Deep learning identified glioblastoma subtypes based on internal genomic expression ranks.

BACKGROUND: Glioblastoma (GBM) can be divided into subtypes according to their genomic features, including Proneural (PN), Neural (NE), Classical (CL) and Mesenchymal (ME). However, it is a difficult task to unify various genomic expression profiles which were standardized with various procedures from different studies and to manually classify a given GBM sample into a subtype. METHODS: An algorithm was developed to unify the genomic profiles of GBM samples into a standardized normal distribution (SND), based on their internal expression ranks. Deep neural networks (DNN) and convolutional DNN (CDNN) models were trained on original and SND data. In addition, expanded SND data by combining various The Cancer Genome Atlas (TCGA) datasets were used to improve the robustness and generalization capacity of the CDNN models. RESULTS: The SND data kept unimodal distribution similar to their original data, and also kept the internal expression ranks of all genes for each sample. CDNN models trained on the SND data showed significantly higher accuracy compared to DNN and CDNN models trained on primary expression data. Interestingly, the CDNN models classified the NE subtype with the lowest accuracy in the GBM datasets, expanded datasets and in IDH wide type GBMs, consistent with the recent studies that NE subtype should be excluded. Furthermore, the CDNN models also recognized independent GBM datasets, even with small set of genomic expressions. CONCLUSIONS: The GBM expression profiles can be transformed into unified SND data, which can be used to train CDNN models with high accuracy and generalization capacity. These models suggested NE subtype may be not compatible with the 4 subtypes classification system.

Hypoxia Regulated Gene Network in Glioblastoma Has Special Algebraic Topology Structures and Revealed Communications Involving Warburg Effect and Immune Regulation.

Hypoxia regulated genes (HRGs) formed a complex molecular interaction network (MINW), contributing to many aspects of glioblastoma (GBM) tumor biology. However, little is known about the intrinsic structures of the HRGs-MINW, mainly due to a lack of analysis tools to decipher MINWs. By introducing general hyper-geometric distribution, we obtained a statistically reliable gene set of HRGs (SR-HRGs) from several datasets. Next, MINWs were reconstructed from several independent GBM expression datasets. Algebraic topological analysis was performed to quantitatively analyze the amount of equivalence classes of cycles in various dimensions by calculating the Betti numbers. Persistent homology analysis of a filtration of growing networks was further performed to examine robust topological structures in the network by investigating the Betti curves, life length of the cycles. Random networks with the same number of node and edge and degree distribution were produced as controls. As a result, GBM-HRGs-MINWs reconstructed from different datasets exhibited great consistent Betti curves to each other, which were significantly different from that of random networks. Furthermore, HRGs-MINWs reconstructed from normal brain expression datasets exhibited topological structures significantly different from that of GBM-HRGs-MINWs. Analysis of cycles in GBM-HRGs-MINWs revealed genes that had clinical implications, and key parts of the cycles were also identified in reconstructed protein-protein interaction networks. In addition, the cycles are composed by genes involved in the Warburg effect, immune regulation, and angiogenesis. In summary, GBM-HRGs-MINWs contained abundant molecular interacting cycles in different dimensions, which are composed by genes involved in multiple programs essential for the tumorigenesis of GBM, revealing novel interaction diagrams in GBM and providing novel potential therapeutic targets.

Advances in the delivery of antisense oligonucleotides for combating bacterial infectious diseases.

Discovery and development of new antibacterial drugs against multidrug resistant bacterial strains have become more and more urgent. Antisense oligonucleotides (ASOs) show immense potential to control the spread of resistant microbes due to its high specificity of action, little risk to human gene expression, and easy design and synthesis to target any possible gene. However, efficient delivery of ASOs to their action sites with enough concentration remains a major obstacle, which greatly hampers their clinical application. In this study, we reviewed current progress on delivery strategies of ASOs into bacteria, focused on various non-virus gene vectors, including cell penetrating peptides, lipid nanoparticles, bolaamphiphile-based nanoparticles, DNA nanostructures and Vitamin B12. The current review provided comprehensive understanding and novel perspective for the future application of ASOs in combating bacterial infections.

A potent and selective antimicrobial poly(amidoamine) dendrimer conjugate with LED209 targeting QseC receptor to inhibit the virulence genes of gram negative bacteria.

The pandemic of multidrug-resistant Gram negative bacteria (GNB) is a worldwide healthcare concern, and very few antibiotics are being explored to match the clinical challenge. Recently, amino-terminated poly(amidoamine) (PAMAM) dendrimers have shown potential to function as broad antimicrobial agents. However, PAMAM displays a generation dependent cytotoxicity to mammalian cells and low selectivity on bacterial cells, which limits PAMAM to be developed as an antibacterial agent for systemic administration. We conjugated G3 PAMAM with LED209, a specific inhibitor of quorum sensor QseC of GNB, to generate a multifunctional agent PAMAM-LED209. Intriguingly, PAMAM-LED209 showed higher selectivity on GNB and lower cytotoxicity to mammalian cells, yet remained strong antibacterial activity. PAMAM-LED209 also inhibited virulence gene expression of GNB, and did not induce antibiotic-resistance. The present work firstly demonstrated that PAMAM-LED209 conjugate had a highly selective anti-GNB activity and low cytotoxicity, which offered a feasible strategy for combating multidrug-resistant GNB infections. FROM THE CLINICAL EDITOR: This research team demonstrated that a novel PAMAM-LED209 conjugate had highly selective activity against Gram-negative bacteria, coupled with low cytotoxicity, offering a potential strategy for combating multidrug-resistant infections.

LGR5 is a proneural factor and is regulated by OLIG2 in glioma stem-like cells.

The biological functional roles of LGR5 (leucine-rich repeat containing G protein-coupled receptor 5, also known as GPR49), a novel potential marker for stem-like cells in glioblastoma (GSCs), is poorly acknowledged. Here, we demonstrated that LGR5 was detected in glioblastoma tissues and GSCs. Bioinformatics analysis revealed that LGR5 is closely related to neurogenesis and neuronal functions, and preferentially expressed in Proneural subtype of GBMs. Furthermore, LGR5 is regulated by Proneural factor OLIG2, which is important for both neurogenesis and GSC maintenance. Biological experiments in GSC cells validated the bioinformatics analysis results and revealed that LGR5 regulated the tumor sphere formation capacity, an important stem cell property for GSCs. Therefore, LGR5 expression may be functionally correlated with the neurogenic competence, and be regulated by OLIG2 in GSCs.

CEBPD is a master transcriptional factor for hypoxia regulated proteins in glioblastoma and augments hypoxia induced invasion through extracellular matrix-integrin mediated EGFR/PI3K pathway.

Hypoxia contributes to the initiation and progression of glioblastoma by regulating a cohort of genes called hypoxia-regulated genes (HRGs) which form a complex molecular interacting network (HRG-MINW). Transcription factors (TFs) often play central roles for MINW. The key TFs for hypoxia induced reactions were explored using proteomic analysis to identify a set of hypoxia-regulated proteins (HRPs) in GBM cells. Next, systematic TF analysis identified CEBPD as a top TF that regulates the greatest number of HRPs and HRGs. Clinical sample and public database analysis revealed that CEBPD is significantly up-regulated in GBM, high levels of CEBPD predict poor prognosis. In addition, CEBPD is highly expressed in hypoxic condition both in GBM tissue and cell lines. For molecular mechanisms, HIF1α and HIF2α can activate the CEBPD promotor. In vitro and in vivo experiments demonstrated that CEBPD knockdown impaired the invasion and growth capacity of GBM cells, especially in hypoxia condition. Next, proteomic analysis identified that CEBPD target proteins are mainly involved in the EGFR/PI3K pathway and extracellular matrix (ECM) functions. WB assays revealed that CEBPD significantly positively regulated EGFR/PI3K pathway. Chromatin immunoprecipitation (ChIP) qPCR/Seq analysis and Luciferase reporter assay demonstrated that CEBPD binds and activates the promotor of a key ECM protein FN1 (fibronectin). In addition, the interactions of FN1 and its integrin receptors are necessary for CEBPD-induced EGFR/PI3K activation by promoting EGFR phosphorylation. Furthermore, GBM sample analysis in the database corroborated that CEBPD is positively correlated with the pathway activities of EGFR/PI3K and HIF1α, especially in highly hypoxic samples. At last, HRPs are also enriched in ECM proteins, indicating that ECM activities are important components of hypoxia induced responses in GBM. In conclusion, CEPBD plays important regulatory roles in the GBM HRG-MINW as a key TF, which activates the EGFR/PI3K pathway through ECM, especially FN1, mediated EGFR phosphorylation.

Promoting effects of chemical permeation enhancers on insulin permeation across TR146 cell model of buccal epithelium in vitro.

To find potential enhancers for facilitating the buccal delivery of insulin, a TR146 cell-culture model of buccal epithelium, cultured on commercially available insert plates, was used to evaluate the permeability-enhancing effects of several traditional and new types of chemical enhancers, including N-acetyl-L-cysteine (NAC), sodium deoxycholate (SDC), sodium nitroprusside (SNP), reduced glutathione (GSH), glutamine (Gln), chitosan (CS), L-arginine (Arg), 1-dodecylazacycloheptan-2-one (Azone), and soybean lecithin (SPC) (50 and 10 µg/mL respectively). Permeability studies were performed to determine the enhancing effects of these compounds on insulin permeation across the cell-culture model. The enhancing effects of the enhancers were assessed by calculating the apparent permeability coefficients and enhancement ratio. Cytotoxicity of the permeation enhancers at different concentrations was investigated by using the methylthiazolydiphenyl-tetrazolium bromide (MTT) assay. Results showed that 50 µg/mL of NAC, SDC, GSH, CS, Arg, Azone, SPC, SNP, and 10 µg/mL of SNP had a significant enhancing effect on promoting the transport of insulin across the TR146 cell model. MTT assays showed that 50 µg/mL of Gln, Azone, SDC, SNP, Arg, 10 µg/mL SDC, and Arg had obvious toxic effects on TR146 cells. Therefore, NAC, GSH, CS, SPC, and SNP appear to be safe, effective permeability enhancers that promote the transport of insulin across the TR146 cell-culture model of buccal epithelium and may be potential enhancers for buccal delivery of insulin with both low toxicity and high efficiency.

Protein-protein interaction network of E. coli K-12 has significant high-dimensional cavities: new insights from algebraic topological studies.

As a model system, Escherichia coli has been used to study various life processes. A dramatic paradigm shift has occurred in recent years, with the study of single proteins moving toward the study of dynamically interacting proteins, especially protein-protein interaction (PPI) networks. However, despite the importance of PPI networks, little is known about the intrinsic nature of the network structure, especially high-dimensional topological properties. By introducing general hypergeometric distribution, we reconstruct a statistically reliable combined PPI network of E. coli (E. coli-PPI-Network) from several datasets. Unlike traditional graph analysis, algebraic topology was introduced to analyze the topological structures of the E. coli-PPI-Network, including high-dimensional cavities and cycles. Random networks with the same node and edge number (RandomNet) or scale-free networks with the same degree distribution (RandomNet-SameDD) were produced as controls. We discovered that the E. coli-PPI-Network had special algebraic typological structures, exhibiting more high-dimensional cavities and cycles, compared to RandomNets or, importantly, RandomNet-SameDD. Based on these results, we defined degree of involved q-dimensional cycles of proteins (q-DCprotein ) in the network, a novel concept that relies on the integral structure of the network and is different from traditional node degree or hubs. Finally, top proteins ranked by their 1-DCprotein were identified (such as gmhB, rpoA, rplB, rpsF and yfgB). In conclusion, by introducing mathematical and computer technologies, we discovered novel algebraic topological properties of the E. coli-PPI-Network, which has special high-dimensional cavities and cycles, and thereby revealed certain intrinsic rules of information flow underlining bacteria biology.

Overexpression of ZNF217 in glioblastoma contributes to the maintenance of glioma stem cells regulated by hypoxia-inducible factors.

Glioblastoma multiforme (GBM) is the most aggressive and common kind of primary brain tumor in adults, and is thought to be driven by a subpopulation of glioma stem cells (GSCs). GSCs reside in a specialized hypoxic niche, which can regulate the tumorigenic capacity of GSCs primarily through the hypoxia-inducible factors (HIFs), HIF1α and HIF2α. ZNF217 is an oncogene frequently amplified in many kinds of tumors. It is associated with aggressive tumor behavior and poor clinical prognosis, but its role in gliomas is poorly known. Gene expression and copy number analysis from TCGA data reveal that ZNF217 is amplified in 32% and overexpressed in 71.2% of GBMs. Quantitative RT-PCR and western blotting of a cohort of glioma samples showed that ZNF217 was highly expressed in gliomas and increased with tumor grade. Analysis of a molecular database demonstrated that ZNF217 expression correlated with poor survival of glioma patients. Investigation of ZNF217 expression in GSCs, non-GSCs and normal neural stem cells (NSCs) indicated that ZNF217 was more highly expressed in GSCs than in non-GSCs and NSCs. Knockdown of ZNF217 in GSCs by small-interfering RNA (siRNA) inhibited their growth and promoted their differentiation. Interestingly, ZNF217 was upregulated in GSCs and the GBM cell line U87 when exposed to the hypoxic environment of 1% oxygen. Knockdown of either HIF1α or HIF2α, which has a central role in the hypoxia-induced responses of these cells, inhibited ZNF217 expression. In addition, ZNF217 upregulation was compromised under hypoxia in U87 and GSCs when either HIF1α or HIF2α was targeted by siRNA. HIF2α knockdown inhibited ZNF217 expression more efficiently in both normoxia and hypoxia than HIF1α knockdown. Therefore, ZNF217 is overexpressed in GBMs and contributes to the maintenance of GSCs, which is regulated by HIFs released by the hypoxic environment of the tumor.

Maintenance of critical properties of brain tumor stem-like cells after cryopreservation.

It would be very useful to be able to classify brain tumor stem cells (BTSCs) by certain criteria to afford the design of specific or individualized treatment. Here, we studied two BTSC lines with differing biological and molecular features and whose respective features were well preserved after cryopreservation as single cells in SFM or 90% serum with 10% DMSO, a method not previously reported. The resuscitated BTSCs shared properties indistinguishable from their respective parental cells, including tumor sphere forming potentials, growth and differentiation properties, and tumorigenesis in vivo. The two cell lines also had differing molecule profiles, which can be well preserved after cryopreservation, similar to that of their respective primary tumors. Therefore, BTSCs from different patients, that have their own properties, were well retained by the present cryopreservation method, which might be a useful and reliable method for preserving BTSCs for long-term studies, such as classification and specific therapy design.

Progress on potential strategies to target brain tumor stem cells.

The identification of brain tumor stem cells (BTSCs) leads to promising progress on brain tumor treatment. For some brain tumors, BTSCs are the driving force of tumor growth and the culprits that make tumor revive and resistant to radiotherapy and chemotherapy. Therefore, it is specifically significant to eliminate BTSCs for treatment of brain tumors. There are considerable similarities between BTSCs and normal neural stem cells (NSCs), and diverse aspects of BTSCs have been studied to find potential targets that can be manipulated to specifically eradicate BTSCs without damaging normal NSCs, including their surface makers, surrounding niche, and aberrant signaling pathways. Many strategies have been designed to kill BTSCs, and some of them have reached, or are approaching, effective therapeutic results. Here, we will focus on advantages in the issue of BTSCs and emphasize on potential therapeutic strategies targeting BTSCs.

Malignant Intracranial High Grade Glioma and Current Treatment Strategy.

Malignant high-grade glioma (HGG) is the most common and extremely fatal type of primary intracranial tumor. These tumors recurred within 2 to 3 cm of the primary region of tumor resection in the majority of cases. Furthermore, the blood-brain barrier significantly limited the access of many systemically administered chemotherapeutics to the tumor, pointing towards a stringent need for new therapeutic patterns. Therefore, targeting therapy using local drug delivery for HGG becomes a priority for the development of novel therapeutic strategies. The main objectives to the effective use of chemotherapy for HGG include the drug delivery to the tumor region and the infusion of chemotherapeutic agents into the vascular supply of a tumor directly, which could improve the pharmacokinetic profile by enhancing drug delivery to the neoplasm tissue. Herein, we reviewed clinical and molecular features, different methods of chemotherapy application in HGGs, especially the existing and promising targeting therapies using local drug delivery for HGG which could effectively inhibit tumor invasion, proliferation and recurrence of HGG to combat the deadly disease. Undoubtedly, novel chemical medicines targeting these HGG may represent one of the most important directions in the Neuro-oncology.

The transcriptional modulator HMGA2 promotes stemness and tumorigenicity in glioblastoma.

Glioblastoma (GBM) contains a population of stem-like cells that promote tumor invasion and resistance to therapy. Identifying and targeting stem cell factors in GBM may lead to the development of more effective therapies. High Mobility Group AT-hook 2 (HMGA2) is a transcriptional modulator that mediates motility and self-renewal in normal and cancer stem cells. We identified increased expression of HMGA2 in the majority of primary human GBM tumors and cell lines compared to normal brain. Additionally, HMGA2 expression was increased in CD133+ GBM neurosphere cells compared to CD133- cells. Targeting HMGA2 with lentiviral short hairpin RNA (shRNA) led to decreased GBM stemness, invasion, and tumorigenicity. Ectopic expression of HMGA2 in GBM cell lines promoted stemness, invasion, and tumorigenicity. Our data suggests that targeting HMGA2 in GBM may be therapeutically beneficial.

Hypoxia upregulates HIG2 expression and contributes to bevacizumab resistance in glioblastoma.

Hypoxia contributes to the maintenance of stem-like cells in glioblastoma (GBM), and activates vascular mimicry and tumor resistance to anti-angiogenesis treatments. The present study examined the expression patterns and biological significance of hypoxia-inducible protein 2 (HIG2, also known as HILPDA) in GBM. HIG2 was highly expressed in gliomas and was correlated with tumor grade, and high HIG2 expression independently predicted poor GBM patient prognosis. HIG2 was upregulated during hypoxia and by hypoxia mimics, and HIG2 knockdown in GBM cells inhibited cell proliferation and invasion. HIF1α bound to the HIG2 promoter and increased its expression in GBM cells, and HIG2 upregulated HIF1α expression. Reconstruction of a HIG2-related molecular network using bioinformatics methods revealed that HIG2 is closely correlated with angiogenesis genes, such as VEGFA, in GBM. HIG2 levels positively correlated with VEGFA in GBM samples. In addition, treatment of transplanted xenograft nude mice with bevacizumab (anti-angiogenesis therapy) resulted in HIG2 upregulation at late stages. We conclude that HIG2 is overexpressed in GBM and upregulated by hypoxia, and is a potential novel therapeutic target. HIG2 overexpression is an independent prognostic indicator and may promote tumor resistance to anti-angiogenesis treatments.

LGR5/GPR49 is implicated in motor neuron specification in nervous system.

The biological roles of stem cell marker LGR5, the receptor for the Wnt-agonistic R-spondins, for nervous system are poorly known. Bioinformatics analysis in normal human brain tissues revealed that LGR5 is closely related with neuron development and functions. Interestingly, LGR5 and its ligands R-spondins (RSPO2 and RSPO3) are specifically highly expressed in projection motor neurons in the spinal cord, brain stem and cerebral. Inhibition of Notch activity in neural stem cells (NSCs) increased the percentage of neuronal cells and promoted LGR5 expression, while activation of Notch signal decreased neuronal cells and inhibited the LGR5 expression. Furthermore, knockdown of LGR5 inhibited the expression of neuronal markers MAP2, NeuN, GAP43, SYP and CHRM3, and also reduced the expression of genes that program the identity of motor neurons, including Isl1, Lhx3, PHOX2A, TBX20 and NEUROG2. Our data demonstrated that LGR5 is highly expressed in motor neurons in nervous system and is involved in their development by regulating transcription factors that program motor neuron identity.

Glioblastoma vasculogenic mimicry: signaling pathways progression and potential anti-angiogenesis targets.

Glioblastoma (GBM) is a highly angiogenic malignancy that is resistant to standard therapy; neo-formed vessels of this aggressive malignancy are thought to arise by sprouting of pre-existing brain capillaries. However, the conventional anti-angiogenic therapy, which seemed promising initially, shows transitory and incomplete efficacy. The discovery of vasculogenic mimicry (VM) has offered a new horizon for understanding tumor vascularization. VM is a tumor cell-constituted, matrix-embedded fluid-conducting meshwork that is independent of endothelial cells and is positively correlated with poor prognosis. Therefore, a better understanding of GBM vasculature is needed to optimize anti-angiogenic therapy. This review focuses on the signaling molecules and cascades involved in VM in relation to ongoing glioma research, as well as the clinical translational advances in GBM that have been offered by the development of optimized anti-angiogenesis treatment modalities.

Overexpression of FRAT1 is associated with malignant phenotype and poor prognosis in human gliomas.

Glioma is the most common malignancy of the central nervous system. Approximately 40 percent of intracranial tumors are diagnosed as gliomas. Difficulties in treatment are associated closely with the malignant phenotype, which is characterized by excessive proliferation, relentless invasion, and angiogenesis. Although the comprehensive treatment level of brain glioma is continuously progressing, the outcome of this malignancy has not been improved drastically. Therefore, the identification of new biomarkers for diagnosis and therapy of this malignancy is of significant scientific and clinical value. FRAT1 is a positive regulator of the Wnt/ß-catenin signaling pathway and is overexpressed in many human tumors. In the present study, we investigated the expression status of FRAT1 in 68 patients with human gliomas and its correlation with the pathologic grade, proliferation, invasion, angiogenesis, and prognostic significance. These findings suggest that FRAT1 may be an important factor in the tumorigenesis and progression of glioma and could be explored as a potential biomarker for pathological diagnosis, an indicator for prognosis, and a target for biological therapy of malignancy.

Disrupting LIN28 in atypical teratoid rhabdoid tumors reveals the importance of the mitogen activated protein kinase pathway as a therapeutic target.

Atypical teratoid rhabdoid tumor (AT/RT) is among the most fatal of all pediatric brain tumors. Aside from loss of function mutations in the SMARCB1 (BAF47/INI1/SNF5) chromatin remodeling gene, little is known of other molecular drivers of AT/RT. LIN28A and LIN28B are stem cell factors that regulate thousands of RNAs and are expressed in aggressive cancers. We identified high-levels of LIN28A and LIN28B in AT/RT primary tumors and cell lines, with corresponding low levels of the LIN28-regulated microRNAs of the let-7 family. Knockdown of LIN28A by lentiviral shRNA in the AT/RT cell lines CHLA-06-ATRT and BT37 inhibited growth, cell proliferation and colony formation and induced apoptosis. Suppression of LIN28A in orthotopic xenograft models led to a more than doubling of median survival compared to empty vector controls (48 vs 115 days). LIN28A knockdown led to increased expression of let-7b and let-7g microRNAs and a down-regulation of KRAS mRNA. AT/RT primary tumors expressed increased mitogen activated protein (MAP) kinase pathway activity, and the MEK inhibitor selumetinib (AZD6244) decreased AT/RT growth and increased apoptosis. These data implicate LIN28/RAS/MAP kinase as key drivers of AT/RT tumorigenesis and indicate that targeting this pathway may be a therapeutic option in this aggressive pediatric malignancy.