Immortalized human podocytes were obtained from Dr. Moin Saleem (Saleem et al., 2002). Podocytes were cultured in RPMI (Wisent Inc., 350–000-CL) supplemented with 10% FBS and 1% penicillin/streptomycin (PS) and were maintained under permissive conditions (33 °C) with 5% CO₂. HEK293 cells were maintained in DMEM containing 10% FBS and 1% PS at 37 °C with 5% CO₂.

Proximity-dependent biotin identification (BioID) bait constructs were gifts from Drs. Anne-Claude Gingras (Lunenfeld-Tanenbaum Research Institute) and Jean-Francois Coté (Montreal Clinical Research Institute). Wild-type Rho GTPases (RhoA WT, Rac1 WT, and Cdc42 WT), nucleotide-free (NF) mutants (RhoAG17A, Rac1G15A, and Cdc42G15A), and constitutively active mutants (RhoAG14V, Rac1G12V, and Cdc42G12V) were subcloned into pSTV6 (gift from Dr. Anne-Claude Gingras), a tetracycline-inducible lentiviral vector that contains BirA with puromycin N-acetyltransferase (PAC) as a reporter. Empty MYC-BirA vector and GFP subcloned into pSTV2-N-BirA*-Flag were used as negative controls (Samavarchi-Tehrani et al., 2018).

He et al. (2021) dataset: RPKM data of scRNA-seq on human glomeruli from the He et al. (2021) dataset (GSE160048) were downloaded from the GEO database, and podocytes were identified based on the expression of NPHS1 and NPHS2 genes. Muto et al. (2021) dataset: for the single-nucleus RNA sequencing (snRNA-seq) data from healthy kidneys, normalized Seurat objects were downloaded from the GitHub repository. Diabetic kidney disease dataset: log2 expression data were extracted from Wilson et al. (2019). Human adult kidney dataset: the expression levels of BioID-identified GEFs were examined in various human kidney cell types using the human adult kidney data from Subramanian et al. (2019), available on the Single Cell Portal (https://singlecell.broadinstitute.org/single_cell).

Podocytes were cultured in a 96-well IncuCyte® ImageLock microplate coated with collagen at 33 °C and then serum-starved at 37 °C in RPMI containing 1% FBS for 2 h before wound induction. Confluent monolayers were scratched using the IncuCyte 96-well Wound Maker (Sartorius, Ann Arbor, MI, USA), and cells were kept at 37 °C up to 24 h. The podocyte migration rate was calculated using the percentage of wound confluence generated by IncuCyte Analysis Software (Sartorius, Ann Arbor, MI, USA). The percentage of the scrambled control was calculated based on the 6- to 8-hour time points.

Podocytes were cultured in 12-well plates on day 0. On day 1, podocytes were treated with EGF (100 ng) and placed in a 37 °C incubator. Ten hours after EGF treatment, 16 snapshots of each well were taken using IncuCyte S3 (Essen BioScience, Ann Arbor, MI, USA). Cellular projections were counted manually and normalized to the percentage of cell confluence, as calculated by IncuCyte software.

Mouse glomeruli were isolated through differential sieving, as described previously (Zhu et al., 2008), and lysed in IP buffer for 30 min on ice. Proteins were eluted in Laemmli buffer and then assayed by Western blot.

HEK293 cells were transfected with the KIAA1522-mCherry plasmid for 16 h and then lysed in IP buffer. Cell lysates were incubated with KIAA1522 antibody or rabbit IgG overnight at 4 °C. Protein A agarose beads (Santa Cruz, sc-2001, Dallas, TX, USA) were added for 1 h at 4 °C on rotation. Beads were washed three times, and proteins were eluted in Laemmli buffer and then assayed by Western blot.

Cells were lysed in TNE buffer. Cell lysates were incubated with streptavidin beads fused to GST-empty or GST-IRSp53 for 1 h at 4 °C on rotation. Beads were washed three times, and proteins were eluted in Laemmli buffer and then assayed by Western blot.

Cells were lysed in IP buffer. Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes; the membranes were then blocked with 5% BSA and incubated with primary antibodies at 4 °C overnight. Next, the membranes were washed three times and then incubated with secondary antibodies for 1 h at room temperature. Quantitative densitometry was performed using ImageJ.

Three podocyte scrambled controls and three KIAA1522 KOs were generated using CRISPR guides listed in Supplementary Table S1C. KO efficiency was validated by immunoblotting (Supplementary Figure S8). RNA extraction was performed using an RNeasy Mini Kit from QIAGEN, Germantown, MD, USA. The reads were trimmed with fastp and aligned using the STAR aligner. Raw read counts were obtained with HTSeq. Batch effects were corrected using the sva R package. The DESeq2 R package was used to normalize counts and perform differential expression (DE) analysis between the conditions. Gene set enrichment analysis (GSEA) was performed using the clusterProfiler R package.

Sixty thousand podocytes were cultured on collagen-coated coverslips in 12-well plates (collagen type I (Sigma)) and then differentiated at 37 °C for 5–7 days. Cells were fixed with 4% PFA in PBS and permeabilized using 0.5% Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA) in PBS. Following blocking with 3% bovine serum albumin in PBS, cells were immunostained with the respective antibodies and stained with phalloidin. Images of the cells were taken using a Zeiss LSM 780, Oberkochen, Baden-Württemberg, Germany Laser Scanning Confocal Microscope. Cell size was quantified using ImageJ software. Vinculin count was quantified using the “analyze particle” command in ImageJ. Particle size was selected between 1 µm²–8 µm². The filopodia localization ratio of KIAA1522 and IRSp53 was quantified using ImageJ software and calculated as (integrated density in the whole cell − integrated density in the inner cell body)/integrated density in the whole cell, following an approach analogous to the one we previously used to assess membrane localization of a protein (Matsuda et al., 2022).

Podocytes were either untreated or treated with EGF (100 ng/mL) for 5 min. Lysates were prepared, and GLISA was conducted according to the manufacturer’s instructions. Basal RhoA activity (unstimulated) of KO lines was calculated as the percentage of basal RhoA activity in the scrambled control.

Zebrafish AB and Tupfel long fin (TL) lines were used for KIAA1522 and ARHGEF12 experiments, respectively. Zebrafish were maintained at 28.5 °C in E3 media (NaCl = 0.287 g/L, KCl = 0.0132 g/L, CaCl₂·2H₂O = 0.0479 g/L, MgCl₂·6H₂O = 0.0807 g/L, 0.00005% methylene blue, pH 7.2). Antisense morpholino oligonucleotides (MOs) were designed for translation blocking and purchased from Gene Tools (Gene Tools LLC, Philomath, OR, United States). Working MO dilutions were prepared in water (at 150 µM for KIAA1522 and 250 µM for arhgef12a + arhgef12b) and heated at 65 °C for 5 min to disrupt potential secondary structures. Zebrafish embryos were microinjected into the yolk with the MO solution at the single-cell stage. Injected embryos were maintained in E3 media under standard conditions until 5 days post-fertilization (5 dpf). Live images were taken at 3 and/or 5 dpf while embryos were under anesthesia using Tricaine (168 mg/L).

Zebrafish were fixed in 2% glutaraldehyde. Transmission electron micrographs of glomerular tufts were captured using a 120 kV Hitachi H-7650 Transmission Electron Microscope (Hitachi, Tokyo, Japan).

The antibodies and reagents are listed in Supplementary Table S2.

To map the interactome of Rho GTPase in human podocytes, we performed BioID analysis in in vitro-cultured immortalized podocytes. We used baits representing different nucleotide loading statuses of the three prototypical Rho GTPases: the active, GTP-bound forms (RhoAG14V, Rac1G12V, and Cdc42G12V), the NF forms that bind GEFs with high affinity (RhoAG17A, Rac1G15A, and Cdc42G15A), and WT, which cycles between GTP- and GDP-bound forms. The nine baits revealed 1,927 significant protein–protein interactions in podocytes (Supplementary Table S3). We first compared the hits captured by the active baits with those from WT baits and found that 40%–59% of the interactions were unique for active baits (Figure 1A). Similarly, comparing the hits captured by the NF baits to those from WT baits revealed that 24%–48% of NF baits’ interactions were unique (Figure 1B). Thus, our approach of using baits with distinct nucleotide loading statuses allowed the sensitive detection of specific interactors, enabling the discovery of distinct groups of interactors, including potential activators, inactivators, and effectors.

To assess the cell-type specificity of Rho GTPase interactions, we next compared our results to previously published datasets from HEK293 and HeLa cells using a similar BioID approach (Bagci et al., 2020). Among the six baits (active and NF), 11%–25% of all the hits were common between podocytes and HEK293/HeLa cells, 7%–21% of the hits were unique to podocytes, and 56%–82% of the hits were unique to HEK293/HeLa cells (Figure 1C; Supplementary Table S4). Notably, approximately 50% of the Rho GTPase interactors identified in podocytes were absent in HEK293 and HeLa cells. This suggests the existence of podocyte-specific Rho GTPase regulatory mechanisms and signaling networks.

To gain insights into the functional roles of active bait interactors, we next performed Gene Ontology (GO) enrichment analysis as these interact or are expected to interact with downstream effectors. Our analysis revealed a significant enrichment of GO biological processes (GO-BP) related to cell morphology and cytoskeleton regulation, including cell–matrix interaction, actin filament-based process, and cell–cell adhesion (Figure 2A; Supplementary Figure S1A). Further analysis of the GO cellular components (GO-CC) indicated that these interactors are enriched in key pathways involved in cellular structures such as stress fiber, lamellipodium, filopodium, and focal adhesion (Figure 2B; Supplementary Figure S1B). Moreover, the identified hits included known interactors of Rho GTPases. For instance, active RhoA interacts with ROCK1, ROCK2, DIAPH2, and DIAPH3, which are essential for stress fiber formation (Shi et al., 2013; Watanabe et al., 1999). Similarly, active Rac1 interacted with ABI1, NCKAP1, and CYFIP2, which are key components in WAVE complexes and crucial for lamellipodia formation (Rottner et al., 2021). Furthermore, active Cdc42 identified WIPF1, WIPF2, WASL, and BAIAP2, all of which contribute to filopodia formation (Derivery and Gautreau, 2010) (Supplementary Table S3). In addition, by using WD scores to measure the hit specificity, we identified the top 30 interactors for each active bait, revealing potentially critical components and effectors of their respective signaling pathways (Supplementary Figure S2).

Together, these results uncover the interactomes of Rho GTPases in human podocytes and suggest that Rho GTPase signaling in podocytes is closely linked to cytoskeletal dynamics. Our findings are consistent with established Rho GTPase functions, further validating the robustness of our approach.

Our BioID results identified KIAA1522, a protein with a minimally characterized function, as an interactor of active Rac1 and Cdc42. Since the role of KIAA1522 in cytoskeletal regulation had not been explored, we sought to investigate its function in podocytes. First, we investigated the expression pattern of KIAA1522. The overexpression of mCherry-tagged KIAA1522 in HEK293 cells revealed membrane localization and co-localization with F-actin (phalloidin) (Figure 3A). Immunofluorescence staining of KIAA1522 in immortalized human podocytes also showed a clear localization at the plasma membrane, particularly at cell–cell junctions (Figure 3B). In addition, immunofluorescence staining of human kidney sections showed that KIAA1522 is expressed in the glomerulus, with partial co-localization with the podocyte-specific transmembrane protein nephrin (Figure 3C).

To identify potential interacting partners of KIAA1522 and gain mechanistic insight into its role, we searched the BioGRID database for potential KIAA1522 interactors predicted by Affinity Capture-MS (https://thebiogrid.org/). BioGRID identified IRSp53, a key adaptor protein involved in filopodia formation (Krugmann et al., 2001), and WAVE/SCAR, a Rac1 effector implicated in lamellipodia formation (Abou Kheir et al., 2005), as putative interactors of KIAA1522. We next performed a pull-down assay, which confirmed the interaction between KIAA1522 and IRSp53 (Figure 4A). Immunoprecipitation further demonstrated that KIAA1522 interacts with WAVE/SCAR (Figure 4B). This suggests that KIAA1522 may play an active role in the dynamics of the actin cytoskeleton.

Notably, our BioID data showed that, similar to KIAA1522, IRSp53 interacted with Cdc42 and Rac1 (Supplementary Table S3). IRSp53 promotes filopodia formation by acting as the Cdc42 effector, where the latter causes conformational changes to IRSp53, enabling Mena recruitment and subsequent actin filament assembly (Krugmann et al., 2001). In podocytes, the formation of Cdc42: IRSp53: Mena complexes was found to be suppressed by synaptopodin, which is a key protein known for its role in maintaining podocyte integrity and protecting against proteinuria (Yanagida-Asanuma et al., 2007). To investigate whether KIAA1522 acts as a Cdc42 effector in podocytes, we examined whether the Cdc42 activation status induced its specific recruitment. For this, we overexpressed active and dominant-negative forms of Cdc42 in podocytes and assessed KIAA1522 localization by immunofluorescence. When the active Cdc42 was expressed, KIAA1522 was specifically recruited to Cdc42-induced filopodia. In contrast, when the dominant-negative form of Cdc42 was expressed, KIAA1522 was strictly localized in the cytoplasm (Figure 4C). These results suggest that KIAA1522 may function in the filopodium complex at the plasma membrane downstream of Cdc42, possibly via IRSp53. To test this further, we used pooled CRISPR KO podocyte lines (see section 2.11). When KIAA1522 KO podocytes were transfected with active Cdc42, IRSp53 localization to filopodia was reduced significantly compared to that of the control podocytes (Figure 4D). These findings support that KIAA1522 interacts with IRSp53 and likely facilitates its relocalization to the plasma membrane, thereby contributing to filopodium formation (Figure 4E).

We next examined the response of KIAA1522 KO podocytes to EGF stimulation, which activates Rho GTPases and induces cell motility. KIAA1522 KO podocytes exhibited a reduced cell projection formation following EGF treatment (Figure 5A). In addition, KIAA1522 KO significantly decreased podocyte migration, with an average migration rate of 74% compared to that in the control (p-value = 0.0011; Figure 5B). Together, these findings confirm that KIAA1522 plays a role in cytoskeletal regulation and podocyte motility. Next, to investigate the signaling pathways associated with KIAA1522, we performed RNA sequencing of KIAA1522 KO HPs and controls. While differential expression (DE) analysis showed significant but modest changes at the gene level (Supplementary Tables S5, S6), we utilized GSEA to identify broader pathway-level alterations. GSEA of GO pathways revealed a downregulation of genes associated with the RNAi effector complex pathway and the RISC complex compared to that in the control (Figure 5C, left panel). In addition, GSEA of the Reactome pathways showed a downregulation of genes involved in extracellular matrix organization (Figure 5C, right panel). These findings suggest that KIAA1522 may influence post-transcriptional regulation and play a role in regulating cell motility, likely by modulating actin rearrangement, promoting filopodia and cell projection formation, and indirectly affecting the extracellular matrix composition.

To investigate the functional role of KIAA1522 in vivo, we transiently knocked down its homolog in zebrafish larvae, a well-established model for studying podocyte biology and screening for proteinuric phenotypes (Schindler and Endlich, 2024). Between 62% and 65% of KIAA1522 morphants (n = 69) exhibited pericardial effusion/edema starting at 3 dpf compared with 13%–18% of control morphants (n = 46). Yolk sac edema was also observed in KIAA1522 morphants (Figure 6A). This pericardial edema phenotype may indicate proteinuria as similar phenotypes have been observed following knockdown of proteins critical for glomerular function, including nephrin, podocin, and Apol1 (Fukuyo et al., 2014; Kotb et al., 2016; Lee et al., 2022). Notably, pericardial edema has also been reported in the metronidazole-induced podocyte injury model (Siegerist et al., 2017). Next, we analyzed the ultrastructure of the glomerular filtration barrier of these morphants by transmission electron microscopy (TEM). We identified significant foot process effacement of podocytes in KIAA1522 morphants compared to that in the control, indicating podocyte injury and subsequent disruption of the glomerular function (Figure 6B). Foot process effacement is a biomarker of proteinuric diseases, and thus, our findings reveal that KIAA1522 is crucial for maintaining podocyte integrity and normal glomerular function.

Rho GTPase activities are regulated by three groups of proteins: 1) activator GEFs (82 members), 2) inactivator GAPs (69 members), and 3) guanine nucleotide dissociation inhibitors (GDIs), which bind to the GDP-bound inactive forms, maintaining a pool of inactive Rho GTPases (three members). The interplay between these numerous GEFs/GAPs and the 21 Rho GTPase members creates the complexity of the Rho GTPase signaling network. To dissect this network in podocytes, we analyzed our BioID results, focusing on GEFs and GAPs. The NF baits are known to bind to GEFs with high affinity and are commonly used to identify GEFs for a specific Rho GTPase, whereas the active baits are more likely to interact with GAPs. Additionally, WT control baits are expected to interact with GEFs and GAPs at lower affinity than their mutated counterparts.

Our results showed that NF-RhoA (RhoAG17A) interacted with 11 GEFs, while NF-Rac1 (Rac1G15A) and NF-Cdc42 (Cdc42G15A) interacted with 8 and 5 GEFs, respectively. Although some GEFs exhibited specificity for a single Rho GTPase, others were promiscuous. Specifically, eight GEFs were unique to RhoA (ECT2, ARHGEF5, ARHGEF2, ARHGEF18, ARHGEF12, ARHGEF11, AKAP13, and ARHGEF1), two were specific to Cdc42 (DNMBP and DOCK11), and three were specific to Rac1 (FGD6, ARHGEF6, and DOCK1). In contrast, VAV2 interacted with both RhoA and Rac1, ARHGEF7 and DOCK9 were shared between Rac1 and Cdc42, and ARHGEF26 interacted with all three Rho GTPases (Figure 7A, right panel). Notably, the majority of the identified GEFs interacted exclusively with the NF forms, which is consistent with previous reports showing that the NF forms have a high affinity for GEFs. On the other hand, ABR and ARHGEF40 interacted only with WT-Rac1 and WT-Cdc42, respectively, although the reason and significance of this unexpected interaction remain unclear. The GAPs detected with the active baits were fewer in number than the GEFs detected with the NF baits, and they were found to interact with the WT forms. In addition, GAPs were less specific to a single Rho GTPase, with five GAPs interacting specifically with one Rho GTPase and five others interacting with multiple Rho GTPases (Figure 7A, left panel). Overall, the BioID results identified the panel of GEFs and GAPs that interact with RhoA, Rac1, and Cdc42 in human podocytes.

BioID experiments yielded numerous GEFs and GAPs as potential interactors of Rho GTPases in podocytes. To refine the focus of our study, we analyzed publicly available transcriptomic databases to assess the expression profile of GEFs and GAPs in human podocytes. First, we examined the expression of 21 Rho GTPases, 69 Rho GAPs, and 82 Rho GEFs in two independent single-cell and single-nucleus RNA sequencing (scRNA-seq and snRNA-seq) datasets from healthy kidney tissues. In both datasets, podocytes expressed the three prototypical Rho GTPases (RHOA, RAC1, and Cdc42) along with two additional members, RHOQ and RHOBTB2 (Supplementary Figure S3). Among the top 20 most abundantly expressed GEFs in podocytes from each dataset, 10 were common in both datasets. Similarly, 12 GAPs were commonly ranked among the top 20 most abundant GAPs in both datasets (Figure 7B, left panel; Supplementary Figures S4, S5). To identify the dominant Rho GTPase signaling pathways in podocytes, we analyzed the interaction of the identified 10 GEFs and 12 GAPs with a comprehensive panel of Rho GTPases using the STRING database. Our analysis revealed that six or more of these GEFs and GAPs interacted with RhoA, Cdc42, and Rac1, which represented the most frequent interactions (Figure 7B, right panel).

Specifically, RhoA interacted with five GEFs and ten GAPs, Cdc42 interacted with five GEFs and six GAPs, and Rac1 interacted with two GEFs and four GAPs. Notably, two RhoA GEFs (ARHGEF3 and ARHGEF26) also interacted with RhoB and RhoC (which are structurally similar to RhoA) (Wheeler and Ridley, 2004) and with RhoG (which resembles Rac1 and Cdc42). Overall, the high abundance of RhoA modulators suggests that RhoA is a major Rho GTPase in healthy human podocytes.

Next, to determine whether the expression levels of identified GEFs and GAPs are altered in human podocytes under kidney disease conditions, we analyzed the snRNA-seq dataset from Wilson et al. (2019), which compares proteinuric and diabetic kidneys to healthy controls. Notably, the RhoA-targeting GAP, DLC1, was significantly upregulated in podocytes from diabetic kidneys (log2 FC: 0.39, p-value = 1.01E-09). Moreover, DLC1 expression was significantly higher in podocytes from proteinuric DKD patients than in those from non-proteinuric DKD patients (log2 FC = 0.56, p-value = 2.1E-07). Although the significance of these changes remains to be determined, they support the notion that regulation of RhoA activity in podocytes is crucial in both health and disease.

Based on the BioID results and database analysis, we proceeded to screen the function of 11 RhoA-targeting GEFs identified by BioID in human podocytes in vitro. Using CRISPR/Cas9, we generated pooled KO lines for each RhoA-GEF gene in immortalized human podocytes (Supplementary Figure S6A, B).

We first studied the impact of each KO on basal RhoA activity using GLISA, which revealed that the knockout of 9 out of 11 RhoA GEFs reduced basal RhoA activity. Among these, ARHGEF12 KO had the most pronounced effect on RhoA activity, followed by ARHGEF1 KO and ARHGEF11 KO (Supplementary Figure S6C). In addition, the depletion of VAV2, DOCK10, ARHGEF12, ARHGEF11, and ARHGEF1 resulted in reduced cell size (Supplementary Figure S6D).

Interestingly, analysis of a healthy human kidney dataset from the Broad Institute’s Single-Cell Portal revealed that ARHGEF12 and ARHGEF26 are more specifically enriched in podocytes than in other kidney cell types (Supplementary Figure S6E). In addition, ARHGEF12 had the highest average spectral counts (AvgSpec) among the RhoA GEFs interacting with NF RhoA in our BioID results (Supplementary Figure S6B). ARHGEF12 was also among the most highly expressed GEFs in podocytes from the two independent primary single-cell transcriptomics datasets examined above (Figure 7B).

Given its prominent expression profile and activity, we chose to further investigate the role of ARHGEF12 in podocytes. First, we validated ARHGEF12 protein expression by immunoblotting in human podocytes in vitro and mouse glomerular lysates (Figure 8A). Immunofluorescence staining confirmed ARHGEF12 expression in the mouse glomerulus, with partial co-localization observed between ARHGEF12 and nephrin in podocytes (Figure 8B). Functionally, ARHGEF12 KO cells exhibited a reduced migration rate, as determined by the wound healing assay (Figure 8C).

In addition, ARHGEF12 KO cells displayed an increase in vinculin complex number per cell area, indicating that ARHGEF12 may play a role in the turnover and/or maturation of focal adhesion complexes (Figure 8D). This increase in focal adhesion at both the leading and trailing edges likely impairs cell motility by preventing the cells from forming new complexes or detaching properly. Live imaging of podocytes transfected with mCherry-paxillin further demonstrated impaired cell motility and revealed the abundance of small and persistent focal adhesions in ARHGEF12 KO cells compared to those in the control (Supplementary Videos S1, S2).

Finally, to investigate the functional role of ARHGEF12 in vivo, we transiently knocked-down ARHGEF12 homologs (arhgef12a and arhgef12b) in zebrafish larvae. This induced pericardial edema in zebrafish at 5 dpf in 75%–87% ARHGEF12 morphants (n = 190) compared with 2%–14% of control morphants (n = 182) (Figure 9A). Analysis of the ultrastructure of the glomerular filtration barrier by TEM revealed foot process effacement of podocytes in ARHGEF12 morphants compared to that in the control (Figure 9B). Together, these results indicate that ARHGEF12 is the main RhoA-GEF in podocytes and is crucial for their integrity and motility.