anion exchange, and size exclusion chromatography according to published procedures\u003csup\u003e42\u003c/sup\u003e. In spite of multiple attempts and in contrast to all other proteins mentioned here, the plasmid of mutant W207A could not be successfully expressed as a stable triple-labeled construct in deuterated minimal media, precluding its structural and dynamic interrogation by NMR. Peak assignments were achieved via 3D HNCA spectra for both wild-type and mutant constructs, complemented by TROSY-HSQC used as fingerprint spectra (Fig. S16), upon which existing assignments (BMRB entry: 17471) could be transferred in a stepwise manner. To complement the missing fourth mutant, we newly introduced V38A, in this case, however, specifically as an experimental negative control. V38 lies in a folded element of the tertiary structure, but outside the predicted dynamic network, allowing us to compare any \u0026ldquo;regular\u0026rdquo; alterations induced locally to global changes entailed by a tight dynamic network.\u003c/p\u003e\n\u003cp\u003eWe first employed chemical-shift perturbation (CSP) analysis to probe changes in the average chemical environment of backbone amides across the protein. In the absence of allosteric communication, the chemical shifts of very distant residues would not be expected to change even in the case of slight structural changes around the mutation site. Instead, the analysis revealed widespread deviations across the backbone for all functional mutants involved in the allosteric coupling. Notably, D168A, the \u0026quot;DFG-loop mutant\u0026quot;, exhibited the most dramatic global extent of CSPs across the structure (\u003cstrong\u003eFig. 4a, c, and d\u003c/strong\u003e), including long-range effects such as the complete disappearance of the W207 peak in the C-lobe. As mentioned above, apart from its identification as a spatially distant key residue involved in the interdomain communication, W207 also represents a marker for the lipid pocket. This corroborates both, D168A\u0026rsquo;s sensitive influence on allosteric interactions and domain coupling as well as the participation of W207 in the allosteric network. The distributed nature of dynamic perturbations throughout the protein can be recognized from \u003cstrong\u003eFig. 4b\u003c/strong\u003e, exemplarily depicting chemical-shift perturbations for this variant compared to the wildtype. (CSPs and spectral overlays for the other mutants are shown in Fig. S17.) In contrast to the mutants focused on hotspots found for the dynamic network, the negative control V38A showed indeed very limited chemical-shift changes (see \u003cstrong\u003eFig. 4d\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eFig. S17). The stark differences in chemical-shift perturbations between mutations within the dynamic network and the negative control is also seen from the violin plots shown in \u003cstrong\u003eFig. 4c\u003c/strong\u003e, supporting V38A\u0026rsquo;s role as a mutation benign to the conformational ensemble and underscoring the global conformational remodeling induced by the actual network-disruptive mutations on the contrary. These findings provide experimental support for mutations within hotspots identified by dynamical network analysis to indeed induce substantial allosteric rewiring of the kinase conformational ensemble.\u003c/p\u003e\n\u003cp\u003eTo determine whether the observed chemical-shift perturbations really reflect changes in global protein dynamics, we employed solution-state \u003csup\u003e15\u003c/sup\u003eN relaxation, more specifically, [\u003csup\u003e15\u003c/sup\u003eN, \u003csup\u003e1\u003c/sup\u003eH] heteronuclear NOE (hetNOE) and \u003csup\u003e15\u003c/sup\u003eN transverse relaxation rates (\u003cem\u003eR\u003csub\u003e2\u003c/sub\u003e\u003c/em\u003e). These partly complementary relaxation parameters probe site-specific motion on the ps-ns timescale, with \u003cem\u003eR\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e being additional influenced by motion on the \u0026micro;s timescale. Among all mutants tested, D150A exhibited the most pronounced deviations from wild-type behavior across both, hetNOE and \u003cem\u003eR₂\u003c/em\u003e datasets (\u003cstrong\u003eFigs. 5\u003c/strong\u003e and S18). The further reduction in hetNOE values compared to the wildtype at several positions throughout the protein suggests a further increase in fast-timescale local flexibility. In parallel, increased \u003cem\u003eR\u003c/em\u003e₂ values, especially in the C-lobe (residues around 200 \u0026ndash; 250), indicate an additional increase of conformational-exchange contributions, stemming from increased \u0026micro;s timescale fluctuations and suggesting a widely altered energy landscape introduced by the mutation. These observations also align with significant exchange broadening, which is usually incurred by enhanced \u0026micro;s\u0026ndash;ms conformational dynamics. The differences to the wildtype are also visualized through structural mapping of per-residue deviations (\u003cstrong\u003eFig. 5a-b\u003c/strong\u003e), witnessing dynamic perturbations extending well beyond the respective mutation sites. Intriguingly, D168A and R49A also show vivid alterations in the C-lobe regarding the \u003cem\u003eR\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e rates, and more slightly for hetNOE values, again supporting that these sites represent key nodes in long-range allosteric coupling. Also the extent of dynamic reorganization in D150A appears to be globally distributed. In disparity, the control variant V38A, with its mutation site outside the predicted dynamic network, exhibited again only minimal, local changes. Both its hetNOE and \u003cem\u003eR₂\u003c/em\u003e profiles remain largely aligned with the wild-type except for subtle shifts near the site of mutation (Fig. S18), confirming its neutrality regarding the kinase\u0026rsquo;s conformational ensemble. Boxplot comparisons across the variants (\u003cstrong\u003eFig. 5c-d\u003c/strong\u003e) support a general enhancement in both, fast-timescale flexibility (decreased hetNOE) and a concurrent rise in \u0026micro;s timescale dynamics (higher \u003cem\u003eR\u003c/em\u003e₂ values), suggesting that mutations in dynamic network positions generally increase protein plasticity and conformational exchange. These shifts correlate with a loosening of the dynamic network upon mutation, upon which motion of the individual residues becomes less coherent and less restricted overall.\u003c/p\u003e\n\u003cp\u003eTo probe conformational-exchange processes on the microsecond-to-millisecond timescale, we finally performed \u003csup\u003e15\u003c/sup\u003eN CPMG relaxation dispersion (RD) measurements on wildtype protein and the mutants (Fig. S19). Data quality was insufficient to perform a more comprehensive analysis, but exchange contributions \u003cem\u003eR\u003c/em\u003e\u003csub\u003eex\u003c/sub\u003e to the effective transverse relaxation values could be obtained with reasonable accuracy, offering residue-specific identification of transient conformational substates. In RD experiments, generally, the exchange contribution \u003cem\u003eR\u003csub\u003eex\u003c/sub\u003e\u003c/em\u003e is derived from the difference in \u003cem\u003eR\u003csub\u003e2eff\u003c/sub\u003e\u003c/em\u003e measured at varying CPMG refocusing efficiency. High\u003cem\u003e\u0026nbsp;R\u003csub\u003eex\u003c/sub\u003e\u003c/em\u003e values are indicative of residues undergoing chemical exchange (allosteric transitions, functional loop dynamics, or domain breathing motions) between conformers in particular on the fast ms timescale, being stronger both, for elevated excited-state populations and increased chemical-shift differences for those transient states. Globally, wild-type samples exhibited the largest number of residues with significant \u003cem\u003eR\u003csub\u003eex\u003c/sub\u003e\u003c/em\u003e contributions throughout the N-lobe and the C-lobe, indicating widespread conformational plasticity (Fig. S19A-B). As expected, a highly similar pattern as for the wild type is observed for V38A, both with respect to the residue level (Fig. S19A) and regarding the distribution of \u003cem\u003eR\u003c/em\u003e\u003csub\u003eex\u003c/sub\u003e contributions depicted in the form of a histogram (Fig. S19B), consistent with its position outside the dynamic network and confirming its dynamically silent nature in terms of allosteric communication. The mutant construct R49A shows a slight loss of ms timescale exchange in the N-lobe and a slight increase in the C-lobe, confirming the above reorganization of motional behavior (Fig. S19C). For mutants D150A and D168A, by contrast, a slight decrease of \u003cem\u003eR\u003c/em\u003e\u003csub\u003eex\u003c/sub\u003e in the C-lobe and a slight increase in the N-lobe was observed (Fig. S18C). All of these changes are, however, rather sporadic and any trends are difficult to discern. The apparent lack of strong differences between the samples results on the one hand from a rather noticeable error associated with the individual \u003cem\u003eR\u003c/em\u003e\u003csub\u003e2eff\u003c/sub\u003e rates. Fig. S18D exemplifies this for A172, a residue lying in between the DFG motif and the activation loop, via comparison of dispersion curves for the different samples. A number of other dispersion curves are shown in Fig. S19. However, the higher overall similarities across the \u003cem\u003eR\u003c/em\u003e\u003csub\u003eex\u003c/sub\u003e histograms in Fig. S18B suggests that the differences induced by the perturbations to the dynamic networks, as consistently seen upon CSPs, hetNOEs, and \u003cem\u003eR\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e analyses, are rather limited to motions faster than the ms regime.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe results presented here demonstrate that p38\u0026alpha; kinase harbors a dynamic allosteric network that connects the N- and C-lobes and explicitly incorporates the lipid-binding domain and its associated lipid pocket. Mutations at residues identified as central nodes within this network induce substantial perturbations in both local and global dynamics, as evidenced by a consistent set of computational and experimental observations, including molecular dynamics simulations, dynamical network analysis, pathway-level correlation changes, chemical-shift perturbations, heteronuclear NOE measurements, and transverse relaxation rates. While ligand binding partially modulates these effects, the dominant impact of network disruption is observed in the apo ensembles. Collectively, these findings provide experimental validation for the simulation-derived prediction that the p38\u0026alpha; scaffold supports long-range dynamic communication between distinct functional regions of the protein. Whereas the firm connectivity between dynamic-network residues tethers the plastic kinase architecture tightly together, thereby conducting concerted dynamics across long distances, mutations at key network sites abrogate long-range communication, entailing increased fast-timescale flexibility and conformational exchange. This conclusion can rationalize prior observations in a more biological context in a residue-specific manner. For example, local perturbations, such as activation-loop phosphorylation or binding of upstream regulators, can induce dynamical changes at distal sites within the kinase, albeit on slower timescales than those primarily interrogated here\u003csup\u003e15\u003c/sup\u003e. With the revelation of a widespread dynamic network prominently involving the activation loop and encompassing in particular the active site, the top part of the C-lobe, and the lipidic pocket, the various prior observations hinting to dynamic allostery across the kinase architecture and regulation come to no surprise.\u003c/p\u003e\n\u003cp\u003eSeveral residues identified as key network nodes coincide with sites previously implicated in pathological or regulatory phenotypes, despite not being directly involved in catalysis. D150, located in proximity to the catalytic site\u003csup\u003e44\u003c/sup\u003e of the protein, forms an intramolecular H-bond with T185 that is required for TAB1-induced autoactivation\u003csup\u003e45\u003c/sup\u003e. R49K/A is used as a convenient tool mutation for studying PRMT1-mediated control of p38\u0026alpha;\u003csup\u003e46\u003c/sup\u003e. Also D168 mutation is known to render the enzyme dysfunctional and has been used as a valuable mechanistic probe\u003csup\u003e47, 48, 49\u003c/sup\u003e. Its prominent functional role, being key for the in/out conformational switch of the DFG motif, required for exchanging ATP/ADP at the active site, clearly exceeds the role as a mere dynamics-transducing element of the network. It is still important to realize that this site can also be ascribed a prominent role in transducing information across the protein scaffold, which would mean an impact on the motional properties and allosteric sensing associated with the conformational changes at the DFG motif. Importantly, our data establish the lipid-binding domain as an integral component of the p38\u0026alpha; dynamic network, which provides further prospects for pharmacological avenues based on lipid-pocket-binding allosteric modulators.\u003csup\u003e7\u003c/sup\u003e Taken together, these observations demonstrate how wide-spread dynamic networks, rather than the local mobility associated with the individual structural elements, underly the integration of regulatory signals in p38\u0026alpha; and likely in other kinases as well.\u003c/p\u003e\n\u003cp\u003eBeyond the shear identification/mapping of communication pathways between the p38a N- and C-lobes, in particular including the so-far largely unexploited lipid pocket, in the presence and absence of the inhibitor, the above results may also offer a conceptual framework for the in-silico identification of possible allosteric sites in a kinase drug discovery context. The knowledge about, in particular, the existence and whereabouts of surface sites associated to dynamic networks could facilitate the design of next-generation allosteric modulators that do not target the known pockets, such as the ATP binding site or the \u0026ldquo;allosteric pocket\u0026rdquo;. It is so far unclear whether the lipid pocket will ever be addressed pharmacologically. The confirmation that a dynamic network extends to this site and in particular the identification of W207 at the rim of the lipid pocket suggests the site may serve as a useful starting point for new allosteric p38\u0026alpha; inhibitors. With first-generation, low-affinity binders (lipid pocket ligands), lipid pocket occupation (at the low populations obtained) has not been found to bear a strong impact on kinase activity.\u003csup\u003e7\u003c/sup\u003e However, given that for a desired systemic effect even a weak molecular impact can suffice, the above data encourage further work, involving bulkier or sterically optimized ligands, which might result in new perspectives to overcome the off-site effects usually associated with targeting kinases. More generally, awareness and spatial mapping of dynamic networks in enzymes may guide the identification of regulatory sites amenable to therapeutic intervention or rational engineering. It remains to be seen whether dynamic allostery will prove valuable not only for pharmacological inhibition but also for fine-tuning of enzymatic activity in biotechnological applications through targeted allosteric mutations\u003csup\u003e50, 51, 52\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eIn summary, we identify and map a dynamic allosteric network within p38\u0026alpha; MAP kinase that spans both lobes, prominently involves inter-lobe segments of the architecture and the activation loop, and explicitly includes the lipid-binding domain and its associated lipid pocket. By integrating computational and experimental structural biology approaches, we reveal specific residues that act as key information relays within a coordinated motional framework, transmitting dynamics across spatially distant regions of the protein. Mutations at these hotspots largely corrupt the integrity of the network, thereby significantly reducing inter-lobe communication and inducing global changes in residue-specific dynamics, underscoring their central role in long-range allostery. Together, these findings provide new insights into the dynamic basis of kinase regulation and pathology. With the lipid pocket surface, in particular residue W207, found to form an integral part of the dynamic network, the study also highlights opportunities for identifying and targeting allosteric sites in future pharmacological strategies.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cem\u003eDynamical Network Analysis\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAll atomic molecular dynamics (MD) simulations were performed using NAMD 3.0\u003csup\u003e27\u003c/sup\u003e in conjunction with the QwikMD plugin\u003csup\u003e53\u003c/sup\u003e. To quantify ensemble-level differences associated with the reported ligand-dependent conformational changes, simulations were initiated from two crystallographic structures of p38\u0026alpha; in which the inhibitor ring flip was observed (PDB IDs: 3HEG\u003csup\u003e13\u003c/sup\u003e and 3GCS\u003csup\u003e14\u003c/sup\u003e). For apo p38\u0026alpha; (1WFC\u003csup\u003e29\u003c/sup\u003e), missing segments were rebuilt by comparative modeling. NOESY-derived restraints were applied during energy minimization (20,000 steps), slow heating to 300 K, and a ~1 \u0026mu;s restrained equilibration; restraints were then released and five independent 1 \u0026mu;s production replicas were generated. Dynamical network analysis identified betweenness-centrality hotspots, which were alanine-scanned using QwikMD; each mutant underwent the same minimize\u0026ndash;anneal\u0026ndash;equilibrate protocol, followed by five 1 \u0026mu;s replicas for correlation and network metrics. Sorafenib was docked into the allosteric pocket (Table S6), prepared identically, and simulated in five 1\u0026mu;s replicas. Correlations and communication pathways were compared across apo, mutant, and ligand-bound ensembles. A total of ~50 \u0026mu;s simulations are used for this study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAll simulations employed the CHARMM36\u003csup\u003e29\u003c/sup\u003e force field for proteins and ligands. Covalent bonds involving hydrogen atoms were constrained using SHAKE, and equations of motion were integrated using a 1 fs time step. Long-range electrostatics were treated using the particle-mesh Ewald method under periodic boundary conditions, with a real-space cutoff of 12 \u0026Aring; and a pair-list distance of 14 \u0026Aring;. Systems were equilibrated under NVT conditions for thermalization and NPT conditions for pressure coupling prior to production. Temperature was maintained at 300 K using Langevin dynamics, and pressure was controlled at 1 atm using the Langevin piston method under periodic boundary conditions.\u003c/p\u003e\n\u003cp\u003eThe topology and trajectory files from the MD simulations were incorporated into the Python notebooks of Dynetan\u003csup\u003e25\u003c/sup\u003e. The software uses the incorporation of MDAnalysis to analyze the trajectory simulation files\u003csup\u003e54\u003c/sup\u003e. Contact detection optimization of the code was already performed via Numba\u003csup\u003e55\u003c/sup\u003e and Cython\u003csup\u003e56\u003c/sup\u003e. Network statistics and the determination of optimal paths were carried out using the Floyd\u0026ndash;Warshall algorithm, provided by the NetworkX package. For Floyd\u0026ndash;Warshall calculations, the \u0026ldquo;distance\u0026rdquo; between nodes was defined as\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u0026nbsp;d = \u0026minus;log(r\u003csub\u003eMI\u003c/sub\u003e),\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003ewhich uses mutual information as a method for contact detection consistent with previous applications of this method which can be seen as:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003er\u003csub\u003eMI\u003c/sub\u003e\u003c/em\u003e\u003csub\u003e\u0026nbsp;\u003c/sub\u003e[\u003cem\u003ei,j\u003c/em\u003e]\u003csub\u003e\u0026nbsp;\u003c/sub\u003e =\u003csub\u003e\u0026nbsp;\u003c/sub\u003e(1- \u003cem\u003ee\u003c/em\u003e\u003csup\u003e-2/3\u003cem\u003eI\u003c/em\u003e[\u003cem\u003ei,j\u003c/em\u003e]\u003c/sup\u003e)\u003csup\u003e1/2\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003ewhere i and j are the position of the alpha carbon of each residue and \u003cem\u003eI\u003c/em\u003e[\u003cem\u003ei,j\u003c/em\u003e] are the mutual information between them which is computed by the density estimator described as:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eI\u003c/em\u003e[\u003cem\u003ei,j\u003c/em\u003e] = \u003cem\u003e\u0026psi;\u003c/em\u003e\u003cem\u003e(k)\u0026nbsp;\u003c/em\u003e-\u003cem\u003e1/k\u003c/em\u003e - \u0026lsaquo;\u003cem\u003e\u0026psi;\u003c/em\u003e(\u003cem\u003en\u003csub\u003ei\u003c/sub\u003e\u003c/em\u003e) + \u003cem\u003e\u0026psi;\u003c/em\u003e(\u003cem\u003en\u003csub\u003ej\u003c/sub\u003e\u003c/em\u003e)\u0026rsaquo; + \u003cem\u003e\u0026psi;\u003c/em\u003e(\u003cem\u003eN\u003c/em\u003e)\u003c/p\u003e\n\u003cp\u003ewhere here N is the total number of simulation frames, \u0026psi; (x) is the digamma function, n\u003csub\u003ei\u003c/sub\u003e is the number of frames in which residue i is close to the one in the reference, and stands for the average of the trajectory, using a neighboring parameter k of 6.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEach amino acid residue is regarded as a node. The network is structured by these nodes and nodes lie within a cutoff distance (4.5 \u0026Aring;) for at least 75% of an MD trajectory. The residues which lie farther than the cutoff value are excluded which mainly includes some solvent and ion molecules. The nodes are connected via links known as \u0026ldquo;edges\u0026rdquo;.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBetweenness centrality was computed as :\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"262\" height=\"40\"\u003e\u003c/p\u003e\n\u003cp\u003ewhere \u0026sigma; (s, t) is the number of shortest paths between nodes s and t, V is the ensemble of graph nodes, and C is a normalization factor.\u003c/p\u003e\n\u003cp\u003eThe simulations were analyzed using in-house Python and TCL scripts along with VMD\u003csup\u003e57\u003c/sup\u003e. Contact maps for residue\u0026ndash;residue interactions were generated with VMD and PyContact. For the statistical testing, we normalized all measurements by the cumulative of the WT (or LWT) row. For each variant we report the per-variant mean and plot error bars as the sample standard deviation across replicates. Group comparisons to WT used a non-parametric permutation bootstrap: we pooled the two groups, permuted labels, recomputed the mean difference, and repeated this 10,000 times (two-sided). The p-value is the proportion of permuted differences whose absolute value exceeded the observed absolute difference. No parametric assumptions (normality or equal variances) were required\u0026nbsp;and plotted using python notebooks.\u0026nbsp;All the structural figures for simulations are rendered through VMD. The figures showing residue-wise properties are extracted through Chimera attributes.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSample Preparation\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe plasmid map of the kinase was elucidated by nanopore sequencing to entail the specific design of mutagenesis primers. The primers for the mutants were designed such that the mutation residue lies at the center for the forward and reverse primers. \u003cem\u003ePhusion\u0026trade; High-Fidelity DNA Polymerase\u003c/em\u003e, a fusion polymerase containing a Mastermix of nucleotides, enzymes, and buffers were used to carry out the polymerase chain reaction. The PCR product was digested using the DpnI enzyme at 37 \u0026deg;C for 1 hr and was immediately transformed overnight at 37 \u0026deg;C to \u003cem\u003eE.coli XL 10 gold\u003c/em\u003e cells. The transformed colonies were grown in an overnight culture at 37 \u0026deg;C and the plasmid was extracted using a PEG gold miniprep kit and sent to Eurofins for Sanger Sequencing. The wildtype as well as the mutants was expressed into BL21(DE3) \u003cem\u003eE. coli\u003c/em\u003e cells using \u003csup\u003e13\u003c/sup\u003eC and \u003csup\u003e15\u003c/sup\u003eN labelling for assignment and relaxation experiments. D\u003csub\u003e2\u003c/sub\u003eO adaptation was carried out and the cultures were grown in 1\u0026nbsp;L at 37 \u0026deg;C to reach the OD\u003csub\u003e600\u003c/sub\u003e of 0.6-0.7. They were then cooled for 30 min to room temperature, and induced with 1 mM IPTG overnight (\u0026sim;20 h) at 18 \u0026deg;C while shaking at 180 rpm. For the mutant W207A, expression in deuterated minimal media could not be achieved. The cells were lysed using a 50 mM Tris Buffer, benzonase, and lysozyme in the presence of a protease inhibitor, centrifuged, and the supernatant was exposed to affinity chromatography (Binding Buffer: 50 mM Tris, 500 mM NaCl, 25 mM Imidazole and 5 % Glycerol; Elution Buffer: 50 mM Tris, 500 mM NaCl, 500 mM Imidazole and 5 % Glycerol) using a nickel column. The eluted fractions were dialyzed to a pH of 7 and were subjected to an overnight cleavage in the presence of thrombin. Anion exchange (Binding Buffer: 25 mM Hepes and 5 % Glycerol; Elution Buffer: 25 mM Hepes, 1 M NaCL, 5 % Glycerol) and Size Exclusion Chromatography (SEC Buffer: 20 mM Hepes, 50 mM NaCl, 100 mg/L methionine and \u0026nbsp;5 % Glycerol) was then used to obtain pure protein.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eHTRF Analysis\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eWild-type and mutant p38\u0026alpha; constructs (unlabelled) were activated with constitutively active MKK6\u003csup\u003eS207E/T211E\u003c/sup\u003e (Thermo Scientific, Lot 877061F) in activation buffer (50 mM Tris, 10 mM MgCl₂, 1 mM ATP, 1 mM DTT, 0.001% Tween-20, pH 7.4) at 37 \u0026deg;C for 90 min with shaking at 400 rpm. Reactions were dialyzed overnight at 4 \u0026deg;C into storage buffer (20 mM HEPES, 50 mM NaCl, 5% glycerol, pH 7.1), concentrated to ~0.2 mg/mL, and stored at \u0026minus;80 \u0026deg;C. ATP \u003cem\u003eK\u003csub\u003eM\u003c/sub\u003e\u003c/em\u003e was determined using the HTRF\u0026reg; KinEASE\u0026trade; kit (Cisbio) following the manufacturer\u0026rsquo;s instructions. Activated p38\u0026alpha; was added per well at 0.04 ng (WT), 3.5 ng (mutant/unlabelled), or 20 ng (mutant/labelled) and incubated for 10/10/20 min, respectively, with 1 \u0026micro;M GST-ATF2 substrate across 0.4\u0026ndash;900 \u0026micro;M ATP in reaction buffer (50 mM HEPES, 0.1 mM Na₃VO₄, 0.02% NaN₃, 0.01% w/v BSA, 10 mM MgCl₂, 1 mM MnCl₂, 1 mM DTT, 0.01% Triton X-100, pH 7.0) in 384-well black flat-bottom plates (Greiner Bio-One). Reactions were stopped with detection solution (50 mM HEPES, 0.1% w/v BSA, 800 mM KF, 20 mM EDTA, 0.666 nM anti-phospho-ATF2-Eu(K) antibody, 100 nM anti-GST-d2 antibody, pH 7.0) and incubated 60 min at room temperature. Time-resolved fluorescence was read on an EnVision 2104 (PerkinElmer) at 620 nm and 665 nm, 60 \u0026micro;s after excitation at 317 nm. The 665/620 signal ratio was plotted versus ATP concentration and fitted to the Michaelis\u0026ndash;Menten equation in Origin (OriginLab).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSolution NMR\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe protein from the SEC Buffer was exchanged to \u0026ldquo;NMR Buffer\u0026rdquo; containing 50 mM of HEPES and 150 mM NaCl (pH ~6.8), and the sample was concentrated to approximately 450 \u0026micro;M. NMR experiments were recorded on an 800 MHz Bruker NEO spectrometer. 2D \u003csup\u003e15\u003c/sup\u003eN-\u003csup\u003e1\u003c/sup\u003eH TROSY HSQC and 3D TROSY HNCA were recorded to gain the assignments for the mutants via assignment transfer from the BMRB (entry: 17471). CCPN version 3.1 was used for assigning the mutant peaks\u003csup\u003e58\u003c/sup\u003e. \u003csup\u003e15\u003c/sup\u003eN-\u003csup\u003e1\u003c/sup\u003eH TROSY hetNOE was performed to assess fast-timescale dynamics and was analysed using a CCPN macro for relaxation. \u003csup\u003e15\u003c/sup\u003eN TROSY-CPMG experiments were acquired in an interleaved manner as follows for the CPMG frequencies: 0, 2000, 25, 1500, 1000, 100, 750, 200, 1250, 500, 200 Hz with a CPMG delay of a total of 0.022 s and a recycling delay of 1.5 seconds. The experiments were analyzed with NESSY software\u003csup\u003e59\u003c/sup\u003e. Graphs and analysis were produced using in house python scripts.\u003c/p\u003e\n\u003cp\u003eAll the raw data has been deposited in the public repository of Technical University Dortmund (https://data.tu-dortmund.de/previewurl.xhtml?token=ff748a60-a45c-4c14-8f09-b7e397880aca) and can be freely accessed\u003csup\u003e60\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFunded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany\u0026apos;s Excellence Strategy - EXC 2033 \u0026ndash; 390677874 \u0026ndash; RESOLV, and EXC-114 \u0026ndash; 24286268 \u0026ndash; CiPS-M. Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) \u0026ndash; 27112786, 325871075 and the Emmy Noether program. Funded/co-funded by the European Union (ERC, 101082494 bypassNMR). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. The authors gratefully acknowledge the computing time provided on the Delta-AI clusters, the clusters of the Bernardi group in Auburn University and the Linux HPC cluster at Technical University Dortmund (LiDO3), which is partially funded in the course of the Large-Scale Equipment Initiative by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) as project 271512359. R.C.B. was supported by the National Science Foundation under Grant MCB-2143787. Computational resources were provided in part through NSF ACCESS (Allocation BIO250251). This research made use of the DeltaAI advanced computing and data resource, supported by the National Science Foundation (Award OAC-2320345) and the State of Illinois, as well as the Delta advanced computing and data resource, supported by the National Science Foundation (Award OAC-2005572) and the State of Illinois. Delta and DeltaAI are joint efforts of the University of Illinois Urbana-Champaign and the National Center for Supercomputing Applications.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003ePerry JJ, Harris RM, Moiani D, Olson AJ, Tainer JA. p38alpha MAP kinase C-terminal domain binding pocket characterized by crystallographic and computational analyses. \u003cem\u003eJ. Mol. Biol.\u003c/em\u003e \u003cstrong\u003e391\u003c/strong\u003e, 1-11 (2009).\u003c/li\u003e\n\u003cli\u003eUng PM, Thompson AD, Chang L, Gestwicki JE, Carlson HA. Identification of key hinge residues important for nucleotide-dependent allostery in E. coli Hsp70/DnaK. \u003cem\u003ePLoS Comput. Biol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e1003279 (2013).\u003c/li\u003e\n\u003cli\u003eWang Z, Harkins PC, Ulevitch RJ, Han J, Cobb MH, Goldsmith EJ. The structure of mitogen-activated protein kinase p38 at 2.1-A resolution. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e94\u003c/strong\u003e, 2327-2332 (1997).\u003c/li\u003e\n\u003cli\u003eWrobleski ST, Doweyko AM. Structural comparison of p38 inhibitor-protein complexes: a review of recent p38 inhibitors having unique binding interactions. \u003cem\u003eCurr. Top. Med. Chem.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1005-1016 (2005).\u003c/li\u003e\n\u003cli\u003eGrossi V\u003cem\u003e, et al.\u003c/em\u003e Sorafenib inhibits p38\u0026alpha; activity in colorectal cancer cells and synergizes with the DFG-in inhibitor SB202190 to increase apoptotic response. \u003cem\u003eCancer Biol. Ther.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1471-1481 (2012).\u003c/li\u003e\n\u003cli\u003eDiskin R, Engelberg D, Livnah O. A novel lipid binding site formed by the MAP kinase insert in p38 alpha. \u003cem\u003eJ. Mol. Biol.\u003c/em\u003e \u003cstrong\u003e375\u003c/strong\u003e, 70-79 (2008).\u003c/li\u003e\n\u003cli\u003eB\u0026uuml;hrmann M, Hardick J, Weisner J, Quambusch L, Rauh D. Covalent Lipid Pocket Ligands Targeting p38\u0026alpha; MAPK Mutants. \u003cem\u003eAngew. Chem., Int. Ed.\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 13232-13236 (2017).\u003c/li\u003e\n\u003cli\u003eZhang XH\u003cem\u003e, et al.\u003c/em\u003e Targeting the non-ATP-binding pocket of the MAP kinase p38\u0026gamma; mediates a novel mechanism of cytotoxicity in cutaneous T-cell lymphoma (CTCL). \u003cem\u003eFEBS Lett.\u003c/em\u003e \u003cstrong\u003e595\u003c/strong\u003e, 2570-2592 (2021).\u003c/li\u003e\n\u003cli\u003eCooper A, Dryden DT. Allostery without conformational change. A plausible model. \u003cem\u003eEur. Biophys. J.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 103-109 (1984).\u003c/li\u003e\n\u003cli\u003eMonod J, Wyman J, Changeux JP. ON THE NATURE OF ALLOSTERIC TRANSITIONS: A PLAUSIBLE MODEL. \u003cem\u003eJ. Mol. Biol.\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 88-118 (1965).\u003c/li\u003e\n\u003cli\u003eKuriyan J, Eisenberg D. The origin of protein interactions and allostery in colocalization. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e450\u003c/strong\u003e, 983-990 (2007).\u003c/li\u003e\n\u003cli\u003eKoshland DE, Jr., N\u0026eacute;methy G, Filmer D. Comparison of experimental binding data and theoretical models in proteins containing subunits. \u003cem\u003eBiochemistry\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 365-385 (1966).\u003c/li\u003e\n\u003cli\u003eNamboodiri HV, Bukhtiyarova M, Ramcharan J, Karpusas M, Lee Y, Springman EB. Analysis of imatinib and sorafenib binding to p38alpha compared with c-Abl and b-Raf provides structural insights for understanding the selectivity of inhibitors targeting the DFG-out form of protein kinases. \u003cem\u003eBiochemistry\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 3611-3618 (2010).\u003c/li\u003e\n\u003cli\u003eSimard JR\u003cem\u003e, et al.\u003c/em\u003e Development of a fluorescent-tagged kinase assay system for the detection and characterization of allosteric kinase inhibitors. \u003cem\u003eJ. Am. Chem. Soc.\u003c/em\u003e \u003cstrong\u003e131\u003c/strong\u003e, 13286-13296 (2009).\u003c/li\u003e\n\u003cli\u003eKumar GS\u003cem\u003e, et al.\u003c/em\u003e Dynamic activation and regulation of the mitogen-activated protein kinase p38. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e115\u003c/strong\u003e, 4655-4660 (2018).\u003c/li\u003e\n\u003cli\u003eMedina G\u0026oacute;mez S, Homberg LT, B\u0026uuml;hrmann M, Rauh D, Linser R. Lipid Pocket Binders Impose Allosteric Changes of Protein Dynamics Around the Active Site of the Protein Kinase p38\u0026alpha; \u003cem\u003eAngew. Chem., Int. Ed.\u003c/em\u003e, DOI: 10.1002/anie.202522665 (2026).\u003c/li\u003e\n\u003cli\u003eAoto PC, Martin BT, Wright PE. NMR Characterization of Information Flow and Allosteric Communities in the MAP Kinase p38\u0026gamma;. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 28655 (2016).\u003c/li\u003e\n\u003cli\u003eAoto PC, Stanfield RL, Wilson IA, Dyson HJ, Wright PE. A Dynamic Switch in Inactive p38\u0026gamma; Leads to an Excited State on the Pathway to an Active Kinase. \u003cem\u003eBiochemistry\u003c/em\u003e \u003cstrong\u003e58\u003c/strong\u003e, 5160-5172 (2019).\u003c/li\u003e\n\u003cli\u003eAvitzour M, Diskin R, Raboy B, Askari N, Engelberg D, Livnah O. Intrinsically active variants of all human p38 isoforms. \u003cem\u003eFEBS J.\u003c/em\u003e \u003cstrong\u003e274\u003c/strong\u003e, 963-975 (2007).\u003c/li\u003e\n\u003cli\u003eAhuja LG, Kornev AP, McClendon CL, Veglia G, Taylor SS. Mutation of a kinase allosteric node uncouples dynamics linked to phosphotransfer. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e114\u003c/strong\u003e, E931-E940 (2017).\u003c/li\u003e\n\u003cli\u003eFoda ZH, Shan Y, Kim ET, Shaw DE, Seeliger MA. A dynamically coupled allosteric network underlies binding cooperativity in Src kinase. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 5939 (2015).\u003c/li\u003e\n\u003cli\u003eZhang X, Gureasko J, Shen K, Cole PA, Kuriyan J. An Allosteric Mechanism for Activation of the Kinase Domain of Epidermal Growth Factor Receptor. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e125\u003c/strong\u003e, 1137-1149 (2006).\u003c/li\u003e\n\u003cli\u003eSkora L, Mestan J, Fabbro D, Jahnke W, Grzesiek S. NMR reveals the allosteric opening and closing of Abelson tyrosine kinase by ATP-site and myristoyl pocket inhibitors. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e110\u003c/strong\u003e, E4437-E4445 (2013).\u003c/li\u003e\n\u003cli\u003eHu J\u003cem\u003e, et al.\u003c/em\u003e Kinase regulation by hydrophobic spine assembly in cancer. \u003cem\u003eMol. Cell. Biol.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 264-276 (2015).\u003c/li\u003e\n\u003cli\u003eMelo MCR, Bernardi RC, de la Fuente-Nunez C, Luthey-Schulten Z. Generalized correlation-based dynamical network analysis: a new high-performance approach for identifying allosteric communications in molecular dynamics trajectories. \u003cem\u003eJ. Chem. Phys.\u003c/em\u003e \u003cstrong\u003e153\u003c/strong\u003e, 134104 (2020).\u003c/li\u003e\n\u003cli\u003eErin S, George PL. Analysis of coordinated NMR chemical shifts to map allosteric regulatory networks in proteins. \u003cem\u003eMethods\u003c/em\u003e \u003cstrong\u003e209\u003c/strong\u003e, 40-47 (2023).\u003c/li\u003e\n\u003cli\u003ePhillips JC\u003cem\u003e, et al.\u003c/em\u003e Scalable molecular dynamics on CPU and GPU architectures with NAMD. \u003cem\u003eJ. Chem. Phys.\u003c/em\u003e \u003cstrong\u003e153\u003c/strong\u003e, 044130 (2020).\u003c/li\u003e\n\u003cli\u003eHaque A\u003cem\u003e, et al.\u003c/em\u003e Interaction Analysis of MRP1 with Anticancer Drugs Used in Ovarian Cancer: In Silico Approach. \u003cem\u003eLife\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 383 (2022).\u003c/li\u003e\n\u003cli\u003eWilson KP\u003cem\u003e, et al.\u003c/em\u003e Crystal structure of p38 mitogen-activated protein kinase. \u003cem\u003eJ. Biol. Chem.\u003c/em\u003e \u003cstrong\u003e271\u003c/strong\u003e, 27696-27700 (1996).\u003c/li\u003e\n\u003cli\u003eKrishnan K, Tian H, Tao P, Verkhivker GM. Probing conformational landscapes and mechanisms of allosteric communication in the functional states of the ABL kinase domain using multiscale simulations and network-based mutational profiling of allosteric residue potentials. \u003cem\u003eJ. Chem. Phys.\u003c/em\u003e \u003cstrong\u003e157\u003c/strong\u003e, 245101 (2022).\u003c/li\u003e\n\u003cli\u003eTaylor SS, Zhang P, Steichen JM, Keshwani MM, Kornev AP. PKA: lessons learned after twenty years. \u003cem\u003eBiochim. Biophys. Acta\u003c/em\u003e \u003cstrong\u003e1834\u003c/strong\u003e, 1271-1278 (2013).\u003c/li\u003e\n\u003cli\u003eAhuja LG, Aoto PC, Kornev AP, Veglia G, Taylor SS. Dynamic allostery-based molecular workings of kinase:peptide complexes. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e116\u003c/strong\u003e, 15052-15061 (2019).\u003c/li\u003e\n\u003cli\u003ePanjarian S, Iacob RE, Chen S, Engen JR, Smithgall TE. Structure and dynamic regulation of Abl kinases. \u003cem\u003eJ. Biol. Chem.\u003c/em\u003e \u003cstrong\u003e288\u003c/strong\u003e, 5443-5450 (2013).\u003c/li\u003e\n\u003cli\u003eGriswold IJ\u003cem\u003e, et al.\u003c/em\u003e Kinase domain mutants of Bcr-Abl exhibit altered transformation potency, kinase activity, and substrate utilization, irrespective of sensitivity to imatinib. \u003cem\u003eMol. Cell. Biol.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 6082-6093 (2006).\u003c/li\u003e\n\u003cli\u003eShan Y\u003cem\u003e, et al.\u003c/em\u003e Oncogenic mutations counteract intrinsic disorder in the EGFR kinase and promote receptor dimerization. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e149\u003c/strong\u003e, 860-870 (2012).\u003c/li\u003e\n\u003cli\u003eShah NG\u003cem\u003e, et al.\u003c/em\u003e Novel Noncatalytic Substrate-Selective p38\u0026alpha;-Specific MAPK Inhibitors with Endothelial-Stabilizing and Anti-Inflammatory Activity. \u003cem\u003eJ. Immunol.\u003c/em\u003e \u003cstrong\u003e198\u003c/strong\u003e, 3296-3306 (2017).\u003c/li\u003e\n\u003cli\u003eTaylor SS, Kornev AP. Protein kinases: evolution of dynamic regulatory proteins. \u003cem\u003eTrends Biochem. Sci.\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 65-77 (2011).\u003c/li\u003e\n\u003cli\u003eKornev AP, Taylor SS. Dynamics-Driven Allostery in Protein Kinases. \u003cem\u003eTrends Biochem. Sci.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 628-647 (2015).\u003c/li\u003e\n\u003cli\u003eHuse M, Kuriyan J. The conformational plasticity of protein kinases. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 275-282 (2002).\u003c/li\u003e\n\u003cli\u003eBIOVIA. BIOVIA Discovery Studio Visualizer v21.1. \u003cem\u003eDassault Syst\u0026egrave;mes, San Diego, SA, USA\u003c/em\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eDegorce F, Card A, Soh S, Trinquet E, Knapik GP, Xie B. HTRF: A technology tailored for drug discovery - a review of theoretical aspects and recent applications. \u003cem\u003eCurr. Chem. Genomics\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 22-32 (2009).\u003c/li\u003e\n\u003cli\u003eKeesler GA\u003cem\u003e, et al.\u003c/em\u003e Purification and activation of recombinant p38 isoforms alpha, beta, gamma, and delta. \u003cem\u003eProtein Expr. Purif.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 221-228 (1998).\u003c/li\u003e\n\u003cli\u003eHolliday MJ, Camilloni C, Armstrong GS, Vendruscolo M, Eisenmesser EZ. Networks of Dynamic Allostery Regulate Enzyme Function. \u003cem\u003eStructure\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 276-286 (2017).\u003c/li\u003e\n\u003cli\u003eSok P\u003cem\u003e, et al.\u003c/em\u003e MAP Kinase-Mediated Activation of RSK1 and MK2 Substrate Kinases. \u003cem\u003eStructure\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1101-1113.e1105 (2020).\u003c/li\u003e\n\u003cli\u003eThapa D\u003cem\u003e, et al.\u003c/em\u003e TAB1-Induced Autoactivation of p38\u0026alpha; Mitogen-Activated Protein Kinase Is Crucially Dependent on Threonine 185. \u003cem\u003eMol. Cell. Biol.\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eLiu MY, Hua WK, Chen CJ, Lin WJ. The MKK-Dependent Phosphorylation of p38\u0026alpha; Is Augmented by Arginine Methylation on Arg49/Arg149 during Erythroid Differentiation. \u003cem\u003eInt. J. Mol. Sci.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eAmbrosino C\u003cem\u003e, et al.\u003c/em\u003e Negative feedback regulation of MKK6 mRNA stability by p38alpha mitogen-activated protein kinase. \u003cem\u003eMol. Cell. Biol.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 370-381 (2003).\u003c/li\u003e\n\u003cli\u003eHuang S, New L, Pan Z, Han J, Nemerow GR. Urokinase plasminogen activator/urokinase-specific surface receptor expression and matrix invasion by breast cancer cells requires constitutive p38alpha mitogen-activated protein kinase activity. \u003cem\u003eJ. Biol. Chem.\u003c/em\u003e \u003cstrong\u003e275\u003c/strong\u003e, 12266-12272 (2000).\u003c/li\u003e\n\u003cli\u003eMcLaughlin MM\u003cem\u003e, et al.\u003c/em\u003e Identification of mitogen-activated protein (MAP) kinase-activated protein kinase-3, a novel substrate of CSBP p38 MAP kinase. \u003cem\u003eJ. Biol. Chem.\u003c/em\u003e \u003cstrong\u003e271\u003c/strong\u003e, 8488-8492 (1996).\u003c/li\u003e\n\u003cli\u003eBhattacharya S\u003cem\u003e, et al.\u003c/em\u003e NMR-guided directed evolution. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e610\u003c/strong\u003e, 389-393 (2022).\u003c/li\u003e\n\u003cli\u003eSaavedra HG, Wrabl JO, Anderson JA, Li J, Hilser VJ. Dynamic allostery can drive cold adaptation in enzymes. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e558\u003c/strong\u003e, 324-328 (2018).\u003c/li\u003e\n\u003cli\u003eSingh H, Das CK, Buchmuller BC, Sch\u0026auml;fer LV, Summerer D, Linser R. Epigenetic CpG duplex marks probed by an evolved DNA reader via a well-tempered conformational plasticity. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 6495-6506 (2023).\u003c/li\u003e\n\u003cli\u003eRibeiro JV\u003cem\u003e, et al.\u003c/em\u003e QwikMD\u0026thinsp;-\u0026thinsp;Integrative Molecular Dynamics Toolkit for Novices and Experts. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 26536 (2016).\u003c/li\u003e\n\u003cli\u003eMichaud-Agrawal N, Denning EJ, Woolf TB, Beckstein O. MDAnalysis: a toolkit for the analysis of molecular dynamics simulations. \u003cem\u003eJ. Comput. Chem.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 2319-2327 (2011).\u003c/li\u003e\n\u003cli\u003eLam SK, Pitrou A, Seibert S. Numba: a LLVM-based Python JIT compiler. In: \u003cem\u003eProceedings of the Second Workshop on the LLVM Compiler Infrastructure in HPC\u003c/em\u003e). Association for Computing Machinery (2015).\u003c/li\u003e\n\u003cli\u003eBehnel S, Bradshaw R, Citro C, Dalcin L, Seljebotn DS, Smith K. Cython: The Best of Both Worlds. \u003cem\u003eComput. Sci. Eng. \u003c/em\u003e\u003cstrong\u003e13\u003c/strong\u003e, 31-39 (2011).\u003c/li\u003e\n\u003cli\u003eWilliam H, Andrew D, Klaus S. VMD: Visual molecular dynamics. \u003cem\u003eJ. Mol. Graph. Model.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 33-38 (1996).\u003c/li\u003e\n\u003cli\u003eSkinner SP, Fogh RH, Boucher W, Ragan TJ, Mureddu LG, Vuister GW. CcpNmr AnalysisAssign: a flexible platform for integrated NMR analysis. \u003cem\u003eJ. Biomol. NMR\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 111-124 (2016).\u003c/li\u003e\n\u003cli\u003eBieri M, Gooley PR. Automated NMR relaxation dispersion data analysis using NESSY. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 421 (2011).\u003c/li\u003e\n\u003cli\u003eRoy Acharyya S. Identification of allosteric communication pathways within p38 alpha kinase from dynamical network analysis and NMR spectroscopy.). DRAFT VERSION edn. TUDOdata (2026).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8958159/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8958159/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Kinases are major drug targets especially in cancer therapy. However, the high degree of conservation of their active sites hinders the development of selective inhibitors, motivating a deeper understanding of kinase conformational ensembles and allosteric communication pathways. Here, we use dynamical network analysis to identify key residues involved in a dynamic allostery between the N- and C-lobes that connects the major functional units of the MAP kinase p38α. By combining NMR spectroscopy, activity assays, and in silico analysis of wildtype protein and mutants in the presence or absence of an active-site inhibitor, we experimentally validate the obtained architecture with respect to global protein motion and long-range allosteric modulation. Notably, the identified network highlights communication pathways across several functional sites, prominently involving the allosteric site, the activation loop, and the lipid-binding domain with its embedded cryptic pocket in the C-lobe. These findings provide mechanistic insight into p38α allostery and suggest new opportunities for the rational design of allosteric modulators of MAP kinases.","manuscriptTitle":"Mapping of dynamic allostery within p38 alpha kinase via network analyses and NMR spectroscopy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-09 05:16:57","doi":"10.21203/rs.3.rs-8958159/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1aae4021-5076-4614-a311-9fb14567f3a2","owner":[],"postedDate":"March 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":64071487,"name":"Biological sciences/Biophysics/Molecular biophysics/Molecular conformation"},{"id":64071488,"name":"Physical sciences/Physics/Techniques and instrumentation/NMR spectroscopy"}],"tags":[],"updatedAt":"2026-04-20T11:04:27+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-09 05:16:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8958159","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8958159","identity":"rs-8958159","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}