Generative artificial intelligence (AI) is now widely applied in medicinal chemistry, with detailed case studies emerging in the literature. Here, we describe an early application of REINVENT, AstraZeneca's in-house generative molecular design platform, to identify new inhibitor scaffolds for hematopoietic progenitor kinase 1 (HPK1). REINVENT was deployed at two stages of the project to address distinct design objectives. For hit identification, transfer learning on kinase-active compounds, followed by reinforcement learning guided by QSAR-based scoring, led to the discovery of three active chemotypes. Subsequently, REINVENT was applied to scaffold hopping, using 3D pharmacophore and docking models as scoring functions, which enabled the identification of two additional active chemotypes. Optimization of one of these scaffolds delivered a compound with potent cellular activity, kinase selectivity, and favorable rat pharmacokinetics. These results demonstrate the value of integrating generative AI with medicinal chemistry expertise and support broader application of the approach in future discovery programs.

Hematopoietic progenitor kinase 1 (HPK1) is a negative regulator of the T cell receptor signaling pathway and is therefore a target of interest for immunooncology. Nonselective HPK1 inhibitors may affect other kinase components of T cell activation, blunting the beneficial impact of enhanced T cell activity that results from HPK1 inhibition itself. Here, we report the discovery of pyrazine carboxamide HPK1 inhibitors and their optimization through structure-based drug design to afford a highly selective HPK1 inhibitor, compound 24 (AZ3246). This compound induces IL-2 secretion in T cells with an EC50 of 90 nM without inhibiting antagonistic kinases, exhibits pharmacokinetic properties consistent with oral dosing, and demonstrates antitumor activity in the EMT6 syngeneic mouse model.

Targeting CDK2 with first generation CDK2 inhibitors suffered from a reduced therapeutic index likely due to toxicity stemming from lack of selectivity against the CDK family and other kinases. Recently, CDK2 has been identified as a mediator of resistance to CDK4/6 inhibitors in the context of high levels of cyclin E expression. Discovery of highly selective CDK2 inhibitors may minimize off-target effects, reduce toxicity observed with first generation CDK2 inhibitors, and allow precise targeting of aberrant cell cycle progression and resistance mechanisms mediated by high cyclin E/CDK2 activity. To this end, we report the discovery of AZD8421, a potent and highly selective CDK2 inhibitor, which exhibits superior selectivity for CDK2 over CDK1, other CDK family members, and the broader human kinome. AZD8421 demonstrates favorable pharmacokinetic properties, including excellent solubility and robust in vitro stability. Demonstrated efficacy in an ovarian cancer patient-derived xenograft model further supports its potential as a therapeutic agent.

During annual influenza epidemics, influenza B viruses (IBVs) co-circulate with influenza A viruses (IAVs), can become predominant and cause severe morbidity and mortality. Phylogenetic analyses suggest that IAVs (primarily avian viruses) and IBVs (primarily human viruses) have diverged over long time scales. Identifying their common and distinctive features is an effective approach to increase knowledge about the molecular details of influenza infection. The virus-encoded RNA-dependent RNA polymerases (FluPolB and FluPolA) are PB1-PB2-PA heterotrimers that perform transcription and replication of the viral genome in the nucleus of infected cells. Initiation of viral mRNA synthesis requires a direct association of FluPol with the host RNA polymerase II (RNAP II), in particular the repetitive C-terminal domain (CTD) of the major RNAP II subunit, to enable "cap-snatching" whereby 5'-capped oligomers derived from nascent RNAP II transcripts are pirated to prime viral transcription. Here, we present the first high-resolution co-crystal structure of FluPolB bound to a CTD mimicking peptide at a binding site crossing from PA to PB2. By performing structure-based mutagenesis of FluPolB and FluPolA followed by a systematic investigation of FluPol-CTD binding, FluPol activity and viral phenotype, we demonstrate that IBVs and IAVs have evolved distinct binding interfaces to recruit the RNAP II CTD, despite the CTD sequence being highly conserved across host species. We find that the PB2 627 subdomain, a major determinant of FluPol-host cell interactions and IAV host-range, is involved in CTD-binding for IBVs but not for IAVs, and we show that FluPolB and FluPolA bind to the host RNAP II independently of the CTD. Altogether, our results suggest that the CTD-binding modes of IAV and IBV may represent avian- and human-optimized binding modes, respectively, and that their divergent evolution was shaped by the broader interaction network between the FluPol and the host transcriptional machinery.

Human lysine methyltransferase 2D (hKMT2D) is an epigenetic writer catalyzing the methylation of histone 3 lysine 4. hKMT2D by itself has little catalytic activity and reaches full activation as part of the WRAD2 complex, additionally comprising binding partners WDR5, RbBP5, Ash2L, and DPY30. Here, a detailed mechanistic study of the hKMT2D SET domain and its WRAD2 interactions is described. We characterized the WRAD2 subcomplexes containing full-length components and the hKMT2D SET domain. By performing steady-state analysis as a function of WRAD2 concentration, we identified the inner stoichiometry and determined the binding affinities for complex formation. Ash2L and RbBP5 were identified as the binding partners critical for the full catalytic activity of the SET domain. Contrary to a previous report, product and dead-end inhibitor studies identified hKMT2D as a rapid equilibrium random Bi-Bi mechanism with EAP and EBQ dead-end complexes. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-ToF MS) analysis showed that hKMT2D uses a distributive mechanism and gives further insights into how the WRAD2 components affect mono-, di-, and trimethylation. We also conclude that the Win motif of hKMT2D is not essential in complex formation, unlike other hKMT2 proteins.

The broad-spectrum antiviral pseudobase T705, a fluorinated pyrazinecarboxamide, is incorporated via its triphosphate form into nascent viral RNA by viral RNA-dependent RNA polymerases. Since it mimics guanine or adenine it can act as a mutagen, whereas consecutive incorporation leads to chain termination. Here we examine the structural basis for incorporation and stalling for the case of influenza polymerase, using T1106-TP, the nucleotide form of T1105, the de-fluoro analogue of T705. We used a specially designed template that allows single T1106-MP incorporation at a defined site followed by consecutive T1106-MP incorporation and stalling four nucleotides later, as demonstrated by biochemical analysis. A high-resolution cryoEM structure of influenza A/H7N9 polymerase, stalled after transcribing this template, revealed that the entire product-template duplex has backtracked by five nucleotides. Consequently, the singly incorporated T1106-MP resides at the +1 position and forms an unexpected wobble base-pair with a U in the template. The relative stability of the canonical and wobble T1106:U base-pairs in different contexts is investigated by molecular dynamics simulations. Using a different template and influenza B polymerase we also observe stalling after double incorporation of T1106-MP and structural analysis showed again that backtracking occurs, this time by four nucleotides. These results show that, at least in early elongation, consecutive T1106-MP incorporation into the product destabilises the proximal end of the product-template duplex, promoting irreversible backtracking until a more favourable overall configuration is achieved. These results give new insight into the unusual mechanism of chain termination by pyrazinecarboxamide base analogues.

Treatment recommendations for fragility fractures of the pelvis (FFP) have been provided along with the good reliable FFP classification but they are not proven in large studies and recent reports challenge these recommendations. Thus, we aimed to determine the usefulness of the FFP classification determining the treatment strategy and favored procedures in six level 1 trauma centers. Sixty cases of FFP were evaluated by six experienced pelvic surgeons, six inexperienced surgeons in training, and one surgeon trained by the originator of the FFP classification during three repeating sessions using computed tomography scans with multiplanar reconstruction. The intra-rater reliability and inter-rater reliability for therapeutic decisions (non-operative treatment vs. operative treatment) were moderate, with Fleiss kappa coefficients of 0.54 (95% confidence interval [CI] 0.44-0.62) and 0.42 (95% CI 0.34-0.49). We found a therapeutic disagreement predominantly for FFP II related to a preferred operative therapy for FFP II. Operative treated cases were generally treated with an anterior-posterior fixation. Despite the consensus on an anterior-posterior fixation, the chosen procedures are highly variable and most plausible based on the surgeon's preference.

Mer is a member of the TAM (Tyro3, Axl, Mer) kinase family that has been associated with cancer progression, metastasis, and drug resistance. Their essential function in immune homeostasis has prompted an interest in their role as modulators of antitumor immune response in the tumor microenvironment. Here we illustrate the outcomes of an extensive lead-generation campaign for identification of Mer inhibitors, focusing on the results from concurrent, orthogonal high-throughput screening approaches. Data mining, HT (high-throughput), and DECL (DNA-encoded chemical library) screens offered means to evaluate large numbers of compounds. We discuss campaign strategy and screening outcomes, and exemplify series resulting from prioritization of hits that were identified. Concurrent execution of HT and DECL screening successfully yielded a large number of potent, selective, and novel starting points, covering a range of selectivity profiles across the TAM family members and modes of kinase binding, and offered excellent start points for lead development.

Inhibition of Mer and Axl kinases has been implicated as a potential way to improve the efficacy of current immuno-oncology therapeutics by restoring the innate immune response in the tumor microenvironment. Highly selective dual Mer/Axl kinase inhibitors are required to validate this hypothesis. Starting from hits from a DNA-encoded library screen, we optimized an imidazo[1,2-a]pyridine series using structure-based compound design to improve potency and reduce lipophilicity, resulting in a highly selective in vivo probe compound 32. We demonstrated dose-dependent in vivo efficacy and target engagement in Mer- and Axl-dependent efficacy models using two structurally differentiated and selective dual Mer/Axl inhibitors. Additionally, in vivo efficacy was observed in a preclinical MC38 immuno-oncology model in combination with anti-PD1 antibodies and ionizing radiation.

Influenza polymerase uses unique mechanisms to synthesize capped and polyadenylated mRNAs from the genomic viral RNA (vRNA) template, which is packaged inside ribonucleoprotein particles (vRNPs). Here, we visualize by cryoelectron microscopy the conformational dynamics of the polymerase during the complete transcription cycle from pre-initiation to termination, focusing on the template trajectory. After exiting the active site cavity, the template 3' extremity rebinds into a specific site on the polymerase surface. Here, it remains sequestered during all subsequent transcription steps, forcing the template to loop out as it further translocates. At termination, the strained connection between the bound template 5' end and the active site results in polyadenylation by stuttering at uridine 17. Upon product dissociation, further conformational changes release the trapped template, allowing recycling back into the pre-initiation state. Influenza polymerase thus performs transcription while tightly binding to and protecting both template ends, allowing efficient production of multiple mRNAs from a single vRNP.

The activation loop (A-loop) plays a key role in regulating the catalytic activity of protein kinases. Phosphorylation in this region enhances the phosphoryl transfer rate of the kinase domain and increases its affinity for ATP. Furthermore, the A-loop possesses autoinhibitory functions in some kinases, where it collapses onto the protein surface and blocks substrate binding when unphosphorylated. Due to its flexible nature, the A-loop is usually disordered and untraceable in kinase domain crystal structures. The resulting lack of structural information is regrettable as it impedes the design of drug A-loop contacts, which have proven favourable in multiple cases. Here, we characterize the binding with A-loop engagement between type 1.5 kinase inhibitor 'example 172' (EX172) and Mer tyrosine kinase (MerTK). With the help of crystal structures and binding kinetics, we portray how the recruitment of the A-loop elicits a two-step binding mechanism which results in a drug-target complex characterized by high affinity and long residence time. In addition, the type 1.5 compound possesses excellent kinome selectivity and a remarkable preference for the phosphorylated over the dephosphorylated form of MerTK. We discuss these unique characteristics in the context of known type 1 and type 2 inhibitors and highlight opportunities for future kinase inhibitor design.

BACKGROUND:The fragility fractures of the pelvis (FFP) classification was established to address the specific fracture morphology and dynamic instability in the elderly. Although this system is frequently used, data on the intra-rater and inter-rater reliabilities are lacking. METHODS:Six experienced and 6 inexperienced surgeons and 1 surgeon trained by the originator of the FFP classification ("gold standard") each used the FFP classification 3 times to grade the computed tomography (CT) scans of 60 patients from 6 hospitals. We assessed intra-rater and inter-rater reliabilities using Fleiss kappa statistics and the percentage of agreement using the "gold standard," the submitting hospital, and the majority vote as references. RESULTS:The intra-rater reliability for the FFP classification was mainly moderate, with a mean Fleiss kappa coefficient (and 95% confidence interval) of 0.46 (0.40 to 0.50) for the complete classification (i.e., both the main-group FFP ratings [I through III] and the subgroup ratings [a, b, and c]) and 0.60 (0.53 to 0.65) for the main group only. The inter-rater reliability was substantial for the main group classification (0.61 [0.54 to 0.66]) and moderate for the complete classification (0.53 [0.48 to 0.58]). The percentage of agreement was 68% to 80%. The lowest agreement was found for FFP II and III. CONCLUSIONS:The FFP classification displayed moderate and substantial intra-rater and inter-rater reliabilities. CLINICAL RELEVANCE:With moderate to substantial intra-rater and inter-rater reliabilities, the FFP classification forms a solid basis for future clinical investigations. The differentiation of FFP II from FFP III should be evaluated thoroughly, as the initial treatment changes from nonoperative for II to operative for III.

Influenza virus is a segmented, negative strand RNA virus with each genome segment being packaged in a distinct ribonucleoprotein particle (RNP). The RNP consists of the heterotrimeric viral RNA-dependent RNA polymerase bound to the conserved 5' and 3' ends of the genome segment (the viral promoter) with the rest of the viral RNA (vRNA) being covered by multiple copies of nucleoprotein. This review focusses on the new insights that recent crystal structures have given into the detailed molecular mechanisms by which the polymerase performs both transcription and replication of the vRNA genome. Promoter binding, in particular that of 5' end, is essential to allosterically activate all polymerase functions. Transcription is initiated by the hijacking of nascent, capped host transcripts by the process of 'cap-snatching', for which the viral polymerase makes an essential interaction with the C-terminal domain (CTD) of cellular RNA polymerase II. The structures allow a coherent mechanistic model of the subsequent cap-snatching, cap-dependent priming, elongation and self-polyadenylation steps of viral mRNA synthesis. During replication, the vRNA is copied without modification into complementary RNA (cRNA) which is packaged into cRNPs. A priming loop located in the polymerase active site is required for the unprimed synthesis of cRNA from vRNA, but is not required for cRNA to vRNA replication due to differences in the mode of initiation of RNA synthesis. Overall a picture emerges of influenza polymerase being a highly complex, flexible and dynamic machine. The challenge remains to understand in more detail how it functions within the RNP and how interacting host factors modulate its activity in the cellular context. Finally, these detailed insights have opened up new opportunities for structure-based antiviral drug design targeting multiple aspects of polymerase function.

Influenza polymerase uses short capped primers snatched from nascent Pol II transcripts to initiate transcription of viral mRNAs. Here we describe crystal structures of influenza A and B polymerase bound to a capped primer in a configuration consistent with transcription initiation ('priming state') and show by functional assays that conserved residues from both the PB2 midlink and cap-binding domains are important for positioning the capped RNA. In particular, mutation of PB2 Arg264, which interacts with the triphosphate linkage in the cap, significantly and specifically decreases cap-dependent transcription. We also compare the configuration of the midlink and cap-binding domains in the priming state with their very different relative arrangement (called the 'apo' state) in structures where the potent cap-binding inhibitor VX-787, or a close analogue, is bound. In the 'apo' state the inhibitor makes additional interactions to the midlink domain that increases its affinity beyond that to the cap-binding domain alone. The comparison suggests that the mechanism of resistance of certain mutations that allow virus to escape from VX-787, notably PB2 N510T, can only be rationalized if VX-787 has a dual mode of action, direct inhibition of capped RNA binding as well as stabilization of the transcriptionally inactive 'apo' state.

The crystal structure of bat influenza A polymerase bound to a serine-5 phosphorylated peptide mimic from the C-terminal domain of cellular RNA polymerase II shows how the two polymerases are directly coupled and suggests that the interaction site could be targeted for antiviral drug development.

The acknowledged potential of small-molecule therapeutics targeting disease-related protein-protein interactions (PPIs) has promoted active research in this field. The strategy of using small molecule inhibitors (SMIs) to fight strong (tight-binding) PPIs tends to fall short due to the flat and wide interfaces of PPIs. Here we propose a biligand approach for disruption of strong PPIs. The potential of this approach was realized for disruption of the tight-binding (KD = 100 pM) tetrameric holoenzyme of cAMP-dependent protein kinase (PKA). Supported by X-ray analysis of cocrystals, bifunctional inhibitors (ARC-inhibitors) were constructed that simultaneously associated with both the ATP-pocket and the PPI interface area of the catalytic subunit of PKA (PKAc). Bifunctional inhibitor ARC-1411, possessing a KD value of 3 pM toward PKAc, induced the dissociation of the PKA holoenzyme with a low-nanomolar IC50, whereas the ATP-competitive inhibitor H89 bound to the PKA holoenzyme without disruption of the protein tetramer.

Influenza virus polymerase transcribes or replicates the segmented RNA genome (vRNA) into respectively viral mRNA or full-length copies and initiates RNA synthesis by binding the conserved 3' and 5' vRNA ends (the promoter). In recent structures of promoter-bound polymerase, the cap-binding and endonuclease domains are configured for cap snatching, which generates capped transcription primers. Here, we present a FluB polymerase structure with a bound complementary cRNA 5' end that exhibits a major rearrangement of the subdomains within the C-terminal two-thirds of PB2 (PB2-C). Notably, the PB2 nuclear localization signal (NLS)-containing domain translocates ∼90 Å to bind to the endonuclease domain. FluA PB2-C alone and RNA-free FluC polymerase are similarly arranged. Biophysical and cap-dependent endonuclease assays show that in solution the polymerase explores different conformational distributions depending on which RNA is bound. The inherent flexibility of the polymerase allows it to adopt alternative conformations that are likely important during polymerase maturation into active progeny RNPs.

Influenza virus polymerase uses a capped primer, derived by 'cap-snatching' from host pre-messenger RNA, to transcribe its RNA genome into mRNA and a stuttering mechanism to generate the poly(A) tail. By contrast, genome replication is unprimed and generates exact full-length copies of the template. Here we use crystal structures of bat influenza A and human influenza B polymerases (FluA and FluB), bound to the viral RNA promoter, to give mechanistic insight into these distinct processes. In the FluA structure, a loop analogous to the priming loop of flavivirus polymerases suggests that influenza could initiate unprimed template replication by a similar mechanism. Comparing the FluA and FluB structures suggests that cap-snatching involves in situ rotation of the PB2 cap-binding domain to direct the capped primer first towards the endonuclease and then into the polymerase active site. The polymerase probably undergoes considerable conformational changes to convert the observed pre-initiation state into the active initiation and elongation states.