key: cord-0282182-dma4wl7e
authors: Klocke, Jan; Kim, Seung Joon; Skopnik, Christopher M.; Hinze, Christian; Boltengagen, Anastasiya; Metzke, Diana; Grothgar, Emil; Prskalo, Luka; Wagner, Leonie; Freund, Paul; Görlich, Nina; Muench, Frédéric; Schmidt-Ott, Kai M.; Mashreghi, Mir-Farzin; Kocks, Christine; Eckardt, Kai-Uwe; Rajewsky, Nikolaus; Enghard, Philipp
title: Urinary single-cell sequencing captures intrarenal injury and repair processes in human acute kidney injury
date: 2022-02-17
journal: bioRxiv
DOI: 10.1101/2022.02.15.479234
sha: 7afa7176187ce77125348ba20ef469a2ef5cfd3f
doc_id: 282182
cord_uid: dma4wl7e

Acute kidney injury (AKI) is a major health issue, the outcome of which depends primarily on damage and reparative processes of tubular epithelial cells (TEC). Mechanisms underlying AKI remain incompletely understood, specific therapies are lacking and monitoring the course of AKI in clinical routine is confined to measuring urine output and plasma levels of filtration markers. Here we demonstrate feasibility and potential of a novel approach to assess the cellular and molecular dynamics of AKI by establishing a robust urine-to-single cell RNA sequencing (scRNAseq) pipeline for excreted kidney cells via flow cytometry sorting. We analyzed 42,608 single cell transcriptomes of 40 urine samples from 32 AKI patients and compared our data with reference material from human AKI post-mortem biopsies and published mouse data. We demonstrate that TEC transcriptomes mirror intrarenal pathology and reflect distinct injury and repair processes, including oxidative stress, inflammation, and tissue rearrangement. We also describe an AKI-specific abundant urinary excretion of progenitor-like cells. In conclusion, single cell transcriptomics of kidney cells excreted in urine provides non-invasive, unprecedented insight into cellular processes underlying AKI, thereby opening novel opportunities for target identification, AKI sub-categorization and monitoring of natural disease course and interventions.

Acute kidney injury (AKI) is a major health concern associated with significant morbidity and mortality (1, 2) . Causal therapies for preventing AKI or improving its recovery are still missing.

Renal tubular epithelial cells (TEC) are the primarily affected cells in AKI and also a key component in the resulting inflammation and healing processes (3) (4) (5) (6) .

Research on AKI and on the role of TEC must overcome three major challenges: a) finding a suitable way to study human disease pathomechanisms or effectively translate findings from mouse models to the human setting, b) provide clinicians with meaningful biomarkers to predict outcomes and c) understanding TEC repair to enable strategies to enhance regeneration and reduce scarring (7) . It is reasonable to believe that answers to all three challenges may lie in the urine.

Transcriptomic analyses in mouse AKI models on the single-cell level have rapidly enhanced and transformed our understanding of injured and recovering TEC states (5, 6, (8) (9) (10) . Recently the first biopsy-based human single-cell studies of the kidney and AKI have been reported (8, 11, 12) . However, the invasive nature of kidney biopsies limits scalability and represents a hurdle for translation into clinical routine. In contrast, analysis of kidney cells excreted in urine may provide non-invasive insights into altered cell physiology both in inflammation (13) (14) (15) (16) and

injury (17) . The feasibility of urinary single-cell RNA sequencing (scRNAseq) has been proven in chronic kidney diseases such as diabetic nephropathy (DN) and focal-segmental glomerulosclerosis (FSGS) (18) (19) (20) and indicates that the urine has untapped potential as a source for the non-invasive study of renal epithelial cells. Moreover, urine-derived progenitor or stem cells, which were first isolated in pediatric patients (21) , have since been studied extensively (22) . They are now a common source for inducible pluripotent stem cells (23) and

can be detected even in healthy individuals (20) .

Therefore, we hypothesized that urinary TEC, which are excreted with the urine after kidney damage, may be an apt target to study inflammation and regeneration in AKI via urinary scRNAseq. By analyzing 40 urine samples from 32 individuals with AKI, we observed different injury-related cell states that mirror findings from mouse disease models and human AKI kidney tissue. Our data show that urinary progenitor-like cells are a prominent feature in AKI patients and reveal dynamic changes in urinary cell composition over time.

The first goal of our study was to understand what cell types are to be expected in the urine sediment after AKI. Therefore, we performed scRNAseq on sediment derived from 40 fresh urine samples of 32 patients (Suppl. Table 1 , Suppl. Fig. 1 ). We collected samples at different timepoints 0 to 21 days after onset of AKI due to cardiac surgery (n=7), pneumonia (n=15, among them 14 cases of Coronavirus disease 2019 (COVID-19)), or other prerenal causes (n=10). After quality control filtering, we obtained a total of 42608 single cells with a median of 472 cells per sample as well as 1436 detected genes and 3991 unique transcripts per cell.

Data analysis (Uniform manifold approximation and projection (UMAP) Visualization, unbiased clustering and annotation based on known marker genes) revealed three cell type categories to be the main features of the AKI urine sediment (Fig. 1A+D , Suppl. Table 2 Interestingly, all three major cell types were detectable across most samples (Suppl. Fig. 4D ), despite the diversity of the overall quantity of captured high-quality single cell transcriptomes across samples, with a median of 472 (range 8 -5,900) cells/sample (Suppl. Fig.4D ). UGEC were featured more prominently in female patients (Fig. 1B+D, Suppl. Fig 5) . Thus, we asked whether other factors like differing AKI etiology (Fig. 1C+D ) or an infection with SARS-Cov-2 also influence the cellular urine composition in AKI. Therefore, we examined the data of different AKI entities separately and compared their cellular composition: Individuals with cardiac-surgery (CS) or pneumonia / COVID-19 associated AKI showed a similar pattern of urinary composition with many renal cells (33 and 36% of all single-cell transcriptomes) and differed only in a more pronounced (B) lymphocyte signal in the pneumonia patient pool (Cluster 17 in Suppl. Fig. 6 .1A, 6.2). The prerenal AKI pool contributed mostly to myeloid (clusters 10-15) and urogenital (clusters 19, 20) cells and had much less kidney parenchymal cells (7%, clusters 1-9), consistent with less severe tubular damage.

SARS-Cov-2 infection, however, had only marginal effects on the AKI urine sediment, despite being discussed as an active driver of kidney inflammation and fibrosis (27) : Urinary cells from COVID-19 AKI patients showed no transcriptional activity for the SARS-Cov-2 genome.

Although previously reported COVID-19 associated genes (27, 28) (PLCG2, AFDN) as well as inflammation (CCL2, IFITM3) and damage markers (HSPA1A) were differentially expressed in TEC of COVID-19 patients in comparison to the Non-Covid AKI group (Suppl. Fig. 7 ), the differences were marginal, none of these signatures were exclusive for COVID-19 patients and no distinct TEC population was detected.

Overall, urine samples from AKI patients regularly featured immune and most importantly epithelial cells from the kidney, potentially reflecting kidney damage type and severity.

To use urinary cellular analysis as a proxy for AKI pathophysiology, it is crucial to know which kidney features are released into the urine and how detailed those cells can be characterized.

For that reason, we re-analyzed the 12,853 AKI single cell transcriptomes which were identified as TEC or PDC before ( Fig.2A, Suppl. Fig. 8 , Suppl. Table 3 ): In this focused analysis on renal parenchymal cells, ten cell subsets resembled cells from different segments of the renal tubule system with expression of specific marker genes (Fig. 2B, Suppl. Fig. 9A ). However, those All injury related subsets weakly expressed either proximal tubule (PT) markers (GPX3), thick ascending limb (TAL) markers SLC12A1/UMOD or collecting duct (CD) markers AQP2/FXYD4, without co-expression of these markers in single cells (Suppl. Fig. 10AB ), indicating a preserved injury reaction from all nephron segments, that is not limited to the much-investigated PT.

Next, we wanted to know whether any of these TEC phenotypes are specific for the damage introduced by AKI: We compared our data with recently published urinary scRNAseq data of chronic glomerular diseases DN (18) and FSGS (19) (Suppl. Fig. 6 ): TEC_emt and TEC_str were also the most prominently featured TEC in DN and FSGS (Suppl. Fig. 6B ). Interestingly, TEC_prg were detectable in DN and FSGS urine only to a much lesser extent than in AKI (Suppl. Fig. 6 .1+6.3), which may indicate a possible AKI-specific abundant excretion of TEC_prg. Taken together, the phenotype of urinary TEC seemed to be primarily shaped by injury-related dedifferentiation processes and only to a lesser extent by their tubule segment origin.

It is important to investigate to what extent urinary cells indeed reflect kidney pathophysiology. Therefore, we used single nuclei RNA seq (snRNAseq) data from human post-mortem AKI kidney biopsies generated previously(12) (Fig. 3B, Suppl. Fig. 11A ) and compared it to our urinary data. Briefly, data of human AKI kidney tissue contained 106,971 single nuclei with transcriptomes indicating tubular, leukocyte, endothelial and fibrocyte identities. In most nephron segments, Hinze et al. describe several injury-related altered tubular phenotypes, most notably inflammation, oxidative stress, and signs of EMT, here annotated as PT_New, TL_New, TAL_New, DCT_New, CNT_New, CD-PC_New and CD-IC_New (12) .

We compared the gene expression between urine and biopsy datasets. Correlations of expression profiles showed molecular similarity of corresponding urine and tissue cell types ( Fig. 3C ) and most urinary tubular subsets had a better expression correlation with injuryrelated "New" biopsy clusters than healthy counterparts.

Encouraged by these findings, we then integrated biopsy AKI data with urine data (Fig. 3AB ).

This data overlap of urine and tissue validated our prior findings with few urinary TEC clustering with healthy tissue TEC, while most injury related urinary subsets clustered separately, overlapping only partially with biopsy "New" clusters. Integration also revealed very low abundance of urinary fibrocytes and endothelial cells (Fig. 3A ) that had previously been clustered with TEC and leukocytes in the non-integrated urine data (Fig. 1 ).

bias toward distal and injured nephron segments (Fig. 3D) indicates that the urinary renal cell signature is influenced by injured TEC being excreted following loss of tubular integrity and/or detachment from the basal membrane, with viable TAL and CD cells being more prone to final urinary excretion.

Working with urine derived cells always brings forth two major skeptic questions: "Are those cells really derived from the kidney and not from the urogenital tract?" and "To what extent does prolonged contact to urine alter the phenotype of these cells?". We addressed these questions by comparison to two further references: In an integrated dataset with tumor adjacent healthy human kidney tissue (KPMP (38), Suppl. Fig. 13C ) and healthy bladder tissue Fig. 13D ), urinary TEC clustered with kidney derived TEC, while urinary UGEC clustered with bladder cells (Suppl. Fig. 13 ), providing further evidence for the renal origin of urinary TEC. We then compared TEC from all tissues and urinary datasets at hand for the occurrence of gene signatures associated with salt-, fluid shear-, osmotic-or pH-stress. Urine samples uniformly had a higher mean expression of the salt-stress signature genes, but all other stress related gene sets were not enriched in the urinary datasets, suggesting that the impact of exposure to urine was confined. (Suppl. Fig. 14) .

In conclusion, mostly injured and distal TEC were excreted with urine. Even though the urinary environment seemed to alter the transcriptional signature to a certain extent, information about tubular segment origin and pathophysiology was preserved in these cells.

The occurrence of multiple injury related TEC subsets with inflammatory, fibrotic and regenerative traits motivated us to ask whether our urinary data of human AKI are in accordance with recent findings in mouse AKI models describing "maladaptive" (6) or "failedrepair" (5, 9) TEC, which are suspected to drive fibrosis post AKI. We downloaded the mouse ischemia reperfusion injury dataset by Kirita et al (5) , who initially described these cell states, and reconstructed the investigated PT clusters showing the emergence of injured, repairing and "failed repair" out of healthy PT cells after ischemia-reperfusion injury (Fig. 4A ). The urinary TEC were compared with automated cell annotation (SingleR) using the mouse TEC as reference (Fig. 4B) . Unsurprisingly, most urinary TEC showed a phenotype most reminiscent of injured PT cells. However, especially in the progenitor-like clusters, a fraction of cells correlated most with repairing and "failed-repair"-like PT phenotypes (Fig. 4CD ). This indicated that TEC subsets with potentially beneficial or detrimental effects on the disease course may be detected via urine.

The subset TEC_prg remained the most elusive because no evidence for a tissue correlate was found, even though similar SOX4+SOX9+ cells occur in the reference kidney tissue samples after AKI (Suppl. Fig. 11B-D) . Several hints indicated that these cells might be a "snapshot" of (re-)differentiating tubular cells: These cells were reminiscent of repairing and partially "failed-repair" cell states in the mouse data. Furthermore, Phospholipase C gamma 2 (PLCG2) -together with proliferation markers FOS and JUN -were among the most highly expressed genes in these clusters (Fig. 5A ). Although PLCG2 is mostly associated with immune cells (40) , recent evidence for PLCG2-dependent signaling in epithelia (33, 41) and in in vitro proliferation of hepatocytes (42) suggest a role of PLCG2 in epithelial regeneration, possibly via interactors in the fibroblast growth factor (FGF)/ERK pathway (43) or WNT pathways (44) . Fittingly, activity for AKI regeneration associated pathways PI3K (45, 46) , Estrogen (47) and WNT (44) (inferred by PROGENy) was more pronounced in the TEC_prg clusters (Fig. 5C ). c) We noticed a divide between embryogenesis/developmental transcription factors PAX2, SOX4 and SOX9, which were predominantly expressed in the presumed progenitor clusters and other known surface stem cell markers like CD24, CD44 and PROM1, which were expressed most in the EMT clusters (Fig. 5B) . Appropriately, gene signatures in PAX2+ cells were similar to developmental terms found in the Gene Ontology (GO) database, many of them kidney specific like "metanephric tubule development" with additional stem cell terms like "somatic stem cell population maintenance", indicating a developmental stage (Fig.   4D ). PROM1+ cell signatures yielded similar GO terms like "nephron tubule development" but in contrast also "extracellular matrix organization" and "regulation of TGF-ß production", hinting at potential EMT (Suppl. Fig. 15 ). Distinct cellular crosstalk of TEC_prg with epithelial and immune cells in comparison to other TEC further emphasized these findings (Suppl. Fig. 16 ).

One of the major advantages of cellular urinalysis is its repeatability, allowing to evaluate temporal changes in the emergence of TEC after AKI onset more easily than biopsy-based approaches. Therefore, we compared absolute TEC counts in different phases (Fig. 6 ) after onset of AKI (early = up to five days after AKI onset, mid = day six to ten, late = after day ten).

Overall, the highest urinary TEC excretion seemed to happen from day six onward, with total TEC counts being significantly increased in patients sampled between day six and ten (p < 0.018, Unpaired Two-Sample Wilcoxon Test). This rise in TEC counts seems to be driven mainly by urinary shedding of TEC_emt and TEC_str after AKI, which both significantly increased in the mid phase compared to early and late AKI (p < 0.005 and p < 0.0059 respectively). On the contrary, amounts of excreted TEC_prg remained unchanged in the analyzed timeframe. TEC_cnn seemed to be a prominent urinary feature mostly in the early course of AKI (Fig. 6B) , although time-dependent differences did not reach statistical significance in this dataset. The interindividual variability in the AKI disease course is difficult to control for but these dynamic changes also observed in several repeatedly sampled patients ( Fig. 6C ) hint at a common underlying temporal shift in the urine cell signature after AKI.

This study positions urinary scRNAseq as a new, powerful tool to study human AKI. Kidney parenchymal cells including TEC and podocytes can reliably be detected in urine of AKI patients. These urine-excreted kidney cells are mainly of distal tubular origin and show diverse injury reactive phenotypes mirroring findings in transcriptional analyses of human AKI tissue samples and mouse ischemia reperfusion injury models. Furthermore, urine TEC quantity and composition change over time, enabling temporally resolved insights of the disease course.

Together with data from DN (18) The urinary cellular signal -independent from disease type or severity -is dominated by leukocytes and urogenital epithelia. (18, 19) . Therefore, most urinary cellular studies, singlecell (48) or otherwise (14, 16, 49) , have focused on analyses of urinary immune cells. Here, we demonstrate that by flow cytometric pre-selection and gathering of sufficient viable cell counts, the focus of urinary cellular analysis can be shifted towards renal parenchymal cells.

Urinary TEC transcriptomes correlate well with respective reference cells from tissue and mirror the damaged and injured TEC phenotypes in post-mortem AKI biopsy samples. This result is essential as it shows that the molecular makeup of TEC is sufficiently preserved in urine to infer data informative about molecular expression patterns in a patient's kidney without biopsies. Moreover, together with the data by Hinze et al. 12 , our results challenge the proximal tubular centricity of (murine) AKI research, with other nephron segments like TAL and CD-PC seemingly contributing to injury related cell pools as much as PT cells. Urine appears to be especially suited to study changes in the more distal nephron segments, as those TEC seem to be preferentially excreted into urine.

Our current understanding of TEC injury reaction assumes a dedifferentiation and redifferentiation after successful repair, which has been demonstrated for PT cells in several mouse models (5, 8, 9) and cells in human AKI biopsies have been reported to mirror this process (8) . A central question is how this repair takes place and why it can fail. The Humphreys group first described a "failed-repair" subset in mouse AKI (5), characterized by downregulation of differentiation markers of PT cells, expression of VCAM1 and persistence after AKI. These cells have since been reported in several other mouse models for AKI, indicating an inflammatory signature, which may fuel progression to CKD (50) as well as a potential recovery of these cells governed by ferroptotic stress (9) .

The most prominent urinary TEC subset in our data, TEC_emt, is also the most reminiscent of the "failed repair" state. Although mostly VCAM1-negative, (VCAM1 expression in our data is mostly confined to fractions of the descending thin limb cluster), this subset has proinflammatory and EMT phenotypes similar to the failed state and shows a dynamic excretion after injury, potentially indicating wound healing processes.

Though not unprecedented (21, 51) , urinary tubular progenitor-like cells (TEC_prg) and their high abundance after AKI are a particularly interesting finding. They show markers of specific TEC subsets, especially CD-PC and TAL, and expression of dedifferentiation markers SOX4, SOX9 and PAX2 while not expressing urogenital markers. Similar cells in urine of FSGS and DN patients can be detected, albeit to a much lesser extent. SOX4+SOX9+ TEC can also be found in post-mortal kidney samples with increased frequency in TAL and CD clusters after AKI compared to control, suggesting an injury-related increased excretion of these cells.

The origin of the regenerative progenitors (dedifferentiation (5) vs. fixed stem cells (52)) cannot be delineated from this dataset. RNA velocity analysis was tested to infer (de-)differentiation trajectories in the urine TEC and suggested, among other findings, a MALAT1 / PRMT1 dependent kinetic potentially linked to progenitor self-renewal as shown in embryonic nephron progenitors (53) . However, more research is needed to validate these findings, as robustness of velocity analysis in urine samples is still lacking.

Though speculative, based on the urine cell signature's bias toward distal nephron segments, those progenitor-like cells -if not dedifferentiated distal TEC -may also be a separate entity of distal/medullary/papillary tip origin, like the recently characterized distal multipotent Aqp2+ progenitor cells in mice (54) , and therefore harder to detect via biopsy.

Although urinary scRNAseq enables safe and easily repeatable examination of kidney cells in AKI, it is limited to patients with preserved diuresis, therefore excluding a substantial part of severe, oligo-to-anuric AKI cases from analysis. On a similar note, because healthy donors usually excrete almost no tubular cells, urinary data cannot be compared to a healthy urinary control group. While our findings are overlapping quite well with reference tissue data, it is not yet evident how exact urinary analysis can reflect the individual disease course, as gathering of urine and concurrent biopsy tissue from the same patient was not possible. Lastly, our method currently requires immediate processing of fresh urine samples, thus limiting its use for prospective clinical and research applications.

With the present dataset, we demonstrate feasibility of urinary scRNAseq in human AKI. The epithelial cells detected in the urine originate mainly from the distal nephron segments and reflect changes observed in kidney tissue of mice and men. With further standardization of sample collection and conservation, non-invasive and temporally resolved pathophysiological insights through urinary scRNAseq may facilitate kidney diagnostics and monitoring.

Between 2019 and 2021 we collected 40 urine samples of 32 patients with AKI as defined by Kidney Disease: Improving Global Outcomes (KDIGO) criteria (Suppl. Fig. 1 , Suppl. Table 1 ).

Patients were sampled at a variable timepoint within the first 21 days after AKI onset. Five patients were sampled at two separate timepoints and a single patient on four occasions. Samples were sorted using a Sony MA900 (Sony, Tokyo, Japan) with a 100µm nozzle into phosphate buffered saline (PBS) / 1 % BSA gating To-Pro3 -AnnexinV -Hoechst33342 + CalceinAM + singlets for obtaining single viable urinary cells (Suppl. Fig. 1 ).

Urinary granulocytes, which make up a large fraction of the urinary cellular signal (median 58% in FACS gating, Suppl. Fig. 1 ) but for which a renal tissue origin is not verifiable, and which may interfere with downstream analysis, were excluded. In the initial cohort of patients (n=7) only CD45cells were obtained, excluding most leukocytes, to maximize the amount of analyzed renal parenchymal cells; for the subsequent patients (n=25) CD45 -CD66band CD45 + CD66bwere sorted, including lymphocytes and monocytes/macrophages in the measurements (Suppl. Fig. 2 ). To adjust for low total cell counts in urine (median 3455 viable cells /100ml), samples P008-016 and P048-119 (Suppl. Fig. 4D ) were labelled with TotalSeq B Hashtag antibodies 1-4 (Biolegend, clone LNH-94) during the staining procedure described above and pooled in groups of four.

Relative single cell counts in sorted suspensions were counted using a MACSQuant Analyzer (Miltenyi). Cells were centrifuged, resuspended, and re-counted on a hemocytometer. Cell suspensions were subjected to single-cell sequencing following the 10x Genomics protocol for Chromium Next GEM Single Cell 3' v3.1 chemistry targeting between 1000 -10000 cells depending on sample cell count. In brief, cells were partitioned into a droplet with a barcoded gel bead using the 10x Chromium controller (10x Genomics, Pleasanton, California, USA) and lysed. RNA was reverse transcribed into complementary DNA (cDNA) within each droplet.

cDNA was amplified and fragmented, followed by the addition of Illumina adapters using Single Cell 3′ Library & Gel Bead Kit v3.1 (10x Genomics). Libraries were sequenced on Illumina HiSeq 4000 (Illumina, San Diego, California, USA) sequencers.

Demultiplexing, barcode processing, read alignment and gene expression quantification was carried out using Cell Ranger software (v3.1.0, 10x Genomics). First, Cell Ranger mkfastq demultiplexed the sequencing by sample index. The quality of the data was checked using FastQC (v0.11.5) and all samples showed high quality RNAseq data with good median perbase quality across most of the read length. Cell Ranger count used STAR software to align sequenced reads to the reference genome (GRCh38.p12). Samples from COVID-19 patients included SARS-CoV1/2 genomic information to reference genome.

Demultiplexing of unfiltered hashed cell sample pools was performed in Seurat v4(57) (Suppl. Fig. 3 ). Hashtag oligo (HTO) libraries were added to the Seurat object as an independent assay and normalized ("NormalizeData(SO, assay = "HTO", normalization.method = "CLR")").

Demultiplexing was performed with "HTODemux(SO, assay = "HTO")" with "positive.quantile =" between 0.90-0.99. Cross sample doublets were excluded. HTO negative cells were further analyzed as "POOL" samples (Suppl. Fig. 3+4 ).

Initial quality control was done by removing cells with less than 200 or more than 4,000 genes detected as well as cells with > 10% mitochondrial (mt) reads. For renal cells, a cut-off of 20% mt reads was applied to adjust for higher mitochondrial abundance in the kidney (58) . ScDblFinder (59) was used for removing multiplets. Seurat v4 was used for downstream analysis. We combined all AKI sample datasets. SCTransform (60) was used for normalization, scaling, and determining variable features. After principal component analysis ("RunPCA" function), integration and batch correction were done by Harmony (61) ("RunHarmony" function). Dimensional reduction by UMAP (62) (dims=1:45) and clustering was performed using reduction="harmony". Cluster resolution was determined manually. "FindAllMarkers()" was applied to determine differentially expressed genes (DEG) in all clusters. Cluster annotation and rearrangement was done based on review of specific lineage markers.

For further detailed analysis of kidney parenchymal cells in urine, cells of corresponding clusters were subsetted and re-analyzed using the same workflow as above. Cluster resolution and annotation was again performed manually based on DEG and known lineage markers.

Clustering revealed 25 clusters, two of which were comprised of few remaining leukocytes and urogenital cells, which were excluded from further analysis. were also downloaded from the NCBI/GEO repository. Data integration/analysis/dimensional reduction and clustering of these datasets was done exactly as with in urine AKI samples.

Mouse ischemia reperfusion injury PT data(5) (GSE139107) was downloaded from NCBI/GEO database. Ischemia reperfusion injury datasets were subsetted for PT cells based on the provided metadata, merged, normalized and scaled with base Seurat functions ("FindVariableFeatures()", "ScaleData()", "RunPCA()"). UMAP dimensional reduction and clustering was followed by differentially expressed gene discovery through "FindAllMarkers()" function and cluster annotation based on marker genes provided by the original paper. Human orthologous genes were generated using Biomart (63) .

For gene expression correlation, the top 3000 variable genes each per urine and reference datasets were used. Integration of urine and reference datasets was done by Harmony.

SingleR (64) was used for automatic unbiased cell type annotation in the urine TEC dataset using Human primary cell atlas bulk data (29) as well as aforementioned datasets as reference.

Briefly, SingleR computes Spearman's rank correlation coefficient for the gene expression of each cell with set reference groups. Wilcoxon ranked sum test is used to identify the top markers used for each comparison. A score is defined for each cell based on the reference correlation in comparison to the other cells. Scores are determined across all reference groups.

Cells are labeled according to their highest score.

We used CellPhoneDB (2.1.1) (65) to assess cellular crosstalk between different cell types.

As input data, we calculated normalized data with scale factor 10,000. To identify putative cellcell interactions via multi-subunit ligand-receptor complex pairs, label permutation was performed. Finally, we conducted statistical analyses by randomly permuting the cluster labels of each cell 10,000 times. Tests were conducted intraindividually.

GO term enrichment analysis was performed with topGO (66) using all DEG with an average log-fold change of > 0.5 in PAX2 and PROM1 positive cells ("runTest(GOdata, statistic = "fisher", algorithm = "elim")"). GO set visualization was done with rrvgo (67). Rrvgo groups the provided GO terms based on their similarity. GO group representatives were selected by rrvgo based on adjusted p value. Pathway activity across TEC subsets was calculated with PROGENy (68).

All RNA-sequencing data were analyzed using R (69) . Normalization, logarithmic transformation, identification of highly variable features, scaling, principal component analysis and extraction of differentially expressed features were done using Seurat. Significant differences between cell counts were tested using two-tailed Unpaired Two-Sample Wilcoxon Test.

The ethics committee of Charité University Hospital (Charité EA2/141/19) approved the study.

Informed consent was obtained from all patients or next of kin before participation. . Urinary tubule cells partly reflect "repairing" and "failed-repair" mouse ischemia model phenotypes A. UMAP and marker genes for proximal tubule (PT) dynamic cell states after ischemia reperfusion injury in a mouse model (5) . B. Heatmap of automatic annotation scores (SingleR) of urinary cells with mouse model cell types as reference. Most, but not all cells are labeled as "Injured". C. Urine TEC labeled as healthy, repairing or "failed-repair" by automatic annotation plotted in UMAP. Note how "repairing and "failed repair" labels are predominantly found in presumed progenitor cells (bottom middle clusters) and EMT clusters (top middle clusters) D. Relative abundance of healthy, repairing or "failedrepair" annotation labels in urinary TEC subsets (compare Fig. 2 ). The mean proportion of cell type across samples in each bin was calculated). C. Individual AKI course over time with serum creatinine, urinary output, single cell transcriptome yield and relative amount of urinary renal parenchymal cell subsets. AKI disease course is individual diverse but temporal changes of cell type proportions are similar. cnn -canonical, emt -epithelial to mesenchymal transition, stroxidative stress, prg -progenitor-like., prlf -proliferating.

Acute Kidney Injury Associates with Increased Long-Term Mortality

Intermediate and Long-term Outcomes of Survivors of Acute Kidney Injury Episodes: A Large Population-Based Cohort Study

Renal tubular epithelial cells: the neglected mediator of tubulointerstitial fibrosis after injury

How Tubular Epithelial Cell Injury Contributes to Renal Fibrosis

Cell profiling of mouse acute kidney injury reveals conserved cellular responses to injury

Single-Cell Profiling of AKI in a Murine Model Reveals Novel Transcriptional Signatures, Profibrotic Phenotype, and Epithelial-to-Stromal Crosstalk

Regrow or Repair: An Update on Potential Regenerative Therapies for the Kidney

Single cell profiling in COVID-19 associated acute kidney injury reveals patterns of tubule injury and repair in human. Pathology

Ferroptotic stress promotes the accumulation of pro-inflammatory proximal tubular cells in maladaptive renal repair

FOXM1 drives proximal tubule proliferation during repair from acute ischemic kidney injury

An atlas of healthy and injured cell states and niches in the human kidney

Transcriptomic responses of the human kidney to acute injury at single cell resolution

Kidney transplant monitoring by urinary flow cytometry: Biomarker combination of T cells, renal tubular epithelial cells, and podocalyxin-positive cells detects rejection

Mapping urinary chemokines in human lupus nephritis: Potentially redundant pathways recruit CD4+ and CD8+ T cells and macrophages

Deep Phenotyping of Urinary Leukocytes by Mass Cytometry Reveals a Leukocyte Signature for Early and Non-Invasive Prediction of Response to Treatment in Active Lupus Nephritis

Urinary CD8+ T-cell counts discriminate between active and inactive lupus nephritis

Monitoring tubular epithelial cell damage in AKI via urine flow cytometry2022

Urinary Single-Cell Profiling Captures the Cellular Diversity of the Kidney

Urine single cell RNA-sequencing in focal segmental glomerulosclerosis reveals inflammatory signatures

Single-cell RNA-Seq analysis identified kidney progenitor cells from human urine

Human Urine-Derived Renal Progenitors for Personalized Modeling of Genetic Kidney Disorders

A Comprehensive Molecular Portrait of Human Urine-derived Renal Progenitor Cells

Urine-Derived Stem Cells: Applications in Regenerative and Predictive Medicine

How to Find a Resident Kidney Macrophage: the Single-Cell Sequencing Solution

Proliferating SPP1/MERTK-expressing macrophages in idiopathic pulmonary fibrosis

Inhibition of 5-lipoxygenase decreases renal fibrosis and progression of chronic kidney disease

SARS-CoV-2 infects the human kidney and drives fibrosis in kidney organoids

COVID-19 tissue atlases reveal SARS-CoV-2 pathology and cellular targets

An expression atlas of human primary cells: inference of gene function from coexpression networks

The transcription factor Sry-related HMG box-4 (SOX4) is required for normal renal development in vivo

Cited1 and Cited2 are differentially expressed in the developing kidney but are not required for nephrogenesis

Hes1: a key role in stemness, metastasis and multidrug resistance

Signatures of plasticity, metastasis, and immunosuppression in an atlas of human small cell lung cancer

Inactivation of the transcription factor Elf3 in mice results in dysmorphogenesis and altered differentiation of intestinal epithelium

Elf3 plays a role in regulating bronchiolar epithelial repair kinetics following Clara cell-specific injury

The RNA helicases DDX5 and DDX17 facilitate neural differentiation of human pluripotent stem cells NTERA2

Identification of immediate early response protein 2 as a regulator of angiogenesis through the modulation of endothelial cell motility and adhesion

Rationale and design of the Kidney Precision Medicine Project

Single-Cell Transcriptomic Map of the Human and Mouse Bladders

Are BTK and PLCG2 mutations necessary and sufficient for ibrutinib resistance in chronic lymphocytic leukemia

Cells of the human intestinal tract mapped across space and time

PLCG2 promotes hepatocyte proliferation in vitro via NF-κB and ERK pathway by targeting bcl2, myc and ccnd1

FGF/FGFR2 Protects against Tubular Cell Death and Acute Kidney Injury Involving Erk1/2 Signaling Activation

Role of CD133 Molecule in Wnt Response and Renal Repair

PI3K/AKT activation attenuates acute kidney injury following liver transplantation by inducing FoxO3a nuclear export and deacetylation

Tempol Protects Against Acute Renal Injury by Regulating PI3K/Akt/mTOR and GSK3β Signaling Cascades and Afferent Arteriolar Activity

A single dose of estrogen during hemorrhagic shock protects against Kidney Injury whereas estrogen restoration in ovariectomized mice is ineffective

The immune cell landscape in kidneys of patients with lupus nephritis

The cellular signature of urinary immune cells in Lupus nephritis: new insights into potential biomarkers

Single-nuclear transcriptomics reveals diversity of proximal tubule cell states in a dynamic response to acute kidney injury

Human Urine as a Noninvasive Source of Kidney Cells

Endocycle-related tubular cell hypertrophy and progenitor proliferation recover renal function after acute kidney injury

Single cell RNA sequencing uncovers cellular developmental sequences and novel potential intercellular communications in embryonic kidney

Generation of Distal Renal Segments Involves a Unique Population of Aqp2+ Progenitor Cells

Unbiased transcriptomes of resting human CD4+CD45RO+ T lymphocytes

Integrated analysis of multimodal single-cell data

A Protocol for the Parallel Isolation of Intact Mitochondria from Rat Liver, Kidney, Heart, and Brain

Doublet identification in single-cell sequencing data using scDblFinder2021

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

Fast, sensitive, and accurate integration of single cell data with Harmony

UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. ArXiv180203426 Cs Stat

Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt

Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage

CellPhoneDB: inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes

Enrichment Analysis for Gene Ontology

Perturbation-response genes reveal signaling footprints in cancer gene expression

The R Project for Statistical

FGF/FGFR signaling in health and disease

Activation of β-adrenergic receptors is required for elevated α1A-adrenoreceptors expression and signaling in mesenchymal stromal cells

Prognostic role of the beta-2 adrenergic receptor in clear cell renal cell carcinoma

Expression of CD44 in kidney after acute ischemic injury in rats

Upregulated renal tubular CD44, hyaluronan, and osteopontin in kdkd mice with interstitial nephritis

Receptor-ligand interaction between CD44 and osteopontin (Eta-1)

Diverse Role of TGF-β in Kidney Disease

Induction of renal fibrotic genes by TGF-β1 requires EGFR activation, p53 and reactive oxygen species

The epidermal growth factor receptor pathway in chronic kidney diseases

Sema4C-Plexin B2 signalling modulates ureteric branching in developing kidney

This work was supported by grants from the Berlin Institute of Health (BIH), the Jackstädt-