Histopathological assessment of transplant biopsies is currently the standard method to diagnose allograft rejection and can help guide patient management, but it is one of the most challenging areas of pathology, requiring considerable expertise, time, and effort. We aimed to analyse the utility of deep learning to preclassify histology of kidney allograft biopsies into three main broad categories (ie, normal, rejection, and other diseases) as a potential biopsy triage system focusing on transplant rejection. We performed a retrospective, multicentre, proof-of-concept study using 5844 digital whole slide images of kidney allograft biopsies from 1948 patients. Kidney allograft biopsy samples were identified by a database search in the Departments of Pathology of the Amsterdam UMC, Amsterdam, Netherlands (1130 patients) and the University Medical Center Utrecht, Utrecht, Netherlands (717 patients). 101 consecutive kidney transplant biopsies were identified in the archive of the Institute of Pathology, RWTH Aachen University Hospital, Aachen, Germany. Convolutional neural networks (CNNs) were trained to classify allograft biopsies as normal, rejection, or other diseases. Three times cross-validation (1847 patients) and deployment on an external real-world cohort (101 patients) were used for validation. Area under the receiver operating characteristic curve (AUROC) was used as the main performance metric (the primary endpoint to assess CNN performance). Serial CNNs, first classifying kidney allograft biopsies as normal (AUROC 0·87 [ten times bootstrapped CI 0·85-0·88]) and disease (0·87 [0·86-0·88]), followed by a second CNN classifying biopsies classified as disease into rejection (0·75 [0·73-0·76]) and other diseases (0·75 [0·72-0·77]), showed similar AUROC in cross-validation and deployment on independent real-world data (first CNN normal AUROC 0·83 [0·80-0·85], disease 0·83 [0·73-0·91]; second CNN rejection 0·61 [0·51-0·70], other diseases 0·61 [0·50-0·74]). A single CNN classifying biopsies as normal, rejection, or other diseases showed similar performance in cross-validation (normal AUROC 0·80 [0·73-0·84], rejection 0·76 [0·66-0·80], other diseases 0·50 [0·36-0·57]) and generalised well for normal and rejection classes in the real-world data. Visualisation techniques highlighted rejection-relevant areas of biopsies in the tubulointerstitium. This study showed that deep learning-based classification of transplant biopsies could support pathological diagnostics of kidney allograft rejection. European Research Council; German Research Foundation; German Federal Ministries of Education and Research, Health, and Economic Affairs and Energy; Dutch Kidney Foundation; Human(e) AI Research Priority Area of the University of Amsterdam; and Max-Eder Programme of German Cancer Aid.

Background: At border sites, and in internal organs, tissue resident memory T cells (T_RM) contribute to the immune barrier against pathogens like viruses, bacteria, fungi, and cancer. However, information on the presence and function of these cells in the human kidney is scant. In order to better understand the T cell-mediated immunological defense in this organ, we aimed to determine phenotypic and functional aspects of CD8 and CD4 T cells present in healthy and allograft kidney tissue. Methods: Using multichannel flow cytometry, we assessed the phenotype and function of T cells in healthy renal tissue samples (n = 5) and kidney allograft tissue (n = 7) and compared these aspects to T cells in peripheral blood from healthy controls (n = 13). Results: Kidney tissue samples contained substantial amounts of CD8 and CD4 T cells. In contrast to the circulating cells, kidney T cells frequently expressed CD69 and CD103, and were more often actively cycling. Furthermore, nearly all kidney T cells expressed CXCR3, and often expressed CXCR6 compared to T cells in the circulation. Markedly, kidney T cells produced greater quantities of IFNγ than circulating cells and were frequently polyfunctional. Conclusion: Functional T cells with the characteristic traits of T_RM reside in human kidney tissues. These cells are more often actively cycling and frequently express CXCR3 and CXCR6.

The application of proteomics to fresh frozen (FF) and formalin-fixed paraffin-embedded (FFPE) human tissues is an important development spurred on by requests from stakeholder groups in clinical fields. One objective is to complement current diagnostic methods with new specific molecular information. An important goal is to achieve adequate and consistent protein recovery across and within large-scale studies. Here, we describe development of several protocols incorporating mass spectrometry compatible detergents, including Rapigest, PPS, and ProteaseMax. Methods were applied on 4 and 15 μm thick FF tissues, and 4 μm thick FFPE tissues. We evaluated sensitivity and repeatability of the methods and found that the protocol containing Rapigest enabled detection of 630 proteins from FF tissue of 1 mm² and 15 μm thick, whereas 498 and 297 proteins were detected with the protocols containing ProteaseMax and PPS, respectively. Surprisingly, PPS-containing buffer showed good extraction of the proteins from 4 μm thick FFPE tissue with the average of 270 protein identifications (1 mm²), similar to the results on 4 μm thick FF. Moreover, we found that temperature increases during incubation with urea on 4 μm thick FF tissue revealed a decrease in the number of identified proteins and increase in the number of the carbamylated peptides.

The PRO-C6 assay, a reflection of collagen type VI synthesis, has been proposed as a non-invasive early biomarker of kidney fibrosis. We aimed to investigate cross-sectional and longitudinal associations between plasma and urine PRO-C6 and proven histological changes after kidney transplantation. The current study is a post-hoc analysis of 94 participants of the MECANO trial, a 24-month prospective, multicenter, open-label, randomized, controlled trial aimed at comparing everolimus-based vs. cyclosporine-based immunosuppression. PRO-C6 was measured in plasma and urine samples collected 6 and 24 months post-transplantation. Fibrosis was evaluated in biopsies collected at the same time points by Banff interstitial fibrosis/tubular atrophy (IF/TA) scoring and collagen staining (Picro Sirius Red; PSR); inflammation was evaluated by the tubulo-interstitial inflammation score (ti-score). Linear regression analyses were performed. Six-month plasma PRO-C6 was cross-sectionally associated with IF/TA score (Std. β = 0.34), and prospectively with 24-month IF/TA score and ti-score (Std. β = 0.24 and 0.23, respectively) (p < 0.05 for all). No significant associations were found between urine PRO-C6 and any of the biopsy findings. Fibrotic changes and urine PRO-C6 behaved differentially over time according to immunosuppressive therapy. These results are a first step towards non-invasive fibrosis detection after kidney transplantation by means of collagen VI synthesis measurement, and further research is required.

IMPORTANCE: Sepsis-related host-response anomalies contribute to acute kidney injury (AKI) duration. Data on the host-response specific to COVID-19-associated AKI (COVID-AKI) in critically ill patients is limited. OBJECTIVES: We postulated that persistent COVID-AKI (> 48 hr) differs in host response from transient (< 48 hr) or no COVID-AKI. DESIGN, SETTING, AND PARTICIPANTS: This prospective biomarker study observed patients with severe acute respiratory syndrome coronavirus 2 infection, without chronic kidney disease, in three ICUs from March 2020 to July 2020. AKI was assessed by hourly urine output and daily plasma creatinine. MAIN OUTCOMES AND MEASURES: Luminex and enzyme-linked immunosorbent assay were used to analyze 48 plasma protein biomarkers across six pathophysiological domains, which were tested with mixed-effects models. RESULTS: Of 177 included patients, 106 (59.9%) had AKI within the first 48 hours of admission, of whom 76 (71.7%) had persistent AKI and 30 (28.3%) transient AKI. Those with persistent AKI often had obesity, hypertension, and a higher Sequential Organ Failure Assessment score due to the renal component. Longitudinal analyses revealed that seven proteins were elevated in persistent AKI compared with no AKI. These were related to inflammation (triggering receptor expressed on myeloid cells 1, p < 0.001; tumor necrosis factor receptor 1, p < 0.001; procalcitonin, p = 0.001), complement activation (mannan-binding lectin serine protease-2, p = 0.001), kidney dysfunction (cystatin C, p < 0.001; neutrophil gelatinase-associated lipocalin, p < 0.001), and lung dysfunction (Clara cell secretory protein 16, p < 0.001). AKI (duration) was not associated with differences in the cytokine signaling, endothelial cell activation, or coagulation domains. CONCLUSIONS AND RELEVANCE: In contrast with sepsis-associated AKI, primarily inflammation-related biomarker levels correlated with COVID-AKI persistence. This study offers insights into COVID-AKI and may guide approaches to mitigate its persistence.

Multiclass segmentation of microanatomy in kidney biopsies is an important and non-trivial task in computational renal pathology. In a multicenter study, we densely annotated basic anatomical objects (glomeruli, tubules, and vessels) in 261 regions of interest of 147 kidney biopsy WSIs sourced from the archives of hospitals in Amsterdam, Utrecht, and Leiden (Netherlands). And we trained multiple UNet- and Mask2Former-based models on WSI-level and patch-level splitting methods, and compared their performance across training strategies. Test performance was assessed on 24 annotated renal WSIs from Leuven (Belgium) with sensitivity analysis on the extent of fibrosis and inflammation. At the patch-level, UNet-ResNet18 achieved comparable performances to M2F-Swin-B with average Intersection over Union of all classes (A-IoU, 0.84 vs 0.94), as well as per-class IoU. However, at the WSI-level, M2F-Swin-B significantly surpassed UNet-ResNet18 with large margins on A-IoU (0.84 vs 0.48), with similar observed in per-class IoU. Notably, M2F-Swin-B outperformed UNet-ResNet18 in scenarios characterized by a higher degree of fibrosis and inflammation (A-IoU, 0.76 vs 0.66). Furthermore, at the WSI-level, M2F-Swin-B achieved IoU score of arteries to 0.58, whereas UNet-ResNet18 only achieved 0.33. In this study, we found that the attention mechanism in Mask2Former enables visibly crisper and more uniform segmentation, particularly when data is inadequate. Mask2Former-based models outperform UNet-based models in challenging areas from inflamed and fibrotic renal biopsies.