TGF-β1 is involved in senescence-related pathways in glomerular endothelial cells via p16 translocation and p21 induction

p16 inhibits cyclin-dependent kinases and regulates senescence-mediated arrest as well as p21. Nuclear p16 promotes G1 cell cycle arrest and cellular senescence. In various glomerular diseases, nuclear p16 expression is associated with disease progression. Therefore, the location of p16 is important. However, the mechanism of p16 trafficking between the nucleus and cytoplasm is yet to be fully investigated. TGF-β1, a major cytokine involved in the development of kidney diseases, can upregulate p21 expression. However, the relationship between TGF-β1 and p16 is poorly understood. Here, we report the role of podocyte TGF-β1 in regulating the p16 behavior in glomerular endothelial cells. We analyzed podocyte-specific TGF-β1 overexpression mice. Although p16 was found in the nuclei of glomerular endothelial cells and led to endothelial cellular senescence, the expression of p16 did not increase in glomeruli. In cultured endothelial cells, TGF-β1 induced nuclear translocation of p16 without increasing its expression. Among human glomerular diseases, p16 was detected in the nuclei of glomerular endothelial cells. In summary, we demonstrated the novel role of podocyte TGF-β1 in managing p16 behavior and cellular senescence in glomeruli, which has clinical relevance for the progression of human glomerular diseases.

Induction of cellular senescence in the glomeruli of PodCre(+) TGF mice. To confirm the involvement of TGF-β1 in senescence, we determined the activity of senescence-associated β-galactosidase. β-galactosidase activity was significantly increased in the glomeruli of PodCre(+) TGF mice (Fig. 2a). In addition, we detected the upregulation of senescence-related proteins such as Rb2 and p27 in the glomeruli of Pod-Cre(+) TGF mice (Fig. 2b) 30,31 . Expression sites and levels of p16 and p21 in the glomeruli of PodCre(+) TGF mice. We investigated p16 and p21 expression in PodCre(+) TGF mice. The expression site was distinguished by double immunofluorescence staining with the antibody against collagen IV. Both p16 and p21 were significantly immunostained mainly in the nuclei of endothelial cells (Fig. 2c,d). These findings were confirmed by the immunostaining of CD34 (endothelial cell marker) and p16 or p21 using the serial kidney sections ( Supplementary Fig. S1 online).
However, western blot analysis showed that compared to control mice, the expression of p16 did not change significantly in the glomeruli of PodCre(+) TGF mice, which seemed inconsistent with the immunohistochemical analysis shown in Fig. 2c. On the other hand, p21 expression was increased significantly in PodCre(+) TGF mice (Fig. 2e).
The activation of the TGF-β1-Smad3 pathway can induce p21 expression in the late phase, while it can translocate p16 to the nuclei in the early phase in endothelial cells in vitro.. Further, we investigated the effect of TGF-β1 on the expression of p16 and p21 in cultured endothelial cells. TGF-β1 could increase p21 expression, not within 30 min, but 24 h (late phase). SB431542, a TGF-β1 receptor antagonist, could suppress the increase in p21 expression. Transfection of constitutive active Smad3 could also increase p21 expression, suggesting p21 expression is controlled by the TGF-β1-Smad3 pathway (Fig. 3a,c). However, TGF-β1 did not affect p16 expression (Fig. 3a,b). www.nature.com/scientificreports/ www.nature.com/scientificreports/ Finally, to clarify the inconsistency between in vivo immunohistochemical and western blot analyses, we analyzed the expression of p16 in the nucleus and cytoplasm in cultured endothelial cells separately. p16 was translocated to nuclei by TGF-β1 stimulation in 30 min (early phase), and the effect continued until 24 h later. Analyses using SB431542 and constitutive active Smad3 confirmed that the translocation was induced via the TGF-β1-Smad3 pathway (Fig. 4).
Expression of β-galactosidase and p16 in the glomeruli of patients with kidney diseases. TGF-β1 is involved in the development and progression of various kidney diseases. Therefore, we evaluated the expression of β-galactosidase and p16 using paraffin-embedded human renal biopsy samples. β-galactosidase expression was detected in the glomeruli of patients with kidney diseases, suggesting the induction of cellular senescence (Fig. 5a). In addition, p16 was expressed in endothelial cells of patients with representative glomerulonephritis and nephrotic syndrome except minimal change disease and diabetic nephropathy, suggesting the common pathological significance of p16 in glomerular diseases (Fig. 5b). The expression site was confirmed by double immunofluorescent staining of p16 with CD34 (endothelial cell marker), collagen IV and nephrin (podocyte marker) using human renal biopsy frozen samples ( Supplementary Fig. S2 online). Collagen IV was chosen to show the mesangial matrix because there is no specific mesangial cell marker. www.nature.com/scientificreports/

Discussion
In this study, we demonstrated that podocyte TGF-β1 could affect the behavior of p16 in glomerular endothelial cells in vivo and in vitro. We also clarified the pathological and clinical phenotypic changes in kidneys induced by the practical level of TGF-β1 expression. The most crucial finding in this study is that TGF-β1 was involved in senescence-related pathways via not only p21 but also p16 in glomeruli in vivo and in vitro, because p16 and p21 are major molecules responsible for cellular senescence. The relationship between the TGF-β1-Smad3 pathway and p21 has been well clarified [24][25][26][27] . However, the direct effect of the TGF-β1-Smad3 pathway on p16 behavior had not been investigated, especially in vivo. Several reports suggested the contribution of TGF-β1 to cellular senescence via a p16 mediated mechanism in vitro 32 . However, to our knowledge, the connection between TGF-β1 and p16 in cellular senescence had not been proved in vivo. We reveal the novel mechanism of TGF-β1 involvement in p16 behavior, which is different from the interaction between TGF-β1 and p21. TGF-β1 could not increase the expression of p16 in glomeruli, but could induce nuclear translocation of p16 in glomerular endothelial cells.
p16 expression in kidney tubules and interstitial cells has been studied in mouse models and human kidney diseases with respect to the progression of kidney fibrosis and aging of kidney 33 . For example, high phosphate activates senescence in renal tubular cells through distinct but interconnected mechanisms: upregulation of p16/p21, elevation of plasminogen activator inhibitor-1 and downregulation of Klotho, followed by fibrosis 34 . In www.nature.com/scientificreports/ addition, the acute kidney injury-to-chronic kidney disease transition may involve a wide range of mechanisms, including the action of scar-forming myofibroblasts, microvascular rarefaction, mitochondrial dysfunction, or cell cycle arrest by the involvement of the epigenetic, gene, and protein alterations leading to common final www.nature.com/scientificreports/ signaling pathways such as TGF-β1, p16, Wnt/β-catenin pathway involved in renal aging 35 . However, to our knowledge, p16 expression in each cell comprised of glomeruli has not been thoroughly investigated. In human kidney diseases such as IgA nephropathy, nephrotic syndrome, and diabetic kidney disease, as well as aging kidney, p16 expression was found in mesangial, endothelial cells, and podocytes [10][11][12][13] . On the other hand, there were few reports evaluating p16 expression in glomeruli using the in vivo animal kidney disease model. Aratani et al. showed that p16 is involved in radiation-induced kidney disease by immunohistochemical analysis 36 . In the diabetic kidney disease model, western blot analysis revealed that p16 expression increased significantly in glomeruli [37][38][39] . So far, our animal study is the first to evaluate p16 expression in glomeruli quantitatively using both western blot and immunohistochemical analyses in vivo. We revealed that the increase in nuclei positive immunostaining of p16 does not always coincide with the upregulation of p16 expression in vivo.
In this investigation, we would like to clarify the sole and glomerulus-specific TGF-β1 effect on glomerular diseases. The role of TGF-β1, especially in the interstitial area of the kidney, has been well evaluated using conventional fibrosis models such as the UUO model 40 . This model causes a relatively rapid progression of fibrosis. However, the model does not cause glomerulopathy, and we cannot determine the role of TGF-β1 in glomerular diseases using this model. Diabetic nephropathy is one of the representative and important glomerular diseases. The animal models of diabetic nephropathy display glomerular lesions such as mesangial matrix expansion, glomerular basement membrane thickening, and mild tubulointerstitial damage 41 . However, these changes are induced by hyperglycemia. Hyperglycemia triggers many types of cytokines, chemokines, and several signaling pathways including protein kinase C cascade, Janus kinase/signal transducer and activator of transcription signaling, mitogen-activated protein kinase, mammalian target of rapamycin, and Smad 42 . Although TGF-β1 is involved in the progression of diabetic nephropathy 43,44 , it is just one of the cytokines involved in the progression of diabetic nephropathy. We cannot determine the sole effect of TGF-β1 on glomerular diseases. Therefore, our model has the strength to clarify the sole effect of TGF-β1 on glomerular diseases in vivo.
Several studies have previously reported the role of TGF-β1 by using the TGF-β1 overexpression mouse model. Kopp et al. reported that TGF-β1 overexpression in the liver can cause kidney glomerulosclerosis 19 . However, this glomerulosclerosis model has an eight times higher expression. Hathaway et al. revealed that TGF-β1 expression level could influence the kidney manifestation in the mice, especially under diabetic conditions 18 . These findings mean that TGF-β1 can cause glomerulosclerosis if its concentration is exceptionally high or any other risk factors such as cytokines and metabolic conditions concur with TGF-β1 stimulation. In human kidney diseases, the reports evaluating plasma TGF-β1 concentration were limited. Plasma TGF-β1 concentration can increase according to kidney dysfunction or diabetic kidney injury. In older community-dwelling adults, the levels of median plasma TGF-β1 were higher for those with eGFR < 60 mL/min/1.73 m 2 compared to those with eGFR > 60 mL/min/1.73 m 2 45 . In patients with diabetic kidney disease, baseline median plasma TGF-β1 level was two times higher in participants with progressive kidney disease than participants whose kidney disease had not progressed 46 . Iwano et al. investigated intraglomerular TGF-β1 mRNA in patients with human kidney diseases. TGF-β1 mRNA was significantly elevated in patients with mesangial proliferative glomerulonephritis having a moderate increase in the mesangial matrix, diabetic nephropathy and lupus nephritis compared to participants with normal glomeruli. TGF-β1 mRNA expression levels in patients with diffuse proliferative lupus nephritis were more than five times higher than those with normal glomeruli 47 . Unfortunately, in our mouse model, we could not estimate the local expression level of total TGF-β1 in glomeruli quantitatively, because we used an overexpression model of porcine TGF-β1, which does not have the same potency as mouse TGF-β1. However, our mice had similar plasma TGF-β1 concentration as the control mice, which is consistent with the previous report using the same mice 48 . Therefore, we believe that our mouse model clarifies the development of primary background lesions in various human kidney diseases, because TGF-β1 is involved in the development and progression of these diseases 14,15 . Our mice would represent the early stage of glomerular lesions considering pathological changes such as mild mesangial expansion, podocyte injury, and albuminuria. Moreover, we could show that the practical level of TGF-β1 per se causes the expression of senescence-related molecules in the nuclei of glomerular endothelial cells. Therefore, endothelial senescence can be triggered in the early stage of various human kidney diseases, as p16 expression was found in the nuclei of endothelial cells in human renal biopsy samples from many kinds of kidney diseases in this study (Fig. 5b, Supplementary Fig. S2 online), consistent with the previous findings of glomerular TGF-β1 mRNA expression in patients with kidney diaseases 47 . Endothelial senescence could be one of the important mechanisms in the progression of arteriosclerosis in glomeruli 24 .
Probably, in addition to the TGF-β1-related basic alterations of pathology and molecular behavior in mesangial, endothelial cells, and podocytes shown in this study, various cytokines and growth factors modify kidney lesions, followed by the establishment of complex and disease-specific kidney manifestation.
In this study, we investigated the phenotype in podocyte-specific TGF-β1 overexpression mice, which have glomerulus-specific TGF-β1 overexpression. Regarding the expression site of TGF-β1 in the glomeruli of human kidney diseases, both Yamamoto et al. and Ito et al. reported that TGF-β1 is expressed in podocytes as well as mesangial, endothelial cells of glomeruli in patients with proliferative nephritis 14,49 . In patients with advanced diabetic nephropathy, TGF-β1 is immunostained in both matrix and remnant cells of glomeruli 43,50 . The limitation of our mouse model is that the podocyte-specific TGF-β1 overexpression mouse model can only partly explain the pathogenic role of TGF-β1 in these glomerulopathies. However, considering the phenotype of the mice having TGF-β1 overexpression in glomerular endothelial cells for 28 days in the previous report 20 , which resembled our results in terms of podocyte injury and proteinuria, TGF-β1 could cause a podocyte-endothelial crosstalk 51 . In addition, it is technically impossible to investigate the effect of glomerular endothelial-specific or mesangial-specific TGF-β1 overexpression in vivo.
In conclusion, we found the involvement of the TGF-β1-Smad3 pathway in the behavior of p16 in glomeruli in vivo and in vitro. These findings will be one of the common and novel molecular mechanisms in the progression of various human kidney diseases. www.nature.com/scientificreports/

Methods
Ethics statement. All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki. All patients provided informed written consent for participation in and publication of the study. The animal experiment was carried out in compliance with the ARRIVE guidelines. All experiments were performed following the institutional guidelines and regulations of Tokushima University. The study, including this human study and the animal experiments, was approved by the Research Ethics Committee of Tokushima University.
Subjects. Renal biopsy samples derived from different human glomerular diseases such as minimal change disease, lupus nephritis, IgA nephropathy, purpura nephritis, membranous proliferative glomerulonephritis, membranous nephropathy, ANCA glomerulonephritis, and diabetic nephropathy diagnosed at Tokushima University Hospital were analyzed in this study. Renal biopsy tissues were fixed in Dubosque-Brazil's solution.
Biopsy samples from patients with asymptomatic hematuria served as controls and showed minor glomerular abnormalities and negative immunofluorescence. The profiles of control and patients with human kidney diseases are shown in Table 1.
Mice. Podocin-Cre mice and Cre-dependent HA-tagged TGF-β1 overexpression mice were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). Before starting this experiment, all mice were backcrossed 10 times to ICR (CLEA Japan Inc., Tokyo, Japan). In PodCre(+) TGF mice, HA-tag was conjugated with bioactive porcine TGF-β1. Urine and plasma were collected from the mice, and the mice were sacrificed at one year of age to analyze the pathological changes in the kidney.  Western blotting. In vivo, glomeruli from one-year-old mice were collected by magnetic beads-based isolation 53 . Briefly, transcardiac perfusion was performed using phosphate-buffered saline containing precleaned beads (Dynabeads, Invitrogen). The perfused renal cortex was briefly digested with collagenase A (Roche, Basel, Switzerland) and deoxyribonuclease I (Invitrogen), and the glomeruli stuffed with beads were isolated by DYNAL (Invitrogen). Glomeruli were lysed using Mammalian Cell Extraction Kit (BioVision Inc., Milpitas, CA, USA). Lysates of glomeruli were subjected to SDS-PAGE and immunoblotted with the following primary antibodies: rabbit antibody against p16 (ab108439, Abcam), p21 (ab109199, Abcam), phospho-Smad3 (ab52903, Abcam), and Smad3 (ab28379, Abcam), and mouse antibody against Rb2 (610262, BD Biosciences, San Jose, CA, USA), p27 (610241, BD Biosciences), α-tubulin, and β-actin (T6199, A5316, Sigma-Aldrich, St. Louis, MO, USA). In vitro, lysates of cultured endothelial cells were immunoblotted with the antibodies mentioned above and goat anti-Histon H3 (sc-8654, Santa Cruz Biotechnology). Immobilon ECL Ultra Western HRP Substrate (Merck Millipore) was used to detect the blotting signals using LAS-3000 (FUJIFILM, Tokyo, Japan). The immunohistochemical signal was quantified using Image J 52 . Mean values were calculated using data obtained from four to six mice or three to four independent in vitro experiments.

Immunohistochemical analysis.
Electron microscopy. Tissues used for electron microscopy were fixed with 2.5% glutaraldehyde. We entrusted electron microscopy analysis to a specialized company (BML Inc. Tokyo, Japan.) 54 . Glomerular basement membrane width was measured using Image J 52 . Mean values were calculated using data obtained from three to four mice. For each sample, six glomerular basement membrane widths were measured.
Urine albumin and creatinine. Urinary albumin and creatinine were determined using Albuwell M and Creatinine Companion kits (Exocell Inc., Philadelphia, PA, USA).

Statistical analysis.
All values are expressed as mean ± SD. Statistical analysis was performed using SPSS for Windows version 13.0 (SPSS Inc., Chicago, IL, USA). If data were normally distributed, the results were compared using Student's t-test or Welch's t-test. Non-normal data were analyzed by Mann-Whitney's U test. F-test was used for comparing the factors of total deviation, and the significance was set at P less than 0.05.