More Like this

1. Longitudinal dynamics and behavioural correlates of telomeres in male wire‐tailed manakins

   https://doi.org/10.1111/1365-2435.13715  

   Vernasco, Ben J. ; Dakin, Roslyn ; Majer, Ariana D. ; Haussmann, Mark F. ; Brandt Ryder, T. ; Moore, Ignacio T. ; Angelier, ed., Frédéric ( December 2020 , Functional Ecology)

   Building on the predictions of state‐dependent life‐history theory, telomeres are hypothesized to either correlate with or function as an adaptive, proximate mediator of an individual's behaviour and life‐history strategy. To further understand the relationship between telomeres, behaviour and life‐history strategies, we measured male behaviour, telomere lengths and telomere dynamics in a free‐living population of known‐age, male wire‐tailed manakinsPipra filicauda.

   Male wire‐tailed manakins perform coordinated displays with other males at leks and these displays form the basis of long‐term coalition partnerships. Males exhibit consistent individual differences in the number of social partners within their social network and the frequency of social interactions. Male sociality is also positively correlated with both social rise and reproductive success.

   We measured male behaviour using a telemetry‐based, proximity datalogging system and blood telomere lengths were quantified using qPCR. We examined the relationships between telomere length, telomere dynamics, social status, and male behaviour. We also quantified the repeatability of telomere lengths, examined age‐related changes in telomere length, and tested for instances of telomere elongation that exceed residual error in telomere length.

   Telomere length was found to be highly repeatable. More social males exhibited shorter telomeres and higher rates of telomere attrition. Telomeres did not significantly vary with age within or between individuals in either of the male social classes. Two out of 25 individuals exhibited patterns telomere elongation that exceeded residual error in telomere measurements.

   Here we show that telomeres consistently vary between male wire‐tailed manakins and these differences are related to variation in male social behaviour. In this relatively long‐lived species, telomeres appear to be flexible traits that can increase or decrease in length. Overall, this study provides observational support for the hypothesis that telomeres act as a molecular marker that relates to behaviour in a state‐dependent manner. We also provide insight into the molecular consequences of individual variation in male social behaviour.

   A freePlain Language Summarycan be found within the Supporting Information of this article.

    

   more » « less
2. Urinary neopterin reflects immunological variation associated with age, helminth parasitism, and the microbiome in a wild primate

   https://doi.org/10.1038/s41598-022-25298-9  

   Schneider-Crease, India A. ; Feder, Jacob A. ; Baniel, Alice ; McCann, Colleen ; Haile, Abebaw Azanaw ; Abebe, Belayneh ; Fitzgerald, Lauren ; Gomery, Megan A. ; Simberloff, Ruth A. ; Petrie, Zack L. ; et al ( December 2022 , Scientific Reports)

   Neopterin, a product of activated white blood cells, is a marker of nonspecific inflammation that can capture variation in immune investment or disease-related immune activity and can be collected noninvasively in urine. Mounting studies in wildlife point to lifetime patterns in neopterin related to immune development, aging, and certain diseases, but rarely are studies able to assess whether neopterin can capture multiple concurrent dimensions of health and disease in a single system. We assessed the relationship between urinary neopterin stored on filter paper and multiple metrics of health and disease in wild geladas (Theropithecus gelada), primates endemic to the Ethiopian highlands. We tested whether neopterin captures age-related variation in inflammation arising from developing immunity in infancy and chronic inflammation in old age, inflammation related to intramuscular tapeworm infection, helminth-induced anti-inflammatory immunomodulation, and perturbations in the gastrointestinal microbiome. We found that neopterin had a U-shaped relationship with age, no association with larval tapeworm infection, a negative relationship with metrics related to gastrointestinal helminth infection, and a negative relationship with microbial diversity. Together with growing research on neopterin and specific diseases, our results demonstrate that urinary neopterin can be a powerful tool for assessing multiple dimensions of health and disease in wildlife.

    

   more » « less
3. The Temple University Digital Pathology Corpus: The Breast Tissue Subset

   https://doi.org/10.1109/SPMB52430.2021.9672275  

   Wevodau, Z. ; Doshna, B. ; Jhala, N. ; Akhtar, I. ; Obeid, I. ; Picone, J. ( December 2021 , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   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
4. Imaging Mass Cytometry Reveals Predominant Innate Immune Signature and Endothelial–Immune Cell Interaction in Juvenile Myositis Compared to Lupus Skin

   https://doi.org/10.1002/art.42283  

   Turnier, Jessica L. ; Yee, Christine M. ; Madison, Jacqueline A. ; Rizvi, Syed M. ; Berthier, Celine C. ; Wen, Fei ; Kahlenberg, J. Michelle ( October 2022 , Arthritis & Rheumatology)

   Cutaneous inflammation can signal disease in juvenile dermatomyositis (DM) and childhood‐onset systemic lupus erythematosus (cSLE), but we do not fully understand cellular mechanisms of cutaneous inflammation. In this study, we used imaging mass cytometry to characterize cutaneous inflammatory cell populations and cell–cell interactions in juvenile DM as compared to cSLE.

   We performed imaging mass cytometry analysis on skin biopsy samples from juvenile DM patients (n = 6) and cSLE patients (n = 4). Tissue slides were processed and incubated with metal‐tagged antibodies for CD14, CD15, CD16, CD56, CD68, CD11c, HLA–DR, blood dendritic cell antigen 2, CD20, CD27, CD138, CD4, CD8, E‐cadherin, CD31, pan‐keratin, and type I collagen. Stained tissue was ablated, and raw data were acquired using the Hyperion imaging system. We utilized the Phenograph unsupervised clustering algorithm to determine cell marker expression and permutation test by histoCAT to perform neighborhood analysis.

   We identified 14 cell populations in juvenile DM and cSLE skin, including CD14+ and CD68+ macrophages, myeloid and plasmacytoid dendritic cells (pDCs), CD4+ and CD8+ T cells, and B cells. Overall, cSLE skin had a higher inflammatory cell infiltrate, with increased CD14+ macrophages, pDCs, and CD8+ T cells and immune cell–immune cell interactions. Juvenile DM skin displayed a stronger innate immune signature, with a higher overall percentage of CD14+ macrophages and prominent endothelial cell–immune cell interaction.

   Our findings identify immune cell population differences, including CD14+ macrophages, pDCs, and CD8+ T cells, in juvenile DM skin compared to cSLE skin, and highlight a predominant innate immune signature and endothelial cell–immune cell interaction in juvenile DM, providing insight into candidate cell populations and interactions to better understand disease‐specific pathophysiology.

    

   more » « less
5. Mitochondrial DNA: a novel indicator of active inflammation in ANCA-associated vasculitides

   https://doi.org/10.1093/rheumatology/kead015  

   Giaglis, Stavros ; Daoudlarian, Douglas ; Thiel, Jens ; Rizzi, Marta ; Kyburz, Diego ; Venhoff, Nils ; Walker, Ulrich A. ( January 2023 , Rheumatology)

   ANCA-associated vasculitis (AAV) includes granulomatosis with polyangiitis (GPA) and microscopic polyangiitis (MPA). ANCA triggers neutrophil extracellular trap formation, which releases either mitochondrial (mt) DNA or nuclear DNA (n) DNA, contributing to inflammation. Our aim was to prospectively examine the extent and nature of circulating DNA in AAV and the clinical utility of DNA quantification.

   DNA was isolated from platelet-free plasma of consecutive GPA and MPA patients and healthy controls (HCs). mtDNA and nDNA copy numbers were quantified by PCR. Clinical data, including the BVAS, were collected.

   Ninety-two HCs (median age 51 years, 58.7% female) and 101 AAV patients (80 GPA, 21 MPA, median age 64 years, 50.5% female, BVAS range: 0–30) were included. Median mtDNA copies were 13-fold higher in patients with AAV than in HCs; nDNA concentrations did not differ. Patients with active AAV (BVAS > 0) had 4-fold higher median mtDNA copies than patients in remission (P = 0.03). mtDNA, unlike nDNA, correlated with BVAS (r = 0.30, P = 0.002) and was associated with AAV activity at multivariable analysis. Receiver operating characteristic curve analysis indicated that mtDNA quantification differentiates patients with active AAV (BVAS > 0) from HCs with 96.1% sensitivity and 98.9% specificity (area under the curve 0.99). In 27 AAV patients with follow-up, mtDNA changes but not CRP or ANCA-titres correlated with BVAS changes (r = 0.56, P = 0.002).

   mtDNA, unlike nDNA, is elevated in the plasma of AAV patients and may contribute to systemic inflammation. mtDNA could be superior to established biomarkers in the laboratory monitoring of AAV activity.

    

   more » « less