TGFβ modulates cell-to-cell communication in early epithelial-to-mesenchymal transition
- First Online:
- Cite this article as:
- Hills, C.E., Siamantouras, E., Smith, S.W. et al. Diabetologia (2012) 55: 812. doi:10.1007/s00125-011-2409-9
A key pathology in diabetic nephropathy is tubulointerstitial fibrosis. The condition is characterised by increased deposition of the extracellular matrix, fibrotic scar formation and declining renal function, with the prosclerotic cytokine TGF-β1 mediating many of these catastrophic changes. Here we investigated whether TGF-β1-induced epithelial-to-mesenchymal transition (EMT) plays a role in alterations in cell adhesion, cell coupling and cell communication in the human renal proximal tubule.
Whole-cell and cell compartment abundance of E-cadherin, N-cadherin, snail, vimentin, β-catenin and connexin-43 was determined in human kidney cell line (HK)2 and human proximal tubule cells with or without TGF-β1, using western blotting and immunocytochemistry, followed by quantification by densitometry. The contribution of connexin-43 in proximal tubule cell communication was quantified using small interfering RNA knockdown, while dye-transfer was used to assess gap junctional intercellular communication (GJIC). Functional tethering was assessed by single-cell force spectroscopy with or without TGF-β1, or by immunoneutralisation of cadherin ligation.
High glucose (25 mmol/l) increased the secretion of TGF-β1 from HK2 cells. Analysis confirmed early TGF-β1-induced morphological and phenotypical changes of EMT, with altered levels of adhesion and adherens junction proteins. These changes correlated with impaired cell adhesion and decreased tethering between coupled cells. Impaired E-cadherin-mediated adhesion reduced connexin-43 production and GJIC, these effects being mimicked by neutralisation of E-cadherin ligation. Upregulation of N-cadherin failed to restore adhesion or connexin-43-mediated GJIC.
We provide compelling evidence that TGF-β1-induced EMT instigates a loss of E-cadherin, cell adhesion and ultimately of connexin-mediated cell communication in the proximal tubule under diabetic conditions; these changes occur ahead of overt signs of renal damage.
KeywordsCell adhesionCell communicationDiabetic nephropathyEpithelial-to-mesenchymal transitionFibrosisGap junctionsProximal tubule
Gap junctional intercellular communication
Human kidney cell line
Human proximal tubule cells
Small interfering RNA
Small mothers against decapentaplegic
Tubular basement membrane
Transforming growth factor beta receptor
The crucial pathology underlying progressive chronic kidney disease in diabetes is tubulointerstitial fibrosis [1, 2]. Central to this process is epithelial-to-mesenchymal transition (EMT) or the trans-differentiation of tubular epithelial cells into myofibroblasts [3–5]. Overwhelming evidence implicates TGF-β1 as the predominant cytokine mediating these phenotypical fibrotic changes [6, 7]. In diabetes, production of TGF-β1 in the proximal tubule is stimulated by high glucose [8, 9]. TGF-β1 modulates the production of several epithelial cell recognition and organisational proteins, while contributing to the reciprocal loss of tubular epithelial cells and accumulation of interstitial fibroblasts, changes associated with declining excretory function [10–12]. In EMT, the loss of epithelial characteristics, e.g. epithelial (E)-cadherin and the zonula occludens protein-1, coincides with the acquisition of proteins associated with a mesenchymal phenotype, e.g. α-smooth muscle actin, fibroblast-specific protein and vimentin, a process that culminates in cytoskeletal remodelling and disruption of the tubular basement membrane [13, 14]. Loss of cell adhesion, associated with reduced E-cadherin levels, represents a pivotal step in those early phenotypical and morphological changes previously observed in response to TGF-β1-induced tubular injury . Cadherins have a central role in forming the multi-protein adherens junction that links cell-to-cell contact to the actin cytoskeleton and various other signalling molecules . The extracellular domain mediates ligation with E-cadherin on adjacent cells , while the cytoplasmic domain binds to β-catenin, linking cadherin to the actin cytoskeleton via α-catenin. Interaction, via the catenins, of cadherin with F-actin not only increases the adhesive strength of the junction, but also acts as a signalling ‘node’ for proteins that influence adhesiveness and/or initiate intracellular signalling. Co-localised with E-cadherin and β-catenin at the sites of cell-to-cell contact , connexins oligomerise into hexameric hemichannels (connexons) that connect the cytoplasm of adjoining cells and form gap junctions (GJs). Gap junctions allow transfer of solutes, metabolic precursors and electrical currents , and are essential for synchronising activity to ensure appropriate function. Inhibition of cadherin-based cell adhesion inhibits GJ assembly , while production of recombinant cadherins into cells lacking strong coupling increases connexin phosphorylation at the adherens junction  and increases cell-to-cell communication . Since intercellular adhesion precedes GJ formation and inhibition of cadherin-based cell adhesion is known to inhibit GJ assembly, we hypothesised that glucose-evoked increases in TGF-β would compromise cell communication and function in the proximal tubule.
In retinal capillaries of diabetic mice, connexin-43 production is reduced and apoptosis increased, resulting in a loss of cell communication, and in a decline in the number of pericytes and acellular capillaries . This suggests that a loss of connexin production may be crucial in the development of the vascular lesions observed in diabetic retinopathy. Similar findings from vascular endothelial cells confirm that high glucose decreases connexin production and function, and that this is an early trigger of apoptosis . These data highlight the importance of GJ-mediated cell coupling and suggest that a loss of cell-to-cell communication may contribute to some microvascular complications of the disease. Glucose decreases GJ conductance and disrupts cellular homeostasis in various cell systems [25, 26], and glucose-dependent downregulation of connexin-43 and of GJ communication has been reported in bovine retinal pericytes , and in endothelial [28, 29] and epithelial cells . While the presence of GJs in the kidney has long been known, details on their function in the proximal tubule are sparse. Studies on renal vasculature have confirmed a role for various connexins in renin secretion and the regulation of blood pressure , but minimal data exist on their role in tubular function, where levels are also high. The novel findings presented here demonstrate a link between high glucose, TGF-β1, impaired cell adhesion and reduced GJ abundance in the proximal tubule. These changes have profound effects on overall cellular integrity and function, and may be among the key events orchestrating loss of function in diabetic nephropathy.
Materials for tissue culture
Tissue culture supplies were purchased from Invitrogen (Paisley, UK). Immobilon P membrane was from Millipore (Watford, UK) and electrochemoluminescence from Amersham Biosciences (Amersham, UK). A Qproteome kit was obtained from Qiagen (Crawley, UK). Antibodies and small interfering RNA (siRNA) were obtained from Santa Cruz (Santa Cruz, CA, USA), R&D Systems (Abingdon, UK) and Affinity Bioreagents (Cambridge, UK). Recombinant human TGF-β1, fibronectin, lipofectamine and Lucifer yellow were obtained from Sigma (Poole, UK), as were all other general chemicals. The anti-TGF-β1 ELISA was from R&D Systems.
Model cell line
Human kidney cell line (HK)2 cells (passages 18–30) were maintained in DMEM/Hams F12 medium, which was supplemented with 10% FCS wt/vol, glutamine (2 mmol/l) and EGF (5 ng/ml), and were cultured at 37°C in a humidified atmosphere with 5% CO2. Prior to treatment, cells were transferred to DMEM/F12 low glucose (5 mmol/l) for 48 h as described previously . Cells were serum-starved overnight before applying either TGF-β1 (2–10 ng/ml), anti-E-cadherin-neutralising antibody (20 μg/ml) or anti-N-cadherin-neutralising antibody (10 μg/ml) for 48 h. To assess the effect of high glucose, cells were treated with 5 mmol/l (control), 25 mmol/l (high) glucose or 25 mmol/l mannitol (osmotic control) for 48 h or 7 days.
Human proximal tubular cells
Following patient consent and ethical approval from South Staffordshire Research Ethics Committee (application number 08/H1203), cells were obtained from anonymised nephrectomy procedures for renal carcinoma. Renal cortex was longitudinally sectioned, the fibrous capsule removed and 1 cm3 portions cut from the outer region. Pieces were placed into DMEM and further cut into 1 mm3 sections. Each piece was placed into a well of a 24-well plate that had been previously coated with gelatine for 20 min, followed by incubation in FCS overnight. Sections were cultured in DMEM/Nutrient Ham’s F12 1:1, which was supplemented with 5 mg/ml insulin, 5 mg/ml transferrin, 5 ng/ml sodium selenite, 36 ng/ml hydrocortisone, 4 pg/ml tri-iodothyronine, 10 ng/ml EGF, 2 mmol/l glutamine, 10,000 U/ml penicillin and 10,000 mg/ml streptomycin. Culture was at 37°C in atmosphere with 5% CO2. Immunohistochemical staining showed cells to be positive for cytokeratin, human epithelial antigen and alkaline phosphatase, but negative for factor VIII-related antigen and actin.
Quantification of TGF-β1
Total TGF-β1 was measured by specific ELISA of cell culture supernatant fractions collected from growth-arrested HK2 cells stimulated under serum-free conditions. Active TGF-β1 was measured directly, while latent TGF-β1 was measured indirectly following acid activation of samples. The assay used has <1% cross-reactivity for TGF-β2 and TGF-β3. The TGF-β1 concentration was normalised to mg/ml of protein. Quantities of TGF-β1 are expressed as pg ml−1 (mg protein)−1.
Cytosolic proteins were prepared and separated by gel electrophoresis and electro-blotting on to Immobilon P membranes as described previously . For determination of protein localisation, proteins were collected using the Qproteome (Qiagen) cell compartment kit. Membranes were probed with specific polyclonal antibodies against anti-E-cadherin (R&D Systems), N-cadherin (Sigma), snail (R&D Systems), vimentin (Affinity Bioreagents), β-catenin (Santa Cruz) and connexin-43 (Santa Cruz) at dilutions of 1:1,000, 1:500, 1:500, 1:800, 1:1,000 and 1:400 respectively.
Cells were grown to 40% confluence in six-well plates or on cover-slips treated with 3-aminopropyltriethoxy-silane. Knockdown of CX43 (also known as GJA1) expression was achieved using siRNA. Transfection of siRNAs was carried out using lipofectamine as described previously . Cells were collected and assayed at 48, 72 and 96 h after transfection. Negative controls included untransfected cells, lipid alone, and two scrambled siRNAs, one of which was fluorescein-conjugated. CX43 knockdown was confirmed by Western blot analysis.
Cells at 80% confluence were fixed with 4% paraformaldehyde. After blocking, the nuclear stain DAPI (1 mmol/l) was added for 3 min. Cells were then either incubated for 1 h at 25°C with tetramethyl rhodamine isothiocyanate (TRITC)-conjugated phalloidin (Sigma), diluted at 1:100, in PBS-Triton, or they were incubated overnight at 4°C with primary antibody (1:100) diluted in PBS-Triton. Candidate proteins were visualised using Alexa488-conjugated secondary antibody (1:400) in PBS-Triton for 1 h at 25°C. Fluorescence was visualised using a fluorescence microscope (Axiovert 200; Carl Zeiss, Welwyn Garden City, UK).
Lucifer yellow was dissolved in 250 μl fresh LiCl (150 mmol/l) with HEPES (10 mmol/l; pH 7.2). Individual cells within a cell cluster were injected using a delivery system (Injectman/Femtojet 5247; Eppendorf, Hamburg, Germany). The duration of injection was set at 1 s, with an injection pressure of 14,000 Pa and a compensation pressure of 4,800 Pa. Dye transfer between coupled cells was recorded over 4 min using a software package (Metamorph; Molecular Devices, Sunnyvale, California, USA) and a Cool Snap HQ CCD camera (Roper Scientific, Gottingen, Germany).
Single cell force spectroscopy
Atomic force microscope force spectroscopy (CellHesion module; JKP Instruments, Berlin, Germany) was used to measure cell-to-cell adhesion and the separation forces required to uncouple these cells. A single HK2 cell was bound to a cantilever using fibronectin (20 mg/ml) and poly-l-lysine (25 μg/ml), and subsequently brought into contact with an adherent cell (in a cluster of coupled cells) using a known force (1 nN). The two cells remained in contact for a defined period of time (10 s) while bonding formed. The cantilever was then retracted at a constant speed (5 μm/s), and force (in nanonewton) versus displacement (deflection of the cantilver) was measured using a laser until the cells were completely separated (pulling length 60–80 μm). Each cell–cell recording was repeated in triplicate with a 30 s interval between successive measurements. Retraction recordings from multiple cells (approx. 50) in separate experiments (n = 5) were made and the maximum unbinding force (in nanonewton) and detachment energy (in femtojoules) calculated.
Autoradiographs were quantified by densitometry using a device (TotalLab 2003; NonLinear Dynamics, Durham, NC USA). Where data were quantified, the non-stimulated, low-glucose control condition was normalised to 100% and data from all other experimental conditions compared with this. Statistical analysis of data was performed using a one-way ANOVA test with Tukey’s multiple comparison post-test. Data are expressed as mean ± SEM, with ‘n’ denoting the number of experiments. A value of p < 0.05 was taken to signify statistical significance.
HK2 and human proximal tubule cells produce adherens junction proteins
High glucose increases secretion of TGF-β1
TGF-β1 alters adherens junction protein abundance in HK2-cells
TGF-β1 alters adherens junction protein abundance in hPTC
TGF-β1 decreases connexin-43 production
Loss of connexin-43 decreases gap junctional intercellular communication in proximal tubule cells
Loss of E-cadherin ligation replicates TGF-β1-induced changes in connexin-43 production
TGF-β1 reduces GJIC in proximal tubule cells
Decreased E-cadherin increases N-cadherin production
TGF-β1 reduces adherence between coupled cells of the proximal tubule
TGF-β1 is important in many tubulointerstitial diseases where disassembly of the adherens junction represents the initial overt change in epithelial organisation, well before any suggested cellular migration associated with EMT [34, 35]. The cadherin–catenin complex is crucial in epithelial cell-to-cell adhesion and facilitates cell communication via GJs. We confirm that TGF-β1 reduces membrane abundance of E-cadherin, and show for the first time that the cytokine decreases functional tethering between cells of the proximal tubule. While TGF-β1 fails to alter whole-cell production of β-catenin, accumulation in the nucleus is synonymous to that shown previously, and may signify release and subsequent re-localisation of β-catenin in response to reduced levels of cadherin . Our data provide compelling evidence that cell-to-cell adhesion and connexin-43 GJIC are dramatically reduced in the presence of TGF-β1, events likely to represent an early stage in glucose-induced renal damage in the proximal tubule.
To establish a direct link between E-cadherin and connexin-43 production, a neutralising antibody against E-cadherin ligation was used to mimic changes in cell adhesion seen in response to TGF-β1. Negation of E-cadherin ligation did not affect production of the protein, nor did it upregulate the transcriptional repressor snail, a process that accompanied the downregulation of E-cadherin production in response to TGF-β1. Importantly, however, loss of ligation reduced connexin-43 levels to a comparable degree to that observed in response to the cytokine, suggesting that it is a loss of tethering between cells, rather than a change in protein abundance as such that controls connexin production and GJIC. Retraction force–displacement curves confirmed that TGF-β1 reduced the maximum unbinding force required to begin separation of two cells by 20%, while halving the detachment energy required to completely separate them. The greater decrease in the detachment energy could be partly explained by the increase in cell rigidity following TGF-β1 treatment, as demonstrated by the re-arrangement of the cytoskeleton into peripheral stress fibres (ESM Fig. 1).
A switch in cadherin isoform from E-cadherin to neural (N)-cadherin is associated with EMT . TGF-β1 dramatically increased N-cadherin production in our model. However, this switch was unable to reverse morphological changes or the reduction in cell-to-cell adhesion in response to the cytokine. Surprisingly, neutralisation of E-cadherin ligation actually increased the maximum unbinding force. Counter-intuitively, this observation suggests that the modest increase in N-cadherin evoked by blocking E-cadherin ligation (Fig. 9e) can maintain tethering between coupled cells when the cadherin–catenin complex is intact, i.e. when only ligation is impaired. However, TGF-β1 dramatically reduced E-cadherin production and forced β-catenin to move away from the membrane. In this scenario, and in the absence of a catenin binding partner, upregulation of N-cadherin is redundant and the switch is unable to maintain tethering. These data suggest that it is decreased E-cadherin production and dissolution of the catenin–cadherin complex that drive the detachment of cells in EMT.
Type 2 EMT is commonly defined as the ability of adult epithelial cells to undergo de-differentiation, traverse the tubular basement membrane (TBM) into the interstitium and trans-differentiate into a myofibroblast phenotype that is capable of synthesising and increasing the deposition of extracellular matrix. While these activated myofibroblasts are thought of as key effector cells in the pathogenesis of renal fibrosis, it is clear that they originate from multiple lineages. Accumulating evidence suggests that local interstitial fibroblasts , pericytes , local mesenchymal stem cells  or the injured epithelium itself  may contribute to this pool, and there is considerable debate for and against a role of EMT in renal fibrosis . The established criteria supporting a role of EMT in fibrosis are based on the identification of morphological changes and altered levels of key epithelial/mesenchymal markers. Failure of fibroblasts to fully migrate and traverse the TBM is more commonly known as partial EMT, a phenomenon where cells produce epithelial and mesenchymal markers, yet lack migratory capacity. The argument against full phenotypic transformation has been fuelled by data from Humphreys et al., who suggest that not only is EMT unlikely to occur in vivo, but that vascular pericytes are the source of fibrosis-generating myofibroblasts .
As summarised in a recent review, TGF-β1 binds to a trans-membrane TGF-β receptor II (TBRII) and initiates several intracellular signalling cascades, including small mothers against decapentaplegic (SMAD) and mitogen-activated protein kinases such as extracellular regulated kinase, p38 and Jun kinase . SMADs are subdivided into three classes: (1) receptor regulated SMADs (SMAD1, -2, -3, -5 and −8); (2) the common SMADs (SMAD4); and (3) the inhibitory SMADs (SMAD6 and −7) . Following TBRII activation, receptor-regulated SMADs form oligomeric complexes with the common SMAD prior to translocation into the nucleus and regulation of gene transcription. The majority of TGF-β1-targeted genes that are regulated in EMT rely on SMAD3-dependent transcriptional regulation. Recent studies in cells from the proximal tubule have demonstrated that angiotensin II-induced tubular EMT was SMAD3-dependent . Similarly, (β)1-integrin gene expression, a potential therapeutic target for renal fibrosis, is also upregulated in unilateral obstruction and in chronic tubulointerstitial fibrosis via a SMAD3-dependent mechanism . However, despite the predominant involvement of SMAD3, a role for SMAD2 should not be discounted .
The hyperactive SMAD-signalling observed in certain types of renal disease reflects aberrant levels of both SMAD co-repressors and their subsequent regulators . The inhibitory SMADs (SMAD6 and SMAD7) inhibit receptor-regulated SMAD phosphorylation by blocking their access to TBRI, and/or by promoting degradation of the receptor complexes. SMAD7 represents a general antagonist of TGF-β1 and bone morphogenic protein signalling, with reports showing that induction of SMAD7 blocks tubular EMT and the development of fibrotic lesions . The role and regulation of SMAD signalling in regulation of GJ expression and GJIC in the proximal tubule remain to be confirmed; however, it is highly likely that these effects are SMAD-dependent and subject to regulation via endogenous inhibitors. The potential of exogenous agonist application to reverse these disrupted changes in cell-to-cell communication represents an area of therapeutic interest and forms the basis of our continuing research.
Reduced cell adhesion, cell coupling and cell-to-cell communication have profound effects on overall integrity and function of the proximal tubule, and altered GJIC and renal haemodynamics have recently been reported in a Zucker fatty rat model of type2 diabetes . In the current study, we concede that our in vitro data provide a minimalistic model of the early events in EMT and thus recommend caution in translating these novel findings to the in vivo situation, where the multifactorial molecular pathology of renal fibrosis may modify responses. However, despite this caveat, our current data provide a compelling foundation for the identification of future therapies aimed at maintaining or restoring renal function in diabetes, a supposition supported by recent data from mesangial cells, where glucose-induced hypertrophy was reversed by overproduction of connexin-43 .
We would like to thank JPK Instruments for their professional support in the use of single-cell force spectroscopy.
This work was supported wholly or in part by the generous support of Diabetes UK (11/0004215) and the Diabetes Research and Wellness Foundation. The work is also supported by an EFSD/Janssen grant and a University of Warwick Research Development Strategic Award. C.E. Hills was a Wellcome Trust ‘Value in People’ Research Fellow.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
All authors contributed to the conception, design and analysis of the data, and to the drafting of the manuscript, and have approved the final version for publication.