Resistance to EGFR inhibitors (EGFRi) presents a major obstacle in treating non-small cell lung cancer (NSCLC). One of the most exciting new ways to find potential resistance markers involves running functional genetic screens, such as CRISPR, followed by manual triage of significantly enriched genes. This triage process to identify 'high value' hits resulting from the CRISPR screen involves manual curation that requires specialized knowledge and can take even experts several months to comprehensively complete. To find key drivers of resistance faster we build a recommendation system on top of a heterogeneous biomedical knowledge graph integrating pre-clinical, clinical, and literature evidence. The recommender system ranks genes based on trade-offs between diverse types of evidence linking them to potential mechanisms of EGFRi resistance. This unbiased approach identifies 57 resistance markers from >3,000 genes, reducing hit identification time from months to minutes. In addition to reproducing known resistance markers, our method identifies previously unexplored resistance mechanisms that we prospectively validate.

Non-alcoholic steatohepatitis (NASH) is a common form of chronic liver disease characterised by lipid accumulation, infiltration of immune cells, hepatocellular ballooning, collagen deposition and liver fibrosis. There is a high unmet need to develop treatments for NASH. We have investigated how liver fibrosis and features of advanced clinical disease can be modelled using an in vitro microphysiological system (MPS). The NASH MPS model comprises a co-culture of primary human liver cells, which were cultured in a variety of conditions including+/− excess sugar, fat, exogenous TGFβ or LPS. The transcriptomic, inflammatory and fibrotic phenotype of the model was characterised and compared using a system biology approach to identify conditions that mimic more advanced clinical disease. The transcriptomic profile of the model was shown to closely correlate with the profile of patient samples and the model displayed a quantifiable fibrotic phenotype. The effects of Obeticholic acid and Elafibranor, were evaluated in the model, as wells as the effects of dietary intervention, with all able to significantly reduce inflammatory and fibrosis markers. Overall, we demonstrate how the MPS NASH model can be used to model different aspects of clinical NASH but importantly demonstrate its ability to model advanced disease with a quantifiable fibrosis phenotype.

Evasion of apoptosis is a hallmark of cancer wherein overexpression and amplification of pro-survival BCL2-family genes like MCL1 is a common observation. MCL1 is frequently amplified in many hematological cancers like Multiple Myeloma (MM) that depend on it for survival. BH3 mimetic drugs, like the BCL2-specific inhibitor Venetoclax, have been successfully used in the clinic to treat certain cancers, and MCL1-selective inhibitors are currently in clinical development. While inhibition of MCL1 displays promising preclinical activity, many cancer models display acquired or intrinsic resistance to MCL1 inhibitors (MCL1i). As MCL1-targeted therapies progress clinically, understanding mechanisms that lead to resistance will be important to not only identify therapeutically-exploitable targets to combat resistance, but to also determine if these biomarkers could stratify patients most likely to respond to an MCL1i.

Abstract Drug resistance is ultimately the cause of death for most cancer patients – even initially strong responses to treatment are usually followed by the emergence of resistance over time. This suggests the existence of residual or persistent cancer cells, creating a reservoir that ultimately gives rise to stable resistance. These drug tolerant persisters (‘the deadly survivors’) have been described for over a decade in numerous studies; they are often present as a minor fraction of the total tumour population and may exploit non-genetic (transcriptional) programs to allow the cells to survive drug treatment. 10-20% of lung adenocarcinoma patients harbour activating mutations in EGFR. Although treatment with the EGFR kinase inhibitor osimertinib has improved overall survival in such patients, almost all patients ultimately develop drug resistance. We carried out parallel genome-wide CRISPR gain and loss of function screens in EGFR mutant lung cancer cell lines treated with EGFR inhibitors, to identify the genes and pathways that may be important in enabling the survival of persister cells. We observed recurrent resistant genes in previously identified resistance pathways including PI3K (PTEN, TSC1, TSC2), MAPK (KRAS, NF1, MET), cell death (BCL2L11, BAX), the mediator complex (MED24, MED19) and ubiquitination (KCTD5, KEAP1). A secondary screen of 63 resistance genes that combined high content microscopy with CRISPR gene knockout demonstrated that 21% (13/63) of genes were associated with increased nuclear localisation of YAP1/WWTR1, key activators of the Hippo pathway. A closer review of the CRISPR screen data confirmed that many resistance hits are members of this pathway - upstream regulators (NF2, AMOTL2), core signalling genes (LATS1, LATS2), main effectors (WWTR1, YAP1), transcriptional co-effectors (TEAD3, FOSL1, VGLL4) and the SWI/SNF complex (ARID2, SMARCA4, SMARCB1, PBRM1). Hippo signalling is mediated through YAP1 and WWTR1 which bind to TEAD transcription factors and activate transcriptional programs affecting cell proliferation and apoptosis. We confirmed using CRISPR that knockout (NF2) or overexpression (YAP1, WWTR1) of key Hippo genes in the EGFR mutant lung cancer cell lines PC-9, HCC827 and HCC4006 resulted in up to 60-fold increased resistance to osimertinib and increased expression of canonical Hippo transcriptional targets. We therefore reasoned that the Hippo pathway might be involved in maintaining the survival of drug tolerant persister cells in this setting. Acute treatment of EGFR mutant cell lines with osimertinib was associated with increased nuclear localisation of YAP1 and WWTR1 and increased expression of canonical Hippo transcriptional targets. Furthermore, the combination of osimertinib and a TEAD inhibitor (K-975) almost completely abolished the survival of drug tolerant persister cells following treatment, indicating that this pathway is an important survival mechanism following drug treatment. Consequently, we propose Hippo signalling as an important target mechanism for the prevention of resistance to osimertinib. Citation Format: Matthias Pfeifer, Jonathan Brammeld, Stacey Price, Matthew Martin, Hannah Thorpe, Aurelie Bornot, Ercia Banks, Nin Guan, Shanade Dunn, Maria Lisa Guerriero, Daniel O'Neill, James Pilling, Davide Gianni, James Brownell, Paul Smith, Ultan McDermott. Gain and loss of function genome-wide CRISPR screens identify Hippo signalling as an important driver of resistance in EGFR mutant lung cancer [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2021 Oct 7-10. Philadelphia (PA): AACR; Mol Cancer Ther 2021;20(12 Suppl):Abstract nr P066.

Introduction and Objectives Pulmonary surfactant homeostasis is critical to lung function as it coats the vast air-liquid interface, reducing surface tension and preventing alveolar collapse. surfactant proteins (SP)-A and D play essential anti-pathogenic defence and immunoregulation roles to maintain lung homeostasis. SP-D is reduced in COPD. We hypothesised lipid and protein surfactant dysregulation could contribute to pathological COPD mechanisms, this requires detailed characterisation. Methods We conducted mass spectrometry-based multiomic (proteomic, lipidomic and metabolomic) analysis of bronchoalveolar lavage to comprehensively characterize surfactant dysregulation in a well-characterised ex-smoking mild/moderate COPD cohort (n=31), with median (IQR) age of 70 (9.5) and FEV1% predicted 73 (21), compared with healthy ex-smoking volunteers (n=20), with age 67.5 (6.75) and FEV1% predicted 100.5 (11.75). To identify the multiomic signatures of COPD, we fitted linear mixed-effects models to the data, accounting for the effects of confounding variables, such as age, gender and individual differences simultaneously. We further characterised purified alveolar macrophage gene expression differences. Results SP-A, SP-B, and SP-D levels were lower in COPD vs control with log2fold changes (FC) of -0.7 (p=1.1x10-6), -0.6 (p=4.8x10-6) and -0.6 (p=3.0x10-5), and showed positive correlations of 0.65 (p=1.5x10-12), 0.62 (p=7.7x10-11) and 0.43 (p=2.3x10-4) with lung function (FEV1/FVC), respectively. SP-C was below detection limit. NAPSA and CTSH, responsible for SP-B synthesis, were lower in COPD, log2FC of -0.1 (p=2.8x10-6) and -0.4 (p=4.3x10-4), respectively. Fatty acid binding protein 4 (FABP4) and CD44, involved in surfactant regulation, were decreased in COPD, log2FC of -0.4 (p=8.0x10-3) and -0.7 (p=1.1x10-5), respectively. Histamine and hypoxanthine metabolites were increased in COPD, log2FC of 0.8 (p=2.5x10-2) and 1.0 (p=4.0x10-2), respectively, which could indicate immune-dysregulation and inflammation. Surfactant lipid levels were decreased in COPD, specifically, PC, PG, and PI, log2FC of -2.1 (p=2.6×10-7), -2.4 (p=5.3×10-8) and -1.5 (p=9.0×10-6), respectively. Conclusions We used a multiomics approach to comprehensively describe surfactant dysregulation in a well-characterised COPD cohort, with an aim to generate novel insights into key aspects of this biology within the COPD lung. Further mechanistic studies could help to understand the role this plays in pathological mechanisms and allow novel therapeutic target identification. Please refer to page A190 for declarations of interest related to this abstract.

Genome-wide arrayed CRISPR screening is a powerful method for drug target identification as it enables exploration of the effect of individual gene perturbations using diverse highly multiplexed functional and phenotypic assays. Using high-content imaging, we can measure changes in biomarker expression, intracellular localization, and cell morphology. Here we present the computational pipeline we have developed to support the analysis and interpretation of arrayed CRISPR screens. This includes evaluating the quality of guide RNA libraries, performing image analysis, evaluating assay results quality, data processing, hit identification, ranking, visualization, and biological interpretation.

Immunotherapies including PD-L1 blockade have shown remarkable increases in the T cell-directed antitumor response; however, efficacy is seen only in a minority of patients. Recently, pooled CRISPR-Cas9 knockout (CRISPRn) screens in tumor/immune co-culture systems have identified a number of genes that confer resistance to T cell killing in pathways including antigen presentation and cytokine signaling, providing insight into tumor mechanisms that cause resistance to immunotherapies. The development of an arrayed CRISPRn screen in a tumor/immune co-culture system would allow the identification of novel targets for immuno-oncology, characterization of hits from pooled screens, and multiple assay endpoints to be measured per gene. Here, a small-scale arrayed CRISPRn screen was successfully developed to investigate the effects on a co-culture of T cells and Cas9-expressing PC9 lung adenocarcinoma cells modified to express anti-CD3 antibody on the cell surface (PC9-OKT3 T cell system). A focused CRISPRn library was designed to target genes involved in known resistance mechanisms (including antigen presentation, cytokine signaling, and apoptosis) as well as genes involved in immune synapse interactions. The viability of PC9 cells was assessed in two-dimensional adherent co-cultures via longitudinal imaging analysis. Knockout of epidermal growth factor receptor (EGFR) and PLK1 in tumor cells cultured alone or with T cells resulted in increased tumor cell death, as expected, whereas knockout of the test gene ICAM1 showed subtle donor-specific resistance to T cell killing. Taken together, these data provide proof of concept for arrayed CRISPRn screens in tumor/immune co-culture systems and warrant further investigation of in vitro co-culture models.

Modified messenger RNAs (mRNAs) hold great potential as therapeutics by using the body's own processes for protein production. However, a key challenge is efficient delivery of therapeutic mRNA to the cell cytosol and productive protein translation. Lipid nanoparticles (LNPs) are the most clinically advanced system for nucleic acid delivery; however, a relatively narrow therapeutic index makes them unsuitable for many therapeutic applications. A key obstacle to the development of more potent LNPs is a limited mechanistic understanding of the interaction of LNPs with cells. To address this gap, we performed an arrayed CRISPR screen to identify novel pathways important for the functional delivery of MC3 lipid-based LNP encapsulated mRNA (LNP-mRNA). Here, we have developed and validated a robust, high-throughput screening-friendly phenotypic assay to identify novel targets that modulate productive LNP-mRNA delivery. We screened the druggable genome (7795 genes) and validated 44 genes that either increased (37 genes) or inhibited (14 genes) the productive delivery of LNP-mRNA. Many of these genes clustered into families involved with host cell transcription, protein ubiquitination, and intracellular trafficking. We show that both UDP-glucose ceramide glucosyltransferase and V-type proton ATPase can significantly modulate the productive delivery of LNP-mRNA, increasing and decreasing, respectively, with both genetic perturbation and by small-molecule inhibition. Taken together, these findings shed new light into the molecular machinery regulating the delivery of LNPs into cells and improve our mechanistic understanding of the cellular processes modulating the interaction of LNPs with cells.

Obesity results from a chronic imbalance between energy intake and energy output but remains difficult to prevent or treat in humans. Adenosine monophosphate (AMP)-activated protein kinase (AMPK) is an important regulator of energy homeostasis1–3 and is a molecular target of drugs used for the treatment of metabolic diseases, including obesity4,5. Here we show that mice expressing a gain-of-function AMPK mutant6 display a change in morphology of subcutaneous white adipocytes that is reminiscent of browning. However, despite a dramatic increase in mitochondrial content, Ucp1 expression is undetectable in these adipocytes. In response to a high-fat diet (HFD), expression of skeletal muscle–associated genes is induced in subcutaneous white adipocytes from the gain-of-function AMPK mutant mice. Chronic genetic AMPK activation results in protection against diet-induced obesity due to an increase in whole-body energy expenditure, most probably because of a substantial increase in the oxygen consumption rate of white adipose tissue. These results suggest that AMPK activation enriches, or leads to the emergence of, a population of subcutaneous white adipocytes that produce heat via Ucp1-independent uncoupling of adenosine triphosphate (ATP) production on a HFD. Our findings indicate that AMPK activation specifically in adipose tissue may have therapeutic potential for the treatment of obesity. AMPK is a master regulator of cellular metabolism. Here the authors show that a constitutively active AMPK mutation protects mice fed a high-fat diet from obesity by increasing energy expenditure in subcutaneous white adipocytes, possibly as a result of the emergence of a hitherto-unknown type of adipocyte.

We have previously developed a high-throughput bioengineered human cardiac organoid (hCO) platform, which provides functional contractile tissue with biological properties similar to native heart tissue, including mature, cell-cycle-arrested cardiomyocytes. In this study, we perform functional screening of 105 small molecules with pro-regenerative potential. Our findings reveal surprising discordance between our hCO system and traditional 2D assays. In addition, functional analyses uncovered detrimental effects of many hit compounds. Two pro-proliferative small molecules without detrimental impacts on cardiac function were identified. High-throughput proteomics in hCO revealed synergistic activation of the mevalonate pathway and a cell-cycle network by the pro-proliferative compounds. Cell-cycle reentry in hCO and in vivo required the mevalonate pathway as inhibition of the mevalonate pathway with a statin attenuated pro-proliferative effects. This study highlights the utility of human cardiac organoids for pro-regenerative drug development, including identification of underlying biological mechanisms and minimization of adverse side effects.

Flexibility is an intrinsic essential feature of protein structures, directly linked to their functions. To this day, most of the prediction methods use the crystallographic data (namely B-factors) as the only indicator of protein's inner flexibility and predicts them as rigid or flexible. PredyFlexy stands differently from other approaches as it relies on the definition of protein flexibility (i) not only taken from crystallographic data, but also (ii) from Root Mean Square Fluctuation (RMSFs) observed in Molecular Dynamics simulations. It also uses a specific representation of protein structures, named Long Structural Prototypes (LSPs). From Position-Specific Scoring Matrix, the 120 LSPs are predicted with a good accuracy and directly used to predict (i) the protein flexibility in three categories (flexible, intermediate and rigid), (ii) the normalized B-factors, (iii) the normalized RMSFs, and (iv) a confidence index. Prediction accuracy among these three classes is equivalent to the best two class prediction methods, while the normalized B-factors and normalized RMSFs have a good correlation with experimental and in silico values. Thus, PredyFlexy is a unique approach, which is of major utility for the scientific community. It support parallelization features and can be run on a local cluster using multiple cores.

A substantial challenge in phenotypic drug discovery is the identification of the molecular targets that govern a phenotypic response of interest. Several experimental strategies are available for this, the so-called target deconvolution process. Most of these approaches exploit the affinity between a small-molecule compound and its putative targets or use large-scale genetic manipulations and profiling. Each of these methods has strengths but also limitations such as bias toward high-affinity interactions or risks from genetic compensation. The use of computational methods for target and mechanism of action identification is a complementary approach that can influence each step of a phenotypic screening campaign. Here, we describe how cheminformatics and bioinformatics are embedded in the process from initial selection of a focused compound library from a large set of historical small-molecule screens through the analysis of screening results. We present a deconvolution method based on enrichment analysis and using known bioactivity data of screened compounds to infer putative targets, pathways, and biological processes that are consistent with the observed phenotypic response. As an example, the approach is applied to a cellular screen aiming at identifying inhibitors of tumor necrosis factor–α production in lipopolysaccharide-stimulated THP-1 cells. In summary, we find that the approach can contribute to solving the often very complex target deconvolution task.

The traditional drug discovery strategy of pursuing "one compound-one target" has had difficulties delivering novel therapies for complex diseases currently lacking adequate treatments. An alternative and complementary approach is the design of multitargeted modulators simultaneously addressing multiple pathological mechanisms or overcoming pathway robustness. In this study, we propose a methodology to increase the probability of success for developing dual-acting modulators by systematically and rationally evaluating all dual-acting modulator opportunities within a specific disease area. This approach employs a combination of a five-step medicinal chemistry evaluation and a two-step biological analysis to help select the optimal target combination. It provides a novel methodology suitable for widespread application across disease areas. To exemplify the power of this approach, we focus on an analysis of the gastrointestinal (GI) disease area to identify opportunities supported by current literature data.

Protein structures are valuable tools for understanding protein function. However, protein structures are often considered as rigid bodies while protein dynamics is a key element in protein function. Full understanding of protein function at the molecular level requires accounting for protein flexibility. Protein structure can be described by a limited set of recurring local structures. We established a library composed of 120 overlapping long structural prototypes (LSPs) representing fragments of 11 residues in length and covering all known local protein structures. We developed a prediction method that proposes structural candidates in terms of LSPs along a given sequence (Bornot et al., Proteins, 2009). We utilized this methodology to predict protein flexibility. X-ray structure coordinates do not necessarily reflect all the diversity of conformations adopted by proteins in cells. On the other hand, X-ray B-factors are considered as good indicators of flexibility. Alternatively, molecular dynamics simulations (MD) are used in routine to account for dynamics. So we performed MD on a set of proteins. We examine flexibility according two different descriptors, root mean square fluctuations and B-factors. We define three flexibility classes and propose a method based on the LSP prediction method for predicting flexibility along the sequence. This method competes rather efficiently with the most recent, cutting-edge methods based on true flexibility data learning with sophisticated algorithms (Bornot et al., Proteins, 2011). Flexibility prediction correlates well with disorder prediction.

Protein structures are necessary for understanding protein function at a molecular level. Dynamics and flexibility of protein structures are also key elements of protein function. So, we have proposed to look at protein flexibility using novel methods: (i) using a structural alphabet and (ii) combining classical X-ray B-factor data and molecular dynamics simulations. First, we established a library composed of structural prototypes (LSPs) to describe protein structure by a limited set of recurring local structures. We developed a prediction method that proposes structural candidates in terms of LSPs and predict protein flexibility along a given sequence. Second, we examine flexibility according to two different descriptors: X-ray B-factors considered as good indicators of flexibility and the root mean square fluctuations, based on molecular dynamics simulations. We then define three flexibility classes and propose a method based on the LSP prediction method for predicting flexibility along the sequence. This method does not resort to sophisticate learning of flexibility but predicts flexibility from average flexibility of predicted local structures. The method is implemented in PredyFlexy web server. Results are similar to those obtained with the most recent, cutting-edge methods based on direct learning of flexibility data conducted with sophisticated algorithms. PredyFlexy can be accessed at http://www.dsimb.inserm.fr/dsimb_tools/predyflexy/.

Protein structures are valuable tools for understanding protein function. However, protein dynamics is also considered a key element in protein function. Therefore, in addition to structural analysis, fully understanding protein function at the molecular level now requires accounting for flexibility. However, experimental techniques that produce both types of information simultaneously are still limited. Prediction approaches are useful alternative tools for obtaining otherwise unavailable data. It has been shown that protein structure can be described by a limited set of recurring local structures. In this context, we previously established a library composed of 120 overlapping long structural prototypes (LSPs) representing fragments of 11 residues in length and covering all known local protein structures. On the basis of the close sequence-structure relationship observed in LSPs, we developed a novel prediction method that proposes structural candidates in terms of LSPs along a given sequence. The prediction accuracy rate was high given the number of structural classes. In this study, we use this methodology to predict protein flexibility. We first examine flexibility according to two different descriptors, the B-factor and root mean square fluctuations from molecular dynamics simulations. We then show the relevance of using both descriptors together. We define three flexibility classes and propose a method based on the LSP prediction method for predicting flexibility along the sequence. The prediction rate reaches 49.6%. This method competes rather efficiently with the most recent, cutting-edge methods based on true flexibility data learning with sophisticated algorithms. Accordingly, flexibility information should be taken into account in structural prediction assessments.

Protein structures are classically described in terms of secondary structures. However, even if the regular secondary structures have relevant physical meaning, their recognition based on atomic coordinates has a number of important limitations, such as uncertainties in the assignment of the boundaries of the helical and β-strand regions. In addition, an average of about 50% of all residues are assigned to an irregular state, i.e., the coil. These limitations have led different research teams to focus on abstracting the conformation of the protein backbone in the localized short stretches. To this end, different geometric measures are being used to cluster local stretches in protein structures in a chosen number of states. A prototype representative of the local structures in each cluster is then generally defined. These libraries of local structure prototypes are named "structural alphabets". We have developed a structural alphabet, denoted protein blocks, not only to approximate the protein structure but also to predict them from the sequence. Since its development, we and others have explored numerous new research fields using this structural alphabet. Here, we review some of the most interesting applications of this structural alphabet.

α-Helical transmembrane proteins (TMPα) are composed of a series of helices embedded in the lipid bilayer. Due to technical difficulties, few 3D structures are available. Therefore, the design of structural models of TMPα is of major interest. We study the secondary structures of TMPα by analyzing the influence of secondary structures assignment methods (SSAMs). For this purpose, a published and updated benchmark databank of TMPα is used and several SSAMs (9) are evaluated. The analysis of the results points to significant differences in SSA depending on the methods used. Pairwise comparisons between SSAMs led to more than 10% of disagreement. Helical regions corresponding to transmembrane zones are often correctly characterized. The study of the sequence–structure relationship shows very limited differences with regard to the structural disagreement. Secondary structure prediction based on Bayes’ rule and using only a single sequence give correct prediction rates ranging from 78 to 81%. A structural alphabet approach gives a slightly better prediction, i.e., only 2% less than the best equivalent approach, whereas the prediction rate with a very different assignment bypasses 86%. This last result highlights the importance of the correct assignment choice to evaluate the prediction assessment.

A relevant and accurate description of three‐dimensional (3D) protein structures can be achieved by characterizing recurrent local structures. In a previous study, we developed a library of 120 3D structural prototypes encompassing all known 11‐residues long local protein structures and ensuring a good quality of structural approximation. A local structure prediction method was also proposed. Here, overlapping properties of local protein structures in global ones are taken into account to characterize frequent local networks. At the same time, we propose a new long local structure prediction strategy which involves the use of evolutionary information coupled with Support Vector Machines (SVMs). Our prediction is evaluated by a stringent geometrical assessment. Every local structure prediction with a Cα RMSD less than 2.5 Å from the true local structure is considered as correct. A global prediction rate of 63.1% is then reached, corresponding to an improvement of 7.7 points compared with the previous strategy. In the same way, the prediction of 88.33% of the 120 structural classes is improved with 8.65% mean gain. 85.33% of proteins have better prediction results with a 9.43% average gain. An analysis of prediction rate per local network also supports the global improvement and gives insights into the potential of our method for predicting super local structures. Moreover, a confidence index for the direct estimation of prediction quality is proposed. Finally, our method is proved to be very competitive with cutting‐edge strategies encompassing three categories of local structure predictions. Proteins 2009. © 2009 Wiley‐Liss, Inc.