Targeting of bone morphogenetic protein complexes to heparin/heparan sulfate glycosaminoglycans in bioactive conformation

Abstract

Bone morphogenetic proteins (BMP) are powerful regulators of cellular processes such as proliferation, differentiation, and apoptosis. However, the specific molecular requirements controlling the bioavailability of BMPs in the extracellular matrix (ECM) are not yet fully understood. Our previous work showed that BMPs are targeted to the ECM as growth factor‐prodomain (GF‐PD) complexes (CPLXs) via specific interactions of their PDs. We showed that BMP‐7 PD binding to the extracellular microfibril component fibrillin‐1 renders the CPLXs from an open, bioactive V‐shape into a closed, latent ring shape. Here, we show that specific PD interactions with heparin/heparan sulfate glycosaminoglycans (GAGs) allow to target and spatially concentrate BMP‐7 and BMP‐9 CPLXs in bioactive V‐shape conformation. However, targeting to GAGs may be BMP specific, since BMP‐10 GF and CPLX do not interact with heparin. Bioactivity assays on solid phase in combination with interaction studies showed that the BMP‐7 PD protects the BMP‐7 GF from inactivation by heparin. By using transmission electron microscopy, molecular docking, and site‐directed mutagenesis, we determined the BMP‐7 PD‐binding site for heparin. Further, fine‐mapping of the fibrillin‐1‐binding site within the BMP‐7 PD and molecular modeling showed that both binding sites are mutually exclusive in the open V‐ versus closed ring‐shape conformation. Together, our data suggest that targeting exquisite BMP PD‐binding sites by extracellular protein and GAG scaffolds integrates BMP GF bioavailability in a contextual manner in development, postnatal life, and connective tissue disease.

Abbreviations

BMP

bone morphogenetic protein

BM

basement membrane

CD

circular dichroism spectroscopy

CPLX

complex

EBNA

Epstein–Barr virus nuclear antigen

ELISA

enzyme linked immunosorbent assay

ECM

extracellular matrix

FGF

fibroblast growth factor

FMF

fibrillin microfibrils

GAG

glycosaminoglycans

GDF

growth and differentiation factor

GF

growth factor

HEK

human embryonic kidney

HS

heparan sulfate

HSPG

heparan sulfate proteoglycan

MAB

monoclonal antibody

MMP

matrix metalloproteinase

NTA

Nitrilotriacetic acid

PD

prodomain

PDB

protein data bank

RT

room temperature

RU

resonance unit

SPR

surface plasmon resonance

TEM

transmission electron microscopy

TGF

transforming growth factor

TMB

tetramethylbenzidine

VEGF

vascular endothelial growth factor

1. INTRODUCTION

Bone morphogenetic proteins (BMPs) are pluripotent growth factors of the TGF‐β superfamily that drive a multitude of biological programs during embryogenesis and tissue regeneration. ¹ , ² Due to their high potency as cellular morphogens tightly regulated, extracellular control mechanisms of their bioavailability are mandatory. BMPs are stored and sequestered by the extracellular matrix (ECM) microenvironment of tissues from which they are specifically activated upon cellular utilization. ³ , ⁴ , ⁵ This view is supported by the original discovery of BMPs as cytokines extractable from bone matrix under denaturing conditions. ⁶ These extracts contain the liberated, bioactive growth factor (GF) moiety as an injection into tissues induces strong ectopic bone formation. ⁷ , ⁸

Currently, it is not fully understood how the ECM regulates the bioavailability of GFs of the TGF‐β superfamily. However, this question is not only important to better understand how embryonic patterning occurs but also how remodeling and regeneration processes are regulated in tissues. TGF‐β and BMP growth factors are known to play crucial roles in growth and differentiation processes, but the specific molecular requirements to fine tune their bioactivity within the respective cellular microenvironment remain elusive. So far the impact of the ECM on BMP GF activity has been mostly studied in absence of their prodomains (PDs) that are proteolytically cleaved from the GFs and noncovalently associate with them to form PD–GF complexes (CPLXs) before secretion. ⁹ Thereby, several interactions of TGF‐β GFs to ECM components such as procollagen II, ¹⁰ , ¹¹ oligomeric matrix protein (COMP), ¹² and glycosaminoglycan (GAG) carrying heparan sulfate proteoglycans (HSPG) such as decorin ¹³ , ¹⁴ have been investigated. However, depending on the investigated physiological context GF binding to ECM proteins seems to either promote or limit their bioactivity. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹

Until now, it is not clear how BMP targeting to ECM proteins versus GAGs integrates their bioavailability. GAGs are supramolecular oligomers consisting of repeating units of sulfated disaccharides that can strongly associate with several other GFs such as VEGF or FGF and were suggested to act as general sinks or reservoirs for GFs required for their spatial concentration. ²⁰ A more active role for GAGs has been described by their function as co‐receptors thereby assisting in the presentation of GFs to their cell surface signaling receptors. ²¹ , ²² Generally, BMP GFs are predicted to interact with their basic, electropositive surface binding sites with long/extended, highly sulfated/negatively charged disaccharides of GAGs. ²³ The GAG heparin is chemically similar to heparan sulfate (HS) but contains more fully sulfated sequences resulting often times in increased binding affinity to heparin‐/HS‐binding proteins. ²⁴ Several studies have shown that BMP GF signaling and gradient formation is controlled by HS during developmental processes. ²⁵ , ²⁶ Thereby, HS traps BMP GFs within the ECM by binding to positively charged residues within their N‐ or C‐terminal regions. ²⁶ This way GFs of BMP‐2, ‐4, ‐5, ‐6, ‐7, ‐15, and GDF‐5 were shown to bind to heparin and HS. ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹

To better understand how targeting of BMPs by the ECM controls their bioavailability, the presence of BMPs as PD‐GF CPLXs has to be considered. ³² , ³³ , ³⁴ , ³⁵ , ³⁶ Three modes have been proposed for how PDs may regulate GF activity. ³⁴ For BMP‐2, it was shown that it does not form stable CPLXs with its cognate PD. ³³ Therefore, studying BMP‐2 GF bioactivity in absence of the PD may be still relevant for physiological processes. However, other TGF‐β superfamily members form stable CPLXs with their PDs of which some are bioactive and some remain latent when soluble. For instance, BMP‐7, ‐9, and ‐10 form stable, bioactive CPLXs in solution, ³⁵ , ³⁶ , ³⁷ while TGF‐βs, GDF‐8 (myostatin), and GDF‐11 are latent. ³⁸ , ³⁹ , ⁴⁰ For bioactive CPLXs such as BMP‐7 targeting the PD to the ECM protein, fibrillin‐1 was shown to confer latency to the GF. Thereby, BMP‐7 CPLX changes its conformation from a bioactive V‐shape into a latent ring‐shape similar to the latent TGF‐β1 CPLX. ⁴⁰ , ⁴¹ Recently, we demonstrated that bioactive BMP GFs can be liberated from ECM‐bound latent CPLX pools via specific matrix metalloproteinase (MMP) mediated proteolytic processing of the alpha2‐helix of the PD, ⁴² that blocks access of BMP type II receptors in the closed ring‐shape conformation of the CPLX. ⁴¹ Proteolytic PD degradation was also reported to activate GDF‐8, ³⁸ GDF‐11, ³⁹ and TGF‐β ⁴³ , ⁴⁴ , ⁴⁵ from the latent state, and in addition, integrin‐mediated mechanical pulling on the TGF‐β PD to liberate the GF was demonstrated as an activation mechanism. ⁴⁶

However, so far, it has not been investigated whether BMP CPLXs are targeted to GAGs and how the presence of the PD may influence the bioavailability of GAG‐bound BMP CPLXs. In this study, we show for the first time how BMP CPLXs facilitate interactions with GAGs via their PD thereby allowing their spatial concentration in bioactive conformation. Further, we show for BMP‐7 that the state of bioactivity of BMP CPLXs is controlled by the targeting of specific PD binding sites for heparin/HS and the microfibril component fibrillin‐1.

2. MATERIALS AND METHODS

2.1. Antibodies and proteins

For ELISA‐style binding assays, the following antibodies were used: anti‐His₆‐HRP conjugated (#130‐092‐785, Miltenyi Biotec, Germany), Strep‐Tactin HRP conjugated (#2‐1502‐001, IBA, Germany), anti‐His₆ Ab (MAB050, R&D systems), goat antibiotin HRP‐conjugated (#ASB‐OAIA00064, Biozol, Germany), rabbit anti‐human BMP‐7 GF (#500‐P198, PeproTech, Rocky Hill, NJ), mouse anti‐human/mouse/primate BMP‐9 GF (MAB3209, R&D Systems, Minneapolis, MN), mouse anti‐human BMP‐10 GF (MAB2926, R&D Systems). The following proteins were commercially obtained: recombinant human BMP‐7 GF (#354‐BP‐010/CF, R&D Systems), recombinant human BMP‐9 GF (#553104, carrier‐free, Biolegend, San Diego, CA), recombinant human BMP‐9 CPLX (#9624‐BP‐025/CF, R&D Systems), recombinant human BMP‐10 GF (#2926‐BP‐025/CF, R&D Systems). BMP‐7 PD fragments D³⁰‐L⁷², G⁶³‐G¹⁰², and V²³⁶‐R²⁹² (Figure S1) were generated by Custom Peptide Synthesis Services (Thermo Fisher Scientific, Waltham, MA). BMP‐7 PD and truncation variants D³⁰‐R¹⁸⁴, G¹⁶⁶‐R²⁹², F¹⁸⁵‐R²⁹², and G⁷⁴‐R¹⁸⁴ were expressed in E. coli and purified as previously described. ⁴¹

2.2. Glycosaminoglycans

The following GAGs were commercially obtained: biotinylated chondroitin sulfate (#CS‐106, MW: 40 kDa) (Creative PEGWorks, Chapel Hill, NC), biotinylated hyaluronic acid (MW 700 kDa) (#B1557, Sigma‐Aldrich, St. Louis, MO), biotinylated heparin, NANOCS (#HRN1‐BN‐1, Boston, MA). In addition, heparin was biotinylated according to a previously reported protocol. ⁴⁷ HS attached to HSPG core protein fragments which enabled quantitative direct coating to microtiter plates was isolated from mouse embryo preparations following established protocols. ⁴⁸ For quality control, HS was isolated from mouse embryos at embryonic day E18 by overnight digestion with pronase followed by purification over DEAE and PD‐10 columns as previously described. ⁴⁹ The purified HS was then coupled to NHS‐activated sepharose by amine coupling via its attached peptides. The growth factor binding activity of columns packed with HS‐coupled sepharose was assessed by subjecting FGF‐8 and VEGF as positive controls to HS affinity chromatography. Elution of bound growth factors was initiated by using a 0–1 M NaCl gradient under native conditions (Figure S2). To analyze the disaccharide content of the isolated and purified HS from mouse embryos, the material was treated with heparin lyases, further purified over a 3‐kDa spin column, and subjected to quantitative LC/MS using internal tritiated standards as previously described. ⁴⁹ , ⁵⁰ The results of the analysis are presented in Table S1.

2.3. Cell culture

HEK293 EBNA and C2C12 cells (ATCC, Wesel, Germany) were cultivated in high glucose DMEM media (#31966047, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; F7524‐500M, Merck, Germany) and 1% penicillin–streptomycin (Pen/Strep) (#15140‐130, Life Technologies, Carlsbad, CA). BMP‐7 CPLX overexpressing HEK293 cells were cultivated as previously described. ³² , ⁴¹

2.4. Transfection

Constructs with N‐terminal truncations of rF87 containing a C‐terminal 2 × Strep‐tag II were transfected into HEK293 EBNA cells using FuGENE® transfection reagent (#E2312, Promega, Madison, WI) according to the manufacturer's protocol at a ratio of 3 μl FuGENE® to 1 μg DNA at 70% confluency. After 24 h of incubation, the FuGENE®/vector‐containing medium was exchanged by high glucose DMEM medium containing 10% v/v FBS, 1% v/v Pen/Strep, and 1 μg/ml puromycin (P7255, Sigma‐Aldrich) to select for positive clones. Subsequently, the puromycin concentration was adjusted to 0.5 μg/ml.

2.5. Protein expression and purification

cDNA sequences encoding for human proBMP‐7 precursor, BMP‐7 PD variants R182A/R184A/R204A/R214A/K260A, R150A/R154A, R204A/R214A, R260A/R267A, and the human BMP‐10/9 hybrid PD with C‐terminal His₆‐tag were generated by gene synthesis as Invitrogen GeneArt Strings DNA Fragments (Life Technologies). For the generation of the human BMP‐10/9 hybrid PD, the following specific amino acid stretches in BMP‐10 PD were replaced by corresponding stretches in BMP‐9 PD:

BMP‐10 PD amino acid stretches	Replaced by BMP‐9 PD amino acid stretches
179GERNMLVLV187	184SATETKTFLV193
211QKSGSSTHQ219	215VRSDSTKSKNK225
227KHDEAEDASSGRLEIDTSAQ246	233HRKGCDTLDISVPPG247
279LPELDNLGL287	281ESVLKKLSK289

All cDNA sequences were cloned via NdeI/BamHI sites into the pET11a expression vector as previously described. ³³ , ³⁴ , ⁴¹ , ⁴²

N‐terminal fibrillin‐1 fragments rF90 (S¹⁹‐V¹⁵²⁷), start_cbEGF4 (S¹⁹‐E⁵⁷¹), rF87 (S¹⁹‐Ile³²⁹), as well as indicated N‐terminal truncation variants of rF87 (A: G⁵⁵‐Ile³²⁹, B: R⁶²‐Ile³²⁹, C: G⁷⁰‐Ile³²⁹, D: G⁷⁷‐Ile³²⁹) as well as BMP‐7 CPLX were expressed and purified as previously described. ³² , ³³ , ³⁴ , ⁴¹ All fibrillin‐1 protein fragments designed to start with S¹⁹ commenced instead with R⁴⁵ as shown by N‐terminal Edman degradation (Proteome Factory AG, Berlin, Germany). ⁵¹ The untagged processed BMP‐10 CPLX was produced according to a previously reported protocol. ³⁶ A detailed description of the BMP‐10 CPLX overexpression and purification will be published elsewhere. All TGF‐β superfamily PDs and variants as well as the proBMP‐7 precursor were expressed in E. coli BL21‐CodonPlus Competent Cells (Agilent Technologies, Santa Clara, CA, USA) and purified by Ni‐NTA affinity chromatography as previously described ⁴¹ , ⁴² (Figure S1).

2.6. ELISA‐style binding assays

For testing the interaction of BMP‐7 proteins with HS, HS was coated at 5 μg/ml in 1 × PBS and the plate was blocked with Pierce protein‐free blocking buffer (#37570, Thermo Fisher Scientific) according to the manufacturer's recommendation. After several washes with 1 × TBS, BMP‐7 proteins were incubated at 32 nM following serial 1:2 dilutions in 1 × TBS for 2 h, at RT. Antibodies were applied in 20% Pierce protein‐free blocking buffer diluted in 1 × TBS. Washes after BMP‐7 protein incubation and between antibody incubations were in 1 × TBS.

For heparin immobilization, neutravidin protein (#A2666, Thermo Fisher Scientific) was coated at 5 μg/ml in 1 × PBS on Nunc MaxiSorp® flat‐bottom 96 well plates (#735‐0083, VWR, Germany) overnight at 4°C. Wells were washed with 1 × TBS and biotinylated heparin was applied at 5 μg/ml in 1 × TBS for 1 h, at RT. Wells were washed again with 1 × TBS and blocked with 3% BSA/1 × TBS for 1 h, at RT. After washing, GFs were applied at 160 nM following serial 1:2 dilutions in 1 × TBS supplemented with 0.6 M urea for 2 h, at RT. After washing with 1 × TBS/0.6 M urea and subsequently with plain 1 × TBS, primary antibodies were applied in 2% BSA/1 × TBS for 1 h, at RT. Next, secondary antibody incubation in 2% BSA/1 × TBS was conducted for 1 h, at RT. After a final wash with 1 × TBS, the signal was developed by the addition of 50 μl 1‐Step Ultra TMB‐ELISA substrate (#34028, Thermo Fisher Scientific) and the reaction stopped by the addition of 50 μl 10% H₂SO₄. OD was measured at 450 nm using a Sunrise microplate reader (Tecan). For binding of BMP‐2, BMP‐7 GF, or CPLX to immobilized biotinylated heparin (#HRN1‐BN‐1, Nanocs), the plate was blocked with 1× animal‐free blocker (#SP‐5030, Vector Labs, Burlingame, CA) and proteins and antibodies were applied in 1 × animal‐free blocker at 48 nM for BMP‐2 or 32 nM for BMP‐7. Washing steps were with 1 × TBS and incubations were as previously described.

For determining the heparin chain length requirement for sufficient BMP targeting, biotinylated heparin was immobilized on Nunc MaxiSorp® flat‐bottom 96 well plates as described above. The wells were blocked with Pierce protein‐free blocking buffer and BMP proteins were applied for 2 h, at RT on the blocked wells after premixing with heparin oligos of defined disaccharide lengths (Iduron, Cheshire, UK) in 1 × TBS for 30 min at RT. Antibodies were applied in 20% Pierce protein‐free blocking buffer diluted in 1 × TBS and all washing steps were in 1 × TBS.

For interaction studies of BMP‐7 PD variants with heparin, freshly purified proteins were coated in 1 × PBS for 24 h at 4°C. Wells were blocked in 3% BSA/1 × TBS for 1 h, at RT and biotinylated heparin was applied in 1 × TBS for 1 h, at RT after several washes with 1 × TBS. Next, the wells were washed with 1 × TBS before application of the antibiotin HRP‐conjugated antibody (ASB‐OAIA00064, Biozol, Germany) at 1:10 000 in 3% BSA/1 × TBS for 1 h, at RT. For interaction studies of BMP‐7 PD truncation variants with fibrillin‐1, PD variants were coated at 0.1 μM in 1 × PBS, overnight at 4°C and blocked with 5% milk in 1 × TBS for 1 h, at RT. Subsequently, fibrillin‐1 fragments were applied in 2% milk in 1 × TBS for 2 h, at RT. Next, anti‐strep or anti‐His₆ HRP conjugates were applied at 1:10000 or 1:2000, respectively, in 2% milk/1 × TBS for 1 h, at RT.

For the GAG interaction screen with BMP PDs, streptavidin (PI‐21122, Thermo Fisher Scientific) was coated at 5 μg/ml in 1 × PBS, O/N at 4°C. After washes with 1 × TBS‐T, biotinylated heparin, hyaluronic acid, or chondroitin sulfate were incubated for 1 h at RT followed by blocking for 1 h at RT. After several washes, the BMP PDs were provided in solution for 2 h at RT. Bound BMP PDs were detected with the primary anti‐His₆ Ab (MAB050) in the blocking buffer, for 1 h at RT followed by secondary anti‐mouse HRP incubation for 1 h at RT in the blocking buffer.

To assess the coating efficiency of the coated ELISA wells after overnight incubation with BMP PDs, the supernatants were TCA precipitated and after 2 × ice‐cold acetone washes, were resuspended in purified 8 M urea buffered with 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4.

2.7. Bioactivity assays on heparin

To assess BMP bioactivity while bound to heparin, we developed a combined ligand binding/BMP activity assay relying on the mRNA expression of the endogenous BMP response element Id3. ⁴¹ This BMP response is rapid (5hs) and does not face experimental variability due to required transfection and cell lysis steps. For bioactivity assays of heparin‐bound BMPs, cell culture wells (#3599, sterile 96‐well plates, Corning Costar, Corning, USA) were prepared as described for ELISA‐style binding experiments. After incubation of BMP‐2, BMP‐7, or BMP‐9 proteins at 48 nM or 32 nM with immobilized heparin for 2 h wells were washed and blocked before seeding of reporter cells. C2C12 cells resuspended in serum‐free high glucose DMEM media were seeded at 30000 cells in 0.1 ml per well for 5 h at 37°C in a humidified incubator with 5% CO₂. Subsequently, serum‐free DMEM was discarded and 0.1 ml TRIzol reagent (Thermo Fisher Scientific) was added to the cell layer. RNA extraction, reverse‐transcription followed by qPCR to measure the Id3 BMP response was as previously described. ³⁷ , ⁴¹ Four wells of the 96‐well plate were analyzed per indicated condition.

2.8. Heparin affinity chromatography

For Ni‐NTA affinity chromatography, proBMP‐7 precursor, BMP‐7 PD, BMP‐9 PD, BMP‐10 PD, and the BMP‐10/9 hybrid PD were dialyzed into freshly filtered 8 M urea, 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4, by stirring overnight at 4°C using 10 kDa cutoff mini dialysis caps (#69572, Thermo Fisher Scientific). Heparin columns (#17040601, Cytiva, Marlborough, MA, USA) were calibrated with calibration buffer (8 M urea, 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4). Ten milliliters of each protein sample at 1 μM in calibration buffer was flown over the heparin column three times at RT. Columns were washed with 10 ml calibration buffer, and bound protein was eluted at RT in a 0–1 M NaCl gradient in calibration buffer. Obtained flow‐through fractions was condensed over Ni‐NTA columns (#74105, PureCube 100 Ni‐NTA Agarose, Cube Biotech, Germany) and eluted in calibration buffer supplemented with 250 mM imidazole, at RT.

For heparin affinity chromatography of mixtures of BMP‐7/BMP‐10 CPLXs (molar ratio: 1:1), samples were dialyzed into 150 mM NaCl, 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4, overnight, at 4°C. Heparin columns were calibrated with native calibration buffer (150 mM NaCl, 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4), and 10 ml of sample was flown over three times over the heparin column, at 4°C. Next, the heparin column was washed with native calibration buffer for 10 ml and stepwise elution was conducted in 20 mM NaH₂PO₄ × 2 H₂O, pH = 7.4 supplemented with 0.25 M NaCl (eluted in 6 fractions), 0.5 M NaCl (eluted in 7 fractions), 1 M NaCl (eluted in 6 fractions), and 2 M NaCl (eluted in 11 fractions), at 4°C. The volume of each fraction was 0.5 ml. Heparin column flow through was concentrated over an anion exchanger column (#29051325, Cytiva, Marlborough, MA, USA) and subsequently over a Ni‐NTA column, at 4°C.

2.9. Circular dichroism spectroscopy

BMP‐7 PD, the BMP‐7 R182A/R184A/R204A/R214A/K260A PD variant, and BMP‐10 PD were dialyzed into 5 mM HClO₄ in the cold room. BMP‐10/9 hybrid PD required dialysis into 20 mM HClO₄ in the cold room to retain solubility. CD spectra were recorded using a Jasco J‐715 spectropolarimeter in a 1 mm path length quartz cell (Hellma, Germany) from 260 to 170 nm at 20°C. BMP‐7 PD and mutant were measured at 0.2 mg/ml, whereas BMP‐10 PD and the BMP‐10/9 hybrid PD were measured at 0.1 mg/ml. Buffer contribution was subtracted, theta/machine units were converted to Δε and percentage secondary structure was calculated using the CDSSTR algorithm (Dichroweb server).

2.10. Surface plasmon resonance

SPR experiments were performed as described previously ⁴¹ using a BIAcore 2000 system (Cytiva, Marlborough, MA). CM5 chips (#BR100012, Cytiva, Marlborough, MA) were calibrated with running buffer HBS‐EP (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20) (#BR100188, Cytiva, Marlborough, MA), and BMP‐7 PD, the BMP‐7 PD mutant R182A/R184A/R204A/R214A/K260A, BMP‐10 PD and the BMP‐10/9 hybrid PD were immobilized via NHS/EDC amine coupling (#BR100050, Cytiva, Marlborough, MA) at 400 RU. BMP‐7 GF (0–80 nM) or BMP‐10 GF (0–160 nM) was flown over as an analyte in the HBS‐EP buffer. Kinetic constants were calculated as described previously. ⁴¹

For assessing how heparin binding to BMP‐7 GF and CPLX impacts the interaction with BMP receptors, the ActRIIB receptor ectodomain (#340‐R2‐100/CF; R&D Systems) was immobilized as previously explained at 800 RUs to the CM5 chip surface. Twenty nanomolars of BMP‐7 proteins were preincubated with 0–80 nM heparin and subsequently injected into the immobilized receptor.

To narrow down the BMP‐7‐binding site within the FUN domain of fibrillin‐1, a CM5 chip was calibrated with HBS‐EP buffer and BMP‐7 PD was immobilized via amine coupling at 1000 RUs. N‐terminal truncation variants of the FUN domain were injected at 100 nM in the HBS‐EP buffer.

For the BMP‐7 or BMP‐10 GF/heparin SPR‐binding study, the Sierra SPR‐32 instrument (Bruker, Hamburg, Germany) was employed. PEG chips (Xantec, Düsseldorf, Germany) were calibrated in HBS‐EP buffer supplemented with 1 M urea (running buffer) and neutravidin was coupled via amine coupling. Next, biotinylated heparin (experimental surface) or biotin (control surface) were flown over in 1 × PBS. After calibration of the surfaces with HBS‐EP/1 M urea buffer, 0–60 nM (serial 1:2 dilution in HBS‐EP/1 M urea buffer) BMP‐7 or BMP‐10 GFs were injected. The final NaCl concentration in the HBS‐EP/1 M urea buffer was 150 mM to mimic the physiological condition. RU responses from control surfaces were subtracted from the RU responses of the experiment surfaces and graphs were plotted using the Sierra analysis software (Bruker).

2.11. Negative‐staining transmission electron microscopy and quantification

Purified, cell culture‐derived BMP‐7 CPLX or purchased BMP‐9 CPLX were premixed with heparin chains (#B9806, Sigma) and subsequently adsorbed to negative‐staining EM grids for transmission electron microscopy (TEM). To calculate the percentage of molecules binding to heparin chains utilizing the PD arm region, 500 molecules/field and 50 fields in total were analyzed for BMP‐7 or BMP‐9 CPLX samples.

2.12. Molecular modeling

The structural model of the processed BMP‐7 CPLX was built on the activin CPLX (processed proactivin) atomic model (PDB: 5HLZ) ⁵² by modeling individual BMP‐7 PDs on the activin PDs using the Phyre2 server. ⁵³ Next, the modeled BMP‐7 PDs as well as the crystal structure of the BMP‐7 GF (PDB: 1LXI) ⁵⁴ were aligned chain‐by‐chain to the PDs or GF of the activin CPLX atomic model (PDB: 5HLZ) using the “Match Maker” tool of the UCSF Chimera software. ⁵⁵ The activin CPLX atomic model was then deleted to reveal the resulting processed BMP‐7 CPLX structural model. The human BMP‐9 CPLX structural model was built on the mouse BMP‐9 CPLX atomic model (PDB: 4YCG) ³⁵ by modeling individual human BMP‐9 PDs on the mouse BMP‐9 PDs in 4YCG using the Phyre2 server. Subsequently, the modeled human BMP‐9 PDs were aligned chain‐by‐chain to the mouse BMP‐9 PDs of the 4YCG atomic model using the “Match Maker” tool of the UCSF Chimera software. Next, the mouse BMP‐9 PDs were deleted to reveal the resulting processed human BMP‐9 CPLX structural model. To visualize the identified heparin‐binding BMP‐7 PD residues, the surface of the BMP‐7 CPLX model was shown and the identified residues were labeled. Similarly, to visualize the identified heparin‐binding BMP‐9 PD residues, the surface of the BMP‐9 CPLX atomic model (PDB: 4YCG) or human BMP‐9 CPLX structural model was shown and the identified residues were labeled. All model manipulations were performed using the UCSF Chimera software. The necessary heparin chain length for spanning all identified heparin‐binding sites, was estimated by comparing the length of a structurally idealized heparin chain with the mentioned BMP‐7 and BMP‐CPLX structures. Structures where manually positioned using open‐source PyMol 2.6 guided by the previously identified binding sidechains. Secondary structure elements were detected using the STRIDE algorithm ⁵⁶ and visualized by the Azahar plugin. ⁵⁷

2.13. VEGF heparin‐binding domain predictions

To identify heparin‐binding sites within the processed PDs of BMP CPLXs, the VEGF heparin‐binding domain NMR solution structure (PDB: 2VGH) was aligned chain‐by‐chain to the respective processed PD to be analyzed and subsequently, a mutual sequence alignment was performed using the UCSF chimera software. The method identified stretches within the PD with the highest‐homology to the VEGF heparin‐binding domain.

2.14. Molecular docking

BMP‐7 PD modeled on activin PD, the activin CPLX atomic model (PDB: 5HLZ), the BMP‐9 CPLX atomic model (PDB: 4YCG), the human BMP‐9 CPLX structural model, the proGDF‐8 (promyostatin) atomic model (PDB: 5NTU), ⁵⁸ the BMP‐7 GF unified model (PDB: 1LXI), and the BMP‐2 GF unified model (PDB: 6OMN) ⁵⁹ were docked to heparin that was used as a ligand. To visualize the docked heparin chains in the BMP‐2/BMPRI interaction model (PDB: 1REW), ⁶⁰ the generated docking model of BMP‐2 GF to heparin was overlaid on the BMP‐2/BMPRI interaction model using the “Match Maker” tool of the UCSF chimera software. The closed‐ring BMP‐7 CPLX model ⁴² was docked to the NMR solution structure of the fibrillin‐1 FUN domain (PDB: 2M74) ⁶¹ that was used as a ligand. All molecular docking experiments were performed by employing the ClusPro 2.0 server. ⁶² Visualization of the heparin/protein and protein/protein interactions after docking was with the UCSF chimera software.

2.15. Electrostatic surface coloring

The atomic models of proGDF‐8 (promyostatin) (PDB: 5NTU) and BMP‐9 CPLX (PDB: 4YCG) were visualized using the Swiss‐PdbViewer 4.1.0 and electrostatic potential was computed by employing the Coulomb method.

3. RESULTS

3.1. BMP‐7 GF and CPLX interact with heparin/heparan sulfate

Previous data showed that the BMP‐7 GF interacts with high affinity with heparin/HS. ²⁸ , ³¹ To test whether also the BMP‐7 CPLX interacts with sulfated disaccharides that naturally exist in tissues, we subjected BMP‐7 CPLX to an ELISA‐style binding assay with HS that was isolated from mouse embryo preparations following established protocols ⁴⁸ and was immobilized to the solid phase. Our results revealed that both BMP‐7 GF and CPLX showed a strong interaction signal with HS that saturated at 16 nM for BMP‐7 GF or CPLX (apparent K_D = 4 nM) (Figure 1A). Also, both BMP‐7 GF and CPLX showed strong binding to heparin that saturated at 32 nM for BMP‐7 GF or CPLX (apparent K_D = 8 nM) (Figure 1A). To determine the disaccharide chain length required for sufficient BMP‐7 GF or CPLX targeting, we performed competition binding experiments of soluble heparin chains of defined lengths against immobilized heparin for BMP‐7 GF or CPLX binding (Figure 1B). Our results showed that a minimal chain length of 10 disaccharide units is required to significantly reduce binding to immobilized full‐length heparin by 30% (Figure 1B).

FIGURE 1.

Heparin/HS interacts with BMP‐7 GF and BMP‐7 CPLX requiring at least a length of 10 disaccharide units. (A) Solid‐phase ligand‐binding assay shows that BMP‐7 GF and CPLX interact with immobilized HS and heparin. (B) Determination of heparin disaccharide chain length requirement for sufficient BMP‐7 GF and CPLX targeting. Competition binding studies of soluble heparin chains of defined lengths against immobilized heparin for BMP‐7 binding indicated that at least 10 disaccharide units are required for significant in‐solution targeting (reduction of binding signal to immobilized heparin by 30%). Representative binding assays are shown. Data points represent averages of duplicates ± SD. Each binding experiment was confirmed by two independent experiments.

3.2. Bioactivity of heparin‐bound BMP‐7 GF is abolished while it is preserved in heparin‐bound CPLX

To assess the bioactivity of BMP‐7 when immobilized to heparin/HS, we used biotinylated heparin in BMP bioactivity assays due to its commercial availability. Biotinylated heparin was immobilized in streptavidin‐coated wells followed by incubation of BMP‐7 GF or CPLX at equal molarity. Comparable immobilization levels of the GF were demonstrated by direct ELISA using a polyclonal anti‐BMP‐7 GF antibody (Figure 2A). C2C12 cells were seeded on heparin‐immobilized BMP‐7 GF or CPLX for 5 h followed by measurement of mRNA levels of the BMP response element Id3. Interestingly, BMP‐7 GF interaction with heparin resulted in abolished bioactivity, whereas the heparin‐bound BMP‐7 CPLX showed comparable bioactivity to the GF administered in solution (Figure 2B). Molecular docking of the atomic model of the BMP‐7 GF dimer (PDB: 1LXI) ⁵⁴ on heparin indicated that the proposed interaction of heparin chains to the characterized heparin‐binding site of the BMP‐7 GF ³¹ renders crucial residues of the ActRII receptor binding site within the GF (S¹¹³, L¹²⁵) inaccessible (Figure S3A). This explains the diminished bioactivity observed for the heparin‐bound BMP‐7 GF since it needs to associate with the type II receptor first to trigger the BMP response. In contrast to BMP‐7 GF, immobilization of BMP‐2 GF to heparin did not confer latency (Figure 2C). Molecular docking of heparin to the characterized heparin‐binding site in the atomic model of BMP‐2 GF (PDB: 6OMN) ³¹ showed that heparin binding still allows for receptor engagement since the crucial residues within the BMPRI (L⁵¹ and D⁵³) are still accessible in the atomic model of the BMP‐2/BMP type I receptor complex (PDB: 1REW) (Figure S3B). ⁵⁵ In summary, our activity assays of heparin‐bound BMPs revealed that targeting different GF binding sites via GAGs modulates their bioactivity. Thereby, the presence of the BMP‐7 PD protects the GF from its GAG‐mediated inactivation (Figure 2D).

FIGURE 2.

Heparin/HS confers latency to BMP‐7 GF while BMP‐7 CPLX and BMP‐2 GF remain bioactive. (A) Direct ELISA using anti‐BMP‐7 GF as detector antibody indicated that incubation of 32 nM of BMP‐7 GF or CPLX with immobilized heparin results in the binding of comparable BMP‐7 GF amounts to wells. (B) Bioactivity measurements of heparin‐bound BMP‐7 showed that the GF is rendered latent while the CPLX remains bioactive. (C) Bioactivity measurements revealed that binding to heparin does not affect the bioactivity of BMP‐2 GF. (D) Scheme summarizing bioactivity results of BMP‐2 and BMP‐7 when bound to heparin. Representative binding assays are shown. Data points represent averages of triplicates ± SD. Each binding experiment was confirmed by three independent experiments.

3.3. Heparin‐binding to BMP‐7 GF blocks BMP receptor access while BMP‐7 PD protects the GF from heparin inactivation in BMP‐7 CPLX

Since we observed that the bioactivity of heparin‐bound BMP‐7 CPLX was maintained, we wanted to investigate whether the presence of the PD protects the GF from heparin‐mediated inactivation and thereby allows BMP receptor access. To examine this, we employed surface plasmon resonance (SPR) and measured the binding activity of BMP‐7 GF or CPLX to the immobilized ectodomain of ActRII after preincubation with increasing concentrations of heparin (Figure 3). Our binding studies showed that the addition of heparin to the BMP‐7 GF resulted in a reduction of its interaction signal with ActRII in a concentration‐dependent fashion. However, this reduction in binding was not observed when BMP‐7 CPLX was preincubated with heparin (Figure 3), suggesting that the PD protects the GF from heparin‐mediated blockage of its type II receptor binding site.

FIGURE 3.

Heparin inhibits BMP‐7 GF but not BMP‐7 CPLX interaction with ActRIIB. (A) BMP‐7 GF or CPLX (20 nM) were preincubated with heparin in increasing concentrations (0–80 nM) before flown over the ActRIIB receptor ectodomain that was immobilized onto a CM5 chip. (B) (left) Percentage of binding signal for BMP‐7 GF and CPLX to ActRIIB in the presence of increasing heparin concentrations. Data points represent averages obtained from three independent experiments. Error bars indicate ± SD. (right) Cartoon illustrates BMP‐7 GF and CPLX interaction with ActRII in presence of heparin.

3.4. Mapping of the heparin‐binding site in BMP‐7 CPLX

To investigate the interaction interface of BMP‐7 CPLX and heparin, purified CPLX was preincubated with heparin chains and analyzed by transmission electron microscopy (TEM) after negative staining. The quantitative analysis of electron micrographs in 50 different fields revealed that 95% of all imaged CPLX particles make contact with heparin chains via the tip of the PD arm region (Figure 4A), suggesting that the characterized heparin‐binding site of the GF moiety ³¹ is masked within the CPLX. Molecular modeling of the BMP‐7 CPLX using the activin CPLX (PDB: 5HLZ) as a template provided further insight into the potential heparin/BMP‐7 PD interaction interface (Figure 4B).

FIGURE 4.

BMP‐7 CPLX interacts with heparin via its PD arm regions. (A) Negative staining TEM of BMP‐7 CPLX incubated with heparin chains (arrows indicate interaction points of PD moieties with heparin chains). For the quantification, 500 BMP‐7 CPLX particles were quantified per field in 50 different fields. Scale bars: 30 nm (overview) or 20 nm (magnified micrographs). (B) Structural model of BMP‐7 CPLX built on processed activin CPLX atomic model (PDB: 5HLZ). PD: beige, GF: light gray, heparin chain: black dotted line.

In addition, when the proBMP‐7 full‐length precursor was produced in E. coli and subjected to heparin affinity chromatography, we detected a small degradation fragment of about 15 kDa that eluted very late in the salt gradient thereby indicating a high‐binding affinity (Figure S4). Mass spectrometry analysis after SDS‐PAGE revealed that the small fragment localized within the C‐terminal part of the BMP‐7 PD (Figure S4). These data confirmed that the high‐affinity heparin‐binding site in the BMP‐7 PD resides within its C‐terminal region.

To fine map the BMP‐7 CPLX heparin‐binding site within the C‐terminal region of the BMP‐7 PD, we performed sequence and structural alignments of known GF heparin‐binding motifs with TGF‐β superfamily members. Of the several growth factors tested, the vascular endothelial growth factor (VEGF) heparin‐binding domain was mapped correctly to the characterized heparin/HS binding site within the activin PD. ⁶³ Chain‐by‐chain sequence alignment of the VEGF heparin‐binding domain to the activin PD mapped the heparin/HS binding site to the PD arm region within a structural motif composed of one alpha‐helix and three beta‐sheets that closely resemble the VEGF heparin‐binding domain structure. This finding could be supported by molecular docking experiments which also localized the heparin chains to the same site of the activin CPLX atomic model (PDB: 5HLZ) (Figure 5A). By applying the same methodology to the BMP‐7 CPLX structural model, we aligned the VEGF heparin‐binding domain chain‐by‐chain to the BMP‐7 PD (Figure 5B). This approach led to the identification of specific heparin‐binding residues within the C‐terminal region of the PD, thereby agreeing with the obtained TEM results and mapping data (Figure 4, Figure S4). More specifically, the identified heparin‐binding residues localize within three beta‐sheets of a structural motif within the BMP‐7 PD arm that is similar to the VEGF heparin‐binding domain. Manual positioning of heparin chains within this motif guided by our TEM data, while simultaneously including the predicted heparin‐binding residues, confirmed their possible participation in the PD–heparin interaction interface (Figure 5B and Figure S5). Furthermore, molecular docking of heparin to the atomic model of proGDF‐8 (promyostatin) (PDB: 5NTU) also indicated that heparin interacts with GDF‐8 via its PD arm region (Figure S6B), thereby confirming our previous finding that the GDF‐8 PD is targeted to GAG chains of the heparan sulfate proteoglycan (HSPG) perlecan. ³⁴

FIGURE 5.

VEGF heparin‐binding domain sequence alignment predicts the heparin/HSbinding sites within activin and BMP‐7 CPLX. (A) (top) Sequence alignment of the VEGF heparin‐binding domain to the activin PD. (bottom) Chain‐by‐chain alignment of the VEGF heparin‐binding domain to the activin PD and molecular docking of heparin to the atomic model of the processed activin CPLX (PDB: 5HLZ). (B) (top) Sequence alignment of the VEGF heparin‐binding domain to the BMP‐7 PD. VEGF residues participating in heparin binding ⁸² are marked in yellow boxes. (bottom) Chain‐by‐chain alignment of VEGF heparin‐binding domain to BMP‐7 PD within the BMP‐7 CPLX structural model and identification of specific heparin‐binding residues within the PD arm region.

To validate the participation of the identified PD residues in heparin binding, we generated a PD variant in which all positively charged residues of the interaction interface were replaced by alanines (Figure 6A). Circular dichroism (CD) analysis of the resulting BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant showed that the secondary structure remained unaltered by the introduced mutations (Figure 6B). ELISA‐style binding studies showed that heparin binding was abolished for the BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant (Figure 6D), while the capability to interact with the BMP‐7 GF was preserved as shown by SPR interaction studies (Figure 6C). In addition, visualization of the BMP‐7 CPLX structural model surface indicated that four (R¹⁸², R¹⁸⁴, R²⁰⁴, and R²¹⁴) out of five mutated residues are localized on the surface of the PD arms, where they are accessible by heparin chains (Figure 6E), whereas K²⁶⁰ is hidden at the back of the structure and further stabilizes the interaction with penetrating heparin chains (Figure 5 and Figure S5B). Thereby, this heparin‐binding interface can be viewed as a pocket, where R204 and R214 control the entry and R182 and R184 the exit site for the heparin chains (Figure S5B). By examining the BMP‐7 PD stretch that participates in this pocket (P¹⁴⁹‐R²⁶⁷) an involvement of additional positively charged residues such as R¹⁵⁰, R¹⁵⁴, and R²⁶⁷ that flank the identified five critical residues may be possible. However, when generating double mutants that neutralize the positive charge of the flanking residues R150A/R154A or R260A/R267A as well as the charge of the predicted entry site R204A/R214A, no significant reduction of the heparin‐binding affinity was observed (Figure S5C). These results indicate that all identified five PD residues participate in the interaction interface with heparin and have an equal contribution to the overall affinity.

FIGURE 6.

Characterization of the BMP‐7 CPLX heparin‐binding site. (A) SDS‐PAGE and Coomassie staining after purification of the BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant. (B) CD analysis showed that the BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant assumes a similar secondary structure as wild‐type BMP‐7 PD. (C) The SPR‐binding study showed that the immobilized BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant has a similar binding affinity to flown over BMP‐7 GF as wild‐type BMP‐7 PD. Representative sensorgrams of binding experiments are shown. K_Ds ± SD was derived from three independent experiments. (D) (left) Coating efficiency of mutant and wild‐type BMP‐7 PD to wells of microtiter plates was comparable as assessed by SDS‐PAGE and Coomassie staining of supernatant before and after coating (right) BMP‐7 PD R182A/R184A/R204A/R214A/K260A variant does not bind to heparin in solid‐phase ligand binding assay. Data points ± SD from representative binding assays are shown. Each binding experiment was confirmed by three independent experiments. (E) Surface model of the BMP‐7 CPLX interacting with heparin. PD residues are marked in blue, and GF residues are marked in orange. Positively charged residues within the BMP‐7 PD heparin‐binding interface that are surface‐exposed are marked in magenta.

3.5. Comparison of the BMP‐7 PD‐binding sites for heparin and fibrillin‐1 suggests a structural integration mode to control GF bioavailability

In order to better understand how heparin binding and fibrillin‐1 binding can be integrated to control BMP‐7 CPLX bioavailability, the fibrillin‐1/BMP‐7 CPLX interaction interface was further characterized (Figure 7). We therefore further narrowed down the BMP‐7 PD‐binding site within the fibrillin unique N‐terminal domain (FUN) of fibrillin‐1 but also the fibrillin‐1 binding site within the BMP‐7 PD. Previously, we determined via SPR that the fibrillin‐1‐binding site for the BMP‐7 PD resides within Gly⁷⁴–Arg¹⁸⁴, ⁴¹ and that the BMP‐7 PD‐binding site within the fibrillin‐1 localizes to the unique N‐terminal domain (FUN). ³³ Further N‐terminal truncation of FUN by 10 residues resulted in 50% reduction of binding to the immobilized BMP‐7 PD as assessed by SPR‐binding studies, while the additional deletion of the subsequent 15 residues completely abolished the interaction (Figure 7A). This suggests that the BMP‐7 PD‐binding site is mostly contained within ⁴⁴RGGGGHDALKGPNVCGSRYNAYCCPGWKTLPG⁷⁶ of fibrillin‐1. ELISA‐style binding studies with BMP‐7 PD truncation variants confirmed the previously characterized fibrillin‐1‐binding site within BMP‐7 PD and further narrowed it down to Gly¹⁰²–Gly¹⁶⁶ (Figure 7B). ⁴¹ The site was further narrowed down by employing molecular docking of the closed‐ring BMP‐7 CPLX ⁴² to the NMR solution structure of the FUN domain (PDB: 2M74). ⁶¹ After careful examination of 30 docking models generated by the online server, the fibrillin‐1‐binding site within the BMP‐7 PD was mapped to the stretch ¹⁵⁶FRFDLSK¹⁶² (Figure 7B). Further, in the closed‐ring/FUN domain interaction model, the identified heparin‐binding sites are hidden within the interior of the BMP‐7 CPLX closed‐ring structure, whereas the fibrillin‐1 identified stretch is surface exposed (Figure 8A). However, in the model of the bioactive, V‐shape BMP‐7 CPLX bound to heparin, the identified fibrillin‐1 interaction stretch is masked, whereas the heparin‐binding sites are available for interaction (Figure 8B). This suggests that the conformational change of the BMP‐7 PD integrates the bioavailability of the GF by restricting access to fibrillin‐1 or GAG‐binding sites. Interestingly, SPR‐binding studies showed that the soluble BMP‐7 CPLX binds with the same molecular affinity (K_D ~ 7 nM) to the immobilized N‐terminal half of fibrillin‐1 or immobilized heparin (Figure S7). This suggests tight and specific targeting of BMP‐7 CPLX to the ECM microenvironment in both conformations.

FIGURE 7.

Characterization of the fibrillin‐1/BMP‐7 PD‐binding site. (A) (top) Domain structure of used fibrillin‐1 proteins spanning the N‐terminal region. (bottom, left) SDS‐PAGE and Coomassie staining of rF87 and N‐terminal truncation variants after streptavidin affinity chromatography. (middle) SPR interaction study of rF87 and N‐terminal truncation variants with immobilized BMP‐7 PD. (right) Percentage of the binding signal detected in shown sensorgrams. (B) (top) ELISA detection of fibrillin‐1 fragments rF11 and start_EGF4 incubated in increasing concentrations with coated BMP‐7 PD (full length) and truncation variants (bottom). Residues of the identified fibrillin‐1‐binding stretch (G¹⁰²‐G¹⁶⁶) within the BMP‐7 PD sequence based on binding data are marked in purple. BMP‐7 PD residues of the fibrillin‐1‐binding site identified by molecular docking are highlighted in red. Data from representative binding assays are shown. Each binding experiment was confirmed by three independent experiments.

FIGURE 8.

Conformational change of BMP‐7 CPLX from open bioactive to closed latent conformation controls surface exposure of binding sites for heparin and fibrillin‐1. (A) Molecular docking of BMP‐7 CPLX in closed‐ring conformation to the NMR solution structure of the fibrillin‐1 FUN domain (PDB: 2M74). Visualization of the BMP‐7 PD‐binding sites of fibrillin‐1 (red) and heparin (magenta) within the surface of the docking model. (B) Visualization of the fibrillin‐1 and heparin‐binding sites within the surface model of BMP‐7 CPLX in open V‐shape conformation. PD residues: blue, GF residues: orange, BMP‐7 PD‐binding sites within FUN: black and yellow, residues of FUN domain binding site in BMP‐7 PD: red, heparin‐binding residues in BMP‐7 PD: magenta.

3.6. Interaction screen of PDs of TGF‐β superfamily members with immobilized GAGs

It is a widely accepted notion that GAGs act as reservoirs that concentrate several BMPs locally within the ECM. ²⁸ To investigate whether other BMPs can also interact with GAGs via their PDs, we performed an interaction screen with soluble BMP PDs incubated in increasing concentrations with immobilized GAGs on microtiter wells (Figure 9). We found that the PDs of TGF‐β1, BMP‐4, and ‐9 not only showed robust binding to heparin but also interaction with hyaluronic acid (HA) and chondroitin sulfate (CS) with varying affinities (Figure 9). In contrast, BMP‐10 PD did not show a significant interaction signal with any of the tested GAGs. However, its closest relative BMP‐9 PD showed a similar strong interaction with heparin, HA, and CS. BMP‐4 PD also showed binding with reduced affinity to HA while binding to CS was significantly weaker. Weak binding to HA and CS could also be detected for the TGF‐β1 PD (Figure 9).

FIGURE 9.

Interaction screen with soluble BMP PDs reveals binding to several immobilized GAGs. TGF‐β1, BMP‐4, BMP‐9, and BMP‐10 PDs were given in 1:2 serial dilutions on immobilized heparin, hyaluronic acid (HA) and chondroitin sulfate (CS) on microtiter wells. BMP PDs were detected using an anti‐His₆ antibody. Data from representative binding assays are shown. Each binding experiment was confirmed by three independent experiments.

3.7. Characterization of the BMP‐9 CPLX heparin‐binding site

Sequence alignments of the BMP‐7, BMP‐9, and BMP‐10 PDs suggest that the identified heparin‐binding stretches within the PD arm region of BMP‐7 may also be conserved in BMP‐9 PD (Figure 10A). This is in line with the observation that the potentially conserved heparin‐binding stretches between BMP‐7 and BMP‐9 PDs are absent in the BMP‐10 PD where negatively charged amino acid insertions are also present (Figure 10A). In agreement with this finding, heparin affinity chromatography of the BMP‐9 PD yielded a clear peak after salt gradient elution at 227 mM NaCl indicating robust binding similar to the BMP‐7 PD, whereas the BMP‐10 PD failed to bind heparin (Figure 10B). To further investigate whether the BMP‐10 CPLX is capable to interact with heparin, we purified the processed BMP‐10 CPLX from overexpressing HEK293 cells and subjected it to heparin affinity chromatography after premixing with BMP‐7 CPLX. SDS‐PAGE analysis followed by Coomassie staining of elution fractions and flow through showed that only BMP‐7 CPLX was eluted from the heparin column at 1 M NaCl (Figure S8), whereas BMP‐10 CPLX was exclusively detected in the heparin flow through (Figure 10D). Next, we wanted to investigate whether the BMP‐9 and ‐10 GFs in absence of their cognate PDs are principally capable to interact with heparin. Sequence alignment of the BMP‐7, ‐9, and ‐10 GFs suggested that the described BMP GF heparin‐binding motif XBBXBX ³¹ is disturbed in the BMP‐9 and ‐10 GFs by a single glutamic acid (E⁴¹⁹ or E⁴¹⁴, respectively) substitution (Figure 10A). In line with this finding, our SPR and ELISA interaction assays showed no binding of BMP‐9 and ‐10 GFs to immobilized heparin (Figure 10C).

FIGURE 10.

BMP‐9 PD but not BMP‐10 PD interacts with heparin. (A) (left) Multiple sequence alignment of BMP‐7, BMP‐9, and BMP‐10 PDs. (right) Multiple sequence alignment of BMP‐7, BMP‐9, and BMP‐10 GFs. Depicted heparin‐binding motifs were reported previously. ³¹ (B) Heparin affinity chromatography and salt gradient elution of BMP‐7, BMP‐9, and BMP‐10 PDs under denaturing conditions (fraction volume: 0.5 ml). (C) SPR‐ (left) and ELISA‐binding (right) studies of soluble BMP‐7, BMP‐9, and BMP‐10 GFs with immobilized heparin. (D) Heparin affinity chromatography and stepwise salt elution under native conditions of BMP‐7 CPLX premixed with BMP‐10 CPLX followed by SDS‐PAGE and Coomassie staining. Samples before chromatography (input) and the condensed flow through were analyzed. Data from representative binding assays are shown. Each binding experiment was confirmed by three independent experiments.

To investigate the BMP‐9 CPLX/heparin interaction interface, commercially available BMP‐9 CPLX was incubated with heparin followed by negative staining TEM. Consistent with the findings for BMP‐7 CPLX (Figure 4A,B), also BMP‐9 CPLX showed contact with heparin chains via its PD arm regions in 95% of particles examined in 50 different fields (Figure 11A). Bioactivity measurements of BMP‐9 CPLX, when immobilized to heparin, showed that it remains bioactive similar to BMP‐7 CPLX (Figure S9). To narrow down the heparin‐binding site of the BMP‐9 CPLX, the VEGF heparin domain was chain‐by‐chain aligned to the structure of the BMP‐9 PD of the human BMP‐9 CPLX structural model, or the BMP‐9 CPLX atomic model (PDB: 4YCG) (Figure 11B,C and Figure S10A). The alignment revealed a similar structural motif upon vertical rotation of the human BMP‐9 CPLX structural model by 90° that resembled the VEGF heparin‐binding domain structure and was composed of one alpha‐helix and three beta‐sheets that created a binding pocket within the BMP‐9 PD arm. By employing molecular docking of heparin to the human BMP‐9 CPLX structural model or the BMP‐9 CPLX atomic model (PDB: 4YCG), it was possible to narrow down the BMP‐9 CPLX heparin‐binding site to specific stretches within the PD arm region at the entry of the binding pocket (Figure 11C and Figure S10A) that are compatible with electrostatic surface coloring showing the presence of positive charge locally (Figure 11D). Hence, to identify the complete set of the human BMP‐9 PD heparin‐binding residues, the heparin chain was manually positioned within the binding pocket based on the negative staining TEM results of the BMP‐9 CPLX (Figure 11E, left)

FIGURE 11.

BMP‐9 CPLX interacts with heparin chains via the PD arm region. (A) Negative staining TEM of BMP‐9 CPLX incubated with heparin chains (arrows indicate interaction points of PD moieties with heparin chains). Scale bars: 30 nm (overview) or 20 nm (magnified micrographs). For quantification, 500 BMP‐9 CPLX particles per field were quantified in 50 different fields. (B) Sequence alignment of the VEGF heparin‐binding domain to the human BMP‐9 PD. VEGF residues participating in the interaction ⁸² with heparin are marked in the yellow box. (C) Molecular docking of heparin to the human BMP‐9 CPLX structural model and chain‐by‐chain alignment of VEGF heparin‐binding domain to the modeled BMP‐9 PD. (D) Electrostatic surface coloring to identify positively and negatively charged patches in the BMP‐9 CPLX surface model (PDB: 4YCG). Positive charge is marked in blue, and negative charge is marked in red. (E) (left) Manual positioning of the heparin chain within the potential binding pocket of BMP‐9 PD as suggested by VEGF alignments and molecular docking. (right) Surface model of the BMP‐9 CPLX indicates that the heparin‐binding residues at the entry and exit points of the heparin‐binding pocket are exposed. PD residues: cyan, GF residues: brown, positively charged residues of the characterized BMP‐9 CPLX heparin‐binding site: magenta, conserved residues between BMP‐9 and BMP‐10 PDs: blue.

To experimentally assess whether these stretches harbor heparin‐binding potential (Figure S10B), we introduced the predicted heparin‐binding stretches of the BMP‐9 PD into the BMP‐10 PD by sequence swapping of the corresponding regions. CD analysis revealed that the secondary structure of the BMP‐10/9 hybrid PD was similar to that of BMP‐10 PD (Figure 12A). Also, SPR analysis showed that the BMP‐10/9 hybrid PD has a similar capability to interact with the BMP‐10 GF as the BMP‐10 PD (Figure 12B). To test the heparin‐binding capability of the BMP‐10/9 hybrid PD, heparin affinity chromatography was employed after pre‐mixing it with the BMP‐10 PD. As expected, only the BMP‐10/9 hybrid PD was eluted from the heparin column in a salt gradient at 242 mM, whereas the BMP‐10 PD was exclusively detected in the flow through (Figure 12C). In addition, visualizing the surface of the human BMP‐9 CPLX structural model confirmed that the positively‐charged residues of the identified heparin‐binding stretches within the BMP‐9 PD are surface‐exposed and therefore available for heparin interaction (Figure 11E, right). These residues appear to create a pocket that heparin chains can penetrate to facilitate efficient contact with BMP‐9 PD residues (Figure 11E).

FIGURE 12.

Characterization of the BMP‐9 CPLX heparin‐binding site. (A) CD analysis of BMP‐10/9 hybrid PD and BMP‐10 PD shows similar secondary structures. (B) SPR interaction study shows that soluble BMP‐10 GF has a similar affinity to immobilized BMP‐10/9 hybrid PD and BMP‐10 PD. Representative sensorgrams of binding experiments are shown. K_Ds ± SD was derived from three independent experiments. (C) (left) Heparin affinity chromatography of the BMP‐10/9 hybrid PD premixed with BMP‐10 PD under denaturing conditions. Elution was performed in a NaCl gradient and elution fractions were analyzed by SDS‐PAGE followed by Coomassie staining. (right) Flow through of experiment (left) was condensed by nickel chelate affinity chromatography and run on separate SDS‐gel together with sample inputs.

3.8. In silico determination of heparin chain length requirement for sufficient BMP targeting

Next, we wanted to determine the minimal heparin chain length requirement for sufficient BMP targeting by utilizing the known BMP GF and our newly identified BMP PD interaction interfaces with heparin. By manually positioning a heparin chain glycan to the heparin‐binding residues of BMP‐7 GF, BMP‐7 CPLX, and BMP‐9 CPLX, we observed that a minimum of 10 heparin disaccharide units are required for sufficient BMP targeting (Figure S11) which is consistent with our experimental binding data for BMP‐7 GF and CPLX (Figure 1B).

4. DISCUSSION

In this study, we identified for the first time the heparin/HS interaction interface of BMP‐7 and BMP‐9 CPLXs and investigated the molecular requirements for bioactivity and latency of complexed and free GF in the presence and absence of heparin. Interactions of recombinantly generated free BMP GFs to GAGs have already been reported in several studies ²⁷ , ²⁸ , ³¹ but in which conformation and bioactivity state physiologically relevant BMP PD‐GF CPLXs are targeted to GAGs remained so far unknown.

Our results suggest a new general mechanism of how BMP GFs and CPLXs are targeted and concentrated in the ECM in the bioactive or latent state by GAGs such as heparin/HS (Figure 13). Thereby, BMP CPLXs specifically interact in bioactive open V‐shape conformation with GAGs via positively charged stretches at the tip of the PD arm region (Figures 4A, 11A). These heparin/HS‐binding stretches appear conserved between the in‐solution bioactive activin, BMP‐7, and BMP‐9 CPLXs (Figure 10A, Figure S6A) and the latent GDF‐8 CPLX (Figure S6A) that interacts via its PD with the side‐chains of the HSPG perlecan. ³⁴ Structural conservation is also present since the heparin‐binding residues of activin, BMP‐7, BMP‐9, and GDF‐8 CPLXs reside within a motif at the PD arm that resembles the VEGF heparin‐binding domain and is composed of one alpha‐helix and three beta‐sheets creating a binding pocket that heparin chains can penetrate within the PD (Figures 5, 11, Figure S6B). Interestingly, our findings demonstrate that spatial concentration of BMP‐7 and ‐9 CPLXs to heparin does not compromise GF bioactivity (Figure 2B, Figure S9). This suggests that PD displacement upon BMP type II receptor binding to the GF as shown for soluble CPLXs of BMP‐7, ‐9, ‐10, and anti‐Müllerian hormone (AMH) ³⁶ , ³⁷ , ⁶⁴ , ⁶⁵ also occurs while being immobilized to heparin. In addition, the heparin‐binding site within the PD arm region of the BMP‐7 CPLX seems to stabilize the bioactive open V‐shape conformation and thereby renders the fibrillin‐1 binding site less accessible (Figure 8B). In contrast, in the closed‐ring BMP‐7 CPLX conformation ⁴¹ access to heparin chains is restricted by maintaining the heparin‐binding sites towards the interior of the structure (Figure 8A and Figure 13). Interestingly, molecular modeling approaches also suggested a closed‐ring shape conformation for BMP‐9 CPLX in which the PD limits receptor access to the BMP‐9 GF and the conserved BMP‐9 PD heparin‐binding sites are inaccessible. ³⁵

FIGURE 13.

Model of BMP targeting to the ECM in bioactive and latent conformation. Free BMP GFs such as BMP‐2 may be targeted and concentrated in bioactive form by GAGs. Thereby, heparin/HS binding sites within BMP‐2 GF (marked in magenta) are distant from BMP type I receptor binding sites (marked in green). However, when BMP‐7 GF is targeted to GAGs via the characterized binding sites (marked in magenta) access of BMP type II receptor binding sites is blocked within the GF (marked in green). In addition, the interaction of specific binding sites of the BMP‐7 PD to heparin/HS (marked in magenta) allows the presentation of the BMP‐7 CPLXs to cells in open, bioactive V‐shape conformation. In contrast, targeting of specific BMP‐7 PD binding sites (marked in red) to the FUN domain of fibrillin‐1 induces a conformational change into a closed, latent ring shape that inadvertently hides the heparin/HS binding sites to the interior of the latent ring. PD residues: blue, GF residues: orange.

ELISA‐style interaction studies with immobilized heparin and soluble BMP‐7 CPLX revealed a high molecular affinity (apparent K_D = 8 nM) (Figure 1A). A similar binding constant was obtained when soluble fibrillin‐1 was flown over immobilized BMP‐7 PD in SPR experiments (K_D = 10 nM). ³³ In addition, we directly compared the affinities of soluble BMP‐7 CPLX with immobilized fibrillin‐1 and heparin which showed very similar K_Ds of around 7 nM for both interactions (Figure S7). This suggests that both CPLX conformations may be equally well stabilized by FMF or GAGs when respective binding sites within the BMP‐7 PD are exposed. These observed binding properties may be crucial for targeting of BMPs to different cellular microenvironments where a different state of bioactivity is required. For instance, heparin interferes with fibrillin‐1 fiber formation ⁶⁶ which may represent an important mechanism to control the assembly form of fibrillin‐1 in basement membranes (BMs). Fibrillin microfibrils (FMF) insert into BMs, but although being also localized within BMs, fibrillin‐1 does not maintain a ~56 nm periodicity characteristic of FMF there. ⁶⁷ One possibility for this observation is that heparan sulfate proteoglycans such as perlecan interact with fibrillin‐1 at BM zones ⁶⁸ and prevent FMF assembly within BMs via their GAG side chains. Therefore, it can be assumed that BMPs are predominantly targeted to BMs in bioactive conformation via GAGs while at the same time fibrillin‐1‐targeting sites may be compromised to sequester BMPs. This hypothesis is in line with observations in the skin where BMPs are known to be active within specialized ECM microenvironments such as stem‐cell niches, for example, in BMs engulfing hair follicles, ⁶⁹ , ⁷⁰ , ⁷¹ while they are also found to be targeted as PD‐GF CPLXs to fibrillin fibers in the dermis in the inactive sequestered state. ³²

It is an important question how BMP GF activity is regulated by the ECM in the absence of the PD. Such mechanisms may be relevant for GFs that either do not well associate with their cognate PDs or that were liberated from ECM‐stored pools via proteolytic degradation of their PD. ³⁸ , ³⁹ , ⁴² BMP GF binding to GAGs has been studied for years, ²⁷ , ²⁸ , ³¹ but how these interactions modulate GF bioactivity is not fully understood. Our finding that immobilization of BMP‐2 GF to heparin does not negatively impact its bioactivity (Figure 2C) is in line with previous studies. ⁷² , ⁷³ Our molecular docking results showed in agreement with previous reports ³¹ that heparin interacts with the N‐terminal region of BMP‐2 and therefore does not block access to GF residues that participate in BMPRI receptor binding ⁶⁰ (Figure S3B). A similar binding mechanism can be assumed for BMP‐4 GF which belongs to the same TGF‐β subgroup. In contrast, utilizing the heparin/HS‐binding site of BMP‐7 GF, which is conserved in members of the same subgroup (also including BMP‐5 and ‐6) ³¹ confers latency by masking crucial residues known to be required for ActRII receptor engagement ⁵⁴ (Figure S3A). Our binding data also showed that GFs of the BMP‐9 and ‐10 subgroup do not interact with heparin, which may be due to a conserved negative charge insertion in BMP‐9 or ‐10 GF sequences (E⁴¹⁹ or E⁴¹⁴, respectively) thereby replacing R⁴²¹ found in the heparin‐binding site of BMP‐7 GF (Figure 10A). Since also the BMP‐10 PD and the processed BMP‐10 CPLX were found not to interact with heparin (Figure 10B,D), it can be assumed that only BMP‐9 CPLX can be concentrated to heparin/HS via the PD in bioactive conformation. This finding is highly relevant for a better understanding of diseases implicating the involvement of dysregulated BMP‐9 and ‐10 signaling via ALK‐1 such as hereditary hemorrhagic telangiectasia (HHT). In the current view, BMP‐9 and ‐10 CPLXs are produced by the liver and circulate with the blood to target endothelial cells of the vasculature. ⁷⁴ , ⁷⁵ Our findings suggest that BMP‐9 can be preferentially concentrated in the endothelial glycocalyx, which may potentiate its signaling capability via prolonged contact with co‐receptors such as endoglin. ⁷⁶ This could also explain why BMP‐9 zebrafish mutants and knockout mice are viable and demonstrate no vascularization defects, ⁷⁷ , ⁷⁸ whereas BMP‐10 zebrafish mutants and knockout mice show defects in vascularization and a severe HHT phenotype. ⁷⁸ , ⁷⁹

In the present study, it was possible to identify the heparin‐binding sites of the BMP‐7 and BMP‐9 CPLXs by utilizing chain‐by‐chain sequence alignments of the VEGF heparin‐binding domain to the respective BMP PDs (Figure 5B, 11B, Figure S10A,B). However, BMP‐7 GF sequence alignments chain‐by‐chain to the BMP‐7 PD failed to map the BMP‐7 CPLX heparin‐binding site within the BMP‐7 PD arm region. This is because the BMP‐7 PD does not contain the typical heparin‐binding consensus motifs XBBXBX and XBBBXXBX (B: basic residue, X: uncharged residue) identified in BMP GFs, ³¹ further suggesting different interaction interfaces of heparin/HS with BMP‐7 GF or BMP‐7 PD. Further, the VEGF alignment method was not only able to predict correctly the heparin/HS‐binding site within the BMP‐7 and BMP‐9 PD arms but also within the activin PD arm of the activin CPLX despite the fact that part of it was deleted for activin CPLX crystal formation ⁵² (Figure 5A). This suggests that the identified motif composed of one alpha‐helix and three beta‐sheets within the PD arm region participating in heparin binding of BMP CPLXs is structurally conserved.

The VEGF affinity to heparin was reported to have a K_D of 80 nM ⁸⁰ with the heparin‐binding domain of VEGF (VEGF₅₅, PDB: 2VGH) ⁸¹ that was used to predict the PD heparin‐binding sites having even a lower affinity to heparin (K_D = 197 nM). ⁸⁰ In contrast, the affinities of BMP‐7 GF and CPLX to heparin measured by us were significantly higher (4–8 nM). A reason for this could be the number of charged residues involved in the interaction interface. While it was shown that VEGF binds to heparin via three charged residues, ⁸² five charged residues in the BMP‐7 PD were found to be crucial to mediate interactions with heparin. A similar requirement of multiple basic residues for proper coordination of the interaction with heparin has already been described for other heparin‐binding growth factors, cytokines, and chemokines. ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ , ⁸⁶ In agreement with this, blunting charge of only two residues did not result in an overall reduction of BMP‐7 PD affinity to heparin (Figure S5), suggesting that specific deletion of residues at different sites of the heparin‐binding interface can be fully compensated by the other contributing residues. Based on this data, we conclude that all identified residues participate with equal strength to the overall affinity of the BMP‐7 PD to heparin. Another important contribution to the strength of the interaction could be the binding stoichiometry. While VEGF appears to bind with a 1:1 stoichiometry (VEGF: heparin), ⁸⁰ our EM results showed that BMP CPLXs can bind heparin with a 1:2 stoichiometry (BMP CPLX: heparin chains) (Figures 4, 11). The binding constant of BMP‐7 GF was previously determined to be 40 nM which indicates a considerably lower affinity to our measurements (apparent K_D = 4 nM). However, a reason for this could be the different detection methods applied. Billings et al. used biotinylated BMP GFs in a solid phase ligand binding assay and detected them after incubation with immobilized HS via a neutravidin‐HRP conjugate (NA‐HRP). ³¹ In contrast, we used a polyclonal anti‐BMP‐7 GF antibody in a similar ELISA‐style binding assay as a detector of bound BMP‐7 GF to coated heparin/HS. As biotinylation eliminates positive charges of lysyl residues, it is likely to negatively affect the binding of BMP‐7 GF to HS.

Our GAG interaction screen with PDs from different subgroup representatives of the TGF‐β superfamily suggests that also other PD‐GF CPLXs may be targeted to specific GAGs in different tissues via specific PD‐binding sites (Figure 9). These binding sites could constitute different GAG‐binding interfaces depending on the TGF‐β superfamily ligand examined. Thereby, different BMP PDs may have different affinities to different GAGs depending on the degree of the negative charge, length, and structure of the respective GAG chains. For instance, while the TGF‐β‐1 PD only showed considerable affinity to the densely sulfated heparin, the BMP‐9 PD showed equally strong binding to heparin, CS, but also to HA which is not sulfated (Figure 9). This may be due to the presence of negatively charged acetyl groups within the HA disaccharide. For instance, it was shown that BMP‐2 GF binding to heparin is abolished when heparin was desulfated, however, when de‐N‐sulfated heparin was acetylated at free amine groups, BMP‐2 GF binding was restored to 51%. ⁸⁷ These results suggest that introducing negatively charged acetyl groups is sufficient for BMP GF and PD targeting. In addition, it was reported that the sulfation pattern (2‐O‐, 6‐O‐ or N‐) plays a crucial role in GF targeting to heparin/HS. Thereby, interaction studies with desulfated heparin forms revealed that 6‐O‐ and N‐sulfations have the most significant contribution in BMP‐2 and BMP‐7 GFs binding to heparin. ⁸⁷ , ⁸⁸ To add a further layer of complexity, the length of the GAG chain can also modulate BMP targeting with a minimum of 10 heparin disaccharides being essential for BMP‐2 GF heparin binding. ⁸⁷ This is consistent with our experimental and in silico determination of a heparin chain length requirement for sufficient targeting of BMP‐7 GF, BMP‐7 CPLX, or BMP‐9 CPLX of at least 10 disaccharide units (Figure 1B, and Figure S11).

However, since purified HS is not readily available at consistent qualities with known sulfation status from commercial sources the comparability of our results to those of other studies is difficult. However, our quality control analysis shows that the HS preparation from E18 mouse embryos used in our interaction studies is endowed with all relevant sulfations (Table S1). Further, our interaction data show no difference between the binding of heparin and HS to BMP‐7 GF or CPLX despite the fact that HS is slightly less sulfated than heparin (Figure 1A). We, therefore, trust that using heparin for most of our experiments sufficiently mimics physiologically sulfated HS in tissues. Considering negative charge, length and structure of GAGs are important since, in each specific cellular microenvironment, differences in composition and arrangement of GAGs may target exquisite sets of growth factors in close proximity to cell surface receptors of specialized tissue‐resident cells. This is especially important in developmental or regenerative processes that are regulated by controlled growth factor action. For instance, the BMP‐7 PD mutation L198P was reported to cause unilateral microphthalmia, optic disc, and chorioretinal coloboma in a patient. ⁸⁹ L198 resides between the heparin‐binding stretches ¹⁷⁸KDYIRER¹⁸⁴ and ²⁰³GRESDLFLLDSR²¹⁴ and its conversion to proline may potentially alter the PD conformation and therefore the interaction interface with heparin/HS GAGs. Therefore, future studies are required to investigate how targeting to different GAGs impacts the bioactivity of free BMP GFs versus CPLXs. This knowledge will be also extremely useful for the future development of GAG‐based biomaterials that have emerged as attractive candidates for growth factor delivery in tissue engineering applications. ⁹⁰

Recently, we suggested that MMP‐mediated PD cleavage of FMF‐targeted latent BMP CPLXs represents a release mechanism of bioactive GFs from ECM‐stored BMP pools. ⁴² Freed BMP GFs could then be locally concentrated by specific, HS‐enriched extracellular microenvironments to either promote or inhibit BMP signaling (Figure 13). In addition, we previously showed that MMP‐13 cleavage of the open V‐shape BMP‐7 CPLX results in partial CPLX unfolding characterized by a four subparticle appearance. ⁴² Such MMP‐mediated unfolding may compromise the bioactivity of partially cleaved CPLXs as soon as they are targeted to heparin/HS. This may be plausible since the characterized heparin‐binding sites within the C‐terminus of BMP‐7 or BMP‐9 PDs (Figure 10A, Figure S6A) do not overlap with the conserved MMP cleavage site within its N‐terminal regions. ⁴² In addition, it is possible that MMP cleavage of heparin‐bound BMP CPLXs may function as a de‐activation mechanism of GAG‐targeted, bioactive BMP CPLXs.

Our study provides novel insight into how BMP PDs integrate the bioavailability of their cognate GFs by targeting exquisite sets of PD‐binding sites to ECM proteins or GAGs and thereby allowing targeting and spatial concentration of BMP CPLXs to extracellular microenvironments in bioactive or latent conformation. This information has a high significance for therapeutic interventions that currently rely on optimized BMP‐based treatments in the context of non‐healing bone fractures, osteoporosis, and connective tissue disease.

AUTHOR CONTRIBUTIONS

Chara E. S. Spanou and Gerhard Sengle conceived the study. Chara E. S. Spanou, Alexander P. Wohl, Sandra Doherr, Annkatrin Correns, Niklas Sonntag, Steffen Lütke, Matthias Mörgelin, and Jan M. Gebauer performed research and analyzed data. Thomas Imhof, Kay Grobe, and Manuel Koch generated essential reagents. Jan M. Gebauer and Ulrich Baumann provided critical advice for molecular docking and modeling approaches. Chara E. S. Spanou and Gerhard Sengle wrote and edited the manuscript. Gerhard Sengle acquired funding for this study.

DISCLOSURES

The authors declare that they have no conflict of interest.

Supporting information

Appendix S1. Supporting Information

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in the results, methods, and/or supplementary material of this article.