Dissecting Normal and Abnormal Human Kidney Development Using Multiomics

Abstract

Congenital kidney anomalies frequently lead to development of CKD in children and adults, with limited possibility for accurate prognostication and successful intervention. Causal genetic variants are identified in a minority of individuals, while the biologic interpretation of putative genetic variants and their effect on kidney development and CKD remains incompletely understood. Advances in single-cell and spatial multiomics now enable a deeper biologic understanding and interpretation of disease-causing mechanisms of congenital kidney anomalies, holding promise for precise diagnoses, prognostication, and treatment for patients. In this review, we provide an overview of multiomics approaches, including transcriptomics, epigenomics, proteomics, and metabolomics, for characterizing and understanding the biology of human kidney development and disease. We will discuss the technical capabilities and challenges in mapping the spatial distribution of normal and abnormal developmental processes in the kidney. Moreover, we present three key multiomics case studies and discuss their experimental design considerations. Finally, future directions and the potential effect of multiomics approaches on the biologic understanding of kidney disease in development and adulthood are discussed. This review highlights that by integrating molecular insights, multiomics has the potential to transform our understanding of genetic (or variant-driven) and nongenetic kidney disease mechanisms and to advance precision diagnostics, prognostics, and therapies for congenital kidney anomalies.

Introduction

Congenital kidney anomalies represent a heterogeneous group of clinical phenotypes that arise from aberrant embryonic kidney development, primarily driven by genetic alterations and environmental exposures.¹ They account for more than half of all pediatric CKD cases and place a substantial burden on affected children and their families, with only limited therapeutic options available.² Most cases are detected by routine prenatal ultrasound, which enables early diagnosis but rarely allows successful prognostication and intervention.^3,4 A deeper understanding of the molecular mechanisms underlying normal and abnormal kidney development is therefore critical for advancing precision medicine in congenital kidney anomalies.⁵

Kidney embryogenesis depends on tightly coordinated spatiotemporal interactions between the metanephric mesenchyme, derived from the nephrogenic cord, and the ureteric bud, branching from the nephric duct (for detailed information, see refs. 6–14). Disruptions in this process can result in agenesis, dysplasia, hypoplasia, or multicystic dysplastic kidneys, which may manifest unilaterally or bilaterally.^15,16 Despite the identification of pathogenic variants in more than 50 nephrogenesis-associated genes, a clear genetic diagnosis remains elusive in over 80% of affected patients.^17,18 Animal models and population-based studies have provided valuable insights but often fail to capture the cellular complexity of human kidney development and the phenotypic heterogeneity seen in patients.¹⁹

The advent of single-cell and spatial multiomics technologies has begun to transform the field of kidney research.²⁰ These approaches enable unbiased profiling of distinct molecular layers such as DNA (e.g., assay for transposase-accessible chromatin using sequencing [ATAC-seq] for chromatin accessibility), RNA (RNA-seq), proteins (proteomics), and metabolites (metabolomics) at bulk, single-cell, and tissue-resolved scales (for a description of technologies: Table 1). When two or more omics modalities are integrated, they are considered as a multiomics approach, which allows the mapping of developmental processes at unprecedented resolution. Because no single technology can capture all data modalities, integration across datasets is essential—linking chromatin states to gene expression, or embedding spatial context into inferred cell–cell interactions, which represents a major challenge in the field.²¹ Since human kidney development occurs in utero,^19,22 studies in fetal tissues are particularly powerful for defining regulatory networks, lineage trajectories, and microenvironmental interactions driving nephrogenesis and maldevelopment^23–25 (Table 2). Multiomics applications have additionally yielded insights into novel CKD mechanisms, including the reactivation of developmental programs in injury²⁶ and fibrosis²⁷ as well as tumorigenesis.²⁸

Table 1.

Brief overview of multiomics technologies applied in human nephrogenesis studies

Technology	Description	Reference
Epigenome
ATAC	Uses a hyperactive transposase enzyme to cut open regions of DNA and insert sequencing adapters, identifying areas of accessible chromatin	82
ChIP-seq	Antibodies are used to pull down DNA-bound proteins (e.g., histones). The associated DNA is sequenced to map binding sites	83
Genome-editing
CRISPR	Uses guide RNAs and Cas proteins to modify or cut specific DNA sequences. This enables targeted genome editing	84
Transcriptome
EEL-FISH	Applies an electric field to accelerate probe hybridization, which improves the sensitivity and speed of RNA detection in tissues	85
HDST	Uses a dense array of barcoded spots on a slide to capture mRNA, which allows high-resolution spatial mapping of gene expression	86
MERFISH	Encodes RNA species with combinatorial barcodes read out by rounds of FISH imaging, enabling highly multiplexed transcripts in situ	87
scRNA-seq	Captures both chromatin accessibility (ATAC) and RNA expression from the same nucleus, linking gene regulation to expression	88
SeqFISH	Leverages fluorescence labeling of designed RNA probes in a sequential approach to spatially determine the expression of hundreds to thousands of genes	37
Proteome
4i	Multiplexed imaging method that allows visualization of many proteins (and protein modifications) in the same biologic sample at subcellular resolution	89
CODEX	Cyclic fluorescent antibody labeling and imaging, which allows multiple proteins to be visualized in the same tissue section	90
DVPs laser microdissection+MALDI-MSI	Combines high-resolution microscopy with mass spectrometry to identify and map proteins directly in tissue sections	38
Metabolome
MALDI-MSI	A laser, which ionized biomolecules from tissue coated with a matrix. This creates spatial maps of proteins, lipids, or metabolites	91
Integrated multiome
SHARE-seq	Shared high-throughput ATAC and RNA sequencing technology for measurement of chromatin accessibility and gene expression to predict single-cell fate outcomes in different tissues	33
SNARE-seq2	Isolated nuclei instead of whole cells, useful for tissue that are hard to dissociate. Nuclear RNA is sequenced	32

4i, iterative indirect immunofluorescence imaging; ATAC, assay for transposase-accessible chromatin; ChIP-seq, chromatin immunoprecipitation sequencing; CODEX, co-detection by indexing; CRISPR, clustered regularly interspaced short palindromic repeats; DVP, deep visual proteomic; EEL-FISH, enhanced electric fluorescence in situ hybridization; FISH, fluorescence in situ hybridization; HDST, high-definition spatial transcriptomics; MALDI-MSI, matrix-assisted laser desorption/ionization—mass spectometry imaging; MERFISH, multiplex error robust fluorescent in situ hybridization; scRNA-seq, single-cell RNA sequencing; SeqFISH, sequential fluorescence in situ hybridization; SHARE-seq, simultaneous high-throughput ATAC and RNA expression with sequencing; SNARE-seq2, single-nucleus chromatin accessibility and mRNA expression sequencing (version 2).

Table 2.

Multiomics-driven advances in human kidney developmental biology

Knowledge Gap	Integrated Omics Technologies	Key Insight(s)	References
Defining nephron progenitor cell states and fate decisions	scRNA-seq+snATAC-seq/ChIP-seq	Within the nephron progenitor cell pool, some cells are self-renewing, while others are primed for differentiation Epigenetic priming of differentiation genes occurs before they are transcriptionally activated	10,73,74
Reconstructing 3D nephron patterning and segment identity	Spatial transcriptional profiling+snATAC-seq+proteomics	Multiomics integration reconstructed how the nephron architecture emerges. Each segment has its own identity regarding transcription factors, open chromatin, and signaling pathway activation	10,75
Understanding metabolic control of progenitor fate	scRNA-seq+spatial metabolomics	Nephron progenitor self-renewal is linked to a glycolytic metabolic state, while differentiation requires a metabolic shift to oxidative phosphorylation	40,76,77
Linking GWAS variants to functional elements in nephrogenesis	GWASs+scRNA-seq+ATAC-seq/ChIP-seq	Identification of gene regulatory networks of normal and abnormal kidney development. Noncoding GWAS hits for congenital kidney anomalies are found in enhancer regions for key kidney development genes within developing kidney cells	34,78
Testing functional model fidelity for kidney development	scRNA-seq+snATAC-seq+proteomics	Different gene regulation patterns between human and mice Validation of human kidney organoids as a model for kidney development	7,79–81

For a description of technologies and abbreviations, see Table 1. ATAC-seq, assay for transposase-accessible chromatin using sequencing; ChIP-seq, chromatin immunoprecipitation sequencing; GWAS, genome-wide association study; scRNA-seq, single-cell RNA sequencing.

In this review, we highlight key examples where integrated multiomics have advanced our understanding of human kidney development and disease. We discuss the technical challenges of integrating multilayered data in human kidney development before discussing three use cases: (1) the development of the Human Nephrogenesis Atlas,¹⁰ (2) human–mouse kidney developmental comparisons, and (3) insights into the reactivation of developmental programs during injury and disease.

Finally, we outline a translational research trajectory, from genetic variant through mechanism to clinic. Developmentally resolved multiomics maps can link (non)coding genomic variation to stage type and cell type–specific regulatory elements, nominate transcriptional regulators and targets, and inform functional models such as kidney organoids. These frameworks hold promise to improve prenatal diagnostics, refine prognostication by aligning fetal signatures with postnatal trajectories, and identify therapeutic entry points within defined developmental windows. With these advances, multiomics is shifting kidney developmental biology from descriptive to mechanistic understanding, laying the foundation for precision diagnostics, prognostics, and therapies for children with congenital kidney anomalies.

Overview of Recent Multiomics Development

Over the past decade, methodologic innovations have dramatically expanded the scope, resolution, and applicability of single-cell and spatial omics.^29,30 These advances have enabled researchers to not only profile individual molecular layers but also interrogate them in an integrated fashion across developmental stages and disease contexts.³¹ Joint multiomics profiling of RNA and chromatin accessibility (assay for transposase-accessible chromatin [ATAC]) from the same cell (10X Multiome, single-nucleus chromatin accessibility and mRNA expression sequencing [version 2],³² simultaneous high-throughput ATAC and RNA expression with sequencing³³) has linked transcriptional output to upstream regulatory elements and, importantly, facilitated the functional interpretation of noncoding genome-wide association study (GWAS) variants by connecting them to gene regulatory networks and cell type–specific expression programs.³⁴ In parallel, high-resolution spatial imaging–based methods have been developed for both the transcriptome (sequential fluorescence in situ hybridization, Slide-seqV2, Stereo-seq, multiplex error robust fluorescent in situ hybridization, Xenium) and the proteome (codetection by indexing, iterative indirect immunofluorescence imaging), enabling the direct mapping of molecular programs into their tissue context.^35–37 Whereas these spatial proteomics approaches rely on multiplex antibody stainings, the recent introduction of deep visual proteomics has extended proteomic profiling at cellular resolution to unbiased, mass spectrometry–based analyses³⁸ and has already been applied successfully to several human tissues.³⁹ Additional breakthroughs include the development of spatial metabolomics, beginning to uncover metabolic compartmentalization within tissue microenvironments,⁴⁰ and spatial RNA+ATAC co-sequencing, which allows in situ mapping of chromatin accessibility and links regulatory landscapes directly to tissue architecture and metabolism.⁴¹

A persistent limitation of multiomics studies has been the scarcity of high-quality fresh-frozen human tissue.⁴² The recent adaptation of single-cell and spatial omics technologies to formalin-fixed paraffin-embedded (FFPE) material has significantly expanded the availability of samples for analysis, in some cases producing transcriptomic data of even higher quality than from fresh-frozen specimens^43,44 and introducing spatial ATAC-seq from FFPE tissues.⁴⁵ This development is particularly valuable for studies of the developing kidney, where access to tissue is extremely limited. The ability to apply multiomics approaches to archival FFPE samples has greatly increased the potential to investigate rare human developmental and disease states, providing new opportunities to study fetal kidney biology in unprecedented detail.

Barriers to Defining Single-Cell and Spatial Circuits of Kidney Development

A nonuniversally conserved combination of molecular and phenotypic traits defines developmental cell states of the kidney. Unlike terminally differentiated cells, developing cell types often lack stable, universal markers and instead display context-dependent molecular signatures that change as differentiation progresses.⁴⁶ This heterogeneity, both within and across tissues, poses major challenges for identifying and distinguishing specific cell populations under physiologic conditions. The dynamic nature of nephrogenesis further complicates this issue, as the same progenitor cell type may display plasticity and different transcriptional or epigenetic states depending on its developmental stage or microenvironmental niche.¹⁰ Thus, several limitations have to be overcome (Figure 1).

Figure 1.

Barriers for integrated multiomics approaches in human kidney development and disease. Besides the inherent sparsity of high-quality fetal kidney tissues due to limited availability and ethical constraints, the integration of different multiomics technologies can be hampered by loss of information, e.g., due to technical bias caused by batch effects, loss of spatial context, as well as insufficient data capture at single-cell resolution, e.g., due to overrepresentation of (developing) PT cells and transient molecular upregulation in other developmental cell types. It is pivotal to account for these specific barriers when designing multiomics approaches to advance our understanding of human kidney development. FFPE, formalin-fixed paraffin embedded; PT, proximal tubular.

Molecular Information Bias at the Single-Cell Level

Omics data at the single-cell level are inherently sparse, with only a small fraction of the transcriptome, epigenome, or proteome captured in each cell. This sparsity is further compounded by variability across omics modalities, as each method has its own technical biases and capture inefficiencies. For example, single-cell RNA sequencing (scRNA-seq) often fails to detect low-abundance transcripts, whereas ATAC-seq preferentially identifies highly accessible chromatin regions, leaving other regulatory elements underrepresented.⁴⁷ Furthermore, single-nuclei RNA sequencing data tend to be particularly noisy and contaminated with ambient RNA, necessitating the use of sophisticated computational tools to remove it.⁴⁸ As a result, reconstructed molecular circuits provide only a partial view of the underlying biology, thereby limiting the precision with which developmental trajectories can be theoretically defined.

Limited Availability of Fetal Kidney Tissue

Another significant barrier is the rarity of human fetal kidney tissue, which is the only context in which human nephrogenesis can be directly studied. Access is limited both ethically and practically, and when tissue becomes available, it is often preserved as FFPE material rather than as fresh-frozen samples. While recent methodologic advances have expanded the usability of FFPE samples for single-cell and spatial assays, fresh-frozen tissue still produces higher-quality data for certain modalities, creating trade-offs that researchers must consider. While FFPE isolated nuclei can be used for scRNA-seq from archival samples, ATAC-seq is best performed from nuclei isolated from frozen tissues. The same is true for single-cell (epigenetic) methylation studies.⁴⁹ For the combined analysis of RNA+ATAC, currently, only frozen tissues can be used as input to yield sufficient data quality. This scarcity of high-quality developmental kidney tissue is further affected given that kidney biopsies are clinically contraindicated in congenital kidney anomalies and, as such, remains a key limitation for large-scale multiomics studies.

Technical and Sampling Biases

Even when tissue is available, technical and sampling biases can distort the representation of cellular diversity. For developmental studies, comparable with working with kidney biopsies, the volume of the study material is typically challenging. In addition, in the kidney, proximal tubular cells are heavily enriched in mitochondria, which frequently leads to a high proportion of mitochondrial transcripts, noisy data, and reduced nonmitochondrial sequencing depth compared with other cell types. As such, this typical overrepresentation within datasets (approximately 80% of total cells) requires correction.²⁷ Immune populations such as neutrophils are often lost during nuclei isolation from frozen tissues due to their nuclear architecture, while other populations, including fibroblasts embedded in fibrotic tissue regions or tissue structures such as blood vessels, may be underrepresented because of difficulties in dissociation or nuclei recovery. These biases complicate the interpretation of single-cell datasets and may lead to systematic underestimation of specific cell states that are critical for developmental or disease processes.

Challenges in Spatial Profiling

Bulk and dissociative single-cell methods cannot capture the precise tissue localization or niche-specific interactions of developing kidney cells. While emerging spatial transcriptomic and proteomic platforms offer means to retain positional information, they also come with limitations.⁵⁰ Many suffer from restricted gene or protein coverage (1000s for genes and typically 10s for proteins), signal dropout in low-quality tissue regions, and batch effects that hinder cross-sample comparison and data integration. Resolution varies substantially across platforms, ranging from approximately 2–8 µm for 10XGenomics Visium HD⁵¹ to submicrometer scale for Stereo-Seq⁵² or open-source spatial transcriptomics,⁵³ which affects the ability to resolve individual cells or even subcellular compartments. As a result, while spatial omics is beginning to reveal the cellular architecture of nephrogenesis, technical variability and resolution limits continue to constrain its full potential.

Challenges in Integrating Multiomics Datasets

Finally, the integration of multiomics datasets comes with distinct challenges.⁵⁴ Data are high-dimensional and heterogeneous and often contain substantial proportions of missing values across modalities. These issues make it challenging to build coherent maps of developmental circuits without advanced computational strategies for data harmonization and imputation. Because developmental phenotypes are encoded across transcriptional, epigenetic, proteomic, metabolic, and morphologic layers and no single modality is sufficient to define cellular states holistically on its own, there is an additional need for standardized definitions, consistent marker panels, and shared data to enable comparison across laboratories and technologies. Recent computational advances, including algorithms for joint embedding, machine-learning algorithms such as manifold alignment,⁵⁵ and graph-based integration,⁵⁶ offer promising solutions, but their application to kidney development data remains unexplored.

Examples of Multiomics Use Cases in Understanding Human Kidney Development

Spatiotemporal Profiling of Nephron Structures—The Human Nephrogenesis Atlas

Understanding the cellular composition of the developing kidney and how these cells differentiate to form the intricate kidney architecture has been a long-standing focus in developmental biology.^6,57 The Human Nephrogenesis Atlas established by Lindström et al. represents the state-of-the-art reference for human kidney development.¹⁰ By integrating scRNA-seq with multiplexed 3D protein imaging of the nephrogenic zone, they generated a comprehensive temporal and spatial model of nephron formation. This multiomics integration was critical because it enabled 3D spatial transcriptional mapping of nephron structures generated after tissue clearing and multiplex-IF imaging, something neither approach could achieve alone. In doing so, the study reconstructed the trajectories of progenitor cells as they progressed through pretubular aggregates and renal vesicles into fully patterned nephron compartments, providing a dynamic framework for cell fate decisions during human nephrogenesis (Figure 2).

Figure 2.

Integration of scRNA-seq and multiplexed 3D protein imaging yields unprecedented views of nephron segment patterning in human nephrogenesis molecular programs. Multiomics integration of RNA sequencing and proteomics of human fetal kidneys enabled the reconstruction of nephron progenitor cell trajectories from pretubular aggregate through renal vesicle into fully patterned nephron compartments.¹⁰ This nephron patterning was visualized across developmental time, revealing the stereotypical architecture of nephron segmentation during human nephrogenesis. scRNA-seq, single-cell RNA sequencing.

A second major insight from this work was the visualization of nephron patterning in three dimensions across developmental time. The integration of transcriptomic states with protein-based models revealed how proximal–distal identities are established and refined within the emerging nephron, and how progenitor plasticity underlies the stereotypical architecture of nephron assembly. These breakthroughs were only possible because of the multiomics integration, which directly connected molecular trajectories to morphogenetic events. Together, these findings define the Human Nephrogenesis Atlas as the benchmark for studying normal kidney development and as a foundation for investigating how its disruption contributes to congenital kidney anomalies.

From Kidney Single-Cell Multiome to GWAS Integration

Recent multiomics approaches have been essential in redefining the validity of murine kidney development as a model for human health and disease. By comparing integrated chromatin accessibility and gene expression profiles, generated through ATAC-seq and scRNA-seq/single-nucleus RNA sequencing, Kim et al. identified both conserved and divergent molecular features of nephrogenesis between developing human (n=10; 79,629 nuclei with a median of 6476 fragments per cell) and mouse kidneys (n=13; 33,235 nuclei with a median of 12,457 fragments per cell⁷) (Figure 3). Transcriptional and epigenomic signatures correlated closely across species, as reflected by shared cell type–specific transcription factor activity, binding motifs, and accessible promoter regions. Yet, important differences were uncovered in the velocity of developmental programs. For instance, human markers of podocyte (POD) differentiation such as phospholipase A2 receptor and decorin showed limited or absent gene expression and chromatin accessibility in mice, suggesting species-specific glomerular maturation. By contrast, developmental time trajectories for proximal and distal tubular epithelia were largely conserved. Understanding these spatiotemporal differences is particularly relevant for modeling congenital kidney anomalies, where disrupted developmental timing could contribute to abnormal kidney architecture.

Figure 3.

Multiomics integration uncovers similarities and differences in molecular trajectories of human and murine kidney development. Although the mouse and human kidney differ in size, anatomy, and nephron number, combining RNA-seq and ATAC-seq data of human and murine nephrogenesis demonstrated shared cell type–specific transcription factor activity, binding motifs, and accessible promoter regions across species.⁷ In addition, striking differences exist between species in the velocity and regulatory elements of these molecular programs. Subsequent connection of these distinct chromatin regions to noncoding variants in CAKUT-associated genes identified by GWAS datasets creates functional regulatory networks essential to discern abnormality from normal human kidney development. ATAC-seq, assay for transposase-accessible chromatin using sequencing; CAKUT, congenital anomalies of the kidney and urinary tract; GWAS, genome-wide association study; POD, podocyte.

In total, the authors identified 1390 cell type–specific gene expression signatures that were conserved across human and mouse nephrogenesis, while 885 gene profiles were enriched in human kidney cells. Strikingly, the chromatin landscape diverged more strongly: 61% of accessible chromatin regions were human-specific, underscoring that transcriptional programs may appear conserved even when their upstream regulatory architecture is not. This integrative multiomics approach thus suggests that, despite millions of years of divergent evolution, mammalian kidney development is driven by deeply conserved molecular mechanisms but fine-tuned by species-specific regulatory elements.

The integration of RNA and ATAC data is particularly powerful for the functional interpretation of GWAS signals. Computational tools such as Multimarker Analysis of GenoMic Annotation⁵⁸ and related approaches can statistically link GWAS variants to genes by using genomic proximity, expression quantitative trait locus/loci information, or coexpression networks even using spatial transcriptomic approaches.⁵⁹ While valuable, these inference-based methods remain limited, as they cannot directly identify which regulatory element in which cell type is driving disease associations. By contrast, multiome datasets provide a direct mechanistic bridge, as they capture both chromatin accessibility and transcriptional output from the same cell, thereby connecting noncoding variants to the regulatory regions they affect and to the genes they control. Kim et al. demonstrated this principle by overlaying human-specific accessible chromatin with GWAS datasets, identifying risk variants in regulatory regions of kidney developmental genes such as UMOD and GREB1L,⁷ which have been implicated in congenital kidney anomalies.^60,61 This finding aligns with recent studies in adult kidney, such as the study by Liu et al., which demonstrated how joint multiome profiling enables the fine mapping of noncoding GWAS variants into functional regulatory networks of the human kidney.³⁴

Linking Adult Kidney Disease with Developmental Gene Regulatory Circuits

A recurring theme in kidney disease is that injury responses frequently reuse developmental gene programs,^62,63 and dissecting these programs requires approaches that can connect regulatory mechanisms to transcriptional outputs.^64–66 During kidney development, the transcription factor WT1 orchestrates the specification of nephron progenitors and their differentiation into PODs.⁶⁷ Pathogenic variants in WT1 cause congenital anomalies of the kidney and urinary tract, including Wilms tumor, Aniridia, Genitourinary anomalies, and Range of developmental delays (syndrome), Denys–Drash, and Frasier syndromes.⁶⁸ Ettou et al. demonstrated that these developmental programs are re-engaged during POD injury, providing evidence that embryonic signatures are recapitulated in disease⁶⁹ (Figure 4).

Figure 4.

Cellular responses to kidney injury recapitulate human nephrogenesis molecular programs. Integration of ChIP-seq and RNA-seq data from PODs subjected to injury revealed dynamic expression and chromatin binding of WT1 (right panel), a key transcription factor in the differentiation from nephron progenitor cells into PODs during nephrogenesis (left panel), which subsequently results in transient upregulation of pivotal genes for POD architecture such as Nphs2 and Synpo.⁶⁹ This multiomics approach linking epigenomic regulation to transcriptional outcomes demonstrated the identification of developmental programs that are typically briefly reactivated and then silenced. Defining the causal regulatory relationships during injury may additionally enable the identification of molecular targets for prognostication and treatment for individuals affected by kidney injury as well as congenital kidney anomalies. ChIP-seq, chromatin immunoprecipitation sequencing.

The strength of this study was the use of a multiomics framework combining RNA-seq and chromatin immunoprecipitation sequencing in murine PODs subjected to injury. RNA-seq alone could have revealed changes in transcript levels of POD genes such as Nphs2 and Synpo but would not have explained how these genes were regulated. Similarly, chromatin immunoprecipitation sequencing alone would have shown genome-wide WT1 occupancy, but without information on whether binding events were associated with transcriptional activation or repression. By integrating both datasets, the authors demonstrated that WT1 directly controls injury-induced gene expression programs through dynamic chromatin binding. Specifically, WT1 initially increased binding to promoter regions of POD genes, leading to transient upregulation of Nphs2 and Synpo, but over time, binding intensity decreased, chromatin accessibility was lost, and these transcripts were downregulated, revealing a two-step regulatory process that drives POD dysfunction.

This integrated view could not have been achieved by either method alone. Only the multiomics approach linked epigenomic regulation to transcriptional outcomes, uncovering the temporal sequence of events in which developmental programs are briefly reactivated and then collapse.

These findings illustrate the power of multiomics in defining causal regulatory relationships rather than descriptive correlations and highlight how developmental transcriptional circuits are redeployed during injury.

Future Perspective

In the near term, we anticipate the development of consensus, developmental stage–resolved reference maps that integrate genetic variants, RNA, chromatin, protein, metabolites, as well as spatial information across key renal compartments at single-cell resolution. Although transcriptomic, metabolomic, and chromatin accessibility datasets are now available at high resolution (Table 2), proteome and spatial profiling of kidney development remain relatively underdeveloped, particularly at high data quality. Adding these missing dimensions will be essential to capture the niche architecture, signaling gradients, and cell–cell interactions that cannot be resolved in dissociated single-cell assays. Another potential molecular layer that now could be explored is genomic organization of developing kidney cells, by visualizing chromatin organization at the single nuclear level with two-layer DNA sequential fluorescence in situ hybridization+.⁷⁰ These references will form the foundation for systematic benchmarking tools for kidney organoids to recapitulate human nephrogenesis,⁷¹ as current models, e.g., demonstrate incomplete maturation, lack of vascular cell types, and limited 3D architecture.⁷² Enabling such systematic benchmarking of differentiation fidelity and refinement of protocols are pivotal for future modeling of congenital kidney anomalies and CKD.

At the same time, the integration of artificial intelligence (AI) approaches, such as variational autoencoders and emerging agent-based generative models, will allow the comodeling of multiomics data across modalities, developmental stages, and perturbations, while explicitly handling uncertainty. Such models are expected to improve the resolution of rare cell states, predict hidden regulatory interactions, and simulate variant effects in silico before experimental testing.

In the medium term, the routine use of perturbation-based multiomics in organoid systems should enable mechanistic dissection of developmental trajectories, variant effects, and candidate therapeutic pathways directly in humanized disease models. By combining kidney organoid fidelity maps with AI-driven modeling, it will become feasible to not only to recapitulate known developmental programs but also predict alternative differentiation routes and their consequences for disease susceptibility.

In the longer term, the assembly of integrated atlases spanning human kidney development from initiation to completion will clarify which developmental programs are partially reactivated during injury and which are injury-specific. Coupled with AI-driven inference, these atlases will provide a dynamic framework for identifying windows of therapeutic opportunity and designing interventions for affected individuals with congenital kidney anomalies that align with developmental timing and cell type context.

Conclusions

In summary, we have highlighted how multiomics have transformed our understanding of normal and abnormal human nephrogenesis. By integrating transcriptomic, epigenomic, proteomic, and metabolomic data with advanced in vitro and in vivo models, the field will move beyond descriptive catalogs toward a new integration era in which regulatory logic, spatial architecture, and functional validation are considered together. Yet, critical gaps remain: Spatial profiling of human kidney development, especially at high resolution and data quality, is still limited, even though this dimension is indispensable for capturing niche organization and cell–cell interactions.

For patients and families affected by congenital kidney anomalies and childhood CKD, these innovations hold promise. Multiomics approaches can connect genetic variants—especially those in noncoding regions identified by GWAS or sequencing—to their regulatory elements, target genes, and affected developmental programs. This capacity to move from variant discovery to functional interpretation opens the door to molecularly precise diagnostics, individualized prognostic tools, and eventually genotype-informed therapies.

Looking ahead, organoids benchmarked against developmental reference maps will serve as powerful models to study disease-causing variants in human tissue contexts, while AI-driven frameworks, including variational autoencoders and emerging agent-based models, will accelerate the integration of diverse omics layers, predict hidden regulatory states, and simulate the effects of specific mutations before clinical intervention. Ultimately, assembling integrated atlases spanning fetal to postnatal stages will clarify which developmental programs are reactivated during injury and which are disease-specific, guiding rational therapeutic strategies that align with developmental timing and cell type context.

Meeting the biologic, technical, and computational challenges will determine whether multiomics can deliver on this promise. But if successful, integrative multiomics platforms embedded in large-scale translational research frameworks will transform how we diagnose, stratify, and treat patients with congenital kidney anomalies and genetically driven kidney diseases, ensuring that discovery is meaningfully translated into improved patient outcomes.

Supplementary Material

jasn-37-585-s001.pdf ^{(1.4MB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/JSN/F528.

Author Contributions

Conceptualization: Luna S. Klomp, Christoph Kuppe, Elena Levtchenko, Lampros Mavrommatis, Rik Westland.

Supervision: Christoph Kuppe, Elena Levtchenko, Rik Westland.

Visualization: Luna S. Klomp, Christoph Kuppe, Lampros Mavrommatis, Rik Westland.

Writing – original draft: Luna S. Klomp, Christoph Kuppe, Rik Westland.

Writing – review & editing: Fanny O. Arcolino, Luna S. Klomp, Christoph Kuppe, Hildo C. Lantermans, Elena Levtchenko, Lampros Mavrommatis, Rik Westland.

Funding

R. Westland: Nierstichting (24OK1076, 20OC002, and 24OM+012). L.S. Klomp: Amsterdam University Medical Centers (MD PhD grant). E. Levtchenko: European Research Council (ERC-CoG-101045467). C. Kuppe: Deutsche Forschungsgemeinschaft (445703531, 459969915, and 545524314), European Research Council (ERC-StG-101040726), Else Kröner-Fresenius-Stiftung, and BMBF.