Activated GDF11/8 subforms predict cardiovascular events and mortality in humans

Abstract

Circulating Growth Differentiation Factors 11 and 8 (GDF11/8) exist in both latent and active forms, and it is unclear if specific forms can predict disease outcomes. Our data suggest that a dual-specific aptamer selectively binds GDF11/8 after prodomain activation. In 11,609 patients at risk for future cardiovascular events, low dual-specific aptamer-detected GDF11/8 levels strongly predicted adverse outcomes, including cardiovascular events (HR = 0.43, p = 9.1 × 10⁻⁶³) and all-cause mortality (HR = 0.33, p = 4.8 × 10⁻⁴⁰). Use of selective aptamers suggested that results observed with the dual-specific aptamer for cardiovascular and mortality risk replicated with a GDF8 aptamer although with a smaller effect size. In a second cohort of 4110 individuals (ARIC), low dual-specific aptamer-detected GDF11/8 levels also predicted increased 8 year dementia risk (HR = 0.66, p = 0.00148). Our findings reveal that activation of GDF11/8 may be a factor in future aging-related cardiovascular and cognitive decline.

Fig. 1. Aptamer only pulls-down a fraction of Growth differentiation factor 11/8 (GDF11/8) from healthy human serum.

A−C The amount of GDF11/8 ligands in healthy human serum samples, measured by LC-MS/MS. D−F The amount of GDF11/8 ligands pulled-down by the aptamer from the same human serum samples, measured by LC-MS/MS. G−I the ratio of the amount pulled-down aptamer to the LC-MS/MS measure. Young: 18−25 years old, Middle Aged: 40−60 years old, and Aged: older than 60 years old. Two-way ANOVA with Turkey’s multiple comparison test, n = 10 in each group. Statistical analysis: two-way ANOVA (two-sided) with Tukey’s post-hoc test for multiple comparisons. Data are presented as mean ± SEM. n = 10 biological replicates per group. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. LC-MS/MS = immunoplexed liquid chromatography with tandem mass spectrometry.

Fig. 2. The dual-specific aptamer did not recognize the inactive GDF11/8 complex.

A Unprocessed GDF8/11 (inactive) undergoes Furin cleavage to form the latent complex (inactive). TLD processing converts the latent complex into the triggered state (active), and dissociation of the remaining prodomain yields an unbound mature ligand (active). The binding of an antagonist renders the ligand inactive. B, C Anti-GDF8 prodomain (B), and anti-GDF11/8 mature ligand (C) western blot of the latent complex (Latent GDF8 and Latent GDF8P11M; GDF11 mature ligand with GDF8 prodomain), latent complex treated with BMP1 (Latent GDF8 / BMP1 and Latent GDF8P11M / BMP1), and latent complex incubated with the same digestion conditions but without BMP1 (Latent GDF8 / BMP1 and Latent GDF8P11M / no BMP1). Black labels are samples prior to the pull-down and arrows are following the aptamer pull-down. In (B), BMP1 processing of the latent complexes was confirmed via the detection of the ~20 kDa band (arrows). Non-pulled-down samples are shown in black, and the samples are pulled-down by aptamers below the “pull-down samples” labeling. Immunoblotting of latent complex is first used for anti-prodomain blotting, then stripped and reblotted for anti-mature ligand. All aptamer pull-downs were processed simultaneously. Experiments were independently repeated twice with similar results. D Schema of BMP1 processing required for aptamer binding. Source data are provided as a Source Data File.GDF11/8 ligand (monomer): 12.4 kDa, GDF11/8 ligand (dimer): 25 kDa, BMP1 processed prodomain: ~20 kDa, latent GDF11/8: ~40 kDa, mixture of unprocessed and latent GDF11/8: ~50 kDa. BMP1 = bone morphogenetic protein 1; TLD = Tolloid protease; spec = specific. Experiments performed in duplicate.

Fig. 3. GDF11-specific aptamer did not recognize GDF11/8 mature ligand in the presence of their inhibitors.

A, B Anti-GDF11/8 mature ligands western blot of GDF8 (A) and GDF11 (B) mature ligands with inhibitors. The input shows the proteins before the pull-down. The proteins after the pull-down are below the “pulled-down sample” labeling, with non-pulled-down proteins on the left. Mature ligands and inhibitors are mixed in 1:2.5 molar ratio respectively. C Summary of ligand states captured by the aptamers. Experiments were independently repeated twice with similar results. GDF11/8 ligand (monomer): 12.4 kDa, GDF11/8 ligand (dimer): 25 kDa, FS288 = follistatin 288; FSTL3 = follistatin-like 3; WFIKKN1 = WAP, Kazal, immunoglobulin, Kunitz and NTR domain-containing protein 1; WFIKKN2 = WAP, Kazal, immunoglobulin, Kunitz and NTR domain-containing protein 2. Experiments performed in duplicate. Source data are provided as a Source Data File.

Fig. 4. Aptamer recognition of GDF8 following BMP1 activation of latent GDF8.

A previously performed on latent GDF8. B latent GDF8 treated with BMP1 for 24 h; (C) latent GDF8 treated with BMP1 for 48 h. The aptamer was coupled to a streptavidin biosensor and dipped into a dilution series ranging from 400 nM to 3.13 nM of latent GDF8 or BMP1-treated latent GDF8. The binding of the ligand to the aptamer is shown in black and the 1:1 binding model fit is shown below the “pulled-down samples” labeling. D Ligand activity was assessed at 24 h post BMP1 digestion using the HEK293 CAGA luciferase assay. Data shown are from at least 3 technical replicates. Source data are provided as a Source Data File.BLI = Biolayer interferometry; BMP1 = bone morphogenetic protein 1.

Fig. 5. Incidence of the composite end-point and heart failure hospitalization, stroke, myocardial infarction, and death in the meta-cohort unadjusted.

A Stratified by quartile of active GDF11/8 [Q1 (0–2.70) n = 2,905, Q2 (2.71-2.76) n = 2904, Q3 (2.77-2.83) n = 2899, Q4 (2.84-3.53) n = 2,901]. P-Values for trend are p = 2.0e-60 for the composite endpoint, p = 2.5e-39 for death, p = 2.8e-32 for heart failure, p = 6.3e-03 for myocardial infarction, and p = 2.3e-02 for stroke. B Stratified by quartile of active GDF8 [Q1 (0–2.27) n = 2903, Q2 (2.28-2.32) n = 2,903, Q3 (2.33-2.37) n = 2,901, Q4 (2.38-4.17) n = 2,902]. P-Values for trend are p = 2.6e-32 for the composite endpoint, p = 1.3e-17 for death, p = 3.7e-26 for heart failure, p = 6.8e-01 for myocardial infarction, and p = 1.3e-02 for stroke. C Stratified by quartiles of active GDF11 [Q1 (0–1.86) n = 2,930, Q2 (1.87–1.91) n = 2,884, Q3 (1.92–2.00) n = 2894, Q4 (2.01–4.94) n = 2901]. There were no significant trend associations between active GDF11 quartiles with the composite endpoint of any of its individual components.Statistical analysis: Chi-squared test for trend in proportions (two-sided) was used to assess each event group distribution for each respective GDF quartile set. Data are presented as mean ± SEM. Sample size reflects independent biological replicates, each representing a unique patient from the meta-cohort. No technical replicates were used. The unit of study is the individual patient. Groups compared were stratified by quartiles of circulating ligand levels. No pooling of samples occurred. Source data are provided as a Source Data File.

Fig. 6. Active GDF11/8, GDF8, and GDF11 ligand levels by age.

Box plots represent the distribution of circulating ligand levels measured by aptamer-based assays. The center line indicates the median, the bounds of the box represent the 25th–75th percentiles, and the whiskers extend to the 5th–95th percentiles. Units on the vertical axis are log10 transformed relative fluorescence units (RFU). Statistical analysis was performed using the Jonckheere–Terpstra test for trend (two-sided). A significant inverse relationship with age was observed for active GDF11/8 (p = 5.63e-130) and active GDF8 (p = 5.87e-253), while no significant trend was observed for active GDF11 (p = 0.331). Subjects per age group (years old): <50 (n = 497), 50-54 (n = 571), 55-59 (n = 830), 60−64 (n = 1,111), 65−69 (n = 1708), 70−74 (n = 2830), 75−79 (n = 2232), 80+ (1,830). Each data point represents an independent biological replicate corresponding to a unique individual. No technical replicates or pooled samples were used. The unit of study is the individual participant. Source data are provided as a Source Data File.