The Cyclin-G Associated Kinase (GAK) Is a Critical Dependency for Cell-Cycle Progression in Diffuse Large B-Cell Lymphoma Associated with Deregulation of the Retinoblastoma Tumor Suppressor

Background: Recurrent genomic alterations in specific cancers reveal tumor drivers that inform the development of rational therapeutic approaches. In diffuse large B-cell lymphoma (DLBCL), such oncoprotein-focused drug development has achieved only limited success. Phenotypic screening, less explored in DLBCL drug discovery, is an emerging alternative that can be leveraged especially through the application of machine-learning analyses to raw screening data. Under this paradigm, we reveal strong therapeutic potential in DLBCL of inhibiting the cyclin-G associated kinase (GAK). GAK participates in clathrin-mediated endocytosis and regulates mitotic-spindle alignment during cell division. It has been studied as an antiviral target that interrupts internalization at cell membranes but is minimally explored in cancer therapeutics and is a novel target in DLBCL.

Methods and Results: We performed a first-of-its-kind screen of kinase inhibitors in DLBCL followed by machine learning-based analysis using our in-house target identification platform, idTRAX. The screen revealed GAK as a strong target for both activated B-cell (ABC) and germinal center B-cell (GCB) derived cell lines with complete sparing of peripheral-blood mononuclear cells (PBMCs), included as non-malignant controls. We analyzed gene-expression data available for 414 previously untreated DLBCL patients (Lenz cohort) and found tumors with high GAK expression had dramatically worse overall survival (OS) compared to those with low expression (hazard ratio=1.84 (95% conf. int. 1.33-2.53), p=0.0002). Mechanistically, we explored GAK’s role in endocytosis but found no change in cell-surface B-cell receptor (BCR) expression after treatment with the GAK tool compound SGC-GAK-1. In addition, there was no evident perturbation of signaling pathways downstream of BCR by immunoblotting. We found instead that GAK inhibition in DLBCL results in profound disruption of progression through mitosis, potently triggering the spindle assembly checkpoint (SAC) and causing accumulation in G2 before onset of apoptosis. We explored whether GAK dependency is induced due to cell cycle deregulation. Analysis of gene-expression data in the Lenz cohort showed a striking correlation between high GAK expression and low expression of RB1, encoding the tumor suppressive master cell-cycle regulator retinoblastoma-associated protein (RB, p<0.0001). The ABC-DLBCL cell lines RIVA and U2932 completely lack RB expression and show high sensitivity and accumulation in G2 in response to SGC-GAK-1 treatment. Immunofluorescent confocal microscopy revealed profound disruption of mitotic spindles and shattering of chromosomes in these cells and, to a lesser extent, in RB-competent lines. SGC-GAK-1 is not optimized for use as a chemical probe in vivo, but target validation through shRNA knockdown in DLBCL patient-derived xenograft (PDX) models is underway and expected at time of presentation.

Conclusions: GAK is a novel kinase therapeutic target in DLBCL revealed by our target identification platform, idTRAX. GAK’s role in mitotic-spindle alignment represents a critical kinase dependency for DLBCL tumors, in particular those with cell-cycle deregulation due to loss of RB function. Our ongoing efforts to synthesize novel GAK inhibitors and reveal synergistic targets may reveal new therapeutic strategies for translation to clinical trials for DLBCL patients with unmet clinical needs.