Identifying Common Molecular Mechanisms in Experimental and Human Acute Kidney Injury

Summary

Acute kidney injury (AKI) is a highly prevalent, heterogeneous syndrome, associated with increased short- and long-term mortality. A multitude of different factors cause AKI including ischemia, sepsis, nephrotoxic drugs, and urinary tract obstruction. Upon injury, the kidney initiates an intrinsic repair program that can result in adaptive repair with regeneration of damaged nephrons and functional recovery of epithelial activity, or maladaptive repair and persistence of damaged epithelial cells with a characteristic proinflammatory, profibrotic molecular signature. Maladaptive repair is linked to disease progression from AKI to chronic kidney disease. Despite extensive efforts, no therapeutic strategies provide consistent benefit to AKI patients. Since kidney biopsies are rarely performed in the acute injury phase in humans, most of our understanding of AKI pathophysiology is derived from preclinical AKI models. This raises the question of how well experimental models of AKI reflect the molecular and cellular mechanisms underlying human AKI? Here, we provide a brief overview of available AKI models, discuss their strengths and limitations, and consider important aspects of the AKI response in mice and humans, with a particular focus on the role of proximal tubule cells in adaptive and maladaptive repair.

Acute kidney injury (AKI) impacts more than 13 million people per year, resulting in a markedly increased risk of in-hospital and post-discharge mortality, but no therapeutic strategies to prevent or treat AKI have been identified to date.¹ AKI is currently defined according to the 2012 Kidney Disease: Improving Global Outcomes guidelines as an acute increase in serum creatinine level and/or decrease in urine output, and patients are stratified into three categories reflecting increasing severity of AKI (Fig. 1).² However, this definition only captures AKI events that result in a substantial decrease of excretory renal function, because the serum creatinine level may not increase until 50% or more of functional nephrons are lost.^3–5 Therefore, an alternative definition of AKI that includes the category subclinical AKI, which describes tubular injury indicated by increased renal biomarker levels without changes in serum creatinine level or urine output, has been proposed.^6,7

Figure 1.

Definition and staging of acute kidney injury (AKI) according to 2012 Kidney Disease: Improving Global Outcomes guidelines. Creatinine criteria indicate serum creatinine values. Abbreviations: RRT, renal replacement therapy.

Many different factors, such as sepsis, urinary tract obstruction, cardiac surgery–related ischemia, and nephrotoxic drugs, can cause AKI.⁸ At the cellular level, an acute renal insult triggers cell death through apoptosis and necrosis, dedifferentiation and proliferation of proximal tubule cells (PTCs), inflammation, and fibrosis. Globally, these processes can lead to an adaptive repair after the acute injury in which kidney function is restored through effective replicative repair of PTCs, or a maladaptive scenario with persistent proinflammatory PTCs, fibrosis, and disease progression to chronic kidney disease (CKD). Although there are similar features, the pathophysiologic mechanisms underlying different forms of AKI might not be uniform: bulk RNA sequencing studies of kidneys exposed to transient renal ischemia or severe depletion of extracellular fluid volume highlighted major transcriptional differences between each AKI model despite a similar increase in serum creatinine levels.⁹ The often multifactorial genesis of AKI in humans and the tight interconnection between the kidney and other organ systems, manifested, for example, in cardiorenal or hepatorenal syndromes, add additional layers of complexity, particularly in an aging and increasingly multimorbid patient population. None of these aspects are captured by the current Kidney Disease: Improving Global Outcomes definition of AKI. Focused efforts, such as the Kidney Precision Medicine Project, aim to define molecular subtypes of human AKI and CKD using various transcriptomics, proteomics, metabolomics, and imaging approaches, and to identify novel AKI biomarkers that may contribute to a more granular categorization of AKI.^5,10

Modeling a heterogeneous syndrome such as AKI in the laboratory is challenging and failures in moving laboratory model studies to translational AKI trials may result at least in part from inadequate modeling of AKI in preclinical studies.^11,12 This poses the question of whether current experimental models of AKI reflect the molecular and cellular mechanisms underlying human AKI? Given the complexity of the AKI syndrome discussed earlier and a lack of systematic comparative studies, this question cannot be answered comprehensively at present, yet the answer is critical for effective target discovery and drug development. This review attempts to provide some insight, presenting an overview of available AKI models, discussing their strengths and limitations (Fig. 2), and considering selected aspects of the AKI response in mice and humans. In addition, we highlight recent single-cell sequencing studies of experimental and human AKI that open new possibilities to identify conserved and divergent mechanisms across species.

Figure 2.

Model systems of acute kidney injury (AKI). Abbreviations: CABG, coronary artery bypass graft; iPSC, induced pluripotent stem cell. Elements of Figure 2 were created with BioRender.com.

EXPERIMENTAL MODELS OF AKI

Zebrafish

AKI was first modeled in goldfish in 1990 through the intraperitoneal injection of hexachlorobutadiene.¹³ Subsequently, AKI has been explored in several other fish species using gentamicin treatment to induce nephrotoxic damage, bacterial infection to model sepsis, and laser ablation to induce targeted cell death.^14–16 Given practical and experimental considerations such as husbandry, size, fecundity, life span, and, importantly, genetics, the zebrafish is now the most widely used fish in AKI research.

Zebrafish larvae rely on paired pronephric tubules, while adult zebrafish have a more extensive mesonephric kidney comprising several hundred nephrons.¹⁶ The segmentation of zebrafish nephrons is similar to human nephrons except for an absence of the thin limb of the loop of Henle and a connecting segment.¹⁶ Pronephric nephrons are functional within the first 2 days of development; consequently, studies of the pronephric kidney are particularly well suited for high-throughput, high-content genetic screens in which the optical clarity of the larva greatly facilitates observational and experimental analysis of the pronephric kidney.¹⁷

Nephrotoxic AKI in zebrafish larvae is characterized by a loss of apical–basal polarity in renal epithelia, and flattening, dedifferentiation, and apoptosis of tubular cells—features also observed in mammalian AKI.¹⁸ Loss of pronephric kidney activity results in pericardial, or more generalized, edema.¹⁸ In addition, up-regulation of the injury marker kidney injury molecule-1 (encoded by Havcr1 and first described in a rat model of AKI¹⁹) has been observed in adult zebrafish treated with gentamicin.²⁰ In striking contrast to mammals, AKI induces neonephrogenesis by resident nephron stem cells in the mesonephric kidney of the adult zebrafish, resulting in effective nephron replacement without extensive development of kidney fibrosis.^14,21 Although the zebrafish injury–repair response limits cross-species comparability given marked differences with human AKI responses, the zebrafish provides an aspirational model to identify regenerative principles, and compounds that may promote, favor, or enhance a regenerative process.¹⁴ As an example, small-molecule screens performed on zebrafish larvae identified a histone deacetylase inhibitor, 4-(phenylthio)butanoic acid (PTBA), that expands the renal progenitor pool.²² Surprisingly, PTBA showed beneficial effects on AKI outcomes not only in zebrafish larvae, but also in mice despite very different reparative processes.¹⁸

Rodents

Mice and rats have been used extensively in AKI research. The mouse kidney comprises approximately 14,000 nephrons, while the human kidney contains approximately 1 million, with considerable variability in the nephron number in the human population.²³ Structurally, the unilobar rodent kidney resembles a single lobe of the multilobar human kidney in most parts, although the mouse kidney only has cortical and juxtamedullary nephrons, while the human kidney additionally contains midcortical nephrons, resulting in a predominance of long loops of Henle in the mouse and short loops of Henle in humans.²³ In addition, the loops of Henle in mice intermingle with large bundles of vasa recta, whereas they are more separated from the vascular bundles in the human kidney.²³ The anatomic differences correlate with an enhanced urine-concentrating capability by the mouse kidney.²³

AKI is induced and modeled in rodents through several procedures: temporary clamping of the renal arteries resulting in an ischemia-reperfusion injury (IRI), invoking sepsis through cecal ligation and puncture or LPS treatment, and surgical ligation of the ureter resulting in an obstructive injury. Nephrotoxicity can be modeled by drug administration (gentamicin, cisplatin, folic, or aristolochic acid), rhabdomyolysis by intramuscular glycerol injection, and cardiorenal syndromes can be surgically induced, for example, by coronary artery ligation (see Hukriede et al¹⁴ and Heyman et al²⁴ for in-depth reviews). A clear advantage of rodent models, particularly mice, are the advanced genetic tools that can aid directly in the modeling and the analysis of AKI responses, including lineage tracing and fate mapping of cell types and constitutive and conditional modification of global or tissue-specific gene activity. Genomic studies are complemented by proteomic strategies such as enzyme-catalyzed proximity labeling to assess interorgan communication or cell type–specific translational profiling using translating ribosome affinity purification.^25,26 In addition, rodent models enable the investigation of AKI mechanisms in the context of common comorbidities such as diabetes or pre-existing CKD.^14,27,28 AKI outcomes in rodents can vary significantly between different strains and depending on the exact conditions under which AKI is induced. Modeling AKI in rodents therefore requires careful titration of the assay, control over variable conditions, and extensive replication to ensure reproducibility.

Large Animal Models

The anatomy of the kidney in large animal models (LAMs), such as pigs, is characterized by a multilobar structure, more closely resembling human than rodent kidney anatomy.²⁹ In addition to AKI models described earlier, LAMs enable investigation of AKI causes such as coronary artery bypass graft surgery, hemorrhage, or burns, which are more difficult to assess in rodent or other small animal models. An important advantage of AKI modeling in LAMs is the much greater similarity of the immune system with the human immune system.³⁰ For example, neutrophils comprise 10% to 25% of leukocytes in the peripheral blood in mice, but 50% to 70% in humans and pigs, and some chemokines that are up-regulated in response to AKI in humans are absent in mice, but present in pigs.^30,31 However, surgical procedures can be challenging, LAMs are costly and fewer genetic and analytical (eg, antibodies) tools are available for LAMs. As a result, LAMs have not been widely used in exploratory research and much less is known about the molecular mechanisms underlying AKI responses in LAMs than in rodents, but LAMs play an important role in translational studies.³⁰

Human Kidney Organoids

Over recent years human kidney organoids, derived from human pluripotent stem cells in vitro, have emerged as a tool to model AKI in the laboratory. Cisplatin (single high-dose or repetitive low-dose treatment) and gentamicin have been used to model nephrotoxic AKI, and hemin has been used to model hemolysis-induced AKI in kidney organoids.^32–39 In response to hemin treatment, up-regulation of HAVCR1, mitochondrial dysfunction, apoptosis, and reduced expression and function of organic anion and cation transporters have been reported.³² Similarly, treatment with nephrotoxic drugs resulted in up-regulation of the injury markers HAVCR1 and NGAL, DNA damage, apoptosis, and increased proliferation.^33–38 Figure 3 shows the presence of HAVCR1 (kidney injury molecule-1)+ injured PTCs in human kidney organoids treated with cisplatin or gentamicin.³⁵ A recent study suggested that organoids treated with a single low dose of cisplatin can be used to model intrinsic repair characterized by a normal tubular morphology and a reduction of DNA damage and cell-cycle activity after cisplatin treatment, while repeated low-dose cisplatin treatment can be used to model incomplete kidney repair defined by the development of fibrosis and tubular atrophy.³⁹ Human kidney organoids also have been used to show that severe acute respiratory syndrome coronavirus 2 (SARS-Cov-2) can infect renal epithelial cells and trigger kidney injury and fibrosis.⁴⁰ In addition, the observation that hemin,³² cisplatin,³⁹ or interleukin 1β⁴¹ treatment result in enhanced fibrosis in human kidney organoids suggest that organoid models may provide valuable insight into mechanisms of kidney fibrosis, a common feature in most kidney diseases.

Figure 3.

Human kidney organoids to model acute kidney injury. (A) Immunofluorescence staining of human kidney organoids treated with gentamicin (5 mg/mL) from days 21 to 23 or cisplatin (5 μmol/L) from days 21 to 22. Scale bars: 50 μm. The injury marker kidney injury molecule-1 (KIM-1) is detected in gentamicin- or cisplatin-treated organoids, but not in controls. LTL marks proximal tubules, CDH1 marks the loop of Henle and distal tubules. CDH1 levels are reduced in cisplatin-treated organoids compared with controls. (B). Low-magnification image of gentamicin-treated organoids. Scale bar: 100 μm. Abbreviations: CDH1, cadherin-1; DAPI, 4’,6-diamidino-2-phenylindole, dihydrochloride; HAVCR1, hepatitis A virus cellular receptor 1; LTL, lotus tetragonolobus lectin. Reproduced with permission from Morizane et al.³⁵

An understanding of the current limitations of kidney organoid models is important when considering their use in modeling AKI. These include the following: (1) the immature state of kidney organoids, which is comparable with trimester 1 or 2 human fetal kidneys⁴²; (2) the presence of inappropriate (nonkidney) cell types; (3) lack of flow, perfusion, and filtration; and (4) an absence of collecting duct and immune cells, as well as a paucity or poor organization of vascular endothelium in most kidney organoids. Large initiatives, such as the National Institute of Diabetes and Digestive and Kidney Diseases–funded (Re-)Building a Kidney consortium, are working to improve kidney organoid development and function, and disease and injury modeling.⁴³ The feasibility of medium-scale, high-content drug screens using kidney organoids, has recently been demonstrated using a kidney organoid differentiation protocol that can readily generate tens of thousands of miniature organoids (comprising 1–2 nephrons) modeling polycystic kidney disease.⁴⁴ Further improvement of AKI kidney organoid models holds much promise for enhancing efforts to discover AKI therapeutics and to prescreen drugs under clinical development for potential trial-ending renal toxicity.^45,46 Unanticipated organ toxicities account for failure in approximately 50% of phase II clinical trials.⁴⁷

CELLULAR AND MOLECULAR MECHANISMS IN EXPERIMENTAL AND HUMAN AKI

Most of our knowledge of the pathophysiology underlying mammalian AKI is derived from rodent models, because kidney biopsies are rarely performed in the acute phase of AKI in humans. In the following, we discuss selected aspects of the AKI response, highlighting common mechanisms between mice and humans. We focus on IRI as a highly prevalent cause of AKI, and its effect on PTCs, since PTCs are the most abundant cell type in the kidney and play a central role in both adaptive and maladaptive renal repair. Readers can benefit from several excellent recent reviews for a discussion of other aspects of the AKI response, such as hypoxia, autophagy and mitochondrial quality-control mechanisms, innate and adaptive immunity, and inflammation.^48–51

Transcriptional Trajectories From AKI to CKD

Bulk RNA sequencing of mouse kidney tissue at 10 different time points between 2 hours and 12 months after IRI showed that the molecular response to AKI in mice follows a distinct temporal pattern, initially dominated by genes related to regulation of transcription and cell death, followed by activation of the cell cycle in adaptive repair, and inflammation, immune, and fibrotic responses related to repair, and ongoing maladaptive responses in a transition to chronic disease.⁵² Similarly, bulk RNA sequencing of human kidney transplant protocol biopsy specimens taken before transplantation, after reperfusion, and at 3- and 12-months post-transplantation showed specific transcriptional changes early and late after human IRI induced by transplantation.⁵³ Early after injury, up-regulation of immediate-early response genes, including multiple transcription factors, was observed. Comparison with gene expression changes in mice 2 hours after IRI highlighted a strong conservation of the immediate-early response across species.^52,53 The human transplant biopsy data showed two divergent transcriptional directories months after transplantation, one associated with recovery and the other associated with chronic injury.⁵³ Gene expression changes observed in chronically injured human kidneys were partially conserved when compared with mouse kidney tissue 12 months post-IRI. For example, expression of genes encoding various collagens (COL1A1, COL1A2, COL3A1), cytokines (CCL2, CCL19), and vascular cell adhesion molecule 1 (VCAM1) were up-regulated across species.^52,53

Tubular Regeneration

Proliferation is a central factor in the adaptive repair process to replenish tubular cells lost through cell death, and has been observed in all experimental AKI models and in human AKI. Discussion continues on alternative views of tubular regeneration mediated by the proliferation of a pre-existing PTC progenitor population and/or dedifferentiated PTCs.^42,54,55 In addition, it remains unclear if the origin(s) of proliferating PTCs are conserved across species. A renal epithelial cell population demarcated by the cell markers CD133 and/or CD24, further subdivided into VCAM1+ and VCAM1− populations, has been identified in Bowman’s capsule, or scattered throughout the proximal tubule.^56–60 Renal CD133+ CD24+ cells were shown to have limited self-renewal capacity and could differentiate into multiple cell types, including renal epithelial cells in vitro.^56,57 In mice, a Pax2+ PTC population was suggested to represent tubular progenitors and lineage-tracing studies indicated that these Pax2+ PTCs regenerate injured proximal tubules, while other PTCs were reported to hypertrophy through endocycling in response to AKI.⁶¹ To our knowledge, Pax2+ tubular progenitor cells have not been described in single-cell or single-nuclear RNA sequencing (scRNA-seq or snRNA-seq) data sets of healthy mouse kidneys, but an up-regulation of Pax2 was observed in injured PTCs.^62,63 Another lineage tracing study reported clonal expansion of tubule cells restricted to each tubule compartment under homeostasis and in response to injury, indicating that tubule repair is locally regulated.⁶⁴

A different view of PTC repair to dedicated progenitors comes from lineage-tracing studies in mice using either Six2^GC or Slc34a1^GCE mice to label PTCs before injury and the administration of different thymidine analogs to label proliferating cells and Havcr1^GCE mice, which label injured PTCs. These data showed the following: (1) proliferating PTCs are derived from differentiated PTCs that subsequently up-regulate Cd133, Cd24, and vimentin after AKI; (2) the proliferative response to AKI is initiated stochastically by mature PTC types; and (3) the cells that proliferate after AKI are injured PTCs.^65–68 Taken together, these results support the concept that all PTCs have the potential to dedifferentiate and proliferate in response to AKI, which is corroborated by evidence from an additional independent lineage tracing study in mice.⁶⁹

Metabolic Changes

PTCs are mitochondria-rich and highly metabolically active. Septic and ischemic AKI lead to a loss of mitochondria and structural alterations in remaining mitochondria in both model organisms and humans.⁷⁰ Fatty acid oxidation in mitochondria and peroxisomes is the primary energy source for PTCs. Down-regulation of fatty acid oxidation enzymes was observed in experimental and human AKI and CKD, and linked to fibrosis development, whereas genes encoding glycolysis enzymes were up-regulated in response to AKI.^71–75 In addition, the polyol pathway, which leads to endogenous generation of fructose from glucose, was shown to be activated in injured mouse kidneys and urine fructose levels were increased in AKI patients, indicating potential conservation of this AKI response across species.⁷⁶ Interestingly, knock out or pharmacologic inhibition of fructokinase, a fructose-metabolizing enzyme, ameliorated outcomes after IRI in mice, suggesting a detrimental role of endogenous fructose generation in the AKI response.⁷⁶ Similar effects of endogenous fructose generation were described in diabetic nephropathy in mice.⁷⁷

Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1a), an important transcriptional co-activator of fatty acid oxidation and mitochondrial biogenesis, is down-regulated in response to AKI in mice and humans.^78,79 Using elegant overexpression and knock-out studies in mice, PGC-1α was shown to confer renoprotective effects by increasing de novo nicotinamide adenine dinucleotide (NAD) biosynthesis from its progenitor nicotinamide, whereas PGC-1α knock out reduced de novo NAD biosynthesis.⁷⁹ Analysis of human transplant biopsy specimens early and late after transplantation confirmed impairment of de novo NAD biosynthesis in human AKI and CKD.⁸⁰ Nicotinamide treatment in mice improved AKI outcomes and, encouragingly, a small phase I clinical trial in patients undergoing cardiac surgery showed that oral nicotinamide increased nicotinamide levels in blood and urine and was associated with fewer AKI events in a secondary analysis.^79,81

PTCs are also an important site of gluconeogenesis.⁸² A recent study showed that AKI in mice led to down-regulation of genes relevant for gluconeogenesis in PTCs, and AKI in mice and humans resulted in a reduction of glucose production and lactate clearance, which was associated with mortality.⁷⁵ In contrast, thiamine treatment increased gluconeogenesis in PTCs in vitro.⁷⁵ In a retrospective analysis of critically ill patients with AKI, thiamine supplementation was associated with a faster decrease in impaired metabolism (defined as high blood lactate and low glucose levels) and reduced mortality compared with matched controls, suggesting thiamine supplementation as a potential therapeutic strategy.⁷⁵

Maladaptive Repair and Tubular Senescence

AKI can progress to CKD in mice and humans.^52,83,84 Correlating with this, maladaptive injured PTCs have been detected weeks to months after a single AKI event in mice.^62,63,85 Figure 4 shows the presence of injured PTCs, identified by the injury markers Havcr1 and Vcam1, in a mouse kidney 4 weeks after IRI. Repeated selective diphtheria toxin–induced injury of renal epithelial cells from the nephron lineage resulted in the development of fibrosis, pointing to a causal role of PTC injury and maladaptive tubular repair in the AKI-to-CKD transition.⁸⁶

Figure 4.

Maladaptive proximal tubule cells persist weeks to months after ischemia-reperfusion injury (IRI). Immunofluorescence staining of mouse kidneys 4 weeks after sham surgery or acute kidney injury induced by 18 minutes of bilateral IRI surgery. LTL marks proximal tubules, Vcam1 and Havcr1 mark injured proximal tubule cells, and Hoechst marks nuclei. The white boxes in the low-magnification images indicate the region shown in the high-magnification images in the lower part of the figure. Scale bar: 1,000 μm (low-magnification images), 20 μm (high-magnification images). Abbreviations: Havcr1, hepatitis A virus cellular recepto; LTL, lotus tetragonolobus lectin;; Vcam1, vascular cell adhesion molecule 1.

Chronic cellular senescence, a cell state defined by cell-cycle arrest and a proinflammatory secretome, the senescence-associated secretory phenotype (SASP), has emerged as a conserved feature of aging and of maladaptive injury repair across species and organ systems.⁸⁷ Several lines of evidence point to a pathogenic role of senescent PTCs in maladaptive renal repair in mice and humans: (1) senescent PTCs are present in diseased and aged human and mouse kidneys and the senescence load correlates with loss of kidney function; (2) levels of the senescence marker cyclin-dependent kinase inhibitor 2A (CDKN2A) in human kidney transplants are predictive of serum creatinine values 1 year after transplantation; and (3) PTCs with features of senescence (cell-cycle arrest, profibrotic phenotype) are implicated in fibrosis development after AKI in different mouse models.^84,87–91 Transgenic approaches modifying senescence-induction pathways in mice have yielded conflicting results depending on the AKI model and transgenic approach used, while transgenic depletion of senescent cells was shown to promote renal repair.^48,87 A recent study using the senolytic drug ABT-263, a B-cell lymphoma (Bcl) 2/w/xL inhibitor, to remove senescent cells in aged or irradiated murine kidneys before kidney injury was induced by IRI, showed a reduced number of senescent cells and increased kidney regeneration compared with controls.⁹⁰ A clinical trial evaluating senolytic treatment in patients with diabetic nephropathy is currently recruiting (NCT02848131), and a phase I pilot study of this trial reported that senolytic treatment reduced senescent cell burden in human adipose tissue and skin and decreased the levels of circulating SASP factors.⁹²

Single-Cell Sequencing Approaches Characterize Injured Proximal Tubule Cells in Mice and Humans

The advent of single-cell sequencing approaches has revolutionized our understanding of different cell types within complex tissues such as the kidney and allowed the characterization of injured PTC states with unprecedented resolution.^93,94 Two independent snRNA-seq studies described diverse and temporally dynamic injured PTC states after IRI in the mouse and identified a population of failed-repair PTCs (FR-PTCs) characterized by a proinflammatory, profibrotic, transcriptional profile including expression of Ccl2 and Vcam1 and present in the kidney several weeks after IRI.^62,63 Transcriptional regulators critical to normal PTC differentiation and function, such as hepatocyte nuclear factor 4-alpha (Hnf4a), were down-regulated in FR-PTCs, while members of the activator protein-1 and nuclear factor-κB families were activated.⁶² Deconvolution analysis of bulk RNA sequencing data indicated an increase in FR-PTCs in the aging rat kidney and human kidney transplants 1 year after transplantation.⁶³ Ferroptotic stress was shown to contribute to the accumulation of proinflammatory injured PTCs after IRI in the mouse and pharmacologic inhibition of ferroptosis reduced the number of maladaptive PTCs in the kidney and resulted in improved renal repair and reduced fibrosis.^85,95 In keeping with a potential role of the cell death pathways ferroptosis and pyroptosis in maladaptive repair/fibrosis development in humans, genes related to both types of cell death were correlated with fibrosis in human kidney biopsy specimens.⁹⁵

A single-cell sequencing atlas including scRNA-seq and snRNA-seq data of human AKI, CKD, and control samples identified 53 human kidney cell types providing much-needed molecular granularity to cellular diversity in the normal and diseased human kidney.⁹⁶ The study described multiple altered cell states for cell types of the nephron, stroma, and vasculature, some mostly present in diseased kidneys, others derived from both healthy and diseased kidneys. A subpopulation of altered PTCs shared features with the FR-PTC type identified in murine AKI, while also expressing CD133, a marker suggested to distinguish putative renal progenitor cells in human kidneys,⁵⁷ although here the linkage is to cells with a senescent phenotype. Interestingly, a similar altered, proinflammatory cell state was identified in the thick ascending limb of the loop of Henle.

A recent scRNA-seq study of human AKI samples from critically ill patients with and without COVID-19 similarly identified injury-induced cell states among different cell types along the nephron including a VCAM1 + injured PTC population.⁹⁷ Cross-species comparison highlighted strong similarities between this injured PTC population and FR-PTCs in the mouse.⁹⁷ The AKI response did not differ substantially between patients with and without COVID-19.⁹⁷ Another snRNA-seq data set of human and murine AKI ordered PTCs along trajectories from differentiated to injured/dedifferentiated PTCs, highlighting dynamic transcriptional profiles of PTCs responding to AKI, with significant conservation in this response across species.⁷³ scRNA-seq of urine from AKI patients showed several injured tubule clusters derived from the proximal tubule, the thick ascending limb of the loop of Henle, or the collecting duct, with similarities to injured tubule cell types identified by sc/snRNA-seq of mouse and human AKI kidneys, although urine tubule epithelial cells showed some enrichment for genes related to salt stress.⁹⁸ These data suggest urine scRNA-seq analysis as a useful additional tool to study human AKI pathophysiology, but future studies are needed to assess how well the urine cell composition reflects the kidney cell composition because simultaneous scRNA-seq analysis of urine and kidney tissue from the same patient were not performed in this study.

CONCLUSIONS

Animal models have greatly enhanced our understanding of the molecular mechanisms underlying AKI and many features of both adaptive and maladaptive renal repair are conserved between mice and humans. However, none of the available model systems fully recapitulates the astounding complexity of human AKI, and failures of translational AKI studies highlight the need for a better understanding of human AKI pathophysiology on the one hand, and of the strengths and limitations of experimental AKI models on the other hand, to inform and improve preclinical study design. Recent sc/snRNA-seq studies of human AKI showed several novel injury-associated cell states and opened the possibility for future systematic cross-species analyses to identify conserved and divergent features of the AKI response.^73,96,97 Moreover, human kidney organoids represent a promising new addition to the AKI research toolbox, although factors such as the incomplete differentiation of epithelial cells and the absence of immune cells in organoids are currently limiting. Improving and standardizing human kidney organoid protocols is therefore a priority for the coming years. Once refined, human kidney organoids will provide a versatile tool to unravel human AKI mechanisms in vitro and perform large-scale drug and nephrotoxicity screens at low cost.