Biomimetic fractal topography enhances podocyte maturation in vitro

Chuan Liu; Praful Aggarwal; Karl T Wagner; Shira Landau; Teng Cui; Xin Song; Laleh Shamaei; Naimeh Rafatian; Yimu Zhao; Sonia Rodriguez-Ramirez; Keith Morton; Elizabeth Virlee; Chen Yu Li; Dawn Bannerman; Simon Pascual-Gil; Sargol Okhovatian; Anastasia Radisic; Sergi Clotet-Freixas; Teodor Veres; Mohtada Sadrzadeh; Tobin Filleter; Ulrich Broeckel; Ana Konvalinka; Milica Radisic

doi:10.1038/s41467-025-66037-8

. 2025 Dec 15;16:11116. doi: 10.1038/s41467-025-66037-8

Biomimetic fractal topography enhances podocyte maturation in vitro

Chuan Liu ^1,², Praful Aggarwal ³, Karl T Wagner ^2,⁴, Shira Landau ^1,², Teng Cui ⁵, Xin Song ⁴, Laleh Shamaei ⁶, Naimeh Rafatian ¹, Yimu Zhao ^2,⁷, Sonia Rodriguez-Ramirez ^8,⁹, Keith Morton ^10,¹¹, Elizabeth Virlee ³, Chen Yu Li ⁴, Dawn Bannerman ^2,⁴, Simon Pascual-Gil ², Sargol Okhovatian ^1,², Anastasia Radisic ¹, Sergi Clotet-Freixas ^2,^9,¹², Teodor Veres ^5,^10,¹¹, Mohtada Sadrzadeh ⁶, Tobin Filleter ⁵, Ulrich Broeckel ³, Ana Konvalinka ^2,^8,^9,^13,¹⁴, Milica Radisic ^1,^2,^4,^7,^11,^13,^15,^16,^✉

PMCID: PMC12706010 PMID: 41398144

Abstract

In their native environment, tissues are organized into intricate fractal structures, rarely recapitulated in their culture in vitro. The extent to which fractal (self-similar) patterns that resemble complex topography in vivo influence cell maturation remain inadequately elucidated. Yet, the application of fractal topographical stimulation may address the challenge of improving the differentiated cell phenotype in vitro. Here, we show fractality in the kidney glomerulus and podocytes, branching highly differentiated cells within the glomerulus. Biomimetic fractal patterns derived from glomerular histology are used to generate topographical (2.5-D) substrates for cell culture. Podocytes grown on fractal topography exhibit higher expression of functional markers and enhanced cell polarity. RNA sequencing suggests podocytes’ enhanced ECM deposition and remodeling on fractal versus flat topography, and enhanced maturation accompanied by stress on fractal versus non-fractal topography. The incorporation of fractal topography into standard well plates may serve as a user-friendly bioengineered platform for high-fidelity cell culture.

Subject terms: Tissue engineering, Bioinspired materials, Tissues

Liu and colleagues demonstrate that biomimetic fractal patterns derived from glomerular histology can enhance the maturation of podocytes (highly differentiated glomerular cells) that are grown in culture. This work presents a bioengineered platform for improved cell culture fidelity.

Introduction

Podocytes are highly specialized, terminally differentiated epithelial cells in the kidney glomerulus that cover the outer surfaces of glomerular capillaries and form a selective barrier for blood filtration. They possess multiscale foot processes branching from the main cell body, which compose slit diaphragms that serve as a filtration barrier. During development, podocytes’ morphology gradually becomes increasingly more complex as their foot processes turn from a primitive to a mature state¹. This intricate, branching morphology is an important indicator of podocyte health and function, as the deterioration of this interdigitating pattern, referred to as foot process effacement, leads to glomerular disease characterized by significant proteinuria.

Given the definition of fractals as geometry with recurring self-similar patterns at continuously smaller scales, the branching of foot processes at multiple levels/scales suggests that podocytes may be fractal objects. As such, their morphology could be characterized by fractal dimension (Df), a ratio of the change in detail to the change in scale, used for quantifying complex geometry. Similar to podocytes, a kidney glomerulus has a complex architecture, with fractal analysis performed on itself² or its components³ under healthy and pathological/injury conditions. Not surprisingly, the state of tissue health has been reported to correlate with its fractality⁴.

It continues to pose a difficulty in recreating physiologically pertinent podocytes⁵. In addition to gene expression diverging considerably from the native conditions and phenotypes exhibiting immaturity, there is a lack of functional readouts that allow for injury modeling. Various methods have been developed to help mature cultured podocytes or generate more advanced glomerulus models in vitro, including biochemical cues⁶, organoid differentiation⁷, decoration of substrates with extracellular matrix (ECM)⁸, cyclic mechanical stimulation⁹, and flow with vascularization^9–11. However, mimicking the complex glomerular architecture has not yet been accomplished. Flat surfaces or membranes are utilized for cell and tissue culture, even in microfluidic systems. On the other hand, it is well known that topographical cues play an important role in cell behavior. Specifically, simple round-shaped topography or microcurvature has been shown to improve the fidelity of podocytes cultured in vitro^11,12. Building upon the idea of shape stimulation^13–15, we were interested in investigating whether more biomimetic patterns could further enhance podocyte maturation.

In this work, we delineate and establish fractality of podocyte and glomerulus structure. We explore the use of glomerulus-mimicking fractal topography to support higher fidelity podocyte culture in vitro by generating topographical (2.5-D) cell culture substrates covered with biomimetic fractal patterns derived from glomerular histology. We demonstrate that podocytes grown on fractal topography express higher levels of functional markers and exhibit enhanced cell polarity. To track morphological complexities of differentiated podocytes, we employ a fluorescent labeling assay where labeled individual cells are tracked within an otherwise optically silent confluent cell monolayer to reveal single cell borders, which in turn supports drug testing of podocytes cultured on fractal topography, upon incorporation of fractal patterns into standard tissue-culture well plates. RNA sequencing (RNA-seq) suggests enhanced ECM deposition and remodeling in podocytes grown on fractal substrates versus flat surfaces, and enhanced maturation accompanied by stress in podocytes grown on fractal versus non-fractal topography. Lastly, we demonstrate the involvement of Yes-associated protein (YAP) in podocyte shape control on fractal substrates via the inhibition of YAP signaling.

Results

Native glomerulus and podocyte demonstrate fractal patterns

A typical fractal shape shows self-similarity over multiple scales of measurement, such as the famous Koch snowflake, whose size expands 4 times as the scale decreases by one-third, corresponding to a fractal dimension of 1.26 (Fig. 1a). Both the glomerulus and podocyte demonstrate a complex structure under X-ray nanotomography and focused ion beam/scanning electron microscopy (SEM), respectively, suggesting fractal dimension as a suitable index for the quantification of their geometry (Fig. 1b). Fractal analysis of histology images of glomeruli from healthy and pathological mouse (Fig. 1c) and human samples (Supplementary Fig. 1) demonstrated the existence of fractality and a decrease in fractal dimension between healthy and diseased states.

Previously, histological glomerular tracing was developed^13–15 to obtain the outline of a glomerulus from histology images. To confirm the occurrence of fractality, we compared the planar projection of an SEM image of a podocyte¹ (Fig. 1d, e) and a glomerular trace derived from refs. ^13–15 (Fig. 1d, e) to objects that are widely accepted to be fractal, specifically a branching tree and Koch snowflake (Fig. 1d, e). We contrasted these to the objects that are known not to be fractal, specifically the perfect circle and the straight line (Fig. 1d, e). Relying on the box-counting method we determined that both the mature podocyte and the outline of a glomerulus exhibit multifractality, similar to that of a branching tree and of the Koch snowflake (Fig. 1f). Whereas the non-fractal objects, a circle and a line, exhibited a Df = 1 regardless of the box size, the podocyte (Df ranging from 1.56 to 1.81) and the glomerulus (Df ranging from 1.38 to 1.73) exhibited non-integer Df values similar to those observed for the tree (Df from 1.70 to 1.87) and the Koch snowflake (Df from 1.28 to 1.37).

Importantly, whereas dominant frequencies were easily identifiable for the circle and the line (Fig. 1g–i) upon fast Fourier transform (FFT), no dominant frequency was apparent for the podocyte, the glomerulus, a tree and the Koch snowflake (Fig. 1g–i). Together with the fractal analysis (Fig. 1c–f), these results further confirm the presence of fractality both in the adult, mature podocyte shape, as well as the histological outline of the glomerulus. In the remainder of the manuscript, we focus on a single Df value to balance the experimental feasibility of constructing physical scaffolds of clearly delineated patterning categories with the biology of the cells cultured thereon.

To investigate the effect of fractal topography on podocyte culture in vitro, we used topographical cell culture substrates with low and high biomimetic fractal patterns derived from the podocyte microenvironment observed in histology slices from pathological and healthy samples according to refs. ^13–15 and as shown in Fig. 1c, j (mouse) and Supplementary Fig. 1 (mouse and human). For fractal substrate fabrication, an outline of a glomerulus was used to create a design unit, which was then packed across a 2D surface to form a patterned area of a desirable size (Fig. 1j). The fractal patterns were transferred to a master mold using direct laser writing to ultimately fabricate polydimethylsiloxane (PDMS) topographical inserts in a standard 24-well plate (Fig. 1j). To demonstrate generalizability of the approach, we used conditionally immortalized mouse podocytes as a main cell type fully characterized by immunofluorescence staining, SEM, Western blotting and RNA sequencing (Figs. 2–7, Supplementary Figs. 2–4 and 10–12), with select assays confirmed by the non-conditionally immortalized human podocytes (Supplementary Figs. 6–9, Supplementary Fig. 15) and primary fetal human podocytes (Supplementary Figs. 13–15, Fig. 1j).

Fig. 2 — a Scanning electron micrographs (SEM) and 3D display of digital micrographs of PDMS scaffolds showing successful transfer of non-topographic (NT), round non-fractal topographic (RT), and biomimetic low and high fractal (LF and HF) topographic patterns to substrates. b Fractal dimension (Df) of RT, LF, and HF patterns from 3 topographical regions on n = 3 scaffold samples. Kruskal–Wallis test with Dunn’s multiple comparisons test performed. Data plotted as mean ± SD. c Relative gene expression of *Nphs1* to *Gapdh* in mouse podocytes cultivated on n = 5 NT and n = 4 HF scaffolds. Data normalized to NT. Student’s t test (two-sided) for statistical testing. Data plotted as mean ± SD. d Immunostaining of slit diaphragm protein podocin (green) counterstained with WGA (red) and DAPI (blue) from mouse podocytes grown on NT, RT, LF and HF substrates and differentiated for 3 and 14 days, showing expression of podocin increases on fractal substrates with time in culture. Maximum intensity projection of immunofluorescence images shown. e Quantification of podocin immunofluorescence signal from mouse podocytes grown on NT, RT, LF and HF substrates and differentiated for 3 and 14 days. n = 3 scaffold samples. Data normalized to day 3 NT. Two-way ANOVA with Tukey’s multiple comparisons test performed. Data plotted as mean ± SD. WGA wheat germ agglutinin, DAPI 4’,6-diamidino-2-phenylindole, NT no topography (non-fractal non-topographic control group), RT round topography (non-fractal topographic control group), LF low fractal topography (biomimetic low fractal topographic group derived from pathological histological glomerular tracing), HF high fractal topography (biomimetic high fractal topographic group derived from healthy histological glomerular tracing). Source data are provided as a Source Data file.

Fig. 7 — a Hierarchical heatmap of unsupervised clustering of the genes belonging to the YAP interactome, provided the genes were differentially expressed, in the HF, RT and NT groups. RNA-seq data (GSE298416) from E11 mouse podocytes after 14 days of differentiation on 5 NT, 6 RT, 6 HF substrates were used for analysis. b, c YAP localization in the nucleus vs. cytoplasm in mouse podocytes cultured on NT and HF scaffolds (n = 3). b Localization analysis of YAP immunofluorescence signal in c via the ratio of mean intensity of nucleic vs. cytoplasmic signals. Student’s t-test for b. c Confocal microscope images (maximum intensity projection in z-direction) of Mosaic mouse podocytes cultured on NT and HF scaffolds and stained for YAP (green) and DAPI (blue). d–g Mosaic mouse podocytes differentiated on d, e NT and f, g HF scaffolds (n = 3) for 14 days, followed by treatment with YAP inhibitor verteporfin at 0 µM and 2 µM for 2 or 4 days. Confocal microscope images (maximum intensity projection in z-direction) of Mosaic mouse podocytes on d NT and f HF scaffolds with and without YAP inhibition. Lactate dehydrogenase (LDH) release of Mosaic mouse podocytes on e NT and g HF scaffolds (each represented by a dot) was inhibited with verteporfin at 0 µM and 2 µM for 2 days or 4 days. Quantification of cell morphology (Df, aspect ratio) and cell size (normalized to control) of Mosaic mouse podocytes on e NT and g HF scaffolds treated with verteporfin at 0 µM and 2 µM for 2 or 4 days. Each dot represents an individual mCherry-labeled podocyte (29–52 cells per group, see Source Data file). Student’s t-test, Welch’s t-test, or Mann–Whitney test for (e, g). Two-sided statistical tests for (b, e, g). Data plotted as mean ± SD in (b, e, g). Df fractal dimension, DAPI 4’,6-diamidino-2-phenylindole, NT no topography (non-fractal non-topographic control group), RT round topography (non-fractal topographic control group), HF high fractal topography (biomimetic high fractal topographic group derived from healthy histological glomerular tracing). Source data are provided as a Source Data file.

Fractal topography increases expression of slit diaphragm markers

To test the hypothesis that fractal topographical cues could mediate assembly of podocyte structures, four substrate groups were used^13–15: the non-fractal non-topographic control (NT), the non-fractal round topographic control (RT), the biomimetic low (LF) and high (HF) fractal topographic groups described above, each with a diameter of ~150 µm, resembling a glomerular slice. The surface features of the substrates from these different groups can be clearly visualized under SEM and 3D digital microscopy (Fig. 2a), with an increasing Df from the non-fractal RT control group to the fractal LF and HF groups as expected (Fig. 2b).

A conditionally immortalized mouse podocyte cell line, E11, is commonly used for studying podocyte biology since human podocyte cells are scarce. E11 podocytes have a thermoswitching mechanism that could deactivate the forced cell cycle progression and switch from a proliferation to a differentiation mode. A cultivation period of at least 14 days is commonly used and regarded as required for sufficient differentiation. As such, we selected day 3 and day 14 as the early (immature) and late (mature) endpoints, respectively.

Next, we evaluated the expression of a slit diaphragm marker, podocin, from podocytes cultured on flat (NT) and topographical (RT, LF, HF) substrates over time. Due to the use of a conditionally immortalized mouse podocyte cell line, the cells grown on the NT substrates demonstrated increased coverage as a result of faster proliferation, compared to the cells grown on topographical substrates (RT, LF, HF) (Fig. 2d). This proliferative behavior is not consistent with the podocyte behavior in vivo, where adult mature podocytes should not proliferate. Besides less proliferation, the cells grown on fractal topography (LF, HF) resembled the expected in vivo phenotype more closely, as shown by the increased expression of podocin, a podocyte-specific marker essential for the slit diaphragm formation. Although in early cultures (Day 3) HF substrates did not appear to have substantial podocin expression (Fig. 2d, e), as the culture time increased, fractal substrates exhibited a clear increase in podocin expression over flat substrates, whereas cells on flat substrates exhibited low levels of podocin expression, and no significant improvement in podocin expression between Day 3 and Day 14 (Fig. 2d, e). Both fractal substrates (LF and HF) had greater podocin expression than the non-fractal topographical control (RT) on Day 14 (Fig. 2d, e), suggesting the potential role of biomimetic fractal patterns in stimulating E11 mouse podocytes’ differentiation to a more mature state. We additionally assessed the gene expression of another slit diaphragm marker, nephrin (Nphs1), from E11 podocytes grown on NT and HF substrates by quantitative polymerase chain reaction (qPCR) and found a significantly enhanced expression on HF compared to the NT control (Fig. 2c).

Fractal topography elicits enhanced signs of cell polarity

Apical surfaces of epithelial cells are decorated with glycans that contribute to the regulation of cell structure, function, and communication via protruding forms such as charged villi, blebs, and secreted extracellular vesicles (EVs)¹⁶. We hypothesized that the arrangement of ECM proteins, facilitated by fractal topographical cues, could mediate the assembly of podocyte structures involved in polarization. SEM underscored the interdigitated 3D morphology of cells on fractal substrates vs NT and RT controls (Fig. 3a). The protruding apical structures appeared to progress from rounded (mushroom regime) in the NT group to elongated finger-like protrusions (brush regime) in the HF group, according to the glycocalyx development model (Fig. 3a)¹⁶.

Fig. 3 — a SEM of mouse podocytes cultivated on NT, RT, LF and HF scaffolds (n = 3) from low to high magnifications. Arrows point to finger-like apical protrusions. Schematic of a glycocalyx development model¹⁶ created in BioRender⁸². b Immunofluorescence images (orthogonal projection) of mouse podocytes cultivated on NT, RT, LF and HF scaffolds (n = 3) showing podocin (green) presence in z-stacks. c Western blot of podocalyxin (PODXL) from mouse podocytes cultivated on NT, RT, LF and HF scaffolds. d Quantification of PODXL signal from western blot (normalized to GAPDH) from mouse podocytes cultivated on NT, RT, LF and HF scaffolds (n = 3). e Western blot of collagen type IV alpha 4 chain (COL4A4) from mouse podocytes cultivated on NT, RT, LF and HF substrates. f Quantification of COL4A4 signal from western blot (normalized to GAPDH) from mouse podocytes cultivated on NT, RT, LF and HF scaffolds (n = 6). g TEM of EVs isolated from culture media of mouse podocytes cultivated on NT and HF scaffolds. h Western blot of ALIX and CD63 from EV samples isolated from culture media of mouse podocytes cultivated on NT and HF scaffolds. i, j Quantification of i ALIX and j CD63 signal from western blot (normalized to total protein) from EV samples isolated from culture media of mouse podocytes cultivated on NT and HF scaffolds. n = 3, each pooled from media collected from 8 scaffolds, for (g, i, j). One-way ANOVA for (d, f). Student’s t test (two-sided) for (i, j). Data plotted as mean ± SD for (d, f, i, j). f, i, j Two blots containing the samples derived from the same experiment were processed in parallel and used for quantification. SEM scanning electron micrograph, TEM transmission electron micrograph, EV extracellular vesicle, WGA wheat germ agglutinin, DAPI 4’,6-diamidino-2-phenylindole, NT no topography (non-fractal non-topographic control group), RT round topography (non-fractal topographic control group), LF low fractal topography (biomimetic low fractal topographic group derived from pathological histological glomerular tracing), HF high fractal topography (biomimetic high fractal topographic group derived from healthy histological glomerular tracing). Source data are provided as a Source Data file.

Fractal topography was critical for podocyte marker localization, as confocal z-stacks demonstrated greater podocin localization on top of the topography (Fig. 3b) and its enhanced presence in the fractal groups (Supplementary Fig. 2). Podocalyxin, a negatively charged sialylated glycoprotein involved in polarization¹⁷ and basal organization¹⁸, exhibited stable protein expression and an increasing trend in the HF group (Fig. 3c, d, Supplementary Fig. 3a). Western blotting validated the presence of an important glomerular basement membrane protein expressed by mature podocytes, COL4A4 (Fig. 3e, f, Supplementary Fig. 3b).

Epithelial polarity is often associated with a dense glycocalyx^18,19, and a dense glycocalyx is likewise associated with the budding of EVs¹⁶. TEM revealed the cup-like shape of EV’s (Fig. 3g) and expression of EV markers ALIX and CD63 (Fig. 3h–j, Supplementary Fig. 3c, d). Collectively, these observations point to the physiological relevance of podocyte apical surface morphology on fractal substrates^16,20.

Biomimetic fractal topography influences cytoskeleton arrangement

As it is known that the wavelength of a substrate impacts biological responses^21,22 and the sharp microcurvature of the multiscale-curved HF pattern is not well represented by the RT control, we designed a non-fractal ordered “spiky” topography (ST) that aimed to capture the pointy features at the tips of the HF pattern (Fig. 4a, b). To ensure that all conditions remained the same, we fabricated the ST and HF master molds on the same wafer. The resulting substrates have similar feature heights (Fig. 4c–e), with the peaks in the radius of curvature distribution overlapping at approximately 2μm (Fig. 4f).

Fig. 4 — a, b Low (left) and high (right) magnification digital microscope images of a ST and b HF scaffolds (n = 3). c, d Surface characterization of c ST and d HF scaffolds. e, f Feature e height and f radius of curvature measured from different regions on 3 ST (15 regions) and 3 HF (18 regions) scaffolds. The box-plot elements in e are defined as: center line, median; box limits, upper and lower quartiles; whiskers, between the minimum and the maximum values; dots, each individual height value from a region plotted as a dot superimposed on the graph. Mann-Whitney test (two-sided) performed for (e). g, h Immunofluorescence images (maximum intensity projection in z-direction) of mouse podocytes cultured on g 3 ST and h 4 HF scaffolds and stained for nephrin (red), nestin (orange), f-actin (green) and DAPI (blue). i–l Quantification and directionality analysis of immunofluorescence signals from i DAPI (nuclei number and size), j nephrin (intensity and directionality), k nestin (intensity and directionality), and l f-actin (intensity and directionality) within the topographical ST and HF pattern units (as shown by the right image in (a, b) selected by enclosing single pattern units with circles on the z-direction maximum intensity projected fluorescence images) of 3 ST (43-45 regions, see Source Data file) and 4 HF (52-54 regions, see Source Data file) scaffolds from mouse podocyte culture. Student’s t-test (two-sided), Welch’s t-test (two-sided), or Mann-Whitney (two-sided) test performed for (i–l). Data plotted as mean ± SD in (i–l). Each dot represents a topographical region in (i–l). DAPI 4’,6-diamidino-2-phenylindole, ST spiky topography (non-fractal topographic control group with curvature designed to match that of pointy features at the tips of a HF unit pattern); HF high fractal topography (biomimetic high fractal topographic group derived from healthy histological glomerular tracing). Source data are provided as a Source Data file.

Podocytes cultivated on both the ST and HF substrates formed a confluent layer; however, areas without cells (“holes”) were observed on the ST substrates (Fig. 4g) but not on the HF substrates (Fig. 4h). Thus, HF substrates support formation of a cellular layer in full confluency as reflected by the higher nuclei number in the topographical regions covering the patterns, despite a smaller nuclei size (Fig. 4i). Notably, we found stronger fluorescence signals of a slit diaphragm marker (nephrin) and two filament components (f-actin, nestin) in the topographical regions of HF versus ST patterns (Fig. 4j–l), suggesting higher expression of nephrin, f-actin, and nestin in cells cultured on the HF substrates. As an intermediate filament, nestin maintains podocyte foot process²³ and actin structures²⁴, which is consistent with its increased expression in the HF group, accompanied by the increased expression of nephrin and f-actin. Interestingly, nephrin on the HF topography exhibited a lower extent of both elongation and alignment compared to that on the ST topography (Fig. 4j); similar results were observed for nestin and f-actin with respect to alignment (Fig. 4k, l). Decreased elongation and alignment of the cytoskeleton structures indicate a more in vivo-like arrangement since native podocytes possess an arborized as opposed to an elongated morphology.

We additionally performed FFT on the fluorescence images visualizing f-actin (cytoskeleton structure), nestin (intermediate filament), and nephrin (slit diaphragm marker), as well as on the brightfield images showing the ST and HF substrate patterns to determine if the crystallinity of the substrate pattern was reflected in the organization of these subcellular components. It did not appear that the subcellular structures of f-actin, nestin, or nephrin strictly conformed to the substrate patterns of either ST or HF, thus, we believe the main benefit of fractality was in guiding the cell attachment points, which enabled the cells on HF to spread more and thus re-arrange their cytoskeleton (Supplementary Fig. 4).

Mosaic assay reveals single-cell morphology of podocytes on fractal substrates, supporting drug testing

We further engineered fractal well plates with densely packed HF topographies and minimal inter-unit spacing to enhance biophysical cues, enabling successful culture of E11 podocytes with Matrigel and promoting marker expression and polarity (Fig. 5a–d; Supplementary Fig. 5). To enable scalable fabrication beyond drug-absorbing PDMS, we demonstrated clear imprinting of fractal features and podocyte branching morphology on hot-embossed thermoplastic elastomer (TPE) substrates assembled with bottomless well plates (Fig. 5a–c; Supplementary Fig. 5).

Mouse E11 podocytes exhibited a higher gene expression of Nphs1 on the fractal TPE plate, compared to the flat plate (Fig. 5b). The biomimetic fractal topography not only promoted podocyte marker expression, but also resulted in a visible cytoskeleton arrangement when E11 podocytes were seeded on substrates that transitioned between flat surfaces and fractal patterns in a single well (Fig. 5c).

It has been a challenge to observe podocyte morphology in a confluent monolayer due to the difficulty in delineating the border of a single cell from those of its neighbors. Thus, we generated a mixed population of podocytes labeled with mCherry (E11 mouse podocyte cell line) scarcely distributed among their optically silent counterpart (Fig. 5d), enabling the enumeration of the Df of a single podocyte within a confluent monolayer, an approach termed Mosaic assay^13–15. We then applied a compound known to cause podocyte injury, puromycin aminonucleoside (PAN), as proof-of-concept for the use of this fractal topography for drug testing (Fig. 5e, f). For podocyte cell morphology, a higher Df indicates more branching morphology, whereas a lower aspect ratio indicates less elongation of the cells, both suggesting a higher degree of similarity to in vivo morphology. Upon PAN-induced injury, foot process effacement shown by a decrease in cell morphology Df (less arborized morphology) was observed, as expected (Fig. 5e, f). Cells on the fractal substrates appeared to be more sensitive to PAN treatment, with the Df of their morphology decreasing to a greater extent and the aspect ratio of their morphology increasing to a greater extent (Fig. 5f). Nearly identical results were obtained when the non-conditionally immortalized human podocytes (PODO/TERT256) were used in the Mosaic assay (green fluorescent protein (GFP) labeled) on fractal plates (Supplementary Results, Supplementary Figs. 6–8).

In vivo glomeruli provide convex fractal topographical cues, motivating the investigation of the effect of concave vs convex patterns on cell morphology. We observed a comparable cell morphology as assessed by Df measurements in non-conditionally imortalized human podocytes on convex vs concave fractal substrates (Supplementary Fig. 9) with a significantly higher cell coverage on convex versus concave fractal patterns (Supplementary Fig. 9). These results were not podocyte specific as a stromal cell type, dental pulp stem cells (DPSCs) cultured on the concave HF substrates exhibited a lower density compared to those on the convex HF substrates (Supplementary Fig. 9).

RNA sequencing points to podocyte maturation on fractal substrates

To investigate transcriptomic changes underlying the effects of fractal topography, we performed bulk RNA-seq on E11 mouse podocytes cultured for 14 days on NT, RT, and HF substrates. Upon principal component analysis (PCA), the NT group showed a separable cluster on the PC1 vs PC2 plot, compared to the topography groups, RT and HF, which further separated on the PC1 vs PC3 plot, despite some overlap (Fig. 6a). Unsupervised clustering analysis further substantiated that topographical groups were distinct from the NT group (Supplementary Fig. 10). When examining the differences in gene expression among the three culture substrates, we identified 418 genes significantly altered only in the RT vs NT comparison, 146 genes significantly altered only in HF vs NT, and 96 genes significantly different in only HF vs RT (Fig. 6b).

Volcano plots further demonstrated notable significant differences in gene expression amongst the groups (Fig. 6c). Comparison of HF vs NT substrates revealed that HF promoted podocyte differentiation, marked by downregulation of developmental regulators and cytoskeletal genes associated with immature states, specifically downregulation of embryonic Nr2f2²⁵, and upregulation of Dusp6²⁶. Yet, there were signs of incomplete maturation and possible stress induced by topographic stimulation, as observed through the activation of Havcr1²⁷ and Sel1l3^28,29, indicating that not all cells achieved a mature podocyte phenotype. Topographical groups exhibited a significantly higher expression of genes associated with cytoskeleton organization, Actn4 (connecting the focal adhesion to the actin cytoskeleton in foot processes^30,31) and Nes (intermediate filament³²), both of which were reported to be enriched in human and mouse late-stage podocytes³³. Among the genes that had the highest expression in HF was Srgap1 (SLIT-ROBO Rho GTPase-activating protein 1), which encodes a mechanosensitive protein important in foot process maintenance³⁴. Furthermore, the gene expression changes observed on the HF vs RT topography suggest enhanced maturation with fractal vs round topography, marked by downregulation of Ncam1 (neural cell adhesion molecule 1), which is high in renal precursors but downregulates in mature kidney epithelia^35–37. Significant downregulation of Dkk3 in HF vs RT also supports enhanced maturation, as it is expressed in the developing kidney but largely silenced in mature podocytes^38–40.

Concurrently, several gene changes were characteristic of podocyte stress, e.g., the increase in Ndrg1⁴¹, Ppbp (Cxcl7)⁴², and Agt⁴³ with a decrease in Ace2^44,45 and Cbs⁴⁶ on HF. Mechanical strain on podocytes is known to trigger local angiotensinogen production and cytoskeletal strain, similar to what we see with enhanced expression of Agt and Actn3 in HF vs RT. Collectively, this implies that although the cells are differentiating, the fractal substrate might be imposing mechanical stress.

Compared to the NT group, the HF group exhibited enriched gene ontology (GO) terms among upregulated genes related to angiogenesis, cell–substrate adhesion, wound response, positive regulation of migration and motility, basement membrane organization, ECM components, stress fibers, actin filaments, adhesion junctions, integrin binding, and growth factor binding (Fig. 6d, Supplementary Fig. 11), all of which are consistent with the improved morphology observed in HF vs NT. Many of these GO terms enriched among upregulated genes were similarly observed in the RT vs NT comparison, pointing to the importance of topography in general (Supplementary Fig. 11). Yet, when comparing HF vs RT, the HF group uniquely expressed signs of podocyte developmental maturation, such as enrichment of GO terms from upregulated genes related to kidney development, metanephros development, branching involved in ureteric bud morphogenesis (Fig. 6d) and enrichment of non-podocyte GO terms among the downregulated genes such as mesenchymal cell differentiation, neural development, ossification, stem cell development and tubule development (Fig. 6d). Overall, the RT group appeared to support the highest ECM deposition, as ECM, anchoring junction and adherens junction GO terms were enriched among downregulated genes in HF vs RT comparison (Fig. 6d). Yet, there were also some signs of stress in the HF group, such as complement activation GO term enrichment from upregulated genes in the HF vs RT comparison, requiring further investigation (Fig. 6d).

The gene expression of major and minor components in the glomerular basement membrane (GBM)⁴⁷ was increased in the topographical groups (RT and HF) compared to the NT group (Fig. 6e). The HF group consistently clustered between NT and RT, which may be beneficial in preventing over-deposition of the matrix that is often related to fibrosis and glomerulosclerosis^48,49. Among the genes that encode an important collagen IV heterotrimer col(IV)α3α4α5, in adult kidneys, only produced by podocytes, both the RT and the HF groups showed higher expression than the NT group, with a significant increase in Col4a4 and Col4a5 expression in the RT group.

Furthermore, we compared transcriptomes from NT, RT, and HF groups to developing and adult mouse podocytes by generating a hierarchical heatmap of genes elevated ≥5-fold in adult vs embryonic (E13.5) podocytes or podocytes vs kidney cortex⁵⁰, and differentially expressed across topographies (Fig. 6f). The HF group exhibited balanced, uniformly elevated expression of these significantly upregulated genes, whereas the RT and the NT groups had some of these genes upregulated and others downregulated (Fig. 6f). It appeared that genes elevated in the RT group, were downregulated in the NT group and vice-versa.

On the HF platform, many genes were expressed at a level between adult and embryonic podocytes; however, some important adult podocyte genes, such as Srgap1, were increased on the HF platform. Thus, HF provided finely tuned and uniformly activated gene expression. Similar findings were observed with significant integrin gene expression (Supplementary Fig. 12).

Comparison of RNA-seq data from mouse cell line (Fig. 6, Supplementary Figs. 10–12) and primary human podocytes (Supplementary Figs. 13 and 14) revealed concordant trends in PCA clustering and GO enrichment, including genes related to ECM organization and cell adhesion (Supplementary Figs. 11a, c and 13j, k), with the kidney specific GO terms appearing in the HF vs RT comparison for mouse (Fig. 6d) and in the HF vs NT comparison for human (Supplementary Fig. 13j). In the top enriched GO terms among the genes downregulated in HF compared to NT, several metabolic processes showed up in both mouse and human data (Supplementary Figs. 11b, d and 13o). Despite similarities, some differences emerged. In a curated podocyte-related significant gene list from the human RNA-seq dataset (Supplementary Fig. 13q), only five makers were also significant in the mouse podocyte RNA-seq (Mmp2, Lama5, Timp1, Lama2, Angpt1). This gap in matching findings between late and early mouse vs human podocytes has been recognized before ref. ³³. Taken together, our RNA-seq analyses from both the conditionally immortalized mouse podocyte line and primary fetal human podocytes showed upregulation of genes associated with ECM, cell adhesion, and kidney development in the HF group, compared to the RT and NT groups.

YAP signaling pathway plays a role in maintaining podocyte morphology on fractal substrates

YAP, a key effector of the Hippo signaling pathway, is critical for podocyte biology, with detrimental consequences of its deletion, including apoptosis and development of focal segmental glomerulosclerosis⁵¹. Given YAP’s role in mechanotransduction and the ability of native podocytes to sense and respond to tensile and shear stress, we investigated whether YAP signaling modulates podocyte behavior on fractal topographies. In the RNA-seq data from E11 mouse podocyte cell line, we identified key members of the YAP interactome Wwtr1 and Amot (generated using GeneMANIA algorithm^52,53) to be differentially regulated on the HF substrates in comparison to NT and RT (Fig. 7a).

Consistent with previous studies, YAP expression was observed in the nucleic region of mouse podocytes cultivated on both NT and HF substrates (Fig. 7c)⁵⁴. Although findings vary on whether YAP is primarily localized in the nucleus or cytoplasm of healthy podocytes, our results show more prominent YAP presence in the nuclei on HF (Fig. 7c), indicative of increased YAP activity⁵⁴. Furthermore, the nucleic vs cytoplasmic ratio of mean intensity of the YAP immunofluorescence signal was found to be significantly higher in the HF group compared to the NT group (Fig. 7b), suggesting a more activated YAP signaling pathway in the mouse podocytes on the HF substrates.

Upon YAP inhibition in the mosaic assay on both NT and HF substrates (Fig. 7d, f) using a suppressor of YAP–TEAD complex (Verteporfin⁵⁵), cell size significantly decreased (Fig. 7e, g), despite no significant changes in either Df or aspect ratio of podocyte morphology under 2 days of inhibition (Fig. 7e, g). When the presence of Verteporfin was extended to 4 days, there was no further decrease in cell viability measured by lactate dehydrogenase (LDH) secretion (Fig. 7e, g) in either the NT or HF group. After 4 days of inhibition, podocytes on HF substrates showed reduced fractal dimension and cell size, whereas on NT substrates, fractal dimension increased without changes in aspect ratio and with reduced cell size (Fig. 7e, g). Importantly, the involvement of the YAP pathway on fractal substrates was confirmed in human podocytes as well (Supplementary Results, Supplementary Fig. 15). Collectively, these results suggest the role of YAP in podocyte shape control on fractal substrates.

Discussion

We highlight here the fractal nature of glomerular podocytes and the glomerulus itself, which may ultimately prove to have a significant contribution to the functional identification of healthy vs. diseased states. Our findings suggest how fractal topographical cues could enhance cell maturation: fractal topography organizes ECM proteins and facilitates the localization and spatial arrangement of subcellular structures, resulting in higher-order branching morphology and polarity. Introducing fractal dimensions as an external variable may be pivotal for augmenting in vitro cell differentiation. Here, we emulated the fractal designs found in glomerular histology to fabricate 2.5-dimensional (2.5-D) substrates for cell culture.

Notably, in vitro podocyte biology displays limited resemblance to its native counterparts, with significantly reduced maturation and challenges in recapitulating key physiological features. To verify the podocyte behavior in response to fractal cues in a comprehensive manner, we used multiple techniques (immunostaining with confocal imaging, qPCR, Western blotting and RNA-seq) to examine the key markers, important phenotypic characteristics, their morphology, and their interactions with the substrates.

Podocytes on the fractal substrate appear to be in a delicate balance—initiating maturation while activating some stress responses to deal with biomechanical strain. Gene expression changes in the HF group reflect this duality: downregulation of Ncam1 and upregulation of Actn3 and St3gal1 suggest enhanced maturation and function, while elevated Ppbp and Agt, alongside Ace2 downregulation, indicate a stress-alert state.

YAP activation, through nuclear localization is required for podocyte function⁵⁶, and nephron development, as conditional knockdown of YAP in the mouse kidney led to reduced nephrogenesis and morphogenesis, in a manner that was independent of proliferation and apoptosis⁵⁷, consistent with our observation of YAP1 involvement in shape patterning. Wwtr1 encoding transcriptional coactivator with PDZ-binding motif (TAZ), known as the paralog of YAP, is critical in maintaining podocyte integrity⁵¹. High WWTR1 (TAZ) expression with nuclear localization in podocytes usually indicates active mechanotransduction, such as in response to stiff substrates, matrix cues or tension. In mouse podocyte line, Wwtr1 expression was similar between HF and NT but was upregulated in RT. The other differentially expressed gene from the YAP interactome was Amot, which encodes angiomotin (AMOT). Whether AMOT inhibits or promotes YAP activity appears to be context dependent⁵⁸. Higher Amot in the HF group compared to the NT and RT may suggest regulatory control with efforts at enhanced cytoplasmic sequestration of YAP. Collectively, this fits with the idea that although the HF topography mechanically stimulates podocytes, they are also invoking regulatory controls (like AMOT) to prevent prolonged pathological YAP activation. Yet mRNA data of YAP pathway members need to be interpreted with caution, since it is the localization and protein expression, rather than the transcript level or the total amount of protein, that regulates cell responses.

The marked differences in cell coverage between concave and convex substrates were not podocyte-specific and likely stem from the intrinsic unfavorability of concave geometries for cell adhesion and proliferation, as a result of mass transfer limitations⁵⁹, rather than from fractal features. Similar escape from concave topographies has been reported in other cell types⁶⁰, with proposed mechanisms involving altered stress fiber organization, cell migration directionality, and focal adhesion dynamics^61,62.

We validated our findings here by considering three cell sources: conditionally immortalized mouse podocyte line, non-conditionally immortalized human podocyte line, and primary human fetal podocytes⁶³, with varied proliferation rates, genetic fidelity, and physiological relevance. The conditionally immortalized mouse podocyte cell line has a thermoswitching mechanism to deactivate the forced cell cycle progression and allow differentiation over longer cultivation periods, e.g., 14 days. The non-conditionally immortalized human cells do not have such a mechanism; they keep proliferating past confluence, limiting the ability to achieve a fully differentiated phenotype and requiring a shorter cultivation period before overcrowding (~ 5 days). The primary podocytes are genetically unmodified; however, they are limited in number. The published RNA-seq data used for the analysis here were generated from human gestation week 19 podocytes (Lonza)^13–15,63, which are still relatively immature and proliferative. Thus, achieving ~80% to full confluence—where phenotypes stabilize without overgrowth—required cell type-specific seeding and culture protocols. Mouse podocytes were cultured for ≥14 days at a seeding density of 50,000 cells/cm², while human lines reached confluence within 5 days at a seeding density of 50,000 or 100,000 cells/cm². Notably, for each cell type, all topographies received identical seeding densities. Whereas conditionally immortalized podocyte lines have been widely used to study integrin function and cell–matrix interactions^64–67, these transformed cells differ substantially from native podocytes in vivo. Nonetheless, RNA-seq analyses of human fetal primary podocytes and a non-conditionally immortalized mouse podocyte line revealed consistent upregulation of ECM and cell–ECM-related GO terms in HF vs. NT and RT vs. NT comparisons, supporting the relevance of our findings.

Recapitulating native podocyte features—such as interdigitations and full maturation—remains a major challenge in vitro. Here, we captured aspects of fractal organization by applying a single power law to inherently multifractal podocyte and glomerular morphologies. Future studies should dissect this multifractality and integrate it into experimental design with greater precision. Advancing from 2.5D to fully 3D fractal scaffolds at physiologically relevant scales may further approximate the native microenvironment and enhance podocyte maturation.

In summary, podocytes cultured on the fractal matrices demonstrated elevated levels of functional biomarkers and more sophisticated cellular architecture. RNA sequencing indicated that podocytes on fractal substrates show enhanced ECM synthesis and alteration compared to their counterparts on flat surfaces, and enhanced maturation together with stress, in contrast to counterparts cultured on non-fractal surfaces. Our integration of fractal topography into standard tissue-culture plates could offer a straightforward and effective approach for higher fidelity cell culture.

Methods

Extraction of histological features of the glomerulus

Kidney histology images from the literature were imported to Fiji and cropped to select the glomerular region. Next, the images were thresholded using the Otsu filter in Fiji to create a binary mask that covers the tissue sections, followed by despeckling once to reduce the background noise. The outline of the binary mask was then collected in Fiji for fractal dimension analysis.

Fractal dimension analysis

The fractal dimension analysis was performed using either the box-counting algorithm from the Fraclac plugin in Fiji or the box-counting method through a MATLAB code developed by Frederic Moisy⁶⁸ (v1.0.0.0). Fraclac was used to generate single-value Df; the MATLAB code was used for multifractal analysis. Briefly, outlined binary images were input to Fraclac and box counting with the white background selected was applied to calculate fractal dimension. The method developed by Frederic Moisy involves covering the image with a set of boxes of size r, and counting the number of boxes (N) that are required to cover the entire image. The relationship between N and r can be expressed as:

N = N_{0} r^{- D f}

D f \leq D

where Df is the fractal dimension or the Minkowski-Bouligand dimension, and D is the space dimension. By plotting the local exponent, we can identify the local fractal properties within a limited range of box size r using the equation:

D f = - d \frac{\ln (N)}{\ln (r)}

FFT analysis

We employed the FFT analysis to investigate the frequency components present within our images. The FFT function available in ImageJ was used to compute the FFT image of our data. The resulting FFT image was then subjected to surface plotting and profile plotting of a central horizontal line (for the images in Fig. 1, for the fluorescence images of subcellular structures on ST and HF substrates as well as the brightfield images of HF substrate patterns in Supplementary Fig. 4) or a straight line passing through the spikes (for the brightfield images of ST substrate patterns in Supplementary Fig. 4).

Topographical pattern fabrication

The design of the topographical patterns RT, LF, and HF is reported previously^13–15. The ST pattern was designed in AutoCAD to approximate the sharp and pointy features of the HF pattern by arranging lattices of spikes (circles of 10 μm diameter, 12.5 μm between centers of adjacent circles) within a circular region (150 μm in diameter) in close proximity to the size of a repeating unit of the HF pattern. The ST unit design, packed hexagonally similar to the HF pattern, was imported to a direct laser writer (MicroWriter ML3 Baby) and replicated to span a square large enough to cover the entire area of a well in a standard tissue-culture 24-well plate. The master molds with topographical patterns were fabricated using a method developed and reported previously^13–15. Briefly, a silicon wafer was spin-coated with hexamethyldisilazane (HMDS) at 3000 rpm for 30 seconds, followed by soft baking at 150 °C for 1 minute, to increase the adhesion of the subsequent photoresist layer. Positive photoresist AZ P4620 was then spin-coated on top of the HMDS layer at 1500 rpm for 30 seconds, followed by soft baking at 110 °C for 80 seconds and rehydrating for approximately 30 minutes at room temperature. The design patterns were written onto the wafer by exposing the photoresist to a laser at 770 mJ/cm², followed by developing and reflow (120 °C for 1–2 minutes) to smooth the edges of the features.

PDMS substrate fabrication

Patterns from master molds fabricated by direct laser writing of AZ P4620 positive photoresist on a silicon wafer were copied to PDMS negative molds by inverse molding at a ratio of 1:10. The PDMS negative molds were then fully cured by baking at 120 °C overnight and used as master molds to fabricate PDMS substrates used for cell culture. Circular PDMS substrates that fit in standard tissue-culture well plates were hole-punched using a biopsy punch.

Substrate pattern characterization

Patterned substrates were imaged with a 3D digital microscope (Keyence, VHX-7000) in z-stacks to acquire their surface profiles via the associated VHX software (v3.0). The stacked microscope images were then reconstituted into a 3D display. Feature height was measured on the surface profile plots from regions of interest and radius of curvature was measured by fitting a circle on these plots. Patterned substrates were also imaged under SEM (Hitachi TM4000 or SU7000) at 5.0 kV voltage acceleration to confirm feature transfer from master molds to substrates.

Podocyte culture

The conditionally immortalized murine podocyte E11 cell line (STR authenticated) was a generous gift from GSK. Murine podocyte E11 cell line was expanded in flasks coated with 0.1 mg/ml rat tail collagen I (Corning, #354236) under the proliferative condition at 33 °C with 10 units/mL mouse interferon gamma (Thermo Scientific, PMC4031) supplemented in fresh RPMI 1640 basal media containing HEPES and GlutaMAX (Gibco, #72400120) with 10% FBS and 1% pen-strep. To induce differentiation, cells were cultured under the differentiation condition at 37 or 38 °C in DMEM-F12 basal media (Gibco, #11330032) with 10% FBS and 1% pen-strep. The immortalized PODO/TERT256 human podocyte cell line (Evercyte, CHT-033-0256, STR authenticated) was expanded in flasks coated with 6 µg/cm² human collagen I (Sigma, C7774) at 37 °C in MCDB131 basal media (Gibco, #10372019) supplemented with 20% FBS, 2 mM GlutaMAX (Gibco, #35050061), 12 µg/mL Bovine Brain Extract (Lonza, CC-4098), 10 ng/mL hEGF (Sigma-Aldrich, E9644), 25 ng/mL hydrocortisone (Sigma-Aldrich, Cat# H0396), and 100 µg/mL G418 (InvivoGen, ant-gn-5). For both mouse and human podocyte cell lines, the media were changed every 2–3 days.

Substrate preparation for cell seeding

PDMS substrates were sterilized by autoclave and inserted into well plates to allow for standard cell culture techniques to be applied for cell seeding and cultivation on the scaffolds. In addition to scaffolds, a blunt tweezer was autoclaved at the same time and then used to insert the scaffolds into the wells of a tissue-culture well plate. Each scaffold was pushed to the bottom of the well using the tweezers to ensure all edges of the scaffold were flat in the well and to eliminate the gap space between the scaffold and the well. By doing so, fewer cells would be lost in the gap space and not settle on the substrate surface when the cells were seeded. After inserting the scaffolds, PBS was added to the wells to wet the scaffold surface and the well plate was incubated at 37 °C for at least 30 minutes. Prior to cell seeding, ECM proteins were coated on top of the substrates to ensure adequate cell adhesion. Briefly, Matrigel (Corning, #354234) diluted 1:60 was coated on substrates used for mouse podocyte cultivation. Matrigel (Corning, #354234) diluted 1:60 or porcine kidney ECM (Xylyx Bio, NativeCoat Kidney ECM surface coating kit #MTSKY201) diluted according to manufacturer’s instructions at 0.08 mg/mL was coated on substrates used for human podocyte cell line cultivation. Appropriate coating is critical for successful cell culture, and kidney ECMs may exhibit batch-to-batch variability that will require optimization of the coating procedure and cell seeding density. Most consistent and reproducible results are obtained if Matrigel is used for coating.

Podocyte seeding

Mouse podocyte cell line was seeded at 50,000 cells/cm² and allowed to adhere and reach ~80% to full confluence under the proliferative condition before thermoswitching and culturing under the differentiation condition as described in the Podocyte culture section. For the early study endpoint (immature cells), mouse podocytes were cultivated under the differentiation conditions for 3 days. For the late study endpoint (mature cells), mouse podocytes were cultivated under the differentiation condition for 14 days. Human podocyte cell line was seeded at 50,000 or 100,000 cells/cm², and grown until ~80% to full confluence as the study endpoint (3–5 days). To seed the cells, harvested cells from the expansion cultures were resuspended at a concentration corresponding to the target seeding density. After that, the cell suspension was pipetted up and down a few times to mix evenly and then added to each well. Finally, the well plate was left in the incubator (37 °C for human podocytes, 33 °C for mouse podocytes) for the cells to attach. Media change was performed the next day or in two days. The same culture media were used for human podocytes, whereas the media were switched to differentiation media for mouse podocytes. To ensure a fair comparison, for each experiment, the same seeding and culture procedures were applied across the different substrate groups to examine the effect of fractal topography on either human or mouse cell behaviors.

Glycocalyx imaging

For glycocalyx imaging, mouse podocytes were fixed with 4% PFA, 1% glutaraldehyde, and 1% Alcian Blue overnight. Next, the samples were post-fixed with 0.5% osmium and dehydrated in ethanol in a serial manner, followed by critical point drying (CO₂) and gold coating. Samples were imaged using a Hitachi SU8230 or SU7000 scanning electron microscope at 2.0 kV voltage acceleration.

Immunofluorescent staining

At the study endpoint, cells were fixed with 2% or 4% PFA for 20 minutes at room temperature and rinsed with PBS three times. Next, cells were blocked in 5% normal goat serum (NGS) with 0.1% Triton X-100 for 1 hour under gentle shaking. Without rinsing, cells were incubated with primary antibodies, anti-podocin (Abcam, ab50339, 1:200 dilution), anti-nephrin (Invitrogen, PA5-20330, 1:500 dilution), anti-nestin (Invitrogen, MA1-110, 1:500 dilution), or anti-YAP1 (Invitrogen, PA1-46189, 1:500 dilution), diluted in 2% NGS with 0.04% Triton X-100 overnight at 4 °C. After that, cells were washed with PBS for three times, followed by incubating with secondary antibodies, goat anti-rabbit IgG FITC (LifeTech, F2765, 1:500 dilution), goat anti-rabbit IgG Alexa Fluor 647 (Invitrogen, A21245, 1:500 dilution), or goat anti-mouse IgG Alexa Fluor 568 (Invitrogen, A11004, 1:500 dilution), and stains, Alexa Fluor 488 Phalloidin (Invitrogen, A12379, 1:500–1:1000 dilution), or rhodamine labeled wheat germ agglutinin (Vector Laboratories, RL-1022, 1:500 dilution) in 2% NGS with 0.04% triton X-100 for 1 hour at room temperature under gentle shaking. Samples were then washed three times with PBS and carefully removed from the wells. Excess liquid was removed from the bottom of each scaffold with a paper towel. Two to three scaffolds were placed on each microscope slide. 6 µL of DAPI-containing mounting media (Vector Laboratories, Vectashield #H-1200) was added to the surface of each scaffold. Lastly, a cover slip was placed on top of each scaffold to spread the mounting media over the entire surface of the scaffold. Care was taken when handling a scaffold to make sure not to disturb the delicate sheet of cells on top. When taking a scaffold out of the well, efforts were made to minimize any contact with the cell side of the scaffold and to avoid twisting or bending the scaffold during handling. During the staining procedure, scaffolds were rinsed with solutions gently added from the side as opposed to on top. Full details of the antibodies and fluorescent stains used are listed in Supplementary Table 1.

Immunofluorescent imaging and analysis

Samples were imaged via confocal microscopy using a Leica Lightsheet or Nikon A1R confocal microscope. Z-stacked images were taken to capture cell morphology on topographic substrates. Maximum intensity projection images were acquired from the z-stacks for image analysis. Regions of interest in the images from the experiment comparing the HF group with the ST control were selected in Fiji as circular areas containing the HF and ST unit patterns. To quantify fluorescence signals, the channel of interest was thresholded by a filter in Fiji and the integrated and mean intensity were measured. Directionality analysis was performed using MATLAB code developed by Landau et al.⁶⁹ to compute eccentricity and orientation variance of the signals from subcellular structures. Eccentricity is defined as a measurement of the elliptical shape of structures, where a score of 0 is considered a circle and a score approaching 1 indicates elongation of structures. Briefly, the channels of interest were converted to a binary format and the regionprops function was used to measure the average eccentricity value of all elements and orientation variance within each region of interest. The integrated intensity, mean intensity, averaged eccentricity, and orientation variance were reported as values combined from two batches of experiments after each value was normalized to the mean value of the ST control from the corresponding batch. The images from the experiment with four substrates (NT, RT, LF, HF) were analyzed as a whole without selecting regions of interest. Integrated intensity of the fluorescence signals from these images was quantified in Fiji as described above, averaged by cell numbers, and reported as values normalized to the average value from the day 3 NT group. The same settings were applied to all the samples during image acquisition, processing and analysis.

Mosaic assay

Mosaic PODO/TERT256 human podocytes (a mixture of GFP-labeled cells and non-fluorescent cells at a ratio of 1:11) or Mosaic E11 mouse podocytes (a mixture of mCherry-labeled cells and non-fluorescent cells with approximately 5% mCherry-positive cells) were cultured and seeded in the same way as regular human or mouse podocyte cell lines as described above. At the study endpoint, Mosaic cells were fixed with PFA, followed by staining with rhodamine-labeled wheat germ agglutinin (WGA) and mounting with DAPI-containing mounting media as described above. Samples were imaged via confocal microscopy using a Zeiss LSM 880 Super Resolution or Leica Lightsheet Confocal microscope. Z-stacked images spanning the entire section with visible signals were taken for each sample, followed by processing via maximum intensity projection. The images acquired on the Zeiss LSM 880 Super Resolution microscope were subjected to contrast increase using the Zen 2.3 desk software to increase the contrast of GFP-positive cells’ morphology before cell selection in Photoshop by setting gamma in the FITC channel to 2.5, with the white point adjusted to 30 and adjusting the white point of the DAPI channel to 80. The wand tool in Photoshop (for analyzing contrast enhanced Zeiss images as described above) or Fiji (for analyzing images acquired on the Leica Lightsheet Confocal microscope) was used to select fluorescent cells from which binary masks of single cells were created (in Fiji, the auto function was used to enhance contrast to facilitate cell selection and the masks obtained subsequently were despeckled once to reduce noise). Single cells were verified by overlaying the DAPI channel with the FITC (Mosaic human podocytes) or mCherry (Mosaic mouse podocytes) channel to ensure each cell only contained one nucleus. The outline of each binary mask was obtained in Fiji, and fractal dimension was calculated using the FracLac plugin as described above.

EV isolation

At the study endpoint (14 days of differentiation on flat and fractal substrates), mouse podocytes were washed with PBS twice and cultured in serum-free culture media for a period of 48 hours. Media from 8 scaffolds for each topographical group was combined and collected in a 15 mL centrifuge tube. Cell debris was removed by centrifuging at 3000 × g for 10 minutes at 20 °C. EVs were then isolated by the miRCURY Exosome Isolation Kit (Qiagen, #76743) according to the manufacturer’s protocol. A precipitation buffer was applied at a ratio of 4:10 to the samples, followed by rotating overnight at 4 °C for the precipitation of EVs. Precipitated EVs were then centrifuged at 3200 × g for 30 minutes at 20 °C. After removing the supernatant, the pellet was gently rinsed with PBS. 100 µL resuspension buffer was added to isolated EVs prior to downstream analysis.

Western blotting

Mouse podocytes were cultivated on PDMS substrates that fit in 24-well plates, with six wells per plate dedicated to each of the four surface topographies (NT, RT, LF, HF). For PODXL western blotting, cell lysates for each topography were pooled from six wells. Upon 14 days of differentiation, cells were washed with ice-cold PBS. 150 μl of RIPA buffer supplemented with a protease inhibitor tablet (Thermo Scientific) was added to a single well for podocyte lysis, pipetting up and down several times while scraping the surface of the well with the pipette tip to release cells. The full volume of the RIPA-lysate mixture was then transferred sequentially to the next five wells of like-topography, one at a time, repeating the process of pipetting and scraping each time. After all six wells of a single topography were lysed, the RIPA lysate mixture was transferred to a microcentrifuge tube, and lysis was repeated for the remaining three topographies. For COL4A4 western blotting, cell lysates for each of the four topographical groups were collected from a single well instead of being pooled in the same way as described above.

EVs were isolated from conditioned media of day 14 differentiation of mouse podocytes grown on NT and HF topographies as described above. EV pellets were lysed in 50 μl RIPA buffer supplemented with protease inhibitor. Cell and EV lysates were centrifuged at 10,000 × g and 4 °C for 15 minutes. Supernatants were transferred to fresh tubes, and pellets were discarded.

The Pierce BCA Protein Assay Kit (Thermo-Fisher Scientific) was used to analyze protein concentration in lysates. Initial western blots were performed to determine the linear regime for sample loading and target detection. For podocyte cell lysate western blots, samples were loaded at 7 µg (PODXL blot) or 4 µg (COL4A4 blots) total protein adjusted to a volume of 35 μl per well of a 15-well gel; EV blots were loaded at 3.7 µg (replicate 1) or 10 µg (replicates 2 and 3) total protein adjusted to a volume of 40μl per well of a 10-well gel.

Appropriate sample volumes were diluted with 4x Sample Loading Buffer (LI-COR), 10x BOLT Sample Reducing Agent (Thermo Fisher Scientific), and DI water, then incubated at 70 °C for 10 minutes. Samples were loaded into Bolt 4–12% Bis-Tris Gels (Invitrogen). Electrophoresis was performed in Bolt MES SDS Running Buffer (Invitrogen) at 200 V for 35 minutes or until the protein front reached the bottom of the gels. Completed gels were rinsed in DI water, then transferred to Immobilon-FL PVDF membranes (Millipore Sigma) at 20–25 V for 7 minutes using the iBlot Gel Transfer Device and transfer stacks (Invitrogen).

REVERT Total Protein Stain (LI-COR) was applied to PVDF membranes according to the manufacturer's protocol before imaging on the LI-COR Odyssey Fc Imager (LI-COR). Blocking was then performed by incubating membranes in skim milk buffer for 1 hour. Membranes were washed and stained overnight at 4 °C with anti-PODXL (Invitrogen, PA5-28116, 1:1000 dilution), anti-COL4A4 (BiCell Scientific, 05004wb, 1:500 dilution), anti-CD63 (Abcam, ab217345, 1:1000 dilution), or anti-ALIX (Invitrogen, MA5-32773, 1:1000 dilution). Membranes were then washed, donkey anti-rabbit IgG IRDye800CW secondary antibody (LICORbio, 925-32213, 1:10,000 dilution) was added for 2 hours, washing was repeated, and images were captured with a 10-minute exposure. Before staining for GAPDH, the PODXL blot was stripped using a mild stripping buffer consisting of 1.5% w/v glycine, 0.1% w/v SDS, and 1% v/v Tween 20, titrated to pH 2.2 in deionized water. Blot was incubated in stripping buffer twice for 5 min, washed four times, then blocked as previously described. PODXL blot was then washed and stained as before using anti-GAPDH (Invitrogen, PA1-987, 1:1000 dilution) and IRDye800CW secondary antibody prior to imaging. The COL4A4 blots were stained for GAPDH as described above without stripping. The blot containing EV replicate 1 from NT and HF was used for ALIX detection, followed by stripping and reusing for CD63 detection. Full details of the antibodies used for western blotting are listed in Supplementary Table 1. Quantification of western blot signals was performed using LICORbio’s Image Studio software (v5.5.4).

Transmission electron microscopy of EVs

Plasma treatment was used to pre-treat electron microscopy grids (Electron Microscopy Sciences). Grids were incubated with 5 µl of 100× EV concentrate for 1 minute before wicking with a wet filter paper. To wash grids, 5 µl of DI water was applied, then wicked away. 5 µl of 2% uranyl acetate was then added to grids for 30 seconds as a negative stain. Grids were once again wicked, and imaging was subsequently performed on a Talos L120C TEM (Thermo-Fisher Scientific) at ×22,000, ×45,000, and ×92,000.

Fractal topographic well plate fabrication

To produce 384-well plates with a high-density fractal pattern, extruded, thermoplastic elastomer sheets (230 μm thick, Vertex FK79, Gel-Pak Inc, USA) were hot embossed for 10 minutes at 160 °C (EVG520HE, EV Group, Austria). The base was then sealed to a self-adhesive bottomless well plate from Grace Bio-Labs. 96-well plates with a high-density fractal pattern and TPE bottoms were fabricated in a similar method using a master mold designed for a 96-well blueprint.

PAN injury model

For the PAN injury model on human cells, Mosaic PODO/TERT256 human podocytes were seeded on the standard and fractal 96-well plates at a concentration of 50,000 cells/cm². PAN (10 μg/ml, Sigma-Aldrich, P7130) was added to the culture media 24 hours post-seeding. Fresh media containing 10 μg/ml PAN was changed on day 3. Cells were fixed on day 5. For the PAN injury study on mouse cells, Mosaic E11 mouse podocytes were seeded on the NT and HF substrates at a concentration of 50,000 cells/cm². PAN (100 μg/ml, Sigma-Aldrich, P7130) was added to the culture media after 14 days of differentiation. Fresh media containing 100 μg/ml PAN was changed every 3 days until day 8 post-introducing PAN, on which day cells were fixed. Mosaic assay imaging and analysis were performed as described above.

DPSC culture and seeding

DPSCs (Lonza, PT-5025, not authenticated) were expanded in low-glucose DMEM (Gibco, #11054020), supplemented with 10% FBS, 1% MEM Non-Essential Amino Acids (NEAA) Solution (Thermo Scientific, #11140050), 1% GlutaMAX (Gibco, #35050061), and 1% pen-strep. The media was changed every two days. DPSCs were seeded at 50,000 cells/cm² on concave and convex HF substrates coated with Matrigel. Cells were fixed on day 2 and stained with phalloidin and DAPI. Confocal microscope images were acquired in z-stacks as described above. Cell coverage on the concave and convex topography was quantified by counting the number of nuclei per image.

RNA isolation

RNA extraction and isolation of mouse podocytes were performed using the PicoPure RNA isolation Kit (Thermo Fisher Scientific, #KIT0204) according to the manufacturer’s protocol (100 µL instead of 50 µL of lysis buffer was applied to extract RNA), including DNase treatment with RNase-Free DNase Set (Qiagen, #79254). At the final step, 12 µL of elution buffer was used. The concentration and quality of isolated RNA were measured using a NanoDrop spectrophotometer. Isolated RNA was stored at −80 °C.

Quantitative polymerase chain reaction

Isolated RNA was reverse transcribed into cDNA using the SuperScript™ II RT kit (Invitrogen, #18064-014) according to the manufacturer’s protocol. Briefly, random hexamer primer (ThermoFisher, #SO142) was mixed with total RNA (300 ng), dNTP mix (Invitrogen, #18427-013) and water according to the specified quantities. The mixture was heated to 65 °C for 5 min, then chilled on ice. Each of the remaining contents of the SuperScript™ II RT kit was then added and incubated as specified in the protocol, with the addition of 1 µL of RNaseOUT™ (Invitrogen, 10777-019) to each preparation. The RT mixture was incubated at 25 °C for 10 min, 42 °C for 50 min, and 70 °C for 15 min. Each final 20 µL cDNA prep was diluted 1:3 in RNase-free sterile water. qPCR was carried out following the PowerTrack™ SYBR™ Green Master Mix (ThermoFisher, #A46109) kit protocol and performed on a Bio-Rad CFX384 Touch Real-Time PCR Detection System. A 10 µL reaction volume was used for each PCR well in a 384-well plate, adding diluted cDNA, primers, yellow sample buffer, master mix, and water as specified in the kit protocol. The sequences of Gapdh and Nphs1 primers (Supplementary Table 2) used are listed below. C_t values were obtained from the Bio-Rad CFX Manager 3.1 software. For comparison of gene expression, analysis was performed using the ΔΔC_t method, with Gapdh as the reference gene and averaged ΔC_t values for NT as the control condition from which to determine ΔΔC_t differences and fold changes for NT and HF data points.

Gapdh forward: GGGAAGGTGAAGGTCGGAGTC

Gapdh reverse: TGGAATTTGCCATGGGTGGAA

Nphs1 forward: GTGCCCTGAAGGACCCTACT

Nphs1 reverse: CCTGTGGATCCCTTTGACAT

Bulk RNA sequencing of E11 mouse podocytes

RNA extraction and isolation from E11 mouse podocytes cultivated on the NT, RT, and LF/HF substrates were performed as described above. Bulk RNA sequencing was performed by Novogene. Briefly, following sample quality control, mRNA was isolated from total RNA using poly-T oligo-attached magnetic beads. The mRNA was subsequently fragmented and reverse transcribed into first-strand cDNA using random hexamer primers. Second-strand cDNA synthesis was performed using dTTP for a non-directional library. The library was assessed using Qubit and real-time PCR for quantification and a bioanalyzer for size distribution analysis. Quantified libraries were then pooled and sequenced on the Illumina NovaSeq X Plus platform with 2 × 150 bp paired-end reads, targeting a depth of 20 million reads per direction. Raw sequencing data were processed through in-house software for adapter trimming, quality filtering, and quality control analyses. Cleaned reads were aligned to the mouse reference genome using Hisat2 (v2.0.5). Quantification of gene expression level was performed using featureCounts (v1.5.0-p3) to count the reads mapped to each gene and obtain FPKM of each gene. RNA sequencing data from the mouse podocyte cell line cultivated on the LF substrates were collected; however, due to their similarity with the HF group, analyses were performed using just one fractal topographical group (namely HF) and compared with the non-fractal and non-topographical controls (RT and NT, respectively).

PCA analysis was performed on raw counts of the mouse RNA sequencing data for the NT, RT, and HF groups using PCAGO (v1.0.0, https://pcago.bioinf.uni-jena.de/). The DESeq2 package on the Galaxy web platform public server^70,71 (2.11.40.8+galaxy0) was used to obtain rlog normalized counts of the mouse RNA sequencing data for the NT, RT, and HF groups. Differential expression analysis employing DESeq2 (v1.20.0) was performed using the NovoMagic analyzer (online RNA-seq bioinformatics analysis tool, Novogene). Volcano plots of the rlog normalized counts were generated by GraphPad Prism 10 to show the most significant and the most differentially expressed genes between HF vs NT, RT vs NT, and HF vs RT. K-means clustering module of the pheatmap R package (v1.0.12) was used to classify the 500 most variable genes (in the NT, RT, and HF groups combined) into 3 clusters using the kmeans_k parameter. Genes of interest from the normalized count files were filtered and plotted using SRPlot⁷² (www.bioinformatics.com.cn/srplot). GO enrichment analysis was performed using NovoMagic (full data output can be found in Supplementary Data 1), and the results were plotted using GraphPad Prism 10.

Bulk RNA sequencing data of E11 mouse podocytes have been archived at the Short Read Archive at NCBI. The accession number is GSE298416.

PCA and Pluritest analysis of RNA-seq data from primary fetal human podocytes

Data were analyzed from the GEO database, accession numbers GSE185491 and GSE116471. The R prcomp function was used to perform PCA analysis of the data in GSE185491. The Pluritest⁷³ tool (https://www.pluritest.org/) was used to evaluate the pluripotency quality of the primary fetal human podocytes on different topographical substrates (GSE185491) in comparison to human iPS cell-derived podocytes and sorted human adult podocytes (GSE116471).

KeyGenes analysis of RNA-seq data from primary fetal human podocytes

Data were analyzed from the GEO database, accession number GSE185491. In order to quantify the tissue type and the developmental stage of the differentiated podocytes at the different topographical states, we used the identity score (range 0–1) from the KeyGenes tool⁷⁴ (v2.0.0, https://github.com/chuvalab/KeyGenes). The identity score predicts the “identity” of a test sample to a known feature (tissue or age) by comparing gene expression profiles of classifier genes used in KeyGenes. A higher predicted identity score implicates a greater identity to a specific tissue or differentiation age. The identity score for the topographic replicates against kidney tissue, and the adult stage of development was plotted using GraphPad Prism 10.

Heatmaps and clustering of RNA-seq data from primary fetal human podocytes

Data were analyzed from the GEO database, accession number GSE185491. rlog normalized counts of RNA sequencing data for NT, RT, and HF samples were produced using the DESeq2 package on the Galaxy web platform public server^70,71 (2.11.40.7+galaxy2). Genes of interest from normalized count files were filtered and plotted using the heatmap2 function on Galaxy⁷¹ as well as the pheatmap R package (v1.0.12, https://CRAN.R-project.org/package=pheatmap), employing bidirectional complete Euclidean clustering. K-means clustering module of the pheatmap R package (v1.0.12) was used to classify the 1000 most variable genes (in the Frozen, NT, RT, and HF groups combined), obtained by DESeq2⁷⁰ (v1.34.0), into 8 clusters and ShinyGo⁷⁵ (v0.74) was used for further enrichment analysis on these clusters.

GO analyses of RNA-seq data from primary fetal human podocytes

Data were analyzed from the GEO database, accession number GSE185491. For both of the pairwise comparison sets of HF vs. NT and RT vs. NT, a list of significantly differentially expressed genes (FDR < 0.05) and their respective log₂(Fold Change) values, obtained by DESeq2⁷⁰ (v1.34.0), was uploaded to the PANTHER knowledgebase for GO analysis⁷⁶. The PANTHER statistical enrichment test (release 20221017, GO database 10.5281/zenodo.6799722, released 2022-07-01) was used to generate lists of significantly upregulated and downregulated GO terms from the biological process, cellular component, and molecular function databases using a threshold of FDR < 0.05⁷⁷. Full data output from GO analyses can be found in Supplementary Data 2 and 3.

YAP inhibition study

HF and NT scaffolds were prepared as described above. For the inhibition study on human cells, Mosaic PODO/TERT256 human podocytes were seeded on top of the scaffolds at a concentration of 50,000 cells/cm² (100,000 cells per scaffold). Verteporfin (1 or 2 μg/ml, Tocris) was added to the culture media 24 hours post-seeding. Samples were fixed on day 3. Mosaic assay imaging and analysis were performed as described above. To validate cell viability, the lactate dehydrogenase assay (LDH) assay (Cayman) was performed on media collected on day 5 of culture according to the manufacturer’s protocol. For the inhibition study on mouse cells, Mosaic E11 mouse podocytes were seeded on top of the scaffolds at a concentration of 50,000 cells/cm² (100,000 cells per scaffold) and allowed to differentiate for 14 days before Verteporfin (2 mg/ml, Tocris) was added to the media. The cells were exposed to Verteporfin for either 2 days or 4 days before the media was collected for LDH assay, and the cells were fixed for Mosaic assay imaging and analysis as described above. The inhibitor-free control groups in both the human and the mouse experiments were subjected to the same treatment as their counterparts, except without Verteporfin supplemented in the media.

Additionally, mouse cells in the inhibitor-free control group for 2 days of exposure were immunostained with YAP (Supplementary Table 1) and imaged by confocal microscopy as described above. To quantify the YAP fluorescent signal within the nucleus and cytoplasm, we developed a custom MATLAB script to analyze the images. For each image, the blue channel (DAPI) was used to identify the nuclei by applying Otsu’s thresholding method, followed by binary filling holes and removing small artifacts. The resulting binary mask defined the nuclear region, while its inverse was used to identify the cytoplasmic region. The green channel, representing the YAP signal, was analyzed within each defined compartment. The mean pixel intensity (total intensity divided by the number of pixels) within the respective regions was obtained, and the ratio of the values between the nucleus and the cytoplasm was reported.

YAP1 interactome

The YAP1 interactome was generated using the GeneMANIA (v3.6.0) prediction server plugin in Cytoscape^52,53. Differentially expressed nodes from the interactome were highlighted and plotted on a separate heatmap as described above in the Heatmaps and Clustering section for primary fetal human podocyte RNA-seq data (GSE185491). For E11 mouse podocyte RNA-seq data, the significantly differentially expressed genes from the same YAP1 interactome were plotted using SRPlot⁷² (www.bioinformatics.com.cn/srplot).

Statistics

Statistical analysis was performed in GraphPad Prism 10 where t test (two-sided) was used to compare two experimental groups and one-way or two-way ANOVA was used for comparison of more than two experimental groups, with p < 0.05 considered as significant. Multiple comparisons were performed using Tukey’s test. Normal distribution and homogeneity of variance were tested. When normality was not confirmed, the Mann-Whitney nonparametric test or the Kruskal–Wallis one-way analysis of variance with Dunn’s multiple comparisons test was performed instead. Welch’s t test was performed when equal variance was not confirmed. Outliers in image analyses/quantifications identified by GraphPad Prism’s ROUT method were excluded from endpoint statistical analyses.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(28.5MB, pdf)}

Reporting Summary^{(87.3KB, pdf)}

Supplementary Data 1^{(1.4MB, xlsx)}

Supplementary Data 2^{(36.5KB, xlsx)}

Supplementary Data 3^{(17.6KB, xlsx)}

41467_2025_66037_MOESM6_ESM.pdf^{(27.3KB, pdf)}

Description of Additional Supplementary Files

Transparent Peer Review file^{(3.1MB, pdf)}

Source data

Source data^{(22.7MB, xlsx)}

Acknowledgements

We thank K. Chan and J. Moffat for transducing and establishing a stable fluorescent immortalized human podocyte cell line used in the study, C. Ng for fluorescently labeling mouse conditionally immortalized podocyte cell line, R. John for providing insights on histology, L. Fiddes and Y. Chen for sample preparation for SEM imaging, J. Tam and M. Li for technical help with SEM image acquisition, L. Lukic and K. Soon for technical help with microfabrication of thermoplastic substrates. This study is funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) Idea to Innovation grant (I2IPJ 567631-22), NSERC Discovery Grant (RGPIN 326982-10), Canadian Institutes of Health Research (CIHR) Foundation Grant FDN-167274, Center for Research and Applications in Fluidic Technologies Project Award (CRAFT, 512814) and Canada Foundation for Innovation Grant #36442. M.R. was supported by Killam Fellowship and Canada Research Chair. S.O. was supported by CIHR Canada Graduate Scholarship and Additional Ventures.

Author contributions

C.L. designed and performed experiments, analyzed data, and prepared the manuscript; P.A., K.T.W., and M.R. analyzed RNA sequencing data; K.T.W. performed EV analysis and western blotting, and contributed to qPCR; S.L. performed multifractal and Fourier transform analysis, directionality analysis, and LDH assay, and contributed to fluorescence image analysis; N.R. contributed to qPCR; Y.Z. contributed to RNA sequencing data analysis and fluorescence image acquisition and analysis; K.M. fabricated fractal topographical 96- and 384-well plates; D.B., S.P-G., and S.O. contributed to fluorescence image acquisition; T.C., X.S., L.S., S.R-R., E.V., C.Y.L., A.R., and S.C-F. contributed to review and editing of the manuscript; T.V., M.S., T.F., U.B., and A.K. contributed to conceptualization and co-supervision of the study and review and editing of the manuscript; M.R. conceived and supervised the study, designed experiments, and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA sequencing data of E11 murine podocytes generated and analyzed in this study have been deposited in the GEO repository under accession code GSE298416. The RNA sequencing data of primary fetal human podocytes analyzed in this study are available in the GEO repository under accession code GSE185491. The RNA sequencing data of human iPS cell-derived podocytes and sorted human adult podocytes analyzed in this study are available in the GEO repository under accession code GSE116471. Source data for all the graphs within the paper are provided in the Source Data file and the Supplementary Data files. Raw data (original images) and analysis files for all the figures within the paper are available on the Mendeley Data repository (10.17632/pwwtfbm53h.1). Source data are provided with this paper.

Code availability

Boxcount code used for multifractal analysis in this study is developed by Frederic Moisy and available on MATLAB Central File Exchange at https://www.mathworks.com/matlabcentral/fileexchange/13063-boxcount. MATLAB code used for directionality and localization analyses in this study is developed by Landau et al. and published⁶⁹.

Competing interests

M.R., A.K. and C.L. are inventors on a pending US Patent application (No. 18/216,504) relating to fractal cues for cell maturation. The remaining authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66037-8.

References

1.Ichimura, K. et al. Morphological process of podocyte development revealed by block-face scanning electron microscopy. J. Cell Sci.130, 132–142 (2017). [DOI] [PubMed] [Google Scholar]
2.Jordanova, E. et al. The fractal and textural analysis of glomeruli in obese and non-obese patients. J. Pathol. Inf.13, 100108 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Kojima, K. et al. Podocyte flattening and disorder of glomerular basement membrane are associated with splitting of dystroglycan-matrix interaction. J. Am. Soc. Nephrol.15, 2079 (2004). [DOI] [PubMed] [Google Scholar]
4.Korolj, A., Wu, H.-T. & Radisic, M. A healthy dose of chaos: using fractal frameworks for engineering higher-fidelity biomedical systems. Biomaterials219, 119363 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Shankland, S. J., Pippin, J. W., Reiser, J. & Mundel, P. Podocytes in culture: past, present, and future. Kidney Int.72, 26–36 (2007). [DOI] [PubMed] [Google Scholar]
6.Yoshimura, Y. et al. Manipulation of nephron-patterning signals enables selective induction of podocytes from human pluripotent stem cells. JASN30, 304–321 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Hale, L. J. et al. 3D organoid-derived human glomeruli for personalised podocyte disease modelling and drug screening. Nat. Commun.9, 1–17 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Hu, M. et al. A biomimetic gelatin-based platform elicits a pro-differentiation effect on podocytes through mechanotransduction. Sci. Rep.7, 43934 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Musah, S. et al. Mature induced-pluripotent-stem-cell-derived human podocytes reconstitute kidney glomerular-capillary-wall function on a chip. Nat. Biomed. Eng.1, 0069 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Petrosyan, A. et al. A glomerulus-on-a-chip to recapitulate the human glomerular filtration barrier. Nat. Commun.10, 3656 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Xie, R. et al. h-FIBER: microfluidic topographical hollow fiber for studies of glomerular filtration barrier. ACS Cent. Sci.10.1021/acscentsci.9b01097 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Korolj, A. et al. Curvature facilitates podocyte culture in a biomimetic platform. Lab a Chip18, 3112–3128 (2018). [DOI] [PubMed] [Google Scholar]
13.Korolj, A. Engineering the fractal properties of podocytes. TSpace. https://tspace.library.utoronto.ca/handle/1807/128350 (2022).
14.Korolj, A. et al. Multiscale fractal shape cues support hierarchical maturation of podocytes via curvature induced extracellular matrix patterning. Tissue Eng. - Part A359, 359 (2022). [Google Scholar]
15.Korolj, A., Radisic, M., Konvalinka, A. & Liu, C. Fractal cues support hierarchical maturation of cells via curvature-induced patterning. US Patent US 2024/0002809 A1 (2024).
16.Shurer, C. R. et al. Physical principles of membrane shape regulation by the glycocalyx. Cell177, 1757–1770.e21 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Nielsen, J. S. et al. The CD34-Related molecule podocalyxin is a potent inducer of microvillus formation. PLoS One2, e237 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Economou, C. G. Enhanced podocalyxin expression alters the structure of podocyte basal surface. J. Cell Sci.117, 3281–3294 (2004). [DOI] [PubMed] [Google Scholar]
19.Nielsen, J. S. & McNagny, K. M. The role of podocalyxin in health and disease. JASN20, 1669–1676 (2009). [DOI] [PubMed] [Google Scholar]
20.Burger, D. et al. Urinary podocyte microparticles identify prealbuminuric diabetic glomerular injury. J. Am. Soc. Nephrol.25, 1401–1407 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Pieuchot, L. et al. Curvotaxis directs cell migration through cell-scale curvature landscapes. Nat. Commun.9, 3995 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Luciano, M. et al. Cell monolayers sense curvature by exploiting active mechanics and nuclear mechanoadaptation. Nat. Phys.17, 1382–1390 (2021). [Google Scholar]
23.Eladl, M. A., M Elsaed, W., Atef, H. & El-Sherbiny, M. Ultrastructural changes and nestin expression accompanying compensatory renal growth after unilateral nephrectomy in adult rats. Int J. Nephrol. Renovasc. Dis.10, 61–76 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Wen, D. et al. Upregulation of nestin protects podocytes from apoptosis induced by puromycin aminonucleoside. Am. J. Nephrol.34, 423–434 (2011). [DOI] [PubMed] [Google Scholar]
25.Suh, J. M. et al. The expression profiles of nuclear receptors in the developing and adult kidney. Mol. Endocrinol.20, 3412–3420 (2006). [DOI] [PubMed] [Google Scholar]
26.Chen, L. et al. DUSP6 protects murine podocytes from high glucose‑induced inflammation and apoptosis. Mol. Med. Rep.22, 2273–2282 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Zhao, X. et al. Glomerular expression of kidney injury molecule-1 and podocytopenia in diabetic glomerulopathy. Am. J. Nephrol.34, 268–280 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Wang, H., Ma, X., Li, S. & Ni, X. SEL1L3 as a link molecular between renal cell carcinoma and atherosclerosis based on bioinformatics analysis and experimental verification. Aging15, 13150–13162 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Shen, C.-Y. et al. Tissue proteogenomic landscape reveals the role of uncharacterized SEL1L3 in progression and immunotherapy response in lung adenocarcinoma. J. Proteome Res.22, 1056–1070 (2023). [DOI] [PubMed] [Google Scholar]
30.Perico, L., Conti, S., Benigni, A. & Remuzzi, G. Podocyte–actin dynamics in health and disease. Nat. Rev. Nephrol.12, 692–710 (2016). [DOI] [PubMed] [Google Scholar]
31.Jalanko, H. & Kääriäinen, H. Chapter 64 - Nephrotic disorders. In Emery and Rimoin’s Principles and Practice of Medical Genetics (6th edn) (eds Rimoin, D., Pyeritz, R. & Korf, B.) 1–13 (Academic Press, Oxford, 2013).
32.Karaiskos, N. et al. A single-cell transcriptome atlas of the mouse glomerulus. J. Am. Soc. Nephrol.29, 2060 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Tran, T. et al. In vivo developmental trajectories of human podocyte inform in vitro differentiation of pluripotent stem cell-derived podocytes. Dev. Cell50, 102–116.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Haydak, J. & Azeloglu, E. U. Role of biophysics and mechanobiology in podocyte physiology. Nat. Rev. Nephrol.20, 371–385 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Buzhor, E. et al. Reactivation of NCAM1 defines a subpopulation of human adult kidney epithelial cells with clonogenic and stem/progenitor properties. Am. J. Pathol.183, 1621–1633 (2013). [DOI] [PubMed] [Google Scholar]
36.Abbate, M., Brown, D. & Bonventre, J. V. Expression of NCAM recapitulates tubulogenic development in kidneys recovering from acute ischemia. Am. J. Physiol. Renal. Physiol.277, F454–F463 (1999). [DOI] [PubMed] [Google Scholar]
37.Caza, T. N. et al. Neural cell adhesion molecule 1 is a novel autoantigen in membranous lupus nephritis. Kidney Int.100, 171–181 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Fang, X., Hu, J., Chen, Y., Shen, W. & Ke, B. Dickkopf-3: current knowledge in kidney diseases. Front. Physiol. 11 (2020). [DOI] [PMC free article] [PubMed]
39.Federico, G. et al. Tubular Dickkopf-3 promotes the development of renal atrophy and fibrosis. JCI Insight1, e84916 (2016). [DOI] [PMC free article] [PubMed]
40.Ju, W. et al. Renal gene and protein expression signatures for prediction of kidney disease progression. Am. J. Pathol.174, 2073–2085 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Wakisaka, Y. et al. Cellular distribution of NDRG1 protein in the rat kidney and brain during normal postnatal development. J. Histochem. Cytochem.51, 1515–1525 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Zhang, F. et al. PPBP as a marker of diabetic nephropathy podocyte injury via Bioinformatics Analysis. Biochem. Biophys. Res. Commun.577, 165–172 (2021). [DOI] [PubMed] [Google Scholar]
43.Márquez, E., Riera, M., Pascual, J. & Soler, M. J. Renin-angiotensin system within the diabetic podocyte. Am. J. Physiol. Renal. Physiol.308, F1–F10 (2015). [DOI] [PubMed] [Google Scholar]
44.Harris, R. C. Podocyte ACE2 protects against diabetic nephropathy. Kidney Int.82, 255–256 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Maksimowski, N., Williams, V. R. & Scholey, J. W. Kidney ACE2 expression: implications for chronic kidney disease. PLoS One15, e0241534 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Xu, Z. et al. Ischemia-reperfusion reduces cystathionine-β-synthase-mediated hydrogen sulfide generation in the kidney. Am. J. Physiol. Renal. Physiol.297, F27–F35 (2009). [DOI] [PubMed] [Google Scholar]
47.Naylor, R. W., Morais, M. R. P. T. & Lennon, R. Complexities of the glomerular basement membrane. Nat. Rev. Nephrol.17, 112–127 (2021). [DOI] [PubMed] [Google Scholar]
48.Borza, C. M. & Pozzi, A. The role of cell-extracellular matrix interactions in glomerular injury. Exp. Cell Res318, 1001–1010 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Bülow, R. D. & Boor, P. Extracellular matrix in kidney fibrosis: more than just a scaffold. J. Histochem. Cytochem.67, 643–661 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Brunskill, E. W., Georgas, K., Rumballe, B., Little, M. H. & Potter, S. S. Defining the molecular character of the developing and adult kidney podocyte. PLOS One6, e24640 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Chen, J., Wang, X., He, Q. & Harris, R. C. TAZ is important for maintenance of the integrity of podocytes. Am. J. Physiol. Renal. Physiol.322, F419–F428 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Warde-Farley, D. et al. The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res.38, 214–220 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Shannon, P. et al. Cytoscape: a software environment for integrated models. Genome Res.13, 2498–2504 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.McEvoy, C. M. & Yuen, D. A. Not too much, not too little—just the right amount: the story of YAP in the podocyte. Kidney Int.105, 1157–1159 (2024). [DOI] [PubMed] [Google Scholar]
55.Feng, J. et al. Verteporfin, a suppressor of YAP–TEAD complex, presents promising antitumor properties on ovarian cancer. Onco Targets Ther.9, 5371–5381 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Bonse, J. et al. Nuclear YAP localization as a key regulator of podocyte function. Cell Death Dis.9, 1–14 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Reginensi, A. et al. Yap- and Cdc42-dependent nephrogenesis and morphogenesis during mouse kidney development. PLOS Genet.9, e1003380 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Kang, P. H., Schaffer, D. V. & Kumar, S. Angiomotin links ROCK and YAP signaling in mechanosensitive differentiation of neural stem cells. MBoC31, 386–396 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Malda, J. et al. Chapter 12 - Cell nutrition. In Tissue Engineering (eds Blitterswijk, C. van et al.) 327–362 (Academic Press, Burlington, 2008).
60.Park, J. Y., Lee, D. H., Lee, E. J. & Lee, S.-H. Study of cellular behaviors on concave and convex microstructures fabricated from elastic PDMS membranes. Lab Chip9, 2043–2049 (2009). [DOI] [PubMed] [Google Scholar]
61.Kim, M.-H., Sawada, Y., Taya, M. & Kino-oka, M. Influence of surface topography on the human epithelial cell response to micropatterned substrates with convex and concave architectures. J. Biol. Eng.8, 13 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Bidan, C. M. et al. How linear tension converts to curvature: geometric control of bone tissue growth. PLoS One7, e36336 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.GEO Accession viewer. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE185491 (2022).
64.Madhusudhan, T. et al. Podocyte integrin-β3 and activated protein c coordinately restrict RhoA signaling and ameliorate diabetic nephropathy. J. Am. Soc. Nephrol.31, 1762–1780 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Kang, Y. S. et al. Inhibition of integrin-linked kinase blocks podocyte epithelial–mesenchymal transition and ameliorates proteinuria. Kidney Int.78, 363–373 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Bosquetti, B. et al. The role of α3β1 integrin modulation on fabry disease podocyte injury and kidney impairment. Toxins15, 700 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Dai, R. et al. A vital role for Angptl3 in the PAN-induced podocyte loss by affecting detachment and apoptosis in vitro. BMC Nephrol.16, 38 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Moisy, F. Boxcount. https://www.mathworks.com/matlabcentral/fileexchange/13063-boxcount (2022).
69.Landau, S., Shor, E., Radisic, M. & Levenberg, S. Quantitative image analysis of tissue properties: a MATLAB tool for measuring morphology and co-localization in 2D images. Preprint at 10.1101/2024.04.03.587971 (2024).
70.Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol.15, 1–21 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Jalili, V. et al. The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2020 update. Nucleic Acids Res.48, W395–W402 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Tang, D. et al. SRplot: A free online platform for data visualization and graphing. PLoS One18, e0294236 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Schulze, M. et al. Model testing of pluritest with next-generation sequencing data. Stem Cells Dev.25, 569–571 (2016). [DOI] [PubMed] [Google Scholar]
74.Roost, M. S. et al. KeyGenes, a tool to probe tissue differentiation using a human fetal transcriptional atlas. Stem Cell Rep.4, 1112–1124 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Ge, S. X., Jung, D. & Yao, R. ShinyGO: a graphical gene-set enrichment tool for animals and plants. Bioinformatics36, 2628–2629 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Thomas, P. D. et al. PANTHER: Making genome-scale phylogenetics accessible to all. Protein Sci.31, 8–22 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Mi, H. et al. Protocol Update for large-scale genome and gene function analysis with the PANTHER classification system (v.14.0). Nat. Protoc.14, 703–721 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Wagner, R. et al. High-resolution imaging of kidney vascular corrosion casts with nano-CT. Microsc. Microanal.17, 215–219 (2011). [DOI] [PubMed] [Google Scholar]
79.Ichimura, K. et al. Three-dimensional architecture of podocytes revealed by block-face scanning electron microscopy. Sci. Rep.5, 8993 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Toda, N. et al. Crucial role of mesangial cell-derived connective tissue growth factor in a mouse model of anti-glomerular basement membrane glomerulonephritis. Sci. Rep.7, 42114 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Created in BioRender. Radisic, M. https://BioRender.com/gmz3m4g (2025).
82.Created in BioRender. Radisic, M. https://BioRender.com/tssq2k8 (2025).
83.Created in BioRender. Radisic, M. https://BioRender.com/7e4j3ed (2025).

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(28.5MB, pdf)}

Reporting Summary^{(87.3KB, pdf)}

Supplementary Data 1^{(1.4MB, xlsx)}

Supplementary Data 2^{(36.5KB, xlsx)}

Supplementary Data 3^{(17.6KB, xlsx)}

41467_2025_66037_MOESM6_ESM.pdf^{(27.3KB, pdf)}

Description of Additional Supplementary Files

Transparent Peer Review file^{(3.1MB, pdf)}

Source data^{(22.7MB, xlsx)}

Data Availability Statement

[CR1] 1.Ichimura, K. et al. Morphological process of podocyte development revealed by block-face scanning electron microscopy. J. Cell Sci.130, 132–142 (2017). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Jordanova, E. et al. The fractal and textural analysis of glomeruli in obese and non-obese patients. J. Pathol. Inf.13, 100108 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Kojima, K. et al. Podocyte flattening and disorder of glomerular basement membrane are associated with splitting of dystroglycan-matrix interaction. J. Am. Soc. Nephrol.15, 2079 (2004). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Korolj, A., Wu, H.-T. & Radisic, M. A healthy dose of chaos: using fractal frameworks for engineering higher-fidelity biomedical systems. Biomaterials219, 119363 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Shankland, S. J., Pippin, J. W., Reiser, J. & Mundel, P. Podocytes in culture: past, present, and future. Kidney Int.72, 26–36 (2007). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Yoshimura, Y. et al. Manipulation of nephron-patterning signals enables selective induction of podocytes from human pluripotent stem cells. JASN30, 304–321 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Hale, L. J. et al. 3D organoid-derived human glomeruli for personalised podocyte disease modelling and drug screening. Nat. Commun.9, 1–17 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Hu, M. et al. A biomimetic gelatin-based platform elicits a pro-differentiation effect on podocytes through mechanotransduction. Sci. Rep.7, 43934 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Musah, S. et al. Mature induced-pluripotent-stem-cell-derived human podocytes reconstitute kidney glomerular-capillary-wall function on a chip. Nat. Biomed. Eng.1, 0069 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Petrosyan, A. et al. A glomerulus-on-a-chip to recapitulate the human glomerular filtration barrier. Nat. Commun.10, 3656 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Xie, R. et al. h-FIBER: microfluidic topographical hollow fiber for studies of glomerular filtration barrier. ACS Cent. Sci.10.1021/acscentsci.9b01097 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Korolj, A. et al. Curvature facilitates podocyte culture in a biomimetic platform. Lab a Chip18, 3112–3128 (2018). [DOI] [PubMed] [Google Scholar]

[CR13] 13.Korolj, A. Engineering the fractal properties of podocytes. TSpace. https://tspace.library.utoronto.ca/handle/1807/128350 (2022).

[CR14] 14.Korolj, A. et al. Multiscale fractal shape cues support hierarchical maturation of podocytes via curvature induced extracellular matrix patterning. Tissue Eng. - Part A359, 359 (2022). [Google Scholar]

[CR15] 15.Korolj, A., Radisic, M., Konvalinka, A. & Liu, C. Fractal cues support hierarchical maturation of cells via curvature-induced patterning. US Patent US 2024/0002809 A1 (2024).

[CR16] 16.Shurer, C. R. et al. Physical principles of membrane shape regulation by the glycocalyx. Cell177, 1757–1770.e21 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Nielsen, J. S. et al. The CD34-Related molecule podocalyxin is a potent inducer of microvillus formation. PLoS One2, e237 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Economou, C. G. Enhanced podocalyxin expression alters the structure of podocyte basal surface. J. Cell Sci.117, 3281–3294 (2004). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Nielsen, J. S. & McNagny, K. M. The role of podocalyxin in health and disease. JASN20, 1669–1676 (2009). [DOI] [PubMed] [Google Scholar]

[CR20] 20.Burger, D. et al. Urinary podocyte microparticles identify prealbuminuric diabetic glomerular injury. J. Am. Soc. Nephrol.25, 1401–1407 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Pieuchot, L. et al. Curvotaxis directs cell migration through cell-scale curvature landscapes. Nat. Commun.9, 3995 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Luciano, M. et al. Cell monolayers sense curvature by exploiting active mechanics and nuclear mechanoadaptation. Nat. Phys.17, 1382–1390 (2021). [Google Scholar]

[CR23] 23.Eladl, M. A., M Elsaed, W., Atef, H. & El-Sherbiny, M. Ultrastructural changes and nestin expression accompanying compensatory renal growth after unilateral nephrectomy in adult rats. Int J. Nephrol. Renovasc. Dis.10, 61–76 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Wen, D. et al. Upregulation of nestin protects podocytes from apoptosis induced by puromycin aminonucleoside. Am. J. Nephrol.34, 423–434 (2011). [DOI] [PubMed] [Google Scholar]

[CR25] 25.Suh, J. M. et al. The expression profiles of nuclear receptors in the developing and adult kidney. Mol. Endocrinol.20, 3412–3420 (2006). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Chen, L. et al. DUSP6 protects murine podocytes from high glucose‑induced inflammation and apoptosis. Mol. Med. Rep.22, 2273–2282 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Zhao, X. et al. Glomerular expression of kidney injury molecule-1 and podocytopenia in diabetic glomerulopathy. Am. J. Nephrol.34, 268–280 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Wang, H., Ma, X., Li, S. & Ni, X. SEL1L3 as a link molecular between renal cell carcinoma and atherosclerosis based on bioinformatics analysis and experimental verification. Aging15, 13150–13162 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Shen, C.-Y. et al. Tissue proteogenomic landscape reveals the role of uncharacterized SEL1L3 in progression and immunotherapy response in lung adenocarcinoma. J. Proteome Res.22, 1056–1070 (2023). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Perico, L., Conti, S., Benigni, A. & Remuzzi, G. Podocyte–actin dynamics in health and disease. Nat. Rev. Nephrol.12, 692–710 (2016). [DOI] [PubMed] [Google Scholar]

[CR31] 31.Jalanko, H. & Kääriäinen, H. Chapter 64 - Nephrotic disorders. In Emery and Rimoin’s Principles and Practice of Medical Genetics (6th edn) (eds Rimoin, D., Pyeritz, R. & Korf, B.) 1–13 (Academic Press, Oxford, 2013).

[CR32] 32.Karaiskos, N. et al. A single-cell transcriptome atlas of the mouse glomerulus. J. Am. Soc. Nephrol.29, 2060 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Tran, T. et al. In vivo developmental trajectories of human podocyte inform in vitro differentiation of pluripotent stem cell-derived podocytes. Dev. Cell50, 102–116.e6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Haydak, J. & Azeloglu, E. U. Role of biophysics and mechanobiology in podocyte physiology. Nat. Rev. Nephrol.20, 371–385 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Buzhor, E. et al. Reactivation of NCAM1 defines a subpopulation of human adult kidney epithelial cells with clonogenic and stem/progenitor properties. Am. J. Pathol.183, 1621–1633 (2013). [DOI] [PubMed] [Google Scholar]

[CR36] 36.Abbate, M., Brown, D. & Bonventre, J. V. Expression of NCAM recapitulates tubulogenic development in kidneys recovering from acute ischemia. Am. J. Physiol. Renal. Physiol.277, F454–F463 (1999). [DOI] [PubMed] [Google Scholar]

[CR37] 37.Caza, T. N. et al. Neural cell adhesion molecule 1 is a novel autoantigen in membranous lupus nephritis. Kidney Int.100, 171–181 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Fang, X., Hu, J., Chen, Y., Shen, W. & Ke, B. Dickkopf-3: current knowledge in kidney diseases. Front. Physiol. 11 (2020). [DOI] [PMC free article] [PubMed]

[CR39] 39.Federico, G. et al. Tubular Dickkopf-3 promotes the development of renal atrophy and fibrosis. JCI Insight1, e84916 (2016). [DOI] [PMC free article] [PubMed]

[CR40] 40.Ju, W. et al. Renal gene and protein expression signatures for prediction of kidney disease progression. Am. J. Pathol.174, 2073–2085 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Wakisaka, Y. et al. Cellular distribution of NDRG1 protein in the rat kidney and brain during normal postnatal development. J. Histochem. Cytochem.51, 1515–1525 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Zhang, F. et al. PPBP as a marker of diabetic nephropathy podocyte injury via Bioinformatics Analysis. Biochem. Biophys. Res. Commun.577, 165–172 (2021). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Márquez, E., Riera, M., Pascual, J. & Soler, M. J. Renin-angiotensin system within the diabetic podocyte. Am. J. Physiol. Renal. Physiol.308, F1–F10 (2015). [DOI] [PubMed] [Google Scholar]

[CR44] 44.Harris, R. C. Podocyte ACE2 protects against diabetic nephropathy. Kidney Int.82, 255–256 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Maksimowski, N., Williams, V. R. & Scholey, J. W. Kidney ACE2 expression: implications for chronic kidney disease. PLoS One15, e0241534 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Xu, Z. et al. Ischemia-reperfusion reduces cystathionine-β-synthase-mediated hydrogen sulfide generation in the kidney. Am. J. Physiol. Renal. Physiol.297, F27–F35 (2009). [DOI] [PubMed] [Google Scholar]

[CR47] 47.Naylor, R. W., Morais, M. R. P. T. & Lennon, R. Complexities of the glomerular basement membrane. Nat. Rev. Nephrol.17, 112–127 (2021). [DOI] [PubMed] [Google Scholar]

[CR48] 48.Borza, C. M. & Pozzi, A. The role of cell-extracellular matrix interactions in glomerular injury. Exp. Cell Res318, 1001–1010 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Bülow, R. D. & Boor, P. Extracellular matrix in kidney fibrosis: more than just a scaffold. J. Histochem. Cytochem.67, 643–661 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Brunskill, E. W., Georgas, K., Rumballe, B., Little, M. H. & Potter, S. S. Defining the molecular character of the developing and adult kidney podocyte. PLOS One6, e24640 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Chen, J., Wang, X., He, Q. & Harris, R. C. TAZ is important for maintenance of the integrity of podocytes. Am. J. Physiol. Renal. Physiol.322, F419–F428 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Warde-Farley, D. et al. The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res.38, 214–220 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Shannon, P. et al. Cytoscape: a software environment for integrated models. Genome Res.13, 2498–2504 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.McEvoy, C. M. & Yuen, D. A. Not too much, not too little—just the right amount: the story of YAP in the podocyte. Kidney Int.105, 1157–1159 (2024). [DOI] [PubMed] [Google Scholar]

[CR55] 55.Feng, J. et al. Verteporfin, a suppressor of YAP–TEAD complex, presents promising antitumor properties on ovarian cancer. Onco Targets Ther.9, 5371–5381 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Bonse, J. et al. Nuclear YAP localization as a key regulator of podocyte function. Cell Death Dis.9, 1–14 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Reginensi, A. et al. Yap- and Cdc42-dependent nephrogenesis and morphogenesis during mouse kidney development. PLOS Genet.9, e1003380 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Kang, P. H., Schaffer, D. V. & Kumar, S. Angiomotin links ROCK and YAP signaling in mechanosensitive differentiation of neural stem cells. MBoC31, 386–396 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Malda, J. et al. Chapter 12 - Cell nutrition. In Tissue Engineering (eds Blitterswijk, C. van et al.) 327–362 (Academic Press, Burlington, 2008).

[CR60] 60.Park, J. Y., Lee, D. H., Lee, E. J. & Lee, S.-H. Study of cellular behaviors on concave and convex microstructures fabricated from elastic PDMS membranes. Lab Chip9, 2043–2049 (2009). [DOI] [PubMed] [Google Scholar]

[CR61] 61.Kim, M.-H., Sawada, Y., Taya, M. & Kino-oka, M. Influence of surface topography on the human epithelial cell response to micropatterned substrates with convex and concave architectures. J. Biol. Eng.8, 13 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Bidan, C. M. et al. How linear tension converts to curvature: geometric control of bone tissue growth. PLoS One7, e36336 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.GEO Accession viewer. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE185491 (2022).

[CR64] 64.Madhusudhan, T. et al. Podocyte integrin-β3 and activated protein c coordinately restrict RhoA signaling and ameliorate diabetic nephropathy. J. Am. Soc. Nephrol.31, 1762–1780 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR65] 65.Kang, Y. S. et al. Inhibition of integrin-linked kinase blocks podocyte epithelial–mesenchymal transition and ameliorates proteinuria. Kidney Int.78, 363–373 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Bosquetti, B. et al. The role of α3β1 integrin modulation on fabry disease podocyte injury and kidney impairment. Toxins15, 700 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Dai, R. et al. A vital role for Angptl3 in the PAN-induced podocyte loss by affecting detachment and apoptosis in vitro. BMC Nephrol.16, 38 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR68] 68.Moisy, F. Boxcount. https://www.mathworks.com/matlabcentral/fileexchange/13063-boxcount (2022).

[CR69] 69.Landau, S., Shor, E., Radisic, M. & Levenberg, S. Quantitative image analysis of tissue properties: a MATLAB tool for measuring morphology and co-localization in 2D images. Preprint at 10.1101/2024.04.03.587971 (2024).

[CR70] 70.Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol.15, 1–21 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Jalili, V. et al. The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2020 update. Nucleic Acids Res.48, W395–W402 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Tang, D. et al. SRplot: A free online platform for data visualization and graphing. PLoS One18, e0294236 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Schulze, M. et al. Model testing of pluritest with next-generation sequencing data. Stem Cells Dev.25, 569–571 (2016). [DOI] [PubMed] [Google Scholar]

[CR74] 74.Roost, M. S. et al. KeyGenes, a tool to probe tissue differentiation using a human fetal transcriptional atlas. Stem Cell Rep.4, 1112–1124 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR75] 75.Ge, S. X., Jung, D. & Yao, R. ShinyGO: a graphical gene-set enrichment tool for animals and plants. Bioinformatics36, 2628–2629 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR76] 76.Thomas, P. D. et al. PANTHER: Making genome-scale phylogenetics accessible to all. Protein Sci.31, 8–22 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR77] 77.Mi, H. et al. Protocol Update for large-scale genome and gene function analysis with the PANTHER classification system (v.14.0). Nat. Protoc.14, 703–721 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR78] 78.Wagner, R. et al. High-resolution imaging of kidney vascular corrosion casts with nano-CT. Microsc. Microanal.17, 215–219 (2011). [DOI] [PubMed] [Google Scholar]

[CR79] 79.Ichimura, K. et al. Three-dimensional architecture of podocytes revealed by block-face scanning electron microscopy. Sci. Rep.5, 8993 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR80] 80.Toda, N. et al. Crucial role of mesangial cell-derived connective tissue growth factor in a mouse model of anti-glomerular basement membrane glomerulonephritis. Sci. Rep.7, 42114 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR81] 81.Created in BioRender. Radisic, M. https://BioRender.com/gmz3m4g (2025).

[CR82] 82.Created in BioRender. Radisic, M. https://BioRender.com/tssq2k8 (2025).

[CR83] 83.Created in BioRender. Radisic, M. https://BioRender.com/7e4j3ed (2025).

PERMALINK

Biomimetic fractal topography enhances podocyte maturation in vitro

Chuan Liu

Praful Aggarwal

Karl T Wagner

Shira Landau

Teng Cui

Xin Song

Laleh Shamaei

Naimeh Rafatian

Yimu Zhao

Sonia Rodriguez-Ramirez

Keith Morton

Elizabeth Virlee

Chen Yu Li

Dawn Bannerman

Simon Pascual-Gil

Sargol Okhovatian

Anastasia Radisic

Sergi Clotet-Freixas

Teodor Veres

Mohtada Sadrzadeh

Tobin Filleter

Ulrich Broeckel

Ana Konvalinka

Milica Radisic

Abstract

Introduction

Results

Native glomerulus and podocyte demonstrate fractal patterns

Fig. 1. Fractal metrics characterize the architecture of podocytes and glomeruli in healthy and diseased states.

Fig. 2. Fractal topography increases expression of slit diaphragm markers in mouse podocytes.

Fig. 7. YAP signaling pathway plays a role in maintaining mouse podocyte morphology on fractal substrates.

Fractal topography increases expression of slit diaphragm markers

Fractal topography elicits enhanced signs of cell polarity

Fig. 3. Fractal topographic patterning promotes signs of polarity in mouse podocytes.

Biomimetic fractal topography influences cytoskeleton arrangement

Fig. 4. Mouse podocytes exhibit enhanced in vivo-like morphology on biomimetic fractal topography with multiscale curvature versus equally sharp non-fractal ordered topography with a narrow curvature range.

Mosaic assay reveals single-cell morphology of podocytes on fractal substrates, supporting drug testing

Fig. 5. Fractal topographical substrates can be incorporated into a multi-well plate format for high-throughput applications.

RNA sequencing points to podocyte maturation on fractal substrates

Fig. 6. RNA sequencing data suggest that fractal topography modulates biological processes that include ECM organization, cell adhesion and kidney development in mouse podocytes.

YAP signaling pathway plays a role in maintaining podocyte morphology on fractal substrates

Discussion

Methods

Extraction of histological features of the glomerulus

Fractal dimension analysis

FFT analysis

Topographical pattern fabrication

PDMS substrate fabrication

Substrate pattern characterization

Podocyte culture

Substrate preparation for cell seeding

Podocyte seeding

Glycocalyx imaging

Immunofluorescent staining

Immunofluorescent imaging and analysis

Mosaic assay

EV isolation

Western blotting

Transmission electron microscopy of EVs

Fractal topographic well plate fabrication

PAN injury model

DPSC culture and seeding

RNA isolation

Quantitative polymerase chain reaction

Bulk RNA sequencing of E11 mouse podocytes

PCA and Pluritest analysis of RNA-seq data from primary fetal human podocytes

KeyGenes analysis of RNA-seq data from primary fetal human podocytes

Heatmaps and clustering of RNA-seq data from primary fetal human podocytes

GO analyses of RNA-seq data from primary fetal human podocytes

YAP inhibition study

YAP1 interactome

Statistics

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review