High glucose–induced Smad3 linker phosphorylation and CCN2 expression are inhibited by dapagliflozin in a diabetic tubule epithelial cell model

Abstract Background: In the kidney glucose is freely filtered by the glomerulus and, mainly, reabsorbed by sodium glucose cotransporter 2 (SGLT2) expressed in the early proximal tubule. Human proximal tubule epithelial cells (PTECs) undergo pathological and fibrotic changes seen in diabetic kidney disease (DKD) in response to elevated glucose. We developed a specific in vitro model of DKD using primary human PTECs with exposure to high D-glucose and TGF-β1 and propose a role for SGLT2 inhibition in regulating fibrosis. Methods: Western blotting was performed to detect cellular and secreted proteins as well as phosphorylated intracellular signalling proteins. qPCR was used to detect CCN2 RNA. Gamma glutamyl transferase (GT) activity staining was performed to confirm PTEC phenotype. SGLT2 and ERK inhibition on high D-glucose, 25 mM, and TGF-β1, 0.75 ng/ml, treated cells was explored using dapagliflozin and U0126, respectively. Results: Only the combination of high D-glucose and TGF-β1 treatment significantly up-regulated CCN2 RNA and protein expression. This increase was significantly ameliorated by dapagliflozin. High D-glucose treatment raised phospho ERK which was also inhibited by dapagliflozin. TGF-β1 increased cellular phospho SSXS Smad3 serine 423 and 425, with and without high D-glucose. Glucose alone had no effect. Smad3 serine 204 phosphorylation was significantly raised by a combination of high D-glucose+TGF-β1; this rise was significantly reduced by both SGLT2 and MEK inhibition. Conclusions: We show that high D-glucose and TGF-β1 are both required for CCN2 expression. This treatment also caused Smad3 linker region phosphorylation. Both outcomes were inhibited by dapagliflozin. We have identified a novel SGLT2 -ERK mediated promotion of TGF-β1/Smad3 signalling inducing a pro-fibrotic growth factor secretion. Our data evince support for substantial renoprotective benefits of SGLT2 inhibition in the diabetic kidney.


Abstract Background
In the kidney glucose is freely filtered by the glomerulus and, mainly, reabsorbed by sodium glucose cotransporter 2 (SGLT2) expressed in the early proximal tubule. Human proximal tubule epithelial cells (PTECs) undergo pathological and fibrotic changes seen in Diabetic Kidney Disease (DKD) in response to elevated glucose. We developed a specific in vitro model of DKD using primary human PTECs with exposure to high D-glucose and TGF-1 and propose a role for SGLT2 inhibition in regulating fibrosis.

Methods
Western blotting was performed to detect cellular and secreted proteins as well as phosphorylated intracellular signalling proteins. qPCR was used to detect CCN2 RNA. Gamma glutamyl transferase (GT) activity staining was performed to confirm PTEC phenotype. SGLT2 and ERK inhibition on high D-glucose, 25mM, and TGF-1, 0.75ng/ml, treated cells was explored using dapagliflozin and U0126, respectively.

Results
Only the combination of high D-glucose and TGF-β1 treatment significantly upregulated CCN2 RNA and protein expression. This increase was significantly ameliorated by dapagliflozin. High Dglucose treatment raised phospho ERK which was also inhibited by dapagliflozin. TGF-β1 increased cellular phospho SSXS Smad3 serines 423 and 425, with and without high D-glucose. Glucose alone had no effect. Smad3 serine 204 phosphorylation was significantly raised by a combination of high D-glucose+TGF-β1; this rise was significantly reduced by both SGLT2 and MEK inhibition.

Conclusions
We show that high D-glucoseand TGF-1 are both required for CCN2 expression. This treatment also caused Smad3 linker region phosphorylation. Both outcomes were inhibited by dapagliflozin.
We have identified a novel SGLT2 -ERK mediated promotion of TGF-1/Smad3 signalling inducing a pro-fibrotic growth factor secretion. Our data evinces support for substantial renoprotective benefits of SGLT2 inhibition in the diabetic kidney.

Introduction
Diabetic kidney disease (DKD) is a multifactorial condition and major complication of diabetes.
Data indicates that between 20 and 40% of patients with diabetes develop diabetic kidney disease.
The global prevalence of diabetes is predicted to rise from 450 million to 600 million by 2040 (1).
With the number of patient diagnoses only expected to rise, the battle to prevent established kidney disease remains a prevalent healthcare challenge. The result of chronic exposure to raised glucose and spikes of severe hyperglycaemia results in pathological changes to the kidney including mesangial expansion, glomerular and tubular hypertrophy along with glomerulosclerosis, and tubulointerstitial fibrosis (2). The severity of tubulointerstitial fibrosis is an important determinant of progressive loss of renal function (3,4) and failure to intervene early in the unrelenting fibrotic process can lead to end stage renal disease. The identification of novel biomarkers of renal disease in diabetic patients such as cystatin C and kidney injury molecule-1, have contributed to an important shift to a 'tubulocentric' view on the development and progression of DKD with the Proximal Tubule Epithelial Cell (PTEC) being the major cell type involved in this perspective (5,6).
Sodium glucose transporters (SGLT2) inhibitors, also known as gliflozins, are the most recent class in a long line of anti-diabetic drugs approved by the FDA (7). The underlying principle behind these gliflozins is that by blocking glucose entry at the PTEC via the inhibition of the transporter, glucose reabsorption is reduced, thus inducing glycosuria and decreasing plasma glucose concentrations (8,9). Mutations in the SGLT2 gene are linked to familial renal glycosuria where abnormal glucose excretion into the urine occurs with no adverse effects on their health (10)(11)(12). Evidence from these patients indicated that SGLT2 could be targeted to increase glucose loss in the urine without deleterious effects. Phlorizin was the first naturally occurring gliflozin to be discovered but was found to be a non-specific SGLT Inhibitor and dapagliflozin was later developed as an alternative.
Additionally, a type 2 diabetic db/db mouse study found that treatment with the SGLT2 inhibitor Empagliflozin also reduced tubulointersitial fibrosis, observed through the significant decrease of 1(I) collagen and fibronectin abundance in kidney sections (14). Thus, we hypothesised that reducing PTEC glucose uptake with the SGLT2 inhibitor dapagliflozin would result in a reduction in the fibrotic response in human cells. In addition to a chronic difficulty regulating circulating glucose Downloaded from http://portlandpress.com/bioscirep/article-pdf/doi/10.1042/BSR20203947/911416/bsr-2020-3947-t.pdf by guest on 19 May 2021 patients with diabetes can suffer from rapid spikes in blood glucose levels which may not be reflected in HbA1c levels. We developed a model using primary human PTEC with acute exposure to high glucose to mimic these spikes.
The current diagnostic criteria for DKD include established diabetes and albuminuria. A consequence of PTEC exposure to albumin is the megalin independent expression of TGF-1 (15).
TGF-1 is a pleiotropic protein involved in a plethora of functions including fibrosis. One of the major molecular pathways driving renal fibrosis is believed to be the canonical TGF-1/Smad3 pathway. Results published by Zorena et al indicated that serum TGF-β1 levels was the most "discriminate power" when compared to several other variables including blood pressure, HbA1c and creatinine in predicting diabetic outcomes in a group of "juvenile patients". They determined the cut-off threshold concentration for TGF-β1 where it relates to diabetes induced complications was 0.443ng/ml. We have therefore incorporated a threshold concentration of TGF-1 as part of our in vitro model (16).
Connective tissue growth factor (CCN2), a secreted 36/38kDa pro-fibrotic marker, has also been implicated in DKD (17). It has a characteristic doublet consistent with partial glycosylation and is known to act downstream of TGF-β1 (18). FG-3019, a fully human monoclonal neutralising antibody against CCN2, has been shown to reduce fibrosis and significantly reduce albumin creatinine ratio in diabetic patients, supporting the proposal that CCN2 is an effective target for the treatment of DKD (19). A clinical study showed that the abundance of urinary CCN2 N terminal fragments increased 10 fold when comparing microalbuminuric and normoalbuminuric diabetic patients, directly correlating with the severity of proteinuria and therefore the rate of DKD progression (20).
In this work, we investigate the induction of CCN2 in primary human PTECs and have sought to determine the molecular basis underlying the pro-fibrotic response. We have delineated a role for SGLT2 in the regulation of a pro-fibrotic number of intracellular signalling cascades and hence, offer a novel target for the treatment of tubulointerstitial fibrosis in DKD.
Fibroblasts were cultured in complete fibroblast medium (Cell Biologics). All medium were sterile filtered using a 0.2uM Puradisc™ (Whatman) and changed on alternate days. All cells were grown under 5% CO 2 humidity at 37°C on T75 tissue culture flasks and 35mm culture dishes (Corning).

Collagen IV Coating
Human placenta collagen IV (Sigma) was reconstituted in sterile 0.25% acetic acid and coated onto both tissue culture flasks and dishes (Corning) at 5g/cm 2 . After all culture-ware with collagen IV was exposed to UV radiation for 20 mins, the coating was left to incubate overnight at room temperature. Flasks and dishes were then rinsed once using sterile PBS and air dried before use.

Cell Subculture and Cryopreservation
After cells reached approximately 85% in confluence, the culture medium was removed and cell dissociated with trypsin-EDTA solution for 5 min at 37°C. When detached, the cell suspension was aspirated and centrifuged at 1500rpm(327g) for 6 min. The supernatant was discarded and the pellet re-suspended in fresh medium. Live cells were counted after mixing with trypan blue solution (Sigma) using an automated cell counter (Bio-Rad). PTECs and fibroblasts were seeded onto 35mm collagen IV coated dishes at a density of 2500 and 5000 cells/cm 2 respectively.

Treatment and Inhibitors
Cells were serum starved for 24 h prior to treatment. All subsequent conditions were supplemented with 0.1% BSA. Cells were exposed to 3 different conditions: (i) 7mM D-glucose (control); (ii) 7mM D-glucose + 18mM D-glucose (high D-glucose or D-Glu); and (iii) 7mM D-glucose + 18mM L-glucose (osmotic control or L-Glu), with or without TGF-1 0.75ng/ml (Sigma-Aldrich). We used L-glucose to account for the osmotic stress the cells are subjected to upon >7mM glucose treatment as osmotic stress is known to activate ERK in renal epithelial cells (22). The SGLT2 inhibitor dapagliflozin (Cayman Chemical) was administered at 0.1, 1, or 10nM while MEK/ERK inhibitor U0126 (Sigma-Aldrich) was administered at 10μM. Treatment times ranged from 5 min to 24 h.

Protein Lysis and Assay
The medium was removed from cell cultures and stored at -70°C. Cells were then rinsed with ice cold PBS. To obtain whole cell protein, cells were scraped from dishes and homogenised in ice cold lysis buffer (Tris-HCL 20mM at pH 7.2, EDTA 1mM, SDS 0.1%, sodium deoxycholate 0.5%, Triton X-100 1%) and 10ml/l EASYpack protease inhibitor and PHOSstop phosphatase inhibitor cocktails (Roche). This mixture was placed on ice for 15 min prior to centrifugation for 10 min at 10000rpm (6720g). The lysed protein in the supernatant was then removed and stored at -70°C until use.
Bicinchoninic acid assay (Pierce) was used to determine protein concentration of the lysate as per the manufacturer's protocol.
Immunostained proteins were also captured on the ChemiDoc™ imager. Protein abundance analysis was carried out using ImageLab™ software (Bio-Rad) where target protein density was expressed relative to the amount of total protein. Polyacrylamide gel containing a proprietary trihalo compound to make proteins fluorescent directly in the gel with a short photoactivation, allowing the immediate visualization of proteins. Following polyacrylamide gel elxtropheresis. Gels were placed in a ChemiDoc™ Imaging (Bio-Rad) system fo activation by exposure to UV light for 1 minute. A stainfree image was taken using the imaging system for total protein measurement in each lane. Image data was analysed using Image 4.1 software (Bio-Rad). Stain-Free gels provide a linear dynamic range between 10 and 80 µg of total protein load. The total density for each lane is measured from the blot and a lane profile is obtained. The background is adjusted in such a way that the total background is subtracted from the sum of density of all the bands in each lane (referred to as the rolling disk background subtraction algorithm).The software interprets the raw data in three dimensions with the length and width of the band defined by the "Lanes and Bands" tool in concert with the "Lane Profile" tool such that the chemiluminescent signal emitted from the blot is registered in the third dimension as a peak rising out of the blot surface. The density of a given band was measured as the total volume under the three-dimensional peak, which could be viewed in two dimensions using the "Lane Profile" tool to adjust the precise width of the band to account for the area under the peak of interest. An example total protein gel is presented in figure 1.

Heparin Bead Purification
For the analysis of proteins with heparin binding domains, media from treated cells were mixed together at a 10:1 ratio with heparin-agarose bead suspension (Sigma-Aldrich). Medium and bead suspension were mixed on a roller for 6h. Tubes were then centrifuged at 4°C for 10 min at 10000rpm (10060g) (Hettich Mikro), and the supernatant was discarded. The beads were then rinsed in ice cold PBS twice and heparin bound proteins were subsequently extracted and separated as detailed in the above western blot section.

qPCR
Complementary DNA (cDNA) was produced via a reverse transcription polymerase chain reaction (RT-PCR) using the High Capacity cDNA Reverse Transcription kit (Life Technologies). Sample cDNA, glyceraldehyde 3-phosphate dehydrogenase (GAPDH) housekeeping primers, CCN2 Taqman gene expression assay and Taqman gene expression master mix were then mixed together on a 96-well PCR plate (Bio-Rad). Real time quantification of gene target(s) was carried out using the CFX1000 Real Time system thermocycler (Bio-Rad) under the pre-optimised qPCR thermal cycler program. CFX manager software was used to analyse the results. The fold changes in cDNA expression were presented as values normalised to the housekeeping gene using the ∆Ct method. The following primer probe sets (Life Technologies) were all pre-validated at standardised conditions:

Gamma GT Staining
Confluent dishes of PTECs were stained with a solution of 6mM glutamic acid γ-(4-methoxy-βnaphthylamide), 4mM glycyl-glycine, 1.1mM Fast Blue B Salt, 100mM NaCl, and 25mM Tris at pH 7.4 (Sigma Aldrich) in deionised water for 45 min at room temperature. Cells were then washed with PBS before examination under the microscope. Distinct orange-red staining confirmed presence of the enzyme.

Statistical Analysis
Statistical analysis on all treatments was performed on GraphPad Prism 7 (GraphPad Software Inc) where P values < 0.05 were considered statistically significant. Comparison between two means was carried out using unpaired t-test. Comparison of the means of more than two groups was carried out using analysis of variance (ANOVA) with Bonferroni's post hoc test. Values are reported as mean  standard deviation (SD).
The use of all human tissue was carried out with the approval of the South London and Surrey Research Ethics Committee, REC reference number 08/H0806/8.

Gene
Assay ID Reference Sequence

The Effect of High Glucose and TGF-β1 on CCN2 Production in Human PTECs
Upon treatment, neither exposure to TGF-1 0.75ng/ml nor 25mM D-Glucose (D-Glu) was found to be sufficient to produce a significant increase in CCN2 secretion within 24 h. However, the combination of D-Glu+ TGF-1 significantly increased the secretion of the glycosylated and nonglycosylated isoforms of CCN2, 38 and 36 kDa (P<0.05, n=3) ( Fig. 1A-C).

The Effect of SGLT2 Inhibition on CCN2 Protein Secretion and mRNA Expression
TGF-1 (0.75ng/ml) induction of CCN2 was dependent on high D-glucose; this was seen in both

Investigating the Point of Convergence of High Glucose and TGF-1 stimulation
We next investigated putative signalling pathways that may be involved in facilitating this synergistic upregulation of CCN2. Based on our previous work (22)

Discussion
We have studied the role of extracellular D-glucose in mediating TGF-β1/Smad3 signaling and modulating the expression of CCN2 in human proximal tubule cells. The CCN2 promoter contains a Smad binding element (SBE) which is a well-documented target of TGFβ1/Smad3. There are 2 apical glucose transporters expressed in the proximal tubule, SGLT1 and SGLT2. SGLT2 is a high capacity/low affinity glucose transporter primarily localised in the S1 and S2 segment of the proximal tubule whilst SGLT1 is a low capacity, high affinity glucose transporter found in the S3 segment. Our study specifically investigated SGLT2 using dapagliflozin which has an approximate 1200 fold selectivity for SGLT2 over SGLT1 (13). The primary PTECs used in these experiments were from 3 separate donors and contain cells derived from all three segments. Therefore we expect that our results may differ from those using We chose to normalise our target band densities against the total protein visualised in our electrophoresis gel using the trademarked Stain-Free imaging technology recently developed by Biorad. As briefly mentioned in our results section, this technology employs a patented trihalocompound in a polyacrylamide gel which strengthens the fluorescent signal of tryptophan amino acids after a short photo-activation upon exposure to UV light. As these fluorophores are covalently bound to the protein, they can be easily captured without the need for a laborious staining and de-staining protocol. We opted to use this method instead of the traditional housekeeping protein normalisation, which is often unreliable, as it allows researchers to acquire more accurate data by dividing band densities with total lane protein.
Our study has investigated the efficacy of the diabetic class of drugs gliflozins on ameliorating diabetes associated fibrotic changes in human renal cells by targeting the underlying molecular mechanisms. There is currently no pharmacotherapy for renal fibrosis. We have focused on dapagliflozin, which was licensed for use in the US in October 2019 after the DECLARE-TIMI 58 trial showed that the drug successfully reduced the risk of hospitalization for heart failure (HF) in adults with type 2 diabetes (25). The data we have presented is derived from a cell culture model of DKD. Cells were exposed to raised glucose levels, to replicate hyperglycaemic peaks seen in patients, and TGF-β1. Albuminuria is a common characteristic of DKD and albumin stimulates TGF-β1 expression in PTECs. There is also evidence for significantly elevated urinary TGF-β1 in patients with diabetes when compared to healthy controls (26). We believe this model recreates many important features of diabetic tubular disease observed in patients and can be a valuable tool for the study of some aspects of DKD, particularly the role of SGLT2.
TGF-1 is a pleiotropic protein involved in a plethora of crucial functions including cell differentiation, wound repair and inflammation (27,28 (29). TGF-β1 is a key driver of fibrosis in the kidney and has been proposed as a target for chronic kidney disease since 1998 (30). TGF-β1/Smad3 signalling has been suggested as the main molecular mechanism in renal fibrosis due to several studies with various models (31). However, targeting TGF-β in renal disease has proved difficult (32). Alternative therapies, particularly those using drugs already licensed for use are attractive options.
CCN2 is a modular matricellular protein that acts downstream of TGFβ1 (33). We observed a clear and significant increase in CCN2 mediated by SGLT2 when cells were exposed to a combination of TGF-1 and elevated glucose (25mM) for 24h. SGLT2 involvement was deduced from the experiment using dapagliflozin. There is a discrepancy in the effect of dapagliflozin between protein and mRNA, 0.1nM only significantly inhibited protein expression. We would postulate that this is due to the fact that CCN2 is an early immediate gene with transient RNA expression and the lowest concentration of dapagliflozin used by us flattens the curve but does not result in a significant effect at 24 hours.
Work from ourselves and others have previously shown that TGF-1 is able to induce CCN2 on its own, after 12 to 24 h in primary PTECs and other cell types only when the concentration is 5ng/ml or higher (22,34). This concentration of TGF-1 also induces epithelial mesenchymal transition (EMT) in breast cancer cells and renal PTEC (35,36). There are no published reports of PTEC EMT in human DKD. In this study, we did not observe any loss of phenotype after the cells were subjected to a combination of a lower concentration of TGF-1 in the presence of hyperglycaemia, suggesting PTECs retained their epithelial phenotype at the time of CCN2 production and secretion. The effects we observed were dependent on both glucose concentration and TGF-1, which provides a possible reason why DKD only occurs in a minority of patients with diabetes and strengthens the argument that fibrosis may depend on a two-hit model. In contrast, a research group recently found that high glucose (25mM) treatment alone was sufficient enough to induce a significant accumulation of fibronectin and collagen IV protein, commonly observed during tubulointerstitial fibrosis (37).
However, it is worth noting that they did not use a primary cell line from multiple donors but instead utilised a transformed HK-2 cell line which is less archetypal of the in vivo situation. In addition, Wu and Derynk (38) previously published a glucose induced Smad3 C terminal phosphorylation, not observed in the work presented above. This may also be due to the difference in the cell model as they utilised murine cells grown on plastic culture dishes.
We found that hyperglycaemia alone and in combination with TGFβ1 increased phosphorylated However, the concentration used from this group was 0.25μM (42). The IC50 for dapagliflozin inhibition of SGLT2 is reported to be approximately 1nM (43), and the authors concluded that the effect observed was likely to be independent of SGLT2 inhibition.
In our experiments the presence of 25mM glucose with TGFβ1 induced a cascade of signalling proteins and gene expression that was sensitive to SGLT2 inhibition. Glucose entry through SGLT2 is accompanied by equimolar sodium entry, hence increased glucose entry is accompanied by increased sodium entry. PTEC sodium levels are closely controlled by a number of apical and basolateral sodium channels; SGLT is the only glucose channel in PTEC. Hence, it is unlikely that the effects observed by us are a result of raised intracellular sodium. Furthermore, glucose-induced ERK has been observed in cells that do not express sodium/glucose co-transport, as described above (41,44). Never the less the effect of intracellular sodium on Smad signalling remains undefined and may merit enquiry. promoter activation. The constructs they created that specifically lacked serine 204 also reduced Smad3 2(I) collagen promoter activity, further indicating a key role for it in the development of renal fibrosis (23). In the same cells, TGF-1 was able to phosphorylate the serines within the linker region of Smad3, and this was similarly reversed upon MEK inhibition (46). Taken together these data strongly support the case that targeting SGLT2 could also inhibit the Glucose-TGF-1/ERK/Smad3 signalling cascade and subsequent key elements of renal fibrosis in our PTECs.  (49). We propose a novel TGFβ/Smad3 signalling pathway where a dimer consisting of a full length Smad3 and a truncated form is phosphorylated on the former at the C-terminal SSXS at receptor level and later by ERK on the linker region. This proposed pathway merits further investigation.
In summary, we have identified a mechanism for an SGLT2 mediated fibrogenic pathway in a diabetic milieu and a cellular signalling cascade mediating the pro-fibrotic effect. We propose that high glucose acts on or via SGLT2 to activate MEK/ERK signalling while TGF-1 is simultaneously activating its own receptor. These two events converge on different phosphorylation sites on Smad3 leading to upregulation of pro-fibrotic protein CCN2. Our results provide molecular basis for improvements in renal outcomes demonstrated in patients treated with an SGLT2 inhibitor.
Inhibition of SGLT2 may provide a novel therapeutic approach for the treatment of DKD by limiting tubular interstitial fibrosis.

Conflicts of Interest
No conflicts of interest.

Availability of Data and Materials
The datasets/resources generated during and/or analysed for the current study are available from the corresponding author upon reasonable request.

Funding
This project is jointly funded by Kidney Research UK (Grant no. ST_008_20151127) and the South West Thames Kidney Fund (Grant no. TKF PhD2017).

Competing Interests
The authors declare that they have no competing interests.      There was no expression of SGLT2 expression in the fibroblasts under our specific conditions.