Generation and characterization of GYM329, latent myostatin-specific antibody with pH-dependent binding properties
Anti-latent myostatin antibodies were generated in rabbits by alternatively immunizing the animals with recombinant human and mouse latent myostatin to enrich cross-reactive clones. Screening of B cell supernatants from the immunized rabbits through a binding assay identified clones that specifically bind the latent and not the mature myostatin. They were then functionally screened through the Smad3/4-binding elements-driven secreted alkaline phosphatase (SEAP) reporter gene assay. We assessed the candidate antibodies’ inhibitory activity against BMP1-mediated activation of myostatin, and MST1032 was finally selected as the lead antibody based on its strong activity.
The variable domain of MST1032 was humanized and engineered to confer pH-dependent binding essential for the sweeping function through a comprehensive mutagenesis method described elsewhere41. For the Fc region, human IgG1 was chosen as template and was engineered for selective and enhanced binding to the human FcγRIIb and for stronger affinity to FcRn in acidic pH conditions. The resulting antibody was named GYM329 (Fig. 1A) and was subjected to subsequent characterization.
The representative results of the surface plasmon resonance (SPR) analysis demonstrating the pH-dependent binding of GYM329 to the human latent myostatin are shown in Fig. 1B. A higher dissociation rate was observed at pH 6.0 than at pH 7.4 in the dissociation phase of the analysis. The ka, kd, and KD values at pH 7.4 for the human, cynomolgus monkey, and mouse latent myostatins were nearly equal (Table S1). To assess the binding affinity of GYM329 Fc to various FcgR subtypes, GYM329 and a reference wild type human IgG1 (with trastuzumab Fab) were captured onto an SPR chip. The amount of soluble human FcgR recombinant proteins bound to the antibodies were compared. The ratios (GYM329-bound FcγR divided by the reference IgG1-bound FcγR) are shown in Table S2. FcγRIIb had higher affinity to GYM329 than to IgG1 (>5-folds), and other FcγRs had much lower affinities to GYM329. This data demonstrates that the Fc of GYM329 has enhanced selectivity and affinity to FcγRIIb. The binding affinity of GYM329 and the reference IgG1 to the human and cynomolgus monkey FcRn at pH 6.0 were determined by SPR analysis (Table S3). GYM329 had stronger affinity to FcRn than IgG1 in acidic conditions. This is indicative of a long serum half-life for GYM329 as there is more efficient FcRn-mediated recycling of the internalized antibody from the endosome, as observed in other engineered antibodies with the same property.
The effects of different GYM329 doses on the inhibition of BMP1-facilitated and spontaneous activation of human, cynomolgus monkey, and mouse latent myostatins were assessed using the SEAP reporter gene assay. The half-maximal inhibitory concentrations (IC50) were comparable between the species (Table S4). We also confirmed that the inhibitory effect of GYM329 is specific to latent myostatin, as no inhibition of mature myostatin or latent/mature GDF11 was observed (Fig. 1C).
GYM329-induced muscle mass increase and muscle strength enhancement in three different mouse models of muscle disease
To determine whether GYM329 could increase muscle mass and enhance muscle strength, we tested three different mouse models of muscle disease. To minimize the production of mouse antibodies against human IgG1-derived constant regions of GYM329, we generated its murine functional equivalent, GYM-mFc, by fusing mouse IgG1-derived constant regions to the Fab domain of GYM329. GYM-mFc was confirmed to have identical inhibitory potency against activation of latent myostatin as GYM329 invitro (Fig. S1).
We first evaluated the activity of GYM-mFc in a Duchenne muscular dystrophy (DMD) mouse model, C57BL/10-mdx Jic (mdx mice). We compared the activity of GYM-mFc with the activity of three different clinically tested anti-myostatin blocking antibodies that we generated in-house based on the sequence information in International Immunogenetics Information System (IMGT) database (http://www.imgt.org). These include two mature myostatin-neutralizing antibodies based on the sequences of landogrozumab and domagrozumab and the anti-activin receptor II antibody bimagrumab. Two doses (high and low) of all the agents were evaluated. Almost similar maximum effects on muscle mass increment were observed with the high doses of all tested agents (Fig. 2A). On the other hand, significant grip strength enhancement was observed in mice treated with GYM-mFc (at both doses) and with domagrozumab (at the high dose) but not in mice treated with the other agents (Fig. 2B).
We evaluated the activity of these agents in aged mice (seventy-nine weeks old), which exhibit sarcopenic characteristics, such as decreased muscle mass per bodyweight, impaired muscle strength, and low physical activity (Fig. S2). An increase in muscle mass up to levels comparable to those found in young mice was observed only in the aged mice treated with GYM-mFc or with bimagrumab (Fig. 2C). At a high dose of landogrozumab (50 mg/kg), a small increase in muscle mass was observed, but no effect was observed in treatment with the lower dose of landogrozumab or with any dose of domagrozumab. Intriguingly, a remarkable enhancement in grip strength was observed only in GYM-mFc-treated mice, reaching levels that are comparable to those in young mice even at a lower treatment dose (3 mg/kg, Fig. 2D). Much weaker grip strength enhancement was seen at the high bimagrumab dose (100 mg/kg), and no enhancement was observed in mice treated with landogrozumab and domagrozumab.
Lastly, the activity of these agents was assessed in a muscular atrophy model, which was recreated by suspending the hindlimb of the mice for two weeks. Muscle tissue weight reduction was seen in the vehicle-treated muscular atrophy group compared to the non-suspension control group, which was completely suppressed by treatment with GYM-mFc, domagrozumab, and bimagrumab, but not by landogrozumab (Fig. 2E). Remarkable impairment in grip strength was also induced by hindlimb suspension, and it was significantly ameliorated only by GYM-mFc treatment (Fig. 2F).
In all these studies, GYM-mFc had the most potent effects on muscle mass and muscle strength of all the tested anti-myostatin agents.
Negative contribution of GDF11 signaling blockade to muscle strength enhancement
We next aimed to determine the possible molecular mechanism underlying the superior effects of the GYM329 surrogate over the other agents, particularly for muscle strength enhancement, which is more important therapeutically than muscle mass increment. Two fundamental differences exist between GYM329 and the other agents: specificity to myostatin and the sweeping function. GYM329 can specifically inhibit myostatin signaling by binding latent myostatin and suppressing its activation, whereas landogrozumab and domagrozumab bind mature myostatin and GDF11 with similar affinity to both molecules, thereby equally inhibiting both (Fig. S3). Bimagrumab is an antibody against ActRII that blocks its interaction with multiple TGF-β superfamily ligands including myostatin, GDF11, and activin28,42. In addition, the mechanism of action of all these conventional antibodies includes simple neutralization, while GYM329 has a sweeping function that reduces antigen levels by forcing internalization via FcγRs.
We first determined whether the specificity of GYM329 to myostatin is involved in its superior efficacy in muscle strength enhancement. This was investigated in a hindlimb suspension muscular atrophy model because direct testing of GYM329 with the human IgG sequence is possible in this model with the use of mice with severe combined immunodeficiency (SCID mice). As observed with the surrogate antibody (Fig. 2F), GYM329 (3 mg/kg) significantly enhanced grip strength in the hindlimb suspension mouse model, reaching levels comparable to those in the non-suspended control mice (Fig. 3A). To assess the influence of the GDF11 signaling blockade in this phenomenon, we generated a neutralizing antibody specific to mature GDF11. We have confirmed that this antibody binds mature GDF11 but not myostatin (Fig. S4), and that it specifically neutralizes GDF11-mediated signaling (Fig. 3B). Interestingly, we found that the muscle strength enhancement induced by GYM329 was significantly suppressed by combined treatment with the anti-GDF11 antibody (Fig. 3A). Meanwhile, anti-GDF11 antibody treatment alone did not show any muscle strength enhancement activity.
We next examined the effect of GDF11 itself on muscle strength enhancement. Administration of recombinant mature GDF11 (0.01 mg/kg bodyweight, three times, intraperitoneal) significantly suppressed the reduction of muscle strength induced by hindlimb suspension (Fig. 3C). On the other hand, recombinant myostatin, which was confirmed to have similar potency as the recombinant mature GDF11 protein in the reporter gene assay described above (Fig. S5), did not have any suppressive effects on muscle strength reduction. Taken together, these data suggest that GDF11 and myostatin act in opposing directions in terms of muscle strength enhancement, where inhibition of GDF11 signaling negatively impacts muscle strength enhancement.
Contribution of the sweeping function of GYM329 to its muscle strength enhancement activity
We then explored the possible contribution of the sweeping function of GYM329 to its superior muscle strength enhancing activity over other anti-myostatin agents. We generated a reference antibody, hMST1032-hIgG1, which has a humanized MST1032 Fab without the pH-dependent antigen binding property, and a wild-type human IgG1 Fc. The Smad reporter gene assay showed that hMST1032-hIgG1 has the same potency as GYM329 in inhibiting the BMP1-mediated activation of latent myostatin (Fig. 4A). Due to the lower affinity of the wild-type human IgG1 to FcγRs, particularly to FcγRIIb, compared to the engineered GYM329 Fc, the immune complex formed by hMST1032-hIgG1 and latent myostatin was thought to be captured and internalized into the cell less efficiently compared to the GYM329 and latent myostatin combination. Furthermore, due to the lack of pH-dependent binding in hMST1032-hIgG1, latent myostatin is not expected to dissociate from the antibody in the acidic endosome, and the latent myostatin bound to the antibody is brought back to the extracellular space. This would result in a lower clearance of latent myostatin, even if it is taken up into the cell via a less efficient internalization of the immune complex. In the absence of the two properties essential for the sweeping function, hMST1032-hIgG1 serves as a non-sweeping reference for GYM329 with the same potency in the myostatin functional blockade.
Muscle strength enhancement by GYM329 and hMST1032-hIgG1 was compared in the hindlimb suspension muscular atrophy model (Fig. 4B). We found that hMST1032-hIgG1 treatment led to lower muscle strength enhancement even when administered at a high dose (30 mg/kg), suggesting that the sweeping function of GYM329 contributes to its potent capacity to enhance muscle strength. A pharmacokinetic study of GYM329 and hMST1032-hIgG1 showed that hMST1032-hIgG1 has a slightly longer plasma half-life than GYM329 (Fig. S6); therefore, the superior activity of GYM329 in this experiment is not due to differences in antibody exposure.
Western blot analysis for mature myostatin in quadriceps collected from the animals in the same experiment revealed a reduction in the levels of mature myostatin by GYM329 treatment but not by hMST1032-hIgG1 treatment (Fig. 4C, Fig. S7). This was further confirmed by fluorescent immunostaining (Fig. 4D) that detects both pro-myostatin and latent myostatin. The signals appeared to be stronger in the extensor digitorum longus (EDL) muscle of the hindlimb that had been suspended to induce muscular atrophy (vehicle treatment, middle) compared to non-manipulated animals (left). The signal was greatly suppressed in the muscles treated with GYM329 (right), suggesting that myostatin sweeping occurred in the muscle. GYM329 treatment reduced the baseline plasma myostatin levels (mature + latent myostatin; Fig. 4E), which is another indication of the sweeping function of GYM329. Treatment with the non-sweeping variant hMST1032-hIgG1 enhanced the myostatin staining signal in the muscle and increased the level of total myostatin in the plasma, which indicate lower muscle and systemic clearance of myostatin. The sweeping effect exhibited by GYM-mFc was similar to the effects of GYM329 (Fig. S8). Taken together, the sweeping function of GYM329 and its myostatin specificity contributed to its superior activity.
Muscle mass increase and plasma myostatin sweeping activity of the GYM329 surrogate in cynomolgus monkeys
Finally, we investigated the activity of GYM329 in cynomolgus monkeys using a functional equivalent, GYM-cyFc. Although the affinity of the GYM329 Fc to human FcgRIIb was higher than that of the wild-type human IgG1 (Table S2), its affinity to cynomolgus FcgRIIa and cynomolgus FcgRIIb was similar (Table S5). We therefore generated another engineered Fc with enhanced affinity to cynomolgus FcγRIIa and FcgRIIb from the Fc of GYM329, which is combined with the same Fab of GYM329 resulting in GYM-cyFc. We have confirmed that GYM-cyFc selectively binds cynomolgus monkey FcgRIIa and FcggRIIb (Table S5), with lower binding activity to FcgRI and FcgRIII. The IC50 of GYM-cyFc against BMP1-mediated activation of cynomolgus myostatin in vitro was determined to be 0.199 mg/mL, which is comparable to the IC50 of GYM329 against the activation of human latent myostatin (0.182 mg/mL). These data suggest that GYM-cyFc is a functional equivalent of GYM329 that may be used for studies in cynomolgus monkeys.
Administration (intravenous, i.v.) of 1.25, 2.5, or 5 mg/kg GYM-cyFc to female cynomolgus monkeys (n = 10 per group) every 4 weeks for 2 months resulted in increased muscle section area compared to the vehicle-treated group as detected by magnetic resonance imaging (MRI) (Fig. 5A). In addition, the rate of body weight increase relative to baseline was also higher in the GYM-cyFc group than in the vehicle group (Fig. 5B). Rapid and drastic reduction in the level of total plasma myostatin, which reflects the sweeping function of GYM-cyFc, was observed after the 1st and 2nd doses of the antibody (Fig. 5C). No reduction in plasma myostatin levels was observed in the vehicle treatment group. All doses of GYM-cyFc were well tolerated in the cynomolgus monkeys throughout the duration of the study. Preliminary toxicological investigation revealed no significant adverse pathological effects or abnormalities in hematology and blood chemistry.