Grainyhead-like 2 Deficiency and Kidney Cyst Growth in a Mouse Model.

BACKGROUND: The transcription factor Grainyhead-like 2 (GRHL2) plays a crucial role in maintaining the epithelial barrier properties of the renal collecting duct and is essential for osmoregulation. We noticed a reduction in GRHL2 expression in cysts derived from the collecting ducts in kidneys affected by Autosomal Dominant Polycystic Kidney Disease (ADPKD). However, the specific role of GRHL2 in cystic kidney disease remains unknown. METHODS: The functional role of the transcription factor Grhl2 in the context of cystic kidney disease was examined through analysis of its expression pattern in patient samples with ADPKD and generating a transgenic cystic kidney disease (TCKD) mouse model by overexpressing the human proto-oncogene c-MYC in kidney collecting ducts. Next, TCKD mice bred with collecting duct-specific Grhl2 knockout mice (Grhl2KO). The resulting TCKD-Grhl2KO mice and their littermates were examined by various types of histological and biochemical assays and gene profiling analysis via RNA-seq. RESULTS: A comprehensive examination of kidney samples from patients with ADPKD revealed GRHL2 downregulation in collecting duct-derived cyst epithelia. Comparative analysis of TCKD and TCKD-Grhl2KO mice exhibited that the collecting duct-specific deletion of Grhl2 resulted in markedly aggravated cyst growth, worsened kidney dysfunction, and shortened life span. Furthermore, transcriptomic analyses indicated sequential downregulation of kidney epithelial cyst development regulators (Frem2, Muc1, Cdkn2c, Pkd2, and Tsc1) during cyst progression in kidneys of TCKD-Grhl2KO mice which included presumed direct Grhl2 target genes. CONCLUSIONS: These results suggest GRHL2 as a potential progression modifier, especially for cysts originating from collecting ducts.

Histopathological biomarkers for predicting the tumour accumulation of nanomedicines.

The clinical prospects of cancer nanomedicines depend on effective patient stratification. Here we report the identification of predictive biomarkers of the accumulation of nanomedicines in tumour tissue. By using supervised machine learning on data of the accumulation of nanomedicines in tumour models in mice, we identified the densities of blood vessels and of tumour-associated macrophages as key predictive features. On the basis of these two features, we derived a biomarker score correlating with the concentration of liposomal doxorubicin in tumours and validated it in three syngeneic tumour models in immunocompetent mice and in four cell-line-derived and six patient-derived tumour xenografts in mice. The score effectively discriminated tumours according to the accumulation of nanomedicines (high versus low), with an area under the receiver operating characteristic curve of 0.91. Histopathological assessment of 30 tumour specimens from patients and of 28 corresponding primary tumour biopsies confirmed the score's effectiveness in predicting the tumour accumulation of liposomal doxorubicin. Biomarkers of the tumour accumulation of nanomedicines may aid the stratification of patients in clinical trials of cancer nanomedicines.

A Machine Learning-Driven Virtual Biopsy System For Kidney Transplant Patients.

In kidney transplantation, day-zero biopsies are used to assess organ quality and discriminate between donor-inherited lesions and those acquired post-transplantation. However, many centers do not perform such biopsies since they are invasive, costly and may delay the transplant procedure. We aim to generate a non-invasive virtual biopsy system using routinely collected donor parameters. Using 14,032 day-zero kidney biopsies from 17 international centers, we develop a virtual biopsy system. 11 basic donor parameters are used to predict four Banff kidney lesions: arteriosclerosis, arteriolar hyalinosis, interstitial fibrosis and tubular atrophy, and the percentage of renal sclerotic glomeruli. Six machine learning models are aggregated into an ensemble model. The virtual biopsy system shows good performance in the internal and external validation sets. We confirm the generalizability of the system in various scenarios. This system could assist physicians in assessing organ quality, optimizing allograft allocation together with discriminating between donor derived and acquired lesions post-transplantation.

Inhibition of the renal apical sodium dependent bile acid transporter prevents cholemic nephropathy in mice with obstructive cholestasis.

BACKGROUND & AIMS: Cholemic nephropathy (CN) is a severe complication of cholestatic liver diseases for which there is no specific treatment. We revisited its pathophysiology with the aim of identifying novel therapeutic strategies. METHODS: Cholestasis was induced by bile duct ligation (BDL) in mice. Bile flux in kidneys and livers was visualized by intravital imaging, supported by MALDI mass spectrometry imaging and liquid chromatography-tandem mass spectrometry. The effect of AS0369, a systemically bioavailable apical sodium-dependent bile acid transporter (ASBT) inhibitor, was evaluated by intravital imaging, RNA-sequencing, histological, blood, and urine analyses. Translational relevance was assessed in kidney biopsies from patients with CN, mice with a humanized bile acid (BA) spectrum, and via analysis of serum BAs and KIM-1 (kidney injury molecule 1) in patients with liver disease and hyperbilirubinemia. RESULTS: Proximal tubular epithelial cells (TECs) reabsorbed and enriched BAs, leading to oxidative stress and death of proximal TECs, casts in distal tubules and collecting ducts, peritubular capillary leakiness, and glomerular cysts. Renal ASBT inhibition by AS0369 blocked BA uptake into TECs and prevented kidney injury up to 6 weeks after BDL. Similar results were obtained in mice with humanized BA composition. In patients with advanced liver disease, serum BAs were the main determinant of KIM-1 levels. ASBT expression in TECs was preserved in biopsies from patients with CN, further highlighting the translational potential of targeting ASBT to treat CN. CONCLUSIONS: BA enrichment in proximal TECs followed by oxidative stress and cell death is a key early event in CN. Inhibiting renal ASBT and consequently BA enrichment in TECs prevents CN and systemically decreases BA concentrations. IMPACT AND IMPLICATIONS: Cholemic nephropathy (CN) is a severe complication of cholestasis and an unmet clinical need. We demonstrate that CN is triggered by the renal accumulation of bile acids (BAs) that are considerably increased in the systemic blood. Specifically, the proximal tubular epithelial cells of the kidney take up BAs via the apical sodium-dependent bile acid transporter (ASBT). We developed a therapeutic compound that blocks ASBT in the kidneys, prevents BA overload in tubular epithelial cells, and almost completely abolished all disease hallmarks in a CN mouse model. Renal ASBT inhibition represents a potential therapeutic strategy for patients with CN.

Operational greenhouse-gas emissions of deep learning in digital pathology: a modelling study.

BACKGROUND: Deep learning is a promising way to improve health care. Image-processing medical disciplines, such as pathology, are expected to be transformed by deep learning. The first clinically applicable deep-learning diagnostic support tools are already available in cancer pathology, and their number is increasing. However, data on the environmental sustainability of these tools are scarce. We aimed to conduct an environmental-sustainability analysis of a theoretical implementation of deep learning in patient-care pathology. METHODS: For this modelling study, we first assembled and calculated relevant data and parameters of a digital-pathology workflow. Data were breast and prostate specimens from the university clinic at the Institute of Pathology of the Rheinisch-Westfälische Technische Hochschule Aachen (Aachen, Germany), for which commercially available deep learning was already available. Only specimens collected between Jan 1 and Dec 31, 2019 were used, to omit potential biases due to the COVID-19 pandemic. Our final selection was based on 2 representative weeks outside holidays, covering different types of specimens. To calculate carbon dioxide (CO2) or CO2 equivalent (CO2 eq) emissions of deep learning in pathology, we gathered relevant data for exact numbers and sizes of whole-slide images (WSIs), which were generated by scanning histopathology samples of prostate and breast specimens. We also evaluated different data input scenarios (including all slide tiles, only tiles containing tissue, or only tiles containing regions of interest). To convert estimated energy consumption from kWh to CO2 eq, we used the internet protocol address of the computational server and the Electricity Maps database to obtain information on the sources of the local electricity grid (ie, renewable vs non-renewable), and estimated the number of trees and proportion of the local and world's forests needed to sequester the CO2 eq emissions. We calculated the computational requirements and CO2 eq emissions of 30 deep-learning models that varied in task and size. The first scenario represented the use of one commercially available deep-learning model for one task in one case (1-task), the second scenario considered two deep-learning models for two tasks per case (2-task), the third scenario represented a future, potentially automated workflow that could handle 7 tasks per case (7-task), and the fourth scenario represented the use of a single potential, large, computer-vision model that could conduct multiple tasks (multitask). We also compared the performance (ie, accuracy) and CO2 eq emissions of different deep-learning models for the classification of renal cell carcinoma on WSIs, also from Rheinisch-Westfälische Technische Hochschule Aachen. We also tested other approaches to reducing CO2 eq emissions, including model pruning and an alternative method for histopathology analysis (pathomics). FINDINGS: The pathology database contained 35 552 specimens (237 179 slides), 6420 of which were prostate specimens (10 115 slides) and 11 801 of which were breast specimens (19 763 slides). We selected and subsequently digitised 140 slides from eight breast-cancer cases and 223 slides from five prostate-cancer cases. Applying large deep-learning models on all WSI tiles of prostate and breast pathology cases would result in yearly CO2 eq emissions of 7·65 metric tons (t; 95% CI 7·62-7·68) with the use of a single deep-learning model per case; yearly CO2 eq emissions were up to 100·56 t (100·21-100·99) with the use of seven deep-learning models per case. CO2 eq emissions for different deep-learning model scenarios, data inputs, and deep-learning model sizes for all slides varied from 3·61 t (3·59-3·63) to 2795·30 t (1177·51-6482·13. For the estimated number of overall pathology cases worldwide, the yearly CO2 eq emissions varied, reaching up to 16 megatons (Mt) of CO2 eq, requiring up to 86 590 km2 (0·22%) of world forest to sequester the CO2 eq emissions. Use of the 7-task scenario and small deep-learning models on slides containing tissue only could substantially reduce CO2 eq emissions worldwide by up to 141 times (0·1 Mt, 95% CI 0·1-0·1). Considering the local environment in Aachen, Germany, the maximum CO2 eq emission from the use of deep learning in digital pathology only would require 32·8% (95% CI 13·8-76·6) of the local forest to sequester the CO2 eq emissions. A single pathomics run on a tissue could provide information that was comparable to or even better than the output of multitask deep-learning models, but with 147 times reduced CO2 eq emissions. INTERPRETATION: Our findings suggest that widespread use of deep learning in pathology might have considerable global-warming potential. The medical community, policy decision makers, and the public should be aware of this potential and encourage the use of CO2 eq emissions reduction strategies where possible. FUNDING: German Research Foundation, European Research Council, German Federal Ministry of Education and Research, Health, Economic Affairs and Climate Action, and the Innovation Fund of the Federal Joint Committee.

IgA nephropathy.

IgA nephropathy (IgAN), the most prevalent primary glomerulonephritis worldwide, carries a considerable lifetime risk of kidney failure. Clinical manifestations of IgAN vary from asymptomatic with microscopic or intermittent macroscopic haematuria and stable kidney function to rapidly progressive glomerulonephritis. IgAN has been proposed to develop through a 'four-hit' process, commencing with overproduction and increased systemic presence of poorly O-glycosylated galactose-deficient IgA1 (Gd-IgA1), followed by recognition of Gd-IgA1 by antiglycan autoantibodies, aggregation of Gd-IgA1 and formation of polymeric IgA1 immune complexes and, lastly, deposition of these immune complexes in the glomerular mesangium, leading to kidney inflammation and scarring. IgAN can only be diagnosed by kidney biopsy. Extensive, optimized supportive care is the mainstay of therapy for patients with IgAN. For those at high risk of disease progression, the 2021 KDIGO Clinical Practice Guideline suggests considering a 6-month course of systemic corticosteroid therapy; however, the efficacy of systemic steroid treatment is under debate and serious adverse effects are common. Advances in understanding the pathophysiology of IgAN have led to clinical trials of novel targeted therapies with acceptable safety profiles, including SGLT2 inhibitors, endothelin receptor blockers, targeted-release budesonide, B cell proliferation and differentiation inhibitors, as well as blockade of complement components.

Mapping cardiac remodeling in chronic kidney disease.

Patients with advanced chronic kidney disease (CKD) mostly die from sudden cardiac death and recurrent heart failure. The mechanisms of cardiac remodeling are largely unclear. To dissect molecular and cellular mechanisms of cardiac remodeling in CKD in an unbiased fashion, we performed left ventricular single-nuclear RNA sequencing in two mouse models of CKD. Our data showed a hypertrophic response trajectory of cardiomyocytes with stress signaling and metabolic changes driven by soluble uremia-related factors. We mapped fibroblast to myofibroblast differentiation in this process and identified notable changes in the cardiac vasculature, suggesting inflammation and dysfunction. An integrated analysis of cardiac cellular responses to uremic toxins pointed toward endothelin-1 and methylglyoxal being involved in capillary dysfunction and TNFα driving cardiomyocyte hypertrophy in CKD, which was validated in vitro and in vivo. TNFα inhibition in vivo ameliorated the cardiac phenotype in CKD. Thus, interventional approaches directed against uremic toxins, such as TNFα, hold promise to ameliorate cardiac remodeling in CKD.

Kidney fibrosis: Emerging diagnostic and therapeutic strategies.

An increasing number of patients worldwide suffers from chronic kidney disease (CKD). CKD is accompanied by kidney fibrosis, which affects all compartments of the kidney, i.e., the glomeruli, tubulointerstitium, and vasculature. Fibrosis is the best predictor of progression of kidney diseases. Currently, there is no specific anti-fibrotic therapy for kidney patients and invasive renal biopsy remains the only option for specific detection and quantification of kidney fibrosis. Here we review emerging diagnostic approaches and potential therapeutic options for fibrosis. We discuss how translational research could help to establish fibrosis-specific endpoints for clinical trials, leading to improved patient stratification and potentially companion diagnostics, and facilitating and optimizing development of novel anti-fibrotic therapies for kidney patients.

Multi-institutional PET/CT image segmentation using federated deep transformer learning.

BACKGROUND AND OBJECTIVE: Generalizable and trustworthy deep learning models for PET/CT image segmentation necessitates large diverse multi-institutional datasets. However, legal, ethical, and patient privacy issues challenge sharing of datasets between different centers. To overcome these challenges, we developed a federated learning (FL) framework for multi-institutional PET/CT image segmentation. METHODS: A dataset consisting of 328 FL (HN) cancer patients who underwent clinical PET/CT examinations gathered from six different centers was enrolled. A pure transformer network was implemented as fully core segmentation algorithms using dual channel PET/CT images. We evaluated different frameworks (single center-based, centralized baseline, as well as seven different FL algorithms) using 68 PET/CT images (20% of each center data). In particular, the implemented FL algorithms include clipping with the quantile estimator (ClQu), zeroing with the quantile estimator (ZeQu), federated averaging (FedAvg), lossy compression (LoCo), robust aggregation (RoAg), secure aggregation (SeAg), and Gaussian differentially private FedAvg with adaptive quantile clipping (GDP-AQuCl). RESULTS: The Dice coefficient was 0.80±0.11 for both centralized and SeAg FL algorithms. All FL approaches achieved centralized learning model performance with no statistically significant differences. Among the FL algorithms, SeAg and GDP-AQuCl performed better than the other techniques. However, there was no statistically significant difference. All algorithms, except the center-based approach, resulted in relative errors less than 5% for SUVmax and SUVmean for all FL and centralized methods. Centralized and FL algorithms significantly outperformed the single center-based baseline. CONCLUSIONS: The developed FL-based (with centralized method performance) algorithms exhibited promising performance for HN tumor segmentation from PET/CT images.

Macrophage migration inhibitory factor (MIF) and its homolog D-dopachrome tautomerase (D-DT) are significant promotors of UVB- but not chemically induced non-melanoma skin cancer.

Non-melanoma skin cancer (NMSC) is the most common cancer in Caucasians worldwide. We investigated the pathophysiological role of MIF and its homolog D-DT in UVB- and chemically induced NMSC using Mif-/-, D-dt-/- and Mif-/-/D-dt-/- mice on a hairless SKH1 background. Knockout of both cytokines showed similar attenuating effects on inflammation after acute UVB irradiation and tumor formation during chronic UVB irradiation, without additive protective effects noted in double knockout mice, indicating that both cytokines activate a similar signaling threshold. In contrast, genetic deletion of Mif and D-dt had no major effects on chemically induced skin tumors. To get insight into the contributing mechanisms, we used an in vitro 3D skin model with incorporated macrophages. Application of recombinant MIF and D-DT led to an accumulation of macrophages within the epidermal part that could be reversed by selective inhibitors of MIF and D-DT pathways. In summary, our data indicate that MIF and D-DT contribute to the development and progression of UVB- but not chemically induced NMSC, a role at least partially accounted by effects of both cytokines on epidermal macrophage accumulation. These data highlight that MIF and D-DT are both potential therapeutic targets for the prevention of photocarcinogenesis but not chemical carcinogenesis.

Biallelic variants in CRIPT cause a Rothmund-Thomson-like syndrome with increased cellular senescence.

PURPOSE: Rothmund-Thomson syndrome (RTS) is characterized by poikiloderma, sparse hair, small stature, skeletal defects, cancer, and cataracts, resembling features of premature aging. RECQL4 and ANAPC1 are the 2 known disease genes associated with RTS in >70% of cases. We describe RTS-like features in 5 individuals with biallelic variants in CRIPT (OMIM 615789). METHODS: Two newly identified and 4 published individuals with CRIPT variants were systematically compared with those with RTS using clinical data, computational analysis of photographs, histologic analysis of skin, and cellular studies on fibroblasts. RESULTS: All CRIPT individuals fulfilled the diagnostic criteria for RTS and additionally had neurodevelopmental delay and seizures. Using computational gestalt analysis, CRIPT individuals showed greatest facial similarity with individuals with RTS. Skin biopsies revealed a high expression of senescence markers (p53/p16/p21) and the senescence-associated ß-galactosidase activity was elevated in CRIPT-deficient fibroblasts. RECQL4- and CRIPT-deficient fibroblasts showed an unremarkable mitotic progression and unremarkable number of mitotic errors and no or only mild sensitivity to genotoxic stress by ionizing radiation, mitomycin C, hydroxyurea, etoposide, and potassium bromate. CONCLUSION: CRIPT causes an RTS-like syndrome associated with neurodevelopmental delay and epilepsy. At the cellular level, RECQL4- and CRIPT-deficient cells display increased senescence, suggesting shared molecular mechanisms leading to the clinical phenotypes.

KIT D816V Mast Cells Derived from Induced Pluripotent Stem Cells Recapitulate Systemic Mastocytosis Transcriptional Profile.

Mast cells (MCs) represent a population of hematopoietic cells with a key role in innate and adaptive immunity and are well known for their detrimental role in allergic responses. Yet, MCs occur in low abundance, which hampers their detailed molecular analysis. Here, we capitalized on the potential of induced pluripotent stem (iPS) cells to give rise to all cells in the body and established a novel and robust protocol for human iPS cell differentiation toward MCs. Relying on a panel of systemic mastocytosis (SM) patient-specific iPS cell lines carrying the KIT D816V mutation, we generated functional MCs that recapitulate SM disease features: increased number of MCs, abnormal maturation kinetics and activated phenotype, CD25 and CD30 surface expression and a transcriptional signature characterized by upregulated expression of innate and inflammatory response genes. Therefore, human iPS cell-derived MCs are a reliable, inexhaustible, and close-to-human tool for disease modeling and pharmacological screening to explore novel MC therapeutics.

Platelet-instructed SPP1+ macrophages drive myofibroblast activation in fibrosis in a CXCL4-dependent manner.

Fibrosis represents the common end stage of chronic organ injury independent of the initial insult, destroying tissue architecture and driving organ failure. Here we discover a population of profibrotic macrophages marked by expression of Spp1, Fn1, and Arg1 (termed Spp1 macrophages), which expands after organ injury. Using an unbiased approach, we identify the chemokine (C-X-C motif) ligand 4 (CXCL4) to be among the top upregulated genes during profibrotic Spp1 macrophage differentiation. In vitro and in vivo studies show that loss of Cxcl4 abrogates profibrotic Spp1 macrophage differentiation and ameliorates fibrosis after both heart and kidney injury. Moreover, we find that platelets, the most abundant source of CXCL4 in vivo, drive profibrotic Spp1 macrophage differentiation. Single nuclear RNA sequencing with ligand-receptor interaction analysis reveals that macrophages orchestrate fibroblast activation via Spp1, Fn1, and Sema3 crosstalk. Finally, we confirm that Spp1 macrophages expand in both human chronic kidney disease and heart failure.

Next-Generation Morphometry for pathomics-data mining in histopathology.

Pathology diagnostics relies on the assessment of morphology by trained experts, which remains subjective and qualitative. Here we developed a framework for large-scale histomorphometry (FLASH) performing deep learning-based semantic segmentation and subsequent large-scale extraction of interpretable, quantitative, morphometric features in non-tumour kidney histology. We use two internal and three external, multi-centre cohorts to analyse over 1000 kidney biopsies and nephrectomies. By associating morphometric features with clinical parameters, we confirm previous concepts and reveal unexpected relations. We show that the extracted features are independent predictors of long-term clinical outcomes in IgA-nephropathy. We introduce single-structure morphometric analysis by applying techniques from single-cell transcriptomics, identifying distinct glomerular populations and morphometric phenotypes along a trajectory of disease progression. Our study provides a concept for Next-generation Morphometry (NGM), enabling comprehensive quantitative pathology data mining, i.e., pathomics.

Stain-Independent Deep Learning-Based Analysis of Digital Kidney Histopathology.

Convolutional neural network (CNN)-based image analysis applications in digital pathology (eg, tissue segmentation) require a large amount of annotated data and are mostly trained and applicable on a single stain. Here, a novel concept based on stain augmentation is proposed to develop stain-independent CNNs requiring only one annotated stain. In this benchmark study on stain independence in digital pathology, this approach is comprehensively compared with state-of-the-art techniques including image registration and stain translation, and several modifications thereof. A previously developed CNN for segmentation of periodic acid-Schiff-stained kidney histology was used and applied to various immunohistochemical stainings. Stain augmentation showed very high performance in all evaluated stains and outperformed all other techniques in all structures and stains. Without the need for additional annotations, it enabled segmentation on immunohistochemical stainings with performance nearly comparable to that of the annotated periodic acid-Schiff stain and could further uphold performance on several held-out stains not seen during training. Herein, examples of how this framework can be applied for compartment-specific quantification of immunohistochemical stains for inflammation and fibrosis in animal models and patient biopsy specimens are presented. The results show that stain augmentation is a highly effective approach to enable stain-independent applications of deep-learning segmentation algorithms. This opens new possibilities for broad implementation in digital pathology.

The Role of Platelet-Derived Growth Factor in Focal Segmental Glomerulosclerosis.

BACKGROUND: FSGS is the final common pathway to nephron loss in most forms of severe or progressive glomerular injury. Although podocyte injury initiates FSGS, parietal epithelial cells (PECs) are the main effectors. Because PDGF takes part in fibrotic processes, we hypothesized that the ligand PDGF-B and its receptor PDGFR- ß participate in the origin and progression of FSGS. METHODS: We challenged Thy1.1 transgenic mice, which express Thy1.1 in the podocytes, with anti-Thy1.1 antibody to study the progression of FSGS. We investigated the role of PDGF in FSGS using challenged Thy1.1 mice, 5/6 nephrectomized mice, Col4 -/- (Alport) mice, patient kidney biopsies, and primary murine PECs, and challenged Thy1.1 mice treated with neutralizing anti-PDGF-B antibody therapy. RESULTS: The unchallenged Thy1.1 mice developed only mild spontaneous FSGS, whereas challenged mice developed progressive FSGS accompanied by a decline in kidney function. PEC activation, proliferation, and profibrotic phenotypic switch drove the FSGS. During disease, PDGF-B was upregulated in podocytes, whereas PDGFR- ß was upregulated in PECs from both mice and patients with FSGS. Short- and long-term treatment with PDGF-B neutralizing antibody improved kidney function and reduced FSGS, PEC proliferation, and profibrotic activation. In vitro , stimulation of primary murine PECs with PDGF-B recapitulated in vivo findings with PEC activation and proliferation, which was inhibited by PDGF-B antibody or imatinib. CONCLUSION: PDGF-B-PDGFR- ß molecular crosstalk between podocytes and PECs drives glomerulosclerosis and the progression of FSGS. PODCAST: This article contains a podcast at.

Direct prediction of genetic aberrations from pathology images in gastric cancer with swarm learning.

BACKGROUND: Computational pathology uses deep learning (DL) to extract biomarkers from routine pathology slides. Large multicentric datasets improve performance, but such datasets are scarce for gastric cancer. This limitation could be overcome by Swarm Learning (SL). METHODS: Here, we report the results of a multicentric retrospective study of SL for prediction of molecular biomarkers in gastric cancer. We collected tissue samples with known microsatellite instability (MSI) and Epstein-Barr Virus (EBV) status from four patient cohorts from Switzerland, Germany, the UK and the USA, storing each dataset on a physically separate computer. RESULTS: On an external validation cohort, the SL-based classifier reached an area under the receiver operating curve (AUROC) of 0.8092 (± 0.0132) for MSI prediction and 0.8372 (± 0.0179) for EBV prediction. The centralized model, which was trained on all datasets on a single computer, reached a similar performance. CONCLUSIONS: Our findings demonstrate the feasibility of SL-based molecular biomarkers in gastric cancer. In the future, SL could be used for collaborative training and, thus, improve the performance of these biomarkers. This may ultimately result in clinical-grade performance and generalizability.

Adult human kidney organoids originate from CD24+ cells and represent an advanced model for adult polycystic kidney disease.

Adult kidney organoids have been described as strictly tubular epithelia and termed tubuloids. While the cellular origin of tubuloids has remained elusive, here we report that they originate from a distinct CD24+ epithelial subpopulation. Long-term-cultured CD24+ cell-derived tubuloids represent a functional human kidney tubule. We show that kidney tubuloids can be used to model the most common inherited kidney disease, namely autosomal dominant polycystic kidney disease (ADPKD), reconstituting the phenotypic hallmark of this disease with cyst formation. Single-cell RNA sequencing of CRISPR-Cas9 gene-edited PKD1- and PKD2-knockout tubuloids and human ADPKD and control tissue shows similarities in upregulation of disease-driving genes. Furthermore, in a proof of concept, we demonstrate that tolvaptan, the only approved drug for ADPKD, has a significant effect on cyst size in tubuloids but no effect on a pluripotent stem cell-derived model. Thus, tubuloids are derived from a tubular epithelial subpopulation and represent an advanced system for ADPKD disease modeling.

The fatal trajectory of pulmonary COVID-19 is driven by lobular ischemia and fibrotic remodelling.

BACKGROUND: COVID-19 is characterized by a heterogeneous clinical presentation, ranging from mild symptoms to severe courses of disease. 9-20% of hospitalized patients with severe lung disease die from COVID-19 and a substantial number of survivors develop long-COVID. Our objective was to provide comprehensive insights into the pathophysiology of severe COVID-19 and to identify liquid biomarkers for disease severity and therapy response. METHODS: We studied a total of 85 lungs (n = 31 COVID autopsy samples; n = 7 influenza A autopsy samples; n = 18 interstitial lung disease explants; n = 24 healthy controls) using the highest resolution Synchrotron radiation-based hierarchical phase-contrast tomography, scanning electron microscopy of microvascular corrosion casts, immunohistochemistry, matrix-assisted laser desorption ionization mass spectrometry imaging, and analysis of mRNA expression and biological pathways. Plasma samples from all disease groups were used for liquid biomarker determination using ELISA. The anatomic/molecular data were analyzed as a function of patients' hospitalization time. FINDINGS: The observed patchy/mosaic appearance of COVID-19 in conventional lung imaging resulted from microvascular occlusion and secondary lobular ischemia. The length of hospitalization was associated with increased intussusceptive angiogenesis. This was associated with enhanced angiogenic, and fibrotic gene expression demonstrated by molecular profiling and metabolomic analysis. Increased plasma fibrosis markers correlated with their pulmonary tissue transcript levels and predicted disease severity. Plasma analysis confirmed distinct fibrosis biomarkers (TSP2, GDF15, IGFBP7, Pro-C3) that predicted the fatal trajectory in COVID-19. INTERPRETATION: Pulmonary severe COVID-19 is a consequence of secondary lobular microischemia and fibrotic remodelling, resulting in a distinctive form of fibrotic interstitial lung disease that contributes to long-COVID. FUNDING: This project was made possible by a number of funders. The full list can be found within the Declaration of interests / Acknowledgements section at the end of the manuscript.

Adversarial attacks and adversarial robustness in computational pathology.

Artificial Intelligence (AI) can support diagnostic workflows in oncology by aiding diagnosis and providing biomarkers directly from routine pathology slides. However, AI applications are vulnerable to adversarial attacks. Hence, it is essential to quantify and mitigate this risk before widespread clinical use. Here, we show that convolutional neural networks (CNNs) are highly susceptible to white- and black-box adversarial attacks in clinically relevant weakly-supervised classification tasks. Adversarially robust training and dual batch normalization (DBN) are possible mitigation strategies but require precise knowledge of the type of attack used in the inference. We demonstrate that vision transformers (ViTs) perform equally well compared to CNNs at baseline, but are orders of magnitude more robust to white- and black-box attacks. At a mechanistic level, we show that this is associated with a more robust latent representation of clinically relevant categories in ViTs compared to CNNs. Our results are in line with previous theoretical studies and provide empirical evidence that ViTs are robust learners in computational pathology. This implies that large-scale rollout of AI models in computational pathology should rely on ViTs rather than CNN-based classifiers to provide inherent protection against perturbation of the input data, especially adversarial attacks.