Methanobactins (MBs) are ribosomally produced and post-translationally modified peptides (RiPPs) that are used by methanotrophs for copper acquisition. The signature post-translational modification of MBs is the formation of two heterocyclic groups, either an oxazolone, pyrazinedione or imidazolone group, with an associated thioamide from an X -Cys dipeptide. The precursor peptide (MbnA) for MB formation is found in a gene cluster of MB-associated genes. The exact biosynthetic pathway of MB formation is not yet fully understood, and there are still uncharacterized proteins in some MB gene clusters, particularly those that produce pyrazinedione or imidazolone rings. One such protein is MbnF, which is proposed to be a flavin monooxygenase (FMO) based on homology. To help to elucidate its possible function, MbnF from Methylocystis sp. strain SB2 was recombinantly produced in Escherichia coli and its X-ray crystal structure was resolved to 2.6 Å resolution. Based on its structural features, MbnF appears to be a type A FMO, most of which catalyze hydroxylation reactions. Preliminary functional characterization shows that MbnF preferentially oxidizes NADPH over NADH, supporting NAD(P)H-mediated flavin reduction, which is the initial step in the reaction cycle of several type A FMO enzymes. It is also shown that MbnF binds the precursor peptide for MB, with subsequent loss of the leader peptide sequence as well as the last three C-terminal amino acids, suggesting that MbnF might be needed for this process to occur. Finally, molecular-dynamics simulations revealed a channel in MbnF that is capable of accommodating the core MbnA fragment minus the three C-terminal amino acids.
more »
« less
MbnC Is Not Required for the Formation of the N-Terminal Oxazolone in the Methanobactin from Methylosinus trichosporium OB3b
ABSTRACT Methanobactins (MBs) are ribosomally synthesized and posttranslationally modified peptides (RiPPs) produced by methanotrophs for copper uptake. The posttranslational modification that defines MBs is the formation of two heterocyclic groups with associated thioamines from X-Cys dipeptide sequences. Both heterocyclic groups in the MB from Methylosinus trichosporium OB3b (MB-OB3b) are oxazolone groups. The precursor gene for MB-OB3b is mbnA , which is part of a gene cluster that contains both annotated and unannotated genes. One of those unannotated genes, mbnC , is found in all MB operons and, in conjunction with mbnB , is reported to be involved in the formation of both heterocyclic groups in all MBs. To determine the function of mbnC , a deletion mutation was constructed in M. trichosporium OB3b, and the MB produced from the Δ mbnC mutant was purified and structurally characterized by UV-visible absorption spectroscopy, mass spectrometry, and solution nuclear magnetic resonance (NMR) spectroscopy. MB-OB3b from the Δ mbnC mutant was missing the C-terminal Met and was also found to contain a Pro and a Cys in place of the pyrrolidinyl-oxazolone-thioamide group. These results demonstrate MbnC is required for the formation of the C-terminal pyrrolidinyl-oxazolone-thioamide group from the Pro-Cys dipeptide, but not for the formation of the N-terminal 3-methylbutanol-oxazolone-thioamide group from the N-terminal dipeptide Leu-Cys. IMPORTANCE A number of environmental and medical applications have been proposed for MBs, including bioremediation of toxic metals and nanoparticle formation, as well as the treatment of copper- and iron-related diseases. However, before MBs can be modified and optimized for any specific application, the biosynthetic pathway for MB production must be defined. The discovery that mbnC is involved in the formation of the C-terminal oxazolone group with associated thioamide but not for the formation of the N-terminal oxazolone group with associated thioamide in M. trichosporium OB3b suggests the enzymes responsible for posttranslational modification(s) of the two oxazolone groups are not identical.
more »
« less
- Award ID(s):
- 1912482
- NSF-PAR ID:
- 10339879
- Editor(s):
- Kelly, Robert M.
- Date Published:
- Journal Name:
- Applied and Environmental Microbiology
- Volume:
- 88
- Issue:
- 2
- ISSN:
- 0099-2240
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
more » « less
Summary Proper protein anchoring is key to the biogenesis of prokaryotic cell surfaces, dynamic, resilient structures that play crucial roles in various cell processes. A novel surface protein anchoring mechanism in
Haloferax volcanii depends upon the peptidase archaeosortase A (ArtA) processing C‐termini of substrates containing C‐terminal tripartite structures and anchoring mature substrates to the cell membrane via intercalation of lipid‐modified C‐terminal amino acid residues. While this membrane protein lacks clear homology to soluble sortase transpeptidases of Gram‐positive bacteria, which also process C‐termini of substrates whose C‐terminal tripartite structures resemble those of ArtA substrates, archaeosortases do contain conserved cysteine, arginine and arginine/histidine/asparagine residues, reminiscent of His‐Cys‐Arg residues of sortase catalytic sites. The study presented here shows that ArtAWT‐GFP expressedin trans complements ΔartA growth and motility phenotypes, while alanine substitution mutants, Cys173(C173A), Arg214(R214A) or Arg253(R253A), and the serine substitution mutant for Cys173(C173S), fail to complement these phenotypes. Consistent with sortase active site replacement mutants, ArtAC173A‐GFP, ArtAC173S‐GFP and ArtAR214A‐GFP cannot process substrates, while replacement of the third residue, ArtAR253A‐GFP retains some processing activity. These findings support the view that similarities between certain aspects of the structures and functions of the sortases and archaeosortases are the result of convergent evolution. -
more » « less
ABSTRACT EseN is an
Edwardsiella ictaluri type III secretion system effector with phosphothreonine lyase activity. In this work, we demonstrate that EseN inactivates p38 and c-Jun-N-terminal kinase (JNK) in infected head-kidney-derived macrophages (HKDMs). We have previously reported inactivation of extracellular-regulated kinase 1/2 (ERK1/2). Also, for the first time, we demonstrated that EseN is involved in the inactivation of 3-phosphoinositide-dependent kinase 1 (PDK1), which has not been previously demonstrated for any of the EseN homologs in other species. We also found that EseN significantly affected mRNA expression ofIL-10 , pro-apoptoticbaxa , andp53 , but had no significant effect on anti-apoptoticbcl2 or pro-apoptotic apoptotic peptidase activating factor 1. EseN is also involved in the inhibition of caspase-8 and caspase-3/7 but does not affect caspase-9 activity. Repression of apoptosis was further confirmed with flow cytometry using Alexa Fluor 647-labeled annexin V and propidium iodide. In addition, we found that theE. ictaluri T3SS is essential for the inhibition of IL-1β maturation, but EseN is not involved in this process. EseN did not affect cell pyroptosis, as indicated by the lack of EseN impact on the release of lactate dehydrogenase from infected HKDM. The transmission electron microscopy data also indicate that HKDM infected with WT or aneseN mutant died by apoptosis, while HKDM infected with the T3SS mutant more likely died by pyroptosis. Collectively, our results indicate thatE. ictaluri EseN is involved in inactivation of ERK1/2, p38, JNK, and PDK1 signaling pathways that lead to modulation of cell death among infected HKDMs.IMPORTANCE This work has global significance in the catfish industry, which provides food for increasing global populations.
E. ictaluri is a leading cause of disease loss, and EseN is an important player inE. ictaluri virulence. TheE. ictaluri T3SS effector EseN plays an essential role in establishing infection, but the specific role EseN plays is not well characterized. EseN belongs to a family of phosphothreonine lyase effectors that specifically target host mitogen activated protein kinase (MAPK) pathways important in regulating host responses to infection. No phosphothreonine lyase equivalents are known in eukaryotes, making this family of effectors an attractive target for indirect narrow-spectrum antibiotics. Targeting of major vault protein and PDK1 kinase by EseN has not been reported in EseN homologs in other pathogens and may indicate unique functions ofE. ictaluri EseN. EseN targeting of PDK1 is particularly interesting in that it is linked to an extraordinarily diverse group of cellular functions. -
The iron-containing heterodimeric MbnBC enzyme complex plays a central role in the biosynthesis of methanobactins (Mbns), ribosomally synthesized, posttranslationally modified natural products that bind copper with high affinity. MbnBC catalyzes a four-electron oxidation of a cysteine residue in its precursor-peptide substrate, MbnA, to an oxazolone ring and an adjacent thioamide group. Initial studies of MbnBC indicated the presence of both diiron and triiron species, complicating identification of the catalytically active species. Here, we present evidence through activity assays combined with electron paramagnetic resonance (EPR) and Mössbauer spectroscopic analysis that the active species is a mixed-valent, antiferromagnetically coupled Fe(II)Fe(III) center. Consistent with this assignment, heterologous expression of the MbnBC complex in culture medium containing less iron yielded purified protein with less bound iron but greater activity in vitro. The maximally activated MbnBC prepared in this manner could modify both cysteine residues in MbnA, in contrast to prior findings that only the first cysteine could be processed. Site-directed mutagenesis and multiple crystal structures clearly identify the two essential Fe ions in the active cluster as well as the location of the previously detected third Fe site. Moreover, structural modeling indicates a role for MbnC in recognition of the MbnA leader peptide. These results add a biosynthetic oxidative rearrangement reaction to the repertoire of nonheme diiron enzymes and provide a foundation for elucidating the MbnBC mechanism.more » « less
-
Obeid, I. (Ed.)The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do not have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository (https://www.foxchase.org/research/facilities/genetic-research-facilities/biosample-repository -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York City, New York, USA: Springer, 2020, pp. 67 104. https://www.springer.com/gp/book/9783030368432. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. isip.piconepress.com/projects/nsf_dpath/. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. https://doi.org/10.21437/interspeech.2020-3015. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. https://ieeexplore.ieee.org/document/8675201. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: https://ai.googleblog.com/2020/06/recent-advances-in-google-translate.html. [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. piconepress.com/publications/conference_proceedings/2021/ieee_spmb/eeg_transfer_learning/. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/nsf/mri_dpath/. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. https://ieeexplore.ieee.org/document/9037859. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. https://doi.org/10.5858/arpa.2015-0238-OA.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1128/AEM.01841-21
