Interpretable machine learning-based individual analysis of acute kidney injury in immune checkpoint inhibitor therapy.

BACKGROUND: Acute kidney injury (AKI) is a critical complication of immune checkpoint inhibitor therapy. Since the etiology of AKI in patients undergoing cancer therapy varies, clarifying underlying causes in individual cases is critical for optimal cancer treatment. Although it is essential to individually analyze immune checkpoint inhibitor-treated patients for underlying pathologies for each AKI episode, these analyses have not been realized. Herein, we aimed to individually clarify the underlying causes of AKI in immune checkpoint inhibitor-treated patients using a new clustering approach with Shapley Additive exPlanations (SHAP). METHODS: We developed a gradient-boosting decision tree-based machine learning model continuously predicting AKI within 7 days, using the medical records of 616 immune checkpoint inhibitor-treated patients. The temporal changes in individual predictive reasoning in AKI prediction models represented the key features contributing to each AKI prediction and clustered AKI patients based on the features with high predictive contribution quantified in time series by SHAP. We searched for common clinical backgrounds of AKI patients in each cluster, compared with annotation by three nephrologists. RESULTS: One hundred and twelve patients (18.2%) had at least one AKI episode. They were clustered per the key feature, and their SHAP value patterns, and the nephrologists assessed the clusters' clinical relevance. Receiver operating characteristic analysis revealed that the area under the curve was 0.880. Patients with AKI were categorized into four clusters with significant prognostic differences (p = 0.010). The leading causes of AKI for each cluster, such as hypovolemia, drug-related, and cancer cachexia, were all clinically interpretable, which conventional approaches cannot obtain. CONCLUSION: Our results suggest that the clustering method of individual predictive reasoning in machine learning models can be applied to infer clinically critical factors for developing each episode of AKI among patients with multiple AKI risk factors, such as immune checkpoint inhibitor-treated patients.

Proteinuria frequency and subsequent renal dysfunction in bevacizumab-treated patients: a single center, retrospective, observational study.

BACKGROUND: Proteinuria is a common adverse event observed during treatment with antivascular endothelial growth factor (VEGF) antibodies. Proteinuria is a risk factor for renal dysfunction and cardiovascular complications in patients with chronic kidney disease. However, the association between anti-VEGF antibody-induced proteinuria and renal dysfunction or cardiovascular complications remains unclear. METHODS: This retrospective, observational study included patients with cancer that were treated with bevacizumab (BV) at Kyoto University Hospital (Kyoto, Japan) between January 2006 and March 2018. Adverse event rates were compared between patients who developed qualitative ≥ 2 + proteinuria and those who developed < 1 + proteinuria. Adverse events were defined as renal dysfunction (i.e., ≥ 57% decrease in the eGFR, compared to the rate at the initial treatment) and hospitalization due to BV-associated cardiovascular complications and other adverse events. RESULTS: In total, 734 patients were included in this analysis. Renal dysfunction was more common in patients with ≥ 2 + proteinuria than in those with < 1 + proteinuria (13/199, 6.5% vs. 12/535, 2.3%). Seven of these 13 patients with ≥ 2 + proteinuria had transient reversible renal dysfunction. Only four (2.0%) patients had BV-associated renal dysfunction. Of the 734 patients, six patients, 16 patients, and 13 patients were hospitalized because of the adverse events of cardiovascular complications, thromboembolisms, and cerebrovascular complications, respectively. No relationship was observed between these adverse events and proteinuria. CONCLUSION: BV treatment-induced proteinuria was not associated with renal dysfunction or other adverse events. Continuing BV with caution is a possible treatment option, even after proteinuria develops, in patients with cancer and a limited prognosis.

Evaluation of Kidney Histological Images Using Unsupervised Deep Learning.

INTRODUCTION: Evaluating histopathology via machine learning has gained research and clinical interest, and the performance of supervised learning tasks has been described in various areas of medicine. Unsupervised learning of histological images has the advantage of reproducibility for labeling; however, the relationship between unsupervised evaluation and clinical information remains unclear in nephrology. METHODS: We propose an unsupervised approach combining convolutional neural networks (CNNs) and a visualization algorithm to cluster the histological images and calculate the score for patients. We applied the approach to the entire images or patched images of the glomerulus of kidney biopsy samples stained with hematoxylin and eosin obtained from 68 patients with immunoglobulin A nephropathy. We assessed the relationship between the obtained scores and clinical variables of urinary occult blood, urinary protein, serum creatinine (SCr), systolic blood pressure, and age. RESULTS: The glomeruli of the patients were classified into 12 distinct classes and 10 patches. The output of the fine-tuned CNN, which we defined as the histological scores, had significant relationships with assessed clinical variables. In addition, the clustering and visualization results suggested that the defined clusters captured important findings when evaluating renal histopathology. For the score of the patch-based cluster containing crescentic glomeruli, SCr (coefficient = 0.09, P = 0.019) had a significant relationship. CONCLUSION: The proposed approach could successfully extract features that were related to the clinical variables from the kidney biopsy images along with the visualization for interpretability. The approach could aid in the quantified evaluation of renal histopathology.

Spatiotemporal ATP Dynamics during AKI Predict Renal Prognosis.

BACKGROUND: Depletion of ATP in renal tubular cells plays the central role in the pathogenesis of kidney diseases. Nevertheless, inability to visualize spatiotemporal in vivo ATP distribution and dynamics has hindered further analysis. METHODS: A novel mouse line systemically expressing an ATP biosensor (an ATP synthase subunit and two fluorophores) revealed spatiotemporal ATP dynamics at single-cell resolution during warm and cold ischemic reperfusion (IR) with two-photon microscopy. This experimental system enabled quantification of fibrosis 2 weeks after IR and assessment of the relationship between the ATP recovery in acute phase and fibrosis in chronic phase. RESULTS: Upon ischemia induction, the ATP levels of proximal tubule (PT) cells decreased to the nadir within a few minutes, whereas those of distal tubule (DT) cells decreased gradually up to 1 hour. Upon reperfusion, the recovery rate of ATP in PTs was slower with longer ischemia. In stark contrast, ATP in DTs was quickly rebounded irrespective of ischemia duration. Morphologic changes of mitochondria in the acute phase support the observation of different ATP dynamics in the two segments. Furthermore, slow and incomplete ATP recovery of PTs in the acute phase inversely correlated with fibrosis in the chronic phase. Ischemia under conditions of hypothermia resulted in more rapid and complete ATP recovery with less fibrosis, providing a proof of concept for use of hypothermia to protect kidney tissues. CONCLUSIONS: Visualizing spatiotemporal ATP dynamics during IR injury revealed higher sensitivity of PT cells to ischemia compared with DT cells in terms of energy metabolism. The ATP dynamics of PTs in AKI might provide prognostic information.

Classification of glomerular pathological findings using deep learning and nephrologist-AI collective intelligence approach.

BACKGROUND: Automated classification of glomerular pathological findings is potentially beneficial in establishing an efficient and objective diagnosis in renal pathology. While previous studies have verified the artificial intelligence (AI) models for the classification of global sclerosis and glomerular cell proliferation, there are several other glomerular pathological findings required for diagnosis, and the comprehensive models for the classification of these major findings have not yet been reported. Whether the cooperation between these AI models and clinicians improves diagnostic performance also remains unknown. Here, we developed AI models to classify glomerular images for major findings required for pathological diagnosis and investigated whether those models could improve the diagnostic performance of nephrologists. METHODS: We used a dataset of 283 kidney biopsy cases comprising 15,888 glomerular images that were annotated by a total of 25 nephrologists. AI models to classify seven pathological findings: global sclerosis, segmental sclerosis, endocapillary proliferation, mesangial matrix accumulation, mesangial cell proliferation, crescent, and basement membrane structural changes, were constructed using deep learning by fine-tuning of InceptionV3 convolutional neural network. Subsequently, we compared the agreement to truth labels between majority decision among nephrologists with or without the AI model as a voter. RESULTS: Our model for global sclerosis showed high performance (area under the curve: periodic acid-Schiff, 0.986; periodic acid methenamine silver, 0.983); the models for the other findings also showed performance close to those of nephrologists. By adding the AI model output to majority decision among nephrologists, out of the 14 constructed models, the results of the majority decision showed improvement in sensitivity for 10 models (four of them were statistically significant) and specificity for eight models (five significant). CONCLUSION: Our study showed a proof-of-concept for the classification of multiple glomerular findings in a comprehensive method of deep learning and suggested its potential effectiveness in improving diagnostic accuracy of clinicians.

The relationship between cigarette smoking and the tongue microbiome in an East Asian population.

Background: The oral microbiome, which consists of various habitats, has been shown to be influenced by smoking. However, differences in the tongue microbiomes of current and former smokers, as well as their resultant functional consequences, have rarely been investigated in East Asian populations. Methods: We used 16S rRNA amplicon sequencing of tongue-coating samples obtained from East Asian subjects who were current, former, or never smokers to identify differences in their tongue microbiomes and related metagenomic functions. Two sets of participants from 2016 to 2017 (n = 657 and n = 187, respectively) were analyzed separately. Results: We found significant differences between the overall microbiome compositions of current versus never smokers (p = 0.0015), but not between former versus never smokers (p = 0.43) based on the weighted UniFrac distance. Twenty-nine of 43 investigated genera showed significantly different expression levels in current versus never smokers. Neisseria and Capnocytophaga were less abundant, and Streptococcus and Megasphaera were more abundant in current smokers. Moreover, the abundances of metagenomic pathways, including those related to nitrate reduction and the tricarboxylic acid cycle, were significantly different between current and never smokers. Conclusions: The tongue microbiomes and related metagenomic pathways of current smokers differ from those of never smokers among East Asians.

Metagenomic analysis of bacterial species in tongue microbiome of current and never smokers.

Cigarette smoking affects the oral microbiome, which is related to various systemic diseases. While studies that investigated the relationship between smoking and the oral microbiome by 16S rRNA amplicon sequencing have been performed, investigations involving metagenomic sequences are rare. We investigated the bacterial species composition in the tongue microbiome, as well as single-nucleotide variant (SNV) profiles and gene content of these species, in never and current smokers by utilizing metagenomic sequences. Among 234 never smokers and 52 current smokers, beta diversity, as assessed by weighted UniFrac measure, differed between never and current smokers (pseudo-F = 8.44, R2 = 0.028, p = 0.001). Among the 26 species that had sufficient coverage, the SNV profiles of Actinomyces graevenitzii, Megasphaera micronuciformis, Rothia mucilaginosa, Veillonella dispar, and one Veillonella sp. were significantly different between never and current smokers. Analysis of gene and pathway content revealed that genes related to the lipopolysaccharide biosynthesis pathway in Veillonella dispar were present more frequently in current smokers. We found that species-level tongue microbiome differed between never and current smokers, and 5 species from never and current smokers likely harbor different strains, as suggested by the difference in SNV frequency.

Myofibroblasts acquire retinoic acid-producing ability during fibroblast-to-myofibroblast transition following kidney injury.

Tubular injury and interstitial fibrosis are the hallmarks of chronic kidney disease. While recent studies have verified that proximal tubular injury triggers interstitial fibrosis, the impact of fibrosis on tubular injury and regeneration remains poorly understood. We generated a novel mouse model expressing diphtheria toxin receptor on renal fibroblasts to allow for the selective disruption of renal fibroblast function. Administration of diphtheria toxin induced upregulation of the tubular injury marker Ngal and caused tubular proliferation in healthy kidneys, whereas administration of diphtheria toxin attenuated tubular regeneration in fibrotic kidneys. Microarray analysis revealed down-regulation of the retinol biosynthesis pathway in diphtheria toxin-treated kidneys. Healthy proximal tubules expressed retinaldehyde dehydrogenase 2 (RALDH2), a rate-limiting enzyme in retinoic acid biosynthesis. After injury, proximal tubules lost RALDH2 expression, whereas renal fibroblasts acquired strong expression of RALDH2 during the transition to myofibroblasts in several models of kidney injury. The retinoic acid receptor (RAR) RARγ was expressed in proximal tubules both with and without injury, and αB-crystallin, the product of an RAR target gene, was strongly expressed in proximal tubules after injury. Furthermore, BMS493, an inverse agonist of RARs, significantly attenuated tubular proliferation in vitro. In human biopsy tissue from patients with IgA nephropathy, detection of RALDH2 in the interstitium correlated with older age and lower kidney function. These results suggest a role of retinoic acid signaling and cross-talk between fibroblasts and tubular epithelial cells during tubular injury and regeneration, and may suggest a beneficial effect of fibrosis in the early response to injury.

Membranous nephropathy associated with pregnancy: an anti-phospholipase A2 receptor antibody-positive case report.

Pregnancy and membranous nephropathy (MN) can occur concurrently with nephrotic syndrome. However, the pathophysiology of MN associated with pregnancy remains unclear, including the involvement of anti-M-type phospholipase A2 receptor (PLA2R) antibody, the major antigen of idiopathic MN (iMN). A treatment for the condition is also not established. We present the case of a 43-year-old pregnant female with incidental proteinuria and hypoalbuminemia. We made a diagnosis of nephrotic syndrome at 11 week gestation. Renal biopsy revealed iMN using predominant granular staining of IgG4 along the glomerular basement membrane. No secondary cause was identified. Oral glucocorticoid therapy was started from 17 week gestation and induced complete remission at 28 week gestation. A healthy infant was born at 38 week gestation. Glucocorticoid therapy was stopped postpartum without MN relapse. Anti-PLA2R antibody was later found to be positive using serum reserved from before treatment. In conclusion, we presented the case of a pregnant woman with iMN and anti-PLA2R antibodies, whose nephrotic syndrome was successfully controlled with oral glucocorticoids to reach complete remission, even after tapering off the medication. Pregnancy per se might be associated with iMN onset.

Renal-limited vasculitis with elevated levels of multiple antibodies.

Renal-limited vasculitis (RLV) is a type of anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis that presents with crescentic glomerulonephritis with no other organ involvement. Although several studies reported patients with crescentic glomerulonephritis who were dual positive for proteinase 3 (PR3)-ANCA and myeloperoxidase (MPO)-ANCA or ANCA and anti-glomerular basement membrane (GBM) antibody, patients positive for all three antibodies, i.e., triple-positive patients, were rarely reported. We herein report the case of a male with pauci-immune type crescentic glomerulonephritis positive for MPO-ANCA, PR3-ANCA, and anti-GBM antibody. Renal biopsy led to the definitive diagnosis of RLV with pauci-immune-type crescentic glomerulonephritis. Fluorescence immunostaining showed no linear deposition of IgG on GBM, indicating no involvement of anti-GBM associated diseases. Intensive therapy, including prednisolone, plasma exchange, and intravenous cyclophosphamide, was effective. We report the case of triple-positive patient with crescentic glomerulonephritis, who was successfully treated with glucocorticoid, plasma exchange, and cyclophosphamide, suggesting that treatment for RLV in the patient with serological triple antibodies positivity in the absence of linear IgG deposition could benefit from the combination therapy regimen for plasma exchange and primary induction of remission against microscopic polyangiitis.