Cross-species analysis of enhancer logic using deep learning.

Deciphering the genomic regulatory code of enhancers is a key challenge in biology because this code underlies cellular identity. A better understanding of how enhancers work will improve the interpretation of noncoding genome variation and empower the generation of cell type-specific drivers for gene therapy. Here, we explore the combination of deep learning and cross-species chromatin accessibility profiling to build explainable enhancer models. We apply this strategy to decipher the enhancer code in melanoma, a relevant case study owing to the presence of distinct melanoma cell states. We trained and validated a deep learning model, called DeepMEL, using chromatin accessibility data of 26 melanoma samples across six different species. We show the accuracy of DeepMEL predictions on the CAGI5 challenge, where it significantly outperforms existing models on the melanoma enhancer of IRF4 Next, we exploit DeepMEL to analyze enhancer architectures and identify accurate transcription factor binding sites for the core regulatory complexes in the two different melanoma states, with distinct roles for each transcription factor, in terms of nucleosome displacement or enhancer activation. Finally, DeepMEL identifies orthologous enhancers across distantly related species, where sequence alignment fails, and the model highlights specific nucleotide substitutions that underlie enhancer turnover. DeepMEL can be used from the Kipoi database to predict and optimize candidate enhancers and to prioritize enhancer mutations. In addition, our computational strategy can be applied to other cancer or normal cell types.

Leukemic transformation by the MLL-AF6 fusion oncogene requires the H3K79 methyltransferase Dot1l.

The t(6;11)(q27;q23) is a recurrent chromosomal rearrangement that encodes the MLLAF6 fusion oncoprotein and is observed in patients with diverse hematologic malignancies. The presence of the t(6;11)(q27;q23) has been linked to poor overall survival in patients with AML. In this study, we demonstrate that MLL-AF6 requires continued activity of the histone-methyltransferase DOT1L to maintain expression of the MLL-AF6-driven oncogenic gene-expression program. Using gene-expression analysis and genome-wide chromatin immunoprecipitation studies followed by next generation sequencing, we found that MLL-fusion target genes display markedly high levels of histone 3 at lysine 79 (H3K79) dimethylation in murine MLL-AF6 leukemias as well as in ML2, a human myelomonocytic leukemia cell line bearing the t(6;11)(q27;q23) translocation. Targeted disruption of Dot1l using a conditional knockout mouse model inhibited leukemogenesis mediated by the MLL-AF6 fusion oncogene. Moreover, both murine MLL-AF6-transformed cells as well as the human MLL-AF6-positive ML2 leukemia cell line displayed specific sensitivity to EPZ0004777, a recently described, selective, small-molecule inhibitor of Dot1l. Dot1l inhibition resulted in significantly decreased proliferation, decreased expression of MLL-AF6 target genes, and cell cycle arrest of MLL-AF6-transformed cells. These results indicate that patients bearing the t(6;11)(q27;q23) translocation may benefit from therapeutic agents targeting aberrant H3K79 methylation.

Standardized uptake value by positron emission tomography/computed tomography as a prognostic variable in metastatic breast cancer.

BACKGROUND: In this retrospective, single-institution study, the authors examine the maximum standardized uptake value (SUVmax) on positron emission tomography/computed tomography (PET/CT) images as a prognostic variable in patients with newly diagnosed metastatic breast cancer (MBC). METHODS: Patients with ≥1 metastatic lesion on PET/CT images that were obtained within 60 days of their MBC diagnosis between January 1, 2001 and December 31, 2008 were included. Patients were excluded if they had received chemotherapy ≤30 days before the PET/CT images were obtained. Electronic medical reports were reviewed to determine the SUVmax and overall survival. Because of intraindividual variation in the SUV by body site, separate analyses were conducted by metastatic site. Relationships between site-specific PET/CT variable tertiles and overall survival were assessed using Cox regression; hazard ratios for the highest tertile versus the lowest tertile were reported. RESULTS: In total, 253 patients were identified, and their median age was 57 years (range, 27-90 years). Of these, 152 patients (60%) died, and the median follow-up was 40 months. On univariate analysis, SUVmax tertile was strongly associated with overall survival in patients who had bone metastases (N = 141; hazard ratio, 3.13; 95% confidence interval, 1.79-5.48; P < .001). This effect was maintained on multivariate analysis (HR = 3.19; 95% confidence interval, 1.64-6.20, P = .002) after correcting for known prognostic variables. A greater risk of death was associated with SUVmax tertile in patients who had metastases to the liver (N = 46; hazard ratio, 2.07; 95% confidence interval, 0.90-4.76), lymph nodes (N = 149; hazard ratio, 1.1; 95% confidence interval, 0.69-1.88), and lung (N = 62; hazard ratio, 2.2; 95% confidence interval, 0.97-4.95), although these results were not significant (P = .18, P = .31, and P = .095, respectively). CONCLUSIONS: The current results indicate that PET/CT has value as a prognostic tool in patients with newly diagnosed MBC to bone.

Serum N-telopeptide and bone-specific alkaline phosphatase levels in patients with osteonecrosis of the jaw receiving bisphosphonates for bone metastases.

PURPOSE: Oversuppression of bone turnover can be a critical factor in the pathogenesis of osteonecrosis of the jaw (ONJ). We investigated N-telopeptide of type I collagen (NTX) and bone-specific alkaline phosphatase (BAP) as potential predictors of ONJ onset. PATIENTS AND METHODS: Patients with ONJ and available stored serum were identified retrospectively from the institutional databases. Four approximate points were examined: point of ONJ diagnosis and 12, 6, and 1 month before the diagnosis. NTX and BAP were measured using enzyme-linked immunosorbent assays and examined as possible predictors of ONJ. RESULTS: From March 1998 to September 2009, we identified 122 patients with ONJ. Of these, 56 (46%) had one or more serum samples available. Overall, 55 patients (98%) received bisphosphonates. Using the exact dates, no obvious patterns in either NTX or BAP were noted. Similarly, using the ordinal points, no evidence of suppression of NTX or BAP over time was seen. The consecutive median values were as follows: The median NTX values were 8.0 nmol/L (range 3.8 to 32.9) at 12 months before ONJ; 9.5 nmol/L (range 4.7 to 42.7) at 6 months; 9.5 nmol/L (range 4.5 to 24.6) at 1 month, and 10.4 nmol/L (range 4.4 to 32.5) at the ONJ diagnosis. The median BAP values were BAP 18.0 U/L (range 7.0 to 74) at 12 months before ONJ; 18.0 U/L (range 4.0 to 134) at 6 months; 14.0 U/L (range 4.0 to 132) at 1 month, and 18.0 U/L (range 0.7 to 375) at the ONJ diagnosis. Only 2 patients (4%) had NTX and 17 (30%) had BAP below the normal range at the ONJ diagnosis. CONCLUSIONS: In the present large retrospective study, no trends were seen in the NTX and BAP levels before the ONJ diagnosis.

Recurrent co-alteration of HDGF and SETDB1 on chromosome 1q drives cutaneous melanoma progression and poor prognosis.

A progressive increase in copy number variation (CNV) characterizes the natural history of cutaneous melanoma progression toward later disease stages, but our understanding of genetic drivers underlying chromosomal arm-level CNVs remains limited. To identify candidate progression drivers, we mined the TCGA SKCM dataset and identified HDGF as a recurrently amplified gene whose high mRNA expression correlates with poor patient survival. Using melanocyte-specific overexpression in the zebrafish BRAFV600E -driven MiniCoopR melanoma model, we show that HDGF accelerates melanoma development in vivo. Transcriptional analysis of HDGF compared to control EGFP tumors showed the activation of endothelial/angiogenic pathways. We validated this observation using an endothelial kdrl:mCherry reporter line which showed HDGF to increases tumor vasculature. HDGF is frequently co-altered with the established melanoma driver SETDB1. Both genes are located on chromosome 1q, and their co-amplification is observed in up to 13% of metastatic melanoma. TCGA patients with both genes amplified and/or overexpressed have a worse melanoma specific survival. We tested co-expression of HDGF and SETDB1 in the MiniCoopR model, which resulted in faster and more aggressive melanoma development than either gene individually. Our work identifies the co-amplification of HDGF and SETDB1 as a functional driver of melanoma progression and poor patient prognosis.

SATB2 induction of a neural crest mesenchyme-like program drives melanoma invasion and drug resistance.

Recent genomic and scRNA-seq analyses of melanoma demonstrated a lack of recurrent genetic drivers of metastasis, while identifying common transcriptional states correlating with invasion or drug resistance. To test whether transcriptional adaptation can drive melanoma progression, we made use of a zebrafish mitfa:BRAFV600E;tp53-/- model, in which malignant progression is characterized by minimal genetic evolution. We undertook an overexpression-screen of 80 epigenetic/transcriptional regulators and found neural crest-mesenchyme developmental regulator SATB2 to accelerate aggressive melanoma development. Its overexpression induces invadopodia formation and invasion in zebrafish tumors and human melanoma cell lines. SATB2 binds and activates neural crest-regulators, including pdgfab and snai2. The transcriptional program induced by SATB2 overlaps with known MITFlowAXLhigh and AQP1+NGFR1high drug-resistant states and functionally drives enhanced tumor propagation and resistance to Vemurafenib in vivo. In summary, we show that melanoma transcriptional rewiring by SATB2 to a neural crest mesenchyme-like program can drive invasion and drug resistance in autochthonous tumors.

Zebrafish patient avatars in cancer biology and precision cancer therapy.

In precision oncology, two major strategies are being pursued for predicting clinically relevant tumour behaviours, such as treatment response and emergence of drug resistance: inference based on genomic, transcriptomic, epigenomic and/or proteomic analysis of patient samples, and phenotypic assays in personalized cancer avatars. The latter approach has historically relied on in vivo mouse xenografts and in vitro organoids or 2D cell cultures. Recent progress in rapid combinatorial genetic modelling, the development of a genetically immunocompromised strain for xenotransplantation of human patient samples in adult zebrafish and the first clinical trial using xenotransplantation in zebrafish larvae for phenotypic testing of drug response bring this tiny vertebrate to the forefront of the precision medicine arena. In this Review, we discuss advances in transgenic and transplantation-based zebrafish cancer avatars, and how these models compare with and complement mouse xenografts and human organoids. We also outline the unique opportunities that these different models present for prediction studies and current challenges they face for future clinical deployment.

RNA helicase DDX21 mediates nucleotide stress responses in neural crest and melanoma cells.

The availability of nucleotides has a direct impact on transcription. The inhibition of dihydroorotate dehydrogenase (DHODH) with leflunomide impacts nucleotide pools by reducing pyrimidine levels. Leflunomide abrogates the effective transcription elongation of genes required for neural crest development and melanoma growth in vivo1. To define the mechanism of action, we undertook an in vivo chemical suppressor screen for restoration of neural crest after leflunomide treatment. Surprisingly, we found that alterations in progesterone and progesterone receptor (Pgr) signalling strongly suppressed leflunomide-mediated neural crest effects in zebrafish. In addition, progesterone bypasses the transcriptional elongation block resulting from Paf complex deficiency, rescuing neural crest defects in ctr9 morphant and paf1(alnz24) mutant embryos. Using proteomics, we found that Pgr binds the RNA helicase protein Ddx21. ddx21-deficient zebrafish show resistance to leflunomide-induced stress. At a molecular level, nucleotide depletion reduced the chromatin occupancy of DDX21 in human A375 melanoma cells. Nucleotide supplementation reversed the gene expression signature and DDX21 occupancy changes prompted by leflunomide. Together, our results show that DDX21 acts as a sensor and mediator of transcription during nucleotide stress.

From fish bowl to bedside: The power of zebrafish to unravel melanoma pathogenesis and discover new therapeutics.

Melanoma is the most aggressive and deadliest form of skin cancer. A detailed knowledge of the cellular, molecular, and genetic events underlying melanoma progression is highly relevant to diagnosis, prognosis and risk stratification, and the development of new therapies. In the last decade, zebrafish have emerged as a valuable model system for the study of melanoma. Pathway conservation, coupled with the availability of robust genetic, transgenic, and chemical tools, has made the zebrafish a powerful model for identifying novel disease genes, visualizing cancer initiation, interrogating tumor-microenvironment interactions, and discovering new therapeutics that regulate melanocyte and melanoma development. In this review, we will give an overview of these studies, and highlight recent advancements that will help unravel melanoma pathogenesis and impact human disease.

Efficient Transduction of Zebrafish Melanoma Cell Lines and Embryos Using Lentiviral Vectors.

The establishment of in vitro cultures of zebrafish cancer cells has expanded the potential of zebrafish as a disease model. However, the lack of effective methods for gene delivery and genetic manipulation has limited the experimental applications of these cultures. To overcome this barrier, we tested and optimized vesicular stomatitis virus glycoprotein (VSV-G) pseudotyped lentiviral and retroviral vector transduction protocols. We show that lentivirus successfully and efficiently transduced zebrafish melanoma cell lines in vitro, allowing antibiotic selection, fluorescence-based sorting, and in vivo allotransplantation. In addition, injection of concentrated lentiviral particles into embryos and tumors established the feasibility of in vivo gene delivery.

AF10 regulates progressive H3K79 methylation and HOX gene expression in diverse AML subtypes.

Homeotic (HOX) genes are dysregulated in multiple malignancies, including several AML subtypes. We demonstrate that H3K79 dimethylation (H3K79me2) is converted to monomethylation (H3K79me1) at HOX loci as hematopoietic cells mature, thus coinciding with a decrease in HOX gene expression. We show that H3K79 methyltransferase activity as well as H3K79me1-to-H3K79me2 conversion is regulated by the DOT1L cofactor AF10. AF10 inactivation reverses leukemia-associated epigenetic profiles, precludes abnormal HOXA gene expression, and impairs the transforming ability of MLL-AF9, MLL-AF6, and NUP98-NSD1 fusions-mechanistically distinct HOX-activating oncogenes. Furthermore, NUP98-NSD1-transformed cells are sensitive to small-molecule inhibition of DOT1L. Our findings demonstrate that pharmacological inhibition of the DOT1L/AF10 complex may provide therapeutic benefits in an array of malignancies with abnormal HOXA gene expression.

Prognostic value of quantitative fluorodeoxyglucose measurements in newly diagnosed metastatic breast cancer.

The aim of this study was to determine the prognostic value of quantitative fluorodeoxyglucose (FDG) measurements (maximum standardized uptake value [SUVmax ], metabolic tumor volume [MTV], and total lesion glycolysis [TLG]) in patients with newly diagnosed metastatic breast cancer (MBC). An IRB-approved retrospective review was performed of patients who underwent FDG positron emission tomography (PET)/computed tomography (CT) from 1/02 to 12/08 within 60 days of diagnosis MBC. Patients with FDG-avid lesions without receiving chemotherapy in the prior 30 days were included. Target lesions in bone, lymph node (LN), liver, and lung were analyzed for SUVmax , MTV, and TLG. Medical records were reviewed for patient characteristics and overall survival (OS). Cox regression was used to test associations between quantitative FDG measurements and OS. A total of 253 patients were identified with disease in bone (n = 150), LN (n = 162), liver (n = 48), and lung (n = 66) at the time of metastatic diagnosis. Higher SUVmax tertile was associated with worse OS in bone metastases (highest vs. lowest tertile hazard ratio [HR] = 3.1, P < 0.01), but not in LN, liver or lung (all P > 0.1). Higher MTV tertile was associated with worse OS in LN (HR = 2.4, P < 0.01) and liver (HR = 3.0, P = 0.02) metastases, but not in bone (P = 0.22) or lung (P = 0.14). Higher TLG tertile was associated with worse OS in bone (HR = 2.2, P = 0.02), LN (HR = 2.3, P < 0.01), and liver (HR = 4.9, P < 0.01) metastases, but not in lung (P = 0.19). We conclude measures of FDG avidity are prognostic biomarkers in newly diagnosed MBC. SUVmax and TLG were both predictors of survival in breast cancer patients with bone metastases. TLG may be a more informative biomarker of OS than SUVmax for patients with LN and liver metastases.