Citations

Abstract Purpose: Intratumoral immunosuppression mediated by myeloid-derived suppressor cells (MDSC) and tumor-associated macrophages (TAM) represents a potential mechanism of immune checkpoint inhibitor (ICI) resistance in solid tumors. By promoting TAM and MDSC infiltration, IL1β may drive adaptive and innate immune resistance in renal cell carcinoma (RCC) and in other tumor types. Experimental Design: Using the RENCA model of RCC, we evaluated clinically relevant combinations of anti-IL1β plus either anti-PD-1 or the multitargeted tyrosine kinase inhibitor (TKI), cabozantinib. We performed comprehensive immune profiling of established RENCA tumors via multiparameter flow cytometry, tumor cytokine profiling, and single-cell RNA sequencing (RNA-seq). Similar analyses were extended to the MC38 tumor model. Results: Analyses via multiparameter flow cytometry, tumor cytokine profiling, and single-cell RNA-seq showed that anti-IL1β reduces infiltration of polymorphonuclear MDSCs and TAMs. Combination treatment with anti-IL1β plus anti-PD-1 or cabozantinib showed increased antitumor activity that was associated with decreases in immunosuppressive MDSCs and increases in M1-like TAMs. Conclusions: Single-cell RNA-seq analyses show that IL1β blockade and ICI or TKI remodel the myeloid compartment through nonredundant, relatively T-cell–independent mechanisms. IL1β is an upstream mediator of adaptive myeloid resistance and represents a potential target for kidney cancer immunotherapy.

Abstract Splicing alterations are common in diseases such as cancer, where mutations in splicing factor genes are frequently responsible for aberrant splicing. Here we present an alternative mechanism for splicing regulation in T-cell acute lymphoblastic leukemia (T-ALL) that involves posttranslational stabilization of the splicing machinery via deubiquitination. We demonstrate there are extensive exon skipping changes in disease, affecting proteasomal subunits, cell-cycle regulators, and the RNA machinery. We present that the serine/arginine-rich splicing factors (SRSF), controlling exon skipping, are critical for leukemia cell survival. The ubiquitin-specific peptidase 7 (USP7) regulates SRSF6 protein levels via active deubiquitination, and USP7 inhibition alters the exon skipping pattern and blocks T-ALL growth. The splicing inhibitor H3B-8800 affects splicing of proteasomal transcripts and proteasome activity and acts synergistically with proteasome inhibitors in inhibiting T-ALL growth. Our study provides the proof-of-principle for regulation of splicing factors via deubiquitination and suggests new therapeutic modalities in T-ALL. Significance: Our study provides a new proof-of-principle for posttranslational regulation of splicing factors independently of mutations in aggressive T-cell leukemia. It further suggests a new drug combination of splicing and proteasomal inhibitors, a concept that might apply to other diseases with or without mutations affecting the splicing machinery. This article is highlighted in the In This Issue feature, p. 1241

Abstract Diffuse Large B Cell Lymphoma (DLBCL) is the most common form of B-cell Non-Hodgkin Lymphoma (NHL), representing a third of all new cases. DLBCL is further sub-divided into various molecular sub-types based on gene expression and co-occurring genetic alterations. Gene expression based- subtypes include the germinal center B-cell (GCB) and the activated B-cell type (ABC) subtypes, with the ABC sub-type having a poorer prognosis than the GCB sub-type. Interestingly, 30-40% of all DLBCL patients harbor mutations in key epigenetic regulators, EZH2, KMT2D, CREBBP, EP300, and mutations in histone proteins themselves. Through an unbiased cell-based phenotypic screen, we discovered that inhibition of lysine demethylases, specifically KDM4C and KDM4A, represents a vulnerability across 15 different DLBCL cell lines including germinal center B-cell (GCB) and activated B-cell (ABC) type lines, with a GI50 value between 75nM to ~200nM, while sparing leukemia lines. Consistently, treatment of xenograft-based animal models of DLBCL with a low dose of KDM4A/KDM4C inhibitor delivered intra-peritoneally three times a week, results in a drastic reduction of tumor burden. Both KDM4A and KDM4C catalyze the removal of histone H3 K9 di- and tri-methylation (H3K9me2/3) and H3K36 di- and tri-methylation (H3K36me2/3). H3K9me2/3 is associated with promoter and enhancer repression, while H3K36me2/3 is present in gene bodies during transcription but also functions as a chromatin repressor Consistently, we have identified key enhancers, including those associated with IKZF1, that are "decommissioned" after inhibition of KDM4A/KDM4C. Repressed enhancer activity, through loss of H3K27 acetylation and H3K4me1, and gain of H3K9me2/3, results in a rapid transcriptional downregulation of IKZF1 and its partners including IKZF3. Given the role of IKZF1 in the transcriptional regulation of key B Cell Receptor (BCR) signaling components, we show that KDM4A/KDM4C inhibition leads to a downregulation of SYK, a proximal BCR-signaling component, which likely precedes DLBCL cell apoptosis. In addition, we observed an activation of extra-lineage transcription factors such as CEBPA and CEBPB, which are normally repressed by IKZF1 in the lymphoid lineage. A concomitant downregulation of the B-cell gene expression program and an upregulation of the myeloid (CD14+ monocytic) gene expression program is also observed, implying a "trans-differentiation" of DLBCL cells into the monocyte lineage. This lineage-switch correlates with an increased population of CD14+ expressing cells. Finally, using DLBCL patient data sets, we can show that over-expression of either KDM4A or KDM4C is associated with poor prognosis in DLBCL patients. In summary, we have discovered that KDM4A/KDM4C inhibition results in an increase of repressive histone modifications at several intra-genic "enhancers" of genes that are responsible for the survival and proliferation of DLBCL cells. The elucidation of this unique epigenetic mechanism provides a strong rationale for the development of novel targeted therapies against both multiple subtypes of DLBCL. Figure. Figure. Disclosures Shipp: Bayer: Research Funding; AstraZeneca: Honoraria; Merck: Research Funding; Bristol-Myers Squibb: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding.

Abstract Neuroblastoma is an embryonal tumor arising in the peripheral sympathetic nervous system (PSNS), and often presents with widespread metastasis at diagnosis. We recently demonstrated that overexpression of LMO1 can synergize with MYCN to promote tumor metastasis in our zebrafish model system. Here, we identify a “metastasis suppressor” gene – growth arrest specific 7 (GAS7) – whose loss of function or reduced expression can promote dissemination and migration of neuroblastoma cells with MYCN overexpression, both in vitro and in vivo. GAS7 resides on chromosome 17p13.1, a region that is deleted in high-risk neuroblastomas with higher frequency in MYCN-amplified cases. Expression levels of GAS7 are also significantly downregulated in MYCN-amplified or otherwise MYCN-overexpressing human neuroblastoma cells, as well as MYCN-overexpressing zebrafish PSNS cells, suggesting an inverse relationship between the oncogenic activities of MYCN and GAS7downregulation. To determine whether loss of function or reduced expression of GAS7 could contribute to neuroblastoma pathogenesis in the context of MYCN overexpression, we developed a gas7 knockout zebrafish line using a transcription activator-like effector nucleases (TALEN) method. Loss of function of gas7 promoted hematogenous metastasis of MYCN-overexpressing tumor cells to tissues and organs that paralleled the common sites of metastasis in neuroblastoma patients. Further RNA-sequencing and qRT-PCR analyses showed that representative signature genes encoding proteins involved in tumor cell-cell or cell-extracellular matrix interactions were significantly downregulated in fish tumors with loss of function of gas7, consistent with our electron microscopic finding in the fish model showing a loss of contact among tumor cells in fish lacking functional gas7, suggesting that GAS7 may physically act as a metastasis suppressor. Thus, in high-risk neuroblastoma with MYCNamplification or overexpression, we propose that deletion or reduced expression of GAS7 cooperates with MYCN to promote neuroblastoma cell dissemination. Although the precise mechanism of this interaction remains unclear, we suggest that it relies, in part, on the downregulation of pivotal genes involved in tumor cell-cell or cell-extracellular matrix contacts. Citation Format: Zhiwei Dong, Gonzalo Lepez Garcia, Cheng Zhang, Erin N. Dankert, Jo Lynne Harenza, Xiaonan Hou, Hu Li, John M. Maris, Shizhen (Jane) Zhu. Loss of function of GAS7 promotes neuroblastoma metastasis in the context of MYCN overexpression [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr LB-127.

e14237 Background: Targeting myeloid-derived suppressor cells (MDSCs), key mediators of intrinsic and adaptive immune resistance, remains a significant challenge in kidney cancer immunotherapy. One potential driver of MDSC mediated tumor immunosuppression is the cytokine interleukin-1 beta (IL-1β). Although large cancer prevention studies show that targeting IL1 decreases cancer risk, the exact mechanism of anti-IL-1 mediated anti-tumor efficacy is not yet defined. To understand the immunogenic effects of IL-1β blockade and effects on MDSCs, we used the RENCA model of RCC. Methods: We inoculated BALB/c mice with 1x106 RENCA cells. On Day +10 we treated mice (n = 8/group) with vehicle, αPD1, αIL1β, cabozantinib, combination αPD1/αIL1β, or cabozantinib/αIL1β. We harvested day +18 tumors to quantify tumor size and immune cells by flow cytometry. We then isolated immune cells for differential gene expression profiling by RNAseq. Results: Treatment with αIL1β significantly decreased tumor size compared to vehicle or αPD1 monotherapy. Combining αPD1/αIL1β or cabozantinib/αIL1β potentiated the anti-tumor effects of αPD1 or cabozantinib monotherapy respectively. Anti-IL1β therapy decreased polymorphonuclear (PMN)-MDSC infiltration (αIL1β: 0.76% +/- 0.21 vs. vehicle: 1.89% +/- 0.37, P = 0.014). Relative to PD1 monotherapy, combining αPD1 and αIL1β decreased PMN-MDSCs (αPD1: 2.14% +/- 0.56 vs αPD1/αIL1β: 0.91% +/- 0.15, P = 0.033) and increased M1-like tumor associated macrophages (TAM) (αPD1: 0.798% +/- 0.21 vs. αPD1/αIL1β: 12.04% +/- 2.9, P = 0.0008). Gene expression profiling from sorted immune cells from αIL1β treated tumors revealed multiple differentially regulated cytokines including interleukin-6, IFN-γ, and IL-10. Conclusions: Blockade of IL-1β depletes intratumoral PMN-MDSCs and increases M1-like TAM with minimal effects observed on the T cell compartment. Treatment with cabozantinib or αPD1 in combination with αIL1β led to sustained control of tumor growth. IL-1β represents a promising target for kidney cancer immunotherapy. Based on the preclinical data, a clinical trial of neoadjuvant αPD1 and αIL1β therapy is planned for patients with high risk RCC.

Abstract With the emergence of genome editing technologies and synthetic biology, it is now possible to engineer genetic circuits driving a cell's phenotypic response to a stressor. However, capturing a continuous response, rather than simply a binary ‘on’ or ‘off’ response, remains a bioengineering challenge. No tools currently exist to identify gene candidates responsible for predetermining and fine-tuning cell response phenotypes. To address this gap, we devised a novel Regulostat Inferelator (RSI) algorithm to decipher intrinsic molecular devices or networks that predetermine cellular phenotypic responses. The RSI algorithm is designed to extract gene expression patterns from basal transcriptomic data in order to identify ‘regulostat’ constituent gene pairs, which exhibit rheostat-like mode-of-cooperation capable of fine-tuning cellular response. Our proof-of-concept study provides computational evidence for the existence of regulostats and that these networks predetermine cellular response prior to exposure to a stressor or drug. In addition, our work, for the first time, provides evidence of context-specific, drug–regulostat interactions in predetermining drug response phenotypes in cancer cells. Given RSI-inferred regulostat networks offer insights for prioritizing gene candidates capable of rendering a resistant phenotype sensitive to a given drug, we envision that this tool will be of great value in bioengineering and medicine.

Abstract 5-Fluorouracil (5-FU) is a cornerstone therapy for the treatment of several aggressive tumors. However, nearly 1/3 of patients receiving 5-FU experience severe (grade 3+) adverse 5-FU toxicity. Genetic variations in dihydropyrimidine dehydrogenase (DPD; DPYD gene) are major contributors of 5-FU toxicity. With continued progress in next-generation sequencing technologies, the number of reported genetic variants is substantially increasing; to date, there are more than 450 reported missense variants in the DPYD gene. Only a handful of those variants, including *2A, I560S, D949V, and rs75017182, have been established as predictors of toxicity. Because most of the remaining variants are unstudied, mainly due to their rare occurrence and unequal racial distribution, it is challenging to conduct statistically powered clinical association studies. Therefore, in vitro and computational methods to identify potential toxicity-associated variants have major appeal. The primary goal of the current study was to develop a highly accurate gene-specific in silico classifier to predict deleterious DPYD variants. For this purpose, using a recombinant cellular model, we characterized 69 missense DPYD variants reported in online databases. We then utilized DPD activity data from the 69 variants in combination with data from 87 previously characterized variants (see Offer, SM et al. Cancer Research 2013 & 2014) to develop the DPYD variant classifier model (DPYD-Varifier). DPD protein-specific biochemical and structural features and amino acid conservation scores were integrated into the model's random forest classifier-based algorithm. DPYD-Varifier showed improved accuracy (85% accurate) compared to other in silico tools. Additionally, the model provided mechanistic insights into alterations of DPD function conferred by DPYD variants. In general, deleterious variants were located in closer proximity to important binding/coordinating sites of the DPD protein. Additionally, DPYD-Varifier allowed us to estimate the cumulative allele frequency of all reported deleterious variants in various geographical/racial groups, which can substantially add to our knowledge of DPD deficiency and 5-FU toxicity observed in these populations. In conclusion, the integration of in vitro and accurate computational data from our study permitted classification of relevant DPYD variants, which could be useful to more effectively individualize 5-FU therapy in the future. Citation Format: Shikshya Shrestha, Cheng Zhang, Calvin R. Jerde, Hu Li, Steven M. Offer, Robert B. Diasio. DPYD-Varifier, a computational model to identify 5-FU toxicity-associated DPYD variants [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 3894.

Deleterious variants in dihydropyrimidine dehydrogenase (DPD, DPYD gene) can be highly predictive of clinical toxicity to the widely prescribed chemotherapeutic 5‐fluorouracil (5‐FU). However, there are very limited data pertaining to the functional consequences of the >450 reported no‐synonymous DPYD variants. We developed a DPYD‐specific variant classifier (DPYD‐Varifier) using machine learning and in vitro functional data for 156 missense DPYD variants. The developed model showed 85% accuracy and outperformed other in silico prediction tools. An examination of feature importance within the model provided additional insight into functional aspects of the DPD protein relevant to 5‐FU toxicity. In the absence of clinical data for unstudied variants, prediction tools like DPYD‐Varifier have great potential to individualize medicine and improve the clinical decision‐making process.