We investigated a family with a brachydactyly type A2 and identified a heterozygous arginine to glutamine (R380Q) substitution in the growth/differentiation factor 5 (GDF5) in all affected individuals. The observed mutation is located at the processing site of the protein, at which the GDF5 precursor is thought to be cleaved releasing the mature molecule from the prodomain. In order to test the effect of the mutation, we generated the GDF5-R380Q mutant and a cleavage-resistant proGDF5 mutant (R380A/R381A) in vitro. Both mutants were secreted from chicken micromass cultures, but showed diminished biological activity. Western blot analyses showed that wt GDF5 was processed by the chicken micromass cells, whereas the mutants were not, indicating that the mutations interfere with processing and that this leads to a strong reduction of biological activity. To test the requirements for GDF5 processing in vitro we produced recombinant human (rh) proGDF5 wild-type protein in Escherichia coli. The results show that unprocessed (rh) proGDF5 is virtually inactive but can be proteolytically activated by different enzymes such as trypsin, furin, and MMP3. (rh) proGDF5 could thus be used as a locally administered depot form with retarded release of activity. In contrast to mature rhGDF5, (rh) proGDF5 shows a high solubility at physiological pH, a characteristic that might be useful for therapeutic applications.
Transforming growth factor β (TGFβ) superfamily proteins such as bone morphogenetic proteins (BMPs) promote cellular proliferation, differentiation and regeneration. Growth/differentiation factor 5 (GDF5, also known as CDMP1 or BMP14) is a protein belonging to the GDF-subgroup of BMPs with a characteristic developmental expression pattern in precartilage condensations and in the developing cartilaginous joints (1). The main effects of GDF5 appear to be the induction of cartilage and the patterning of joints, as shown by various in vitro and in vivo experiments (2–4). The ability of GDF5 to induce cartilage and bone has made GDF5 an attractive factor for medical applications such as spine fusions (3) and disc regeneration (5,6). Furthermore, it has been shown that GDF5 has trophic and protective effects on dopaminergic neurons in vitro (7) and in vivo (8). Like other members of the TGFβ protein family, GDF5 is initially synthesized as a large precursor protein (proGDF5) consisting of 501 amino acids including an N-terminal signal peptide of 27 amino acids length (9). Based on data obtained with other BMPs, proGDF5 is thought to undergo proteolytic cleavage at a cluster of basic residues [RX(K/R)R] at position 381 (10–12), resulting in the release of the C-terminal mature protein from the N-terminal prodomain. The released mature protein is active indicating that the bioactivity and signaling range of BMPs is regulated by proteolytic cleavage. The prodomain of GDF5 consists of 354 amino acids with a single N-linked glycosylation site at position 189 and comprises a potential heparin-binding site (239LRILRKKP246) (13). The mature GDF5 protein consists of 120 amino acids, which form homo-, or possibly also heterodimers with other BMPs that are linked via a disulfide-bridge. For BMP4 it was shown that endoproteolytic cleavage of the precursor primarily takes place at the multibasic motif RSKR by specific members of the subtilisin-like proprotein convertase (SPC) family of endoproteases. In mammals seven members of this family have been characterized that show distinct substrate specificity and expression patterns (14). The prototype of the SPC family, furin (also known as SPC1) as well as PACE4 (also known as SPC4) and PC6 (also known as SPC6) do also recognize the consensus sequence RXR/KR but can additionally cleave after the minimal sequence RXXR. This principle applies also for GDF5 and sequence homologies with the BMPs, and the presence of an SPC recognition site suggests a similar mechanism (15). The presence of at least some of these SPCs in the extracellular space, like SPC4 and SPC6, also suggests that the activity of secreted proBMP as well as proGDF5 might be regulated in situ via proteolytic cleavage (16).
Other mechanisms that regulate BMP activity in the extracellular space include binding to antagonistic proteins such as Noggin (NOG) and Chordin that block activation of cell surface receptors. Proteases such as Tolloid can cleave Chordin to liberate BMP4 (17). Furthermore, several BMPs were shown to bind to surface heparin sulfate proteoglycans, a mechanism that can promote or restrict activity (18). Upon binding of BMPs/GDFs to their cognate BMP-receptors SMADs (SMAD1,5,8) are activated and translocated into the nucleus where they bind to specific sequences of the DNA to activate target genes. BMP signaling is negatively regulated by inhibitory SMADs (SMAD6,7) which block transmission of signals from the membrane to the nucleus (19).
The importance of GDF5 is underlined by a number of clinical syndromes associated with mutations in GDF5 or its receptor BMP receptor 1B (BMPR1B) (20,21). Heterozygous loss of function mutations in GDF5 cause brachydactyly type C (BDC), a condition characterized by a specific pattern of shortened digits and supernumerary ossifications of phalanges. Homozygous loss-of-function mutations in GDF5 result in the acromesomelic chondrodysplasias of the Grebe, Hunter-Thompson, or DuPan type. A specific point mutation in GDF5 (L441P) was shown to interfere with receptor binding and consequently has reduced biological activity. This is associated with brachydactyly type A2 (BDA2) (22), a condition that can also be caused by dominant negative mutations in the gene encoding for BMPR1B (23,24). In contrast, activating mutations in GDF5 (R438L) result in symphalangism (fusion of finger joints) which can also be caused by mutations in the BMP-inhibitor NOG (22,25).
In the family described here, a genetic analysis revealed a mutation in GDF5 at position 380 (R380Q), located at the GDF5 processing site. To test the hypothesis that the skeletal defect is due to abnormal cleavage at the proGDF5 precursor maturation site we generated GDF5 mutants and tested them in vitro. To further investigate the processing requirements for proGDF5, we produced recombinant proGDF5 in Escherichia coli(E. coli). We found that unprocessed (rh) proGDF5 is virtually biologically inactive, and, in contrast to mature GDF5, soluble at neutral pH. (rh) ProGDF5 can be proteolytically cleaved by intracellular as well as extracellular proteases thus generating biological active mature GDF5. Our results indicate that full length (rh) proGDF5, like mature rhGDF5, has the potential to be used as a drug in regenerative medicine.
Brachydactyly caused by a mutation in the SPC cleavage site of GDF5
We identified a family with a brachydactyly phenotype that showed a dominant pattern of inheritance (Supplementary Material, Fig. S1). Hands and feet were clinically examined in 28 family members of whom 14 were affected. Radiographs were available from five affected individuals and metacarpophalangeal profiling (Fig. 1) was performed as described previously (26). In four of five radiographically evaluated cases the second mesophalanx was shortened by more than 2 SD (Fig. 1). The metacarpophalangeal profile (MCPP) demonstrated that the second mesophalanx was constantly (five out of five subjects) shorter than mesophalanges 3, 4 and 5 which is in accordance with a diagnosis of BDA2. Other occasional findings were a relatively short first metacarpal (V:2) or first proximal phalanx (V:9). The characteristic MCPP was also evident in VI:6 who otherwise had bones of normal length, suggesting that this analysis is the most certain method for determining carrier status. Hands were more commonly affected than the feet. The phenotype was variable with some affected showing an almost normal pattern while others had additional features such as mild hypoplasia of middle phalanges of fingers 3 and 4 indicating overlap with BDA1 and BDC (Fig. 1). One affected family member in addition had a congenital heart defect (type unknown) which is regarded as unrelated to the identified mutation.
BDA2 is known to be associated with mutations in BMPR1B. However, no mutations in BMPR1B were identified. As GDF5 mutations are known to cause both BDA2 and BDC, we subsequently focused on this gene. Mutation analysis of GDF5 revealed a missense mutation in all affected at position c.1139 from G to A resulting in an exchange of arginine for glutamine at position 380 (R380Q).
Investigation of GDF5 R380Q in micromass culture
In contrast to other TGFβ family protein members, GDF5 has two superimposed SPC processing sites, consisting of two RXXR motifs (RRKRR, Fig. 2). The R380Q mutation is located at the P2 position of the SPC processing site of GDF5 disrupting one of the two general motifs (Fig. 2). To investigate the functional consequences of the R380Q mutation, we performed site directed mutagenesis of cloned full-length human GDF5 cDNA in order to produce (i) the R380Q mutation and (ii) to inactivate the SPC processing site completely by converting the two arginines at positions 380 and 381 to alanines (R380A/R381A). The mutagenized cDNA clones were verified by sequencing and cloned into the replication competent avian sarcoma (RCAS) retroviral vector system. RCAS retrovirus is avian specific and allows the infection of chick cells to almost 100% (22). We used this system to overexpress wt and mutant GDF5 in mesenchymal stem cells prepared from chick limb buds. These cells spontaneously differentiate into chondrocytes, a process that can be quantified by staining the cultures with Alcian blue. We observed a severe reduction of GDF5 activity in both mutants when compared with wt GDF5 infected cells indicating a severe loss of activity in the R380Q as well as in the R380A/R381A mutant (Fig. 3A and B).
To investigate the consequences of both mutations on prodomain cleavage we performed western blots against GDF5 in the supernatant and cell lysates of micromass cells infected with RCAS expressing wt or mutant GDF5. Using the monoclonal antibody aMP5 against mature GDF5, we detected both proGDF5 and mature GDF5 (Fig. 3C). Under non-reducing conditions proGDF5 migrates at about 100 kDa and mature GDF5 at 25 kDa. In addition intermediate cleavage products were visible at 50–80 kDa. We observed strong staining for full-length GDF5 in the cellular supernatant indicating that GDF5 is efficiently secreted in this experimental setting. In addition to the full-length band we observed a smaller band corresponding to the processed mature GDF5. The latter band was absent in the R380Q and R380A/R381A mutants indicating that both mutations interfere with GDF5 processing. Moreover, these results suggest that disrupting the P2 position appears sufficient to inhibit the processing significantly and this effect is not compensated for by an intact P1 position.
Expression and purification of (rh) proGDF5
In order to investigate the properties and function of proGDF5 in more detail we produced recombinant (rh) proGDF5 protein. Because production of (rh) proGDF5 in eukaryotic expression systems such as CHO cells proved to be inefficient and time consuming (Supplementary Material, Fig. S2) we switched to expression of (rh) proGDF5 in E. coli. After optimization of expression vectors, bacterial strains, promoters as well as codon usage optimization, protein expression could be induced after IPTG induction in the bacterial strains BL21 (DE3) and Rosetta2 (DE3) (Fig. 4A). To initiate protein expression, the N-terminus of proGDF5 was modified by introducing a sequence tag.
The purification process comprises a direct refolding step. Thus, after inclusion body preparation the proteins can be used directly in the refolding without prior column purification. The direct refolding process for (rh) proGDF5 was optimized to generate correctly refolded dimeric protein with efficiency greater than 80% (Fig. 4B). The refolding process was almost completed after 2 days of incubation. Non-refolded protein and remaining E. coli host proteins were removed using a combination of size exclusion and reversed phase chromatography (RP HPLC). After successful refolding, correctly refolded protein migrates under non-reducing conditions at around 100 kDa. The resulting purity of (rh) proGDF5 was greater than 90%, as determined by SDS–PAGE followed by Coomassie staining (Fig. 4C).
Solubility testing of (rh) proGDF5
Mature rhGDF5 has an isoelectric point of 7.6 and is therefore nearly insoluble under physiological conditions (27). To assess whether (rh) proGDF5 shows an improved solubility in aqueous solutions at neutral pH, (rh) proGDF5 was dissolved in buffers ranging from pH 2.0 to 12.0. (rh) proGDF5, unlike its mature counterpart, is soluble at physiological pH in buffers such as PBS (Fig. 5). Solubility of (rh) proGDF5 was best at pH 2; (rh) proGDF5 was nearly insoluble at its predicted isoelectric point at pH 10.
Proteolytic cleavage of (rh) proGDF5
Members of the TGFβ protein family are initially synthesized as large precursor proteins which undergo proteolytic cleavage to release the active part of the protein. In analogy to proBMP4 and proBMP2 (10–12,28), it is thought that proGDF5 is processed in a similar way to mature GDF5. We investigated proteases of different classes, which might be involved in protein maturation. (rh) proGDF5 was digested with furin, MMP3 and trypsin and separated under non-reducing conditions on a 10% PAGE and detected via western blotting.
The subtilisin-like protein convertase furin acts intracellularly in the trans golgi network, but furin can also be posttranslationally modified to produce a smaller soluble form of the enzyme that can be secreted (29). Digestion with furin results in cleavage of (rh) proGDF5, and the release of mature GDF5 (Fig. 6A). A fragment of ∼25 kDa was detected by the monoclonal antibody aMP5, which only detects correctly folded GDF5. N-terminal protein sequencing/MALDI-TOF and analysis revealed that the 25 kDa fragment starts with the sequence 381RAPLA and 382APLA, representing the expected N-terminal starting position of the mature GDF5. Thus, we could demonstrate that recombinantly produced (rh) proGDF5 can be processed by furin generating mature GDF5 from its natural precursor protein at the predicted cleavage site (RXXR). Furthermore, we can show that furin cleaves the superimposed RXXR motif twice within the RRKRR sequence.
Matrix metalloproteases (MMPs) are extracellular proteases that modulate the extracellular matrix (ECM), some of them are co-localized with proBMPs. To investigate if proGDF5 can be processed by MMPs, (rh) proGDF5 was incubated with the matrix metalloprotease 3 (MMP3). Three major fragments were detected on a Coomassie Blue stained polyacrylamide gel. A fragment of ∼25 kDa, migrating on the same level as the mature rhGDF5 control protein, was recognized by the GDF5 specific antibody in western blot analysis (Fig. 6A). This cleavage product was further characterized by MALDI analysis. We expected a GDF5 fragment starting with the mature N-terminal sequence 382APLA at the proGDF5 maturation site. However, the MMP3 cleavage occurred after the hydrophobic cleavage stretch APLA liberating a truncated GDF5 starting at position 386. The protein band of about 40 kDa, which is also recognized by the antibody aMP5, is an intermediate proGDF5 cleavage product.
Previous studies have shown that pro-forms of NGF and BMP2 can be digested in vitro by trypsin to obtain the mature counterpart (28,30). In analogy, trypsin was used to digest proGDF5. After trypsin digestion, a truncated mature GDF5 cleavage product was detected by the GDF5 specific monoclonal antibody aMP5 (Fig. 6A). N-terminal peptide sequencing showed that the N-terminus of the mature GDF5 started at the positions 395NLKA and 398ARCS. Figure 6B shows a schematic of the GDF5 protein and the resulting digestion fragments.
Activation of (rh) proGDF5 by proteolytic digestion
After having demonstrated that (rh) proGDF5 can be cleaved by enzymes of different protease families in vitro we wanted to know if the proteolytic cleavage products of proGDF5 are biologically active or inactive. We used the GDF5 responsive cell line MCHT1/26, a mouse stromal calvarium cell line, to investigate the biological activity of undigested (rh) proGDF5 in comparison with proteolytically cleaved proGDF5. MCHT1/26 cells were stimulated with 0.14 to 44 nM (rh) proGDF5 or equal amounts of (rh) proGDF5 subjected to proteolytic digestion with trypsin, MMP3, or furin and determined their biological activity by an alkaline phosphatase (ALP) assay. The in vitro matured GDF5 proteins released from the precursor molecule by proteolytic digestion exhibited biological activity in a dose-dependent manner (Fig. 6 C–E). Similar results were obtained for all three proteases. In contrast, the unprocessed (rh) proGDF5 showed only little activity at high protein concentrations. This background activity was probably due to a partial proteolysis during the 3 days incubation period on the MCHT1/26 cells (data not shown). The proteases trypsin, furin or MMP-3 itself had no negative influence on the ALP assay.
Endogenous activation of (rh) proGDF5 by primary mesenchymal cells
MMP3 and other extracellular proteases are coexpressed with proBMPs at many sites, indicating that extracellular processing and activation of BMPs is possible. We wanted to test if (rh) proGDF5 could be used as recombinant therapeutic protein and if (rh) proGDF5 can be activated by target cells. We again used the micromass culture system of chicken limb buds as a model for primary mesenchymal cells that differentiate into chondrocytes. We stimulated micromass cultures with different dosages of rhGDF5 versus (rh) proGDF5 in a range between 0.37 and 22.2 nM at day 1 and incubated them for 3 more days without changing the media. Alcian blue staining revealed that (rh) proGDF5 as well as rhGDF5 induced cartilage. rhGDF5 was more potent at low concentration in comparison with proGDF5, but this effect disappears at higher concentrations (Fig. 7A and B).
We identified a so-far undescribed mutation in GDF5 that results in an arginine to glutamine substitution at position 380 (R380Q). This amino acid change is located in direct proximity to the cleavage site at which the prodomain is cleaved from the mature GDF5 molecule. The mutation causes hypoplasia of the mesophalanges in finger 2 in all affected individuals leading to the diagnosis of BDA2. Milder hypoplasia of mesophalanges in fingers 3, 4 and 5 indicate a clinical overlap to BDA1 and BDC. BDA2 is known to be caused by specific mutations in BMPR1B, the cognate receptor for GDF5 that result in diminished transphosphorylation by the type II receptor (BMPR2) and reduced SMAD activation (23,24). Recently, GDF5 was identified as a second locus for BDA2. The mutation at position L441P described to be associated with BDA2 is located in the type I receptor binding pocket resulting in a severely diminished receptor binding and thus reduced activation (22). In contrast, loss of function mutations of GDF5 result in BDC, a hand malformation with a distinguishable pattern of digit involvement (21,31). The phenotype–genotype comparison of these conditions indicates that BDA2 is caused by mutations that specifically diminish GDF5 signaling. The R380Q mutation can thus be hypothesized to result in a reduced GDF5 activation and therefore decreased signaling through BMPR1B or other receptors. The variability in the family described here suggests that other factors influence the phenotype as modifiers. Such factors are likely to alter the cleavage conditions of the mutant protein and thus interfere with the bioavailability of GDF5.
We hypothesized that the skeletal malformation was due to a diminished cleavage event at the GDF5 precursor maturation site. Based on previous studies with BMPs, we assumed that proteolytic cleavage of proGDF5 is a prerequisite to liberate biologically active mature GDF5. We therefore generated a cleavage resistant proGDF5 mutant (R380A/R381A) and compared it with the R380Q mutant and wt GDF5. As demonstrated by micromass cultures, proGDF5 has the potency to induce cartilage. In contrast, the mutants R380A/R381A and R380Q only slightly induce cartilage production. Using western blots we can show that the majority of GDF5 is secreted by the cells as proGDF5. This is in line with the observation that unprocessed proGDF5 was the predominant form in rat brain extracts (32). The mutants R380A/R381A and R380Q are secreted but are not properly processed. Our results indicate that the R380Q mutation results in a reduced GDF5-function caused by diminished proteolytic cleavage which is shown here to be a precondition to generate biologically active GDF5.
To further investigate the processing requirements for the generation of biologically active GDF5, we produced proGDF5 recombinantly. The production of recombinant TGFβ superfamily members is time consuming and expensive, because they are not readily obtainable in biologically active and sufficiently pure form in prokaryotes. A limiting factor is the separation of the active renatured dimeric protein from the inactive monomeric form by size exclusion. In this study, we describe the successful production of (rh) proGDF5 in E. coli. The (rh) proGDF5 obtained was of sufficient purity as demonstrated by western blot analysis. In comparison with the mature rhGDF5 dimer, (rh) proGDF5 shows a completely different pH-dependent solubility profile. Whereas mature rhGDF5 is practically insoluble at neutral pH, (rh) proGDF5 was soluble. It has been suggested previously that one principal function of the BMP-prodomains is to solubilize the growth factor dimer by shielding the hydrophobic surfaces present on the mature molecule (33). A similar function is likely to be applied to the GDF5 prodomain.
As expected from our micromass experiments using the cleavage resistant GDF5 variants, (rh) proGDF5 showed no activity in a cell-based assay. In contrast, wt (rh) proGDF5 was fully active in micromass culture indicating that these cells secrete factors that are able to process proGDF5 in the extracellular space. As a rule, TGFβ superfamily proteins are synthesized as precursor proteins that subsequently undergo proteolytic cleavage in the cell releasing the biologically active C-terminal mature protein parts from the N-terminal prodomain (10,34). Responsible for cleavage of the precursor protein within the cell are proteases such as furin (35). We therefore tested several proteases including the intracellular protease furin as well as the extracellular proteases MMP3 and trypsin, for their ability to activate proGDF5. Here we demonstrate that proGDF5 is biologically active after proteolytic activation with intracellular (furin) and extracellular proteases (MMP3). Western blot analysis and N-terminal protein sequencing of the proGDF5 cleavage products confirmed that proteolysis occurred at or near the GDF5 maturation site.
MMPs are important proteolytic enzymes that modulate and rearrange the ECM. Among the large group of the MMPs, MMP3 is an interesting candidate for proGDF5 processing, because MMP3 as well as precursor molecules of BMP2 and BMP6 are expressed in arthritic synovium in patients with rheumatoid arthritis (36). Furthermore, it has been shown that MMP3 and GDF5 play a role in intervertebral disc cells (6,37,38). GDF5 deposited as a proform in the ECM is likely to become activated in degenerated tissue at sites of inflammation where elevated proteolytic enzyme activity is present (37). We were able to generate a biologically active GDF5 molecule after trypsin incubation of (rh) proGDF5. This GDF5-variant was truncated at the maturation site starting at position 395NLKA and 398ARCS. Previously, we demonstrated that a genetically engineered N-terminally truncated version of the mature GDF5 (GDF5-Cys) missing the first 18 amino acids, has full biological activity in the ALP assay (39). We have shown previously that under reducing conditions mature GDF5 is degraded into 13 trypsin cleavage fragments (27). Here we demonstrate that the mature part of proGDF5 is not degraded after trypsin digestion under non-reducing conditions implicating that the respective trypsin cleavage sites are hidden in the cystine-knot structure.
In contrast to furin and MMP3, trypsin digests proteins after basic amino acids and therefore degrades the complete prodomain of proGDF5 and the N-terminal part of mature GDF5 into small peptide fragments. Only the cystine-knot structure of the mature GDF5 domain is protected against the trypsin digestion. In comparison, furin cleavage liberates the prodomain as a larger intact protein fragment. It was shown for BMP2 that the N-terminus of the mature domain has an affinity to heparin and therefore influences its bioavailability or mobility within the ECM (40–42). On the other hand, the non-covalent association of the proteolytically liberated prodomain to the mature growth factor has been shown for other TGFβ family members such as BMP7 (43), GDF8 (44) and TGFβ1 (45), which influences their bioavailability and/or bioactivity.
Considering the lack of conservation within the prodomains of BMPs/GDFs which is in contrast to the high conservation of the mature domains, different and specific functions of the prodomains are likely. Interestingly, proGDF5 has a heparin binding domain in the prodomain suggesting that extracellular proteases and heparin sulfate proteoglycan in the ECM play a role in the regulation of proGDF5’s biological activity.
Taken together we conclude that proGDF5 is soluble under physiological conditions and likely to function as depot form, which is able to build up a morphogen gradient under physiological pH. After cleavage by extracellular proteases the processed mature GDF5 is less soluble; a property that may restrict its range of biological availability. One potential activator in the extracellular space is MMP3, an enzyme that is also expressed in brain and is upregulated in degenerative disorders. In Alzheimer disease amyloid beta 1–42 induces high levels of MMP3 in the microglia (46). In the process of neuronal degradation, MMP3 is released from apoptotic neurons, which may play a role in degenerative human brain disorders, such as Parkinson disease (47). These findings along with pervious reports that GDF5 has survival-promoting effects in vitro and in vivo (7,8,32,48–51) implicate the use of the soluble GDF5 pro-form for neuronal therapeutical applications. The increased solubility of (rh) proGDF5 at physiological pH suggests that it can be used to formulate pharmaceutical compositions for the therapy of tissue disorders. Further studies are needed to show, if activation of this protein can be achieved predictably in vivo by extracellular proteases.
MATERIALS AND METHODS
DNA was extracted by a standard salt-out method from a peripheral venous blood sample in adults, and from a filter blot or a buccal swap in children. Material could be obtained from 14 clinically affected and 16 unaffected individuals. The study was approved by the Danish ethical committee system and written informed consent was obtained for all participants in accordance with the Helsinki II Declaration. For mutation analysis of GDF5 both exons and exon–intron border regions in the GDF5 gene were sequenced and compared with the NCBI database [GDF5, RefSeq NM_000557 (mRNA) or NP_000548 (protein)]. Primers were designed in the intron regions and amplified by PCR using Platinum Taq (Invitrogen, Carlsbad, CA, USA) or NEB Taq-DNA polymerase (New England Biolabs) as described previously (26). PCR and sequencing was done by standard methods with PCR primers used as sequencing primers. Products were evaluated on an automated capillary sequencer (Applied Biosystems). PCR products were separated by 2% agarose, 1×TBE gel electrophoresis. Primers were removed by treatment with 1U shrimp ALP (USB, Cleveland, OH, USA) and 10U exonuclease I (New England Biolabs) followed by sequencing using the BigDye Terminator Kit (Applied Biosystems, Foster City, CA, USA) and analyzed on an ABI 377 sequencer (Applied Biosystems). The mutation was confirmed by MspA1I (NEB) restriction enzyme digestion of the PCR product generated by the primers 5′-ATGAGATTAAGGCCCGCTCTG-3′ and 5′-GCAGAGTCAATGAAGAGGAT-3′. Eighty normal individuals (160 alleles) served as a control group. Digestion of the PCR product from an unaffected individual resulted in two products of 62 and 317 bp, whereas heterozygous mutation carriers had an additional two products of 12 and 305 bp. The digested PCR products were separated on a 20% acrylamide 1×TBE gel.
Chicken limb bud micromass cultures
Chicken limb bud micromass cultures were prepared as described previously (22). Briefly, fertilized SPF eggs were obtained from Tierzucht Lehmann and incubated for 4.5 days at 37.5°C. Limb buds from HH24 chicken embryos were prepared and digested with 3 mg/ml dispase (Gibco) in HBSS (Lonza) to remove the ectoderm followed by a digestion with digestion-solution: (0.1% collagenase type Ia, 0.1% trypsin, 5% FCS or CS in PBS without Ca/Mg). Cells were plated at a density of 2 × 107 cells/ml at 10 µl drops. 1 µl/culture concentrated viral supernatant of RCAS containing the cds of human wt-GDF5, R380Q or R380A/R381A were added directly to the cells before plating. Cells were grown in DMEM:F12 with 10% FCS; 0.2% chicken serum and Pen/Strep. Medium was replaced every 2 to 3 days. Alcian blue staining of micromass cultures were performed over night in 1% Alcian blue in 1 N HCl after fixation in Kahles Fixative [28.9% (v/v) EtOH; 0.37% formaldehyde; 3.9%(v/v) acetic acid]. To quantify Alcian blue staining the dye was extracted overnight by 6 M Guanidine–HCl and photometric measurement at OD 595. For western blot analysis, growth media were replaced with serum-free media and incubated for 24 h. Media/cultures were precipitated with an equal amount of acetone over night at −20°C. Precipitated proteins were solved in sample buffer (0.1 M Tris–HCl, pH 8.8, 2% SDS, 20% glycerol, 0.01% bromophenolblue). Micromass culture was homogenized in lysis buffer [50 mm HEPES, pH 7.5, 50 mm NaCl, 10 mm EDTA, 10% Glycerol, 1% Triton X-100, complete mini (Roche), PMSF]. Proteins were separated under non-reducing SDS–PAGE followed by western blot analysis as described previously (52).
Cloning, expression and purification of proGDF5
CDS of ProGDF5 without signal peptide was integrated into the protein expression vector pET15b (Novagen) using the restriction sites NdeI and BamHI. The vector codes for an N-terminal histidine tag and a thrombin cleavage site.
The expression of proGDF5 was performed in the E. coli strains BL21 (DE3) and Rosetta2 (DE3) (Novagen). The protein expression was induced with 1 mm isopropyl-1-thio-ß-D-galactopyranoside (IPTG). The proteins were expressed in inclusion bodies. These inclusion bodies were isolated using a homogenization buffer (25 mm Tris–HCl, 10 mm EDTA, pH 7.3) and wash buffer (20 mm Tris–HCl, 5 mm EDTA, pH 8.3) according to standard procedures. The inclusion bodies were solubilized in solubilization buffer (4 M GuHCl, 3 mm DTT, 0.1 M Tris–HCl, pH 8.5) and the direct refolding was performed in a 1:10 dilution with refolding buffer (1 M ariginine–HCl, 5 mm oxidized glutathione (GSSG), 1 mm reduced glutathione (GSH), 0.1 M Tris–HCl, 5 mm EDTA, pH 8.0) at room temperature for 5 days.
Ultrafiltration was conducted to an optimum load for the size exclusion chromatography (SEC) column. Purification was carried out on an SEC-column (GE Healthcare, Munich, Germany; hiLoad 26/60; column material: Superdex 200 prepgrade, column volume 319 ml) flow rate 2.5 ml/min. 6 ml protein from the ultrafiltration procedure was loaded on the column. Protein was eluted in 2 M urea, 100 mm Tris–HCl, 5 mm EDTA, pH 8.0.
An additional purification step was carried out on reversed phase HPLC (GE Healthcare, Munich Germany; column HR16/10; column material Source 15RPC, volume 20 ml), flow rate 3 ml/min, System: Äkta Eplorer 100. Gradient was started with 35% of eluent B (0.1% TFA, 90% CH3N, HPLC H2O) then gradient 35–60%, slope 0.38%/min, then gradient 60–90%, slope 1.5%/min, then 90% two column volumes and finally 35%, two column volumes. The fractions containing the dimerized protein were pooled, lyophilized and stored at −80°C.
Western blotting under reducing and non-reducing conditions
Western blotting was performed as described previously (2). The proteins were separated using 10 or 16% SDS–PAGE. Blotting onto PVDF membranes (Millipore) was carried out in the presence of 10 mm DTT, when the polyclonal antibody chicken B pool was used, because this antibody recognizes only reduced GDF5 molecules bound. The monoclonal aMP5 does only recognize refolded GDF5 under non-reducing conditions. The antibodies were detected with a chemiluminescence detection kit (Applied Biosystems, Foster City, CA, USA).
In vitro processing of (rh) proGDF5
The in vitro digestion of (rh) proGDF5 was performed with the proteases trypsin, furin and MMP-3. Proteolytic cleavage with trypsin was carried out with 3 µg proGDF5 dissolved in 1×PBS incubated with 0.3 µg trypsin (Merck, Darmstadt, Germany) over night at 4°C. Furin cleavage was carried out with 3 µg proGDF5 dissolved in 1×PBS supplemented with 1 mm CaCl2 and incubated with 3U Furin (New England Biolabs, Frankfurt, Germany) at 30°C over night. MMP3 cleavage was carried out with 3 µg proGDF5 dissolved in 1×PBS supplemented with 90 mm NaCl, 30 mm Tris–HCl, pH 7.4, 6 mm CaCl2 and incubated with 0.03 µg MMP3 (Sigma, Taufkirchen, Germany) at 37°C over night. The digestion was controlled by Coomassie stained SDS gels and western blot analysis with antibodies directed against the mature rhGDF5.
MALDI-TOF MS and N-terminal protein sequencing after peptide mapping
After digestion of 20 µg proGDF5 with proteases (MMP3, trypsin and furin), the protein fragments were adjusted with 100 mm HCl to reach final concentration of 10 mm HCl and separated by reverse phase HPLC using a VYDAC C18 column (5 µm, order No. 218TP52 protein and peptide, 2.1 × 250 mm; Grace Vydac) developed with a linear gradient ranging from 25 to 100% acetonitrile in 0.1% trifluoroacetic acid (Pierce, Rockford, IL, USA; sequencing grade) within 130 min. The flow rate was 0.3 ml/min and the wavelength of the UV detector was 214 nm. Peak fractions corresponding to the retention time of wt GDF5 were isolated. Proteolytic peptides were N-terminally sequenced on a protein sequencer (Procise 491 HT, Applied Biosystems) and their masses were determined by MALDI-TOF mass spectrometry.
The N-terminal sequencing was done via the Edman degradation using the automatic amino acid sequencer (Procise 491 HT, Applied Biosystems). The identification was performed with reverse phase HPLC by comparison with the phenylthiohydantoin amino acid standard.
MALDI-TOF was performed using a ‘Microflex’ mass spectrometer from Bruker (Bremen, Germany). The instrument was operated in the positive ion linear mode. 0.5 μl matrix, 0.5 µl sample and 0.5 μl matrix (10 mg α-cyano-4-hydroxy-cinnamic acid in 300 μl 0.1% (v/v) TFA and 700 μl 0.1% (v/v) TFA in acetonitrile) were applied as a ‘sandwich’ on the sample carrier. Each layer was dried before the next layer was added. The peptide masses were determined using external mass calibration with insulin and albumin.
Solutions of 21 µg proGDF5 in 10 mm HCl were lyophilized. To determine the solubility of proGDF5, seven different kinds of buffers with different pH values ranging from pH 2.0 to 12.0 (10 mm HCl, pH 2.0; 0.1 M Na-acetate were adjusted to pH 4.0 and pH 5.0; 1×PBS was adjusted to pH 7.0; 0.1 M Tris–HCl were adjusted to pH 8.0 and pH 9.0; 0.1 M Na-carbonate buffers were adjusted to pH 10, pH 11 and pH 12.0) were used to dissolve rhGDF5 or (rh) proGDF5. Respective buffers were added to lyophilized protein, and samples were vigorously agitated for 10 min, followed by centrifugation at 13 000 rpm. The supernatant was carefully removed and quantitatively analyzed by reverse phase HPLC and peak integration.
Alkaline phosphatase assay
The biological activity of (rh) proGDF5 digested with the proteases furin, trypsin and MMP3 was measured using different concentrations of digested (rh) proGDF5 on MCHT1/26 cells. Mouse stromal MCHT1/26 cells were plated at 4.5 × 103 cells per well in 96-multiwell plates in cell culture medium (alphaMEM, Sigma, Taufkirchen, Germany; 2 mm L-glutamine, Invitrogen, Karlsruhe, Germany; 10% fetal calf serum, Invitrogen, Karlsruhe, Germany). After 24 h, cells were stimulated with 0.14 to 44 nM of digested proGDF5 incubated with the proteases trypsin, furin or MMP-3. As a control, undigested proGDF5 in the respective protease reaction buffer without protease and as second control reaction in the respective protease without proGDF5 was used. After 72 h cells were washed with PBS and extracted with alkaline phosphate buffer 1, containing 1% Nonident-40, 0.1 M glycine pH 9.6, 1 mm MgCl2 and 1 mm ZnCl2. To achieve thorough lysis, cells were incubated 15–18 h at 37°C. Enzyme activity was assayed with 10 mm p-nitrophenylphosphate as a substrate in 0.1 M glycine pH 9.6, 1 mm MgCl2 and 1 mm ZnCl2. After 60 min incubation at 37°C, the absorbance was measured with an automatic microplate reader (Tecan Spectra Rainbow, TECAN, Crailsheim, Germany) at 405 nM under consideration of blank value subtraction.
Supplementary Material is available at HMG Online.
This work was funded by the Deutsche Forschungsgemeinschaft grant LE 1851/1-2 (to K.L.) and supported by BMBF 0312114 (to F.P.) as well as BMBF 0313851A/B (to F.P. and S.M.). Wilhelm Johannsen Centre was established by the National Danish Research Foundation.
We thank Inge Ruths, Angela Edelmann and Petra Schramm for technical assistance.
Conflict of Interest statement: There is no conflict of interest.