Deep Learning for Automated Segmentation of Liver Lesions at CT in Patients with Colorectal Cancer Liver Metastases.

PURPOSE: To evaluate the performance, agreement, and efficiency of a fully convolutional network (FCN) for liver lesion detection and segmentation at CT examinations in patients with colorectal liver metastases (CLMs). MATERIALS AND METHODS: This retrospective study evaluated an automated method using an FCN that was trained, validated, and tested with 115, 15, and 26 contrast material-enhanced CT examinations containing 261, 22, and 105 lesions, respectively. Manual detection and segmentation by a radiologist was the reference standard. Performance of fully automated and user-corrected segmentations was compared with that of manual segmentations. The interuser agreement and interaction time of manual and user-corrected segmentations were assessed. Analyses included sensitivity and positive predictive value of detection, segmentation accuracy, Cohen κ, Bland-Altman analyses, and analysis of variance. RESULTS: In the test cohort, for lesion size smaller than 10 mm (n = 30), 10-20 mm (n = 35), and larger than 20 mm (n = 40), the detection sensitivity of the automated method was 10%, 71%, and 85%; positive predictive value was 25%, 83%, and 94%; Dice similarity coefficient was 0.14, 0.53, and 0.68; maximum symmetric surface distance was 5.2, 6.0, and 10.4 mm; and average symmetric surface distance was 2.7, 1.7, and 2.8 mm, respectively. For manual and user-corrected segmentation, κ values were 0.42 (95% confidence interval: 0.24, 0.63) and 0.52 (95% confidence interval: 0.36, 0.72); normalized interreader agreement for lesion volume was -0.10 ± 0.07 (95% confidence interval) and -0.10 ± 0.08; and mean interaction time was 7.7 minutes ± 2.4 (standard deviation) and 4.8 minutes ± 2.1 (P < .001), respectively. CONCLUSION: Automated detection and segmentation of CLM by using deep learning with convolutional neural networks, when manually corrected, improved efficiency but did not substantially change agreement on volumetric measurements.© RSNA, 2019Supplemental material is available for this article.

Absolute Regulatory Small Noncoding RNA Concentration and Decay Rates Measurements in Escherichia coli.

Regulation of RNA turnover is of utmost importance for controlling the concentration of transcripts and consequently cellular protein levels. Among the processes controlling RNA decay, small noncoding regulatory RNAs (sRNAs) have recently emerged as major new players. In this chapter, we describe and discuss protocols that can be used to measure sRNA concentration in vivo and to assess sRNA decay rates in Gram-negative bacteria. Precisely, we focus our analyses on the Escherichia coli Gram-negative bacterium as a model. The information described in this chapter provides a guideline to help develop a protocol in order to assess these important parameters and to identify RNA-processing enzymes involved in sRNA degradation processes.

Astaxanthin from Haematococcus pluvialis Prevents Oxidative Stress on Human Endothelial Cells without Toxicity.

Astaxanthin, a powerful antioxidant, is a good candidate for the prevention of intracellular oxidative stress. The aim of the study was to compare the antioxidant activity of astaxanthin present in two natural extracts from Haematococcus pluvialis, a microalgae strain, with that of synthetic astaxanthin. Natural extracts were obtained either by solvent or supercritical extraction methods. UV, HPLC-DAD and (HPLC-(atmospheric pressure chemical ionization (APCI)+)/ion trap-MS) characterizations of both natural extracts showed similar compositions of carotenoids, but different percentages in free astaxanthin and its ester derivatives. The Trolox equivalent antioxidant capacity (TEAC) assay showed that natural extracts containing esters displayed stronger antioxidant activities than free astaxanthin. Their antioxidant capacities to inhibit intracellular oxidative stress were then evaluated on HUVEC cells. The intracellular antioxidant activity in natural extracts was approximately 90-times higher than synthetic astaxanthin (5 µM). No modification, neither in the morphology nor in the viability, of vascular human cells was observed by in vitro biocompatibility study up to 10 µM astaxanthin concentrations. Therefore, these results revealed the therapeutic potential of the natural extracts in vascular human cell protection against oxidative stress without toxicity, which could be exploited in prevention and/or treatment of cardiovascular diseases.

Pretest probability of a normal echocardiography: validation of a simple and practical algorithm for routine use.

BACKGROUND: Management of increased referrals for transthoracic echocardiography (TTE) examinations is a challenge. Patients with normal TTE examinations take less time to explore than those with heart abnormalities. A reliable method for assessing pretest probability of a normal TTE may optimize management of requests. AIM: To establish and validate, based on requests for examinations, a simple algorithm for defining pretest probability of a normal TTE. METHODS: In a retrospective phase, factors associated with normality were investigated and an algorithm was designed. In a prospective phase, patients were classified in accordance with the algorithm as being at high or low probability of having a normal TTE. RESULTS: In the retrospective phase, 42% of 618 examinations were normal. In multivariable analysis, age and absence of cardiac history were associated to normality. Low pretest probability of normal TTE was defined by known cardiac history or, in case of doubt about cardiac history, by age>70 years. In the prospective phase, the prevalences of normality were 72% and 25% in high (n=167) and low (n=241) pretest probability of normality groups, respectively. The mean duration of normal examinations was significantly shorter than abnormal examinations (13.8 ± 9.2 min vs 17.6 ± 11.1 min; P=0.0003). CONCLUSION: A simple algorithm can classify patients referred for TTE as being at high or low pretest probability of having a normal examination. This algorithm might help to optimize management of requests in routine practice.

The interplay of Hfq, poly(A) polymerase I and exoribonucleases at the 3' ends of RNAs resulting from Rho-independent termination: A tentative model.

Discovered in eukaryotes as a modification essential for mRNA function, polyadenylation was then identified as a means used by all cells to destabilize RNA. In Escherichia coli, most accessible 3' RNA extremities are believed to be potential targets of poly(A) polymerase I. However, some RNAs might be preferentially adenylated. After a short statement of the current knowledge of poly(A) metabolism, we discuss how Hfq could affect recognition and polyadenylation of RNA terminated by Rho-independent terminators. Comparison of RNA terminus leads to the proposal that RNAs harboring 3' terminal features required for Hfq binding are not polyadenylated, whereas those lacking these structural elements can gain the oligo(A) tails that initiate exonucleolytic degradation. We also speculate that Hfq stimulates the synthesis of longer tails that could be used as Hfq-binding sites involved in non-characterized functions of Hfq-dependent sRNAs.

Search for poly(A) polymerase targets in E. coli reveals its implication in surveillance of Glu tRNA processing and degradation of stable RNAs.

Polyadenylation is a universal post-transcriptional modification involved in degradation and quality control of bacterial RNAs. In Escherichia coli, it is admitted that any accessible RNA 3' end can be tagged by a poly(A) tail for decay. However, we do not have yet an overall view of the population of polyadenylated molecules. The sampling of polyadenylated RNAs presented here demonstrates that rRNA fragments and tRNA precursors originating from the internal spacer regions of the rrn operons, in particular, rrnB are abundant poly(A) polymerase targets. Focused analysis showed that Glu tRNA precursors originating from the rrnB and rrnG transcripts exhibit long 3' trailers that are primarily removed by PNPase and to a lesser extent by RNase II and poly(A) polymerase. Moreover, 3' trimming by exoribonucleases precedes 5' end maturation by RNase P. Interestingly, characterization of RNA fragments that accumulate in a PNPase deficient strain showed that Glu tRNA precursors still harbouring the 5' leader can be degraded by a 3' to 5' quality control pathway involving poly(A) polymerase. This demonstrates that the surveillance of tRNA maturation described for a defective tRNA also applies to a wild-type tRNA.

Is the secondary putative RNA-RNA interaction site relevant to GcvB mediated regulation of oppA mRNA in Escherichia coli?

GcvB is a non-coding RNA that regulates oppA mRNA in different bacterial species by binding a GcvB GU-rich region named R1 to oppA mRNA. A secondary putative interaction site (PS1) was identified in this study that is able to form a second nearly perfect 10 base-pair duplex between these two RNAs in Escherichia coli. In this work, we have studied whether the formation of a second interaction site could help stabilize the previously reported GcvB/oppA complex. Several mutations and the full deletion of PS1 were engineered. None of these modifications affected the ability of GcvB to control OppA expression. Therefore the second, putative, interaction site appears to be unnecessary for the regulatory function of GcvB with regard to its oppA target mRNA.

Hfq affects mRNA levels independently of degradation.

BACKGROUND: The bacterial Lsm protein, Hfq, is an RNA chaperone involved in many reactions related to RNA metabolism, such as replication and stability, control of small RNA activity and polyadenylation. Despite this wide spectrum of known functions, the global role of Hfq is almost certainly undervalued; its capacity to bind DNA and to interact with many other proteins are only now beginning to be taken into account. RESULTS: The role of Hfq in the maturation and degradation of the rpsO mRNA of E. coli was investigated in vivo. The data revealed a decrease in rpsO mRNA abundance concomitant to an increase in its stability when Hfq is absent. This indicates that the change in mRNA levels in hfq mutants does not result from its modification of RNA stability. Moreover, a series of independent experiments have revealed that the decrease in mRNA level is not a consequence of a reduction of translation efficiency and that Hfq is not directly implicated in translational control of rpsO expression. Reduced steady-state mRNA levels in the absence of Hfq were also shown for rpsT, rpsB and rpsB-tsf, but not for lpp, pnp or tRNA transcripts. The abundance of chimeric transcripts rpsO-lacZ and rpsB-lacZ, whose expression was driven by rpsO and rpsB promoters, respectively, was also lower in the hfq null-mutants, while the beta-galactosidase yield remained about the same as in the parent wild-type strain. CONCLUSIONS: The data obtained suggest that alteration of rpsO, rpsT and rpsB-tsf transcript levels observed under conditions of Hfq deficiency is not caused by the post-transcriptional events, such as mRNA destabilization or changes in translation control, and may rather result from changes in transcriptional activity. So far, how Hfq affects transcription remains unclear. We propose that one of the likely mechanisms of Hfq-mediated modulation of transcription might operate early in the elongation step, when interaction of Hfq with a nascent transcript would help to overcome transcription pauses and to prevent preliminary transcript release.

Auto-assembly of E. coli DsrA small noncoding RNA: Molecular characteristics and functional consequences.

RNA molecules are important factors involved in different cellular processes and have a multitude of roles in the cell. These roles include serving as a temporary copy of genes used for protein synthesis or functions in translational machinery. Interestingly, RNA is so far the only biological molecule that serves both as a catalyst (like proteins) and as information storage (like DNA). However, in contrast to proteins well known to be able to self-associate in order to maintain the architecture of the cell, such RNA polymers are not prevalent in cells and are usually not favored by the flexibility of this molecule. In this work, we present evidence that such a polymer of a natural RNA, the DsrA RNA, exists in the bacterial cell. DsrA is a small noncoding RNA (87 nucleotides) of Escherichia coli that acts by base-pairing to mRNA in order to control the translation and the turnover of some mRNA, including rpoS mRNA, which encodes the sigma(s) RNA polymerase subunit involved in bacterial stress response. A putative model is proposed for the structure of this RNA polymer. Although the function of this polymerization is not known completely, we propose that the formation of such a structure could be involved in the regulation of DsrA ncRNA concentration in vivo or in a quality control mechanism used by the cell to eliminate misfolded RNAs.

Poly(A)-assisted RNA decay and modulators of RNA stability.

In Escherichia coli, RNA degradation is orchestrated by the degradosome with the assistance of complementary pathways and regulatory cofactors described in this chapter. They control the stability of each transcript and regulate the expression of many genes involved in environmental adaptation. The poly(A)-dependent degradation machinery has diverse functions such as the degradation of decay intermediates generated by endoribonucleases, the control of the stability of regulatory non coding RNAs (ncRNAs) and the quality control of stable RNA. The metabolism of poly(A) and mechanism of poly(A)-assisted degradation are beginning to be understood. Regulatory factors, exemplified by RraA and RraB, control the decay rates of subsets of transcripts by binding to RNase E, in contrast to regulatory ncRNAs which, assisted by Hfq, target RNase E to specific transcripts. Destabilization is often consecutive to the translational inactivation of mRNA. However, there are examples where RNA degradation is the primary regulatory step.

The poly(A)-dependent degradation pathway of rpsO mRNA is primarily mediated by RNase R.

Polyadenylation is an important factor controlling RNA degradation and RNA quality control mechanisms. In this report we demonstrate for the first time that RNase R has in vivo affinity for polyadenylated RNA and can be a key enzyme involved in poly(A) metabolism. RNase II and PNPase, two major RNA exonucleases present in Escherichia coli, could not account for all the poly(A)-dependent degradation of the rpsO mRNA. RNase II can remove the poly(A) tails but fails to degrade the mRNA as it cannot overcome the RNA termination hairpin, while PNPase plays only a modest role in this degradation. We now demonstrate that in the absence of RNase E, RNase R is the relevant factor in the poly(A)-dependent degradation of the rpsO mRNA. Moreover, we have found that the RNase R inactivation counteracts the extended degradation of this transcript observed in RNase II-deficient cells. Elongated rpsO transcripts harboring increasing poly(A) tails are specifically recognized by RNase R and strongly accumulate in the absence of this exonuclease. The 3' oligo(A) extension may stimulate the binding of RNase R, allowing the complete degradation of the mRNA, as RNase R is not susceptible to RNA secondary structures. Moreover, this regulation is shown to occur despite the presence of PNPase. Similar results were observed with the rpsT mRNA. This report shows that polyadenylation favors in vivo the RNase R-mediated pathways of RNA degradation.

The role of RNA chaperone Hfq in poly(A) metabolism methods to determine positions, abundance, and lengths of short oligo(A) tails.

Polyadenylation is a posttranscriptional modification of RNA occurring in prokaryotes, eukaryotes, and organelles. Long poly(A) tails help export eukaryotic mRNAs and promote mRNA stability and translation, whereas the short bacterial tails facilitate RNA decay. The scarcity of polyadenylated RNAs is one of the obstacles for investigators studying bacterial polyadenylation. The two methods described in this chapter were developed to determine how the poly(A) binding protein Hfq affects the polyadenylation of bacterial RNAs. The first is a 3'-RACE protocol specific to oligoadenylated RNA. This method was designed to rapidly collect a large amount of poly(A) containing 3'-terminal sequences to perform statistical analysis. The second method is an RNA sizing protocol to analyze the polyadenylation status of primary transcripts that were not efficiently detected by 3'-RACE. The latter procedure is based on Northern blot analysis of 3'-RNA fragments generated by RNase H. In the presence of a gene-specific methylated chimeric RNA-DNA oligonucleotide, the enzyme is directed to a unique cleavage site. The 3'-RNA fragments, differing by just one nucleotide at their 3'-ends, are then separated in polyacrylamide gels.

Polyadenylation of a functional mRNA controls gene expression in Escherichia coli.

Although usually implicated in the stabilization of mRNAs in eukaryotes, polyadenylation was initially shown to destabilize RNA in bacteria. All the data are consistent with polyadenylation being part of a quality control process targeting folded RNA fragments and non-functional RNA molecules to degradation. We report here an example in Escherichia coli, where polyadenylation directly controls the level of expression of a gene by modulating the stability of a functional transcript. Inactivation of poly(A)polymerase I causes overexpression of glucosamine-6-phosphate synthase (GlmS) and both the accumulation and stabilization of the glmS transcript. Moreover, we show that the glmS mRNA results from the processing of the glmU-glmS cotranscript by RNase E. Interestingly, the glmU-glmS cotranscript and the mRNA fragment encoding GlmU only slightly accumulated in the absence of poly(A)polymerase, suggesting that the endonucleolytically generated glmS mRNA harbouring a 5' monophosphate and a 3' stable hairpin is highly susceptible to poly(A)-dependent degradation.

Spectroscopic observation of RNA chaperone activities of Hfq in post-transcriptional regulation by a small non-coding RNA.

Hfq protein is vital for the function of many non-coding small (s)RNAs in bacteria but the mechanism by which Hfq facilitates the function of sRNA is still debated. We developed a fluorescence resonance energy transfer assay to probe how Hfq modulates the interaction between a sRNA, DsrA, and its regulatory target mRNA, rpoS. The relevant RNA fragments were labelled so that changes in intra- and intermolecular RNA structures can be monitored in real time. Our data show that Hfq promotes the strand exchange reaction in which the internal structure of rpoS is replaced by pairing with DsrA such that the Shine-Dalgarno sequence of the mRNA becomes exposed. Hfq appears to carry out strand exchange by inducing rapid association of DsrA and a premelted rpoS and by aiding in the slow disruption of the rpoS secondary structure. Unexpectedly, Hfq also disrupts a preformed complex between rpoS and DsrA. While it may not be a frequent event in vivo, this melting activity may have implications in the reversal of sRNA-based regulation. Overall, our data suggests that Hfq not only promotes strand exchange by binding rapidly to both DsrA and rpoS but also possesses RNA chaperoning properties that facilitates dynamic RNA-RNA interactions.

Hfq variant with altered RNA binding functions.

The interaction between Hfq and RNA is central to multiple regulatory processes. Using site-directed mutagenesis, we have found a missense mutation in Hfq (V43R) which strongly affects2 the RNA binding capacity of the Hfq protein and its ability to stimulate poly(A) tail elongation by poly(A)-polymerase in vitro. In vivo, overexpression of this Hfq variant fails to stimulate rpoS-lacZ expression and does not restore a normal growth rate in hfq null mutant. Cells in which the wild-type gene has been replaced by the hfqV43R allele exhibit a phenotype intermediate between those of the wild-type and of the hfq minus or null strains. This missense mutation derepresses Hfq synthesis. However, not all Hfq functions are affected by this mutation. For example, HfqV43R represses OppA synthesis as strongly as the wild-type protein. The dominant negative effect of the V43R mutation over the wild-type allele suggests that hexamers containing variant and genuine subunits are presumably not functional. Finally, molecular dynamics studies indicate that the V43R substitution mainly changes the position of the K56 and Y55 side chains involved in the Hfq-RNA interaction but has probably no effect on the folding and the oligomerization of the protein.

Three-dimensional structures of fibrillar Sm proteins: Hfq and other Sm-like proteins.

Hfq is a nucleic acid-binding protein that functions as a global regulator of gene expression by virtue of its interactions with several small, non-coding RNA species. Originally identified as an Escherichia coli host factor required for RNA phage Qbeta replication, Hfq is now known to post-transcriptionally regulate bacterial gene expression by modulating both mRNA stability and translational activity. Recently shown to be a member of the diverse Sm protein family, Hfq adopts the OB-like fold typical of other Sm and Sm-like (Lsm) proteins, and also assembles into toroidal homo-oligomers that bind single-stranded RNA. Similarities between the structures, functions, and evolution of Sm/Lsm proteins and Hfq are continually being discovered, and we now report an additional, unexpected biophysical property that is shared by Hfq and other Sm proteins: E.coli Hfq polymerizes into well-ordered fibres whose morphologies closely resemble those found for Sm-like archaeal proteins (SmAPs). However, the hierarchical assembly of these fibres is dissimilar: whereas SmAPs polymerize into polar tubes (and striated bundles of such tubes) by head-to-tail stacking of individual homo-heptamers, helical Hfq fibres are formed by cylindrical slab-like layers that consist of 36 subunits arranged as a hexamer of Hfq homo-hexamers (i.e. protofilaments in a 6 x 6 arrangement). The different fibrillar ultrastructures formed by Hfq and SmAP are presented and examined herein, with the overall goal of elucidating another similarity amongst the diverse members of the Sm protein family.

Characterization of the molecular mechanisms involved in the differential production of erythrose-4-phosphate dehydrogenase, 3-phosphoglycerate kinase and class II fructose-1,6-bisphosphate aldolase in Escherichia coli.

A gapA-pgk gene tandem coding the glyceraldehyde 3-phosphate dehydrogenase and 3-phosphoglycerate kinase, is most frequently found in bacteria. However, in Enterobacteriaceae, gapA is replaced by an epd open reading frame (ORF) coding an erythrose-4-phosphate dehydrogenase and an fbaA ORF coding the class II fructose-1,6-bisphosphate aldolase follows pgk. Although epd expression is very low in Escherichia coli, we show that, in the presence of glucose, the 3 epd, pgk and fbaA ORFs are efficiently cotranscribed from promoter epd P0. Conservation of promoter epd P0 is likely due to its important role in modulation of the metabolic flux during glycolysis and gluconeogenesis. As a consequence, we found that the epd translation initiation region and ORF have been adapted in order to limit epd translation and to create an efficient RNase E entry site. We also show that fbaA is cotranscribed with pgk, from promoter epd P0 or an internal pgk P1 promoter of the extended -10 class. The differential expression of pgk and fbaA also depends upon an RNase E segmentation process, leading to individual mRNAs with different stabilities. The secondary structures of the RNA regions containing the RNase E sites were experimentally determined which brings important information on the structural features of RNase E ectopic sites.

Fate of mRNA extremities generated by intrinsic termination: detailed analysis of reactions catalyzed by ribonuclease II and poly(A) polymerase.

In all living cells 3' ends of RNA are posttranscriptionally elongated or shortened by nucleotidyl transferases and ribonucleases. The detailed analysis of the rpsO mRNA of Escherichia coli presented here demonstrates that transcription terminates in vivo at two sites located seven and eight nucleotides downstream from the GC-rich hairpin of the intrinsic terminator and that primary transcripts can be shortened by RNase II. The shortest RNA identified in the cell result from nibbling of primary transcripts. Primary transcripts and nibbled molecules can also be adenylated by poly(A) polymerase I (PAP I). In addition, kinetics of decay performed in vitro demonstrate that RNase II rapidly degrades poly(A) tails longer than 7-8 As processively while it slowly nibbles shorter tails and non adenylated RNAs distributively. Comparison of the kinetics of nibbling of oligoadenylated rpsO mRNA in vivo and in vitro lead us to conclude that the rates of shortening and elongation of the oligo(A) tails detected in vivo are very slow: about 0.5-7 nucleotides per min. We finally speculate that the slowness of oligo(A) synthesis may explain why polyadenylation does not affect the stability of mRNAs whose degradation is controlled by RNase E.

Stimulation of poly(A) synthesis by Escherichia coli poly(A)polymerase I is correlated with Hfq binding to poly(A) tails.

The bacterial Lsm protein, host factor I (Hfq), is an RNA chaperone involved in many types of RNA transactions such as replication and stability, control of small RNA activity and polyadenylation. In this latter case, Hfq stimulates poly(A) synthesis and binds poly(A) tails that it protects from exonucleolytic degradation. We show here, that there is a correlation between Hfq binding to the 3' end of an RNA molecule and its ability to stimulate RNA elongation catalyzed by poly(A)polymerase I. In contrast, formation of the Hfq-RNA complex inhibits elongation of the RNA by polynucleotide phosphorylase. We demonstrate also that Hfq binding is not affected by the phosphorylation status of the RNA molecule and occurs equally well at terminal or internal stretches of poly(A).