DeepCOP: deep learning-based approach to predict gene regulating effects of small molecules.

MOTIVATION: Recent advances in the areas of bioinformatics and chemogenomics are poised to accelerate the discovery of small molecule regulators of cell development. Combining large genomics and molecular data sources with powerful deep learning techniques has the potential to revolutionize predictive biology. In this study, we present Deep gene COmpound Profiler (DeepCOP), a deep learning based model that can predict gene regulating effects of low-molecular weight compounds. This model can be used for direct identification of a drug candidate causing a desired gene expression response, without utilizing any information on its interactions with protein target(s). RESULTS: In this study, we successfully combined molecular fingerprint descriptors and gene descriptors (derived from gene ontology terms) to train deep neural networks that predict differential gene regulation endpoints collected in LINCS database. We achieved 10-fold cross-validation RAUC scores of and above 0.80, as well as enrichment factors of >5. We validated our models using an external RNA-Seq dataset generated in-house that described the effect of three potent antiandrogens (with different modes of action) on gene expression in LNCaP prostate cancer cell line. The results of this pilot study demonstrate that deep learning models can effectively synergize molecular and genomic descriptors and can be used to screen for novel drug candidates with the desired effect on gene expression. We anticipate that such models can find a broad use in developing novel cancer therapeutics and can facilitate precision oncology efforts. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Differential expression of full-length and NH2 terminally truncated FAM134B isoforms in normal physiology and cancer.

Selective autophagy of the endoplasmic reticulum (ER), namely ER-phagy, is mediated by ER-localized receptors, which are recognized and sequestered by GABARAP/LC3B-decorated phagophores and transferred to lysosomes for degradation. Being one such receptor, FAM134B plays critical roles in cellular processes such as protein quality control and neuronal survival. FAM134B has also been associated with different cancers, although its exact role remains elusive. We report here that the FAM134B gene encodes not one but at least two different protein isoforms: the full-length and the NH2 terminally truncated forms. Their relative expression shows extreme variation, both within normal tissues and among cancer types. Expression of full-length FAM134B is restricted to the brain, testis, spleen, and prostate. In contrast, NH2 terminally truncated FAM134B is dominant in the heart, skeletal muscle, kidney, pancreas, and liver. We compared wild-type and knockout mice to study the role of the Fam134b gene in starvation. NH2 terminally truncated FAM134B-2 was induced in the liver, skeletal muscle, and heart but not in the pancreas and stomach following starvation. Upon starvation, Fam134b-/- mice differed from wild-type mice by less weight loss and less hyperaminoacidemic and hypocalcemic response but increased levels of serum albumin, total serum proteins, and α-amylase. Interestingly, either NH2 terminally truncated FAM134B or both isoforms were downregulated in liver, lung, and colon cancers. In contrast, upregulation was observed in stomach and chromophobe kidney cancers.NEW & NOTEWORTHY We reported tissues expressing FAM134B-2 such as the kidney, muscle, heart, and pancreas, some of which exhibit stimulated expression upon nutrient starvation. We also demonstrated the effect of Fam134b deletion during ad libitum and starvation conditions. Resistance to weight loss and hypocalcemia, accompanied by an increase in serum albumin and α-amylase levels, indicate critical roles of Fam134b in physiology. Furthermore, the differential expression of FAM134B isoforms was shown to be significantly dysregulated in human cancers.

Cryptochrome deletion in p53 mutant mice enhances apoptotic and anti-tumorigenic responses to UV damage at the transcriptome level.

Previous studies have demonstrated that deletion of cryptochrome (Cry) genes protects p53-/- mutant mice from the early onset of cancer and extends their median life-span by about 1.5-fold. Subsequent in vitro studies had revealed that deletion of Crys enhances apoptosis in response to UV damage through activation of p73 and inactivation of GSK3ß. However, it was not known at the transcriptome-wide level how deletion of Crys delays the onset of cancer in p53-/- mutant mice. In this study, the RNA-seq approach was taken to uncover the differentially expressed genes (DEGs) and pathways following UV-induced DNA damage in p53-/- and p53-/-Cry1-/-Cry2-/- mouse skin fibroblasts. Gene set enrichment analysis with the DEGs demonstrated enrichment in immune surveillance-associated genes regulated by IFN-γ and genes involved in TNFα signaling via NF-κB. Furthermore, protein network analysis enabled identification of DEGs p21, Sirt1, and Jun as key players, along with their interacting partners. It was also observed that the DEGs contained a high ratio of non-coding transcripts. Collectively, the present study suggests new genes in NF-κB regulation and IFN-γ response, as well as non-coding RNAs, may contribute to delaying the onset of cancer in p53-/-Cry1-/-Cry2-/- mice and increasing the life-span of these animals compared to p53-/- mice.

BRD9-containing non-canonical BAF complex maintains somatic cell transcriptome and acts as a barrier to human reprogramming.

Epigenetic reprogramming to pluripotency requires extensive remodeling of chromatin landscapes to silence existing cell-type-specific genes and activate pluripotency genes. ATP-dependent chromatin remodeling complexes are important regulators of chromatin structure and gene expression; however, the role of recently identified Bromodomain-containing protein 9 (BRD9) and the associated non-canonical BRG1-associated factors (ncBAF) complex in reprogramming remains unknown. Here, we show that genetic or chemical inhibition of BRD9, as well as ncBAF complex subunit GLTSCR1, but not the closely related BRD7, increase human somatic cell reprogramming efficiency and can replace KLF4 and c-MYC. We find that BRD9 is dispensable for human induced pluripotent stem cells under primed but not under naive conditions. Mechanistically, BRD9 inhibition downregulates fibroblast-related genes and decreases chromatin accessibility at somatic enhancers. BRD9 maintains the expression of transcriptional regulators MN1 and ZBTB38, both of which impede reprogramming. Collectively, these results establish BRD9 as an important safeguarding factor for somatic cell identity whose inhibition lowers chromatin-based barriers to reprogramming.

EPIKOL, a chromatin-focused CRISPR/Cas9-based screening platform, to identify cancer-specific epigenetic vulnerabilities.

Dysregulation of the epigenome due to alterations in chromatin modifier proteins commonly contribute to malignant transformation. To interrogate the roles of epigenetic modifiers in cancer cells, we generated an epigenome-wide CRISPR-Cas9 knockout library (EPIKOL) that targets a wide-range of epigenetic modifiers and their cofactors. We conducted eight screens in two different cancer types and showed that EPIKOL performs with high efficiency in terms of sgRNA distribution and depletion of essential genes. We discovered novel epigenetic modifiers that regulate triple-negative breast cancer (TNBC) and prostate cancer cell fitness. We confirmed the growth-regulatory functions of individual candidates, including SS18L2 and members of the NSL complex (KANSL2, KANSL3, KAT8) in TNBC cells. Overall, we show that EPIKOL, a focused sgRNA library targeting ~800 genes, can reveal epigenetic modifiers that are essential for cancer cell fitness under in vitro and in vivo conditions and enable the identification of novel anti-cancer targets. Due to its comprehensive epigenome-wide targets and relatively high number of sgRNAs per gene, EPIKOL will facilitate studies examining functional roles of epigenetic modifiers in a wide range of contexts, such as screens in primary cells, patient-derived xenografts as well as in vivo models.

Development of an Androgen Receptor Inhibitor Targeting the N-Terminal Domain of Androgen Receptor for Treatment of Castration Resistant Prostate Cancer.

Prostate cancer patients undergoing androgen deprivation therapy almost invariably develop castration-resistant prostate cancer. Resistance can occur when mutations in the androgen receptor (AR) render anti-androgen drugs ineffective or through the expression of constitutively active splice variants lacking the androgen binding domain entirely (e.g., ARV7). In this study, we are reporting the discovery of a novel AR-NTD covalent inhibitor 1-chloro-3-[(5-([(2S)-3-chloro-2-hydroxypropyl]amino)naphthalen-1-yl)amino]propan-2-ol (VPC-220010) targeting the AR-N-terminal Domain (AR-NTD). VPC-220010 inhibits AR-mediated transcription of full length and truncated variant ARV7, downregulates AR response genes, and selectively reduces the growth of both full-length AR- and truncated AR-dependent prostate cancer cell lines. We show that VPC-220010 disrupts interactions between AR and known coactivators and coregulatory proteins, such as CHD4, FOXA1, ZMIZ1, and several SWI/SNF complex proteins. Taken together, our data suggest that VPC-220010 is a promising small molecule that can be further optimized into effective AR-NTD inhibitor for the treatment of CRPC.

Development of 2-(5,6,7-Trifluoro-1H-Indol-3-yl)-quinoline-5-carboxamide as a Potent, Selective, and Orally Available Inhibitor of Human Androgen Receptor Targeting Its Binding Function-3 for the Treatment of Castration-Resistant Prostate Cancer.

Prostate cancer (PCa) patients undergoing androgen deprivation therapy almost invariably develop castration-resistant prostate cancer (CRPC). Targeting the androgen receptor (AR) Binding Function-3 (BF3) site offers a promising option to treat CRPC. However, BF3 inhibitors have been limited by poor potency or inadequate metabolic stability. Through extensive medicinal chemistry, molecular modeling, and biochemistry, we identified 2-(5,6,7-trifluoro-1H-Indol-3-yl)-quinoline-5-carboxamide (VPC-13789), a potent AR BF3 antagonist with markedly improved pharmacokinetic properties. We demonstrate that VPC-13789 suppresses AR-mediated transcription, chromatin binding, and recruitment of coregulatory proteins. This novel AR antagonist selectively reduces the growth of both androgen-dependent and enzalutamide-resistant PCa cell lines. Having demonstrated in vitro efficacy, we developed an orally bioavailable prodrug that reduced PSA production and tumor volume in animal models of CRPC with no observed toxicity. VPC-13789 is a potent, selective, and orally bioavailable antiandrogen with a distinct mode of action that has a potential as novel CRPC therapeutics.

Beyond the heterodimer model for mineralocorticoid and glucocorticoid receptor interactions in nuclei and at DNA.

Glucocorticoid (GR) and mineralocorticoid receptors (MR) are believed to classically bind DNA as homodimers or MR-GR heterodimers to influence gene regulation in response to pulsatile basal or stress-evoked glucocorticoid secretion. Pulsed corticosterone presentation reveals MR and GR co-occupy DNA only at the peaks of glucocorticoid oscillations, allowing interaction. GR DNA occupancy was pulsatile, while MR DNA occupancy was prolonged through the inter-pulse interval. In mouse mammary 3617 cells MR-GR interacted in the nucleus and at a chromatin-associated DNA binding site. Interactions occurred irrespective of ligand type and receptors formed complexes of higher order than heterodimers. We also detected MR-GR interactions ex-vivo in rat hippocampus. An expanded range of MR-GR interactions predicts structural allostery allowing a variety of transcriptional outcomes and is applicable to the multiple tissue types that co-express both receptors in the same cells whether activated by the same or different hormones.

Taming Oncogenic Signaling at Protein Interfaces: Challenges and Opportunities.

Many key cellular events determining the thin line between healthy and oncogenic behavior rely on the proper functioning of protein-protein interactions (PPIs). Alterations that affect the affinity of a protein-protein binding site may destabilize a desired healthy interaction, or stabilize an oncogenic interaction. The understanding that there are a few key hot-spot residues that are mainly responsible for the binding energy of an interaction greatly widened the prospects of targeting oncogenic protein-protein interfaces enabling the use of small ligands in addition to biological molecules such as peptides and antibodies. Taming oncogenic signaling requires a deep understanding of protein interactions and their networks. Traditional representation of PPIs in signaling pathways as nodes and edges falls short of expressing interaction specific modulation of signals. Structural networks, deciphering which sites on a protein structure are responsible for each of the many interactions it may carry out, help understanding specific oncogenic mutations on signaling. We describe the key features of PPIs and their targeting, together with the advantages of structural networks, and provide four case studies demonstrating different opportunities for the aim of modulating oncogenic interactions.