Gremlin (Grem1) is a member of the DAN family of secreted bone morphogenetic protein (BMP) antagonists. Bone morphogenetic protein-7 (BMP-7) mediates protective effects during renal fibrosis associated with diabetes and other renal diseases. The pathogenic mechanism of Grem1 during diabetic nephropathy (DN) has been suggested to be binding and inhibition of BMP-7. However, the precise interactions between Grem1, BMP-7 and other BMPs have not been accurately defined. In the present study, we show the affinity of Grem1 for BMP-7 is lower than that of BMP-2 and BMP-4, using a combination of surface plasmon resonance and cell culture techniques. Using kidney proximal tubule cells and HEK (human embryonic kidney)-293 cell Smad1/5/8 phosphorylation and BMP-dependent gene expression as readouts, Grem1 consistently demonstrated a higher affinity for BMP-2>BMP-4>BMP-7. Cell-associated Grem1 did not inhibit BMP-2- or BMP-4-mediated signalling, suggesting that Grem1–BMP-2 binding occurred in solution, preventing BMP receptor activation. These data suggest that Grem1 preferentially binds to BMP-2 and this may be the dominant complex in a disease situation where levels of Grem1 and BMPs are elevated.
Bone morphogenetic proteins (BMPs) are glycosylated extracellular matrix-associated members of the transforming growth factor β (TGF-β) superfamily . BMPs were originally identified for their ability to induce bone formation in vivo via osteoblast differentiation [2,3]. BMPs have a key function in morphogenesis, general organogenesis, cartilage and limb formation, as well as cell proliferation, differentiation and apoptosis [1,3–5]. BMPs have a critical role in kidney development as evidenced by data showing that bmp7-null mice are postnatal lethal because of various developmental abnormalities including kidney agenesis [6,7]. Although BMP-4 has been shown to contribute to renal fibrosis, conflicting evidence exists for the pro- or anti-fibrotic role of BMP-2 in this process [8–11]. BMP-7 is thought of as anti-fibrotic, and has been the focus of many groups who demonstrated its anti-fibrotic activity in models of diabetic nephropathy (DN) and other kidney fibrotic diseases [5,12–17]. To date, however, efforts to translate these data into BMP-7-centred treatment of fibrosis have been rather slow to develop [18,19].
Canonical BMPs signalling involves the Smad pathway, where BMP dimers bind to type I and type II BMP receptors leading to the formation of a hexameric complex, triggering receptor phosphorylation. This leads to phosphorylation of R-Smads (Smad1/5/8) and complex formation with co-Smad4 which translocates to the nucleus to regulate BMP target gene expression [1,2]. BMP target genes include inhibitor of differentiation (Id 1-3) genes and inhibitory Smad 6 [1,2]. BMP signalling is regulated on multiple levels: intracellularly by inhibitory Smads (Smad 6 and Smad 7), miRNAs, and methylation, and extracellularly by pseudoreceptors such as BAMBI and BMP antagonists including Gremlin (Grem1), Noggin and twisted gastrulation 1 (Twsg1) [1,2,20].
Grem1 is a 184-amino-acid (25 kDa) cysteine knot superfamily protein that exists in both secreted and cell associated forms . Grem1 exerts an inhibitory effect by directly binding to BMP dimers, preventing their interaction with BMP receptors, as well as blocking BMP secretion and increasing extracellular BMP endocytosis [1,22,23]. Typically, homozygous Grem1 deletion in mice leads to neonatal lethality due to development abnormalities including bilateral agenesis of the kidneys, lung defects and limb malformations [24,25]. However, recently it has been shown that some grem1-null mice survive when generated on a mixed genetic background (C57BL/6/FVB, .) grem1-null mice were smaller, with decreased weight and a shortened femoral length . The lack of kidney development in grem1-null mice was rescued by deletion of one allele of BMP-4 . Conversely, deletion of both alleles of BMP-7 rescued the ureteric branching defect but not the nephrogenesis defect , highlighting the critical nature of the balance between Grem1 levels and individual BMPs during kidney development. Grem1 is induced in response to high glucose in human mesangial cells, human proximal tubule epithelial cells and podocytes [29–33]. Grem1 is implicated in fibrotic diseases including DN [32,33], chronic allograft nephropathy , immune glomerulonephritis  and human idiopathic pulmonary fibrosis . Heterozygous grem1 deletion in mice protects against early DN-like changes in a streptozotocin (STZ) model of diabetes . Additionally, siRNA silencing of Grem1 has shown beneficial effects in an STZ mouse model through maintaining BMP-7 activity and reducing DN-like characteristics . The increased level of Grem1 in fibrotic kidney disease is thought to contribute to pathogenesis via inappropriate inhibition of BMP signalling.
The aim of the present study was to characterize Grem1 binding to BMP-7 and other BMPs. Our data suggest that Grem1 has a low affinity for BMP-7, and preferentially binds to BMP-2 using kidney epithelial cells as a model. Our results will help to refine our model of how Grem1 contributes to kidney fibrosis during diabetes and other diseases.
Detection of rhGrem1, BMP-2, BMP-4 and BMP-7 by Coomassie Violet staining and Western blotting
For Coomassie Violet staining, recombinant human (rh) Grem1, BMP-2, BMP-4, BMP-6 and BMP-7 (R&D Systems) samples were prepared at 0.5 and 1.0 μg via dilution with the appropriate vehicle [PBS (Grem1) or 4 mM HCl (BMPs)] and added to an equal volume of 2× Laemmli buffer without supplemental reducing agent. Samples were separated without boiling on SDS/10% (v/v) PAGE and stained with Coomassie Violet solution
For Western blotting, rhGrem1, BMP-2, BMP-4, BMP-6 and BMP-7 samples were prepared at 100 and 500 ng via dilution with the appropriate vehicle [PBS (Grem1) or 4 mM HCl (BMPs)] and added to an equal volume of 2× Laemmli buffer in the absence of reducing agent. Samples were separated without boiling on SDS/10% (v/v) PAGE and probed via Western blotting using antibodies reactive to Grem1, BMP-2, BMP-4 or BMP-7 (R&D Systems).
Surface plasmon resonance
rhBMP-2, rhBMP-4, rhBMP-6 and rhBMP-7 (R&D Systems) were immobilized on to individual flow channels of XanTec surface plasmon resonance (SPR) Sensorchip HLC 30 m (XanTec bioanalytics) via an amine coupling procedure. The surface of the XanTec SPR Sensorchip HLC 30 m was pre-equilibrated with borate elution buffer (30 μl; flow rate: 10 μl/min) (XanTec bioanalytics) before being activated with a 1:1 mixture of 0.4 M 1-ethyl-3-(3-dimethylaminopropyl)carbodi-imide (EDC) and 0.1 M N-hydroxysuccinimide (NHS) (50 μl; flow rate 10 μl/min) (GE Healthcare). rhBMPs were reconstituted at 100 μg/ml in 10 mM sodium acetate buffer, pH 5.5 (GE Healthcare) and diluted 1:10 with the same buffer. rhBMPs (20 μl) were injected at a flow rate of 10 μl/min until a response of approximately 500 response units (RU) was obtained. Remaining reactive sites on the flow channels were deactivated with 1 M ethanolamine, pH 8.5 (70 μl; flow rate: 10 μl/min).
Recombinant hGrem1 was reconstituted at 200 μg/ml (10 μM) in running buffer (2× PBS, 20 mM phosphate, 5.4 mM KCl, 274 mM NaCl and 0.05% Tween 20). Increasing concentrations of rhGrem1 were injected in the running buffer for 120 s (60 μl at a flow rate of 30 μl/min; dissociation time: 480 s. Regeneration injections were performed with 10 mM glycine/HCl, pH 1.5, after each rhGrem1 injection (20 μl, flow rate 30 μl/min). Sensorgrams were subjected to double referencing with subtraction of their respective control sensorgrams and blank injections. Experiments were performed on a Biacore 3000 instrument.
Size-exclusion chromatography (SEC) was performed at 20°C with a Shimadzu HPLC apparatus connected to Wyatt Dawn Heleos II light scattering and Optilab rEX instruments (Wyatt Technology). SEC was performed in 2× PBS using a 15 ml analytical KW-803 SEC column (Shodex) that was housed inside a thermostatically controlled column oven. Protein samples (5 μl) were injected at protein concentrations of ∼50 μM.
Protein samples were labelled for microscale thermophoresis (MST) measurements using an amine reactive coupling reaction (specifically a NT-647 NHS protein-labelling kit; NanoTemper Technologies). Tween 20 was added to the 2× PBS to a final concentration of 0.05% (v/v) to reduce protein adsorption to plastic ware and the side-walls of the disposable glass capillaries used for MST measurements. One hundred nanomolar labelled BMP-4 was titrated with unlabelled Grem1 using a series of 1:2 serial dilutions (up to a maximum concentration of 1 μM Grem). Each sample (4 μl) was then aspirated into hydrophilic capillaries, and MST measurements performed at 22°C using a Monolith NT.115 (NanoTemper Technologies). The light-emitting diode that excites labelled protein fluorescence was set to 30%, and a heating laser power of 40% was used to induce the temperature gradient that MST measurements employ. Laser on and off times were set to 30 and 5 s, respectively.
Human kidney proximal tubule epithelial cells (HK-2) were cultured as previously described . HEK (human embryonic kidney)-293 T cells were grown in Dulbecco's modified Eagle's medium (DMEM) medium containing 10% (v/v) FBS and 100 μg/ml primocin (Invivogen).
Cell treatments: Western blotting
For BMP stimulation of Smad1/5/8 phosphorylation, HK-2 and HEK-293 cells were plated on 60 mm plates. At 70% confluence, cells were washed with PBS and treated with HK-2 complete medium supplemented with vehicle (4 mM HCl) or increasing concentrations of rhBMP-2, rhBMP-4 (0.5–10 ng/ml) or rhBMP-7 (5–50 ng/ml) (R&D Systems) for 60 min. For rhGrem1 inhibition experiments, HK-2 cells were treated with complete medium supplemented with vehicle (4 mM HCl), 5 ng/ml rhBMP-2, 5 ng/ml rhBMP-4 or 20 ng/ml rhBMP-7 in the absence or presence of increasing concentrations of rhGrem1 (5–400 ng/ml) (R&D Systems) for 60 min. rhGrem1 and BMP proteins were co-incubated in complete medium at 37°C for 15 min prior to adding to cells. To assess the effect of cell associated Grem1, HK-2 cells were pre-treated with HK-2 complete medium supplemented with vehicle (PBS), 25 ng/ml Grem1 (for BMP-2) or 200 ng/ml Grem1 (for BMP-4) for 60 min. The medium was removed and replaced with fresh HK-2 complete medium supplemented with vehicle (4 mM HCl), 5 ng/ml BMP-2 or 5 ng/ml BMP-2 plus 25 ng/ml rhGrem1. A similar approach was employed for BMP-4, with 5 ng/ml BMP-4 and 200 ng/ml rhGrem1 utilized.
Construction of pcDNA3.1/myc-hisA-hGrem1 plasmid
A DNA fragment containing the entire coding sequence for Grem1 was amplified by real-time PCR (RT-PCR) on cDNA from HK-2 cells using forward primer 5′-GACAGTGA-ATTCATGAGCCGCACAGCCTACACG-3′ and reverse primer 3′-GGATTTTCTAGAATCCAAATCGATGGATATGCA-5′ and RedTaq® DNA polymerase at an annealing temperature of 55°C. The fragment was restriction digested using EcoRI and XbaI (Fermentas) to create sticky ends. The fragment was ligated using T4 DNA ligase (Life Technologies) into pcDNA™ 3.1/myc-hisA (Invitrogen) that was cut with EcoRI and XbaI. The ligated product was used to transform competent XL-1 blue Escherichia coli (Agilent Technologies) which were then plated on to LB-ampicillin plates. Positive colonies were identified by PCR, and pcDNA3.1/myc-hisA-hGrem1 plasmid was extracted from cells using Pure Yield™ Plasmid Maxiprep system (Promega). All final plasmid DNA was validated by Sanger sequencing.
Transfection and immunocytochemistry of HEK-293 cells
HEK-293 cells grown on glass coverslips were transfected with empty plasmid (pcDNA3.1) or plasmid containing full-length human Grem1 cDNA (pcDNA3.1-hGrem1) using Lipofectamine™ 2000 (Invitrogen). At 48 h post-transfection, cells were incubated in serum free DMEM for 4 h and treated with vehicle (4 mM HCl) or 5 ng/ml BMP-2 for 60 min. Cells were fixed in 4% PFA, permeabilized with 0.1% Triton-X, blocked in 1% BSA and stained with α-myc (9E10) and pSMAD1/5 (Cell Signalling) primary antibodies. Anti-mouse TRITC and anti-rabbit FITC secondary antibodies were used for detection, together with DAPI (blue) to visualize the nuclei.
Conditioned medium from empty vector and pcDNA3.1-hGrem1 transfected cells of various dilutions was added to fresh, non-transfected HEK-293 cells in the presence of 5 ng/ml rhBMP-2, rhBMP-4 or rhBMP-7 for 60 min. Cells were lysed in supplemented RIPA buffer and Western blotting for p-Smad1/5/8 and β-actin was carried out exactly as described [38,39]. Densitometry analysis was performed using ImageJ software (http://rsbweb.nih.gov/ij/) and pSmad1/5/8 band intensities were expressed as a ratio of β-actin loading control intensity for each sample.
Real-time quantitative PCR
For BMP stimulation and Grem1 inhibition of BMP gene responses, HK-2 cells were plated on 60 mm plates. At 70% confluence, cells were washed with PBS and treated with HK-2 complete medium supplemented with vehicle (4 mM HCl), 10 ng/ml rhBMP-2, 10 ng/ml rhBMP-4 or 20 ng/ml rhBMP-7 in the absence or presence of increasing concentrations of rhGrem1 (25–400 ng/ml) for 2 h. Recombinant proteins were co-incubated in complete medium at 37°C for 15 min prior to adding to the cells. RNA was extracted using an RNeasy RNA extraction kit according to the manufacturers protocol (Qiagen). Total RNA (1 μg) was reverse transcribed and Taqman PCR was performed using specific Taqman probes from Roche Applied Science for Id1 (Assay Id: 104631), Smad6 (Assay Id: 104698) or Grem1 (Assay Id: 105548). Levels of activin receptor-like kinase (ALK) receptor expression were measured in HK-2 and HEK-293 cells using Roche Taqman probes for ACVR1 (ALK 2) (Assay Id: 104525), BMPR1 A (ALK 3) (Assay Id: 104581) and BMPR1B (ALK 6) (Assay Id: 104584). All analysis was carried out using the ΔΔCT method and normalized to an average of 18 S (Assay Id: 104092) and β-actin (Assay Id: 101125) levels. RT-PCR was carried out on a Roche LightCycler 480.
All experiments were carried out a minimum of three times in duplicate, and statistically significant differences were detected using Student's unpaired t test or one-way ANOVA with Bonferroni's multiple comparison test using GraphPad Prism. P values <0.05 were considered significant.
Analysis of commercially available recombinant Grem1 and BMPs
We first assessed the fidelity of commercial sources of recombinant Grem1 and BMPs, which are widely used in the field. Thus rhGrem1 and BMP-2, BMP-4, BMP-6 and BMP-7 samples (all from R&D Systems) were prepared and sub-jected to SDS/10% (v/v) PAGE in the absence of reducing agents and boiling (i.e. non-reducing SDS/PAGE). Protein bands were visualized by both Coomassie Violet staining (Figure 1a) or Western blotting (Figures 1b–1f). Table 1 summarizes the predicted molecular masses of the recombinant proteins versus those detected by SDS/PAGE under reducing and non-reducing conditions.
Detection of rhGrem1 and rhBMPs by Coomassie Violet staining and Western blotting
|Predicted size (kDa) (R&D Systems)||SDS/PAGE (kDa) (reducing)||Observed SDS/PAGE (kDa) (non-reducing)|
|Predicted size (kDa) (R&D Systems)||SDS/PAGE (kDa) (reducing)||Observed SDS/PAGE (kDa) (non-reducing)|
Two predominant bands for rhGrem1 were detected. One had a gel mobility consistent with a mass of 25–27 kDa, close to the expected mass for a Grem1 monomer (∼21 kDa), whereas the other had a mass of ∼14 kDa. Since both species reacted strongly with antibodies specific to rhGrem1, we attributed the lower mass species to a Grem1 breakdown product. This lower mass species could also be a glycosylated Grem1 variant with an idiosynchratic gel mobility. Notably, full length rhGrem1 was present at significantly higher abundance than the minor breakdown product.
We next used SEC in conjunction with multi-angle light scattering (MALS) to further analyse rhGrem1. SEC–MALS is a powerful tool for directly measuring the mass, purity and concentration of biomolecules in solution. Using this approach, we observed just two species eluting from the high resolution SEC column, with relative elution volumes of ∼11.9 and ∼12.3 ml, with relative abundances of ∼10% and ∼90%, respectively, based on the integrated UV peak areas (Figure 2c). The elution profile is consistent with the results obtained using SDS/PAGE and Western blotting, with the most abundant species (native monomeric rGrem1) eluting later from the SEC column than the breakdown product/glycosylated rhGrem1 (presumably due to it being either unfolded or non-globular). Unfortunately, the low molecular mass of rhGrem1 combined with the low protein concentrations available meant that rGrem1 did not scatter light sufficient to determine its molecular mass using MALS. Non-reducing SDS/PAGE and Western blotting showed that monomeric forms of BMP-2 and BMP-7 were the most abundant in commercial, recombinant sources (Figures 1a, 1c and 1e). In contrast, higher order oligomers were observed for BMP-4 and BMP-6, although these were also significantly less abundant than the monomeric forms (Figures 1d and 1f). Most BMPs examined, however, yielded only a single species when examined using silver stained, reducing SDS/PAGE (Supplementary Figure S1). This supports the view that the complexity observed on non-reducing SDS/PAGE is likely to be associated with oligomeric forms of BMPs refractory to denaturation (in non-reducing SDS/PAGE), as opposed to higher-molecular-mass contaminants or unprocessed BMP species. The exception to this was BMP-6, which did not silver stain in our experiments, despite staining with Coomassie Blue (Figure 1a and Supplementary Figure S1 a).
rhGrem1 binds to rhBMPs with different affinities
SEC–MALS analysis showed that rhBMP-2 yielded only a single, symmetrical elution peak in PBS (Figure 2d). These results were consistent with rhBMP-2 being homogeneous, in agreement with the results obtained using SDS/PAGE and Western blotting. As per rhGrem1, the protein concentrations used were too low to obtain a molar mass for BMP-2 using SEC–MALS. BMP-4, BMP-6 and BMP-7 reproducibly failed to elute from SEC columns, precluding any objective assessment of their solution oligomeric state (results not shown).
Recombinant Grem1 binds BMPs with different affinities
Real-time SPR was used to characterize the binding of rhBMPs to rhGrem1. Initially, rhGrem1 was attached as ‘bait’ to Biacore XanTec sensorchips using amine-reactive coupling, and a series of rhBMPs were then used as analytes to probe for binding events. However, little or no BMP binding was detected, suggesting that the coupling of rhGrem1 to the Biacore surface impinged on the interface required for BMP binding (results not shown). Thus the reverse approach was then implemented and BMP-2, BMP-4, BMP-6 and BMP-7 were individually coupled to individual flow cells on the Biacore sensor chip surface. This approach had merit for two reasons: (i) to keep rGrem1 active (which is not the case when it was coupled to an SPR chip, see above); (ii) all BMPs will see the same rhGrem1 preparation, including the 10% low abundance Grem1 species observed by SEC–MALS and Western blot discussed above (Figure 1b).
Increasing concentrations of rhGrem1 (ranging from 10 nM to 3 μM) were injected on to the SPR chip and the relative binding in RU was measured for each immobilized BMP. This approach gave high quality data, with excellent signal to noise ratios and yielded a set of concentration dependent binding kinetics (BMP-7 shown as example, Figure 2). At the lower Grem1 concentrations examined (10-300 nM), there was clear evidence for association kinetics that typically report on ligand binding, and from which on-rates (kon) are determined. Conversely, no convincing dissociation phases were observed after rhGrem1 injections (10–300 nM) were ceased. This was problematic as it is these dissociation phases that are typically used to determine off-rates, and ultimately Kd values. This phenomenon could be due to very tight binding between BMPs and rhGrem1, or irreversible, non-specific binding of rhGrem1 to the SPR chip surface (i.e. in the absence of immobilized BMPs). However, no irreversible binding was observed in ‘empty’ reference channels (i.e. no immobilized BMPs were present). Clear dissociation phases were evident, however, for each BMP when higher concentrations of rhGrem1 (≥1 μM) were injected on to the SPR chip. We believe that this additional phase, which was observed only at higher concentrations of rhGrem1, most likely reflects the reversible binding of rhGrem1 to BMPs.
The SPR sensorgrams were analysed in two different ways: (i) fitting the association and dissociation kinetics to a 1:1 Langmuir analysis, which assumes a single binding site (Figure 2b and Table 2), (ii) assuming a bivalent analyte, where each rhGrem1 can bind multiple BMPs (which seems possible, as Grem1 has been reported to be a functional dimer). We specifically fitted only the 3 μM rhGrem1 injections (which exhibited reversible association-dissociation kinetics), although comparable Kd values were obtained for the 1 μM rhGrem1 datasets (results not shown). For each BMP, the dissociation phases fitted well to the 1:1 Langmuir function (Table 2). The association phases fitted less well, although well within what is normally accepted for publication quality SPR data. Using this approach, the rank order of interaction affinity, from tightest binding to the weakest, was BMP-4=BMP-2 > BMP-6 > BMP-7 (Table 2). Fitting the data to a bivalent analyte function gave excellent fits, and yielded a rank order of binding affinity, from tightest binding to weakest, of BMP-4 > BMP-2 > BMP-6 > BMP-7 for the tighter of the two modelled binding events (although their absolute magnitude varied from those calculated using the Langmuir analysis, Supplementary Table S1). In our opinion, the weaker of the two binding sites modelled using the bivalent analyte analysis did not yield physically sensible parameters, and the high quality fits obtained was merely a consequence of employing a fitting procedure that had more variables than the Langmuir analyses.
|Protein||kon (M−1·s−1)||koff (s−1)||Kd (M)|
|Protein||kon (M−1·s−1)||koff (s−1)||Kd (M)|
We next employed MST to try and obtain objective independent validation of the binding affinities for each BMP for rhGrem1. Since our SPR experiments showed that coupling rhGrem1 to SPR chips inactivated it, we elected to label the BMPs with a red dye (using amine reactive coupling, as per our SPR experiments). Using this approach, we found that we could successfully label BMP-4 with the red dye. By contrast, it was not possible to obtain labelled BMP-2, BMP-6 and BMP-7, as these proteins did not elute from the small SEC columns used to separate labelled proteins from unincorporated dye. This experiment again demonstrate the challenges of studying BMP interactions in vitro, as similar issues occurred in our SEC–MALS studies of BMPs (discussed above).
High-quality binding curves were obtained when labelled BMP-4 was titrated with rhGrem1 (Figure 2e). The apparent Kd of this curve (≤ 50 nM) agreed fairly well with the apparent Kd determined using SPR and the Langmuir analysis (Kd ∼28 nM). By contrast, the Kd determined for BMP-4-rhGrem1 disagreed with the value obtained using a bivalent analyte analysis of the SPR data (Kd ∼ 0.4 nM). Based on this benchmarking of the SPR data, and the problems seen with irreversible binding of BMPs to silica and polymer based SEC columns, we believe that our SPR data were best described by the 1:1 Langmuir analysis, and ignoring the apparently irreversible binding seen at low rGrem1 concentrations (10–300 nM). Irrespective of these issues, in each case (i.e. Langmuir analysis or bivalent analyte), the weakest binder to rhGrem1 was BMP-7, with BMP-2 and BMP-4 having considerably higher affinity for rhGrem1.
Differential BMP-mediated pSmad1/5/8 signalling and Grem1 inhibition in kidney epithelial cells
BMP binding to heterotetrameric type I/II receptor complexes leads to Smad1/5/8 phosphorylation, dimerization with co-Smad4, nuclear translocation and transcriptional activation . To validate the SPR and MST protein interaction data in a cell culture system, human kidney proximal tubule epithelial cells (HK-2) were incubated with increasing concentrations of rhBMPs [BMP-2; 0.5–10 ng/ml, (39–790 pM), BMP-4; 0.5–10 ng/ml, (39–780 pM) or BMP-7; 5–50 ng/ml (325 pM to 3.25 nM)]. Robust Smad1/5/8 phosphorylation was seen in response to BMP-2 at 0.5 ng/ml (Figure 3a) and BMP-4 at 1–2 ng/ml (Figure 3b). In contrast, higher concentrations of BMP-7 (>10 ng/ml) were required to trigger a robust pSmad1/5/8 response in these cells (Figure 3c). BMP-2, BMP-4 and BMP-7 have been reported to utilize similar BMP receptor complexes, with signalling transmitted via BMPRII/Alk3/6 heterotetramers for all three BMPs . One potential difference for the reduced response to BMP-7 may be the ability of BMP-7 to specifically engage with Alk2 type I receptors [41,42]. ACVR1/ALK2 and BMPRIA/ALK3 were the predominant isoforms detected in HK-2 (and HEK-293) cells, with levels of ALK6 lowest in both cases (Figure 3d). Thus the absence of ALK2 receptors cannot explain the lower efficacy of BMP-7 in HK-2 cells.
Human BMPs exhibit different efficacies of Smad 1/5/8 phosphorylation in HK-2 cells
Co-incubation of BMP-2 (5 ng/ml, 395 pM) with 25 ng/ml Grem1 (1.33 nM) completely inhibited BMP-2-mediated Smad1/5/8 phosphorylation (Figures 4a and 4b). In contrast, inhibition of BMP-4-mediated Smad1/5/8 phosphorylation was not detected until 100 ng/ml Grem1 (5.3 nM) was used (Figures 4c and 4d). BMP-7 (20 ng/ml, 1.3 nM) triggered a strong increase in Smad1/5/8 phosphorylation, but this was not inhibited by co-incubation with Grem1 up to 400 ng/ml (21.2 nM) (Figures 4e and 4f). These data demonstrate that the potency of BMP-2 and BMP-4 canonical signalling exceeds that of BMP-7 in HK-2 cells, and that rhGrem1 inhibits BMP-2>BMP-4>BMP-7 when endogenous BMP type I/II receptor activation is measured using pSmad1/5/8 as a readout. BMP signalling is transmitted via pSmad1/5/8 dimer formation with Smad4 which translocates to the nucleus and is recruited to transcriptional complexes on a defined set of BMP-responsive promoters of genes such as Id1-3 and Smad6 . We wished to confirm that increased pSmad1/5/8 levels in response to BMP treatment corresponded to bona fide changes in BMP-mediated gene expression. Incubation of HK-2 cells with BMP-2 and BMP-4 (10 ng/ml), and BMP-7 (20 ng/ml) induced a significant increase in Id1 and Smad6 mRNA (Figure 5), consistent with increased pSmad1/5/8 phosphorylation (Figure 4a). For BMP-2, this transcriptional response was inhibited by rhGrem1 in a concentration-dependent manner, with ∼ 50% inhibition seen with 25 ng/ml Grem1 (Figures 5a and 5d). Higher amounts of Grem1 were required to inhibit BMP-4-mediated gene transcription which was only seen with 200 ng/ml Grem1 (Figures 5b and 5e). Weak inhibition of BMP-7-stimulated Smad6, but not Id1, was detected up to 400 ng/ml Grem1, supporting the hypothesis that the affinity of Grem1 for BMP-7 homodimers is lower than for BMP-2 (Figures 5c and 5f). These data highlight that Grem1 binding to BMPs prevents receptor-mediated activation of BMP gene targets, and that the rank order of affinity of BMP-2>BMP-4>BMP-7 is consistent with data obtained when endogenous BMP gene targets are quantified.
Differential Grem1-mediated inhibition of BMP-stimulated Smad 1/5/8 phosphorylation
Grem1 inhibits BMP-mediated Id1 and Smad6 gene expression with differing affinities
Overexpression of human Grem1 inhibits BMP signalling in HEK-293 cells
Given the critical role of Grem1 post-translational processing and glycosylation in its function , the ability of Grem1 expression to inhibit BMP action when expressed in mammalian cells was then assessed. After optimization, a protocol was developed whereby >80% transfection efficiency was routinely obtained in HEK-293 cells using pcDNA3.1 containing the human Grem1 cDNA (Figure 6). Gremlin expression was detected both intracellularly (Figure 6a) and also in the cell-cell junctions and extracellular space (Supplementary Figure S3). Treatment of HEK-293 cells with BMP-2 increased Smad1/5 phosphorylation (Figure 6a, top right panels) and transfection with myc-hGrem1 abolished this phosphorylation (Figure 6a, bottom right panels). A similar profile was observed when transfected cells were treated BMP-4 (results not shown). These data suggest that the transfection of cells with full-length human Grem1 leads to the production and secretion of hGrem1 from HEK-293 cells, where it binds to the recombinant BMPs preventing receptor activation. Consistently, Grem1 protein was detected when the conditioned medium from Grem1-transfected cells was analysed (Figure 6b).
Overexpression of human Grem1 inhibits BMP stimulation of Smad1/5/8 phosphorylation in HEK-293 cells
To further interrogate the relative affinity of secreted, processed rhGrem1 for BMP-2, BMP-4 and BMP-7, conditioned medium from hGrem1-transfected cells was added to non-transfected HEK-293 cells in the presence of rhBMP-2, rhBMP-4 or rhBMP-7 (Figure 7). ‘Neat’ conditioned medium from Grem1-transfected HEK-293 cells completely inhibited pSmad1/5/8 phosphorylation in response to BMP-2 and BMP-4, but not BMP-7 (Figure 7, ‘100%’). Serial dilution of this conditioned medium demonstrated that recovery of BMP-2-mediated Smad1/5/8 phosphorylation only occurred after four sequential dilutions (5% medium, Figures 7a and 7b). In contrast, Smad1/5/8 phosphorylation was detected in BMP-4 stimulated cells when the conditioned medium was diluted 1:4 (25%, Figures 7a and 7c). Consistent with the fact that 100% conditioned medium modestly inhibited BMP-7-mediated Smad1/5/8 phosphorylation, no further changes were seen with sequential dilution (Figures 7a and 7d). Similar results were obtained when HK-2 cells were used in this assay (results not shown). These data demonstrate that HEK-293-produced rhGrem1 can inhibit BMP action in a rank order of efficacy similar to that seen with commercially available rhGrem1 (R&D Systems), and support the rank order of BMP-2>BMP-4>BMP-7 in terms of Grem1 binding.
Conditioned medium containing human Grem1 inhibits BMP-mediated Smad1/5/8 phosphorylation
Cell-associated Grem1 does not inhibit BMP-mediated signalling responses
As previously reported, Grem1 is a secreted protein but also exists in cell-associated forms [1,21]. Grem1 has been shown to bind to heparin and heparin sulfate proteoglycans, thus modulating its interaction with the vascular endothelial growth factor receptor 2 (VEGFR2) and subsequent pro-angiogenic activity . Other evidence demonstrated that Grem1 acts intracellularly to regulate BMP-4 activation and secretion . In order to assess the ability of ‘cell associated’ versus ‘soluble’ Grem1 to inhibit BMP action, HK-2 cells were pre-treated with medium supplemented with vehicle or 25 ng/ml rhGrem1 for 60 min (Figures 8b and 8d, Grem1 pre-incubation, removal and BMP-2 addition, right panels). The Grem1 containing medium was removed and cells were then treated with medium containing 4 mM HCl (BMP vehicle) or 5 ng/ml BMP-2 (Figures 8b and 8d, Grem1 pre-incubation, removal and BMP-2 addition, right panels) for a further 60 min. Similar to previous data, treatment of HK-2 cells with 5 ng/ml BMP-2 triggered a robust increase in Smad1/5/8 phosphorylation that was completely inhibited by co-incubation with 25 ng/ml Grem1 (Figures 8a and 8c, BMP-2 plus Grem1, left panels). In contrast, pre-treatment of HK-2 cells with 25 ng/ml rhGrem1 followed by removal of the Grem1-containing medium did not inhibit BMP-2-mediated Smad1/5/8 phosphorylation (Figures 8b and 8d, Grem1 pre-incubation, removal and BMP-2 addition, right panels). A similar profile of inhibition was obtained when BMP-4 was used (Supplementary Figure S4). These data suggest that ‘cell associated’ or surface bound rhGrem1 is unable to inhibit BMP-mediated endogenous BMP type I/II receptor activation and Smad1/5/8 phosphorylation, compared with extracellular rhGrem1 in solution.
Cell associated Grem1 does not inhibit BMP-2 mediated Smad1/5/8 phosphorylation
In the present study, we identified that BMP proteins have differing affinities for the antagonist Grem1. Calculation of Kd values from SPR experiments at all concentrations of rhGrem1 examined was somewhat confounded by what appeared to be irreversible, or ultra-tight, binding events at low hGrem1 concentrations (10–300 nM). Nonetheless, when we fitted the reversible rhGrem1-binding phases (i.e. 1 or 3 μM rhGrem1) to either one-site (Figure 2b) or bivalent analyte binding model (results not shown), very similar rank orders of binding affinity were observed in vitro: BMP-4 ≥ BMP-2 > BMP-6 > BMP-7 (Table 2). Although the apparent ‘stickiness’ of BMPs to SEC columns precluded robust SEC–MALS experiments and labelling of most BMPs for MST experiments, the one interaction that we could directly measure in vitro (BMP-4/rhGrem1) to independently benchmark the SPR data, had an apparent Kd value that agreed well with the value using the Langmuir analysis (Figure 2e). Thus we believe that the observed SPR data are a synthesis of irreversible binding (perhaps caused by the stickiness of BMPs) and reversible binding.
The nature and complexity of these interactions, combined with high cost of the recombinant proteins used here, and their challenging properties, conspired to make in vitro biophysical experiments very challenging. Further work will be needed to disentangle fully the precise interaction stoichiometry and ‘true affinities’. Nonetheless, we feel that the good agreement between the rank order of BMP–rhGrem1 interaction affinities reported here, and their associated biological activities and potencies in cellular experiments, represent an important step forward in higher resolution mechanistic and functional studies. Protein related to Dan and Cerberus (PRDC or Grem2) also inhibits BMP-2 action in cell culture and whole Xenopus embryos via the formation of highly stable non-covalent dimers . Kd values estimated by Langmuir equation for Grem1 binding to BMP-2 and BMP-4 were in the same range (low nM) as those calculated for PRDC, with the same caveats in place regarding Grem1–BMP affinities we present here. Other authors have shown that reduction of PRDC (Grem2) using DTT changed its mobility on SDS/PAGE, but not its BMP inhibitory activity .
Using HK-2 kidney epithelial cells, the ability of Grem1 to inhibit BMP-activated Smad1/5/8 phosphorylation and downstream target gene transcription was defined (Figures 4 and 5). In contrast with the SPR assays where only recombinant Grem1 and BMPs were present, this system measures Grem1–BMP competitive binding in the presence of the type I/II BMP receptors in a cell-based assay. The ability of BMP-7 to trigger Smad signalling in these cells was much lower than that of BMP-2 (Figures 3a and 3b compared with Figure 3c). Complete inhibition of BMP-2 activity was seen with a 3.35 molar excess of Grem1, whereas a 13.6-fold molar excess of Grem1 was required to inhibit BMP-4 (Figures 4 and 5). In contrast, no significant inhibition of BMP-7 was seen with a 16.3-fold molar excess of Grem1 (Figures 4 and 5). Given that BMP-7 can strongly activate pSmad1/5/8 in outgrowth endothelial cells in our laboratory, we do not think that the fidelity of the recombinant BMP-7 is an issue here. Importantly, ACVR1/ALK2 and BMPRIA/ALK3 receptor mRNA was detected in HK-2 cells, suggesting that the reduced response to BMP-7 was not due to an absence of cognate receptors.
Transfection of HEK-293 cells with myc-hGrem1 decreased BMP-2-mediated pSmad1/5 phosphorylation (Figure 6a). Similar profiles of inhibition were observed when conditioned medium from these transfected HEK-293 cells was used as a source of Grem1 (Figure 7). These data suggest that the relative affinity profile seen in our experiments is not dependent on the source of Grem1 used. In addition, the consistent results with two independent sources of rhGrem1 suggest that that myc affinity tag present in rhGrem1 produced from HEK-293 cells does not greatly alter Grem1 function or availability in the conditioned medium. How does Grem1 inhibit BMP action? Our data suggest that binding of Grem1 to BMPs in solution is required for its inhibitory activity, as cell-bound Grem1 was not sufficient for inhibition of BMP action (Figure 8). These data also suggest that rhGrem1 is not binding to and preventing access to the BMP receptors. Previous groups have identified that Grem1 can interact with several cell surface receptors including VEGFR2 and heparin sulfate proteoglycans [43,45]. Our data suggest that these cell surface receptor interactions do not ‘capture’ sufficient Grem1 to allow significant inhibition of BMP signalling in HK-2 cells. A different model may exist in endothelial and other vascular cell types that express higher levels of VEGFR2. Apart from direct antagonism of BMP-receptor binding, Grem1 can bind BMP-4 inside C2C12 cells and inhibit BMP-4 maturation and secretion . Recent data have also demonstrated that Grem1 can accelerate the endocytosis of BMP-2 into HeLa cells . Given the relatively short duration of Grem1-BMP treatment of HK-2 cells in our experiments (60 min), we suggest that the main mechanism of Grem1-mediated inhibition of BMP signalling in our experiments is via direct antagonism of BMP in solution, preventing binding to its target receptors at the plasma membrane.
What implications do these data have for those working on the regulation of BMP signalling by Grem1 and other antagonists? We and others have shown that Grem1 levels increase during DN whereas levels of BMP-7 decrease [33,46]. Strategies that stimulate BMP-7 signalling have been shown to attenuate renal fibrosis [5,12–17], and reductions in Grem1 levels have a similar beneficial effect in mouse models of DN [32,37]. In vivo reduction of Grem1 via siRNA targeting recovered BMP-7 levels, suggesting that elevated Grem1 may have a role in repressing BMP-7 in DN . However, our data suggest that this is unlikely to occur via a direct interaction between Grem1 and BMP-7, given their low binding affinity. Small molecules such as tilerone (which induces BMP-7) and Thr123 (which activates the Alk3 BMP receptor) have been shown to reduce kidney fibrosis in vivo [47,48]. Together with a recent report showing that an anti-Grem1 antibody can reduce fibrosis in a model of pulmonary artery hypertension (PAH) , these data illustrate that manipulation of the Grem1–BMP signalling axis may still be a useful therapeutic avenue for the treatment of diabetic kidney disease. A recent paper identified that Twsg1 is the most abundant BMP antagonist expressed in cultured kidney podocytes, and could inhibit both BMP-4 and BMP-7 in these cells . This report highlights the likely involvement of multiple BMP antagonists in the control of BMP action during kidney injury.
In summary, we have demonstrated a differential level of activity of BMPs in HK-2 epithelial cells, and a BMP-2>BMP-4>BMP-7 rank order of Grem1 antagonist binding to these proteins. Our data provide a mechanistic insight into the possible molecular interactions in the diseased kidney in diabetes and other chronic nephropathies.
Rachel H. Church carried out experimental work and wrote sections of the manuscript. Arjun Krishnamkumar and Maria Strömstedt carried out experimental work. Annika Urbanek carried out experimental work for revised manuscript. Stefan Geschwinder, Julie Meneely, Alessandro Bianchi, Barbro Basta and Sean Monaghan carried out experimental work. Christopher Elliot was involved in the planning of SPR experiments. Neil Ferguson was involved in the planning of SEC–MALS and MST experiments and wrote sections of revised manuscript. Finian Martin did project design and planning. Derek Brazil did project funding and design, wrote, edited and revised the manuscript.
We thank Dr John Crean and Dr Gerard Cagney (UCD Conway Institute, University College Dublin) and Dr Christopher M. Johnson (MRC Laboratory for Molecular Biology, Cambridge) for helpful discussions.
R. H. C. is supported by a Biotechnology and Biological Sciences Research Council (BBSRC) CASE Ph. D. studentship, in partnership with AstraZeneca UK. Work in the laboratory of D.P.B. is supported by DEL Northern Ireland, BBSRC, Northern Ireland Kidney Research Fund and Diabetes UK.
activin receptor-like kinase
bone morphogenetic protein
Dulbecco's modified Eagle's medium
human embryonic kidney
multi-angle light scattering
protein related to Dan and Cerberus/Gremlin2
surface plasmon resonance
vascular endothelial growth factor receptor 2