text
string | predicted_class
string | confidence
float16 |
|---|---|---|
We next examined the potency of miR-128 at regulating Tert protein levels in HeLa cells. We determined that miR-128 significantly decreased Tert protein levels, compared to miR controls. In contrast, anti-miR-128 significantly enhanced Tert protein levels, relative to cells expressing miR control (Figure 2B). In addition, analysis of Tert protein levels and localization was analyzed by confocal analysis, demonstrating that miR-128 significantly reduces nuclear staining of Tert protein in HeLa cells (seen as red single channel Tert staining and pink staining in overlay images with DAPI stained nuclei), relative to miR control HeLa cells (Figure 2C). As expected, anti-miR-128 significantly enhanced Tert protein nuclear expression, compared to miR control HeLa cells (Figure 2C). Finally, we verified that miR-128 significantly regulates Tert protein levels in a small panel of cancer cell lines, including lung cancer (A549), colon cancer (SW620) and pancreatic cancer (PANC1), demonstrating that the effect of miR-128 on Tert is not limited to HeLa cells (Figure 2D). Taken together this data demonstrate that miR-128 regulates Tert expression both at the mRNA and protein levels in different cell types.
|
study
| 100.0 |
miR-128 could potentially regulate telomerase activity by directly targeting TERT mRNA, or by regulating expression of other proteins that regulate telomerase, or both. In parallel with this work our laboratory have been investigating the mechanisms by which miR-128 regulate L1-induced mutagenesis, including the repression of L1 RT activity. We have demonstrated that miR-128 repress L1 retrotransposition and genomic integration by directly targeting the L1 RNA, in a similar fashion as when miRs function as an anti-viral defense mechanism in human cells, limiting viral replication of RNA virus [29, 38, 39]. We surprisingly determined that miR-128 targets L1 RNA by binding directly through an imperfect seed match to L1 RNA in the coding sequence (CDS) of ORF2, which encodes the L1 endonuclease and reverse transcriptase (RT). With our new finding that miR-128 can also repress telomerase activity in a functional assay (q-TRAP) and by regulating TERT mRNA and protein levels (Figure 2), we hypothesized that miR-128 might be binding to a shared conserved site between L1 RT and TERT mRNA. We aligned the mRNA sequences of the two cellular RTs (L1 and TERT) and determined that the functional miR-128 binding site in L1 is, in fact, present at two locations in the CDS of TERT mRNA (Figure 3A and 3B). To test if the two non-canonical miR-128 seed sites are functional, we generated TERT CDS luciferase constructs either encoding the wildtype (WT) TERT binding Site #1 or binding Site #2. In addition, we generated a 23nt perfect miR-128 match positive control plasmid (as previously described [29, 40]). HeLa cells were co-transfected with one of the three TERT constructs WT Site #1, WT Site #2 or the positive control, (Wildtype (WT) seed plasmids) and either miR-128 or miR control mimic oligonucleotides. Luciferase activity was modestly, but significantly reduced in HeLa cells transfected with either one of the two WT TERT constructs and miR-128, relative to miR controls (Figure 3C). As expected, luciferase activity was potently repressed in the positive HeLa cell control (Figure 3C). These experiments supports the conclusion that miR-128 can bind to the predicted binding site likely at both location of the TERT mRNA sequence. Next, we generated a plasmid encoding the TERT binding sites (Site #2) in which we included mutations in the putative miR-128–binding site of TERT mRNA (Figure 3D, top). HeLa cells co-transfected with the plasmid encoding WT TERT (WT seed plasmid) and miR-128, showed a significant reduced luciferase activity as previously demonstrated (Figure 3C and 3D). In contrast, HeLa cells co-transfected with the mutant TERT mRNA-binding site (Mutated seed plasmid) and either mature miR-128 or control-miR mimics exhibited luciferase activity at similar levels as in the WT TERT and control-miR cells, consistent with miR-128 no longer binding and repressing luciferase reporter-gene expression (Figure 3D). These experiments determine that miR-128 is dependent on the predicted nucleotide sequence to interact with TERT mRNA.
|
study
| 100.0 |
(A) Schematic representation of the predicted and functional miR-128 binding sites in long-interspaced element-1 (LINE-1, L1) CDS RNA compared to two predicted miR-128 binding sites in the CDS of TERT mRNA. (B) Schematic representation of the two putative non-perfect 7-mer miR-128 binding sites in the coding sequence of TERT mRNA. (C) Relative luciferase levels of HeLa cell transfected with constructs expressing either TERT binding sequences at site #1 or site #2 or a perfect miR-128 seed match (perfect seed) (Wildtype (WT) seed plasmids), along with miR-control (Control) or miR-128 (miR-128) mimics, measured 48 hours post transfection. Results shown as percent change ± SEM (n = 3 independent experiments, *p < 0.05, **p < 0.01, ****p < 0.0001) (D) Relative luciferase levels of HeLa cell transfected with constructs expressing the wildtype (WT seed plasmid) or mutated (Mutated seed plasmid) 7-mer TERT binding sequences, along with miR-control (Control) or miR-128 (miR-128) mimics, measured 48 hours post transfection. Results shown as percent change ± SEM (n = 3 independent experiments, **p < 0.01).
|
study
| 100.0 |
Finally, to test whether miR-128 interacts with TERT mRNA in cells, we isolated Ago complexes containing miRs and target mRNAs by immunopurification from HeLa cells either overexpressing miR-128 or anti–miR-128 and assessed relevant complex occupancy by TERT mRNA (Figure 4A). As determined previously (Figure 2A), miR-128 reduces TERT mRNA levels in HeLa cells (Input). In addition, we determined that despite the lower levels of TERT mRNA in HeLa cells (caused by over-expression of miR-128), Ago-bound TERT mRNA was significantly higher in cells overexpressing miR-128 than in cells in which miR-128 was downregulated by anti–miR-128 (IP) (Figure 4B, left panel). When correcting for the higher levels of TERT mRNA in HeLa cells treated with anti–miR-128, the difference in bound TERT mRNA was even more significant (IP) (anti–miR-128 corrected) (Figure 4B, left panel). A control, constitutively expressed transcript of the GAPDH gene did not show altered levels of total RNA in cells transduced with miR-128 or anti–miR-128 (Input), or relative differences in Ago immunopurification (IP) (Figure 4B, middle panel). Finally, we verified that TERT mRNA could also be immunopurified by miR-128-Ago in Tera cells, showing that miR-128 significantly reduces TERT mRNA levels in Tera cells (Input), and that miR-128 interacts with TERT mRNA in Tera cells, at a significantly manner, when correcting for miR-128 levels (Figure 4B, right panel). This body of work supports the conclusion that miR-128 interacts with TERT mRNA and suggests that the putative miR-128 binding site in the coding region of TERT mRNA is indeed the functional binding site resulting in potent regulation of TERT levels and telomerase activity.
|
study
| 100.0 |
(A) Schematic of Argonaute immunopurification strategy (Ago-RIP) strategy. HeLa or Tera cell lines were generated in which miR-128 was either neutralized or over-expressed (stably transduced with anti-miR-128 or miR-128-expressing constructs). If TERT is a direct target of the miR-128/Ago complex, then Ago immunopurification in cells with neutralized miR-128 will pull out less TERT mRNA compared to miR-128 expressing cells, which will bind TERT mRNA directly. (B) Relative levels of TERT mRNA were determined by q-PCR analysis and (normalized to B2M) in “input samples” of miR modulated HeLa cells. Relative TERT mRNA levels were next determined in IP fractions and normalized to input levels. TERT IP fractions are also shown as “corrected” levels, in which IP TERT levels were corrected for levels of miR-128 in HeLa samples. As a negative control the same samples were analyzed for relative expression of GAPDH levels. Finally, Ago-RIP in Tera cells was performed as described for HeLa cells. Results from 3 independent experiments are shown as mean of IP fraction ± SEM of three independent experiments (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).
|
study
| 100.0 |
This study is the first to identify that miR-128 targets TERT mRNA and reduces TERT mRNA and protein levels resulting in a decrease in telomerase activity in cancer cells. Telomerase is a cellular reverse transcriptase that maintains chromosome health by extending telomeres and protecting chromosome ends. While telomerase is inactive in most adult cells, it is reactivated in cancer cells allowing continuous proliferation. The necessity for telomerase in continued cancer cell growth makes it an attractive therapeutic target.
|
study
| 100.0 |
microRNAs (miRs) have been established as crucial players in cancer initiation and progression by regulating oncogenes or tumor suppressor genes. miR-128 has previously been found to act as a tumor suppressor and it is downregulated in various types of cancer, including glioma , lung cancer , prostate cancer and bladder cancer . Furthermore, miR-128 is highly enriched in the brain, but not detected in glioma cells. Mechanistically, miR-128 has been found to reduce glioma cell proliferation and promote stem cell self-renewal, by the regulation of the BMI-1 oncogene . miR-128 has also been demonstrated to positively regulate p53 by directly targeting SIRT1, and promote apoptosis in a PUMA-dependent manner . In non-small cell lung cancer cells, miR-128 overexpression was observed to suppress invasion and induce cell cycle arrest and apoptosis. Interestingly, when miR-128 was restored, tumorigenicity was greatly suppressed in a mouse model of lung cancer . Taken together, these studies demonstrates that miR-128 functions as a tumor suppressor in many cancer types and our experimental findings add another mechanism by which miR-128 repress the oncogenic phenotype of cancer cells.
|
review
| 70.75 |
The regulation of telomerase activity in healthy cells and in cancer is complex and includes genomic rearrangements, as well as regulation by cellular factors including (but not limited to): Proteinase-activated receptor 1 (PAR1), Telomeric Repeat Factor 1 (TRF1), the Kinase, Endopeptidase and Other Proteins of small Size complex (KEOPS complex) and the Laminin Receptor (LRP) [45–50].
|
study
| 99.75 |
We have previously established that miR-128 interacts with the coding sequence (CDS) of the reverse transcriptase component of long interspersed element 1 (LINE-1) retrotransposons (ORF2), preventing retrotransposition, genomic integration and mutagenesis . Interestingly, the activity of L1 and telomerase is closely intertwined. Specifically, L1 retrotransposons can provide an alternative mechanism to maintain telomere structure in organisms from Drosophila melanogaster to humans and L1 activity has been reported to induce hTERT mRNA levels and telomerase activity in tumor cell lines [51–53].
|
study
| 100.0 |
The finding that the functional miR-128 binding sites are located in both the CDS of telomerase mRNA and the L1 mRNA, responsible for repressing Telomerase activity and L1 mobilization in cancer cells, supports the novel concept that one microRNA (for example miR-128) can function by parallel regulation of groups of enzymes (such as RT) by binding to conserved target sequences in the coding region of the enzymes.
|
study
| 100.0 |
All cells were incubated at 37° C and 5% CO2 and routinely checked for mycoplasma contamination. HeLa cells, which are derived from cervix adenocarcinoma (CCL-2, American Tissue Cell Culture (ATCC)) were cultured in EMEM (SH3024401, Hyclone) supplemented with 10% FBS. 293T cells, which are human embryonic kidney (HEK) cells that express and are transformed with large T antigen (CRL-3216, ATCC) were cultured in DMEM (25–501N, Genesee) with 10% FBS. Tera-1 cells, which originate from metastatic human GCT/teratoma (HTB-105, ATCC) were cultured in McCoy’s 5A (16600–082, Life Technologies) supplemented with 20% cosmic serum (SH3008702, Fisher Scientific).
|
study
| 99.94 |
HeLa cells were transduced with miR Zip Virus Library (MZIPPLVA, System Biosciences), (encoding anti-miRs which neutralizes conserved and/or well-studied miRs) selected for Puromycin resistance and split to single cell dilutions in 96-well plates. Cells were grown to confluency and telomerase activity was measured using the telomeric repeat amplification protocol (q-TRAP) .
|
study
| 99.94 |
VSV-G-pseudotyped lentiviral particles were made by transfecting 293T cells with 0.67 µg of pMD2-G (12259, Addgene), 1.3 µg of pCMV-DR8.74 (8455, Addgene) and 2 µg of mZIP-miR-128 or mZIP-anti-miR-128 using Lipofectamine LTX (15338030, ThermoFisher). Viral supernatants were concentrated using PEG-it (LV810A-1, System Biosciences). Cells were transduced with high titer virus using polybrene (sc-134220, Santa Cruz Biotech) and spinoculated at 800 x g at 32°C for 30 minutes. Transduced cells were selected and maintained using 10 µg/ml puromycin.
|
study
| 99.94 |
HeLa cells were lysed in NP40 Lysis buffer and q-TRAP analysis was carried out. Briefly, lysates were mixed with EGTA (NC9118216), Platinum Taq polymerase (10966034, Life), ACX primer (5′-GCG CGG CTT ACC CTT ACC CTT ACC CTA ACC-3′), TS primer (5′-AAT CCG TCG AGC AGA GTT-3′), and either SYBR Green Master Mix (4367659, Life) or Forget-Me-Not qPCR Master Mix with Rox (31042–1, Biotium). Samples were incubated for 30 min at 30° C for telomerase extension, then 95° C for 10 min to deactivate telomerase and activate Platinum Taq polymerase, then 40 cycles of 95° C for 15 sec and 60° C for 1 min. Amplification was normalized to a standard curve of HeLa lysates diluted 1:5 from 1 µg/well to 8 ng/well.
|
study
| 99.94 |
RNA was extracted with Trizol (15596018, ThermoFisher Scientific) according to manufacturer’s instructions and cDNA synthesis was performed with the High Capacity Reverse Transcriptase Kit (4368813, Life Technologies). cDNA was amplified relative to GAPDH using the Forget-Me-Not qPCR Master Mix with Rox (31042–1, Biotium) according to the manufacturer’s protocol.
|
other
| 99.9 |
Cells were lysed in RIPA buffer (89901, ThermoFisher, Waltham, MA) supplemented with 1× protease inhibitor cocktail (PI78410, ThermoFisher, Waltham, MA) and then mixed with 4x LDS sample buffer (NP0008, ThermoFisher, Waltham, MA) and boiled at 95° C for 10 minutes. Samples were run on NuPAGE Novex 4–12% Bis-Tris Protein Gels (NP0335, ThermoFisher, ThermoFisher Scientific), and transferred to PVDF membranes. Membranes were incubated overnight at 4C with rabbit anti-human Tert antibody, (1:100 dilution) (Y182 ab32020, Abcam, Cambridge, MA) then HRP-linked anti-rabbit IgG antibody (1:2000 dilution) (7074S, Cell Signaling Technology, Danvers, MA) and visualized with Pierce ECL Western Blotting Substrate (32106, ThermoFisher, Waltham, MA) on the Bio-Rad ChemiDoc XRS+ System.
|
study
| 99.9 |
Cells were plated on gelatin-coated coverslips, fixed in 4% paraformaldehyde (Sigma-Aldrich, St. Louis, MO), incubated in blocking buffer (1% bovine serum albumin, 0.3% Triton X-100 (ThermoFisher, Waltham, MA) in PBS), stained with mouse anti-human Tert antibody incubated overnight at 4C (ab5181, Abcam, Cambridge, MA) at 1:25 dilution, followed by PE-conjugated anti-mouse IgM antibody for 2 hours at room temperature (1:250 dilution) (clone eB121-15F9, eBioscience/ThermoFisher, Waltham, MA). Coverslips were mounted on slides with VectaSheild with DAPI (H-1200, Vector Laboratories, Burlingame, CA) and cells imaged at 63x on a Zeiss spinning disk confocal microscope.
|
other
| 91.3 |
Wildtype or mutated TERT sequences were cloned into a dual luciferase reporter plasmid (pEZX-MT05, Genecopoeia). 3 x 105 HeLa cells were forward-transfected with 0.8 µg reporter plasmid and 20 nM control mimic or miR-128 mimic with Attractene transfection reagent (301005, Qiagen) according to the manufacturer’s instructions. Relative Gaussia luciferase and secreted alkaline phosphatase (SEAP) levels were determined with the Secrete-Pair Dual Luminescence Assay Kit (SPDA-D010, Genecopoeia) on a Tecan Infinite F200 microplate reader.
|
study
| 100.0 |
Immunopurification of Argonaute from HeLa and Tera cell extracts was performed using the 4F9 antibody (4F9, Santa Cruz Biotechnology) as described previously . Briefly, 10 mm plates of 80% confluent cultured cells were washed with buffer A [20 mM Tris-HCl pH 8.0, 280 mM KCl, 10 mM EDTA, 1% NP-40, 0.2% Deoxycholate, 2X Halt protease inhibitor cocktail (Pierce), 200 U/ml RNaseout (ThermoFisher Scientific) and 1 mM DTT]. Protein concentration was adjusted across samples with buffer B [20 mM Tris-HCl pH 8.0, 140 mM KCl, 5 mM EDTA pH 8.0, 0.5% NP-40, 0.1% deoxycholate, 100 U/ml Rnaseout (ThermoFisher Scientific), 1 mM DTT and 1X Halt protease inhibitor cocktail (Pierce)]. Lysates were centrifuged at 16,000xg for 15 min at 4° C and supernatants were incubated with 10–20 µg of 4F9 antibody conjugated to epoxy magnetic beads (M-270 Dynabeads, ThermoFisher) for 2 hours at 4° C with gentle rotation. Following magnetic separation, the beads were washed three times five min with 2 ml of buffer C [20 mM Tris-HCl pH 8.0, 140 mM KCl, 5 mM EDTA pH 8.0, 40 U/ml Rnaseout (ThermoFisher Scientific), 1 mM DTT and 1X Halt protease inhibitor cocktail (Pierce)]. Following immunopurification, RNA was extracted using miRNeasy kits (217004, Qiagen), following the manufacturer’s recommendations and qPCR was performed using custom probes/primers (TERT Forward: ACCAAGCATTCCTGCTCAAG and TERT Reverse: GCTGCTGGTGTCTGCTCTC) and Forget-me-not qPCR master mix (31042–1, Biotium). Results were normalized to their inputs and shown as “corrected” values as a proxy for Ago immunopurification efficiency.
|
study
| 100.0 |
Sequence alignment algorithms are a key component of many bioinformatics applications. The NCBI BLAST [1, 2] is a widely used tool that implements algorithms for sequence comparison. These algorithms are the basis for many other types of BLAST searches such as BLASTX, TBLASTN, and BLASTP . The demand for processing large amounts of genomic data that gushes from NGS devices has grown faster than the rate which industry can increase the power of computers (known as Moore’s Law). This fact has raised new challenges for the implementation of scalable and efficient computational systems. In this scenario, MapReduce (and its Hadoop implementation) emerged as a paramount framework that supports design patterns which represent general reusable solutions to commonly occurring problems across a variety of problem domains including analysis and assembly of biological sequences . MapReduce has delivered outstanding performance and scalability for a myriad of applications running over hundreds to thousands of processing nodes . On the other hand, over the last decade, cloud computing has emerged as a powerful platform for the agile and dynamic provisioning of computational resources for computational and data intensive problems.
|
other
| 96.4 |
Despite of its popularity, MapReduce requires algorithms to be adapted according to such design patterns . Although this adaptation may result in efficient implementations for many applications, this is not necessarily true for many other algorithms, which limits the applicability of MapReduce. Moreover, because MapReduce is designed to handle extremely large data sets, its implementation frameworks (e.g. Hadoop and the Amazon’s Elastic MapReduce service) constrains the program’s ability to process smaller data.
|
other
| 99.9 |
More recently, Apache Spark has emerged as a promising and more flexible framework for the implementation of highly scalable parallel applications [10, 11]. Spark does not oblige programmers to write their algorithms in terms of the map and reduce parallelism pattern. Spark implements in-memory operations, based on the Resilient Distribution Datasets (RDDs) abstraction . RDD is a collection of objects partitioned across nodes in the Spark cluster so that all partitions can be computed in parallel. We may think of RDDs as a collection of data objects which are transformed into new RDDs as the computation evolves. Spark maintains lists of dependencies among RDDs which are called “lineage”. It means RDDs can be recomputed in case of lost data (e.g. in the event of failure or simply when some data has been previously discarded from memory).
|
other
| 99.9 |
In this paper we propose SparkBLAST, which uses the support of Apache Spark to parallelize and manage the execution of BLAST either on dedicated clusters or cloud environments. Spark’s pipe operator is used to invoke BLAST as an external library on partitioned data of a query. All the input data (the query file and the database) and output data of a query are treated as Spark’s RDDs. SparkBLAST was evaluated on both Google and Microsoft Azure Clouds, for several configurations and dataset sizes. Experimental results show that SparkBLAST improves scalability when compared to CloudBLAST in all scenarios presented in this paper.
|
other
| 99.75 |
A design goal is to offer a tool which can be easily operated by users of the unmodified BLAST. Thus, SparkBLAST implements a driver application written in Scala, which receives user commands and orchestrates the whole application execution, including data distribution, tasks execution, and the gathering of results in a transparent way for the user.
|
other
| 99.94 |
Two input files must be provided for a typical operation: (i) the target database of bacterial genomic sequences, which will be referred to as target database from now on, for short; and (ii) the query file, which contains a set of query genomic sequences that will be compared to the target’s database sequences for matching. As depicted in Fig. 1, SparkBLAST replicates the entire target database on every computing node. The query file is evenly partitioned into data splits which are distributed over the nodes for the execution. Thus, each computing node has a local deployment of the BLAST application, and it receives a copy of the entire target database and a set of fragments of the query file (splits). Fig. 1Data distribution among n nodes: the target database (D) is copied on every computing node; the query file (S) is evenly partitioned into data splits (S 1,…,S n) which are distributed over the nodes. Each split (S i) can be replicated on more than one node for fault tolerance
|
other
| 99.75 |
Data distribution among n nodes: the target database (D) is copied on every computing node; the query file (S) is evenly partitioned into data splits (S 1,…,S n) which are distributed over the nodes. Each split (S i) can be replicated on more than one node for fault tolerance
|
other
| 99.94 |
Note that it is possible to apply different techniques for task and data partitioning. Each data split (i.e., fragment of the query file) can be replicated by the distributed file system (DFS) on a number of nodes, for fault tolerance purposes. Spark’s scheduler then partitions the whole computation into tasks, which are assigned to computing nodes based on data locality using delay scheduling . For the execution of each task, the target database and one fragment of the query file are loaded in memory (as RDDs). The target database (RDD) can be reused by other local tasks that execute in the same machine, thus reducing disk access .
|
other
| 99.9 |
Spark can execute on top of different resource managers, including Standalone, YARN, and Mesos . We chose YARN because it can be uniformly used by Spark and Hadoop. It is important to avoid the influence of resource scheduling in the performance tests presented in this paper. In fact, YARN was originally developed for Hadoop version 2. With YARN, resources (e.g., cpu, memory) can be allocated and provisioned as containers for tasks execution on a distributed computing environment. It plays better the role of managing the cluster configuration, and dynamically shares available resources, providing support for fault tolerance, inter-, and intra-node parallelism. Other applications which have been written or ported to run on top of YARN include Apache HAMA, Apache Giraph, Open MPI, and HBASE1.
|
other
| 99.9 |
Data processing in SparkBLAST can be divided into three main stages (as depicted in Fig. 2): pre-processing, main processing and post-processing. Such stages are described in the following subsections. Fig. 2The workflow implemented by SparkBLAST: during each of the three stages, parallel tasks (represented as vertical arrows) are executed in the computing nodes. Pre-processing produce the splits of the query file and copy them to the DFS. The main processing execute local instances of BLAST on local data. Finally, the post processing merges output fragments into a unique output file
|
other
| 99.9 |
The workflow implemented by SparkBLAST: during each of the three stages, parallel tasks (represented as vertical arrows) are executed in the computing nodes. Pre-processing produce the splits of the query file and copy them to the DFS. The main processing execute local instances of BLAST on local data. Finally, the post processing merges output fragments into a unique output file
|
other
| 99.94 |
In order to evaluate the performance and the benefits of SparkBLAST, we present two experiments. The first experiment was executed in the Google Cloud, and the second experiment executed in the Microsoft Azure platform. Both experiments executed with 1, 2, 4, 8, 16, 32, and 64 virtual machines as computing nodes for scalability measurement. For the sake of comparison, each experiment was executed on SparkBLAST and on CloudBLAST. The later is a Hadoop based tool designed to support high scalability on clusters and cloud environments. For the experiments, we used Spark 1.6.1 to execute SparkBLAST on both cloud environments. To execute CloudBLAST, we used Hadoop 2.4.1 on the Google Cloud, and Hadoop 2.5.2 on Azure Cloud. In any case, we configures YARN as the resource scheduler, since our experiments focus on performance. Further details on the experimental setup will be provided in the results section.
|
study
| 99.9 |
This work was originally inspired and applied in a radionuclides resistance study. Genome sequences of several radiation-resistant microorganisms can be used for comparative genomics to infer the similarities and differences among those species. Homology inference is important to identify genes shared by different species and, as a consequence, species-specific genes can be inferred. Two experiments are considered in this work. The input data for Experiment 1 was composed of 11 bacterial genome protein sequences, 10 of these are radiation-resistant (Kineococcus radiotolerans - Accession Number NC_009660.1, Desulfovibrio desulfuricans - NC_011883.1, Desulfovibrio vulgaris - NC_002937.3, Rhodobacter sphaeroides - NC_009429.1, Escherichia coli - NC_000913.3, Deinococcus radiodurans - NC_001263.1, Desulfovibrio fructosivorans - NZ_AECZ01000069.1, Shewanella oneidensis - NC_004349.1, Geobacter sulfurreducens - NC_002939.5, Deinococcs geothermalis - NC_008010.2, Geobacter metallireducens - NC_007517.1) for Reciprocal-Best-Hit (RBH) processing.
|
study
| 100.0 |
For Experiment 2, the input query is composed of 10 radiation-resistant bacteria. (i.e., all species listed above but E. coli). This similarity-based experiment consisted on the search of potential protein homologs of 10 radiation-resistant genomes in 2 marine metagenomics datasets.
|
study
| 99.94 |
Each input dataset was concatenated into a single multifasta input file named query1.fa (Experiment 1) and query2.fa (Experiment 2). The files query1.fa and query2.fa had 91,108 and 86,968 sequences and a total size of 36.7 MB and 35 MB, respectively. Two target metagenomic datasets obtained from MG-RAST database2 were used in Experiment 2: (i) Sargaso Sea (Bermuda), coordinates: 32.17,-64.5, 11 GB, 61255,260 proteins (Ber.fasta) and (ii) João Fernandinho (Buzios, Brazil), coordinates: -22.738705, -41874604, 805 MB, 4795,626 proteins (Buz.fasta):
|
study
| 100.0 |
In this stage, implemented by SparkBLAST, the query file is evenly partitioned into splits which are written to the DFS. The splits are then distributed among the computing nodes by the DFS, according to some replication policy for fault tolerance. Each split containing a set of (e.g., thousands of) genome sequences can be processed by a different task. Thus, the query file should be partitioned to enable parallelism. Since the input file can be potentially large, the partitioning operation can be also parallelized as illustrated in the following commands:
|
other
| 99.9 |
This stage starts after all the input data (i.e., the target database and query file splits) are properly transferred to each processing node. Tasks are then scheduled to execute on each node according to data locality. The amount of tasks executed concurrently on each computing node depends on the number of processing cores available. As soon as a computing core completes the execution of a task, it will be assigned another task. This process repeats until the available cores execute all tasks of the job.
|
other
| 99.94 |
In order to measure the scalability and speedup of SparkBLAST we carried out experiments on both the Google Cloud and Microsoft Azure, increasing the platform size from 1 to 64 computing nodes. For the sake of comparison, the same genome searches have been executed with both SparkBLAST and CloudBLAST for each platform size. Every experiment was repeated six times and and the average execution time was considered in results.
|
other
| 99.6 |
During the previous stage each individual task produces a small output file. During the post-processing stage, SparkBLAST merges all these small files into a single final output file. For instance, experiment 1 produces a final output file of 610 MB. All output data is written to the DFS, i.e., the Google Cloud Storage or Microsoft Azure’s Blob storage service.
|
other
| 99.94 |
Added-value to SparkBLAST, similarity results were obtained by (i) performing a Reciprocal Best Hit analysis [14, 15] among pairs of species, or orthology inference (Experiment 1) and (ii) searching for potential radiation-resistant homologous proteins in 2 marine metagenome datasets (Experiment 2), as described in the following section.
|
study
| 99.94 |
In Experiment 1, BLASTP was used to execute queries on a 36 MB database composed of 88,355 sequences from 11 bacterial genomes, in order to identify genes shared by different species. Ten bacteria described in literature as being resistant to ionizing radiation and one species susceptible to radiation were obtained from Refseq database. The same dataset is provided as query and target database, so that an all-to-all bacteria comparison is executed, producing a 610 MB output. BLASTP results were processed to identify RBH among pairs of species.
|
study
| 100.0 |
Experiment 1 was executed on a platform with up to 64 computing nodes plus one master node. Each node is a virtual machine configured as n1-standard-2 instance (2 vCPUs, 7.5 GB memory, CPU Intel Ivy Bridge). The virtual machines were allocated from 13 different availability zones in the Google Cloud: Asia East (3 zones), Europe West (3 zones), US Central (4 zones) e US East (3 zones). For this scalability test, both SparkBLAST and CloudBLAST were executed on platforms with 1, 2, 4, 8, 16, 32, and 64 nodes. The experiment was repeated six times for each platform size. Thus, Experiment 1 encompasses 2×7×6=84 executions in total, which demanded more than 350 h (wall clock) to execute. As an estimate on the amount of the required computational resources, this experiment consumed 2.420 vCPU-hours to execute on the Google Cloud.
|
study
| 58.5 |
The average execution times are presented in Fig. 3. SparkBLAST achieved a maximum speedup (which is the ratio between execution time of the one node baseline over the run time for the parallel execution) of 41.78, reducing the execution time from 28,983 s in a single node, to 693 s in 64 nodes. In the same scenario, CloudBLAST achieved speedup of 37, reducing the execution time from 30,547 to 825 s on 64 nodes. For this set of executions, both SparkBLAST and CloudBLAST used 2 vCPUs per node for tasks execution. The speedup is presented in Fig. 4. As shown, SparkBLAST presents better scalability than CloudBLAST. Fig. 3Total execution time for CloudBLAST vs. SparkBLAST running on the Google Cloud. Values represent the average of six executions for each experiment Fig. 4Speedup for 1 to 64 nodes in the Google Cloud. SparkBlast was executed on virtual machines with one and two cores. CloudBlast was executed on nodes with two cores
|
study
| 100.0 |
The average execution times and standard deviations are presented in Table 1. Table 2 presents the execution times for SparkBLAST when only one vCPU (core) of each node is used for processing. Table 3 presents the total execution times for SparkBLAST when both cores of each node are used for processing. Table 1Execution times for CloudBLAST - Google Cloud# nodes1248163264Exec. time 129,921.4019,018.0011,324.006,204.002,866.001,680.00794.00Exec. time 230,256.2318,550.2513,799.235,779.212,959.651,828.23900.00Exec. time 331.016.8519,221.8112,580.325,700.523,004.521,597.00815.21Exec. time 431.350.2519,102.6810,489.535,850.022,961.231,806.25842.30Exec. time 530.726.8918,981.3212,721.235,780.342,990.811,780.32799.21Exec. time 630.012.1419,118.7211,820.855,900.643,008.151,753.23802.98Mean30,547.2918,998.8012,122.535,869.122,965.061,740.84825.62Std. Dev.576.25235.281.164.02177.7052.7987.2040.32Std.Dev./Mean1.89%1.24%9.60%3.03%1.78%5.01%4.88% Table 2Execution times - SparkBLAST 1 core - Google Cloud# nodes1248163264Exec. time 136,106.8618,845.2310,189.115,556.223,129.201,716.10905.21Exec. time 236,510.1219,120.3210,199.855,540.153,115.121,730.58899.84Exec. time 336,720.8618,952.1510,170.235,560.883,140.011,790.96894.76Exec. time 438,120.2518,998.0610,200.015,543.623,120.581,694.69900.42Exec. time 536,230.5619,112.2310,178.765,552.103,122.151,701.55897.65Exec. time 636,452.5318,880.1110,183.615,565.113,127.581,710.68890.25Mean36,690.2018,984.6810,186.935,553.013,125.771,724.09898.02Std.Dev733.00115.1411.839.738.6235.015.14Std.Dev/Mean2.00%0.61%0.12%0.18%0.28%2.03%0.57% Table 3Execution times - SparkBLAST 2 cores - Google Cloud# nodes1248163264Exec. time 128,915.5214,500.867,935.454,287.852,249.941,260.12695.23Exec. time 229,002.2114,520.237,945.104,290.122,230.261,259.28690.04Exec. time 329,001.8914,515.357,950.014,283.562,255.041,260.10701.50Exec. time 428,989.5214,557.517,942.204,282.212,242.631,259.52710.11Exec. time 528,990.3214,580.017,940.804,310.122,249.261,259.82680.80Exec. time 629,001.1514,520.237,950.124,295.562,251.081,262.15682.10Mean28,983.4414,532.377,943.954,291.572,246.371,260.17693.30Std.Dev33.7829.935.6810.278.851.0311.37Std.Dev/Mean0.12%0.21%0.07%0.24%0.39%0.08%1.64%
|
study
| 99.75 |
Table 4 consolidates results from previous tables and presents mean execution times along with speedup and parallel efficiency figures for the CloudBLAST and SparkBLAST (1 and 2 cores) systems. Table 4Mean execution times, speedups and parallel efficiency (Experiment 1 - query.fasta - 36 MB) - SparkBLAST vs CloudBLAST - Google Cloud# nodes1248163264SparkBLAST1 coreExec. time36,690.2018,984.6810,186.935,553.013,125.771,724.09898.02Speedup11.933.606.6111.7421.2840.86Efficiency10.970.900.830.730.670.64SparkBLAST2 coresExec. time28,983.4414,532.377,943.954,291.572,246.371,260.17693.30Speedup1.001.993.656.7512.9023.0041,81Efficiency1.001.000.910.840.810.720.65CloudBLASTExec. time30,547.2918,998.8012,122.535,869.122,965.061,740.84825.62Speedup1.001.612.525.2010.3017.5537.00Efficiency1.000.800.630.650.640.550.58
|
other
| 99.25 |
Figure 3 compares total execution times of CloudBLAST and SparkBLAST (one and two cores configurations), for platforms composed of 1 up to 64 computing nodes. Execution times presented in correspond to the average for six executions. Parallel efficiency is presented in Fig. 5. Fig. 5Efficiency for CloudBLAST x SparkBLAST running on Google Cloud
|
other
| 99.9 |
Experiment 2 was executed on a total of 66 nodes allocated on the Microsoft Azure Platform, being all nodes from the same location (East-North US). Two A4 instances (8 cores and 14 GB memory) were configured as master nodes, and 64 A3 (4 cores and 7 GB memory) instances were configured as computing nodes. Both SparkBLAST and CloudBLAST executed queries on two datasets (Buz.fasta, and Ber.fasta), varying the number of cores allocated as 1 (BLAST sequential execution), 4, 12, 28, 60, 124 and 252. Every execution was repeated 6 times for CloudBLAST and six times for SparkBLAST. Thus, Experiment 2 encompasses 2×2×7×6=168 executions in total, which demanded more than 8,118 h (wall clock) to execute. An estimate on the amount of computational resources, this experiment consumed more than 139,595 vCPU-hours to execute on the Azure Cloud.
|
other
| 97.75 |
For the Microsoft Azure platform, SparkBLAST outperforms CloudBlast on all scenarios. Both datasets (Buz.fasta and Ber.fasta) were processed, and results are presented in Fig. 6 (speedup), Fig. 7 (total execution time), Fig. 8 (Efficiency), Table 5 (Buz.fasta), and Table 6 (Ber.fasta). It is worth noting that the largest dataset (Ber.fasta - 11 GB) was larger than the available memory in the computing nodes. For this reason, CloudBLAST could not process the Ber.fasta dataset, while SparkBLAST does not have this limitation. It is also worth mentioning that larger speedups were achieved on Microsoft Azure when compared to the Google Cloud. This can be partially explained by the fact that all computing nodes allocated on the Microsoft Azure are placed in the same location, while computing nodes on Google Cloud were distributed among 4 different locations. Fig. 6Speedup - Microsoft Azure Fig. 7Total execution time for CloudBLAST x SparkBLAST on Microsoft Azure Fig. 8Efficiency - CloudBLAST x SparkBLAST - Microsoft Azure Table 5Mean execution times, speedups and parallel efficiency (Experiment 2 - Buz.fasta - 805 MB) - SparkBLAST vs CloudBLAST - Microsoft Azure# cores4122860124252SparkBLAST143,228.9547,031.6224,850.5111,692.456,041.643,138.64Speedup3.8311.6722.0946.9590.86174.89Efficiency0.960.970.790.780.730.69CloudBLAST148,512.9547,950.0526,858.7111,951.116,993.523,879.06Speedup3.711.4520.4445.9378.49141.51Efficiency0.920.950.730.770.630.56 Table 6Mean execution times, speedups and parallel efficiency (Experiment 2 - Ber.fasta - 11 GB) - SparkBLAST vs CloudBLAST - Microsoft Azure# cores4122860124252SparkBLAST2,678,902.06859,687.13458,759.75224,869.12110,222.9856,200.21Speedup3.7311.6121.7644.490.57177.64Efficiency0.930.970.780.740.730.7CloudBLAST------Speedup------Efficiency------
|
study
| 100.0 |
In order to obtain added-value from the SparkBLAST similarity results on the cloud, the output from SparkBLAST processing of Experiment 1 was used to infer orthology relationships with the RBH approach. In Table 7, numbers represent (RBH) orthologs found between 2 species. Numbers in bold represent (RBH) paralogs found in the same species. The higher number of RBH shared by two species was 264 between Desulfovibrio vulgaris and Desulfovibrio desulfuricans, and the lower was 15 between Desulfovibrio fructosivorans and Deinococcus radiodurans. Among the same species, the higher number of RBH was 572 in Rhodobacter sphaeroides and the lower 34 in Deinococcus geothermalis. Regarding experiment 2: 1.27% (778,349/61255,260) of the Bermuda metagenomics proteins and 1.4% (68,748/4795,626) of the Búzios metagenomic proteins represent hits or potential homologs to the 10 radiation-resistant bacteria. Table 7Numbers of RBH found using data from SparkBLAST cloud processing Kineococcus Desulfovibrio Desulfovibrio Rhodobacter Escherichia Deinococcus Desulfovibrio Shewanella Geobacter Deinococcus Geobacter NameAccessionNumber of radiotolerans desulfuricans vulgaris sphaeroides coli radiodurans fructosivorans oneidensis sulfurreducens geothermalis metallireducens Numberproteins 224 43256335532221185220 Kineococcus NC_009660.14,632 radiotolerans 380 264121793816371882760 Desulfovibrio NC_011883.110,443 desulfuricans 362 6246179853472437 Desulfovibrio NC_002937.312,349 vulgaris 572 114464698775044 Rhodobacter NC_009429.120,954 sphaeroides 98 2426155342826 Escherichia NC_000913.34,140 coli 122 15212710217 Deinococcus NC_001263.17,671 radiodurans 84 20401738 Desulfovibrio NZ_AECZ01000069.14,028 fructosivorans 90 362132 Shewanella NC_004349.18,271 oneidensis 146 20120 Geobacter NC_002939.59,340 sulfurreducens 34 16 Deinococcs NC_008010.22,935 geothermalis 50 Geobacter NC_007517.13,592 metallireducens Numbers in bold represent (RBH) 354 paralogs found in the same species
|
study
| 100.0 |
In this paper we investigate the parallelization of sequence alignment algorithms through an approach that employs cloud computing for the dynamic provisioning of large amounts of computational resources and Apache Spark as the coordination framework for the parallelization of the application. SparkBLAST, a scalable parallelization of sequence alignment algorithms is presented and assessed. Apache Spark’s pipe operator and its main abstraction RDD (resilient distribution dataset) are used to perform scalable protein alignment searches by invoking BLASTP as an external application library. Experiments on the Google Cloud and Microsoft Azure have demonstrated that the Spark-based system outperforms a state-of-the-art system implemented on Hadoop in terms of speedup and execution times. It is worth noting that SparkBLAST can outperform CloudBlast even when just one of the vCPUs available per node is used by SparkBLAST, as demostrated by results obtained on the Google Cloud. In the experiments presented in this paper, the Hadoop-based system always used all vCPUs available per node.
|
study
| 99.94 |
From Table 4 it is possible to verify that both Speedup and Parallel Efficiency are better for SparkBLAST when compared to CloudBLAST for experiments executed on both the Google Cloud and Microsoft Azure. This is true for both scenarios of SparkBLAST on the Google Cloud (1 and 2 cores per node). It is worth noting that even when total execution time for CloudBLAST is smaller than the 1-core SparkBLAST configuration, Speedup and Parallel Efficiency is always worse for CloudBLAST. When SparkBLAST allocates two cores per node (as CloudBLAST does) execution times are always smaller when compared to CloudBLAST.
|
other
| 99.06 |
For the Microsoft Azure platform, all measures (processing time, efficiency and speedup) are better on SparkBLAST when compared to the corresponding execution of CloudBLAST for the Buz.fasta (805 MB) dataset. It is worth noting that the speedup difference in favor of SparkBLAST increases with the number of computing nodes, which highlights the improved scalability of SparkBLAST over CloudBLAST. As mentioned in the “Results” section, it was not possible to process the larger Ber.fasta (11 GB) dataset using CloudBLAST due to computing node’s main memory limitation. This constraint does not affect SparkBLAST, which can process datasets even when they are larger than the main memory available on computing nodes. In the case of Spark, every process invoked by a task (each core is associated to a task) can use RDD even when database do not fit in memory, due memory content reuse and the implementation of circular memory . It is worth noting that RDDs are stored as deserialized Java objects in the JVM. If the RDD does not fit in memory, some partitions will not be cached and will be recomputed on the fly each time they are needed . Indeed, one very important loophole of existing methods that we address in SparkBLAST is the capability of processing large files on the Cloud. As described in this paragraph, SparkBLAST can process much larger files when compared to CloudBLAST, due to a more efficient memory management.
|
other
| 83.06 |
The main reason behind the performance of SparkBLAST when compared to Hadoop-based systems are the in-memory operations and its related RDD abstraction. The reduced number of Disk IO operations by SparkBLAST results in a significant improvement on overall performance when compared to the Hadoop implementation.
|
other
| 99.9 |
It is clear that in-memory operations available in SparkBLAST plays a major role both in Speedup and Parallel Efficiency improvements and, as a consequence, also in the scalability of the system. Indeed, the main reason behind the fact that SparkBLAST, even when it allocates only half of nodes processing capacity, achieves performance figures that are superior of those of CloudBLAST is the reduced number of local I/O operations.
|
other
| 99.9 |
Another point to be highlighted is the scalability of SparkBLAST on a worldwide distributed platform such as Google Cloud. For the executions presented in this work, 64 nodes were deployed in 13 zones and it was achieved a speedup of 41.78 in this highly distributed platform. Once again, in-memory operations is a major factor related to this performance.
|
other
| 99.9 |
For applications where the Reduce stage is not a bottleneck, which is the case for SparkBLAST, it is demonstrated in the literature that Spark is much faster than Hadoop. In , those authors state that, for this class of application, MapReduce Hadoop is much slower than Spark in task initialization and is less efficient in memory management. Indeed, the supplementary document “Execution Measurements of SparkBLAST and CloudBLAST”, available in the online version of this paper, presents several measurements performed during SparkBLAST and CloudBLAST executions on the Microsoft Azure Cloud. These measurements show that task initialization in SparkBLAST is considerably faster than CloudBLAST. It is also shown that SparkBLAST is more efficient in memory management than CloudBLAST. The effect of SparkBLAST’s more efficient memory management can be observed in Additional file 1: Figures S5 and S6 of the supplementary information document. These figures show that Hadoop needs to use more memory than Spark, while Spark can maintain a larger cache and less swap to execute. Both factors - task initialization and memory management - are determinant for the improved scalability of SparkBLAST.
|
study
| 99.94 |
Furthermore, CloudBLAST makes use of Hadoop Streaming. In , authors shown that the Hadoop Streaming mechanism used in CloudBLAST can decrease application performance because it makes use of OS pipes to transfer input data to the applications’ (in this case BLAST) standard input and from BLAST standard output to disk storage. Data input to BLAST is done by the option: “-inputreader org.apache.hadoop.streaming. StreamPatternRecordReader”, which send lines from the input file to BLAST one-by-one, which further degrades performance.
|
other
| 99.9 |
Regarding extended scalability over larger platforms than the ones considered in this paper, it should be highlighted that two authors of this paper have proposed a formal scalability analysis of MapReduce applications . In this analysis the authors prove that the most scalable MapReduce applications are reduceless applications, which is exactly the case of SparkBLAST. Indeed, Theorem 5.2 of states that the increase of amount of computation necessary for a reduceless Scalable MapReduce Computation (SMC) application to maintain a given isoefficiency level is proportional to the number of processors (nodes). This is the most scalable configuration over all scenarios analyzed in . Simulation results that goes up to 10000 nodes corroborate the limits stated in this and other theorems of .
|
other
| 87.4 |
Regarding Experiment 1 and RBH inference, we showed that our SparkBLAST results can be post-processed to infer shared genes, then generating added-value to the similarity analysis. That also means that RBH experiments using SparkBLAST are potentially scalable to many more genomes, and can be even used as part of other Blast-based homology inference software such as OrthoMCL . Considering Experiment 2, results indicate that 1.27% of the Bermuda metagenomics proteins and 1.4% of the Búzios metagenomic proteins represent potential homologs to the 10 radiation-resistant bacteria, and as far as we know no related studies have been published to date. Those potential homologs will be further investigated in another study.
|
study
| 100.0 |
In this paper we propose SparkBLAST, a parallelization of BLAST that employs cloud computing for the provisioning of computational resources and Apache Spark as the coordination framework. SparkBLAST outperforms CloudBLAST, a Hadoop-based implementation, in speedup, efficiency and scalability in a highly distributed cloud platform. The superior performance of SparkBLAST is mainly due to the in-memory operations available through the Spark framework, consequently reducing the number of local I/O operations required for distributed BLAST processing.
|
other
| 99.9 |
Chicken anaemia virus (CAV) is a common viral agent in chickens worldwide. CAV belongs to the genus Gyrovirus of the Anelloviridae, which have characteristics of circular single-stranded DNA viruses . This virus frequently results in immunosuppression and anaemia in young chickens due to the destruction of T lymphoid tissue and aplasia of bone marrow, respectively, during virus infection [2–5]. Over 55% of the mortality rate and 80% of the morbidity rate were reported once the chicks were infected with CAV . Therefore, determining how to prevent CAV infection in the poultry industry has becomes an important challenge. The CAV virion lacks an envelope around its capsid coat, and it shows significantly high resistance to environmental stress or chemical agents. Currently, an attenuated live vaccine is available and effective for immunization of chickens for controlling CAV infection. However, young chicks less than 2 weeks-old are susceptible to CAV infection when the live vaccine was used . This consequence has led to the development of a subunit vaccine, including DNA or protein based vaccine, over the past decade.
|
review
| 97.25 |
CAV is a relatively small virus approximately 23 nm in diameter. A total of three open reading frames (ORFs) were involved in the viral genome and have a length of 2.3 kb . These ORFs respectively encode a 51 kDa VP1 protein, a 28 kDa VP2 protein and a 13 kDa VP3 protein. VP2 has dual-specificity phosphatase activity. VP3 is also referred to as apoptin with apoptosis-inducing activity. VP1 is the sole structural protein, which is the major component responsible for capsid assembly [5, 7, 8]. Currently, VP2 and VP3 proteins have been the focus of investigations of virus pathogenicity [9, 10]. In addition to its importance in the viral life cycle, VP3 has also demonstrated apoptosis-inducing activity as well as medical applications for anti-cancer treatments for humans in many previous studies [11–14]. VP1 can interact with VP2 and then significantly elicit the production of virus-neutralizing antibodies in the host in terms of immunogenicity studies . Therefore, VP1 is thought to be a good candidate for an immunogen to develop a subunit vaccine .
|
study
| 99.9 |
DNA replication of DNA viruses usually occurs in the nucleus of infected cells. Thus, to establish a productive infection, viral DNA with a high molecular weight needs to cross the nuclear envelope through protein-mediated nuclear transportation after infecting the cells . Approximately 90% of karyophilic proteins containing nuclear localization signals (NLSs) are directed to the nucleus. The NLS sequences usually overlap with the DNA binding domains. Therefore, proteins for nuclear transport possess both DNA binding and NLS activities [17, 18]. CAV is first Gyrovirus to be discovered and isolated . By cloning and sequencing the viral genome, previous studies have reported an N-terminal 40 amino acid sequence within the predicted amino acid sequence of VP1 that demonstrated a significant (46%) degree of similarity to the protamine protein in Japanese quails. This specific region within the N-terminus of VP1 contains high arginine content and might confer an ability to VP1 to bind and protect DNA . Using online software, including PSORT II (http://psort.hgc.jp) and DP-Bind (lcg.rit.albany.edu/dp-bind/), the VP1 protein was analysed in this study. A total of four putative DNA-binding motifs and two putative NLSs were found and predicted within the CAV VP1, as illustrated in Fig. 1. A previous researcher reported that transient expression of GFP-VP1 in the plant cells has been observed throughout the nucleoplasm . This outcome demonstrated that VP1 protein might be a nuclear protein. Other circular single-stranded DNA virus, such as duck circovirus (DuCV) and beak and feather disease virus (BFDV), have exhibited a pattern of N-terminal amino acid residues within the capsid protein that are highly basic amino acid rich sequences with nuclear localization signals and DNA binding activity [21, 22]. Based on these findings, N-terminal amino acid residues within the capsid protein of circovirus are very similar to the CAV of Gyrovirus. However, there is still a lack of direct evidence to prove and characterize the DNA binding ability or nuclear localization activity of VP1.Fig. 1Prediction results for putative NLS, NES and DNA-binding motifs on CAV VP1 proteins. a Schematic diagram representing the distribution regions of putative functional motifs: nuclear localization signals (NLS), nuclear export signals (NES) and DNA-binding motifs on CAV VP1. Two NLS that separate at the N-terminus of VP1 were predicted by PSORT II software (http://psort.hgc.jp/form2.html) and three NES, which are mainly at the C-terminus of VP1 were predicted by NetNES 1.1 Server software (http://www.cbs.dtu.dk/services/NetNES/), respectively. The software DP-Bind (lcg.rit.albany.edu/dp-bind) was used for putative DNA-binding motif prediction, and the predicted amino acid sequences were described on the diagram. b Predicted amino acid sequence results of NLS and NES for CAV VP1 were indicated by red (NLS)- and green (NES)-labels, respectively
|
study
| 100.0 |
Prediction results for putative NLS, NES and DNA-binding motifs on CAV VP1 proteins. a Schematic diagram representing the distribution regions of putative functional motifs: nuclear localization signals (NLS), nuclear export signals (NES) and DNA-binding motifs on CAV VP1. Two NLS that separate at the N-terminus of VP1 were predicted by PSORT II software (http://psort.hgc.jp/form2.html) and three NES, which are mainly at the C-terminus of VP1 were predicted by NetNES 1.1 Server software (http://www.cbs.dtu.dk/services/NetNES/), respectively. The software DP-Bind (lcg.rit.albany.edu/dp-bind) was used for putative DNA-binding motif prediction, and the predicted amino acid sequences were described on the diagram. b Predicted amino acid sequence results of NLS and NES for CAV VP1 were indicated by red (NLS)- and green (NES)-labels, respectively
|
study
| 100.0 |
In this study, to gain insight into the role of the capsid protein VP1 in the life cycle of CAV, we have investigated the physical interactions of CAV VP1 with the viral DNA. A recombinant E. coli expression system was used to express the recombinant VP1 of CAV following our previous study . The intracellular localization of the CAV VP1 was observed in MDCC-MSB1 cells or CHO-K1 cells using fluorescent green protein in the nucleoplasmic compartment. The DNA-binding activity of VP1 was also systemically examined. To the best of our knowledge, this is the first report to verify the DNA binding activity of the CAV capsid protein, VP1.
|
study
| 100.0 |
Previous studies have shown that only the VP1 protein is located in the CAV virion. Thus, VP1 is also thought to be a DNA-binding protein that is responsible for the encapsidation of a viral genome during virus assembly. Presently, VP3 is the only one of three CAV viral proteins that has exhibited DNA binding activity in previous studies . However, the function of VP1 on nucleic acid binding is still unknown. To gain insight into the role of the capsid protein VP1 in DNA binding, the bioinformatics software DP-Bind (http://lcg.rit.albany.edu/dp-bind/) was applied to analyse the features of DNA binding motifs within the amino acid sequence of the VP1 protein. Computational results of the DNA binding motif from the VP1 protein are shown in Fig. 1. Four potential DNA binding motifs were predicted by the DP-Bind program, and the putative motif position spanned from amino acids residues 3 to 22, 27 to 47, 62 to 67 and 333 to 349. According to these predicted results, VP1 might be having potential activity to bind DNA molecules. However, further investigation is still needed to verify the DNA binding activity of VP1.
|
study
| 100.0 |
To examine the DNA-binding activity of VP1, E. coli was used to express recombinant VP1 protein. Recombinant VP3 protein was also expressed as a positive control for the evaluation of DNA-binding activity. As shown in Fig. 2a, after purification by a GST affinity column, the purity and antigenicity of purified GST-fused VP1 and VP3 were determined by SDS-PAGE and Western blotting, respectively. This result confirmed the integrity of the two recombinant proteins.Fig. 2The VP1 protein has DNA-binding ability with no sequence specificity. The recombinant GST and GST-fused proteins were prepared by E. coli overexpression and purified through GST affinity chromatography. Purified results were analysed by SDS-PAGE with Coomassie blue staining and Western blotting with an anti-GST monoclonal antibody or anti-C-ter-VP1 polyclonal antibody (a). The purified proteins were used for DNA binding ability by an agarose gel shift assay with different DNA sequences of plasmid preparation of pcDNA3.1 (b), of the pGEM-T easy vector (c), and of pCAV containing the whole CAV genome (d). The binding activity of the VP1 protein was determined by comparing the existence of DNA fragments for the protein-DNA complex and DNA patterns from the blank (no-protein used), negative control (GST only) and positive control (GST-VP3). To confirm the observed DNA migration results that were induced by bound recombinant proteins, the protein-DNA experimental samples were mixed with 1% SDS as a protein denaturant (underline lane-labelled 1% SDS). Lane M, DNA ladder marker. Bold triangles indicated the protein-DNA complex formed by tested proteins and plasmids. Asterisks indicated the two conformations of plasmid DNA, including the relaxed form (Form I), and another was the supercoiled form (Form II)
|
study
| 100.0 |
The VP1 protein has DNA-binding ability with no sequence specificity. The recombinant GST and GST-fused proteins were prepared by E. coli overexpression and purified through GST affinity chromatography. Purified results were analysed by SDS-PAGE with Coomassie blue staining and Western blotting with an anti-GST monoclonal antibody or anti-C-ter-VP1 polyclonal antibody (a). The purified proteins were used for DNA binding ability by an agarose gel shift assay with different DNA sequences of plasmid preparation of pcDNA3.1 (b), of the pGEM-T easy vector (c), and of pCAV containing the whole CAV genome (d). The binding activity of the VP1 protein was determined by comparing the existence of DNA fragments for the protein-DNA complex and DNA patterns from the blank (no-protein used), negative control (GST only) and positive control (GST-VP3). To confirm the observed DNA migration results that were induced by bound recombinant proteins, the protein-DNA experimental samples were mixed with 1% SDS as a protein denaturant (underline lane-labelled 1% SDS). Lane M, DNA ladder marker. Bold triangles indicated the protein-DNA complex formed by tested proteins and plasmids. Asterisks indicated the two conformations of plasmid DNA, including the relaxed form (Form I), and another was the supercoiled form (Form II)
|
study
| 100.0 |
To elucidate whether the CAV VP1 protein is a DNA-binding protein, purified recombinant VP1 protein was added to circular dsDNA, pCAV, pcDNA3.1 and pGEM-T plasmids and incubated for 1 hour under 37 °C. After incubation, the occurrence of protein-DNA interaction was analysed by DNA movement on agarose gel. As illustrated in Fig. 2b, c and d, the migrations of pCAV, pcDNA3.1 and pGEM-T plasmid on the agarose gel were significantly reduced and shifted towards a pattern with a higher molecular weight. This result is very similar to the reduction in DNA migration that arose from binding VP3 to DNA (Fig. 2b, c and d). In contrast, no reduction in DNA migration occurred when the VP1 protein was absent or when GST protein was loaded with the addition of circular dsDNA plasmid. Moreover, when VP1 protein was pre-treated with 1% sodium dodecyl sulfate (SDS), the denatured VP1 no longer had DNA binding activity (Fig. 2b, c and d).
|
study
| 100.0 |
With respect to pCAV, which is a pcDNA3.1 plasmid carrying the entire CAV genome, the VP1 protein also displayed its DNA binding activity in terms of altered DNA migration pattern, as illustrated in Fig. 2b. In other words, regardless of whether the plasmid DNA used in the protein-DNA binding reaction was from pcDNA3.1 or pCAV, there was no significant effect on the resulting pattern of DNA migration (Fig. 2b, c and d). However, it is worth noting that the DNA migration of the circular dsDNA plasmid in the agarose gel displayed open circular dsDNA (form I, with a higher molecular weight pattern) and supercoiled dsDNA (form II, with a lower molecular weight pattern), simultaneously (Fig. 2b). The VP1 protein was bound to supercoiled dsDNA. which demonstrated that DNA shifting was more obvious than open circular dsDNA.
|
study
| 100.0 |
Next, to confirm that VP1 not only has DNA binding activity but also has nuclear localization activity, we constructed a transit expression plasmid, pEGFP-VP1, which is a pcDNA3.1 vector carrying the VP1 gene fused to a GFP gene for cell transfection (Fig. 3a). When pEGFP-VP1 was respectively transfected into chicken lymphocytes, MDCC-MSB1 cells and Chinese Hamster Ovary (CHO) K1, the localization of GFP fluorescence was observed using confocal microscopy (Fig. 3b and c). As illustrated in Fig. 3b and c, GFP-VP1 and DAPI staining coincided significantly in the nuclei of MDCC-MSB1 cells (Fig. 3c). Additionally, GFP-VP1 was partially distributed and displayed in the cytoplasm of MDCC-MSB1 cells (Fig. 3c). A similar pattern of the distribution of GFP-VP1 was also presented in the CHO-K1 cells (Fig. 3b). These results clearly demonstrated VP1 is also a nuclear protein and distributed within the nucleocytoplasmic compartment. Taken together, these results indicated that CAV VP1 is a DNA-binding protein with nuclear localization activity, and its DNA binding is not specific to a particular sequence.Fig. 3The nucleocytoplasmic distribution characterization of VP1 protein in CHO-K1 and MDCC-MSB1 cells. To realize the subcellular distribution of VP1, the full-length VP1 gene included the fused whole EGFP gene at the 5′-terminus to generate the EGFP-VP1 expressing plasmid pEGFP-VP1 as illustrated in a schematic diagram (a). After 48 h post-transfection with the above plasmid in CHO-K1 cells (b) and MDCC-MSB1 cells (c), both cell types were fixed and stained with DAPI to reveal the nuclei. The subcellular localization of VP1 was determined by green fluorescence detection through confocal fluorescence microscopy
|
study
| 100.0 |
The nucleocytoplasmic distribution characterization of VP1 protein in CHO-K1 and MDCC-MSB1 cells. To realize the subcellular distribution of VP1, the full-length VP1 gene included the fused whole EGFP gene at the 5′-terminus to generate the EGFP-VP1 expressing plasmid pEGFP-VP1 as illustrated in a schematic diagram (a). After 48 h post-transfection with the above plasmid in CHO-K1 cells (b) and MDCC-MSB1 cells (c), both cell types were fixed and stained with DAPI to reveal the nuclei. The subcellular localization of VP1 was determined by green fluorescence detection through confocal fluorescence microscopy
|
study
| 100.0 |
According to the results in Fig. 2b, the supercoiled dsDNA seems to interact with the VP1 protein more than opened dsDNA. Thus, to further address whether the DNA-binding activity of VP1 is affected by DNA conformation, various species of DNA molecules, such as linear dsDNA, circular ssDNA and linear ssDNA were used to confirm VP1 DNA binding activity. Because the DNA binding activity of VP1 has no sequence specificity as illustrated in Fig. 2b, c and d, the commercial circular single-stranded genome of the M13 phage was used as sample DNA instead of the real circular ssDNA genome of CAV. As illustrated in Fig. 4 with respect to all DNA species, DNA retardation occurred when the recombinant VP1 proteins were added to DNA molecules, such as linear pcDNA3.1 (linear dsDNA, Fig. 4a), the linear single strand of the CAV genome (minus and strand, Fig. 4b) and the genome of the M13 phage (circular ssDNA, Fig. 4c). Similarly, with respect to the VP3 protein, all kinds of DNA molecules were also bound by VP3 and reduced the migration of DNA. However, comparing the significance of DNA patterns between various protein-DNA complexes, different DNA molecules bound by VP1 protein demonstrated there were distinct migration patterns of DNA (Fig. 4). Therefore, to address the binding preferences of the VP1 protein to DNA molecules, various amounts of VP1 protein were added to equal amounts of different DNA molecules for analysis of protein-DNA interactions. By quantifying the amount of VP1 protein required for DNA binding with respect to pCAV (circular dsDNA), especially for its supercoiled form dsDNA, the results showed 200 μg of VP1 were required to initiate VP1 binding to DNA molecules (Fig. 5a, supercoiled, form II). At least 300 μg of VP1 were required for this interaction to occur between VP1 and opened dsDNA (Fig. 5a, opened, form I). Higher amounts of VP1 protein were used to bind circular dsDNA and reduced DNA migration patterns more significantly (Fig. 5a). Similarly, other DNA molecules, such as linear dsDNA (linearized pcDNA3.1, Fig. 5b), a linear single strand of the CAV genome (minus strand, Fig. 6a) and circular ssDNA (genome of the M13 phage, Fig. 6b), showed a similar pattern for protein-DNA interactions in the reaction mixture, with approximately 300 μg of VP1 required for linear dsDNA, 200 μg for the minus strand of linear ssDNA and 100 μg for circular ssDNA. In contrast, residual unbound DNA molecules representing the amount of DNA binding on the gel decreased if protein-DNA interaction occurred. Based on these results, the preferences in order of affinity to DNA with the VP1 protein in terms of the estimation of the percentage of unbound DNA molecules, which were sorted from low to high, were circular ssDNA (46.5% with respect to 300 μg of VP1), the linear minus strand of ssDNA (53.3% with respect to 300 μg of VP1), supercoiled circular dsDNA (82.6% with respect to 300 μg of VP1), linear dsDNA (84.7% with respect to 300 μg of VP1) and opened circular dsDNA (100% with respect to 300 μg of VP1). The comparison of binding preferences of VP1 protein to different conformations of DNA molecules are summarized in Table 1. More unbound DNA existed in the agarose gel, implying that this conformation of DNA exhibited a lower preference to interact with the VP1 protein. These results demonstrated that the interaction of VP1 with the DNA molecule exhibited various binding preferences that were dependent on the structural conformation of DNA.Fig. 4VP1 protein binds to various DNA molecules. Purified GST and GST-fused proteins were used for analysing the interaction of recombinant proteins with various DNA samples, such as linear dsDNA (a), minus-strand ssDNA (b), and M13mp18 phage DNA (c). All DNA samples were generated by different preparations as described in the Materials and Methods. After the agarose gel shift assay, the DNA fragment signals were observed by EtBr staining. The 1% SDS (underline lane-labelled 1% SDS) was also used to confirm the retardation caused by tested proteins. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by the tested protein and DNA molecules. The “pcDNA3.1 x EcoR I” indicated generation of the linear form of pcDNA3.1 DNA digested by EcoR IFig. 5Dose-dependent analysis of VP1-dsDNA binding ability. Various concentrations of GST-VP1 were used to perform the dose-dependent analysis with a consistent concentration of plasmid pCAV (a) or linear dsDNA (b). The DNA fragment signals of the protein-DNA complex and differences in DNA migration patterns were more significant as the protein amount increased. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by the tested protein and plasmids. Asterisks indicate the two conformations of plasmid DNAs, including the relaxed form (Form I) and a supercoiled form (Form II). The "pcDNA3.1xEcoRI" indicated generation of the linear form of pcDNA3.1 DNA digested by EcoRI Fig. 6Dose-dependent analysis of VP1-ssDNA binding ability. Increased concentrations of GST-VP1 were incubated with consistent concentrations of minus-strand ssDNA (a) or circular ssDNA (b) to perform dose-dependent analysis. The disappearance of free DNA signals was more obvious when the protein amount increased. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by tested proteins and DNA moleculesTable 1The ratio of unbound free DNA residue was determined by calculating the image intensity of free DNA fragments on an electrophoretic agarose gel after a VP1-DNA binding assay combining increasing amounts of recombinant GST-VP1 protein with certain DNA molecule conformationsConcentrations of GST-VP1 protein (ng)0100300500700Open circular dsDNA (Form I)100%100%100%59.6%11%Supercoiled dsDNA (Form II)100%100%82.6%0%0%Linear dsDNA100%93.8%84.7%36.5%0%Linear ssDNA (−)100%92.9%53.3%0%0%M13 phage DNA100%90%46.5%0%0%The lower ratios indicate a higher preference of the VP1 protein for specific DNA conformations. The equation used in this study is presented under the table
|
study
| 100.0 |
VP1 protein binds to various DNA molecules. Purified GST and GST-fused proteins were used for analysing the interaction of recombinant proteins with various DNA samples, such as linear dsDNA (a), minus-strand ssDNA (b), and M13mp18 phage DNA (c). All DNA samples were generated by different preparations as described in the Materials and Methods. After the agarose gel shift assay, the DNA fragment signals were observed by EtBr staining. The 1% SDS (underline lane-labelled 1% SDS) was also used to confirm the retardation caused by tested proteins. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by the tested protein and DNA molecules. The “pcDNA3.1 x EcoR I” indicated generation of the linear form of pcDNA3.1 DNA digested by EcoR I
|
study
| 100.0 |
Dose-dependent analysis of VP1-dsDNA binding ability. Various concentrations of GST-VP1 were used to perform the dose-dependent analysis with a consistent concentration of plasmid pCAV (a) or linear dsDNA (b). The DNA fragment signals of the protein-DNA complex and differences in DNA migration patterns were more significant as the protein amount increased. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by the tested protein and plasmids. Asterisks indicate the two conformations of plasmid DNAs, including the relaxed form (Form I) and a supercoiled form (Form II). The "pcDNA3.1xEcoRI" indicated generation of the linear form of pcDNA3.1 DNA digested by EcoRI
|
study
| 100.0 |
Dose-dependent analysis of VP1-ssDNA binding ability. Increased concentrations of GST-VP1 were incubated with consistent concentrations of minus-strand ssDNA (a) or circular ssDNA (b) to perform dose-dependent analysis. The disappearance of free DNA signals was more obvious when the protein amount increased. Lane M, DNA ladder marker. Bold triangles indicate the protein-DNA complex formed by tested proteins and DNA molecules
|
study
| 100.0 |
The ratio of unbound free DNA residue was determined by calculating the image intensity of free DNA fragments on an electrophoretic agarose gel after a VP1-DNA binding assay combining increasing amounts of recombinant GST-VP1 protein with certain DNA molecule conformations
|
study
| 100.0 |
DNA replication in the DNA viruses usually occurs in the nuclei of infected cells. Thus, viral DNA needs to cross the nuclear envelope through protein-mediated nuclear transportation to establish a productive infection after infecting the cells . CAV is a non-enveloped, small DNA virus containing a circular ssDNA genome . VP1, which is a major capsid protein of CAV, interacts with cells during virus infection, and the viral genome should theoretically be carried by VP1 to enter cells. The main question is what protein is conferred a function to direct a viral genome into the nucleus for sequential DNA replication? VP1 is thought to have a functional role to bind and direct the viral genome into the nucleus. However, there is still a lack of direct evidence to support this speculation. In our results from computational prediction, four putative DNA binding motifs were combined with the amino acid sequence of the CAV VP1 protein. Although the exact DNA binding motif was not determined in this study, it did not affect the characterization of VP1 DNA binding activity. After performing a DNA-binding assay to examine DNA migration, the VP1 protein of CAV was confirmed to have DNA-binding activity. Additional experiments are still needed for identifying major DNA binding motifs within VP1 protein. The confirmed DNA-binding activity in VP1 might be useful to further verify the underlying mechanisms of viral DNA replication. In addition, GFP-VP1 has demonstrated nucleo-cytoplasm shuttling activity (Fig. 3). The results imply VP1 is a nuclear protein to binds DNA molecules, such as those in the viral genome and travels into the nucleus during its early life cycle. In fact, we have not only predicted the presence of nuclear localization signals (NLSs) with PSORT II but also predicted nuclear exporting signals (NESs) with NetNES server (http://www.cbs.dtu.dk/services/NetNES/). These putative NLS motifs were within the amino acid sequence of VP1, which spanned from amino acid residues 3 to 19 (NLS1) and 24 to 47 (NLS2). For NESs, the putative motifs within the VP1 protein spanned from amino acid residues 76 to 84 (NES1), 109 to 119 (NES2) and 375 to 387 (NES3) (Fig. 1a). In terms of the observation of fluorescent of GFP-VP1, this result was confirmed according to computational predictions. In the early life cycle of the virus, viral DNA replication is an important stage for the establishment of productive infection . Previous studies have reported circular, negative ssDNA of the CAV genome might be replicated through rolling-circle amplification . During this stage, previous researchers isolated the viral replicative form (RF) DNA, which is an open circular dsDNA obtained from MDCC-MSB1 cells after being infected with the virus for 30 h . In addition to the presence of closed and open circular dsDNA, circular ssDNA with genome-sized and small linear dsDNA of 800 bp were observed in the later stages of DNA replication . In this study, all similar DNA molecules with different conformations, including linear ssDNA, which is derived from circular ssDNA, were used to examine the possible functional roles of the VP1 protein in DNA-binding activities. Additionally, other putative DNA binding motifs, especially for the initiation of rolling-circle amplification (RCR) within the VP1 protein, had also been reported and predicted in a previous study . Three putative motifs were proposed that spanned from amino acid 313 to 320 (FATLTALG), 350 to 358 (GQRWHTLVP), and 399 to 408 (TATYALKEPV). These motifs might be interaction sites, such as the origin (Ori) site of the CAV genome, for interacting with VP1 for regulating DNA replication .
|
study
| 100.0 |
Generally, the DNA binding protein of most DNA viruses showed DNA binding activity with no sequence specificity. In this study, VP1 showed DNA binding characteristics with no sequence specificity similar to other DNA viruses, such as human papillomavirus and polyomavirus [27, 28]. Therefore, some other strategy of VP1 binding to the CAV viral genome should be adopted by the virus. With respect to the binding preferences of the VP1 protein to different DNA molecule conformations, VP1 was found to interact with circular ssDNA and exhibit a higher preference for this conformation, as shown in Table 1. This outcome might truly reflect the conditions of viral encapsidation for coating circular ssDNA of the CAV genome. Actually, circular ssDNA is prone to forming secondary structures with a high probability. This possibility was examined and confirmed with a computational prediction from the Mfold program (http://unafold.rna.albany.edu/?q=mfold/DNA-Folding-Form) (Additional file 1: Figure S1A, B). Similarly, linear ssDNA also has a high probability for forming secondary DNA structures (Additional file 1: Figure S2A, B). Thus, the binding preference of VP1 to linear ssDNA is surpassed only by the preference for circular ssDNA (Table 1). This difference truly meets our expectations. In fact, the sequence of linear ssDNA(−) was complemented with linear ssDNA(+). Then, linear ssDNA(+) was found to have significant VP1-DNA interaction in terms of the results of the DNA migration assay. The binding preference of VP1 to the linear plus-strand of ssDNA(+) is slightly lower than the linear plus-strand of ssDNA(−) (53.1% for 300 μg of VP1) (Additional file 1: Figure S3A, B). Other DNA molecules, such as supercoiled dsDNA, linearized ds DNA and opened dsDNA, displayed lower binding preferences to VP1, which might be a mechanism for VP1 protein to competitively bind various DNA molecules during different stages of the life cycle.
|
study
| 100.0 |
In summary, the characterization of DNA binding activity of the VP1 protein was investigated in this study. VP1 was demonstrated to show DNA binding characteristics with no sequence specificity. In addition, the DNA binding activities of VP1 exhibited a differential preference to interact with various DNA molecules with different conformations. This information could be helpful for determining the biological roles of VP1 in the CAV viral life cycle.
|
study
| 100.0 |
For Chinese Hamster Ovary (CHO-K1) cells, cells were purchased in 2014 from the Bioresource Collection and Research Center (BCRC 6006) in Taiwan. CHO-K1 cells were maintained in Ham’s F12 medium (HyClone, USA) supplemented with 10% FBS (HyClone, USA), 1% P/S (Penicillin/Streptomycin solution) (Gibco, USA). Chicken lymphoblast MDCC-MSB1 cells were purchased from the CLS Cell Lines Service GmbH in Germany in 2015. MDCC-MSB1 cells were grown in RPMI 1640 medium (HyClone, USA) supplemented with 10% FBS (HyClone, USA) and 1% P/S (Penicillin/Streptomycin solution) (Gibco, USA). All cells were cultured in appropriate tissue culture flasks and maintained in a cell culture incubator with 5% CO2 at 37 °C before experiments.
|
study
| 99.94 |
All expression constructs used in this study were maintained in the E. coli strain Top10F’ (Invitrogen, USA). The E. coli strain BL21 (DE3)-pLysS was transformed with protein expression plasmids and followed by IPTG induction to produce recombinant proteins as described in a previous study .
|
study
| 100.0 |
The construction of an expressed plasmid used for detecting subcellular localization was described below. The full-length of CAV VP1 was amplified by PCR using the specific primer sets wt-VP1-f: 5’-CCCGAATTCATGGCAAGACGAGCTCGC-3′, wt-VP1-r: 5’-CGCGTCGACTCAGGGCTGCGTCCCCCAGTA-3′ from the CAV VP1 template that was kindly provided by Dr. Yi-Yang Lien. The PCR product was then cloned into expression vector pEGFP-C2 (#6083–1, Clontech, USA) between EcoR I and Sal I sites to generate a recombinant plasmid named pEGFP-VP1.
|
study
| 100.0 |
To purify the recombinant CAV VP1 and VP3 proteins, the previously created recombinant E. coli strains BL21 (DE3)-pLysS expressing VP1 and VP3 were used to express recombinant proteins . The recombinant E. coli cells were cultured, and the harvested cells were disrupted and prepared following a previously described procedure . Cells were spun down from 50 ml of culture supernatant and resuspended in GST resin binding buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.3). After cell disruption, the resulting cell supernatant was loaded onto a GSTrap FF affinity column (GE healthcare, Piscataway, NJ) for protein purification following the operational conditions described in a previous study . The total protein concentration of recombinant CAV VP1 and VP3 proteins was determined using a Micro BCA kit (Pierce, Rockford, IL) with bovine serum albumin as the reference protein. Purified VP1 and VP3 proteins were dialyzed against DNA-binding buffer (50 mM Tris-HCl, pH 7.5, 120 mM KCl, 1.0 mM EDTA, 0.5 mM DTT, and 30 mg/ml BSA) and analysed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting. Purified proteins were stored at − 20 °C until required.
|
study
| 100.0 |
Different DNA species, including circular dsDNA, linear dsDNA and circular ssDNA were used for assessing DNA-binding activity. The circular dsDNA, pcDNA3.1 (#V80020, Invitrogen, USA), pGEM-T easy vector (#A1360, Promega, USA) and pCAV were used for the DNA binding assay. The pCAV plasmid DNA was composed of a full-length Australian CAV strain CAU269/7 (GenBank #AF227982). The linear dsDNA was prepared from a pcDNA3.1 and pGEM-T plasmid by the cutting restriction enzyme EcoRI. Pure M13mp18 single-stranded DNA along with circular ssDNA materials were purchased from New England BioLabs (#N4040S, NEB, USA). All DNA molecules were diluted to 50 ng/ml with DNA-binding buffer and store at − 20 °C until required.
|
study
| 99.94 |
The linear ssDNA was also used for the DNA binding assay. The preparation of linear ssDNA followed the protocol described in Marimuthu et al. using the biotin-streptavidin separation method. First, a biotinylated DNA fragment containing the whole CAV genome was amplified by PCR using the EmeraldAmp Max PCR Master kit (Takara, Japan) from pCAV with a designed primer set, including the reverse biotinylated primer Biotin-CAV-r: biotin-labelled-GATTGT GCGGTGAACGAATTAG, and the forward regular primer CAV-f: GAATTCCGAGTGGTTACTATTC. After PCR amplification, the biotinylated PCR product was then immobilized on 40 μl Dynabeads M-280 Streptavidin magnetic beads (Invitrogen, USA) and incubated at 4 °C overnight. After washing the DNA-bonded beads twice with B/W buffer (5 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 1 M NaCl), the washed beads were incubated in 150 μl elution buffer (0.1 M NaOH, 1 mM EDTA, pH 13.0) to perform alkaline denaturation. Under the high alkaline environment, the desired non-biotinylated strand can be separated from the biotinylated strand and suspended in the supernatant. After magnet adsorption, the supernatants were collected, and the linear ssDNA was further purified by a PCR clean-up kit (Geneaid, Taiwan). The linear ssDNA was diluted to 50 ng/ml with DNA-binding buffer and store at − 20 °C until required.
|
study
| 100.0 |
Purified GST, GST-VP1 and GST-VP3 proteins were diluted to 500 ng/μl with DNA-binding buffer and 500 ng of proteins were mixed with 100 ng of each DNA variant in a total of 20 μl of DNA-binding buffer. Then, each mixture was incubated for 30 min under 37 °C. The resulting sample was subjected to electrophoresis using a 0.8% agarose gel in a TAE buffer and, then the DNA was stained with ethidium bromide for the analysis of DNA migration.
|
study
| 99.94 |
CHO-K1 cells were transfected by X-tremeGene HP DNA transfection reagent (Sigma, USA) according the manual’s protocol with a mixture containing 2 μg of plasmid pEGFP-VP1 and 4 μl of transfection reagent in 2.5 ml serum-free Opti-MEM medium (Gibco, USA). After incubating the mixture for 20 min at room temperature, the mixture was added drop by drop into cultured CHO-K1 cells in a 6-well plate. The 24 to 48 h post-transfection, the transfection effect was checked with a confocal fluorescent microscope.
|
study
| 99.94 |
For the transfection of MDCC-MSB1 cells, the 4 × 106 log-phase grown MDCC-MSB1 cells were gently pipetted with 15 μg of plasmid pEGFP-VP1 first in serum-free RPMI 1640 medium and then the mixture was transferred into a 0.4-cm gap electroporation cuvette and the cuvette was harvested on ice for 5 min. The electroporation of MDCC-MSB1 cells was performed with a Gene Pulser II (Bio-Rad, USA) with a Time Constant Protocol set at 34 ms and an operating voltage of 300 V. After electroporation, the transfected cells were then cultured into complete medium in a 6-well plate for 24 to 48 h. Post-transfection, the expression of recombinant EGFP-VP1 proteins were analysed by confocal fluorescence microscopy to make sure the transfection was effective.
|
study
| 100.0 |
After transfecting CHO-K1 cells and MDCC-MSB1 cells with plasmid pEGFP-C2 or pEGFP-VP1, the fluorescent images were captured by a confocal fluorescence microscope in terms of the observation of protein fluorescent to verify the EGFP-expressed cells and EGFP-VP1-expressed cells. Transfected cells were collected and fixed with 4% formaldehyde in the dark. After washing the fixed cells twice to remove residual formaldehyde, the cells were stained in 0.1% PBS-T with 1 μg/ml DAPI for 5 min at 37 °C in the dark. Then, the stained cells were mounted with gelvatol medium (Sigma, USA) on a glass slide for confocal observation. Confocal laser scanning microscope (CLSM) images were captured from a Leica TCS SP8 confocal microscope and the images were integrated with LAS X Leica Confocal Software. EGFP fluorescence was observed through excitation at 488 nm and DAPI emitted blue fluorescence upon binding to DNA that was observed through excitation by UV light.
|
study
| 99.94 |
To obtain the ratio of unbound DNA residues, a dose-dependent DNA-binding experiment was performed by combining various concentrations (0 to 700 ng) of GST-VP1 proteins with consistent amounts of different DNA variants. Next, the signal intensities of free DNA from DNA migration images after DNA-binding experiments were obtained using ImageJ software observation. The equation used in this study is presented below.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \mathrm{Unbound}\ \mathrm{DNA}\ \mathrm{residue}\ \left(\%\right)=\frac{\mathrm{Image}\ \mathrm{intensity}\kern0.17em \mathrm{of}\kern0.17em \mathrm{free}\;\mathrm{DNA}\;\left(\mathrm{certain}\kern0.17em \mathrm{protein}\kern0.17em \mathrm{concentration}\right)}{\mathrm{Reference}\kern0.17em \mathrm{image}\kern0.17em \mathrm{intensity}\;\left(\mathrm{no}\;\mathrm{protein}\right)}\times 100\% $$\end{document}UnboundDNAresidue%=Imageintensity of freeDNAcertain protein concentrationReference image intensitynoprotein×100%
|
study
| 100.0 |
After dividing the signal intensity at a certain concentration of GST-VP1 proteins by the signal intensity in the absence of proteins (blank), the calculated ratio of unbound DNA residue was obtained to determine the DNA conformational preference for the VP1 protein.
|
study
| 100.0 |
Additional file 1:Figure S1. The putative secondary structure of the circular CAV genome as predicted by the software Mfold. Figure S2. The putative secondary structure of linear CAV genome. Figure S3. VP1 binds to linear plus-strand ssDNA. Purified GST and GST-fused recombinant proteins were used to analyse the interaction with linear plus-strand ssDNA in an agarose gel shift assay. (DOCX 1965 kb)
|
study
| 100.0 |
Figure S1. The putative secondary structure of the circular CAV genome as predicted by the software Mfold. Figure S2. The putative secondary structure of linear CAV genome. Figure S3. VP1 binds to linear plus-strand ssDNA. Purified GST and GST-fused recombinant proteins were used to analyse the interaction with linear plus-strand ssDNA in an agarose gel shift assay. (DOCX 1965 kb)
|
study
| 100.0 |
Metallic catalysts are involved in 80% of the industrial catalytic processes . These catalysts are of great importance in various fields, such as synthesis chemistry, energy production, but also, environment processes . Among all transition metals, noble metals (or platinum group metals), such as Pd, Ir, and Rh, are of particular interest as catalysts for large scale industrial applications. A non-exhaustive list of applications for Pd include hydrogenation or Suzuki cross-coupling reactions . Rh is commonly used in the preparation of catalysts for the reduction of NOx in automotive applications , and hydrogen production by steam reforming . Iridium is generally used as a catalyst for propulsion applications or ring opening reactions . In catalysis, the activity of catalysts is currently expressed in the literature by the turnover frequency (TOF), exhibiting the activity per active site. In catalysis by metals, the mean metal particle size and the dispersion are required to be known precisely, to determine the TOF.
|
review
| 99.6 |
The hydrogen chemisorption on noble face center cubic (fcc) metals (such as Pt, Pd, Ir, and Rh) is one of the most employed characterization techniques used to determine essential parameters in catalysis, such as metallic accessibility (dispersion), particle size, as well as metallic specific surface area, exposed mostly due to its ease of implementation .
|
other
| 99.44 |
The principle of this technique is to quantify the amount of hydrogen atoms chemisorbed on an atom located on the metal surface (MS) according to the following reaction (R1): (R1)MS+αH2→MS(H)2α where 2α represents the chemisorption stoichiometric factor of H atoms chemisorbed over the number of metal atoms located on the surface of the metallic cluster, which is defined by Equation (1): (1)2α=HMS
|
study
| 99.94 |
If the chemisorption stoichiometric factor 2α is known, the dispersion (D(%)) from H2 chemisorption measurements may be estimated, using the following equation (Equation (2)): (2)D(%)=12α×HM×100=MSM×100 where H/M represents the number of chemisorbed hydrogen atoms per total metal atoms.
|
study
| 99.94 |
Provided that some assumptions are made on chemisorption stoichiometric factor (H/MS) and the nature of atomic planes exposed on the surface, the particle size (d(nm)) and the metallic specific surface area (SM) of noble fcc metals catalysts can be obtained . The common assumption is that the values of H/MS = 1 for Pt, Pd, Ir, and Rh metals . However, some data also report H/MS stoichiometry factor exceeding unity for Pt, Pd, Rh, and Ir supported catalysts. For instance, data compiled by Bartholomew show chemisorption stoichiometric factor (H/MS) values of 1.0–1.2 for Pt, Pd, Rh, and Ir catalysts Kip et al. performed careful characterization of supported platinum, rhodium, and iridium catalysts by hydrogen chemisorption and EXAFS data analysis. They reported H/M ratios exceeding unity for Pt (H/Pt = 1.14) and Rh (H/Rh = 1.98), and even higher than 2 for Ir (H/Ir = 2.68) over highly dispersed metal catalysts supported on Al2O3 and SiO2 . McVicker et al. reported a H/Ir ratio close to 2 for small particle sizes (<0.6 nm) over highly dispersed Ir catalysts on Al2O3 . Krishnamurthy et al. have shown that 0.48 wt% Ir/Al2O3 catalyst adsorbed up to 2.72 hydrogen atoms per iridium atom .
|
study
| 98.0 |
Several explanations have been proposed for H/MS ratios higher than unity, such as (i) spillover of H atoms from the metal to the support , (ii) hydride formation , (iii) the support ionicity (with zeolite) or (iv) multiple adsorption on corners and edges for small metal particles .
|
study
| 99.94 |
In a previous work , we demonstrated that the multiple adsorption assumption is consistent with the H2 chemisorption literature data for the Pt catalysts . For this purpose, a model describing the statistics of the surface atoms and sites (top, bridge, hollow) on perfect cuboctahedron clusters was developed. This model allowed us to assess values of D(%), d and SPt, assuming the most favorable adsorption sites based on DFT calculation from the literature . Thus, it successfully predicted, precisely, the H/PtS stoichiometry, which ranges from 1 to 2 for the smallest cluster (dPt = 0.7 nm), and the experimental values of D, d, and SPt determined from H2 chemisorption data. A set of simple equations was provided for the accurate determination of these parameters from chemisorption experiments on Pt. This approach, based on the combination of identification and quantification of adsorption sites for a given cluster shape, is expected to be valid for other fcc metals, such as Pd, Rh, and Ir.
|
study
| 100.0 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.