Expression and purification of denatured, monomeric MB109
To express the mature domain of hBMP-9, a synthetic, codon-optimized gene, encoding Ser320–Arg429 of NCBI Gene ID:2658 was cloned into a pET21a vector behind T7 promoter (Figure 1A). This recombinant gene product, which we refer to as MB109, contains Met as the N-terminus followed by the coding region of the mature domain of hBMP9. The expression plasmid was transformed into BL21 E. coli cells and the transformants were cultured in LB-broth under aerobic condition in a shaking incubator. After IPTG induction for 20 hours at 37°C, the cells were lysed by microfluidization and the expression of the target protein was analyzed by SDS-PAGE. In the whole cell lysate, MB109 was present in the non-soluble fraction and could be isolated by centrifugation, indicating the formation of inclusion bodies (Figure 1B, lane 5). About 150 mg inclusion bodies were isolated from one liter of cell culture.
After washing three times, the inclusion bodies remained associated with some contaminant proteins (Figure 1B, lane 6). To further improve the purity of the MB109 protein prior to refolding screen, the isolated inclusion bodies were solubilized in reduced and denatured condition and subjected to chromatographic purification by size exclusion chromatography (SEC) using a HiLoad Superdex™ 75 or 200 column (GE Healthcare). We found that even though these two types of gel filtration columns have different separation ranges of molecular size, they gave similar results in terms of final purity and yield of the denatured monomeric MB109 protein. In the Superdex™ 200 column, the solubilized inclusion bodies were eluted as three major peaks at around 46, 72 and 83 mL, with the target MB109 protein present in the last peak (Figure 1C, gray bar). In the Superdex™ 75 column, the solubilized inclusion bodies were eluted as two, well-separated peaks, with the target MB109 protein present in the second peak (Figure 2D, gray bar). The fractions containing the monomeric MB109 protein were pooled and concentrated to 20 mg/mL for refolding. About 50 mg of monomeric MB109 was purified from 100 mg of solubilized inclusion bodies using either column.
Effect of salt concentration on the refolding of MB109
The standard refolding conditions, derived from early studies to readily refold DPP and other BMPs, contain a minimal set of six components: detergent (CHAPS), salt (NaCl), chelator (EDTA), redox agents (reduced and oxidized glutathiones, GSH and GSSG, respectively) and buffer (Tris-HCl) [18, 20, 23–26]. To determine which variables are critical for refolding MB109, we first varied the NaCl concentration from 0 to 4 M and fixed the other refolding variables at their commonly used values (Figure 2A, middle panel). After 7 days incubation at 4°C, visible aggregates were spun down and the supernatants were concentrated to directly visualize the refolding results in non-reduced SDS-PAGE. In the absence of NaCl, most protein aggregated in the refolding solution and thus limited protein remained in the supernatant after centrifugation (Figure 2A, top panel, lane 1). In the presence of NaCl, MB109 refolded into monomer, dimer and higher order multimers (Figure 2A, top panel, lanes 2-6).
In the top panel of Figure 2A, there are two well-separated bands in Lanes 2-6 at around 20–30 kDa, which have a size close to the theoretic molecular weight of dimeric MB109 (24.4 kDa). The protein of the bottom band (black arrow) was purified and its bioactivity was subsequently confirmed (see below). We refer to it as “functional dimer” and we used it as a reference of correctly refolded MB109 on SDS-PAGE. The upper band (gray arrow), which is not bioactive after being purified (see below), is referred to as “chemical dimer.” The densitometry of the functional dimer bands in the SDS-PAGE image (Figure 2A, bottom panel) revealed that the optimal refolding salt concentration for MB109 was around 1.0–2.0 M. At salt concentrations increasingly above 2 M, the refolding efficiency was reduced gradually.
Effect of pH on the refolding of MB109
To analyze the pH effect on refolding MB109, the salt concentration was fixed at 2 M and the buffer pH was varied between 7.0 and 10.0 at a 0.5 interval. As shown in the upper panel of Figure 2B, the functional dimer band was only observed between pH 7.5 and 9.5 (lanes 2-6). The chemical dimer, as well as monomer and multimer, were present in all tested pH conditions (lanes 1-7). The refolding efficiency of the functional dimer had sharp pH dependency. The optimal refolding pH was found to be between 8.0 and 8.5, and the refolding efficiency reduced drastically outside of this range (Figure 2B, bottom panel).
Effect of detergent concentration on the refolding of MB109
In the standard refolding conditions, 1.8–2% CHAPS had been used to refold DPP and other BMPs [18, 20, 23–26]. To test the effect of CHAPS concentration on refolding MB109, 0 to 4% CHAPS was used in the refolding solution. In the absence of the detergent, MB109 formed visible aggregates in the refolding solution after 7 days of incubation, and thus little protein remained in the supernatant after centrifugation (Figure 2C, top panel, lane 1). At a CHAPS concentration above 0.5%, the functional dimer could be refolded; the refolding efficiency was positively correlated with the amount of CHAPS in the refolding solution (Figure 2C, bottom panel). When compared to the commonly used 2% CHAPS concentration, the yield of the functional dimer increased about 37% and 58% in the presence of 3% and 4% CHAPS, respectively.
Effect of redox condition on the refolding of MB109
Because a bioactive BMP molecule contains several disulfide bonds, reduced and oxidized glutathiones (GSH and GSSG, respectively) had been used as a redox system in refolding solution to allow the formation and reshuffling of disulfide bonds. To identify the optimal redox condition, millimolar ratios between 10:1 and 1:10 of GSH and GSSG were tested (Figure 2D). Interestingly, the refolding efficiency of the functional dimer depended mostly upon the amount of GSSG, but not GSH, in the refolding solution. In other words, the more oxidizing “power” (GSSG) in the solution, the functional dimer was refolded less efficiently (Figure 2D, conditions 4-7). In contrast, increasing the reducing “power” (GSH) in the refolding solution did not significantly affect the refolding efficiency (Figure 2D, conditions 1-4). The maximal refolding efficiency was observed at the combination of 2 mM GSH and 1 mM GSSG.
Effect of protein concentration on the refolding of MB109
In all of the refolding conditions tested above, a chemical dimer and higher order multimers of MB109 were prone to form over the functional dimer. This result suggests that the protein concentration (0.2 mg/mL) used for the refolding tests may be too high to form the functional dimer and instead cause the non-specific multimerization observed. To understand the effect of protein concentration on the refolding yield, protein concentrations between 0.05 and 0.4 mg/mL were tested (Figure 2E). Indeed, the refolding yield was reversely correlated with the protein concentration in the refolding solution. At 0.05 mg/mL, the tendency to form higher order multimers was significantly reduced (Figure 2E, top panel, lane 1), and the yield of the functional dimer increased about 55% as compared to the solution refolded at 0.2 mg/mL (bottom panel).
Effect of refolding duration on the refolding of MB109
In addition to the chemical variables tested above, the refolding time course was analyzed for up to 11 days at 4°C to identify the time required to achieve maximal refolding yield. As shown in Figure 2F, the functional dimer band could be seen to after 2 days and the yield plateaued after 9 days of incubation.
Effect of secondary additive on the refolding of MB109
Some chemical additives are known to be effective aids for in vitro protein refolding [27, 28]. To determine whether the refolding yield could be further improved, we tested the commonly used aggregation suppressors and denaturants, including L-arginine, L-proline, glycerol, urea, and guanidine. As shown in Figure 2G, addition of 0.5 M L-arginine, 0.5 M L-proline, 1.5 M urea, and 1 M guanidine significantly reduced the efficiency of refolding the functional dimer, whereas the addition of 10% glycerol provided a minimal increase on the refolding yield.
Effect of host cell contaminants on the refolding of MB109
As shown in the lane 6 of Figure 1B, the isolated inclusion bodies of MB109 contain a visible amount of host cell contaminants on SDS-PAGE. Although these contaminants could be effectively removed by size exclusion chromatography prior to refolding, it would be cost-effective if this purification step were omitted. Therefore, the refolding efficiency was tested in the presence of the host cell contaminants by using the isolated inclusion bodies in the refolding conditions outlined above (variation of pH and salt and CHAPS concentrations). As shown in the upper panels of Figure 3, the functional dimer (black arrows) was refolded in the presence of host cell contaminants. Similar to the size-exclusion purified inclusion bodies, the non-purified inclusion bodies were refolded into functional dimer with an optimal pH of 8.0–8.5 (Figure 3A), an optimal NaCl concentration of 1–2 M (Figure 3B), and a positive correlation with increasing CHAPS detergent concentration, plateauing around 3% (Figure 3C).
Purification of the refolded MB109
In all of the refolding conditions analyzed above, the denatured MB109 was refolded into not only the functional dimer, but also other variants, including the chemical dimer, monomer and multimer (Figure 2). These variants were considered as contaminants because of their lack of bioactivity, and thus we attempted to separate the variants. To separate the functional dimer from the contaminants, we found that the majority of the contaminants could be effectively separated by precipitation when the pH of the refolding mixture was directly titrated to 3.2. As shown in Figure 4A, the majority of contaminants formed water insoluble aggregates at pH 4.2 and 3.2 and could be removed by centrifugation (lanes 4 and 6, respectively), while the functional dimer remained in the supernatant (lane 3 and 5). At pH 2.2, however, the functional dimer started to form water insoluble aggregates as well (Figure 4A lane 8, black arrow).
After acidic fractionation, small amounts of chemical dimer, monomer and multimer remained in the supernatant (Figure 4A lane 5). To further purify the functional dimer, size exclusion chromatography with Superdex 75 and Superdex 200 connected in series was used. As shown in Figure 4B, the multimer and chemical dimer were eluted at 113–128 mL, before the function dimer was eluted at 128–137 mL (gray box). However, some of the monomeric variants were co-eluted with the functional dimer (Figure 4B, bottom panel, lane 6–8).
To remove the remaining monomeric variants, SP Sepharose Fast Flow cation exchange was found to be effective. As shown in Figure 4C, the majority of the monomeric contaminants were washed out around 10.5 mS/cm and the functional dimer was eluted around 13.4 mS/cm. After this step, the purity of the purified functional dimer was greater than 95%, as determined by non-reduced and reduced SDS-PAGE (Figure 4D). The final yield of the purified MB109 was 7.8 mg per 100 mg of SEC purified inclusion bodies under the optimal refolding conditions.
Bioactivity of refolded MB109
The bioactivity of the purified MB109 was first tested by examining its ability to stimulate the Smad1 signaling pathway in an established Smad1-dependent luciferase reporter system in mouse myoblast C2C12 cells (Figure 5A). A CHO-derived hBMP-9 and an E. coli-derived hBMP-2 were used as positive controls. As shown in Figure 5A, the purified MB109 was able to induce dose-dependent Smad1 signaling response with an EC50 of 0.61 ng/mL (25 pM, black circles), which was similar to that induced by CHO-derived hBMP-9 (EC50 of 0.92 ng/mL, 38 pM, gray circles). This is about a 43-fold higher response than that induced by E. coli-derived hBMP-2 (EC50 of 28.1 ng/mL, 1.08 nM, black squares).
To determine receptor binding specificity, purified MB109 was subjected to surface plasmon resonance analysis to determine its binding affinities to immobilized extracellular domains (ECDs) of Type I and Type II receptors, including ALK1, ActRIa (ALK2), ActRIb (ALK4), BMPRIa (ALK3), BMPRIb (ALK6), TGFβRI (ALK5), ALK7, ActRIIa, ActRIIb, BMPRII, TGF-βRII and MISRII. Among these receptor ECDs, MB109 had strong binding affinity only to ALK1, ActRIIb and BMPRII, while the binding to the other receptor ECDs were either very transient (to ActRIIa) or non-detectable (Figure 5B).