The Atypical Cyclin-Dependent Kinase 5 (Cdk5) Guards Podocytes from Apoptosis in Glomerular Disease While Being Dispensable for Podocyte Development

Cyclin-dependent kinase 5 (Cdk5) is expressed in terminally differentiated cells, where it drives development, morphogenesis, and survival. Temporal and spatial kinase activity is regulated by specific activators of Cdk5, dependent on the cell type and environmental factors. In the kidney, Cdk5 is exclusively expressed in terminally differentiated glomerular epithelial cells called podocytes. In glomerular disease, signaling mechanisms via Cdk5 have been addressed by single or combined conventional knockout of known specific activators of Cdk5. A protective, anti-apoptotic role has been ascribed to Cdk5 but not a developmental phenotype, as in terminally differentiated neurons. The effector kinase itself has never been addressed in animal models of glomerular disease. In the present study, conditional and inducible knockout models of Cdk5 were analyzed to investigate the role of Cdk5 in podocyte development and glomerular disease. While mice with podocyte-specific knockout of Cdk5 had no developmental defects and regular lifespan, loss of Cdk5 in podocytes increased susceptibility to glomerular damage in the nephrotoxic nephritis model. Glomerular damage was associated with reduced anti-apoptotic signals in Cdk5-deficient mice. In summary, Cdk5 acts primarily as master regulator of podocyte survival during glomerular disease and—in contrast to neurons—does not impact on glomerular development or maintenance.

In the kidney, Cdk5 is exclusively expressed in glomerular epithelial cells called podocytes [16]. Podocytes are postmitotic epithelial cells arrested in G0/1 phase that reside on the outer aspect of glomerular capillaries. Podocytes protrude primary and secondary cellular processes, called foot processes, which interdigitate with foot processes of neighboring cells. Podocyte foot processes compress the glomerular basement membrane against positive capillary pressure and regulate mesh width of the glomerular basement membrane to limit the passage of macromolecules, specifically proteins larger than 70 kDa [17]. Loss of podocytes either by detachment from the glomerular basement membrane or by apoptosis cannot be replaced by proliferation, which results in loss of function of the glomerular filter, indicated clinically by urinary protein loss (i.e., proteinuria/albuminuria) and histologically by glomerular scarring.
In podocytes, Cdk5 is activated by p25, p35, and/or cyclin I, which also determine the subcellular localization of the kinase [18]. Depending on the binding partners, Cdk5 exerts an anti-apoptotic signal via BCL-2/BCL-XL through diverse mechanisms [7,19]. Loss of Cdk5 activation, by conventional knockout of p35 or cyclin I or both, in mice was associated with higher susceptibility to injury in the mouse model of nephrotoxic nephritis (NTN) [20,21]. In these studies, cyclin I-Cdk5 was shown to activate ERK 1/2, leading to increased BCL-2/BCL-XL mRNA expression, whereas p35-Cdk5 phosphorylated Bcl-2 directly, resulting in stabilization of the protein complex. In addition, p35-Cdk5 phosphorylates the non-selective cation channel TRPC6, thereby increasing channel activity [22]. Enhanced conductance across TRPC6 is associated with podocyte dysfunction in genetic and sporadic diseases [23][24][25]. As both inhibition and promotion of Cdk5 activity may be detrimental to the podocyte, a tight control of Cdk5 kinase activity is warranted in these highly specialized glomerular epithelial cells. Strikingly, knockout of the Cdk5-activators, p35 or cyclin I or both, did not affect podocyte development or function under non-stressed conditions. In contrast, following podocyte stress in the nephrotoxic nephritis model (NTN), apoptosis was higher in the p35/cyclin I double knockout mice compared to a similar stress in the single p35 or cyclin I null mice.
To investigate a potential developmental role of Cdk5, which may be referred by yet unknown activators of the kinase, we generated podocyte-specific Cdk5 knockout mice and analyzed these animals under non-stressed conditions. In addition, we delineated the contribution of Cdk5 in states of glomerular disease, employing inducible podocyte-specific Cdk5 knockout mice and the NTN disease model.
In all animal studies, only male mice were employed. Mice were bred into mixed FVB/N and 129S4/SvJae backgrounds and maintained under standardized, pathogen-free conditions in the University of Washington animal facility, as well as in the University of Cologne animal facility. The experimental protocol was reviewed and approved by

Generation of Inducible Cdk5 Knockout
Temporal control of Cdk5 knockout in podocytes was achieved by mating Cdk5flox mice with hNPHS/rtTA/TetO-Cre mice. In the offspring, transgene transcription was controlled by a podocin-driven promoter which expressed Cre recombinase specifically in podocytes in the presence of doxycycline [30]. Doxycycline (Sigma/Merck, Darmstadt, Germany) was administered for 14 days via the drinking water (0.2 mg/mL in 5% sucrose), between the ages of 8 and 10 weeks. Water was exchanged twice weekly, and the bottles were protected from light to prevent doxycycline degradation. Transgenic mice were identified by PCR amplification. Primer sequences are listed in the following table.

Experimental Glomerular Disease
Nephrotoxic nephritis was induced by intraperitoneal injections of sheep anti-rabbit glomerular antibody, as previously described [31]. Fourteen days after the induction of Cdk5-deletion in podocytes, nephrotoxic serum (20 mg/20 g body weight) was injected intraperitoneally into 10-12 week-old mice on two consecutive days. Mice were sacrificed on day 7 after the second injection.

Immunohistochemistry
Immunohistochemical analyses were performed on kidney sections of podocytespecific Cdk5-knockout (Cdk5 pko ) and wild type mice and stained with primary antibodies listed in the Table 2. Briefly, formaldehyde-fixed, paraffin-embedded kidneys were cut into 4 µm tissue sections, deparaffinized in xylene, and rehydrated in graded alcohol. Antigen retrieval was performed by boiling kidney sections for 10 min in 10 mmol/L citrate buffer, pH 9. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide, and endogenous biotin was inhibited by Avidin/Biotin blocking kit (Vector Laboratories, Burlingam, CA, USA). Kidney sections were incubated overnight at 4 • C, with the respective primary antibody diluted in 1% PBS/BSA buffer. Subsequently, biotin-conjugated anti-rabbit secondary antibody (Jackson ImmunoResearch) was diluted in 1% PBS buffer and incubated for one hour at room temperature. Repeated washing steps were performed with PBS buffer at least three times. The ABC kit (Vector Laboratories, Burlingam, CA, USA) was used for signal amplification. Chromogen 3,3 -diaminobenzamine (DAB; Sigma, St. Louis, MO, USA) was used. Finally, sections were counterstained with hematoxylin (Sigma Aldrich, St. Louis, MO, USA), dehydrated in xylol, and mounted with Histomount (National Diagnostics, Atlanta, GA, USA).

Quantitative Assessment of Podocyte Number and Apoptosis
Podocyte number was quantified on stained paraffin-embedded kidney sections with a specific primary antibody against Wilms' tumor 1 protein (WT1) [4]. Six animals of each genotype were analyzed in a blinded fashion by counting WT1-positive cells in, at a minimum, 50 glomeruli per section. Apoptotic cells were quantified by immunohisto-chemical staining for cleaved caspase-3 on formaldehyde-fixed kidney sections. Cleaved caspase-3-positive cells per glomerulus were counted in at least 50 glomeruli per animal of each cohort (inducible podocyte-specific Cdk5-knockout (Cdk5 ipko ), wt; n = 6 each).

Imaging
All stained tissue sections were scanned with a slide scanner for brightfield images Leica SCN400 (Leica Biosystems, Wetzlar, Germany) and further assessed with ImageScope (Aperio) image-processing software v12.1 (Leica Biosystems, Wetzlar, Germany).

STED Imaging and Computed Analyses
A simplified version of a previously published protocol was used [32]. Formaldehydefixed kidneys were embedded in 3% agarose in DI water and sectioned to 200 µm thickness using a vibratome. Slices were then incubated in clearing solution (200 mM boric acid, 4% SDS, pH 8.5) at 50 • C overnight. Sections were washed in PBST (0.1% Triton-X in 1X PBS, pH 7.4) for 10 min before incubation in a sheep polyclonal antibody against nephrin (R&D systems, Minneapolis, MI, USA, AF4269) diluted 1:50 in 10 mM HEPES, pH 7.5 with 200 mM NaCl and 10% TritonX-100 at 37 • C for 4 h with shaking at 500 rpm. After primary antibody incubation, samples were washed in PBST for 5 min at 37 • C and were then incubated in a donkey anti-sheep secondary antibody conjugated to Abberior STAR635P (Abberior, Goettingen, Germany, 2-0142-007-2, dilution 1:50) at 37 • C for 4 h. Samples were incubated in 80% wt/wt fructose (1 mL of dH 2 O added to 4 g of fructose) at 37 • C for 15 min and then mounted in a MatTek dish with a cover slip on top, prior to imaging with a Leica SP8 3× gSTED system (Leica Biosystems, Wetzlar, Germany) using a 100× 1.4 NA objective. To quantify the slit diaphragm length per area, a previously published ImageJ macro was used [17].

Measurement of Proteinuria
For the assessment of proteinuria, spot urine was collected from animals of each cohort at day 0, as baseline, and at the following days as indicated. Initial analyses were performed by Coomassie blue staining after SDS-PAGE of small volumes of urine. Quantitative assessment was performed by measuring urinary protein using the sulfosalicylic acid method [33] and urinary creatinine using the Creatinine Colorimetric Assay Kit (Cayman Chemical, Ann Arbor, MI, USA). Assays were performed according to manufacturers' instructions.

Statistics
Statistical calculations were performed with PalmPrism. For histological assessment, Image Scope (Aperio Version 12.1, Leica Biosystems, Wetzlar, Germany) was used.
Cdk5 pko mice were born in a regular Mendelian ratio with no apparent developmental defects. Life expectancy was not reduced compared to Cdk5 wild type littermates (data not shown). Phenotyping included screening for proteinuria, and kidney histology was performed in Cdk5 pko mice at 70 weeks of age. Coomassie staining of urine samples showed no detectable proteinuria, specifically, no albuminuria ( Figure 2A). PAS-and AFOG-staining revealed normal kidney morphology without evidence of glomerular disease ( Figure 2B). The regular, garland-like immunostaining pattern of podocin, tracing the glomerular slit diaphragm, was similarly detected in Cdk5 pko and Cdk5 het mice ( Figure  2C). Podocyte number per glomerulus, assessed by staining for WT1, was similar in Cdk5 pko and heterozygous control mice ( Figure 2C,D). To exclude subtle structural defects of foot process architecture at the nanoscale, we utilized STED-imaging on cleared kidney tissue after immunofluorescent labeling of nephrin ( Figure 2E). STED-images revealed an evenly distributed podocin signal and a regular pattern of foot processes in all samples. Computed analysis of STED-images did not identify significant differences in slit diaphragm length between Cdk5 pko , Cdk5 het , and Cdk5 wt littermates ( Figure 2F). In podocyte-specific Cdk5flox-heterozygous mice (Cdk5 het ), Cdk5 expression was reduced to 46.31% (SD 11.86). For podocyte-specific knockout of Cdk5 (Cdk5 pko ), the remaining expression of Cdk5 was 13.1% (SD 2.26). ** p-value < 0.01, In each cohort, n = 3.
Cdk5 pko mice were born in a regular Mendelian ratio with no apparent developmental defects. Life expectancy was not reduced compared to Cdk5 wild type littermates (data not shown). Phenotyping included screening for proteinuria, and kidney histology was performed in Cdk5 pko mice at 70 weeks of age. Coomassie staining of urine samples showed no detectable proteinuria, specifically, no albuminuria ( Figure 2A). PAS-and AFOGstaining revealed normal kidney morphology without evidence of glomerular disease ( Figure 2B). The regular, garland-like immunostaining pattern of podocin, tracing the glomerular slit diaphragm, was similarly detected in Cdk5 pko and Cdk5 het mice ( Figure 2C). Podocyte number per glomerulus, assessed by staining for WT1, was similar in Cdk5 pko and heterozygous control mice ( Figure 2C,D). To exclude subtle structural defects of foot process architecture at the nanoscale, we utilized STED-imaging on cleared kidney tissue after immunofluorescent labeling of nephrin ( Figure 2E). STED-images revealed an evenly distributed podocin signal and a regular pattern of foot processes in all samples. Computed analysis of STED-images did not identify significant differences in slit diaphragm length between Cdk5 pko , Cdk5 het , and Cdk5 wt littermates ( Figure 2F). (E) Visualization of the slit diaphragm pattern on Cdk5 pko , Cdk5 het , and Cdk wt kidney sections using STED microscopy on nephrin-stained kidney samples. Yellow lines denote the slit diaphragm segmented by the ImageJ macro. The blue line denotes the ROI within which the analysis is carried out (some minor areas with poor staining quality are excluded). Scale bar 2 µm. (F) SD length per area for the different genotypes. Each dot/square/triangle shows the SD length per area for one image (5 images per animal, n = 2 per genotype). Tukey's multiple comparison test showed no significant difference between groups. Dots: Cdk5 wt ; squares: Cdk5 het ; triangles: Cdk5 pko . Line represents mean and error bars represent standard deviation.

Podocyte-Specific Knockout of Cdk5 Shows Higher Susceptibility to Glomerulosclerosis and Reduced Kidney Function in the Nephrotoxic Nephritis Model
We next assessed the biological impact of podocyte Cdk5 deficiency in glomerular disease. To exclude undetected developmental aberrations following Cdk5 deficiency, or compensatory mechanisms that might be activated in the absence of Cdk5 during development, inducible Cdk5 knockout mice (Cdk5 ipko ) were generated ( Figure 3A). In these mice, activity of Cre recombinase leading to knockout of Cdk5 was induced at the age of 8-10 weeks by administration of doxycycline in the drinking water over 14 days ( Figure 3B). Baseline urine samples were obtained before doxycycline administration and bi-weekly thereafter up to day 56. Quantitative analysis of urinary protein and creatinine revealed no relevant proteinuria at baseline ( Figure 3C). Experimental glomerular disease was induced after doxycycline-dependent Cdk5 knockout by intraperitoneal injection of nephrotoxic serum (NTS) in mice 10 weeks of age ( Figure 4A). Histological analysis and quantification of proteinuria were performed on day 7 after induction of glomerular disease. PAS-and AFOG-staining revealed glomerular sclerosis in Cdk5 ipko as well as control animals ( Figure 4B). Glomerular scarring was greater in Cdk5 knockout compared to control animals (mean grade Cdk5 ipko : 1.79 (SD Cells 2021, 10, 2464 8 of 14 0.10) vs. Cdk5 contr : 1.37 (SD 0.16); p < 0.05) ( Figure 4C). In addition, the extent of glomerular involvement (percentage of glomeruli with sclerosis) was quantified at 56% (SD 5) in Cdk5 knockout mice, as compared to 30% (SD 7) in control mice ( Figure 4D). No glomerular damage was detected in control Cdk5 ipko and Cdk5 contr given vehicle (saline) (Figure 4C,D). Nephrotic range proteinuria detected in both mouse strains at day 7 of disease was higher in Cdk5 ipko mice (Cdk5 ipko : 62.46 g/g (SD 15.35) vs. Cdk5 contr : 22.11 g/g (SD 5.10); p < 0.05). No significant proteinuria was detected in control saline-treated mice ( Figure 4E).  In each cohort, n = 6.

Cdk5-Deficient Mice Show Higher Rate of Apoptosis and Decreased Anti-Apoptotic Signal
Previously, we and others established the pro-survival signals referred by Cdk5 in podocytes and other post-mitotic cells [7,19,20,34]. In podocytes, cyclin I-Cdk5 activates MAP kinases MEK1/2, followed by promotion by ERK1/2 of the expression of the antiapoptotic genes BCL-2 and BCL-XL. In addition, p35-Cdk5 phosphorylates the Bcl-2 protein to enhance its stability [7].  In each cohort, n = 6. (C) Quantification of glomerular sclerosis according to grades of percentage of the glomerular tuft area involved. The average glomerular fibrosis index per glomerular cross section was 1.79 (SD 0.10) for Cdk5 ipko , as compared to 1.37 (SD 0.16) in Cdk5 contr after NTS application (* p < 0.05). No glomerular damage was detected in Cdk5 ipko and Cdk5 contr treated with saline instead of NTS. In each cohort, n = 6. (D) Extent of glomerular involvement quantified as percentage of glomeruli with sclerosis. Cdk5 ipko mice showed involvement of 56% (SD 5) of the glomeruli, whereas Cdk5 contr showed involvement of 30% (SD 7). In each cohort, n = 6; * p < 0.05. (E) Protein/creatinine ratio (g/g) of Cdk5 ipko compared to Cdk5 contr shows aggravated proteinuria of 62.46 (SD 15.35) g/g creatinine in Cdk5 ipko vs. 22.11 (SD 5.10) in Cdk5 contr (* p < 0.05). No significant proteinuria was detected in mice treated with saline instead of NTS. In each cohort, n = 6.

Cdk5-Deficient Mice Show Higher Rate of Apoptosis and Decreased Anti-Apoptotic Signal
Previously, we and others established the pro-survival signals referred by Cdk5 in podocytes and other post-mitotic cells [7,19,20,34]. In podocytes, cyclin I-Cdk5 activates MAP kinases MEK1/2, followed by promotion by ERK1/2 of the expression of the antiapoptotic genes BCL-2 and BCL-XL. In addition, p35-Cdk5 phosphorylates the Bcl-2 protein to enhance its stability [7].
Podocyte apoptosis, quantitated by cleaved caspase-3 (CC3) staining, was not detected in Cdk5 ipko mice or controls prior to disease induction (day 0) (Figure 5A,B). At day 7 of disease, podocyte apoptosis was threefold higher in Cdk5 ipko mice compared to control mice. At baseline, there was no significant difference in Bcl-2 staining in Cdk5 ipko mice compared to control, although, in Cdk5 knockout mice, a trend to less positive cells was noted ( Figure 5C,D). In diseased Cdk5 ipko mice, Bcl-2 staining was significantly lower compared to Cdk5 control mice. For Bcl-XL, however, already at baseline, a significant reduction of staining-positive glomerular cells was detected in Cdk5 knockout as compared to control ( Figure 5E,F). Even though they increased in both control and Cdk5 ipko mice after NTS challenge, Bcl-XL positive cells were more abundant in control compared to Cdk5 knockout tissue.

Discussion
Cdk5 is an ubiquitously expressed proline-directed serine/threonine kinase implicated in both physiological and pathological cellular processes. Amongst these are cytokine production, regulation of insulin levels, migration and invasion, angiogenesis, myogenesis, apoptosis, and senescence (reviewed in Sharma and Sicinski [35]). Importantly, tight control of Cdk5 activity is crucial since both hyperactivity and inhibition of Cdk5 result in cellular dysfunction. Noteworthy is that cell-specific regulation of Cdk5 activity is controlled by cell type-specific activators. For example, in post-mitotic neurons, p35 and p39 are the major regulators of Cdk5 activity. Combined loss of p35 and p39 mimics the phenotype of Cdk5 deficiency, with perinatal lethality due to defective development of the brain [15,36]. In contrast, mice with a combined deletion of both p35 and cyclin I develop normally and show no phenotype under non-stressed conditions [20]. The results of the present study show that podocyte-specific deletion of Cdk5 does not impact development. In addition, even in advanced age, Cdk5 deficiency in podocytes is not associated with a pathological phenotype under non-stressed conditions. These results are in striking contrast to the severe phenotype in neuronal cells following CDK5 deletion [37,38]. This finding comes not without surprise since the podocytes' branching morphology, supported by a microtubular lattice and F-actin-based foot processes, shares high similarity with the cytoskeleton of dendrite-forming neurons [39,40]. By phosphorylation of Rho-GTPases, Cdk5 plays a crucial role in controlling actin cytoskeletal modification and synaptic plasticity in neuronal cells [41,42]. In podocytes, however, highly active regulation of the actin cytoskeleton appears to be independent of Cdk5 activity and is, therefore, not affected by Cdk5 deficiency, neither during glomerulogenesis nor later in life. Quantification of podocytes per glomerular cross section was performed on the basis of immunostaining for WT1. WT1 staining detects mature podocytes and is not the ideal marker in disease states when podocytes de-differentiate and lose WT1 reactivity. In this study employing non-stressed mice, however, WT1 staining reliably estimated podocyte number. Consistent podocyte number in Cdk5 pko vs. control, reflected by the regular glomerular architecture and function, i.e., absence of proteinuria, argues against a role of Cdk5 in podocyte development.
Recently, several studies in podocytes investigated the beneficial role of the intermediate filament protein nestin in states of stress and its regulation of Cdk5/p35 inhibition of apoptosis [43][44][45]. The concept of apoptosis as a relevant mechanism of cell death in podocytes was often challenged over the past years [46,47]. A common critique arguing against apoptosis in podocytes concerns a lack of evidence of classical morphological signs of apoptosis, such as chromatin condensation, DNA fragmentation, and membrane blebbing in human glomerular disease or animal models. However, this argument disregards the three phases of apoptosis and the implication of the unique anatomical localization of podocytes in this context. The process of apoptosis progresses in three phases: induction, execution, and clearance. Caspase-mediated proteolysis, chromatin condensation, and DNA fragmentation occur during the execution phase. During the entire process of execution, plasma membrane integrity must be maintained to avoid the release of toxic waste products and injury to neighboring cells [48]. However, the first step in the execution phase is the release of focal adhesions and extracellular matrix interactions while actin rearranges to membrane-associated cortical rings [49][50][51]. For the podocyte, residing on the outer aspect of the glomerular tuft, release from the extracellular matrix and focal adhesions during the execution phase of apoptosis will result in detachment from the glomerular basement membrane and, also, from neighboring podocytes, followed by loss of podocytes in the urine. Therefore, podocytes showing classical apoptotic figures that appear later in the process of apoptosis will hardly be detected. In podocytes, the "point of no return", which is usually mediated by permeability transition pores of the mitochondrial membrane and cytochrome c release, is reached during states of cytoskeletal rearrangement and lost ECM anchoring. This emphasizes the necessity of tight control of apoptosis in podocytes during the induction phase. The lack of any direct evidence for regulated, mitochondria-triggered forms of cell death in podocytes also raises the question of which intracellular signaling events control these pathways. Staining for cleaved caspase-3 and TUNEL staining are most commonly used to assess for apoptotic cells in tissue sections. However, TUNEL staining does not discriminate between apoptosis and other modes of cell death reliably [52]. Therefore, in the present study, apoptotic cells were detected by staining for cleaved caspase-3. We acknowledge that cleaved caspase-3 staining may underestimate the number of apoptotic podocytes, due to the loss of podocytes early in the process of apoptosis. Quantification of Bcl-2 and Bcl-XL expression was performed to assess the role of Cdk5 during initiation of apoptosis. Increased expression levels of both anti-apoptotic proteins recapitulated the mechanism proposed previously in which Cdk5/p35 increases BCL-2/BCL-XL expression and, in addition, stabilizes Bcl-2 at the protein level [7].
In conclusion, our study demonstrated that Cdk5 was not implicated in podocyte development but served a central role in the regulation of podocyte apoptosis. Cdk5 activity is crucial after toxic stimuli to avoid initiation of apoptosis, cytoskeletal remodeling, detachment from the glomerular basement membrane, and, eventually, loss of podocytes in the urine. Several studies on Cdk5 in neurological and oncological disorders showed good transferability of results gained in mice to human disease [53,54]. Consequently, it is tempting to speculate that activation of Cdk5 could be a novel therapeutic approach in human inflammatory glomerular disease.