Nephron progenitors rhythmically alternate between renewal and differentiation phases that synchronize with kidney branching morphogenesis

Abstract

The mammalian kidney achieves massive parallelization of function by exponentially duplicating nephron-forming niches during development. Each niche caps a tip of the ureteric bud epithelium (the future urinary collecting duct tree) as it undergoes branching morphogenesis, while nephron progenitors within niches balance self-renewal and differentiation to early nephron cells. Nephron formation rate approximately matches branching rate over a large fraction of mouse gestation, yet the nature of this apparent pace-maker is unknown. Here we correlate spatial transcriptomics data with branching ‘life-cycle’ to discover rhythmically alternating signatures of nephron progenitor differentiation and renewal across Wnt, Hippo-Yap, retinoic acid (RA), and other pathways. We then find in human stem-cell derived nephron progenitor organoids that Wnt/β-catenin-induced differentiation is converted to a renewal signal when it temporally overlaps with YAP activation. Similar experiments using RA activation indicate a role in setting nephron progenitor exit from the naive state, the spatial extent of differentiation, and nephron segment bias. Together the data suggest that nephron progenitor interpretation of consistent Wnt/β-catenin differentiation signaling in the niche may be modified by rhythmic activity in ancillary pathways to set the pace of nephron formation. This would synchronize nephron formation with ureteric bud branching, which creates new sites for nephron condensation. Our data bring temporal resolution to the renewal vs. differentiation balance in the nephrogenic niche and inform new strategies to achieve self-sustaining nephron formation in synthetic human kidney tissues.

Introduction

Collectives of nephron progenitor cells cap the branching ureteric bud tree during kidney development, periodically contributing cells that condense into discrete pre-tubular aggregates (PTAs) — the earliest nephron stage. This process occurs continuously but asynchronously across ureteric bud tips¹. Understanding the rate at which nephrons form is an important factor in addressing person-to-person variability in nephron endowment, which impacts adult kidney function^2-6. A better understanding would also inform in vitro synthetic kidney tissue production using organoids, where contemporary protocols produce nephrons in a single inductive wave rather than in sustained waves^7,8. In vivo, nephron progenitor (‘cap mesenchyme’) niches balance renewal vs. differentiation by interpreting autonomous cues and those from adjacent ureteric bud and stromal cell populations⁹ (Fig. 1A,B). Only a portion of a cap’s progenitors commit to each successive nephron, enabling niches to maintain persistent progenitor pools during development¹. It is known that FGF/EGF/MAPK^10-13, PI3K/Akt¹⁴, Wnt/β-catenin (low/high)^15-18, Hippo^19,20, JAK/STAT^21,22, TGFβ/BMP^23,24, and Notch^25-27 pathways are important for progenitor renewal and differentiation. However, the relative rates of renewal vs differentiation may not be constant in time, since cells commit as transient avalanche-like streams from the cap mesenchyme into discrete PTAs²⁸. Indeed, we recently reported that nephron formation rate is not constant over the ureteric bud branching cycle, but rather nephron formation pauses after each branch event and resumes before the next one²⁹. These observations suggest both a switch-like response of progenitors to inductive cues as single cells and an avalanche-like response to inductive cues as cell collectives. The Wnt ligand Wnt9b presented to the cap by the nearby ureteric bud is thought to be a dominant cue^15,18,30. The switch-like response may be explained by a transition in SIX2/TCF/LEF/β-catenin factor occupancy at target genes upon rising β-catenin levels caused by stochastic exposure of cells to Wnt9b as they migrate in and out of several sub-domains in the niche^{16,17,23,28,31-34}. However, the origin of avalanche-like commitment of batches of cells to early nephron aggregates is unknown. This may relate to feedback associated with the ureteric bud branching ‘life-cycle’¹⁷ (Fig. 1C,D), potentially through the effect of its curvature on Wnt9b concentration gradients³⁵, although this is debated³⁰. Indeed, any life-cycle-dependent change in nephron progenitor contact area with the ureteric bud or stroma²⁰, ‘mixing’ state among these compartments^36-38, or niche mechanical microenvironment²⁹ are reasonable hypotheses yet to be disproven.

Fig. 1: Defining branching ‘life-cycle’-dependent regulation in nephron progenitor niches.

(A) Confocal immunofluorescence renderings of the mouse embryonic kidney ureteric bud tubule tree at embryonic day 14 (E14) and example niches at its tips at E17. Each tip is surrounded by SIX2+ nephron progenitors (the cap mesenchyme) and bordered by PDGFRA+ stroma. (B) Side view of several niches at E17 by confocal immunofluorescence. (C) Schematic of ureteric bud branching, which causes niches traveling with them to divide periodically. Branching can be described as a ‘life-cycle’ due to self-similarity between a branch just beginning to divide and one of the daughters formed after branching. Since branching occurs asynchronously among ureteric bud tips/niches, a ‘pseudo-time’ concept enables their life-cycles to be aligned with each other. (D) At a given real time-point, niches of different age in branching pseudotime co-exist. (E) Schematic hypothesis for ‘just in time’ contribution of committing nephron progenitors to condensation sites newly formed by branching morphogenesis. Nephron progenitors would renew between branching events (cap mes.: cap mesenchyme).

Here we find that the avalanche-like commitment of nephron progenitors coincides with rhythmic transcriptional changes across several pathways synchronized with the ureteric bud branching life-cycle, interspersed with pauses²⁹ in which nephron progenitors renew. This may act to match nephron formation rate to the ureteric bud branching rate. We next transfer this emerging control principle to a human induced pluripotent stem cell (iPSC)-derived nephron progenitor organoid system. Overlapping or staggered activation of the candidate rhythmic pathways YAP or retinoic acid successfully shifts interpretation of an inductive Wnt/β-catenin signaling cue by nephron progenitors to affect renewal vs. differentiation balance and nephron size/segment balance. These data establish a rhythmic homeostatic principle of nephrogenic niche regulation that achieves ‘just in time’ delivery of committing cells to nephron condensation sites newly formed by branching (Fig. 1E). This principle presents the opportunity to create self-sustaining nephrogenic niches in vitro by mimicking cycles of alternating differentiation and renewal cues.

Results

Pseudo-time ordering of niche transcriptomes to enrich for correlations between nephron progenitor state and ureteric bud branching life-cycle

The rates of nephron formation and ureteric bud branching in the mouse embryonic kidney are equivalent from E15-E22¹, suggesting that a type of pace-maker synchronizes the two processes. We wondered if this pace-maker would manifest in a cyclical transcriptional state of nephron progenitors over the branching life-cycle. This required us to 1) sequence nephron progenitors from many separate niches, and 2) assign niches to ‘pre-branching’, ‘mid-branching’, and ‘post-branching’ ‘age’ groups before comparing gene expression between them. To do so we turned to Nanostring GeoMx spatial sequencing technology³⁹, which performs multiplexed bulk RNA sequencing within manually annotated regions of interest (ROIs). GeoMx is performed on 5 μm-thick tissue sections. Since the kidney surface is curved, we partially flattened embryonic day 17 (E17) mouse kidneys prior to sectioning in order to sequence ROIs at the same z-depth from the organ surface (Fig. 2A, Methods). We drew ROIs around SIX2+ nephron progenitor clusters associated with 47 niches and sequenced each niche separately. We next classified the age of the niches associated with each ROI relative to the ureteric bud branching cycle using two qualitative metrics (branch point position and cap mesenchyme connectivity, Fig. 2B,C). As expected, median gene expression scaled with ROI area, which normalization compensated for (Fig. 2D, Methods). The cap mesenchyme ROIs were drawn to exclude stromal cells, nephron connecting segments, and PNA lectin-stained ureteric bud cells that are each readily distinguished from nephron progenitors by their morphology and location in the niche. Other committed cells in pre-tubular aggregates, renal vesicles, and later stages were intrinsically excluded during sample preparation because they are found in deeper sectioning planes. While these off-target tissue compartments were not sampled in the same experiment, expression of canonical markers for them were not significantly differentially expressed between ROIs associated with pre-, mid-, and post-branching ureteric bud tips (Fig. 2E). These data reflect successful recovery of transcriptomes from nephron progenitor collectives associated with individual niches using GeoMx and suggest that any contamination by off-target cells was consistent among ROIs across the age groups.

Fig. 2: Collecting and classifying bulk transcriptomes of nephron progenitor niches from three branching life-cycle stages.

(A) Top, Schematic of E17 kidney flattening, sectioning and processing for GeoMx spatial sequencing. Transcripts from cap mesenchyme cells in 47 individual regions of interest (ROIs) were collected and sequenced. Bottom, fluorescence micrograph of the slice used in this study. Ureteric bud tips from which cap mesenchyme transcriptomes were sampled are annotated. (B) Top, three example cap mesenchyme niches scored for life-cycle stage. Middle, cartoons representing niche morphology in top and side views. Bottom, Qualitative life-cycle metrics (branch point appearance in z and cap mesenchyme connectivity) used to score each ROI as pre-, mid-, or post-branching. (C) Fluorescence micrographs of all niches after scoring. Insets, example niches and GeoMx ROIs. (D) Plots of median gene expression vs. GeoMx ROI area before and after Q3 normalization, showing correction for systematic variation in ROI area between pre-, mid-, and post-branching niches. Points represent individual ROIs (cap mesenchyme niches). (E) Normalized expression of marker genes for cap mesenchyme and surrounding tissue compartments in GeoMx ROIs (PTA: pre-tubular aggregate; cap mes.: cap mesenchyme). Colored squares indicate tissue compartments associated with each marker.

Nephron progenitor niche transcriptomes reveal rhythmic changes in renewal and differentiation pathways over the ureteric bud branching life cycle

Starting from our GeoMx results, we sought to detect transcriptional differences between nephron progenitor collectives in pre-, mid-, and post-branching niches. Non-linear dimensional reduction by tSNE largely grouped ROIs by their age categories, showing some global transcriptional distinction between them (Fig. 3A). We next moved to (‘all-gene’) differential expression analysis and validated expression of certain transcripts in the cap mesenchyme by RNA fluorescence in-situ hybridization (HCR RNA-FISH)^40-42, indicated with a dagger (†) throughout the text (Fig. S1). Retsat, Mucl1, and Mrps17† were hits at false discovery rate (FDR) < 0.1, and Stox2 at FDR < 0.3 (Fig. 3B). Significance testing here was robust to randomizing the age categories of the ROIs (Fig. S2). These hits together suggest possible differential regulation of retinoic acid (Retsat), Wnt/β-catenin (Mucl1)⁴³, TGFβ (Stox2)⁴⁴, and PI3K/Akt (Mrps17)⁴⁵ signaling pathways in nephron progenitors over the branching life-cycle (Note S1).

Fig. 3: The cap mesenchyme transcriptome at E17 varies in several renewal and differentiation pathways depending on ureteric bud branching life-cycle stage.

(A) tSNE plot shaded by niche ‘age’ since branching. Envelopes are a visual aid, bounding >75% of the ROIs in each category. (B) All-gene differential expression analysis and violin plots for hits significant at false discovery rate (FDR) < 0.3. (C) Summary table of false-discovery rates (FDRs) for curated gene set enrichment analysis (GSEA). Gene set details and analyses are provided in Fig. S3. (D) GSEA results for a gene set created from those differentially expressed in ‘committing’ nephron progenitors vs. nephron progenitors based on data from Lawlor et al. 2019. (E) Summary table for differential expression of curated marker and signaling genes relevant to nephron progenitor renewal and early differentiation. (F) tSNE plots of E18.5 mouse kidney scRNA-seq data from Combes et al. 2019 showing: left, Cell type annotations (NP: nephron progenitor, EPT: early proximal tubule, DT/LoH: distal tubule/loop of Henle, RV/SSB: renal vesicle/S-shaped body, PTA: pretubular aggregate, PT: proximal tubule, comm.: committing, NP-STR: stromal-like NP, CC: cell cycle-associated, CnS: connecting segment, podo.: podocyte), middle, feature plots of nephron progenitor (Six2) and differentiation (Wnt4, Pax8) markers. Right, feature and violin plots by cell type of mean normalized expression of rhythmic GeoMx hits from (E) in Combes scRNA-seq clusters.

We moved to gene-set enrichment analysis (GSEA)⁴⁶ to examine these pathways and others with known roles in nephron progenitor renewal, priming, and differentiation at higher sensitivity (Fig. 3C, Fig. S3, Note S2, Introduction). We found that gene sets describing 1) retinol metabolism, 2) Wnt pathway, 3) Wnt targets and, with weaker significance, 4) FGFR signaling and 5) positive regulation of Notch signaling genes were enriched in nephron progenitors from post-branching vs. pre-/mid-branching niches (minimum FDRs = 0.017, 0.14, 0.19, 0.27, and 0.25 respectively). Closer inspection of enriched transcripts suggested that Wnt/β-catenin and retinoic acid signaling may actually be higher in nephron progenitors in pre-/mid-branching niches (Note S3). Gene sets describing 1) Hippo signaling (Note S3, and similarly,) 2) negative regulation of organ growth, 3) JAK-STAT signaling (notably Hras¹⁰), and 4) TGFβ signaling were enriched in nephron progenitors from pre-/mid-branching vs. post-branching niches (minimum FDRs = 0.044, 0.027, 0.076, and 0.188 respectively). These data are evidence that the transcriptional state of cap mesenchyme populations is not monolithic over the ureteric bud branching cycle; rather that nephron progenitors change periodically as a whole or through transitions between sub-populations in pathways relevant to renewal vs differentiation decision-making.

We sought additional evidence for this by testing if, on average, nephron progenitors change their ‘priming state’ as a function of the branching life-cycle. We re-analyzed scRNA-seq data in Lawlor et al. 2019³² to retrieve a list of genes that are differentially expressed between ‘committing’ nephron progenitors vs. naive nephron progenitors in E15.5 mouse kidneys. This gene set was enriched in pre-/mid-branching vs. post-branching niches at p = 0.08 (Fig. 3D, Fig. S4). These data support a rhythmic change in the ‘priming state’ of nephron progenitors synchronized with the branching morphogenesis cycle.

With this evidence for rhythmic regulation of GSEA and progenitor commitment gene sets, we sought to perform gene-by-gene differential expression analysis of a more diverse range of 1) cell identity, and 2) signaling markers relevant to nephron progenitor renewal/differentiation. We created two mutually exclusive lists from literature sources for this ‘curated-gene’ analysis^{10,14,18,19,23,25,28,32,34,36,47-60} (Table S1), scraping genes from scRNA-seq data and triaging them if they lacked a supporting study in nephron development, lacked expression in nephron progenitors, lacked relevance to the correct early commitment stage, and/or were redundant with other marker(s) or isoforms. Differential expression analysis as a function of the branching life-cycle returned the Wnt9b/β-catenin target genes Lef1^18,25 and Pax8^18,61, the TGFβ signaling factor Ltbp1†^58,62, and the cell cycle factor Ttc28†^32,63 as hits at FDR < 0.1 (Fig. 3E, Fig. S5). The first two are of particular note since Lef1 is one of the first markers of nephron induction, followed closely by Pax8^31,64. A range of other genes relevant to or dependent on FGF, TGFβ, Hippo-YAP, and proliferation pathways emerged as branching life-cycle-dependent at FDR < 0.3 (Note S4, Fig. 3E, Fig. S5). This category also included Pdgfra, which we believe indicates rhythmic stromal cell intercalation into the cap mesenchyme niche during post-branching periods based on more detailed analysis (Note S5, Fig. 3E, Fig. S5). These data complement the overall rhythmic priming signal, adding further specificity to branching life-cycle-correlated genes in several key nephrogenesis pathways, particularly downstream of Wnt.

Transcripts enriched in different branching life-cycle phases map to distinct progenitor/nephron sub-populations and exhibit rhythmic expression dynamics

We next sought to determine which cell states and types expressed rhythmic hits from the all-gene and curated-gene differential expression analyses. We first went back to the Lawlor et al. 2019 E15.5 scRNA-seq clusters. On average, hits associated with pre-/mid-branching niches were primarily found in committing nephron progenitors, while hits associated with post-branching niches were primarily found in uncommitted nephron progenitors and nephron progenitors having a stromal-like signature (Note S5, Fig. S6). We repeated this analysis for a higher-resolution E18.5 scRNA-seq dataset reported in Combes et al. 2019⁴⁷ (Fig. 3F, Fig. S7). We recovered similar nephron lineage clusters as Combes and again found a bias in expression of pre-/mid-branching hits vs. post-branching hits along the nephron progenitor differentiation trajectory, but this time without the stromal-like signature (Note S5). Pre-/mid-branching hits (e.g. Ccnd1, Lef1, Pax8, Ltbp1, Mycn, Fgfr2) were enriched in renal vesicle and more mature clusters, while post-branching hits (e.g. Ttc28, Traf1, Fgf1, Tnc, Tgfbi, Col2a1, Gas1, Mest, Eya1, Capn6) were enriched in uncommitted and cell cycle-associated nephron progenitor clusters. These data suggest that all or a fraction of nephron progenitors shift from a differentiating to renewing character between pre-/mid- and post-branching phases.

To create a more continuous picture of transcript dynamics, we used unsupervised clustering using hits across the all-gene and curated-gene sets having FDR < 0.3 and −log₁₀(p) > 2 (Fig. 4A). This roughly clustered nephron progenitor collectives from pre-, mid-, and post-branching niches. The clustering heatmap ordered genes in a wave-like fashion according to their up- and down-regulation along the pseudotime axis from pre- to mid- to post-branching niches. For example, Lef1 and Pax8 clustered together, both being upregulated around the transition from pre- to mid-branching and downregulated in post-branching niches. Several transcripts associated with FGF and TGFβ signaling peaked in mid-/post-branching (Gas1^17,65,66, Fgf1^10,47,48, Fgfr1^47,48,67, Stox2⁴⁴, Tgfbi^47,51,68), and the YAP-dependent gene Traf1¹⁹ along with Ttc28, Retsat, and Mucl1 peaked in post-branching niches, continuing into pre-branching ones (formed as post-branching niches enter another round of branching), Fig. 4A, Fig. S5. The clustering data reinforce the notion of rhythmic, temporally staggered (out-of-phase) transcriptional regulation in several pathways associated with nephron progenitor maintenance and commitment over the branching life-cycle. The temporal relationship between WNT and FGF signaling pathway members/targets along with the GSEA and curated-gene analyses (Fig. 3C-F, Fig. 4A, Fig. S3-6) suggest that nephron progenitor niches may transition through alternating phases of differentiation (in early-middle stages of branching) and renewal (after branching has completed and as the next branching cycle approaches).

Fig. 4: Rhythmic priming/differentiation and renewal in the cap mesenchyme over the niche branching life-cycle.

(A) Unsupervised clustering of differentially expressed curated genes reveals wave-like expression over pre-, mid-, and post-branching nephron progenitor niches (dotted lines added for emphasis). (B-E) Average arbitrary fluorescence (AFU) of hits in cap mesenchymes of pre-, mid-, and post-branching niches, with example confocal immunofluorescence images. Images at bottom show the isolated fluorescence channel for the marker of interest. (F) Schematic model depicting alternating differentiation (pre- and mid-branching) and renewal (post-branching) phases in the cap mesenchyme, annotated by associated GSEA signaling pathways/processes. For B-E each point represents one cap mesenchyme niche, n > 12 niches per category, mean ± S.D. One-way ANOVA, Tukey’s test, *p < 0.05, **p < 0.01, ***p < 0.001.

Protein-level validation confirms rhythmic priming, differentiation, and renewal signatures

We sought to validate a subset of the spatial sequencing hits using immunofluorescence in mouse embryonic kidneys, starting with hits primarily from the all-gene analysis. MUCL1, MRPS17, LTBP1, DHRS3, and CELSR1† antibodies yielded fluorescence consistent with non-specific background. RETSAT showed little protein-level expression in the cap mesenchyme, but did so in early distal nephron/connecting segment cells that are the first to be specified in early nephrons²⁸ (Fig. S8). RETSAT expression there could act to reduce retinoic acid signaling, potentially altering Wnt/β-catenin activation in the distal nephron^25,69,70, a point for future study.

We moved on to focus validation toward priming, differentiation, and renewal markers. Autocrine BMP7 primes nephron progenitors, sensitizing them to subsequent commitment induced by Wnt/β-catenin²³. Prompted by a weak rhythmic signature for Bmp7 (Fig. 3E, Fig. S5) and enrichment of the BMP7 target Id1²⁴ in pre-/mid-branching niches (Figs. S3I, S4), we stained for pSMAD1/5 signaling downstream of BMP7^23,24 (Fig. 4B). pSMAD1/5 was similarly enriched in pre-/mid-branching niches, indicating rhythmic priming of nephron progenitors early in the branching life-cycle.

We turned next to Wnt/β-catenin targets that mark nephron progenitor differentiation^17,18. LEF1 staining peaked in nephron progenitors associated with mid-branching ureteric bud tips, similar to the GeoMx transcript-level data (Fig. 4C). The majority of the signal was contributed by rare LEF1+ ITGA8+ (low) cells typically at the edge of caps adjacent to the cleft between sister ureteric bud tips. These cells are presumably beginning to leave the niche while condensing into nearby pre-tubular aggregates (Note S6). Immunofluorescence for other Wnt9b targets^17,18 PAX8, ITGA8, and CCND1 also showed significant life-cycle correlation. PAX8 peaked in post-/pre-branching niches, lagging the mid-branching peak in Pax8 transcript. This may reflect differences in post-transcriptional timescales (e.g. translation, degradation)⁷¹ between LEF1 and PAX8. These data confirm rhythmic Wnt pathway activity at the protein level in nephron progenitors synchronized with the branching life-cycle, which is relevant to their differentiation¹⁷, mesenchymal-to-epithelial transition⁷², and later establishment of nephron polarity (proximal to distal)²⁵.

As an aside, Pdgfra† (a rhythmic hit in Fig. 3E) is curious since it is a stromal marker. The existence of a stromal-like nephron progenitor population is debated⁴⁷. These cells appear to variable extents in published scRNA-seq datasets^{32,47,66,73,74}, however PDGFRA protein expression is rare in the mouse cap mesenchyme^47,73,75 (Note S5, Fig. S6). PDGFRA staining validated life-cycle-correlated variation in the proportion of stromal-like nephron progenitors and/or stromal cell protrusion into the cap mesenchyme, likely the latter (Fig. 4D, Note S5). Whether stromal intercalation during the post-branching phase causes or results from nephron progenitor differentiation and segregation from the cap mesenchyme is a point for future study³⁸.

Turning to cell renewal, our GSEA and curated-gene analyses suggested that nephron progenitors in post-branching niches may be more proliferative (Fig. 3C,E,F). The GeoMx data showed a decrease in the proliferation marker Mki67 expression in mid-branching niches relative to a peak in post-branching ones (Fig. 4E). We stained for the corresponding protein KI-67, which marks cells in all active phases of the cell cycle, especially S-phase⁷⁶. KI-67 correlated temporally with its transcript and was significantly increased in post-branching niches (Fig. 4E), reflecting an overall proliferation change or a change in the ratio of high/low proliferation sub-populations of nephron progenitors¹. These data reinforce the notion that nephron progenitors favor renewal between branching events and differentiation during them (Fig. 4F).

Rhythmically regulated pathways in the mouse niche are sufficient to modify interpretation of Wnt differentiation signaling in human stem-cell derived nephron progenitors

Having validated rhythmic expression of several targets over the branching life-cycle in nephron progenitors, we sought to reconstitute rhythmic signaling in human iPSC-derived nephron progenitor organoid ‘pucks’ cultured at an air-liquid interface^77-79. Despite some species-specific differences, these cells roughly transit through the same states when differentiated (transcriptionally and in protein marker expression) as nephron progenitors in the mouse niche^75,77,80-82. A ‘pulse’ of Wnt/β-catenin activation via the GSK3β inhibitor CHIR 99021 is commonly used to increase differentiation efficiency of SIX2+ nephron progenitors in these protocols^77,80. We reasoned that combining transient Wnt activation with simultaneous or staggered perturbation of candidate signaling pathways from our GeoMx screen would cause differences in nephron progenitor renewal vs. differentiation if rhythmic signaling was relevant to maintaining this balance. We focused on YAP and retinoic acid signaling due to the rhythmic Hippo signaling and retinol metabolism signatures we uncovered during gene set enrichment analysis (Fig. 3C, Fig. S3A,E), and due to their known roles in nephron progenitor decision-making. Increased YAP signaling in vivo and in vitro appears to drive nephron progenitor renewal at the expense of differentiation, and vice versa (despite some conflicting observations)^{19,20,22,83-86}. According to our GeoMx analysis, Wnt/β-catenin activity is likely higher in nephron progenitors in pre-/mid-branching niches, while YAP activity is likely higher in post-branching niches (Notes S3, S4). Based on this we hypothesized that YAP contributes to nephron progenitor renewal in the post-branching phase. This motivated us to test both simultaneous and staggered Wnt and YAP activation for effects on cell renewal/differentiation via CHIR and the LATS kinase inhibitor (Yap activator) TRULI⁸⁷ in day 10 nephron progenitors derived from two separate iPSC lines (Fig. 5A,B). TRULI has recently been found to recruit YAP to the nucleus of iPSC-derived nephron progenitors and extend their long-term renewal in 2D culture, while retaining their nephron differentiation potential⁸⁴. We verified that TRULI increases YAP nuclear recruitment and target gene expression (CTGF, CYR61) in our day 10 nephron progenitors by immunofluorescence and qPCR respectively (Fig. S9). We then found that TRULI increased the representation of SIX2+ nephron progenitors in organoids at day 12 only when it temporally overlapped the 2 hr CHIR pulse at day 10, but not when it chased the CHIR pulse for 2 hr (Fig. 5C,D). However, YAP activation suppressed differentiation of cells to the JAG1+ renal vesicle state at day 12 in either case (Fig. 5D), which also translated to lower nephron formation at day 25 (Fig. 5E,F). Changes in nephron composition among the conditions were minimal (Fig. S10). These data indicate that the basal YAP state in nephron progenitor organoids permits Wnt-induced differentiation, while YAP activation converts the same Wnt signal into a renewal signal when the two temporally overlap. This interaction may relate to the paradoxical requirement of Wnt/β-catenin for both nephron progenitor renewal and differentiation⁸⁸, namely that a rhythmic YAP signaling context could enable the niche to alternately favor each state given a consistent Wnt/β-catenin input in vivo.

Fig. 5: YAP and RA status modify iPSC-derived nephron progenitor interpretation of Wnt/β-catenin differentiation signaling in ‘puck’ organoids.

(A) Schematic workflow for nephron progenitor organoid differentiation and perturbation. (B) Table of perturbation conditions in the indicated phases of the differentiation schematic. Conditions 2 and 3 stagger or overlap the YAP agonist TRULI (4 μM) with standard 7 μM CHIR-induced differentiation in condition 1, respectively. (C) Confocal immunofluorescence images of representative SIX2^EGFP iPSC-derived organoids at the day 12 endpoint for nephron progenitor (SIX2) and early nephron (JAG1) markers. (D) Top, Plots of marker expression as % of organoid area for two replicates, one for 300,000 cells per puck at day 10 (black markers, MAFB^BFP:GATA3^mCherry iPSC line) and one for 150,000 cells per puck (blue markers, SIX2^EGFP iPSC line), n > 2 pucks per condition per replicate, mean ± S.D. Bottom, schematic of experiment outcomes. (E) Confocal immunofluorescence images of representative MAFB^BFP:GATA3^mCherry iPSC-derived organoids at the day 25 endpoint for nephron markers. (F) Plot of nephron cell area fraction relative to organoid area at day 25 for organoids in (E), n > 5 pucks per condition, mean ± S.D. (G-J) Similar data for MAFB^BFP:GATA3^mCherry iPSC-derived organoids perturbed with the retinoic acid analogue TTNPB, n > 6 pucks per condition, mean ± S.D. (K) Left, Plot of nephron cell area fraction relative to organoid area at day 25 for organoids in (J). Right, plot and schematic of nephron composition by cell type. Arrows indicate increase in connecting segment (CnS) and proximal tubule cells in conditions treated with TTNPB. n > 6 pucks per condition, mean ± S.D. Statistics in D,F,I,K are one-way ANOVA, Tukey’s test, *p < 0.05, **p < 0.01, ***p < 0.001. (L) Schematic model for differential interpretation of Wnt input by nephron progenitors depending on branching life-cycle-associated rhythms in YAP and RA signaling.

While little is known about the role of retinoic acid after specification of nephron progenitors, microarray and qPCR analysis of RA-treated mouse embryonic kidney explants saw up-regulation in early nephron differentiation markers including Lhx1 and Wnt4 relative to untreated controls⁸⁹. Retinoic acid synthesis enzymes are also expressed in early rat kidney connecting segments and renal vesicles⁹⁰. Retinoic acid promotes proximal segment fate at the expense of distal segment in the zebrafish pronephros, but a similar role in the segmentation of mammalian nephrons has not yet been established^91,92. According to our GeoMx analysis, retinoic acid signaling activity is likely higher in nephron progenitors in pre-/mid-branching niches, overlapping the peak of Wnt/β-catenin activity (Note S3). Based on this we hypothesized that RA contributes to nephron progenitor differentiation in the pre-/mid-branching phase. This motivated us to test both overlapping and staggered Wnt and retinoic acid activation for effects on cell renewal/differentiation via CHIR and the retinoic acid analogue TTNPB^93,94 in iPSC-derived nephron progenitors (Fig. 5G). We verified that TTNPB increases retinoic acid target gene expression (RARB, CRABP2, CYP26A1) in these cells by qPCR (Fig. S9C). Staggered or overlapping TTNPB exposure around the 2 hr CHIR pulse at day 10 reduced the SIX2+ cell fraction at day 12 and restricted SIX2 expression to spatially discrete colonies within organoids (Fig. 5H,I). Neither treatment timing affected the fraction differentiated to the JAG1+ state at day 12 (Fig. 5I), reflecting either off-target differentiation of SIX2+ cells or transit through to later differentiation states. Nephron formation by day 25 was robust for both TTNPB treatment timings (Fig. 5J,K). GATA3+ connecting segments appeared and nephrons were notably more spatially discrete, with frequent formation of polarized tubules with correct anatomical connectivity of connecting segment-distal tubule-proximal tubule-podocyte cells (Fig. 5J,K). LTL+ proximal tubule segments were also observed more frequently, suggesting that retinoic acid indeed has a similar proximalizing effect here as in the zebrafish. These data indicate that retinoic acid signaling may increase progenitor exit from the SIX2+ state, restrict the spatial scale of Wnt-induced differentiation and promote an appropriate bias of fates during subsequent nephron segmentation. Together these data explain how a consistent Wnt/β-catenin signal in vivo may be differentially interpreted by nephron progenitors due to rhythmic YAP and retinoic acid signaling dynamics over the branching morphogenesis life-cycle, setting appropriate renewal vs. differentiation balance and segment bias.

Discussion

Our data together reveal that the nephrogenic niche alternates between renewal and differentiation phases in synchrony with the ureteric bud branching program local to each niche. We validated several hits at the protein level, especially downstream of Wnt signaling, that are crucial to nephron progenitor commitment to early nephron stages. The data provide molecular-level insight into our previous inference that nephrogenesis pauses early in each branching cycle²⁹. We also show that incorporation of rhythmic signaling in human iPSC-derived nephron progenitor organoids may aid synthetic control over renewal vs. differentiation decision-making in contexts removed from native branching morphogenesis, essential to the development of organoids and tissue chips for kidney regenerative medicine.

The data leave several questions open for future study. Firstly, are life-cycle-correlated transcriptional changes contributed by all nephron progenitors equally, or through transitions of progenitors between states? The cap mesenchyme niche has spatial structure, with more naive CITED1+/SIX2+ nephron progenitors sitting closer to the organ surface while primed CITED1−/SIX2+ progenitors reside on the medullary side of ureteric bud tips^23,31. Paradoxically, nephron progenitors migrate among these sub-compartments^32,33, such that the connection between commitment state and position in the niche is still being formalized²⁸. The new rhythmic ‘axis’ we uncovered adds another tool that could be used to parse single-cell-resolution spatial sequencing data to resolve this.

Second, what is the molecular nature of the pace-maker orchestrating the transcriptional rhythm? Several possibilities remain since the ureteric bud life-cycle accompanies changes in niche size¹, curvature³⁵, and mechanics²⁹, all of which could feed back into nephron progenitor decision making by modulating reciprocal morphogen signaling between niche cell compartments. The types and timing of signaling pathways revealed through spatial sequencing also have an intriguing overlap with those operating in the somitogenesis clock. Somites (future vertebral, skeletal muscle and cartilage cells) share morphological features with early nephrons – both involve periodic, fractional commitment of mesoderm-derived cells to discrete structures through mesenchymal to epithelial transitions. Discrete blocks of presomitic mesoderm cells condense in response to a molecular clock consisting of alternating cycles of Wnt and Notch/FGF-MAPK signaling^71,95. Our spatial sequencing data show similar alternating Wnt and Notch/FGF cycles in nephron progenitors, including several transcripts previously found to be rhythmically regulated in mouse presomitic mesoderm (e.g. Hes1, Hey1, Nkd1, Id1, Nrarp, Ptpn11, Bcl9l, Tnfrsf19, and Phlda1, see Fig. S11A)⁷¹. A somitogenesis gene set was also enriched in nephron progenitors from pre-branching vs. post-branching niches at p = 0.1 (Fig. S11B). These data suggest that shared developmental mechanisms may be at play. If so, the role of YAP as a mechanosensitive excitability threshold for oscillations⁹⁶ may be transferable to nephrogenesis, perhaps forming a link to our previous discovery of rhythmic mechanical stress in the niche synchronized with branching²⁹. The existence of a somitogenesis-like clock operating in nephron progenitors could be tested via lineage-specific reporter systems in explants or mouse primary cell/human iPSC-derived kidney organoids.

An alternative candidate for the pace-maker is suggested by the approximately 24 hr ureteric bud branch doubling time after ~E15, coincident with the emergence of a circadian transcriptional rhythm⁹⁷. However, we did not see a GSEA signature for circadian regulation of gene expression correlated with the branching life-cycle (minimum p = 0.65 for niche age category comparisons, Fig. S11C). Moreover, since branching is asynchronous among ureteric bud tips it is unlikely that each niche is separately entrained to a different circadian phase.

The rhythmic retinoic acid metabolism and stromal-like nephron progenitor signatures appearing in our data also warrant further investigation. Cyclic retinoic acid metabolism may affect autonomous signaling in nephron progenitors, act as a ‘filter’ that modulates known stroma-ureteric bud signaling⁹⁸, and/or create a gradient involved in somitogenesis-like signaling⁹⁹. Our spatial sequencing and protein-level validation data are consistent with either transient excursions of nephron progenitors through the stromal-like state during differentiation and/or (more likely) increased mixing between nephron progenitor and stromal cell compartments in post-/pre-branching phases. One possibility is that cell biophysical properties change during nephron progenitor priming to disfavor their association with the ureteric bud interface, stabilizing stromal interactions while nephron progenitors condense and undergo MET as they join new nephron aggregates. While requiring validation, this explanation is attractive given the rhythmic variation in cell-cell/cell-matrix adhesion-related transcripts in our cap mesenchyme GeoMx data including Tnc, Itga8, and Tgfbi^{36,47,48,50,51,100}. Alternatively or in conjunction, life-cycle-correlated intercalation of stromal cells via protrusion into the niche may confer contact-dependent pro-differentiation signaling in nephron progenitors^{20,23,38,52,62,101}.

Looking to the future, mimicking rhythmic niche signaling could serve as a strategy to achieve co-existence of nephron progenitor pools and differentiating nephrons, enabling multiple waves of nephrogenesis in a single differentiation experiment. This would ease the lack of ongoing nephrogenesis that has limited the use of human autologous stem-cell derived nephrons in regenerative medicine applications⁸.

Methods

Animal experiments.

Mouse protocols followed NIH guidelines and were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. E17 embryos were collected from timed pregnant CD-1 mice (Charles River) and stages confirmed by limb anatomy as previously described¹⁰². Embryonic kidneys were dissected in chilled Dulbecco’s phosphate buffered saline (DPBS, MT21-31-CV, Corning)¹⁰³.

GeoMx digital spatial profiling for whole transcriptome analysis (WTA).

The NanoString GeoMx Digital Spatial Profiler (DSP) enables spatially resolved RNA expression measurement³⁹. GeoMx DSP analysis was performed according to manufacturer protocols. Freshly dissected E17 mouse kidneys were partially flattened between two 1” x 3” standard microscope slides spaced apart using shims cut from 750 μm-thick plastic (Grainger) and held together using rubber bands during fixation for 2 hr in 4% paraformaldehyde (PFA). Kidneys were then stained with 20 μg ml⁻¹ AlexaFluor 488-labeled peanut (Arachis hypogaea) agglutinin lectin (PNA, L21409, Sigma) for 1 hr in RNAse-free PBS to mark ureteric bud tubules, followed by 3 x 10 min washes in RNAse-free PBS.

Kidneys were embedded in paraffin and sectioned at 5 μm thickness before mounting sections on a positively charged histology slide¹⁰⁴. Sections were deparaffinized and rehydrated by heating to 60°C for 15 min followed by 3 x 2 min washes with xylene, 2 x 2 min washes with 100% ethanol, and 1 min washes each in 95%, 80%, and 70% ethanol in deionized (DI) water, followed by 1 min wash in DI water. Antigen retrieval was then achieved by heating sections in 10 mM citrate buffer (pH 6) using a pressure cooker for 15 min, followed by blocking for 30 min at room temperature in 1x PBS + 0.1% BSA + 0.2% Triton X-100. Sections were then stained with anti-SIX2 antibody (1:600, 11562-1-AP, Proteintech, RRID: AB_2189084) in blocking buffer at 4°C overnight, washed 3 x 5 min in PBS + 0.1% Tween-20, incubated with AlexaFluor 647-conjugated donkey anti-rabbit secondary antibody (1:200, A31573, ThermoFisher, RRID: AB_2536183) in blocking buffer at room temperature for 1 hr, and washed 3 x 5 min in PBS + 0.1% Tween-20.

The tissue sections were then hybridized with Mouse WTA probes overnight at 37°C. Sections were subjected to 2 x 25 min washes in 1:1 4x SSC and formamide, and 2 x 5 min washes in 2x SSC. The slide was then loaded onto the GeoMx DSP and scanned. The regions of interest (ROIs) were manually selected using the polygon tool, guided by SIX2 staining. Oligos from the hybridized probes in the ROIs were released using photo-cleaving UV light and collected into a 96-well plate. After drying overnight, samples were resuspended in nuclease-free water. Library preparation was performed according to the manufacturer’s instructions. i5 and i7 indexes (unique dual indexes) as well as P5 and P7 sequences were added to the photocleaved oligonucleotides through PCR amplification. PCR products were pooled and purified with two rounds of AMPure XP beads (Beckman Coulter). The size and purity of the pooled library were assessed using a BioAnalyzer High Sensitivity DNA chip (Agilent technologies). The concentration of the library was measured using the Qubit 4 Fluorometer (Thermo Fisher Scientific). Paired end (2 x 27 bp reads) sequencing was performed on an Illumina NextSeq 2000 instrument using a P2-100 flow-cell with a 5% PhiX spike-in. Reads were aligned to barcodes as previously described³⁹.

Spatial sequencing data analysis.

GeoMx data were analyzed in R using code modified from Bioconductor¹⁰⁵ and NanoStringNCTools, GeomxTools, and GeoMxWorkflows libraries. An extra column containing the qualitative niche age annotations (‘pre’ = pre-branching, ‘mid’ = mid-branching, ‘post’ = post-branching) were appended to the sample annotation files associated with each analysis. DCC files for each ROI (a.k.a. ‘segment’), PKC, and modified sample annotation files were loaded and quality control performed by first filtering segments using the following parameters - minimum number of reads per segment (1000), minimum % of reads trimmed (80%), minimum % of reads stitched (80%), minimum % of reads aligned (75%), minimum sequencing saturation (50%), minimum negative control counts (1), maximum counts observed in no template control well (9000), minimum segment area (100). Ratio and global probe quality control was performed according to the vignette. Gene-level count data were generated and filtered to remove regions below LOQ. LOQ was defined as two geometric standard deviations above the geometric mean of negative control probes, or a minimum LOQ of 4. Segments having fewer than 20% of genes detected above LOQ were first filtered out, followed by genes detected in fewer than 20% of segments. Count data were then Q3 normalized before differential expression analysis using a linear model (modelDE.R, Roy Lardenoije) suited to comparing regions within the same slide. FDR values were determined via Benjamini-Hochberg multiple test correction.

The following analyses were also performed in R (RStudio v2023.06.1+524) using the indicated libraries: tSNE (Rtsne), unsupervised clustering of log₂-transformed normalized count data with ‘correlation’ option (pheatmap), gene expression and volcano plots (ggplot2, ggrepel).

Clustering and non-linear dimensional reduction was performed using Seurat¹⁰⁶. Briefly, raw GeoMx count data were loaded into R as a Seurat object and log-normalized using the NormalizeData function. The top 500 highly variable features were detected using the FindVariableFeatures function. Data were scaled to mean and variance per gene across ROIs of 0 and 1 respectively. Linear dimensional reduction was performed by principal component analysis using the RunPCA function, taking the first 20 principal components. K-nearest neighbor graph-based clustering with modularity optimization was then performed using FindNeighbors and FindClusters. tSNE was then performed using RunTSNE with a perplexity of 14 (based on 43 ROIs in our experiment).

Kidney immunofluorescence imaging.

Immunofluorescence staining and imaging was performed as previously described¹⁰⁷, using protocols adapted from Combes et al. and O’Brien et al.^36,108. Briefly, dissected kidneys were fixed in ice cold 4% paraformaldehyde in DPBS for 20 min, washed three times for 5 min per wash in ice cold DPBS, blocked for 2 hr at room temperature in PBSTX (DPBS + 0.1% Triton X- 100) containing 5% donkey serum (D9663, Sigma), incubated in primary and then secondary antibodies in blocking buffer for at least 48 hr at 4°C, alternating with 3 washes in PBSTX totaling 12-24 hours. The minimum duration of primary and secondary antibody incubations and washes depended on the age of the kidney, as previously described¹⁰⁸.

Primary antibodies and dilutions included rabbit anti-SIX2 (1:600, 11562-1-AP, Proteintech, RRID: AB_2189084), goat anti-ITGA8 (1:20, AF4076, R&D, RRID: AB_2296280), mouse anti-E-cadherin (1:50, clone 34, 610404, BD Biosciences, RRID: AB_397787), Goat anti-PDGFRA (1:500, AF1062, R&D Systems, RRID:AB_2236897), rabbit anti-phospho-Smad1/5 (1:800, 9516, Cell Signaling Technology, RRID:AB_491015), rabbit anti-Ki67 (0.5 μg/ml, ab15580, Abcam, RRID:AB_443209), mouse anti-MEIS1/2/3 (0.5 μg/ml, 39795, Active Motif, RRID:AB_2750570), rabbit anti-DHRS3 (1:500, 15393-1-AP, Proteintech, RRID:AB_2091855), rabbit anti-CELSR1 (1:500, ABT119, Millipore, RRID:AB_11215810), rabbit anti-Cyclin D1 (1:200, ab16663, Abcam, RRID:AB_443423), rabbit anti-MUCL1 (1:200, PA5-58653, ThermoFisher, RRID: AB_2644299), rabbit anti-RETSAT (1:350, PA5-65443, ThermoFisher, RRID: AB_2663683), rabbit anti-TTC28 (1:200, ab197076, Abcam), rabbit anti-PAX8 (1:250, 10336-1-AP, Proteintech, RRID: AB_2236705), rabbit anti-LEF1 (1:300, 2230, Cell Signaling Technology, RRID: AB_823558), rabbit anti-MRPS17 (1:200, 18881-1-AP, Proteintech, RRID: AB_10597844), rabbit anti-FGF1 (1:200, ab9588, Abcam, RRID: AB_308729), rabbit anti-LTBP1 (1:200, ab78294, Abcam, RRID:AB_1952060), and goat anti-JAG1 (1:150, AF599, R&D Systems, RRID: AB_2128257). Secondary antibodies (all raised in donkey) were used at 1:300 dilution and included anti-rabbit AlexaFluor 647 (A31573, ThermoFisher, RRID: AB_2536183), anti-rabbit AlexaFluor 555 (A31570, ThermoFisher, RRID: AB_2536180), anti-rabbit AlexaFluor 488 (A21206, ThermoFisher, RRID: AB_2535792), anti-mouse AlexaFluor 555 (A31572, ThermoFisher, RRID: AB_162543), and anti-goat AlexaFluor 488 (A11055, ThermoFisher, RRID: AB_2534102). In some experiments, samples were counterstained in 300 nM DAPI (4’,6-diamidino-2-phenylindole; D1306, ThermoFisher), and/or 20 μg ml⁻¹ AlexaFluor 488-labeled peanut (Arachis hypogaea) agglutinin lectin (PNA, L21409, Sigma) diluted in blocking buffer for 2 hours at room temperature, followed by 3 washes in PBS.

Kidneys were imaged in wells created with a 2 mm diameter biopsy punch in a ~5 mm-thick layer of 15:1 (base:crosslinker) polydimethylsiloxane (PDMS) elastomer (Sylgard 184, 2065622, Ellsworth Adhesives) set in 35 mm coverslip-bottom dishes (FD35-100, World Precision Instruments). Imaging was performed using a Nikon Ti2-E microscope equipped with a CSU-W1 spinning disk (Yokogawa), a white light LED, laser illumination (100 mW 405, 488, and 561 nm lasers and a 75 mW 640 nm laser), a Prime 95B back-illuminated sCMOS camera (Photometrics), motorized stage, 4x/0.2 NA, 10x/0.25 NA, 20x/0.5 NA, and 40x/0.9 NA lenses (Nikon), and a stagetop environmental enclosure (OkoLabs).

Average immunofluorescence was quantified for cap mesenchyme regions of interest segmented using Ilastik¹⁰⁹ v1.4.1b6. Segmented outlines were manually refined in FIJI as required.

HCR RNA-FISH.

E17 mouse embryonic kidneys dissected in RNase-free PBS were assayed by whole-mount hybridization chain reaction (HCR v3.0) single-molecule RNA fluorescence in-situ hybridization (smFISH) similar to published protocols^42,110-113. Probe sets were designed using in-house software (AnglerLite) and are provided in Supplementary Files). Buffers and hairpin solutions were procured from Molecular Instruments, Inc. Kidneys were transferred to 1.5 ml Eppendorf tubes and fixed in 4% PFA overnight at 4°C. Kidneys were washed 3 x 5 min in PBST and dehydrated on ice by serial 10 min incubations in 25%, 50%, 75%, and 100% methanol:PBST washes before incubation at −20°C for > 16 hr. Kidneys were rehydrated on ice by serial 10 min incubations in 25%, 50%, 75%, and 100% PBST:methanol, followed by one 10 min PBST wash at room temperature. Kidneys were further permeabilized for 1 hr in 2% SDS, 10 μg/ml proteinase K in PBST. Kidneys were washed 3 x 5 min in PBST, post-fixed 20 min in 4% PFA at room temperature, washed 5 x 5 min in PBST, and incubated in probe hybridization buffer for 5 min. Kidneys were pre-hybridized with pre-warmed probe hybridization buffer for 30 min at 37°C. 100 μl of 16 nM probe solution was then prepared from a 1 μM stock and pre-warmed to 37°C for 15 min before probing kidneys overnight (>12 hr) at 37°C. Probe wash buffer was pre-heated to 37°C and used to wash kidneys for 4 x 30 min at 37°C, followed by 2 x washes with 5x sodium chloride-sodium citrate buffer tween buffer (SSCT) at room temperature. Amplification buffer was equilibrated at room temperature while 2 μl of 3 μM stocks of hairpins 1 and 2 were snap-cooled from 95°C for 90s to 20°C for 30 min in the dark. Kidneys were incubated with pre-warmed amplification buffer for 30 min at room temperature before incubating with 100 μl amplification buffer containing 50x dilutions of each hairpin at room temperature overnight (12-16 hr). Excess hairpins were removed by washing in 5x SSCT at room temperature 2 x 5 min, 2 x 20 min, and 1 x 5 min, followed by 3 x 5 min washes in PBST.

Kidneys were imaged by confocal fluorescence microscopy at 20x and 40x as for antibody assays (see Kidney immunofluorescence imaging.)

Human iPSC-derived nephron progenitors.

Nephron progenitor cell ‘puck’ organoids were generated from SIX2^EGFP transgenic reporter iPSC line (SIX2-T2A-EGFP, Murdoch Children’s Research Institute / Kidney Translational Research Center, Washington University Nephrology) and MAFB^BFP;GATA3^mCherry transgenic reporter iPSC line¹¹⁴ according to published protocols^78,79,115. Briefly, iPSCs were maintained in standard tissue-culture treated 6-well plates in stem cell maintenance medium plus supplement (mTeSR+ kit, StemCell Technologies 100-0276). Cells were passaged using Accutase (StemCell Technologies, 07920) and plated at a density of 5,200 cells/cm². Differentiation was induced with TeSR-E6 Medium plus supplements (TeSR-E6+) (STEMCELL Technologies 05946) and 7μM CHIR 99021 (Tocris, 4423) the following day, and media was refreshed every other day. On day 5, wells were washed with DPBS and media was changed to TeSR-E6+, 1 μg/ml heparin (Sigma-Aldrich H4784) and 200 ng/ml FGF9 (R&D Systems 273-F9-025). Media was refreshed every day. On day 10, cells were lifted using Accutase and TeSR-E6+ was used to quench digestion. Cells were resuspended in TeSR-E6+ and counted, centrifuged at 300g for 3 min and resuspended in the residual media volume remaining after aspiration of TeSR-E6+ to create a dense cell slurry at ~0.3 x 10⁶ cells/μl. 1 μl aliquots of the slurry were then spotted onto 6-well 0.4 μm polyester transwell membranes (CellTreat, 230607) overlying TeSR-E6+ and 7μM CHIR 99021 ‘pulse’ media for 2 hr, which was then exchanged for ‘chase’ media consisting of TeSR-E6+, 1 μg/ml heparin, and 200 ng/ml FGF9. For select conditions, the CHIR pulse was combined (in pulse media) or staggered (in chase media) with 0.1 μM TTNPB (Tocris 0761), 4 μM TRULI (MedChem Express HY-138489 or Sigma-Aldrich SML3634), or equivolume DMSO. In these cases, wells were then washed with DPBS before changing to chase media. Media was refreshed the following day. On day 12, transwells were exchanged with DPBS and media switched to TeSRE6+. Media was refreshed every other day until reaching the experiment endpoint.

Organoid immunofluorescence.

Organoids were fixed in 4% paraformaldehyde for 15 min. Fixed cells were then washed twice for 10 min in PBSG (DPBS + 7.5 g/L glycine) and once in DPBS. Organoids were permeabilized in 0.5% Triton-X-100 for 30 min at room temperature and blocked for 1hr at room temperature in IF Wash (DPBS + 1g/l Bovine Serum Albumin + 0.2% Triton-X-100 + 0.04% Tween-20) + 10% donkey serum. Organoids were then incubated overnight at 4°C in appropriate dilution of primary antibody in IF Wash + 10% donkey serum, washed three times for 1 hour in IF wash at room temperature, and incubated in secondary antibody at overnight at 4°C in IF wash + 10% donkey serum. Organoids were again washed three times in IF wash for 1 hour. Organoids were counterstained in 1 μg/ml DAPI in DPBS and washed once in DPBS before imaging using confocal microscopy (see Kidney immunofluorescence imaging). Primary antibodies and dilutions were rabbit anti-SIX2 (1:400, 11562-1-AP, Proteintech, RRID: AB_2189084), rabbit anti-ECAD (1:300, 3195, Cell Signaling Technology, RRID: AB_2291471), goat anti-JAG1 (1:150, AF599, R&D Systems, RRID: AB_2128257), goat anti-GATA3 (1:20, AF2605, R&D Systems, RRID: AB_2108571), sheep anti-NPHS1 (1:40, AF4269, R&D Systems, RRID: AB_2154851), and biotinylated LTL (1:300, B-1325-2, Vector Laboratories). Donkey secondary antibodies were used at 1:300 dilution: anti-rabbit AlexaFluor 647 (A31573, ThermoFisher, RRID: AB_2536183), anti-goat AlexaFluor 488 (A11055, ThermoFisher, RRID: AB_2534102), anti-sheep AlexaFluor 555 (A21436, ThermoFisher, RRID: AB_2535857), and DyLight 405-Streptavidin (016-470-084, Jackson ImmunoResearch).

For day 12 organoids, nephron progenitor (SIX2+) and early nephron (JAG1+) % area relative to DAPI+ area was determined by thresholding z-stack montages in FIJI. For day 25 organoids, glomeruli (NPHS1+) % area was determined by thresholding z-stack montages while proximal tubule (LTL+) % area, distal tubule (only ECAD+) % area, and connecting segment (nuclear GATA3+ and ECAD+) % area relative to organoid area (determined by manual annotation of background signal) were determined by manual annotation of z-stack montages followed by thresholding. Analyses were performed on a single, central z-plane for each organoid.

Organoid RNA isolation and quantitative PCR.

Organoids were collected 8 hours after seeding for RNA extraction with 8 organoids pooled for each condition. Following the manufacturer protocols, we used the RNeasy Mini Kit (Qiagen 74104) to isolate RNA and the High-Capacity RNA-to-cDNA Kit (Applied Biosystems 4387406) to generate cDNA libraries. For qPCR, we used the PowerUp SYBR Green Master Mix (Applied Biosystems A25742). We used an Applied Biosystems 7300 thermocycler set to manufacturer suggested times and temperatures for PowerUp SYBR Green Master Mix. 25 ng of cDNA were used in each reaction along with the appropriate qPCR primers at 500 nM. qPCR primers for CRABP2, CTGF, CYP26A1, CYR61, GAPDH, HPRT, and RARB were used (Table S2). DeltaDeltaCt values were calculated using either GAPDH or HPRT as housekeeping genes. Samples were run in triplicate for qPCR analysis.

scRNA-seq data analysis.

Gene expression matrices generated from scRNA-seq of dissociated E15.5 mouse embryonic kidneys in Lawlor et al. were used³². These matrices were accessed from the Gene Expression Omnibus (GEO, NCBI) under accession code GSE118486 and sample mk1 was used for further analysis. Analysis was done similarly to Lawlor et al. In short, the Seurat library^116,117 was used to import the gene expression matrix to R and for downstream analyses. Cells with less than 200 genes expressed or with greater than 7.5% mitochondrial gene expression were removed. Genes expressed in less than 3 cells were removed from the dataset. The filtered dataset contained 2,708 cells with an average of 2,441 unique genes detected per cell. Cell cycle scoring was done using the CellCycleScoring function. Scaled data matrices were then generated using the ScaleData function. Shared nearest neighbor (SNN) clustering was performed for whole kidney data using resolution 0.5 for the first 15 principal components calculated from a set of 2,000 variable genes. Differentially expressed marker genes were identified using a Wilcoxon rank sum test and compared to the published list of marker genes from Lawlor et al.³² to assign cell identity. SNN clustering was then performed on cells belonging to the nephron lineage clusters using resolution 0.5 for the first 12 principal components calculated from a set of 2,000 variable genes. Again, differentially expressed marker genes were identified in Seurat using Wilcoxon rank sum test and compared to a list of marker genes for distinct nephron lineage clusters published by Lawlor et al.³² to assign cell identity. Fold-change and Wilcoxon rank sum p-value for individual genes in volcano plots were calculated by comparing the committing cluster to the progenitor cell clusters. A list of 171 highly upregulated genes in the committing cluster compared to the progenitor cell cluster was generated by taking genes with log₂(fold-change) of at least 0.25 and Bonferroni corrected adjusted Wilcoxon rank sum test p-value less than 0.05 between the clusters.

To gain higher resolution in cell types, gene expression matrices generated from scRNA-seq of dissociated E18.5 mouse embryonic kidneys in Combes et al. were used⁴⁷. These matrices were accessed from the Gene Expression Omnibus (GEO, NCBI) under accession code GSE108291 and samples E18.5 kidney 1 through 3 were used for further analysis. Analysis was done similarly to Combes et al. In short, the Seurat library^116,117 was used to import the gene expression matrix to R and for downstream analyses. Cells with less than 200 genes expressed, greater than 95% of genes with zero assigned reads, greater than 4,500 transcript counts, or greater than 6.5% mitochondrial gene expression were removed. Genes expressed in less than 3 cells, mitochondrial and ribosomal genes, and genes without annotations were removed from the dataset. The filtered dataset consisted of 5,494 cells and 15,747 genes, with an average of 2,383 unique genes detected per cell. Cell cycle scoring was done using the CellCycleScoring function, and the filtered dataset was normalized regressing out factors related to cell cycle using the NormalizeData function. Scaled data matrices were then generated using the ScaleData function. Shared nearest neighbor (SNN) clustering was performed for whole kidney data using resolution 0.8 for the first 30 principal components calculated from a set of 2,000 variable genes. Differentially expressed marker genes were identified using a Wilcoxon rank sum test and compared to the published list of marker genes from Combes et al.⁴⁷ to assign cluster identity. Clusters with similar marker profiles were combined for plotting, achieving a similar set to those reported by Combes. However, we did not recover a unique K14 NP-STR cluster and we recovered the extra clusters K1,2,4 STR (with stromal markers that overlapped K1 STR-CD, K2 STR-CS, and K4 STR-MS), and K3,9 UE/CnS (with markers that overlapped K3 DN and K9 UE). SNN re-clustering was then performed on cells belonging to the nephron lineage clusters (K0 NP, K3 DN, K6 SSB, K7 RV, K5 EPT, K8 PT, K3,9 CnS, K12 PTA, K15 Pod) using resolution 0.8 for the first 35 principal components calculated from a set of 2,000 variable genes. Again, differentially expressed marker genes were identified in Seurat using Wilcoxon rank sum test and compared to a list of marker genes for distinct nephron lineage clusters published by Combes et al.⁴⁷ to assign cell identity. Again, we recovered a similar cluster set to those reported by Combes. However, we did not recover a unique N10 NP-STR cluster (instead we found this was a sub-cluster of N6 NP based on Col3a1 expression, which was identified as a marker in Combes), or a unique N12 CnS cluster (we found this to be combined with N8 distal SSB in our analysis).

Gene Set Enrichment Analysis (GSEA).

Gene set enrichment plots were generated using GSEA desktop software⁴⁶ from Q3-normalized GeoMx gene count data. Gene sets consisted of the above differentially expressed set for committing:naive nephron progenitors (see Fig. 3D) and others available in the Broad Institute’s Molecular Signatures Database (MSigDB, Fig. 3C). FDR q-values < 0.25 were considered to be significant per the GSEA user guide¹¹⁸.

Rendering and graphics.

3D reconstruction of E14 kidney in Fig. 1 was generated using ImarisViewer v10.0.1 (Oxford Instruments). Insets were generated by exporting surface .stl files from FIJI 3D viewer and rendering in Rhino 7 (Robert McNeel & Associates). Cartoon schematics with 3D effects were generated in Rhino using Grasshopper and Kangaroo (Daniel Piker) packages.

Statistical analysis.

One-way analysis of variance (ANOVA) with correction for multiple comparisons using Tukey’s honestly significant difference test was performed in MATLAB using anova1.m and multcompare.m functions.