Substrate-Dependent Modulation of SIRT2 by a Fluorescent Probe, 1-Aminoanthracene

Abstract

Sirtuin isoform 2 (SIRT2) is an enzyme that catalyzes the removal of acyl groups from lysine residues. SIRT2’s catalytic domain has a hydrophobic tunnel where its substrate acyl groups bind. Here, we report that the fluorescent probe 1-aminoanthracene (AMA) binds within SIRT2’s hydrophobic tunnel in a substrate-dependent manner. AMA’s interaction with SIRT2 was characterized by its enhanced fluorescence upon protein binding (>10-fold). AMA interacted weakly with SIRT2 alone in solution (K_d = 37 μM). However, when SIRT2 was equilibrated with decanoylated peptide substrate, AMA’s affinity for SIRT2 was enhanced ~10-fold (K_d = 4 μM). The peptide’s decanoyl chain and AMA co-occupied SIRT2’s hydrophobic tunnel when bound to the protein. In contrast, AMA binding to SIRT2 was competitive with myristoylated substrate whose longer acyl chain occluded the entire tunnel. AMA competitively inhibited SIRT2 demyristoylase activity with an IC₅₀ of 21 μM, which was significantly more potent than its inhibition of other deacylase activities. Finally, binding and structural analysis suggests that the AMA binding site in SIRT2’s hydrophobic tunnel was structurally stabilized when SIRT2 interacted with decanoylated or 4-oxononanoylated substrate, but AMA’s binding site was less stable when SIRT2 was bound to acetylated substrate. Our use of AMA to explore changes in SIRT2’s hydrophobic tunnel that are induced by interactions with specific acylated substrates has implications for developing ligands that modulate SIRT2’s substrate specificity.

Graphical Abstract

INTRODUCTION

In mammals, SIRT2 is an NAD⁺-dependent enzyme that removes acyl groups from the epsilon amine of lysine residues.¹ SIRT2 has an accommodating substrate site and uses a conserved reaction to remove lysine modifications that range from acetyl to long fatty groups such as myristoyl.^1–4 SIRT2’s deacylase activity has broad relevance in biological and disease processes, making it desirable to target SIRT2’s activity with small molecules. Indeed, a variety of SIRT2 ligands are being developed to modulate SIRT2’s deacylase activities.^5–9 These compounds have proven efficacious in treating experimental models of cancers and neurodegenerative disorders.^5,6,9 However, most of the reported SIRT2-selective ligands only inhibit its deacetylase activity without affecting its demyristoylase activity.^5,10 In some contexts, selectively inhibiting the long fatty deacylase activity of SIRT2 would appear to be an effective strategy to combat cancer formation.¹¹

High-resolutions structures of SIRT2 are available with the protein bound to synthetic inhibitors, physiologic ligands, and/or acylated peptides. The structures reveal two main conformations for the enzyme. By itself, SIRT2 assumes an “open” conformation until its acylated substrate binds, at which point SIRT2’s zinc binding domain rotates and the enzyme assumes a “closed” conformation.^12–14 Once in the closed conformation, SIRT2’s co-substrate NAD⁺ binds and its deacylation reaction can proceed.³ Although SIRT2’s major conformations are “open” and “closed”, structural heterogeneity exists in each conformation when the enzyme accommodates binding to various substrates and small molecules. SIRT2’s dynamics and flexibility in different conformational states are important for the design of small molecules that target the protein.

In this work, we sought to explore a conformation-specific binding site on SIRT2 that can be occupied by the drug propofol.^12,15 Propofol is the among the most widely-used injectable general anesthetics, and the clinical relevance of propofol-SIRT2 binding is of interest.^12,16,17 Propofol preferentially binds the closed, substrate-bound conformation of SIRT2, which was revealed with photolabeling studies.^12,15 To further investigate the propofol binding site, we examined the interaction of SIRT2 with 1-aminoanthracene (AMA), which has utility as a fluorescent probe unlike propofol. AMA mimics propofol’s chemical properties including mass (193 g/mol for AMA vs. 178 g/mol for propofol), polar surface area (26 cm² vs. 20 cm²), van der Waals volume (179 ų vs. 192 ų), and octanol/water partition coefficient (LogP) (3.7 vs. 3.8).^18–20 Additionally, AMA binds the same sites as propofol on other proteins that include apoferritin,^18,21 GABA_A receptors,²² and LFA-1.²³

AMA has environment-dependent fluorescent properties that facilitate measuring its protein interactions, making it a useful tool to explore small binding sites on proteins.^18,20,24 Our binding assays revealed that AMA preferentially binds closed conformational states of SIRT2, but only when the protein interacts with specific acylated substrates. The formation of AMA’s binding site depends on the ability of substrate acyl groups to facilitate ordering of flexible residues that surround SIRT2’s hydrophobic tunnel. AMA also inhibits the demyristoylase activity of SIRT2 much more potently than its other deacylase activities, including deacetylation. These findings are important for the discovery of new small molecules that modulate SIRT2’s substrate acyl group specificity.

MATERIALS AND METHODS

Materials

AMA (1-aminoanthracene; >98% purity) was purchased from TCI, and propofol (2,6-diisopropylphenol; 97% purity) was purchased from Alfa Aesar. Synthetic peptides purchased from New England Peptide include KGLGKGGAK(acetyl)RHRKGWW, KGLGKGGAK(4-oxononanoyl)RHRKWW, KGLGKGGAK(decanoyl)RHRKGWW, KGLGKGGAK(myristoyl)RHRKGWW, KGLGKGGAK(4-oxononanoyl)RHRK, KGLGKGGAK(decanoyl)RHRK, and KGLGKGGAK(dodecanoyl)RHRK. Synthetic peptides purchased from JPT Peptide Technologies include KGLGKGGAK(acetyl)RHRK and KGLGKGGAK(myristoyl)RHRK. All peptides were purified to >95% with HPLC and their identities were validated with ESI-MS.

Recombinant SIRT2 production

Recombinant human SIRT2 (amino acids 34–356) was expressed and purified from a pET28a vector described previously.⁶ The plasmid encoded a 6xHis-tag and SUMO protein fused to the N-terminus of SIRT2’s catalytic domain, which has been used in previous enzymatic and crystallographic studies.^4,6 The plasmid was transformed into BL21-CodonPlus (DE3) cells, and bacteria was grown and induced using standard procedures.^4,6,25 The encoded protein was purified using a Ni²⁺ column, then the 6xHis-SUMO domain was removed with SUMO protease (ULP1). The mixture was then re-applied to a Ni²⁺ column, and SIRT2 was isolated in the flow-through. SIRT2 was concentrated with a centrifugal filter then further purified with gel filtration using a Bio-Rad ENrich SEC 650 column (10 × 300 mm) with a mobile phase of PBS plus 1 mM DTT. Isolated SIRT2 was >98% pure as indicated by SDS-PAGE and Coomassie staining, and the protein was aliquoted, snap frozen, and stored at −80°C. SIRT2 concentration was determined by its absorbance at 280 nm and its calculated extinction coefficient (Σ=32890 M⁻¹cm⁻¹).²⁶ An identical procedure was used to produce SIRT2(Y139G) after site directed mutagenesis on the plasmid.

Recombinant SIRT6 production

Recombinant human SIRT6 was expressed and purified from a pQE-80 vector described previously.²⁷ The plasmid encoded full-length SIRT6 with an N-terminal 6xHis-tag. The protein was expressed in BL21-CodonPlus (DE3) cells identical to SIRT2, and SIRT6 was purified using a Ni²⁺ column. The SIRT6 protein was concentrated with a centrifugal filter then further purified with gel filtration as described for SIRT2. Isolated SIRT6 was >98% pure as indicated by SDS-PAGE and Coomassie staining, and the protein was aliquoted and snap frozen. SIRT6 concentration was determined by its absorbance at 280 nm and its calculated extinction coefficient (Σ=33460 M⁻¹cm⁻¹).²⁶

SIRT2 binding assays with AMA

All assays were performed at 23°C in a buffer of PBS plus 1 mM DTT. A quartz microcuvette (3 mm pathlength) was used with a Horiba Fluoromax-4 instrument, and the assay volume was 160 μl.

To measure SIRT2’s interaction with AMA, SIRT2 was first diluted to the indicated concentration then scanned with the fluorometer (Excitation λ=380 nm, Emission λ=390–700 nm). Next, 1 μl of AMA dissolved in PBS was added such that the final AMA concentration was 100 nM, the sample was gently mixed by pipetting, then the fluorescence scan was repeated on the protein/AMA mixture. The initial, protein-only background spectra was subtracted from the spectra obtained with the combined protein/AMA mixture. Counts per second (CPS) fluorescence intensities at λ=520 nm were extracted from the background-subtracted spectra. The CPS values were plotted versus SIRT2 protein concentration, and the K_d was determined from a curve fit using the quadratic binding equation

$$\begin{gathered} F=F_{\left(min\right)}-\left[\frac{F_{(\mathrm{m}\mathrm{i}\mathrm{n})}-F_{(\mathrm{m}\mathrm{a}\mathrm{x})}}{2\mathrm{\ast }[AMA]}\right]\mathrm{\ast }[b-\sqrt{b^2-4\mathrm{\ast }[SIRT2]\mathrm{\ast }[AMA]} \\ b=K_d+[SIRT2]+[AMA] \end{gathered}$$ (1)

where F_(min) and F_(max) are the minimum and maximum fluorescence intensities, [AMA] is the total AMA concentration, [SIRT2] is the SIRT2 concentration, and K_d is the SIRT2 concentration at half-maximal saturation.^25,28 We note that quantitatively similar binding affinities can be obtained by using an excitation wavelength at least as high as 410 nm while still collecting peak emission intensities at 520 nm. Additionally, we have performed assays with AMA as the limiting component in the 50–200 nM range and obtained similar binding results with good fluorescence signal above background.

For AMA competition assays containing propofol, peptide, and/or other ligands, SIRT2 was first equilibrated with the competitor and the emission scan was collected with the fluorometer. The concentrations of SIRT2 in competition assays were chosen to ensure sufficient fluorescence signal above background in the absence of competitor while conserving the use of purified protein. Next, 1 μl of AMA in PBS was added to a final concentration of 100 nM, the sample was gently mixed, and the fluorescence scan repeated. The initial background spectra was again subtracted from the spectra obtained in the presence of AMA. Propofol was initially diluted in DMSO for stock solutions because of its limited aqueous solubility (~300–400 μΜ), and the final DMSO concentration in the cuvette for every data point on the propofol competition curve was 0.5% regardless of the propofol concentration. For peptide competition assays, the peptides dissolved readily in PBS and no DMSO was present. AMA binding assays with SIRT6 were performed essentially the same as those with SIRT2.

For some competition assays, the CPS values at 520 nm were plotted versus the log₁₀ concentration of the propofol or peptide competitor. This data was fit with a sigmoidal curve using the equation

$$Y=Y_{min}+\frac{Y_{max}-Y_{min}}{1+{10}^{X-\log (IC50)}}$$ (2)

which assumes a standard Hill slope of −1. The IC₅₀ from this curve was used to calculate the affinity (K_i) of propofol or myristoyl-peptide for SIRT2 using the equation

$$K_i=\frac{{IC}_{50}}{\left(\frac{L_{50}}{K_d}\right)+\left(\frac{P_0}{K_d}\right)+1}$$ (3)

where L₅₀ is the concentration of free AMA at 50% inhibition, P₀ is the concentration of free SIRT2 at 0% inhibition, and K_d is the dissociation constant of the SIRT2-AMA complex.²⁹

SIRT2 deacylase assays

Enzyme assays were performed at 37 °C with recombinant SIRT2, a saturating concentration of NAD⁺ (1 mM),³ and varying amounts of acylated H4K16 peptide. Enzyme concentrations and time points were chosen to capture the steady-state initial velocity of the enzyme reactions. PBS with 1 mM DTT was used as the assay buffer, and the enzyme reactions were quenched with 200 μl of 200 mM HCl and 320 mM acetic acid in methanol.² Precipitated protein was removed by centrifugation, and the supernatant was analyzed by HPLC using a water/acetonitrile gradient that contained 0.05% trifluoroacetic acid. A Kinetex XB-C18 column from Phenomenex was used to separate the acylated and deacylated peptides from the assays,^2,3,30 and the peptide peaks were observed with an in-line UV-Vis detector (Absorbance λ=290 nm). Peptide peaks were quantified by integrating the area under the peaks, and the data was used to calculate initial reaction rates for SIRT2’s deacylation of the peptides. Initial reaction rates were converted to observed rate constants (k_obs) by dividing initial velocities by the enzyme concentration used in the corresponding reaction, and k_obs was plotted versus the peptide substrate concentration [S]. The data was fit with a standard Michaelis-Menten equation to determine parameters k_cat and K_m

$$k_{obs}=\frac{k_{cat}\mathrm{\ast }\left[S\right]}{K_m+\left[S\right]}$$ (4).

Enzyme assays performed in the presence of AMA or propofol were identical to those above except that a single concentration of acylated peptide was used with increasing ligand concentrations. AMA and propofol were initially diluted in DMSO stock solutions, and as such, every activity assay performed in the presence of AMA or propofol also contained 0.4% DMSO. As a control, we determined that this amount of DMSO alone decreased SIRT2 activity by only ~5%. SIRT2’s activity at varying AMA and propofol concentrations were plotted versus the log₁₀ concentration of the inhibitor. A sigmoidal curve was fit to the data using Eq. 2. The reported IC₅₀ values represent the concentration of ligand required to inhibit 50% of the baseline SIRT2 activity in the absence of inhibitor.

Data statistics, structural analysis and figure preparation

All binding and activity assays were performed at least in triplicate, and all curves were fit with GraphPad Prism 7. Structural images were generated using PyMOL or VMD, and electron density data was downloaded from the Protein Data Bank.^31,32 Protein cavity volumes were calculated using fpocket.³³

RESULTS

AMA binds weakly to SIRT2

The structure of AMA and its chemical mimic propofol are shown in Figure 1A. In PBS, AMA can be excited at a broad absorption hump between ~325–430 nm, and its fluorescence emission peaks between ~500–650 nm.^18,20 The emission of AMA has higher intensity and blue-shifts when the ligand is in a nonpolar environment, such as a hydrophobic protein site that is shielded from the surrounding aqueous buffer.^18,24,34

Figure 1.

SIRT2 interacts with AMA and propofol at the same site. (A) Chemical structures of AMA and propofol. (B) Emission spectra of 100 nM AMA equilibrated with the indicated SIRT2 concentrations in PBS. The excitation wavelength was 380 nm. (C) Fluorescence emission intensities at 520 nm were taken from the data in panel B and plotted versus the SIRT2 concentration to derive a binding isotherm. (D) Emission spectra of 100 nM AMA equilibrated with 4 μM SIRT2 and varying propofol concentrations. Propofol caused a dose-dependent decrease in AMA fluorescence and a red-shift of the spectrum, which indicates AMA unbinding SIRT2 and returning to the aqueous solution. (E) Emission intensities at 520 nm were taken from the data in panel D to derive a dose-response curve for propofol’s reduction of AMA’s fluorescence. The IC₅₀ of 27 μM corresponds to a K_d of 25 μM for SIRT2’s interaction with propofol.²⁹ (F) SIRT2(Y139G) interaction with AMA was measured by changes in AMA fluorescence at multiple protein concentrations, as in panel C. In panel F, the K_d is approximate because all measurements were made with protein concentrations below the K_d, which is prior to significant curve saturation.

The fluorescence emission of 100 nM AMA was measured in PBS, and the intensity increased ~10-fold in the presence of 50 μM SIRT2 (Figure 1B). During titration experiments with increasing amounts of SIRT2, the emission spectra blue-shifted from a peak of 566 nm for AMA alone to 520 nm in the presence of SIRT2, indicating specific binding of AMA to a hydrophobic site.¹⁸ The fluorescence intensities of AMA at 520 nm in the presence of multiple SIRT2 concentrations were used to generate a conventional binding isotherm, from which we determined a K_d of 37 μM for SIRT2’s interaction with AMA (Figure 1C).

Next, we performed a competition assay to determine whether AMA binds the previously determined propofol site on SIRT2, which localized to residues Tyr139, Phe190, and Leu206.¹² For the competition assay, 4 μM SIRT2 was equilibrated with 100 nM AMA and increasing amounts of propofol, and fluorescence intensities were acquired (Figure 1D). Propofol dose-dependently decreased the fluorescence of AMA (IC₅₀ = 27 μM) (Figure 1E). The fluorescence decrease likely occurred because AMA was displaced by propofol from a common hydrophobic site on SIRT2, then AMA returned to the polar aqueous solution where its fluorescence was lower. We calculated a K_d of 25 μM for SIRT2’s interaction with propofol using Eq. 3 and the IC₅₀ from the binding competition curve.²⁹

Lastly, we expressed and purified a SIRT2 mutant with a single mutation in the propofol binding site (Tyr139→Gly139). AMA had reduced affinity for SIRT2(Y139G) compared to wild-type protein (K_d ~75 μM), further confirming the location of AMA’s binding site on SIRT2, which was presumably in the open conformation by itself in solution (Figure 1F).

Acyl-peptides affect SIRT2-AMA binding

The fluorescence of 100 nM AMA was measured in the presence of 4 μM SIRT2 and varying amounts of synthetic acylated peptides. The 13mer peptides had a sequence from Histone H4 (amino acids 8–20), and Lys16 was modified with an acyl group that was between two and fourteen carbons long (Figure 2A). This “H4K16” peptide was previously found to be a good substrate of SIRT2.² It was possible that AMA binding to SIRT2 could be prevented by the peptides, which would result in lower fluorescence intensities for AMA. We also reasoned that the peptides could promote AMA’s interaction with SIRT2 if the peptides caused structural changes to the protein that favored AMA binding.

Figure 2.

Varying effects of acylated peptides on SIRT2 interactions with AMA. (A) Chemical structures of lysine acyl modifications on H4K16 peptides. (B) Myristoyl-H4K16 peptide inhibited AMA binding to SIRT2, which was measured by decreased AMA fluorescence in the presence of peptide. The IC₅₀ of 3 μM corresponds to a K_d of 2.7 μM for SIRT2’s interaction with myristoyl-H4K16 peptide. (C) Inhibition of AMA binding to SIRT2 caused by dodecanoyl-H4K16 peptide. (D) Decanoyl-H4K16 peptide significantly enhanced AMA binding to SIRT2. Note the different y-axis scale on this and other panels. (E) Enhancement of AMA binding to SIRT2 mediated by 4-oxononanoyl-H4K16 peptide. (F) AMA’s affinity for SIRT2 in the presence of 10 μM decanoyl-H4K16 peptide was measured by its fluorescence change with increasing SIRT2 concentrations. (G) Propofol inhibited AMA binding to SIRT2 in the presence of 10 μM decanoyl-H4K16 peptide. The IC₅₀ of 8.6 μM corresponds to a K_d of 7 μM for propofol’s interaction with SIRT2 in the presence of 10 μM decanoyl-H4K16 peptide. (H) Decanoyl-H4K16 peptide enhanced AMA binding to SIRT2(Y139G), but to a lesser extent than wild-type SIRT2 (see panel D). (I) Acetyl-H4K16 peptide had no effect on AMA binding to SIRT2, and thus no change in fluorescence was observed. (J) Slight increase in AMA binding to 4 μM SIRT2 when the protein was equilibrated with 100 μM acetyl-H4K16 peptide, 300 μM nicotinamide, and 500 μM ADP-ribose. (K) Weak binding of AMA to 10 μM SIRT6 alone, and no change in binding when 300 μM ADP-ribose or 100 μM decanoyl-H4K16 peptide were added. For all panels, the AMA concentration was 100 nM.

The addition of myristoyl-H4K16 peptide to 4 μM SIRT2 reduced AMA fluorescence in a concentration-dependent manner (IC₅₀ = 3 μM) (Figure 2B). We calculated a K_d of 2.7 μM for the interaction of myristoyl-H4K16 peptide and SIRT2.²⁹ Our analysis of SIRT2 structures presented later suggests that the fourteen-carbon myristoyl group extended into the AMA/propofol binding site and competitively displaced AMA from the protein. Likewise, dodecanoyl-H4K16 peptide reduced AMA binding to SIRT2, but with slightly lower potency than the myrisoylated peptide (Figure 2C).

A striking enhancement of AMA binding was observed when SIRT2 was co-equilibrated with decanoyl-H4K16 peptide (Figure 2D). In that experiment, 10 μM of decanoyl-H4K16 peptide enhanced the peak fluorescence of AMA by 6.3-fold. For unclear reasons, the fluorescence trend was biphasic during this peptide titration experiment, and AMA’s fluorescence decreased at peptide concentrations greater than 30 μM. The fluorescence decrease might be caused by nonspecific interactions between the decanoylated peptide and SIRT2 that occurred after saturation of SIRT2’s peptide substrate site. This weak binding would not necessarily result in significant structural changes to SIRT2 because the flexible decanoyl chain might reach the hydrophobic AMA site from an alternative surface cavity. Similar to decanoyl-H4K16 peptide, we also measured a dose-dependent enhancement of AMA binding to SIRT2 mediated by 4-oxononanoyl-H4K16 peptide,⁴ although peak fluorescence was increased by only 3-fold (Figure 2E).

We further quantified the extent to which decanoyl-H4K16 peptide enhanced SIRT2-AMA binding. We measured the fluorescence of 100 nM AMA in the presence of 10 μM decanoyl-H4K16 peptide and increasing SIRT2 concentrations. The peak emission intensities were used to generate a binding isotherm which determined a K_d of 4 μM for SIRT2’s interaction with AMA in the presence of 10 μM decanoyl-H4K16 peptide (Figure 2F). For comparison, the K_d for SIRT2’s interaction with AMA in the absence of peptide was 37 μM (Figure 1C). Thus, AMA’s affinity for SIRT2 was enhanced by a full order of magnitude when SIRT2 was bound to decanoyl-H4K16 peptide. A subsequent competition assay with propofol determined that the decanoylated peptide also enhanced propofol’s affinity for SIRT2 (K_d = 7 μM) (Figure 2G), which was 4-fold stronger than propofol’s affinity for SIRT2 alone.

We sought to solidify that AMA interacted with the same SIRT2 residues in the presence and absence of decanoyl-H4K16 peptide. We measured the fluorescence of 100 nM AMA equilibrated with 8 μM SIRT2(Y139G) and increasing amounts of decanoyl-H4K16 peptide. We used 8 μM SIRT2(Y139G) here to make the assay quantitatively comparable to our previous titration experiments using 4 μM wild-type SIRT2, because in both cases 13% of AMA was bound to protein in the absence of peptide. AMA’s interaction with SIRT2(Y139G) was maximally enhanced only 2.2-fold by 30 μM decanoyl-H4K16 peptide (Figure 2H). The enhancement of AMA binding to SIRT2(Y139G) was small compared to wild-type SIRT2, suggesting the ligand’s binding site was not fully formed when the mutant protein interacted with the decanoylated peptide. The finding that higher peptide concentrations were required to enhance AMA binding to SIRT2(Y139G) compared to wild-type SIRT2 might also indicate that the decanoyl-H4K16 peptide had lower affinity for the mutant protein.

In contrast to the other peptides examined, AMA’s interaction with SIRT2 was not changed by the addition of acetyl-H4K16 peptide at concentrations as high as 100 μM (Figure 2I). The reported K_d for the interaction of SIRT2 with acetyl-peptides is in the 10–24 μM range,^4,35 suggesting that a large fraction of SIRT2 was likely bound by acetyl-peptide in our experiments. To potentially stabilize a closed conformation of SIRT2 bound to acetylated peptide,^13,36 we also added nicotinamide and ADP-ribose to the cuvette at concentrations that saturate the protein (see Figure Legend).³ These additional ligands did enhance AMA binding to the SIRT2-acetyl peptide complex, but only by a small amount (1.3-fold) (Figure 2J). We note that experiments with SIRT2’s co-factor NAD⁺ were hampered by its absorption of light in the same range as AMA’s excitation wavelengths.

Finally, we tested whether AMA binds SIRT6 to determine if the probe can interact with any sirtuin. SIRT6 is a very poor deacetylase but a strong deacylase, preferring medium to long acyl chains on its substrates including decanoyl modifications.^1,3,37 The fluorescence of 100 nM AMA was only 1.4-fold higher in the presence 10 μM SIRT6 compared to its fluorescence alone (Figure 2K). For comparison, 10 μM of SIRT2 alone causes an ~8-fold enhancement of AMA fluorescence, indicating a very weak or nonspecific interaction for AMA and SIRT6. ADP-ribose and peptide binding also facilitate conformational changes in SIRT6 and had no effect on AMA binding.³⁷ This measure of AMA specificity for SIRT2 adds to previous work showing propofol does not bind SIRT1,¹² which functions like SIRT2 as both an efficient deacetylase and deacylase.^1,3

Peptides induce structural changes in the AMA binding site

We analyzed SIRT2 crystal structures to further understand its interactions with peptides, AMA, and propofol. Binding assays determined that AMA and propofol bind the same residues whether or not SIRT2 is also bound to a peptide. Also, AMA and propofol preferentially interact with SIRT2 bound to peptides containing nine- or ten-carbon acyl groups, and peptide binding is thought to position SIRT2 into a closed conformation.^2,3,12,35

Crystal structures of SIRT2 bound to myristoyl-peptide clearly show the basis for competitive binding between the myristoyl group and AMA. The myristoyl group projects deep into SIRT2’s hydrophobic tunnel and terminates at a wall formed by the residues that interact with propofol (Tyr139, Phe190, and Leu206) (Figure 3A).¹² Occupancy of the hydrophobic tunnel by propofol and AMA would not be surprising given the small size and hydrophobicity of the ligands,³⁸ but there is not adequate space for the ligands and myristoyl group to co-occupy the tunnel.

Figure 3.

Peptide-induced structural changes in the AMA binding site. (A) Co-crystal structure of SIRT2 and myristoyl-peptide (PDB code 4Y6L). The myristoyl modification (green sticks) projects into the AMA/propofol binding site. Tyr139, Phe190, and Leu206 are shown in red spheres. (B) Decanoyl-peptide bound to SIRT2, modeled from PDB code 4Y6L. The decanoyl modification (green sticks) does not quite reach the back of hydrophobic tunnel, allowing the formation of a 182 ų pocket at the AMA/propofol binding site. The pocket is colored blue. (C) Dodecanoyl-peptide bound to SIRT2, modeled as in panel B, with a 102 ų pocket shown at the end of the tunnel. (D) Co-crystal structure of SIRT2, 4-oxononanoyl-peptide, and carba-NAD⁺ (PDB code 5G4C). The pocket at the AMA/propofol binding site is 224 ų in this structure, and for perspective, the pocket is 6 Å from the nearest carba-NAD⁺ atom. Carba-NAD⁺ would be positioned identically in all the peptide-bound structures. (E) Co-crystal structure of SIRT2 and trifluoroacetyl-peptide (PDB code 5FYQ). Select residues that form SIRT2’s hydrophobic tunnel are shown as sticks. Colored red are Tyr139, Phe190, and Leu206, and the electron density of these residues is shown as blue mesh. Note how Tyr139 is not packed with the other residues. In this panel and panels F and G, the contour level of the electron density maps is 0.5 sigma. (F) Co-crystal structure of SIRT2 and 4-oxononanoyl-peptide (PDB code 5G4C), with carba-NAD⁺ removed for clarity. Note the stronger electron density surrounding Tyr139 compared to panel E and its re-positioning near the end of the hydrophobic tunnel. (G) Co-crystal structure of SIRT2 and myristoyl-peptide (PDB code 4Y6L). Residues Tyr139, Phe190, and Leu206 fully form the back of the tunnel and are stabilized (i.e., have stronger electron density) compared to structures in panels E and F.

We modeled SIRT2 bound to a decanoyl-peptide by removing the four terminal carbons of the myristoyl chain from the crystal structure used above.³ In this model, a small pocket formed between the end of the decanoyl chain and the back wall of the hydrophobic tunnel (Figure 3B). The volume of this pocket was 182 ų, which is strikingly similar to the van der Waals volume of AMA (179 ų) and propofol (192 ų). The pocket reduced in volume to 102 ų with a dodecanoyl-peptide modeled (Figure 3C), which is not sufficient to accommodate AMA or propofol, consistent with our binding data (Figure 2C). The pocket was 224 ų in the crystal structure of SIRT2 bound to 4-oxononanoyl-peptide and carba-NAD⁺ (Figure 3D).⁴ We note that additional pockets which are sufficiently large for AMA could be identified in all the structures, but these unlikely binding sites are not adjacent to the residues that bind propofol.

In the crystal structure of SIRT2 bound to trifluoroacetyl-peptide,³⁵ the back wall of the hydrophobic tunnel was not fully formed. In part, this was because the side chain of Tyr139 was not tightly packed with its surrounding hydrophobic residues (Figure 3E). Such packing occurred in the presence of 4-oxononanoyl- and myristoyl-peptides because the longer acyl chains extended into this area and interacted with the hydrophobic residues, stabilizing their positions (Figures 3F and 3G).^2–4 This was reflected in electron density maps at the end of the tunnel, where stronger densities were seen when the acyl chain was longer (myristoyl-peptide > 4-oxononanoyl-peptide > trifluoroacetyl-peptide). The incomplete formation of the tunnel wall with trifluoroacetyl-peptide bound likely equates to incomplete formation of the higher-affinity AMA binding site, which explains why AMA binding to SIRT2 was not enhanced by interacting with acetyl-H4K16 peptide (Figure 2I).

AMA alters SIRT2 deacylase activity

We tested the ability of AMA and propofol to modulate SIRT2’s deacylase activities in HPLC-based assays performed under steady-state conditions.² We used synthetic peptide substrates that had an identical sequence to those used in the binding assays (H4K16), but the peptides also contained two tryptophan residues on the C-termini to facilitate detection with UV-Vis. We first determined SIRT2’s kinetic parameters K_m and k_cat for the peptides in PBS with NAD⁺ levels that saturated the protein (1 mM) (Figures 4A–4D).³ Our kinetic parameters were in reasonable range of published values for deacetylase, demyristoylase, and de-4-oxononanoylase activities, and the K_m values followed the expected trend (myristoyl < 4-oxononanoyl ≈ acetyl).^2,4 Surprisingly, SIRT2 was less efficient at removing the myristoyl modification on the H4K16 peptide compared to the acetyl modification (Figures 4A and 4D). This was unexpected because SIRT2’s demyristoylase activity was 5-fold more efficient than its deacetylase activity when histone H3K9 peptide was the substrate.² Enzyme efficiency therefore depends on the combination of substrate amino acid sequence and acyl modification. We also report that SIRT2’s de-decanoylase activity was remarkably efficient, with 9- to 28-fold higher k_cat/K_m compared to the other deacylase activities (Figure 4C). To our knowledge, kinetic parameters for de-decanoylase activity had not been reported previously under saturating NAD⁺ conditions.

Figure 4.

AMA and propofol differentially modulate SIRT2’s deacylase activities. (A) Kinetics of SIRT2 deacetylase activity using tryptophan-labeled acetyl-H4K16 peptide as the substrate. The buffer contained PBS with 1 mM NAD⁺ and 1 mM DTT, and reactions were run at 37°C. (B) Kinetics of SIRT2 deacylation of tryptophan-labeled 4-oxononanoyl-H4K16 peptide under conditions identical to panel A. (C) Kinetics of SIRT2 de-decanoylase activity using tryptophan-labeled decanoyl-H4K16 peptide. (D) Kinetics of SIRT2 demyristoylase activity using tryptophan-labeled myristoyl-H4K16 peptide. (E) Inhibition of SIRT2 deacetylase activity by AMA and propofol using the indicated peptide substrate and conditions identical to panel A. (F) Enhancement and inhibition of SIRT2 deacylase activity by AMA and propofol, respectively, using H4K16 peptide substrate containing a 4-oxononanoyl modification. (G) Inhibition and enhancement of SIRT2 de-decanoylase activity by AMA and propofol, respectively. De-decanoylase assays were run using a substrate concentration that approximated the K_m (left), and also under k_cat conditions (right bar graph). (H) Inhibition of SIRT2 demyristoylase activity by AMA and propofol.

We performed activity assays in the presence of AMA and propofol using peptide concentrations equal to SIRT2’s K_m for each substrate. AMA and propofol had inhibitory or stimulatory effects on SIRT2 depending on the acyl group on the H4K16 peptide. Propofol inhibited SIRT2’s deacetylase activity with an IC₅₀ of 140 μM (Figure 4E). AMA weakly inhibited the deacetylase activity, but 50% inhibition of SIRT2 was not achieved. This was probably due to the poor solubility of AMA, which was limited to 30 μM, whereas propofol was soluble in the 300–400 μM range.

AMA enhanced deacylation of 4-oxononanoyl-H4K16 peptide by 48%, while propofol had a weak inhibitory effect (Figure 4F). Interestingly, the ligands had opposite effects on SIRT2’s de-decanoylase activity, which was weakly inhibited by AMA but enhanced by propofol (Figure 4G). We also measured how AMA affected de-decanoylase activity under k_cat conditions because of the strong AMA binding we observed when SIRT2 was saturated with 10 μM decanoyl-H4K16 peptide (Figure 2D). Under k_cat conditions, 30 μM AMA inhibited SIRT2’s de-decanoylase activity by 22% (Figure 4G). Finally, AMA and propofol both inhibited SIRT2’s demyristoylase activity, with AMA being ~10-fold more potent with an IC₅₀ of 21 μM (Figure 4H).

DISCUSSION

Equilibrium binding of SIRT2 to AMA and propofol

We show here that AMA and propofol interact with SIRT2 at the same binding site, and that their affinity for SIRT2 is significantly enhanced when the enzyme is bound to a decanoyl-peptide substrate. Two significant structural changes occur when SIRT2 binds decanoylated substrate that are important for SIRT2’s interactions with the small molecules. First, the enzyme undergoes a major change from an open to closed conformation.^2,3 Second, residues that form SIRT2’s hydrophobic tunnel pack near the terminal end of the decanoyl chain and stabilize a small pocket that is optimally sized to accommodate AMA and propofol.

The conformational change that occurs when SIRT2 binds peptide substrate is not by itself sufficient to form the high affinity AMA/propofol binding site. If the substrate acyl chain is too long, it will occlude the ligand binding site in a closed conformation. Alternatively, SIRT2 can be in a closed conformation when bound to substrates containing small acyl groups, such as acetyl or trifluoracetyl, but these small modifications are not long enough to influence the residues that shape the end of the tunnel. Further supporting this are SIRT2 crystal structures where the enzyme forms a dimer in the closed conformation, and the dimerized SIRT2 molecules contain unmodified amino acids in their substrate binding clefts.^13,36 In these crystal structures, which also contain ADP-ribose and nicotinamide, the residues that comprise the end of the tunnel (including Tyr139) are highly flexible, and the side chains have insufficient electron density to be modeled in the crystal structures.^13,36

In the absence of peptides or other substrates, AMA interacted weakly with SIRT2 (Figure 1C), which was similar to previous work reported for propofol.¹² The ligand binding site in the open conformation must be structurally different than the higher affinity site in the closed conformation. However, mutation of Tyr139 to glycine reduced AMA’s interactions with both free and substrate-bound SIRT2, indicating at least some of the residues that form the two sites overlap. It remains puzzling to us what conformation SIRT2 was in when the AMA/propofol binding site was originally discovered.¹² That work showed strong binding of propofol and an analog to endogenous SIRT2 in a somewhat native environment (enriched brain myelin).¹² In those studies, binding was measured in myelin using 4 μM ligand and an endogenous SIRT2 concentration of ~0.3 μM.³⁹ The K_d values we determined in our current manuscript suggest that SIRT2’s structure in myelin must be constitutively altered in a way that promotes binding to propofol or AMA.

Effects of AMA and propofol on SIRT2 activity

AMA and propofol both affected SIRT2’s deacylase activities, indicating that their binding site is druggable. Indeed, several other ligands bind within this area including SirReals,³⁶ CHIC35,⁴⁰ and EX-243.⁴⁰ However, AMA and propofol are generally not potent molecules because of their weak affinities for SIRT2. This is expected for small molecules such as anesthetics that occupy hydrophobic protein sites with low energetic, non-electrostatic interactions.⁴¹

The combined binding and activity assays indicate inhibition of SIRT2’s demyristoylase activity by AMA and propofol through competitive binding with the myristoylated substrate (Figures 2B and 4H). With an IC₅₀ of 21 μM, AMA is especially unique among most small molecules that bind SIRT2 in that it inhibited its demyristoylase activity more potently than its deacetylase activity.⁵ AMA also promoted the activity of SIRT2 at removing a 4-oxononanoyl modification (Figure 4). Thus, it is possible to develop small molecules that selectively target certain acyl substrates of SIRT2. However, it will be challenging to identify new compounds that predictably alter SIRT2’s function through AMA’s binding site considering this site is orthosteric for myristoylated substrates, but allosteric for substrates with smaller acyl groups. This unpredictability was indicated in our deacylase assays using 4-oxononanoyl-peptide and decanoyl-peptide where AMA and propofol had opposite effects, being either inhibitory or stimulatory (Figure 4F and 4G). Although AMA and propofol have similar physical properties, their different pharmacology probably results from their unique molecular geometries and/or the asymmetries related to their hydrogen bond potential and pi stacking ability, which could affect their specific interactions with SIRT2.

We were initially surprised that AMA did not enhance SIRT2’s de-decanoylase activity considering the facilitative interactions between SIRT2, decanoyl-peptide, and AMA (Figure 2D). The apparent disconnect between AMA binding and its effects on activity continued with acetyl-peptide assays, where inhibition of deacetylase activity could not be predicted from our binding data (Figure 2I). Our binding experiments and structural modeling analyzed protein at equilibrium in a specific conformational state. Therefore, these approaches did not represent every step of SIRT2’s enzymatic turnover. If AMA facilitated SIRT2’s association with decanoyl-peptide, this promotion of SIRT2 activity could be offset at later step in SIRT2’s turnover, for example, if AMA slowed the release of the reaction product 2′-O-decanoyl-ADP-ribose.³ Additionally, AMA likely binds SIRT2 at every step throughout its enzymatic reaction, but AMA’s affinity for SIRT2 is different in the presence and absence of substrate, which suggests that the ligand binding site dynamically changes shape during SIRT2’s turnover. Alternative methods are required to understand the dynamics of ligand-protein binding and how they affect reaction rate.

Implications for drug discovery and disease

AMA’s fluorescent properties detected structural variability in SIRT2’s hydrophobic tunnel when the protein interacted with different acyl-peptides. It should now be possible to understand how the shape of the hydrophobic tunnel changes with different acyl chains bound, and to design or modify ligands specific for each pocket. Additionally, the fluorescence changes that occur when AMA is bound to SIRT2 can be exploited in screening approaches to discover new SIRT2 ligands that bind in a substrate-dependent manner. Molecules that alter select deacylase activities of SIRT2 will be important to understand the role of the diverse lysine modifications that occur in the cell, and will assign functions to specific modifications that may not be clear in SIRT2 knockout approaches. This may suggest diseases that can be treated with substrate-dependent SIRT2 modulators that have reduced toxicity compared to nonspecific SIRT2 ligands.