Generation and characterization of podocyte-specific Cldn5 knockout mice
Previous studies demonstrated CLDN5 is highly expressed in podocytes6, suggesting that it might play important roles in maintaining glomerular health. To bypass the postnatal lethality of constitutive deletion and investigate the role of CLDN5 specifically in podocytes, we created Cldn5loxP mice, in which the Cldn5 mutated allele contains exon 1 flanked by loxP sites, in the C57BL/6J background (Fig. 1a). Next, we generated mice with podocyte-specific deletion of Cldn5 by intercrossing Nphs2cre and Cldn5loxP/loxP animals (Fig. 1b). Because Cre is known to have nonspecific effects that could influence podocytes8, we studied 2 groups of mice: Nphs2-Cre+/-/Cldn5loxP/+ and Nphs2-Cre+/-/Cldn5loxP/loxP mice, hereafter referred to as Cldn5ctrl and Cldn5podKO mice. Successful deletion of Cldn5 from Cldn5podKO mice was confirmed by quantitative PCR–based transcript analysis of isolated glomerulus (Fig. 1c). CLDN5 was colocalized with podocyte specific marker podocin (NPHS2) in Cldn5ctrl mice, consistent with a podocyte source (Fig. 1d). Immunofluorescence staining, as indicated by the lack of CLDN5 colocalization with NPHS2 in Cldn5podKO kidneys, but appropriate signal in the endothelial cells of arteriole, confirmed that the Nphs2-Cre–mediated Cldn5 deletion was largely confined to podocytes (Fig. 1d). Likewise, western blot analyses of CLDN5 expression showed a significant decrease in CLDN5 expression (approximately 77%) in glomerulus from Cldn5podKO mice (Fig. 1e), suggesting that CLDN5 expression in podocytes accounts for the majority of CLDN5 in normal glomerulus. The knockout mice didn't show compensated and increased expression of other TJ proteins CLDN1, CLDN3, or CLDN6 in podocytes (Supplemental Fig. 1).
We next investigated whether reduced CLDN5 expression in itself could cause proteinuria directly, using our engineered mice with a podocyte-specific targeted deletion of Cldn5. The age-matched Cldn5ctrl and Cldn5podKO mice were analyzed for the albuminuria abundance levels in spot urine samples at different time points. Cldn5podKO mice showed no albuminuria in early stages and began to appear at around 12 weeks (Fig. 1f). There was no significant difference in body weight, urine volume and urinary osmolality throughout the observation period from 3 to 48 weeks (data not shown). Blood urea nitrogen (BUN) levels were within the normal values in both the groups (Fig. 1g). To investigate whether the appeared albuminuria was due to damage of the glomerular filtration barrier, we examined the kidney ultrastructure by transmission electron microscopy (TEM) in Cldn5podKO and Cldn5ctrl kidneys. TEM studies showed global thickening of the glomerular basement membrane (GBM) in Cldn5podKO mice (Fig. 1h). GBM abnormalities were clear at 8 weeks of age and gradually aggravated by 24 weeks (Fig. 1h). In addition, podocyte foot processes appeared abnormal with broadening and effacement, which were notable in areas of severe GBM thickening (Fig. 1h). Control littermates developed mild GBM and foot processes changes after 24 weeks of age (Fig. 1h). Histological staining with periodic acid-Schiff staining (PAS) identified that Cldn5podKO mice showed mesangial expansion and glomerular matrix accumulation at 24 weeks old compared with their littermate control mice (Fig. 1i). Quantitative PCR–based transcript analysis and immunofluorescence staining revealed that the expression of podocalyxin (PODXL) was reduced in Cldn5podKO mice, which further confirmed podocyte damage (Supplemental Fig. 2). In summary, mice with podocyte-specific deficiency of CLDN5 showed early GBM alterations followed by later development of albuminuria.
Cldn5 deletion in podocytes accelerates DN progression
To determine whether CLDN5 has a role in diabetic kidney disease, we first studied the expression of CLDN5 in 2 mouse models of DN, the unilateral nephrectomy (UNX) combined with streptozotocin (STZ)-induced type I diabetic mice and DB/DB type 2 diabetic mice, by double immunostaining of CLDN5 and NPHS2. In both strains, we found that the expression of CLDN5 was decreased, which was accompanied by an attenuation in nephrin (NPHS1) and NPHS2 expression (Fig. 2a and 2b, Supplemental Fig. 3a and 3b). To determine whether the changes in CLDN5 expression also occur in human glomerular diseases, we queried the published transcriptomic data sets in kidney disease compiled in the Nephroseq database (nephroseq.org). CLDN5 mRNA expression was significantly reduced in the glomerulus of DN patients compared with those of healthy controls (Supplemental Fig. 3c). These findings suggest that loss of CLDN5 may play critical roles in the progression of DN. Then, to further investigate the effects of CLDN5 on DN development, STZ-induced DN mice with or without CLDN5 knockout were used. The Cldn5podKO diabetic mice showed an increase in albuminuria as early 4 week after STZ injection, remaining elevated up to 12 weeks and reaching a difference of more than 4-fold compared with control diabetic mice of the same age (Fig. 2c). PAS staining revealed nodular glomerulosclerosis with increased amounts of extracellular matrix material in two groups of diabetic mice, which was exacerbated in Cldn5podKO diabetic mice (Fig. 2d). TEM analysis demonstrated that GBM thickening and foot process effacement were induced, and these effects were significantly aggravated in Cldn5podKO mice post to STZ treatment (Fig. 2e), which consistent with their more severe albuminuria. Podocyte injury was confirmed with increased expression of podocyte injury indicator desmin (Fig. 2f) and reduced expression of key podocyte markers, NPHS1, NPHS2, and PODXL (Supplemental Fig. 3d-i) in the Cldn5podKO diabetic mice, as compared with the Cldn5ctrl diabetic group. Masson Trichrome staining also showed a significant increase in interstitial fibrosis in the Cldn5podKO diabetic mice (Fig. 2g). These results indicate the higher susceptibility to diabetic injury in the Cldn5podKO mice.
Decreased Wif1 expression is observed in Cldn5 knockout glomerulus
To obtain insights into what might explain the phenotype of Cldn5 deletion in podocytes, we performed RNA-seq of glomerular lysates from Cldn5ctrl and Cldn5podKO mice. This unbiased analysis identified 280 downregulated genes and 102 upregulated genes (Fig. 3a). We found that, among the significantly altered genes by CLDN5 deletion, Wif1 reached remarkably high level (Fig. 3a). Data were validated by qRT-PCR (Fig. 3b) and immunofluorescence (Fig. 3c) performed on the kidney. Comparable results were obtained by qRT-PCR (Fig. 3d) and western blot analysis (Fig. 3e) on glomerulus isolated from mutant and wild-type littermates. WIF1 is a secreted WNT inhibitor, which exerts its inhibitory effect on WNT signaling by binding and inhibiting the activity of extracellular WNT ligands. This finding prompted us to speculate that CLDN5 depletion may lead to downregulation of WIF1 thereby activating WNT signaling. We found that WNT/β-catenin signaling was hyperactivated in Cldn5podKO podocytes based upon upregulated active nuclear β-catenin (CTNNB1) expression in Cldn5podKO podocytes compared to Cldn5ctrl podocytes (Fig. 3f). Notably, we found the same expression fingerprint of β-catenin target genes, including increased expression in CCND1 and CD44 (Fig. 3e, 3g-j). Previous studies in animals and humans have shown that CD44 is not expressed in healthy kidney, and activated parietal epithelial cells, but not podocytes, upregulate their de novo expression of CD44 during glomerular diseases9,10. In our study, although several of the variable CD44 isoforms were detected, mRNA levels of CD44v3 and CD44v5 appeared considerably higher expressed in the glomerulus from knockout mice than in those from control littermates (Fig. 3k). To determine whether dysregulation of WIF1 also occur in glomerulus with podocyte injury, we performed real-time PCR and immunostaining analysis. In comparison to control mouse glomerulus, WIF1 mRNA expression was significantly reduced in DN mice (Fig. 3l and 3m). WIF1 staining was significantly higher in the control while a dramatic loss of WIF1 staining was observed in these mice (data not shown). Collectively, these results indicate that the WNT pathway was activated by WIF1 inhibition in the mutant podocytes, resulting in the subsequent podocyte injury. These results identify CLDN5 as a potentially novel regulator of WNT/β-catenin signaling activity in podocytes.
Wif1 ablation mimics the phenotypes observed in Cldn5-deficient podocytes
To mimic the Wif1 downregulation observed in Cldn5podKO mice, we next constructed conditional knockout mice with podocyte-specific ablation of Wif1 by using the Cre-LoxP system. We generated Wif1loxP mice, in which the Wif1 mutated allele contains exon 3 flanked by loxP sites, in the C57BL/6J background (Supplemental Fig.4a). Next, we generated mice with podocyte-specific deletion of Wif1 by intercrossing Nphs2cre and Wif1loxP/loxP animals (Supplemental Fig. 4b). No residual Wif1-transcript or protein was detectable as determined by qRT-PCR, immunoblot, immunofluorescence in glomerular lysates or kidney sections in Nphs2-Cre+/-/Wif1loxP/loxP (referred to as Wif1podKO) mice (Fig. 4a-c), indicating WIF1 is expressed predominantly in podocytes in the kidney. Our data is in agreement with the single-cell RNA sequencing datasets of mouse kidney which indicated that Wif1 is expressed exclusively in podocytes (Supplemental Fig. 4c and 4d)11,12. To determine whether deletion of Wif1 in podocyte leads to activation of canonical WNT signaling, we studied the expression of several putative WNT/β-catenin target genes in the glomerulus. We found CCND1 and CD44 were upregulated in Wif1podKO mouse glomerulus (Supplemental Fig.4e and 4f). Wif1 KO mice had normal renal histology at 16 weeks of age, but TEM revealed thicker GBM which remains completely covered by the foot processes of the podocytes, but in areas with GBM thickenings foot process effacement was observed (Fig. 4d). Although genetic deletion of Wif1 resulted in the similar glomerular phenotype, the phenotypes observed in these mouse models were less severe than in the Cldn5podKO mice. Consistent with this, podocyte-specific Wif1 mutant mice developed mild albuminuria at 16 weeks of age, 1.7-fold higher than control mice (Fig. 4e). The incomplete phenocopy of podocyte-specific Wif1 KO mice with Cldn5 KO mice led us to conclude that additional pathway maybe involved in the kidney pathogenesis associated with Cldn5 deletion.
AAV9-mediated WIF1 gain of function in podocytes ameliorates the progression of DN in Cldn5 KO mice
To further investigate the relevance of WIF1 to glomerular phenotype in Cldn5podKO mice, we went on to test whether podocyte-specific WIF1 overexpression could rescue the phenotype of Cldn5podKO mice. To overexpress WIF1 in podocytes, we used an AAV9 system with kidney in situ injection which has been proved to primarily transduce cells within the glomerulus of the kidney13. Podocyte-specific WIF1 delivery rescued the glomerular injury phenotype of diabetic Cldn5podKO mice, including profound reduction of urine albumin-to-creatinine ratio (Fig. 4f), reduced foot process effacement, and decreased extracellular matrix deposition, as evidenced by PAS staining (Fig. 4g) and electron microscopy analysis (Fig. 4h). We also observed significantly less podocytes loss in WIF1-treatment group compared with the mutant mice treated with control AAV (podocyte number/glomerulus: AAV-WIF1 versus AAV-CTL: 11.583 ± 0.045 versus 10.167 ± 0.035, P < 0.05, n=10 mice/group) (Fig. 4i). Taken together, these results show a striking normalization of podocytes upon WIF1 administration in Cldn5podKO diabetic mouse model, suggesting new avenues for the development of therapeutic strategies to ameliorate podocyopathy in DN.
Podocyte-specific loss of CLDN5 or WIF1 exacerbates interstitial fibrosis in UUO mouse model
Our data indicated that the phenotype changes in Cldn5podKO podocytes aggravated interstitial fibrosis in DN mouse model (Fig. 2g). Because WIF1 is predominantly expressed in the podocytes, it could play a paracrine role on tubular epithelial cells through constant secretion into the preurine as a carrier. As WNT is an essential modulator of fibrosis development, we reasoned that local production of WIF1 by podocytes might affect WNT pathway tone in proximal tubules and participate in the progression of fibrosis following acute kidney injury. In spite of the fact that genetic deletion is limited to podocytes, both Wif1podKO and Cldn5podKO kidneys showed comparatively higher expression of the tubular damage markers KIM1, αSMA and collagen I compared with their littermates at 14 days of UUO (Fig. 5a and 5b). Kidney histological analysis using Masson’s Trichrome staining showed that histological changes induced by UUO were markedly aggravated in CLDN5 or WIF1 deficient mice (Fig. 5c and 5d). Moreover, this effect was accompanied by up-regulated WNT downstream gene expression including CCND1 and CD44 (Fig. 5e and 5f). To demonstrate directly that podocyte secrete factors capable of silencing WNT signaling in proximal tubular cells, we assessed levels of the WNT target genes in cultured TKPTS exposed to podocyte culture medium. We found that exposure of TKPTS to culture medium from Cldn5podKO podocytes resulted in the increased expression of WNT target genes including Mmp7, Tcf7, Cd44, Lef1, and Ccnd1 in TKPTS, compared with the culture medium from Cldn5ctrl podocytes which has higher concentration of WIF1 (Supplemental Fig. 5b-d). These data suggest that inadequate WIF1 secreted by podocytes in the Cldn5 and Wif1 knockout mouse permits exaggerated kidney damage and fibrosis during UUO via WNT-dependent actions in tubular epithelial cells. Thus, paracrine signals from podocytes that include WIF1 likely interact with proximal tubular cells and are essential to maintain its WNT pathway activity.
CLDN5 interacts with ZO1 and ZONAB in podocytes
The significant transcriptional downregulation of WIF1 by CLDN5 absence promoted us to identify the underlying molecular mechanisms linking them. On the basis of the information currently available, ZO1 form scaffolds to anchor TJ membrane proteins, and it also play very important roles in the control of gene expression via binding to and tuning the activity of transcription factor ZONAB14,15. To directly document the interaction between CLDN5 and ZO1/ZONAB complex, we performed coimmunoprecipitation (CoIP) in sparsely plated HEK293 cells transfected with four genes simultaneously. In HEK293 cells multiply transfected with CLDN5, ZO1, ZONAB variant 1, and ZONAB variant 2, anti-CLDN5 antibody precipitated ZO1 and ZONAB, and reciprocally, ZO1 co-immunoprecipitated with CLDN5 and ZONAB using the anti-ZO1 antibody, ZONAB co-immunoprecipitated with ZO1 and CLDN5 using the anti-ZONAB antibody (Fig. 6a). To test whether CLDN5 and ZO1/ZONAB are associated in native tissue, glomerular extracts were immunoprecipitated with anti-ZO1, anti-CLDN5 and anti-ZONAB, and precipitation of the three proteins was monitored by immunoblotting. We confirmed the endogenous interaction between CLDN5, ZO1, and ZONAB from glomerular extracts (Fig 6b). Together, these data reflect the existence of a complex containing CLDN5, ZO1, and ZONAB in the podocytes.
Podocytes CLDN5 deficiency increases ZONAB nuclear localization
We next asked whether Cldn5 knockout deregulates ZO1/ZONAB signaling in the podocytes. A marked reduction and more fragmented of ZO1 expression were observed in the Cldn5 KO mice compared with their wild-type littermates (Fig. 6c). Real-time quantitative PCR showed that the mRNA level of ZO1 was not significantly changed after Cldn5 deletion (data not shown), suggesting that the downregulation of ZO1 occurred at the protein level, possibly by higher levels of degradation. Confocal imaging showed membrane, cytoplasmic and weak nuclear ZONAB labeling in Cldn5ctrl mouse glomerulus and a distinct colocalization of ZONAB with WT1 in Cldn5podKO mice (Fig. 6d), which indicated CLDN5 deletion in podocytes increased ZONAB nuclear localization. Subcellular fractionation of the isolated glomerulus, followed by western blot analyses, indicated that CLDN5 deletion and ZO1 reduction in podocytes promoted ZONAB to undergo nuclear translocation (Fig. 6e). We then determined whether the effects of CLDN5 deletion on Wif1 expression could be rescued by re-introduction of ZO1. Co-transfection of CLDN5-defiencient primary podocytes with Zo1 and Cldn5, but not transfection with Zo1 alone, showed an increase in Wif1 expression (Fig. 6f), indicating a functional interaction between CLDN5 and ZO1 in the regulation of Wif1. Taken together, our data so far indicate that CLDN5 forms a complex with ZO1 and ZONAB in normal podocytes under physiologic conditions, and this complex is required to sustain ZONAB's subcellular localization and adequate levels of WIF1 to maintain normal WNT signaling activity.
ZONAB regulates Wif1 expression through its 3'-UTR
We next analyzed the molecular mechanism by which ZONAB reduced the expression of Wif1 mRNA. Transcript levels can be altered by changes in transcription, RNA processing, mRNA stability, or a combination thereof, and ZONAB has been suggested to have roles in all these processes. To address these possibilities, we first evaluated whether ZONAB plays a transcriptional role on Wif1 promoter, we performed reporter gene assays by transiently transfecting into primary podocytes and MDCK cells a plasmid construct containing the firefly luciferase gene driven by the Wif1 promoter fragments together with Zonab or its vector control. This analysis showed little to no effect on Wif1 mRNA (Fig. 6g and 6h), suggesting that Wif1 mRNA abundance regulation by ZONAB is not mediated by its 5'-promoter. Next, we set out to understand if ZONAB targets Wif1 3'-UTR, we generated the reporter constructs that had the entire mouse Wif1 3'-UTR sequence cloned downstream of the renilla luciferase gene. The reporter was transfected with Zonab to primary podocytes and MDCK cells. When both cells were co-transfected with a ZONAB expression vector (pCMV6-Zonab-V1 or pCMV6-Zonab-V2), the luciferase activities were decreased (Fig 6i and 6j), thus providing evidence that ZONAB acts as a repressor factor of Wif1 expression by its 3'-UTR. To verify that ZONAB is capable of inhibiting the endogenous WIF1 in native podocytes, we knocked down Zonab with specific siRNA and analyzed the expression level of Wif1 via real-time RT-PCR. Following knockdown of Zonab mRNA, qRT-PCR revealed significantly elevated levels of Wif1 transcript in Cldn5podKO podocytes compared to cells treated with control siRNA (Fig. 6k). Taken as a whole, our results establish that transcriptional regulation of WIF1 by ZONAB at least in part is brought about via repression of the Wif1 3'-UTR.