skip to main content

Attention:

The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 5:00 PM ET until 11:00 PM ET on Friday, June 21 due to maintenance. We apologize for the inconvenience.


Title: Integrated omics endotyping of infants with respiratory syncytial virus bronchiolitis and risk of childhood asthma
Abstract Respiratory syncytial virus (RSV) bronchiolitis is not only the leading cause of hospitalization in U.S. infants, but also a major risk factor for asthma development. While emerging evidence suggests clinical heterogeneity within RSV bronchiolitis, little is known about its biologically-distinct endotypes. Here, we integrated clinical, virus, airway microbiome (species-level), transcriptome, and metabolome data of 221 infants hospitalized with RSV bronchiolitis in a multicentre prospective cohort study. We identified four biologically- and clinically-meaningful endotypes: A) clinical classic microbiome M. nonliquefaciens inflammation IFN-intermediate , B) clinical atopic microbiome S. pneumoniae / M. catarrhalis inflammation IFN-high , C) clinical severe microbiome mixed inflammation IFN-low , and D) clinical non-atopic microbiome M.catarrhalis inflammation IL-6 . Particularly, compared with endotype A infants, endotype B infants—who are characterized by a high proportion of IgE sensitization and rhinovirus coinfection, S. pneumoniae/M. catarrhalis codominance, and high IFN-α and -γ response—had a significantly higher risk for developing asthma (9% vs. 38%; OR, 6.00: 95%CI, 2.08–21.9; P = 0.002). Our findings provide an evidence base for the early identification of high-risk children during a critical period of airway development.  more » « less
Award ID(s):
2028280
NSF-PAR ID:
10332930
Author(s) / Creator(s):
; ; ; ; ; ; ; ;
Date Published:
Journal Name:
Nature Communications
Volume:
12
Issue:
1
ISSN:
2041-1723
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Respiratory syncytial virus (RSV) causes substantial morbidity and mortality in infants, the immunocompromised, and the elderly. RSV infects the airway epithelium via the apical membrane and almost exclusively sheds progeny virions back into the airway mucus (AM), making RSV difficult to target by systemically administered therapies. An inhalable “muco‐trapping” variant of motavizumab (Mota‐MT), a potent neutralizing mAb against RSV F is engineered. Mota‐MT traps RSV in AM via polyvalent Fc‐mucin bonds, reducing the fraction of fast‐moving RSV particles in both fresh pediatric and adult AM by ≈20–30‐fold in a Fc‐glycan dependent manner, and facilitates clearance from the airways of mice within minutes. Intranasal dosing of Mota‐MT eliminated viral load in cotton rats within 2 days. Daily nebulized delivery of Mota‐MT to RSV‐infected neonatal lambs, beginning 3 days after infection when viral load is at its maximum, led to a 10 000‐fold and 100 000‐fold reduction in viral load in bronchoalveolar lavage and lung tissues relative to placebo control, respectively. Mota‐MT‐treated lambs exhibited reduced bronchiolitis, neutrophil infiltration, and airway remodeling than lambs receiving placebo or intramuscular palivizumab. The findings underscore inhaled delivery of muco‐trapping mAbs as a promising strategy for the treatment of RSV and other acute respiratory infections.

     
    more » « less
  2. Objectives/Hypothesis

    Subglottic stenosis (SGS) results from dysregulated extracellular matrix deposition by laryngotracheal fibroblasts causing scar tissue formation following intubation. Recent work has highlighted a relationship between this inflammatory state and imbalances in the upper airway microbiome. Herein, we engineer novel drug‐eluting endotracheal (ET) tubes to deliver a model antimicrobial peptide Lasioglossin‐III (Lasio) for the local modulation of the microbiome during intubation.

    Study Design

    Controlledin vitrostudy.

    Methods

    ET tubes were coated with a water‐in‐oil (w/o) emulsion of Lasio in poly(d,l‐lactide‐co‐glycolide) (PLGA) by dipping thrice. Peptide release was quantified over 2 weeks via fluorometric peptide assays. The antibacterial activity was tested against airway microbes (Staphylococcus epidermidis,Streptococcus pneumoniae, and pooled human microbiome samples) by placing Lasio/PLGA‐coated tubes and appropriate controls in 48 well plates with diluted bacteria. Bacterial inhibition and tube adhesion were tested by measuring optical density and colony formation after tube culture, respectively. Biocompatibility was tested against laryngotracheal fibroblasts and lung epithelial cells.

    Results

    We achieved a homogeneous coating of ET tubes with Lasio in a PLGA matrix that yields a prolonged, linear release over 1 week (typical timeframe before the ET tube is changed). We observed significant antibacterial activity againstS. epidermidis,S. pneumoniae, and human microbiome samples, and prevention of bacterial adherence to the tube. Additionally, the released Lasio did not cause any cytotoxicity toward laryngotracheal fibroblasts or lung epithelial cellsin vitro.

    Conclusion

    Overall, we demonstrate the design of an effective‐eluting ET tube to modulate upper‐airway bacterial infections during intubation which could be deployed to help prevent SGS.

    Level of Evidence

    NALaryngoscope, 132:1356–1363, 2022

     
    more » « less
  3. null (Ed.)
    Human respiratory syncytial virus (RSV) is the most common cause of viral bronchiolitis and pneumonia in infants and children worldwide. Inflammation induced by RSV infection is responsible for its hallmark manifestation of bronchiolitis and pneumonia. The cellular debris created through lytic cell death of infected cells is a potent initiator of this inflammation. Macrophages are known to play a pivotal role in the early innate immune and inflammatory response to viral pathogens. However, the lytic cell death mechanisms associated with RSV infection in macrophages remains unknown. Two distinct mechanisms involved in lytic cell death are pyroptosis and necroptosis. Our studies revealed that RSV induces lytic cell death in macrophages via both of these mechanisms, specifically through the ASC (Apoptosis-associated speck like protein containing a caspase recruitment domain)-NLRP3 (nucleotide-binding domain, leucine-rich-containing family, pyrin domain-containing-3) inflammasome activation of both caspase-1 dependent pyroptosis and receptor-interacting serine/threonine-protein kinase 3 (RIPK3), as well as a mixed lineage kinase domain like pseudokinase (MLKL)-dependent necroptosis. In addition, we demonstrated an important role of reactive oxygen species (ROS) during lytic cell death of RSV-infected macrophages. 
    more » « less
  4. 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
  5. Obesity is now a prevalent disease worldwide and has a multi-factorial etiology. Several viruses or virus-like agents including members of adenoviridae, herpesviridae, slow virus (prion), and hepatitides, have been associated with obesity; meanwhile obese patients are shown to be more susceptible to viral infections such as during influenza and dengue epidemics. We examined the co-factorial role of viral infections, particularly of the persistent cases, in synergy with high-fat diet in induction of obesity. Antiviral interferons (IFNs), as key immune regulators against viral infections and in autoimmunity, emerge to be a pivotal player in the regulation of adipogenesis. In this review, we examine the recent evidence indicating that gut microbiota uphold intrinsic IFN signaling, which is extensively involved in the regulation of lipid metabolism. However, the prolonged IFN responses during persistent viral infections and obesogenesis comprise reciprocal causality between virus susceptibility and obesity. Furthermore, some IFN subtypes have shown therapeutic potency in their anti-inflammation and anti-obesity activity. 
    more » « less