Indirect podocyte injury manifested in a partial podocytectomy mouse model

Abstract

In progressive glomerular diseases, segmental podocyte injury often expands, leading to global glomerulosclerosis by unclear mechanisms. To study the expansion of podocyte injury, we established a new mosaic mouse model in which a fraction of podocytes express human (h)CD25 and can be injured by the immunotoxin LMB2. hCD25⁺ and hCD25⁻ podocytes were designed to express tdTomato and enhanced green fluorescent protein (EGFP), respectively, which enabled cell sorting analysis of podocytes. After the injection of LMB2, mosaic mice developed proteinuria and glomerulosclerosis. Not only tdTomato⁺ podocytes but also EGFP⁺ podocytes were decreased in number and showed damage, as evidenced by a decrease in nephrin and an increase in desmin at both protein and RNA levels. Transcriptomics analysis found a decrease in the glucocorticoid-induced transcript 1 gene and an increase in the thrombospondin 4, heparin-binding EGF-like growth factor, and transforming growth factor-β genes in EGFP⁺ podocytes; these genes may be candidate mediators of secondary podocyte damage. Pathway analysis suggested that focal adhesion, integrin-mediated cell adhesion, and focal adhesion-phosphatidylinositol 3-kinase-Akt-mammalian target of rapamycin signaling are involved in secondary podocyte injury. Finally, treatment of mosaic mice with angiotensin II receptor blocker markedly ameliorated secondary podocyte injury. This mosaic podocyte injury model has distinctly demonstrated that damaged podocytes cause secondary podocyte damage, which may be a promising therapeutic target in progressive kidney diseases.

NEW & NOTEWORTHY This novel mosaic model has demonstrated that when a fraction of podocytes is injured, other podocytes are subjected to secondary injury. This spreading of injury may occur ubiquitously irrespective of the primary cause of podocyte injury, leading to end-stage renal failure. Understanding the molecular mechanism of secondary podocyte injury and its prevention is important for the treatment of progressive kidney diseases. This model will be a powerful tool for studying the indirect podocyte injury.

INTRODUCTION

Accumulating evidence has shown the importance of podocyte injury for the initiation and progression of kidney diseases. Podocyte injury can be caused by various factors, including mechanical stress, oxidative stress, and immunological stress (1). Irrespective of primary causes, once a substantial number of podocytes are lost, the kidneys are irreversibly and progressively injured, showing glomerulosclerosis, which eventually leads to end-stage renal failure (2–4).

We have previously established a mouse model of selective podocyte injury, NEP25 mice (5). In this line, hCD25 is selectively expressed in podocytes, and injection of human (h)CD25-targeting immunotoxin LMB2 induces podocyte injury in a dose-dependent manner. LMB2 damages cells by inhibiting protein translation similarly to diphtheria toxin, which is used in other podocyte injury models (3, 6). Although LMB2 is rapidly cleared from the circulation, progressive podocyte injury over a few weeks occurs after a single injection of LMB2, suggesting the presence of an autonomous deterioration mechanism in podocyte injury. We speculated that injured podocytes hurt other neighboring podocytes. To test this hypothesis, we then generated chimeric mice by aggregating embryos of NEP25 and ROSA(R)26-human placental alkaline phosphatase (hPAP) transgenic mice (7). In NEP25 ↔ R26-hPAP chimeric mice, these two types of podocytes mingle within the same glomerulus. We found that not only hCD25⁺ podocytes but also hPAP⁺ podocytes, which do not carry the hCD25 gene, were secondarily injured after LMB2 injection. This study clearly demonstrated that podocyte damage causes further damage to podocytes. However, the ratio of hCD25⁺ podocytes in each chimeric mouse is determined by chance and not controllable. In addition, as the chimeric mice cannot be reproduced by mating, de novo generation by aggregation is necessary.

In this study, we aimed to establish a new reproducible mosaic mouse model in which a fraction of podocytes can be injured.

MATERIALS AND METHODS

Animal Ethics

Animal experimental protocols were approved by the Animal Experimentation Committee of Tokai University School of Medicine and the Institutional Animal Care and Use Committee of Cold Spring Harbor Laboratory and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Establishment of the hCD25-tdTomato/Enhanced Green Fluorescent Protein Mosaic Model

The targeting construct DNA carrying hCD25, tdTomato, enhanced green fluorescent protein (EGFP), and two types of recombination sites, loxP and lox2272, was introduced into embryonic stem cells of the C57BL/6J line by homologous recombination, and the hCD25-tdTomato/EGFP mouse line was established (Fig. 1A). This line was mated with the Nphs1-Cre transgenic line (8). Hereafter, the double transgenic mice were designated as “Pod-hCD25T/G mosaic mice” or simply “mosaic mice,” since the podocytes of these mice actually showed a mosaic pattern, as they expressed either EGFP or hCD25 and tdTomato.

Figure 1.

Establishment of the mosaic mouse line. A: transgene structures. The targeting construct DNA was introduced into embryonic stem cells by homologous recombination, and the human (h)CD25-tdTomato/enhanced green fluorescent protein (EGFP) mouse line was established. This line carries the promoterless neomycin resistance gene (neo); the EGFP, chloramphenicol acetyltransferase (CAT), hCD25, and tdTomato genes; and an internal ribosome entry site (IRES) in the first intron of the ROSA26 gene. Neo and EGFP-CAT are flanked by lox2272 and loxP sites. The hCD25-tdTomato/EGFP line was mated with the nephrin (Nphs1)-Cre transgenic line, generating Pod-hCD25T/G mosaic mice, which have three types of podocytes. Recombination between the two lox2272 sites generates EGFP-expressing podocytes, whereas recombination between the two loxP sites generates hCD25- and tdTomato-expressing podocytes. These two types of recombination occur randomly and are mutually exclusive. In a minor portion of podocytes, no recombination occurs, generating double-negative podocytes. Ex1, first exon; pA, polyadenylation site; SA, splice acceptor site. B: mosaic distribution of EGFP⁺ and tdTomato⁺ podocytes. In the Pod-hCD25T/G mosaic model, EGFP⁺ and tdTomato⁺ podocytes mingle in a mosaic pattern within the same glomerulus. Scale bars = 20 μm. C: podocyte-specific expression of EGFP and tdTomato. EGFP and tdTomato signals overlapped with staining for nephrin but not for CD31 or desmin. Scale bars = 20 μm. D: flow cytometric analysis of glomerular cells of Pod-hCD25T/G mosaic mice. Among the podocytes identified by podocin staining, 36.2% [interquartile range (IQR): 35.6%–38.7%] were tdTomato⁺ EGFP⁻, 44.9% (IQR: 43.4%–45.7%) were tdTomato⁻ EGFP⁺, and the rest were tdTomato⁻ EGFP⁻. None of the podocin⁻ cells expressed tdTomato or EGFP.

Animal Experiments

Podocyte injury was induced by a single injection of LMB2 (25 ng/g body wt) 2 wk after uninephrectomy under anesthesia with 60 µg/g body wt pentobarbital sodium and 50 ng/g body wt buprenorphine. A lower dose of LMB2 did not cause podocyte injury. Before or after LMB2 injection, mosaic mice were euthanized, and the left kidneys were harvested. Concentrations of albumin and creatinine in 24-h urine were determined by an immunonephelometric method and an enzymatic method, respectively, in an outside laboratory (SRL, Tokyo, Japan).

Male mosaic mice aged 14–34 wk and age-matched wild-type control mice were used for short-term (21-day) longitudinal urinalyses and histological analyses. Female mosaic mice aged 8 wk were used for long-term (120-day) longitudinal urinalyses and histological analyses. Male mosaic mice aged 8 wk were used for flow cytometry analyses, gene expression arrays, and quantitative RT-PCR analyses. To test the effect of angiotensin II receptor blockers (ARBs), we divided male mosaic mice aged 28–34 wk into two groups (each n = 6). One group was treated with losartan (0.5 g/L in the drinking water supplemented with 1% sucrose to mask the taste), and the other was treated with 1% sucrose in the drinking water from 1 day after LMB2 injection until the end of the experiment (day 14).

Podocyte Isolation

Glomerular cells were isolated by a method previously reported by Boerries et al. (9) with some modifications. Kidneys were perfused through the abdominal aorta with Dynabeads (ThermoFisher Scientific, Waltham, MA) suspended in 1 mL of an enzymatic digestion buffer [300 U/mL collagenase type II (Worthington, Lakewood, NJ), 5 U/mL pronase E (P6911, Sigma Aldrich, St. Louis, MO), and 50 U/mL DNase I (No. 11284932001, Roche, Basel, Switzerland) in HBSS]. Kidneys were dissected, minced, incubated in a digestion buffer at 37°C for 15 min on a rotator (100 rpm), and filtered twice through a 100-µm cell strainer. The glomeruli were then harvested with a neodymium magnet and digested on a thermomixer (Eppendorf, Hamburg, Germany) with shaking at 1,400/min for 35 min at 37°C in the digest buffer. After Dynabeads had been removed with the magnet, glomerular cells were harvested through a 40-µm cell strainer.

Approximately 20% of the obtained glomerular cells were stained for podocin using rabbit anti-podocin antibody (P0372, Sigma Aldrich, 1:200), DyLight 649-conjugated anti-rabbit antibody (No. 406406, BioLegend), and Dako IntraStain (K2311, Dako, Glostrup, Denmark) and analyzed with a FACSAria system (BD Biosciences, Franklin Lakes, NJ) to determine the ratio of tdTomato⁺ to EGFP⁺ podocytes. The remaining unstained glomerular cells were sorted into tdTomato⁺ or EGFP⁺ podocytes for DNA and mRNA analyses.

Histological Analysis

For the generation of paraffin sections, kidney tissues were fixed with 4% paraformaldehyde overnight, embedded in paraffin, and cut at 2 μm thickness. For the generation of frozen sections, kidney tissues were fixed with 4% paraformaldehyde for 3 h, incubated in PBS containing 30% sucrose, embedded in optimal cutting temperature (OCT) compound, frozen, and cut at 6 μm thickness. Antibodies against the following were used at the indicated dilution: desmin (M0760, Dako, 1:50, for Fig. 1C, Fig. 2G, and Fig. 4D; ab32362, Y66, Abcam, Cambridge, UK, 1:100, for Fig. 2H), nephrin (Progen GP-N2, Heidelberg, Germany, 1:100, for paraffin sections; AF3159, R&D Systems, Minneapolis, MurN, 1:160, for frozen sections), podocin (P0372, Sigma Aldrich, 1:1,000), GFP (No. 598, MBL, Nagoya, Japan, 1:500, for paraffin sections; No. 04404-26, GF090R, Nacalai, Kyoto, Japan, 1:1,000, for frozen sections), red fluorescent protein (RFP; PM005, MBL, 1:200; this antibody cross-reacts with tdTomato), and CD31 (MEC13.3, BD Biosciences, 1:50). Goat anti-guinea pig antibody (BA-7000, Vector Laboratories, Burlingame, CA), goat anti-rabbit antibody (BA-1000, Vector Laboratories), and Alexa Fluor 488-conjugated, 594-conjugated, or 647-conjugated antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) were used as secondary antibodies. For desmin staining on paraffin sections, the M.O.M. Immunodetection Kit (Vector Laboratories) was used, and sections were stained with periodic acid-Schiff (PAS). In Fig. 1C, desmin antibody was labeled using the Zenon Mouse IgG1 kit (ThermoFisher Scientific). The tdTomato emission signal of podocytes was detected by flow cytometry but not by fluorescence microscopy; therefore, immunostaining for tdTomato was performed for section analysis. The EGFP emission signal of podocytes was detectable by fluorescence microscopy before and after LMB2 injection. In Fig. 2, F and H, sections were stained with anti-GFP antibody to enhance the signal.

Figure 4.

Angiotensin II receptor blocker (ARB) ameliorated podocyte injury in the Pod-human (h)CD25T/G mosaic model. A: urinary albumin-to-creatinine ratio (UACR). With losartan treatment, albuminuria was significantly decreased 14 days after LMB2 injection (D14) in male mosaic mice aged 28−34 wk (n = 6 in each group). D0, D7, and D14, days 0, 7, and 14. B–E: representative morphological phenotypes on D14. Control mice showed glomerular injury, tubular dilatation, and proteinaceous casts (top left image in B, periodic acid-Schiff staining). In glomeruli, vacuolar degeneration of epithelial cells, hyalinosis, mesangiolysis, adhesion, and sclerosis were observed (bottom left image in B). In contrast, kidneys of the ARB group showed an almost normal morphology (top right and bottom right images in B). Semiquantitative analysis showed significantly lower glomerular injury scores in the ARB group. Nephrin (C) staining and enhanced green fluorescent protein (EGFP; E) staining were diminished in the control kidney but improved with ARB treatment. Podocytes in the control kidney showed desmin staining (arrowheads, D), which was attenuated in the ARB group. The dots in E demonstrate the ratio of the number of EGFP⁺ podocytes D14 to D0. Scale bars = 100 μm in the top images in B and 20 μm in the bottom images in B and C–E.

Figure 2.

Kidney injury of the Pod-human (h)CD25T/G mosaic model after injection of LMB2. A: urinary albumin-to-creatinine ratio (UACR). Male mosaic mice aged 14–34 wk developed severe albuminuria starting from 4 days after LMB2 injection (D4), but male age-matched wild-type (WT) mice did not. Mosaic mice: n = 14, 6, 10, 8, and 6 on days 0, 4, 7, 14, and 21 (D0, D4, D7, D14, and D21, respectively); WT mice: n = 3 at each time point. B: renal histology on D21. Mosaic mice showed severe glomerular injury, tubular dilatation, and proteinaceous casts, in contrast to the normal phenotype in WT mice. Scale bars = 100 μm. C: representative glomerular phenotypes on D0–D21 shown by periodic acid-Schiff (PAS) and nephrin staining. Marked injury was observed on D14 and D21, characterized by hyalinosis, mesangiolysis, vacuolar degeneration of epithelial cells, sclerosis, and complete loss of nephrin staining, indicating that all types of podocytes were lost. Scale bars = 20 μm. D: average tdTomato⁺ areas. Each plot represents an average of the tdTomato⁺ area per glomerular area in mosaic mice on D0 (nephrectomized kidney) and D14 (n = 10). tdTomato⁺ podocytes in the glomeruli almost disappeared on D14. Plots from the same mouse are connected with a line. E: average number and area of enhanced green fluorescent protein (EGFP)⁺ podocytes. The average number of EGFP⁺ podocytes per glomerulus was decreased on D14 (left), although the average EGFP⁺ area was unchanged between D0 (nephrectomized kidney) and D14 (n = 10; right). F–H: indirect injury of EGFP⁺ podocytes on D14. Podocin was diminished in EGFP⁺ podocytes (arrows, F). EGFP⁺ podocytes (arrowheads) showed intense desmin staining and diminished nephrin staining on the adjacent paraffin sections (G). Desmin expression in EGFP⁺ podocytes was also shown in frozen sections (arrows, H). Scale bars = 20 μm.

The EGFP⁺ or tdTomato⁺ area per glomerulus was measured (>10 glomeruli for each mouse) in frozen sections with ImageJ 1.52n (10) and averaged. To determine the EGFP⁺ podocyte number, paraffin sections were immunostained for EGFP, and the nuclei of EGFP⁺ cells in 50 glomeruli from each mouse were counted. Uninephrectomized kidneys before LMB2 injection were also analyzed to evaluate baseline status.

To evaluate podocyte injury, we stained paraffin sections for PAS, nephrin, and desmin as previously described (11). For the semiquantification of glomerular injury in PAS-stained sections, each quadrant of each glomerulus was scored as 0 (normal) or 1 (injured), and total glomerular scores ranging from 0 (normal) to 4 (completely injured) were calculated. In addition, for nephrin staining, each quadrant was scored as 0 (no staining), 1 (diminished staining), or 2 (normal staining), and total glomerular scores ranging from 0 (complete loss) to 8 (normal) were calculated. Scores for all of the glomeruli (>85 glomeruli) on a paraffin section for each mouse were averaged and defined as the glomerular injury index and nephrin index. For the semiquantification of desmin staining, the percentage of glomeruli (>90 glomeruli) containing surface podocytes with desmin staining was determined for a paraffin section from each mouse.

Gene Array and Quantitative RT-PCR Assay

Total RNA was extracted from EGFP⁺ or tdTomato⁺ podocytes isolated from male mosaic mice at 8 wk of age before (day 0) and 1, 4, and 7 days after LMB2 injection (days 1, 4, and 7) using TRIzol LS reagent (ThermoFisher Scientific).

For gene expression arrays, RNA samples were labeled using the Low Input Quick Amp Labeling Kit (Agilent Technologies, Santa Clara, CA), hybridized to SurePrint G3 Mouse GE 8 × 60 K arrays (Agilent), and scanned using an Agilent microarray scanner [n = 4 at each time point except for EGFP⁺ podocytes on day 1 (n = 3) and tdTomato⁺ podocytes on day 7 (n = 2)] according to the manufacturer’s protocol. Array data were analyzed using Gene Spring 14.9 (Agilent). For each probe, the average normalized signal value was determined in each condition. In the condition with the highest average signal, if more than 65% of the samples had unacceptable flags, the data of this probe were eliminated. Gene Ontology and pathway analyses were performed with Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 (https://david.ncifcrf.gov/home.jsp) (12) and Enrichr (https://maayanlab.cloud/Enrichr/) (13).

For the validation of gene expression (n = 5 at each time point), the StepOne Real-Time PCR System (ThermoFisher Scientific) was used to perform quantification RT-PCR. TaqMan primer probe sets (ThermoFisher Scientific) were used to detect nephrin (Nphs1), podocin (Nphs2), desmin (Des), and Gapdh mRNA levels. Primers for the other genes are shown in Table 1. Relative mRNA levels were determined by the ΔΔC_T method (where C_T is threshold cycle).

Table 1.

Primer sequences for RT-PCR analysis

Gene	Forward	Reverse
Csf1	5′- GGGGGCCTCCTGTTCTAC-3′	5′- CCCACAGAAGAATCCAATGTC-3′
Glcci1	5′- TGGAAAGGAAGAAGTGTCCAA-3′	5′- TGTCTGGGTATCTATGCTTCGAG-3′
Tgfb2	5′- TCGACATGGATCAGTTTATGCG-3′	5′- CCCTGGTACTGTTGTAGATGGA-3′
Thbs4	5′- CAGACAACTGCAGGCTCGT-3′	5′- GATATCTCCTACCCCGTCATTG-3′
Tnfsf15	5′- AAGCCGAGAGCACACCTG-3′	5′- GAAGGCCATCCCTAGGTCA-3′
Zfp423	5′- GCAACCAGATGTTCGACTCC-3′	5′- GTTGGCCTGGACGAAGACT-3′

Csf1, colony-stimulating factor-1; Glcci, glucocorticoid-induced transcript 1; Tgfb2, transforming growth factor-β2; Thbs4, thrombospondin-4; Tnfsf15, TNF superfamily member 15; Zfp423, zinc finger protein 423.

Statistical Analyses

For comparisons of the urinary albumin-to-creatinine ratio (UACR), values were logarithmically transformed and examined using linear mixed-effect models (Figs. 2A and 4A) or Dunnett’s multiple-comparison test (Fig. 3A). The Wilcoxon signed-rank test was used for comparisons of the tdTomato⁺ or EGFP⁺ area per glomerulus and the numbers of EGFP⁺ podocytes before and after LMB2 injection (Fig. 2, D and E). Quantitative RT-PCR data were logarithmically transformed and examined using linear mixed-effect models (Fig. 3F). The Mann–Whitney U test was used for other comparisons between two groups (Fig. 4, B–D). In each analysis, the level of significance was set at P < 0.05. For microarray analysis, Dunnett’s multiple-comparison tests were performed for EGFP⁺ and tdTomato⁺ samples, and adjusted P values (q values) were calculated with the Benjamini–Hochberg procedure. The level of significance was set at q < 0.025. Statistical analysis was performed with SPSS Statistics v25.0 (IBM, Armonk, NY) and R version 3.6.2.

Figure 3.

Transcriptomics analyses of early injury in podocytes. A: urinary albumin-to-creatinine ratio (UACR). Male mosaic mice aged 8 wk demonstrated albuminuria from 4 days after LMB2 injection (D4) [n = 25, 7, 7, 16, and 17 on days 0, 1, 2, 4, and 7 (D0, D1, D2, D4, and D7, respectively]. B: proportion of podocytes of each type. After immunotoxin treatment, the proportion of tdTomato⁺ podocytes (tdTomato) was decreased, whereas those of enhanced green fluorescent protein (EGFP)⁺ podocytes (EGFP) and tdTomato⁻ EGFP⁻ podocytes (negative) were increased (n = 11, 10, 7, 16, and 18 on D0, D1, D2, D4, and D7, respectively). C: heatmap of normalized array signals of all samples. Hierarchical clustering was performed by Euclidean distance. Individual mouse names (MC-#), time of analysis (D0, D1, D4, and D7), and cell type (EGFP or tdTomato) are shown. [n = 4 except for EGFP⁺ podocytes on D1 (n = 3) and tdTomato⁺ podocytes on D7 (n = 2)]. D: principal component analysis. tdTomato⁺ podocytes on D1, D4, and D7 are clearly distinguished from others. EGFP⁺ podocytes on D7 are weakly distinguished from others. E: heatmap of podocyte-specific genes. Each panel shows the average array signal of a podocyte-specific gene of tdTomato⁺ or EGFP⁺ podocytes at each time point. These genes were significantly downregulated in tdTomato⁺ podocytes after LMB2 injection and tended to be downregulated in EGFP⁺ podocytes on D7. Cd2ap, CD2-associated protein; Podxl, podocalyxin like; Ptpro, protein tyrosine phosphatase receptor type O; Synpo, synaptopodin; Vegfa, VEGF-A; Wt1, Wilms’ tumor 1; F: validation by quantitative RT-PCR analyses of representative genes. Thrombospondin-4 (Thbs4), desmin (Des), TNF superfamily member 15 (Tnfsf15), transforming growth factor-β (Tgfb2), and colony stimulating factor-1 (Csf1) mRNA levels were upregulated, whereas nephrin (Nphs1), podocin (Nphs2), glucocorticoid-induced transcript 1 (Glcci1), and zinc finger protein 423 (Zfp423) mRNA levels were downregulated in both tdTomato⁺ and EGFP⁺ podocytes after LMB2 injection (n = 5 at each time point in both EGFP⁺ podocytes and tdTomato⁺ podocytes).

RESULTS

Generation and Baseline Phenotypes of Pod-hCD25T/G Mosaic Mice

We designed a mosaic mouse line, Pod-hCD25T/G, in which Cre recombinase is expressed under the control of the nephrin promoter and either of two types of alleles is generated: one expressing hCD25 and tdTomato and the other expressing EGFP (Fig. 1A). Only hCD25-expressing podocytes are directly targeted by the immunotoxin LMB2. The two types of recombination are mutually exclusive and occur randomly after podocytes are differentiated and the nephrin promoter becomes active.

Without LMB2 injection, Pod-hCD25T/G mosaic mice showed normal renal morphology and a normal UACR, with a median of 0.048 mg/mg [interquartile range (IQR): 0.044–0.056 mg/mg (n = 14)]. In most glomeruli, a balanced number of EGFP⁺ podocytes and tdTomato⁺ podocytes coexisted, showing a mosaic pattern (Fig. 1B). The EGFP and tdTomato signals overlapped with nephrin but not with CD31 or desmin (Fig. 1C).

We next dissociated glomerular cells from mosaic mice and analyzed them with flow cytometry. Among the podocytes identified by podocin staining (n = 11), 36.2% (IQR: 35.6%–38.7%) were positive for tdTomato and 44.9% (IQR: 43.4%–45.7%) were positive for EGFP (Fig. 1D). As expected from the design of the transgene, no podocytes double positive for tdTomato and EGFP were detected. Furthermore, ∼20% of the podocytes were double negative, indicating that neither type of recombination occurred in these podocytes (see Supplemental Fig. S1; all Supplemental material is available at https://doi.org/10.6084/m9.figshare.13181801.v1). There were no tdTomato⁺ or EGFP⁺ cells in the podocin-negative fraction (Fig. 1D). Quantitative RT-PCR analysis confirmed the detection of hCD25 mRNA in tdTomato⁺ podocytes but not in EGFP⁺ or double-negative podocytes. However, compared with that of NEP25 mice, the expression of hCD25 was very low and undetectable by immunostaining.

Renal Histology of Mosaic Mice after LMB2 Injection

Podocyte injury was induced by a single injection of LMB2, an hCD25-targeting immunotoxin. Because the expression level of hCD25 was low, uninephrectomy was performed before the injection of LMB2 to enhance podocyte injury. Mosaic mice showed albuminuria from day 4, which peaked 1−2 wk after LMB2 injection (Fig. 2A). No histological changes were apparent within the first week (Fig. 2, B and C). Two to three weeks after LMB2 injection, mosaic mice developed severe glomerular damage showing segmental sclerosis, adhesion, and hyalinosis accompanied by tubular dilatation and proteinaceous casts. Sclerotic glomeruli accompanied damage and proliferation of parietal epithelial cells (PECs). CD44, an activation marker of PECs (14–16), was expressed on day 9 before morphological changes of PECs (Supplemental Fig. S2). CD44 was often entirely positive along Bowman’s capsule unrelated to the type of neighboring podocytes, suggesting that diffusible factors probably contained in plasma activate PECs (Supplemental Fig. S3). In a long-term analysis using female mice, most glomeruli showed either a normal appearance or global sclerosis, with almost no segmental sclerosis 4 mo after LMB2 injection.

Indirect Injury of EGFP⁺ Podocytes

Two weeks after LMB2 injection, the density of the tdTomato⁺ area per glomerulus was significantly decreased from 10.3% (IQR: 7.0%–16.3%) to 1.4% (IQR: 0.5%–3.5%) (P = 0.02; Fig. 2D). Notably, the number of EGFP⁺ podocytes, which do not express hCD25, was significantly decreased from 3.91 (IQR: 3.63–4.17) to 1.44 (IQR: 0.13–0.42) podocytes per glomerulus (P = 0.002), although the density of the EGFP⁺ area did not change over this period (Fig. 2E), suggesting hypertrophy of the surviving podocytes. Survived EGFP⁺ podocytes were also injured, as evidenced by diminished podocin and nephrin staining and intense desmin staining (Fig. 2, F–H). Occasionally, detached EGFP⁺ podocytes were identified in the renal tubular lumen and also rarely in the urine. These results indicate that podocytes in the mosaic model were indirectly injured after the injection of LMB2.

Gene Array Analysis of Podocytes at Baseline and Directly and Indirectly Injured Podocytes

We next performed microarray analyses of EGFP⁺ or tdTomato⁺ podocytes isolated from mosaic mice on days 0, 1, 4, and 7. Similar to the findings shown in Fig. 2, no remarkable morphological changes in the glomeruli were observed at these early time points, but the urinary excretion of albumin was increased from day 4 (Fig. 3A). After LMB2 injection, the proportion of tdTomato⁺ podocytes decreased over time (Fig. 3B). On day 7, the yield of tdTomato⁺ podocytes and their RNA levels were very low, and only two samples could be analyzed. Nevertheless, hierarchical clustering and principal component analysis of the data set demonstrated a clear separation of samples of tdTomato⁺ podocytes on days 1, 4, and 7 (Fig. 3, C and D, and Supplemental Fig. S4). EGFP⁺ samples on day 7 were also distinguished from others, although the differences were less remarkable.

Before LMB2 injection, there was no significant difference in the array signal between EGFP⁺ podocytes and tdTomato⁺ podocytes (Fig. 3C). As expected, the expression of known podocyte-specific genes, such as Nphs1 and Nphs2, was high, whereas that of mesangial, endothelial, or extraglomerular genes was low in both types of podocytes (Table 2). This baseline expression profile was also confirmed to be similar to that observed in our previous polysome analysis (Gene Expression Omnibus ID: GSE108629) (17). Thus, among 353 genes that were found to be highly expressed and enriched in normal podocytes in the previous study, 292 or 293 genes were also intensely expressed [normalized array signal > 4 (log₂ ratio)] in EGFP⁺ or tdTomato⁺ podocytes of mosaic mice before injection with LMB2. A notable difference was observed in the stress response genes Fos, Fosb, Junb, Egr1, and Egr2. The signals of these genes were low in normal podocytes in the previous study but high in podocytes before LMB2 injection in the present study. This result was probably caused by proteolytic cellular dissociation at 37°C, which was used only in the present study.

Table 2.

Array signals in EGFP⁺ or tdTomato⁺ podocytes at baseline

Gene	Array Signal (log2)
	EGFP+ Podocytes	tdTomato+ Podocytes
Podocytes
Nphs1	7.72	7.77
Nphs2	9.28	9.58
Podxl	7.06	7.22
Synpo	7.72	7.84
Vegfa	8.02	8.05
Wt1	8.22	8.37
Mesangial cells
Des	−1.45	−1.36
Endothelial cells
Tie1	−4.29	−4.62
Tek	−5.00	−4.19
Kdr	−1.24	−0.612
Vwf	−6.15	−6.28
Pecam1	−2.38	−2.06
Proximal tubules
Cdh1	−6.02	−6.17
Aqp1	−3.22	−1.37

Aqp1, aquaporin-1; Cdh1, cadherin 1; Des, desmin; EGFP, enhanced green fluorescent protein; Kdr, kinase insert domain receptor; Nphs1, nephrin; Nphs2, podocin; Podxl, podocalyxin-like; Pecam1, platelet and endothelial cell adhesion molecule 1; Synpo, synaptopodin; Tek, TEK receptor tyrosine kinase; Tie1, tyrosine kinase with immunoglobulin-like and EGF-like domains 1; Vegfa, VEGF-A; Vwf, von Willebrand factor; Wt1, Wilms’ tumor 1.

After LMB2 injection, the array profile of tdTomato⁺ podocytes changed very rapidly and dramatically, whereas EGFP⁺ podocytes changed more slowly and less dramatically (Fig. 3C). Therefore, the global clustering patterns mainly reflect changes in tdTomato⁺ podocytes. As expected, gene expression changes in tdTomato⁺ podocytes were similar to those observed in our previous polysome analysis of NEP25 mice and other podocyte injury models (17). Thus, the probes that increased in tdTomato⁺ podocytes after LMB2 injection were also increased in NEP25 mice after LMB2 injection (Supplemental Fig. S5). Many gene products characteristic of differentiated podocytes, including Nphs1, Wilms’ tumor 1 (Wt1), synaptopodin (Synpo), CD2-associated protein (Cd2ap), and VEGF-A (Vegfa), were significantly decreased in tdTomato⁺ podocytes, and a similar trend was observed in EGFP⁺ podocytes (Fig. 3E).

Only 14 genes were significantly regulated in EGFP⁺ podocytes on day 4, of which thrombospondin-4 (Thbs4) mRNA showed the greatest change (348.7-fold change; Fig. 3F and Supplemental Table S1). The most significant changes in EGFP⁺ podocytes were observed on day 7. At this time point, 558/113 probes (483/103 gene products) were significantly up/downregulated in EGFP⁺ podocytes (Supplemental Tables S2 and S3). Quantitative RT-PCR analyses showed a decrease in Nphs1 and Nphs2 mRNA levels and an increase in Des mRNA levels (Fig. 3F). Again, these data demonstrated that podocyte damage caused secondary damage to other podocytes.

Gene Ontology analysis of the regulated genes in EGFP⁺ podocytes on day 7 showed that biological process terms such as “inflammatory response,” “positive regulation of apoptotic process,” “cell adhesion,” and “positive regulation of MAPK cascade” were enriched in upregulated genes (Supplemental Table S4).

Fold changes in the expression of these genes in EGFP⁺ podocytes were generally smaller than those in tdTomato⁺ podocytes. This finding is likely because tdTomato⁺ podocytes had both direct and indirect injuries. Nevertheless, approximately one-fifth of the genes were significantly upregulated (>5-fold change) similarly in both types of podocytes on day 7 (Supplemental Table S5). These gene products may strongly reflect indirect podocyte injury. Pathway analysis of this gene set suggested the association of focal adhesion, integrin-mediated cell adhesion, and focal adhesion-phosphatidylinositol 3-kinase (PI3K)-Akt-mammalian target of rapamycin (mTOR) signaling (Supplemental Tables S6 and S7). This gene set included Des, a well-established marker of injured podocytes, and S100a6. Immunostaining revealed that S100a6 was expressed in EGFP⁺ podocytes of mosaic mice after LMB2 injection (Supplemental Fig. S6) and thus may be a good marker for indirect podocyte injury.

One possible mechanism of injury expansion is that molecules secreted from podocytes directly injured by LMB2 act on neighboring EGFP⁺ podocytes. In fact, chemokine (C-X-C motif) ligand 1 (Cxcl1), chemokine (C-X-C motif) ligand 10 (Cxcl10), and TNF superfamily member 15 (Tnfsf15) mRNA levels were increased in tdTomato⁺ podocytes (Fig. 3F). However, their receptors were not expressed in EGFP⁺ podocytes. Combinations of genes encoding ligand molecules that were induced in tdTomato⁺ podocytes and those for the corresponding receptors that were substantially expressed in EGFP⁺ podocytes include 1) heparin-binding EGF-like growth factor (Hbegf)–Erb-B2 receptor tyrosine kinase 2/4 (Erbb2/4), 2) transforming growth factor-β (Tgfb)2/3–TGF-β receptor (Tgfbr)1/2, 3) platelet-derived growth factor-A (Pdgfa)–platelet-derived growth factor receptor A (Pdgfra), 4) connective tissue growth factor (Cntf)–IL-6 receptor α-subunit (Il6ra)/IL-6 signal transducer (Il6st)/leukemia inhibitory factor receptor (Lifr), 5) colony-stimulating factor-1 (Csf1)–colony-stimulating factor-1 receptor (Csf1r), and 6) serpin family F member 1 (Serpinf1)–patatin-like phospholipase domain-containing protein 2 (Pnpla2). Among these ligands, Tgfb2, Pdgfa, Cntf, and Csf1 were upregulated in not only tdTomato⁺ podocytes but also EGFP⁺ podocytes, suggesting a vicious cycle in which injury is spread.

Prevention of Podocyte Injury Expansion with ARB

We next tested whether this new mosaic “partial podocytectomy” model can be used to evaluate the renal-protective effects of candidate drugs. As a proof of concept, we investigated the effect of ARB. Control mosaic mice developed moderate albuminuria and severe glomerular damage with segmental sclerosis, adhesion, and hyalinosis, which were accompanied by dilated tubules with protein casts (Fig. 4, A and B). ARB treatment did not significantly decrease albuminuria on day 7 but ameliorated albuminuria and histological damage on day 14. The glomerular injury index was 3.71 (IQR: 2.93–3.84) in the control group, which was markedly improved to 0.05 (IQR: 0.03–0.10) in the ARB group (P = 0.002). In addition, nephrin staining was better preserved [7.91 (IQR: 7.59–7.95) vs. 1.98 (IQR: 1.57–4.98), P = 0.002], and fewer desmin-positive podocytes were observed [29.3% (IQR: 15.4%–42.4%) vs. 84.2% (IQR: 53.8%–90.8%), P = 0.026] in the ARB group than in the control group, respectively (Fig. 4, C and D). Importantly, EGFP⁺ podocytes were also secondarily injured, and their number per glomerulus in the control group decreased from 3.64 (IQR: 3.56–3.98) to 0.90 (IQR: 0.36–1.04) after LMB2 injection. The number of EGFP⁺ podocytes was significantly better preserved in the ARB group [from 3.74 (IQR: 3.56–3.84) to 1.52 (IQR: 1.02–3.02); Fig. 4E]. In parallel, CD44 staining in PECs was attenuated in the ARB group (Supplemental Fig. S7).

DISCUSSION

Many transcriptomic and proteomic studies of normal or injured podocytes, including our previous study, have been reported (9, 17–25). In these studies, all podocytes were intact or uniformly and directly injured by various factors. The Pod-hCD25T/G mosaic model is unique because secondary podocyte injury can be studied. Thus, the mosaic model has two types of podocytes: one type that expresses the LMB2 receptor, hCD25, and another that does not. Upon injection of LMB2, not only hCD25⁺ podocytes but also hCD25⁻ podocytes were injured. Previously, we found a similar phenomenon in chimeric mice generated by the aggregation of NEP25 and R26-hPAP embryos, the latter of which do not carry the hCD25 gene (7). Our new mosaic model has several advantages over the previous chimeric mouse model. First, this mosaic model can be reproduced by mating, whereas de novo aggregation is required for the generation of chimeric mice. Second, mosaic mice showed balanced chimerism and a relatively constant ratio of hCD25⁺ podocytes, 36.2% (IQR: 35.6%–38.7%), and most glomeruli contained both types of podocytes (Fig. 1, B and C). In contrast, chimeric mice often contained only one type of podocyte. Before experiments, we had to perform renal biopsy and choose mice that contained both hCD25⁺ and hCD25⁻ podocytes. Third, the two types of podocytes in the mosaic model are tagged with EGFP or tdTomato, which enables the separation of podocytes with a cell sorter and further analyses.

After the injection of LMB2, mosaic mice showed severe albuminuria and developed focal segmental glomerulosclerosis (FSGS). EGFP⁺ podocytes, which lack the LMB2 receptor, were indirectly injured by LMB2, as evidenced by a decrease in nephrin and an increase in desmin at both mRNA and protein levels. This finding was essentially compatible with our previous study using chimeric mice (7). One difference is that in the previous study, indirect injury was detected as early as 4 days after LMB2 injection, whereas in the present study, most mRNA changes appeared 7 days after LMB2 injection. This difference was probably caused by the minor degree of direct injury in the mosaic model in the present study due to low-level hCD25 expression. Although most glomeruli contained tdTomato⁺ podocytes before LMB2 injection, most glomeruli showed either a normal appearance or global sclerosis with almost no segmental sclerosis 4 mo after LMB2 injection, suggesting that the fate of injured glomeruli is polarized into either complete recovery or progression to global sclerosis.

Labeling with EGFP or tdTomato allowed us to sort directly or indirectly injured podocytes by flow cytometry, after which RNA analysis was carried out. The gene expression profile of tdTomato⁺ podocytes was similar to that obtained in our previous polysome analysis of NEP25 mice and partly similar to recent proteomic data (17, 22). Following rapid and massive mRNA changes in tdTomato⁺ podocytes, modest but significant RNA changes in EGFP⁺ podocytes were observed within 7 days. The first change in EGFP⁺ podocytes was a dramatic increase in mRNA expression of Thbs4 on day 4. Thbs4 regulates the extracellular matrix and is induced in the injured heart, exerting a protective function by activating an adaptive endoplasmic reticulum stress pathway and other mechanisms (26–30).

Among the downregulated genes in EGFP⁺ podocytes on day 7, the product of glucocorticoid-induced transcript 1 (Glcci1) is a known component of podocyte foot processes, maintaining podocyte function and structure (31–33). Pathway analyses of the upregulated genes in EGFP⁺ podocytes on day 7 suggested that focal adhesion, integrin-mediated cell adhesion, and focal adhesion-PI3K-Akt-mTOR signaling may be involved in secondary injury. Upregulated genes included S100a6, dynamin 1 (Dnm1), and Des, which may be sensitive markers of indirect podocyte injury. In fact, desmin is a well-known marker of injured podocytes. It has been reported that Dnm1 was increased in the glomeruli of model rats and human patients with proteinuric kidney diseases (34).

One possible mechanism of indirect podocyte injury is that molecules secreted from tdTomato⁺ podocytes act on their corresponding receptor on EGFP⁺ podocytes and provoke maladaptive responses. By reviewing the array data, products of the following genes were identified as candidates: Hbegf, Tgfb2/3, Pdgfa, Cntf, Csf1, and Serpinf1. HB-EGF and TGF-β have been reported to be related to podocyte injury and glomerulosclerosis (35–41).

The spread of damage from cells to cells observed in the present study appears to be unique to podocytes. In other cell ablation models, secondary damage in nontargeted cells neighboring target cells has not been reported (42–44). The unique nature of this process may stem from the fact that podocytes are exposed to high filtration pressure (45). Such physical force may augment subtle injurious signals or facilitate detachment in nontargeted podocytes. A recent proteomic study revealed that filamin-B (Flnb gene), a mechanical stress response protein, was induced in podocytes in murine FSGS models and in human FSGS lesions (22). In the present study, expression of Flnb was significantly upregulated in EGFP⁺ podocytes on day 7 (3.67-fold change, q = 0.00884), suggesting that EGFP⁺ podocytes experienced mechanical stress. This possibility led us to assess the effect of ARBs in this model.

ARB treatment dramatically ameliorated glomerular injury and activation of PECs, whereas most podocytes, both tdTomato⁺ and EGFP⁺ podocytes, were lost on day 14 in the control group. This protective effect of ARB was far more remarkable than that observed in the NEP25 model. Notably, urinary albumin was similarly increased in both control and ARB-treated mosaic mice on day 7, suggesting that initial direct damage in the two groups was similar. ARB treatment markedly attenuated albuminuria and histological changes on day 14 and preserved EGFP⁺ podocytes, thus demonstrating that ARB can protect podocytes from secondary injury.

Hypertrophy of podocytes is a compensatory mechanism for podocyte loss to cover the bare glomerular basement membrane. Recent studies have reported that podocyte hypertrophy is mediated by mTOR complex 1 and that inhibition of mTOR complex 1 signaling leads to the development of FSGS lesions (46–48). In the present study, EGFP⁺ podocytes showed hypertrophy after LMB2 injection, and mTOR signaling-related genes, including ribosomal protein S6 kinase A1 (Rps6ka1), were upregulated in EGFP⁺ podocytes (Supplemental Table S2). As hypertrophic changes in podocytes are limited, stresses that induce exacerbated or persistent hypertrophy beyond the limitation cause detachment of podocytes (49). The excessive hypertrophy may also be an important mechanism of indirect podocyte injury.

There are some limitations in this study. First, direct injury was caused by LMB2, which may not generally represent podocyte injury in actual kidney diseases. Second, the sample size of the array experiment was small, which may cause a type 2 error and miss possible candidates for podocyte injury.

The result of the ARB experiment indicates that this mosaic model can be used to evaluate the renal-protective effect of candidate drugs. The subtotal nephrectomy rat model has been widely used to study the progressive loss of nephrons. This mosaic model can be likened to a “partial podocytectomy model” and may be useful to study the progressive loss of podocytes within glomeruli.

Data Availability

All Supplemental material is available at https://doi.org/10.6084/m9.figshare.13181801.v1. The microarray data supporting the findings of this study are openly available in the Gene Expression Omnibus database (GSE151869).

GRANTS

This work was supported by a Grant-in-Aid for Scientific Research [JP18K15986 (to M.Okabe) and JP18H02827, JP19K22628, and 22249033 (to T.M.)] by the Japan Society for the Promotion of Science, the MEXT-Supported Program for the Strategic Research Foundation at Private Universities (2009−2013), and Takeda Pharmaceutical Company and in part by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, and Center for Cancer Research.

DISCLOSURES

T.M. has received grants from Takeda Pharmaceutical Company. No conflicts of interest, financial or otherwise, are declared by the other authors.

AUTHOR CONTRIBUTIONS

M. Okabe, M. Ohtsuka, and T.M. conceived and designed research; M. Okabe, K.Y., and T.M. performed experiments; M. Okabe, K.Y., Y.M., M.M., T.Y., and T.M. analyzed data; M. Okabe, K.Y., Y.M., M.M., T.Y., and T.M. interpreted results of experiments; M. Okabe prepared figures; M. Okabe drafted manuscript; T.M. edited and revised manuscript; M.Okabe, K.Y., Y.M., M.M., M. Ohtsuka, I.P., T.Y., and T.M. approved final version of manuscript.