Abstract 4956: Discovering novel targetable pathways by combining functional and multi-omic data from primary ovarian cancer samples

Background: There is a critical necessity to reveal novel and tractable targets for anti-cancer treatments in indications with high unmet medical need, such as high grade serous ovarian cancer (HGSOC). However, standard process for target discovery using models such as outgrown cell lines and well-averaged readouts has yielded a less than 5% approval rate for drugs entering trials (Thomas et al. 2016 Bio.org).

Here, we describe patient-centric target discovery through the use of disease relevant primary OC samples and single cell functional characterization using a platform with proven hemonc translatability (Kornauth et al. 2021, Snijder et al. 2017). We integrate data from our functional drug testing platform under multiple drug perturbations with matching genomic and transcriptomic data to reveal associations with novel downstream regulators of sensitivity.

Methods: Sensitivity of the cancer cell compartment in primary malignant ascites samples (n = 20; 75% HGSOC) to 85 small molecule drugs, was evaluated using a proprietary and translatable deep learning-driven single cell imaging platform (Vladimer et al. 2017). Cancer cell sensitivity from the drugs was combined with WES, bulk-RNAseq and drug induced changes in phosphoproteome, and single cell RNAseq transcriptome to identify perturbed targets and pathways.

Results: Here we describe a family of TKIs including ALKi that induce cytotoxicity of cancer cells in primary samples, not previously captured in publicly available cell line drug sensitivity screening data (Iorio et al. 2016). We report novel sensitivity of OC driven by non-canonical targets of ceritinib such as FAK1 or IGF1R, mediated by the downstream signaling hub YBX1 (Kuenzi et al. 2017), involved in NFB pathway regulation (Motolani et al. 2021). Indeed, transcriptomic scRNA analysis upon ceritinib treatment of primary OC cells revealed rapid perturbation of numerous NFB pathway members, alongside YBX1 inactivation.

Conclusions: Combining functional endpoints and single cell-based differential expression analysis of primary OC samples, we have identified the NFB pathway and the regulator YBX1 as a promising novel sensitivity for HGSOC treatment development. These and several other important targetable nodes identified, sit outside the recently suggested JAK/STAT pathway (Izar et al. 2020), thereby demonstrating a pipeline towards novel drug target and pathway discovery driven by patient-centric, disease relevant models of high-need indications.

Citation Format: Irene Gutierrez-Perez, Bekir Ergüner, Pisanu Buphamalai, Joost Van Ham, Paul Heinz, Valentin Aranha, Rin Okumura, Elisabeth Waltenberger, Isabella Alt, Claudia Baumgaertler, Maja Stulic, Edgar Petru, Christoph Minichsdorfer, Judith Lafleur, Lukas Hefler, Laudia Hadjari, Lucia Dzurillova, Jozef Sufliarsky, Nikolaus Krall, Thorsten Füreder, Gregory Ian Vladimer, Bojan Vilagos, Robert Sehlke. Discovering novel targetable pathways by combining functional and multi-omic data from primary ovarian cancer samples. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4956.

Abstract 1893: Deep learning supported high content analysis of primary patient samples identifies ALK inhibition as a novel mechanism of action in a subset of ovarian cancers

Background: High unmet need of ovarian cancer (OC) suggests the discovery of new targeted therapeutics is crucial to improve patient prognosis. Unlike artificial model systems such as cell lines, primary cancer samples recapitulate the complexity of the original microenvironment consisting of cancer cells as well as stromal and immune cells; this is especially important when evaluating IO targets and signalling pathways. Supported by our previous success predicting therapy for late stage haematological cancer patients in the EXALT-I trial using AI-supported functional single cell quantification of drug action (Kornauth et. al. 2021) we set out to systematically reveal novel targets and pathways in OC using small molecule drugs (SMDs) as tools.

Single cell phenotypic screening of OC MPAs (malignant pleural effusion and ascites) was enabled by the quantification of drug effects using an end-to-end scalable deep learning driven image analysis tool chain. This custom state-of-the-art AI software is critical to enable robust primary cell screening given the diversity of cells within each sample. This revealed anaplastic lymphoma kinase (ALK), as well as structurally related targets and pathway associated proteins, as being potential novel targets in a subset of OC patient samples. There is sparse literature evidence for therapeutic utilisation of the ALK pathway in OC, and the diversity of responses indicates a further novel patient selection method.

Methods: MPAs from OC patients (n = 20) were collected and the sensitivity of the cancer cells to 85 SMDs was evaluated using high content microscopy. Individual cells were segmented and classified using convolutional neural networks and drug responses were estimated from the resulting cell counts. The integration of these results with whole exome and RNA sequencing guided target and pathway prioritisation.

Results: Screening for novel sensitivities using SMDs as tools uncovered inhibitors of ALK and related targets as having strong cancer cell cytotoxic effects, recapitulated in solid tumour biopsies. Transcriptomic profiling revealed pathway correlations to ALK inhibitor sensitivity, however non-annotated polypharmacological effects of each drug cannot yet be excluded.

Conclusions: Quantifying SMD sensitivity in a disease relevant model system identified ALK as a promising and overlooked target in OC, providing an upstream and potentially more specific target to the recently suggested MEK, PI3K and STAT3 (Papp et. al. 2018, Izar et al. 2020). While further work to confirm the target is required, this study supports a notion of patient-centric drug development using disease relevant models and deep learning. Our work introduces a novel patient-centric tool to advance understanding of the OC target landscape and provides a resource for the development of novel therapeutic approaches.

Citation Format: Irene Gutierrez-Perez, Joost Van Ham, Valentin Aranha, Rin Okumura, Elisabeth Waltenberger, Isabella Alt, Claudia Baumgaertler, Maja Stulic, Edgar Petru, Christoph Minichsdorfer, Lukas Hefler, Judith Lafleur, Nikolaus Krall, Thorsten Füreder, Gregory Ian Vladimer, Robert Sehlke, Bojan Vilagos. Deep learning supported high content analysis of primary patient samples identifies ALK inhibition as a novel mechanism of action in a subset of ovarian cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1893.

Abstract 1303: AI driven single cell analysis of drug action in solid tumor material: An entry point to functional precision medicine

Introduction: The paradigm of precision oncology is the selection of optimal therapy for individual patients. However, disease complexity prevents most cases from having a clear correlation between genetics and response, and programs using driver mutations to stratify patients into treatment arms have shown lackluster results (Middleton et al 2020). In contrast, prior work by us and colleagues demonstrated that functional analysis of drug action on single cells in unmanipulated primary tumor tissues with automated microscopy and advanced image analysis can identify effective therapies for patients with hematological cancers. In a prospective interventional trial (EXALT1 NCT03096821 Snijder et al 2017 N=17, Staber et al 2020 ASH N=56) patients treated with drugs ranked by differential ex vivo response resulted in a 55% ORR and >1.3-fold PFS improvement compared to the prior therapy in 54%. However, the scope has been limited to hematological cancers and solid tumor ascites/effusions.Adapting analysis to tumor biopsies required major advances in wet lab and computation, including robust quantification of 3D confocal images. Here, we present first biological evidence from a newly developed pipeline for the ex vivo treatment and single cell interrogation of singular cell + microaggregate suspensions from ovarian cancer biopsies.

Methods and results: Viable tissue from debulking surgery was obtained after consent and dissociated. The resulting cells and microaggregates were treated with >85 small molecules in 7 concentrations in triplicate for 72h, fixed, and stained for cancer surface markers. Cells were imaged with automated confocal microscopy at 20x across multiple 384 well plates, with z-resolution of <17 planes. Images were segmented with a modular analysis pipeline with custom neural networks in a scalable cloud environment, yielding precise characterization (e.g. viability, size, marker intensity) and localization of each cell, elucidating the complexity of biopsies. Single cell data in turn was used to determine the ability of drugs to induce targeted cytotoxicity and modify the immune system/pathways, cell morphology, and the 3D environment.

Conclusions: High content single cell phenotypic analysis of solid tumors can enable the study of drug action in primary tissues. Compared to organoids where cells are outgrown, working with primary cells in short term incubation enables analysis of drug activity prior to culture adaptation, and the presence of immune cells enables the study of I/O drugs, in a close-to-patient setting. A challenge to achieve single cell resolution of complex mixtures of aggregate/single and adherent/non-adherent cells was overcome by confocal microscopy and AI driven 3D image analysis. The screening and follow-up data indicates this could support preclinical research, biomarker discovery, and precision medicine.

Citation Format: Valentin Aranha, Diogo Tomaz, Irene Gutierrez Perez, Florian Rohrer, Joost Van Ham, Lukas Hefler, Laudia Hadjari, Judith Lafleur, Nikolaus Krall, Robert Sehlke, Bojan Vilagos, Gregory Ian Vladimer. AI driven single cell analysis of drug action in solid tumor material: An entry point to functional precision medicine [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1303.

A crucial role for Jagunal homolog 1 in humoral immunity and antibody glycosylation in mice and humans

Jagunal homolog 1 (JAGN1) has been identified as a critical regulator of neutrophil biology in mutant mice and rare-disease patients carrying JAGN1 mutations. Here, we report that Jagn1 deficiency results in alterations in the endoplasmic reticulum (ER) of antibody-producing cells as well as decreased antibody production and secretion. Consequently, mice lacking Jagn1 in B cells exhibit reduced serum immunoglobulin (Ig) levels at steady state and fail to mount an efficient humoral immune response upon immunization with specific antigens or when challenged with viral infections. We also demonstrate that Jagn1 deficiency in B cells results in aberrant IgG N-glycosylation leading to enhanced Fc receptor binding. Jagn1 deficiency in particular affects fucosylation of IgG subtypes in mice as well as rare-disease patients with loss-of-function mutations in JAGN1. Moreover, we show that ER stress affects antibody glycosylation. Our data uncover a novel and key role for JAGN1 and ER stress in antibody glycosylation and humoral immunity in mice and humans.

Abstract 5233: Towards patient-centric drug discovery: Analyzing drug action in malignant pleural effusions and ascites using high content imaging and deep learning

Introduction:

A vast majority of investigational anticancer drugs found to be active in preclinical development later do not show the desired effect in the clinic (Wong et al., Biostatistics 2019). This suggests that currently used preclinical models do not fully recapitulate the complexity of the human disease.

The study of drug activity in primary human tissue samples, by contrast, could provide a more immediate picture of the activity of a molecule's effect in a patient. Factors that have so far hampered the use of primary tissue samples for drug discovery and development include access in sufficient quantity as well as robust analytical methods.

We hypothesised that malignant pleural effusions and ascites (MPAs) of solid tumour patients are a promising model system to study drug activity in a preclinical setting. They are easily accessible in large quantities and contain cancer cells as well as major recruited immune populations. The latter could render them interesting model systems for studying I/O drug activity.

Following previous successes in studying drug action in primary human tissues of haematological cancer patients with automated microscopy (Snijder et al 2017, Lancet Haem, NCT03096821) we here describe advances in using high content imaging and deep learning-based image analysis to study drug action in MPAs of solid tumour patients.

Methods:

MPAs from patients with metastatic breast, pancreatic, lung and ovarian cancer (at least n=10 of each) were collected under appropriate ethics approval. The response of EpCam+/CD45− and CD45+ cells against small molecule drugs was evaluated using high content microscopy. Drug response was quantified with single cell resolution using regional convolutional neural networks (R-CNNs) comprising an object detection and a single cell classification stage. EpCam+ and live/dead cell classification accuracies on the validation set were >94%.

Results:

MPAs contain both cancer cells (range: 1-37%) and recruited myeloid and lymphoid immune populations with varying activation states (e.g. CD69+/PD-1+ CD8+ T-cells). Ex vivo drug responses from each patient sample were measured and combined to form an overall map of drug susceptibility across the patient population, pan-indication. Sensitivity mirrors drug approvals and also reveals drugs with potential off label use.

Conclusions:

Single-cell phenotypic analysis of MPAs enables the study of anticancer drug action in a setting that is one step closer to the clinic than cell line or outgrown organoid models of solid tumor. While initial response patterns can be observed that mirror current approvals, further biological and clinical validation must occur to understand in how far these data can be used for drug discovery and translational research purposes. Application as a functional tool for selecting salvage therapies for late-stage solid tumor patients (“functional prevision medicine”) can also be envisioned.

Citation Format: Robert Sehlke, Christina Taubert, Isabella Alt, Florian Rohrer, Elisabeth Fuchs, Bojan Vilagos, Nikolaus Krall, Minichsdorfer Christoph, Michael Schumacher, Dominik Maurer, Klaus Hackner, Judith Lafleur, Lukas Hefler, Thorsten Fureder, Gregory Ian Vladimer. Towards patient-centric drug discovery: Analyzing drug action in malignant pleural effusions and ascites using high content imaging and deep learning [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5233.

Towards patient-centric drug discovery: Drug action in malignant pleural effusions and ascites using high content imaging and deep learning

e21633

Background: Many anticancer drugs found to be active in preclinical development later do not show desired effect clinically. This suggests that currently used preclinical models do not fully recapitulate the complexity of the disease. The study of drug activity in primary samples could provide a more immediate picture of a molecule’s activity in a patient. Factors that have so far hampered the use of primary tissue samples for drug discovery and development include access in sufficient quantity as well as robust analytical methods. We hypothesised that malignant pleural effusions and ascites (MPAs) of solid tumour patients are a promising model system to study preclinical drug activity. MPAs are easily accessible and contain cancer cells as well as recruited immune cells. Following previous successes in studying drug action in primary tissues of patients with haematological malignancies with automated microscopy (Snijder et al 2017, Lanc Haem, NCT03096821) we describe advances in using high content imaging and deep learning-based image analysis to study drug action in MPAs of solid tumour patients. Methods: MPAs from patients with metastatic breast, pancreatic and ovarian cancer (at least n = 10 of each) were collected. The response of EpCam+/CD45- and CD45+ cells against small molecule drugs was evaluated using high content microscopy. Drug response was quantified at single cell resolution using regional convolutional neural networks (R-CNNs) comprising object detection and single cell classification. Results: MPAs contain both cancer cells and recruited myeloid and lymphoid immune cells with varying activation. Ex vivo drug responses from each patient sample were measured and the EC50 of each molecule determined by curve fitting. Sensitivity mirrored drug approvals for some indications, and also revealed drugs with potential off label use. On target and off-target response curves, along with integrative scores are used to visualize the effects. Conclusions: Single-cell phenotypic analysis of MPAs enables the study of anticancer drug action in a setting that is one step closer to the clinic than cell line or outgrown organoid models of solid tumor. While initial response patterns can be observed that mirror current approvals, further biological and clinical validation must occur to understand in how far these data can be used for drug discovery and translational research purposes.

The ERBB-STAT3 Axis Drives Tasmanian Devil Facial Tumor Disease

The marsupial Tasmanian devil (Sarcophilus harrisii) faces extinction due to transmissible devil facial tumor disease (DFTD). To unveil the molecular underpinnings of this transmissible cancer, we combined pharmacological screens with an integrated systems-biology characterization. Sensitivity to inhibitors of ERBB tyrosine kinases correlated with their overexpression. Proteomic and DNA methylation analyses revealed tumor-specific signatures linked to the evolutionary conserved oncogenic STAT3. ERBB inhibition blocked phosphorylation of STAT3 and arrested cancer cells. Pharmacological blockade of ERBB or STAT3 prevented tumor growth in xenograft models and restored MHC class I expression. This link between the hyperactive ERBB-STAT3 axis and major histocompatibility complex class I-mediated tumor immunosurveillance provides mechanistic insights into horizontal transmissibility and puts forward a dual chemo-immunotherapeutic strategy to save Tasmanian devils from DFTD. VIDEO ABSTRACT.

Characterization of host proteins interacting with the lymphocytic choriomeningitis virus L protein

RNA-dependent RNA polymerases (RdRps) play a key role in the life cycle of RNA viruses and impact their immunobiology. The arenavirus lymphocytic choriomeningitis virus (LCMV) strain Clone 13 provides a benchmark model for studying chronic infection. A major genetic determinant for its ability to persist maps to a single amino acid exchange in the viral L protein, which exhibits RdRp activity, yet its functional consequences remain elusive. To unravel the L protein interactions with the host proteome, we engineered infectious L protein-tagged LCMV virions by reverse genetics. A subsequent mass-spectrometric analysis of L protein pulldowns from infected human cells revealed a comprehensive network of interacting host proteins. The obtained LCMV L protein interactome was bioinformatically integrated with known host protein interactors of RdRps from other RNA viruses, emphasizing interconnected modules of human proteins. Functional characterization of selected interactors highlighted proviral (DDX3X) as well as antiviral (NKRF, TRIM21) host factors. To corroborate these findings, we infected Trim21^-/- mice with LCMV and found impaired virus control in chronic infection. These results provide insights into the complex interactions of the arenavirus LCMV and other viral RdRps with the host proteome and contribute to a better molecular understanding of how chronic viruses interact with their host.

RNA-dependent RNA-polymerases (RdRps) play a key role in the life cycle of RNA viruses. They interact with cellular proteins during replication and transcription processes and impact the immunobiology of viral infections. This study characterized the host protein interactome of the RdRp-containing L protein of the prototypic arenavirus lymphocytic choriomeningitis virus (LCMV). Several L protein interactors with proviral and antiviral effects were identified in vitro, and mice lacking the identified L protein interactor TRIM21 exhibited impaired control of chronic LCMV infection. Integration of the L protein interactomes with known RdRp interactomes from other RNA viruses highlighted common and virus-specific strategies to interact with the host proteome, which may indicate novel avenues for antiviral interventions.

Superoxide Dismutase 1 Protects Hepatocytes from Type I Interferon-Driven Oxidative Damage

Tissue damage caused by viral hepatitis is a major cause of morbidity and mortality worldwide. Using a mouse model of viral hepatitis, we identified virus-induced early transcriptional changes in the redox pathways in the liver, including downregulation of superoxide dismutase 1 (Sod1). Sod1(-/-) mice exhibited increased inflammation and aggravated liver damage upon viral infection, which was independent of T and NK cells and could be ameliorated by antioxidant treatment. Type I interferon (IFN-I) led to a downregulation of Sod1 and caused oxidative liver damage in Sod1(-/-) and wild-type mice. Genetic and pharmacological ablation of the IFN-I signaling pathway protected against virus-induced liver damage. These results delineate IFN-I mediated oxidative stress as a key mediator of virus-induced liver damage and describe a mechanism of innate-immunity-driven pathology, linking IFN-I signaling with antioxidant host defense and infection-associated tissue damage. VIDEO ABSTRACT.

The methyltransferase Setdb2 mediates virus-induced susceptibility to bacterial superinfection

Immune responses are tightly regulated to ensure efficient pathogen clearance while avoiding tissue damage. Here we report that Setdb2 was the only protein lysine methyltransferase induced during infection with influenza virus. Setdb2 expression depended on signaling via type I interferons, and Setdb2 repressed expression of the gene encoding the neutrophil attractant CXCL1 and other genes that are targets of the transcription factor NF-κB. This coincided with occupancy by Setdb2 at the Cxcl1 promoter, which in the absence of Setdb2 displayed diminished trimethylation of histone H3 Lys9 (H3K9me3). Mice with a hypomorphic gene-trap construct of Setdb2 exhibited increased infiltration of neutrophils during sterile lung inflammation and were less sensitive to bacterial superinfection after infection with influenza virus. This suggested that a Setdb2-mediated regulatory crosstalk between the type I interferons and NF-κB pathways represents an important mechanism for virus-induced susceptibility to bacterial superinfection.

Essential role of EBF1 in the generation and function of distinct mature B cell types

The transcription factor EBF1 is essential for lineage specification in early B cell development. In this study, we demonstrate by conditional mutagenesis that EBF1 is required for B cell commitment, pro-B cell development, and subsequent transition to the pre-B cell stage. Later in B cell development, EBF1 was essential for the generation and maintenance of several mature B cell types. Marginal zone and B-1 B cells were lost, whereas follicular (FO) and germinal center (GC) B cells were reduced in the absence of EBF1. Activation of the B cell receptor resulted in impaired intracellular signaling, proliferation and survival of EBF1-deficient FO B cells. Immune responses were severely reduced upon Ebf1 inactivation, as GCs were formed but not maintained. ChIP- and RNA-sequencing of FO B cells identified EBF1-activated genes that encode receptors, signal transducers, and transcriptional regulators implicated in B cell signaling. Notably, ectopic expression of EBF1 efficiently induced the development of B-1 cells at the expense of conventional B cells. These gain- and loss-of-function analyses uncovered novel important functions of EBF1 in controlling B cell immunity.

B-lymphoid cells with attributes of dendritic cells regulate T cells via indoleamine 2,3-dioxygenase

A discrete population of splenocytes with attributes of dendritic cells (DCs) and coexpressing the B-cell marker CD19 is uniquely competent to express the T-cell regulatory enzyme indoleamine 2,3-dioxygenase (IDO) in mice treated with TLR9 ligands (CpGs). Here we show that IDO-competent cells express the B-lineage commitment factor Pax5 and surface immunoglobulins. CD19 ablation abrogated IDO-dependent T-cell suppression by DCs, even though cells with phenotypic attributes matching IDO-competent cells developed normally and expressed IDO in response to interferon gamma. Consequently, DCs and regulatory T cells (Tregs) did not acquire T-cell regulatory functions after TLR9 ligation, providing an alternative perspective on the known T-cell regulatory defects of CD19-deficient mice. DCs from B-cell-deficient mice expressed IDO and mediated T-cell suppression after TLR9 ligation, indicating that B-cell attributes were not essential for B-lymphoid IDO-competent cells to regulate T cells. Thus, IDO-competent cells constitute a distinctive B-lymphoid cell type with quintessential T-cell regulatory attributes and phenotypic features of both B cells and DCs.