Chemoproteomic target deconvolution reveals Histone Deacetylases as targets of (R)-lipoic acid

Abstract

Lipoic acid is an essential enzyme cofactor in central metabolic pathways. Due to its claimed antioxidant properties, racemic (R/S)-lipoic acid is used as a food supplement but is also investigated as a pharmaceutical in over 180 clinical trials covering a broad range of diseases. Moreover, (R/S)-lipoic acid is an approved drug for the treatment of diabetic neuropathy. However, its mechanism of action remains elusive. Here, we performed chemoproteomics-aided target deconvolution of lipoic acid and its active close analog lipoamide. We find that histone deacetylases HDAC1, HDAC2, HDAC3, HDAC6, HDAC8, and HDAC10 are molecular targets of the reduced form of lipoic acid and lipoamide. Importantly, only the naturally occurring (R)-enantiomer inhibits HDACs at physiologically relevant concentrations and leads to hyperacetylation of HDAC substrates. The inhibition of HDACs by (R)-lipoic acid and lipoamide explain why both compounds prevent stress granule formation in cells and may also provide a molecular rationale for many other phenotypic effects elicited by lipoic acid.

Fig. 1. Chemoproteomics identifies HDACs as targets of lipoic acid and lipoamide.

a An affinity matrix iL was synthesized by immobilizing racemic (R/S)-lipoic acid to sepharose beads. The resulting affinity matrix resembles lipoamide and is reduced to dihydrolipoamide under assay conditions (1 mM DTT). b Schematic representation of the competition pulldown assay used in this study. Lysate containing correctly folded proteins interacting with endogenous cofactors or macromolecular binding partners is incubated with the affinity matrix to pull down target proteins. In a competition experiment, the lysate is first incubated with different doses of the free drug of interest (black droplet symbol) before pull down. LC–MS/MS is used to quantify target proteins. The intensities are plotted against the drug concentration to yield dose-response curves, from which binding EC50s and K_d^app can be derived (cf=correction factor; see “Methods”). c Dose–response curves for lipoic acid and lipoamide using a lysate of SW620 cancer cells. Structures of drugs are shown in the reduced form and the chiral center is indicated by an asterisk. d Dose–response curves for (R/S)-LM using a lysate of MV4–11 cancer cells showing HDACs and HDAC complex partners of the CoREST (blue) and MiDAC (brown) complexes. Source data are provided as a Source Data file.

Fig. 2. HDAC activity assays confirm the inhibitory effects of the reduced forms of (R)-lipoic acid, (R/S)-lipoic acid, and (R/S)-lipoamide.

a Influence of the reducing agent TCEP (0.5 M) on (R/S)-LA mediated HDAC enzymatic activity via reduction and ring opening of the drug (for (R/S)-LM see Supplementary Fig. 2b) (n = 3 technical replicates, data are represented as mean value ± SEM). b HDAC inhibitory effect of the (R)-enantiomer of lipoic acid compared to the (S)-enantiomer (n = 3 technical replicates, data are represented as mean value ± SEM). c Exemplary dose-response profiles of all compounds tested for HDAC1 inhibition in the presence of 0.5 M TCEP ((R/S)-LA red = (R/S)-dihydrolipoic acid). n = 3 technical replicates. d Summary of EC50 values derived from dose-dependent HDAC inhibition curves. Source data are provided as a Source Data file.

Fig. 3. (R/S)-lipoic acid and (R/S)-lipoamide lead to hyperacetylation of HDAC substrates in cells.

a Western blot analysis of acetylation levels of HDAC6 substrates following 12 h treatment of HEK293T cells with (R/S)-LA, (R/S)-LA, and the HDAC6 inhibitor ACY738 (see also Supplementary Fig. 3a). b Western blot for α-Tubulin AcK40 acetylation levels after 12 h treatment of A549 cells with SAHA (Vorinostat), (R)-LA, and (S)-LA. c Western blot analysis for global lysine acetylation levels of HeLa S3 cells treated with (R/S)-LA (16 h; n = 2 independent biological experiments, error bars represent standard deviation; see also Supplementary Fig. 3c). The histograms show hyperacetylation of proteins in the size range of established HDAC substrates, such as Histones (11–16 kDa), Peroxiredoxin (22 kDa), α-Tubulin (50 kDa), and others. d HDAC6 and HDAC10 nano-BRET assays demonstrating in-cellulo target engagement in HEK293T cells (n = 3 independent experiments, data are represented as mean value ± SD; curve fitted with a variable slope; bottom constrained to 0 and top constrained to 100). Source data are provided as a Source Data file.

Fig. 4. Lipoic acid and lipoamide reduce stress granule formation in cells.

a Immunofluorescence detection of the stress granule marker G3BP1 in A549 cancer cells. Stress granules appear as red foci in the DMSO control and cells treated with (HDAC-inactive) (S)-LA. The reduction of defined stress granules in response to (R)-LA and (R/S)-LA is apparent from the blurred red areas. b Quantification of the number of stress granules per cell. Each treatment was performed in n = 3 independent biological experiments and between 140 and 150 cells were submitted to stress granule counting. c Levels of oxidative stress induced by 2 h treatment with 200 µM Tertbutylhydroperoxide (BuOOH) after 1 h pre-treatment with drugs (Vor Vorinostat, NAC N-acetylcysteine) in A549 cells. Oxidative stress levels were assessed using the CellRox assay. Every data point corresponds to one biological replicate and is the mean CellRox intensity from 9 to 10 pictures capturing 60–180 cells in total (n = 2 biologically independent samples for 100 µM (S)-LA, n = 3 biologically independent samples for all other treatments, AU arbitrary units). d Levels of oxidative stress in A549 cells after 1.5 h drug pre-treatment, optionally followed by a 30 min arsenite (1 mM) pulse. Oxidative stress levels were assessed using the CellRox assay. Every data point corresponds to one biological replicate and equals the mean CellRox intensity from 10 to 15 pictures capturing 60–180 cells in total (n = 3 biologically independent samples for each drug dose, AU arbitrary units). b–d Statistical significance was calculated between the control and drug pre-treatments by one-way ANOVA following the Dunnett test for multiple comparisons using the GraphPad Prism software. Data are presented as means ± SD. ns not significant, ***P-value ≤ 0.001, **P-value ≤ 0.01, *P-value ≤ 0.05 in one-way ANOVA after Dunnett’s multiple comparison test). Source data are provided as a Source Data file.