Reversible chemical modification of proteins through the introduction of post-translational modifications (PTMs) is an important biological mechanism for control of protein function. Intrinsically disordered proteins (IDPs) exhibit enhanced accessibility of modifiable residues in comparison with even the surfaces and loops of structured proteins. Due to the lack of stable tertiary structure to impart functional specificity, the chemical properties imparted by the primary structure of IDPs, including charge and local structures imparted by intramolecular interactions, are extremely important to their function (Bah and Forman-Kay, 2016). The influence of PTM-state on these properties allows for dynamic changes to the behavior of IDPs. For example, lysine N_ε acetylation results in both changes to the availability of hydrogen bonding functional groups and charge neutralization of the sidechain (Figure 1A).

Acetylation can modulate protein stability, activity, and complex assembly as well as interact with other PTMs to regulate cellular and developmental processes such signal transduction, cell cycle programming, and gene expression (Narita et al., 2019). Lysine Nε acetylation was first discovered to occur on disordered N-terminal tails of histone proteins (Allfrey et al., 1964), where it has been associated with increased chromatin accessibility (Shogren-Knaak et al., 2006). Lysine N_ε acetylation has since been found to occur on thousands of non-histone proteins including p53 (Sakaguchi et al., 1998; Liu et al., 1999), α-tubulin (L’hernault and Rosenbaum, 1985), and NF-κB (Huang et al., 2010). Thus, methods to characterize the varying molecular mechanisms whereby acetylation impacts local protein structure and biomolecular interactions are imperative due to the high biological prevalence of this regulatory mark.

Despite the urgency of better characterizing lysine N_ε acetylation, protein acetylation is currently difficult to study on a biophysical level due to experimental limitations. Biochemically, the installation of the acetyl group can be monitored using a radioisotope labeled source of acetyl coenzyme A (acetyl-CoA). The use of ³H or ¹⁴C labeled acetyl-CoA provides the benefit of high sensitivity and does not require the use of an exogenous tag but is encumbered by the standard limitations of working with radioactive materials (Allfrey et al., 1964; Nelson et al., 1980; Ait-Si-Ali et al., 1998). Mass spectrometry provides crucial insights into the site specificity of lysine acetylation, and in the proteomic assays that led to the observation of widespread acetylation in the proteome (Hebert et al., 2013; Silva et al., 2013); yet, mass spectrometry is a destructive technique, which can limit its applicability. The downstream impact of acetylation can be studied using techniques such as mutagenesis to glutamine or the use of an acetyllysine mimic, such as the introduction of a thiocarbamate modification (Huang et al., 2010). These techniques may not be applicable in cases where reversibility of the lysine residue’s acetylation status is biologically necessary. Further, non-native modifications have the potential to yield artifacts due to the impact of changing the chemistry of the sidechain (Nagaraj et al., 2012). To study deacetylation, peptide substrates have been conjugated to fluorescent dyes such as 7-amino-4-methylcoumarin, but this impacts deacetylase activity (Toro et al., 2017). Thus, it is desirable to develop a label-free, site specific, and non-destructive technique to monitor lysine N_ε acetylation, readout by downstream proteins or nucleic acids, and reversibility through enzymatic deacetylation.

Nuclear magnetic resonance (NMR) is an advantageous method for studying PTMs of IDPs as it is label-free and non-destructive (Liokatis et al., 2010; Theillet et al., 2012). Conventional heteronuclear 2D NMR, such as the [¹H, ¹⁵N]—HSQC, yields two observables that have been used to monitor acetylation events using ¹⁵N-labeled substrate. Firstly, whereas rapid exchange of the proton with solvent eliminates the lysine sidechain ¹H_ε-¹⁵N_ε resonances from most spectra, this NMR resonance becomes observable upon acetylation, due to the accompanying reduced rate of solvent exchange. Secondly, small changes in backbone chemical shift are often observable upon sidechain acetylation (Liokatis et al., 2010), yielding an indirect readout. These methods have a downside of requiring at the minimum ¹⁵N enriched peptide substrate. To fully benefit from site resolution, these experiments require a full determination of backbone chemical shift assignments. Thus, available techniques are characterized by relatively high cost in materials and time and require expertise that is often beyond many laboratories that might be interested in pursuing these studies.

The [¹H,¹⁵N]-HSQC experiments described above have traditionally been performed without isotopic enrichment of the acetyl group transferred to the substrate protein, yet there are major advantages to be gained by instead transferring a¹³C-acetyl group for direct observation. Previously, we have developed a strategy analogous to this using a¹³C-SAM methyl donor to provide an NMR visible signal to monitor lysine methylation events (Usher et al., 2021). Here we demonstrate enzymatic transfer of uniformly ¹³C-acetyl groups, either with or without concomitant isotopic enrichment of the substrate protein. As the acetyllysine moiety bears chemical similarity to the peptide backbone, we hypothesized that modifying ¹³C direct-detect NMR methods typically used for observation of IDP backbone residues would provide an unambiguous and easily interpreted signal corresponding to an acetylation event. Thus, we report here modifications to the ¹³C direct-detect CACO and CON experiments that yield optimized detection of the acetamide functional group created upon lysine N_ε acetylation. This technique has the capability to dramatically advance knowledge of structure and function impacts from protein acetylation.

Implementing the method described here relies on the enzymatic installation of a¹³C-enriched acetyl group to a peptide construct by a lysine acetyltransferase enzyme. To maximize utility, we provide two methods to produce ¹³C-acetyl CoA, an example protocol for preparation of NMR-quantity IDP, protocols for the production of a lysine acetyltransferase enzyme, and detailed protocols for appropriately implementing our NMR method. To demonstrate the utility of the methods on synthesized, natural abundance isotope peptide, we perform select experiments on H3 1-20 (Genscript). Unless use of this synthetic peptide is noted in the text, all reported spectra were collected on recombinantly expressed histone H3 (1-44).

Control of chromatin structure, promoter and enhancer accessibility, and subsequent levels of transcription at targeted loci through acetylation of multiple lysine residues within the Histone H3 N-terminal tail is one of the most extensively characterized examples of biological regulation by this ubiquitous post-translational mark (Tessarz and Kouzarides, 2014). Therefore, we have chosen to demonstrate the utility of our novel ¹³C direct-detect NMR strategy here using acetylation of H3 (1-44 or 1-20, as indicated) as a model system (Figure 1B). We first demonstrated acetylation of the disordered H3 1–20 by Ada2/Gcn5 using mass spectrometry for confirmation by means orthogonal to NMR. Two acetylation events were observed, corresponding to m/z of 2,225 and 2,268 (Figure 1C). The primary H3 lysine targeted by Gcn5 in vitro is K14 (Grant et al., 1999; Kuo and Andrews, 2013), although the Ada2/Gcn5 complex has previously been shown to also acetylate H3 at K18 in vitro (Balasubramanian et al., 2002). Acetylation of K14 and K18 was verified by MS/MS, thus authenticating our model system through established techniques.

Having verified that H3 (1-20) is acetylated on K14 and K18 by Ada2/Gcn5, we next collected NMR spectra of the modified proteins. Two methods were used to generate modified H3 1-20 tail: a two-step method relying on synthetic generation of ¹³C-acetyl CoA followed by acetyl-lysine modification of the H3 tail by Ada2/Gcn5, and a one-step method coupling enzymatic generation of ¹³C-acetyl CoA to acetyl-lysine modification of the H3 tail by Ada2/Gcn5. Control 1D spectra demonstrate successful transfer of acetate to CoA in both the organic and enzymatic routes (Supplementary Figure S1). [¹H, ¹³C]—HSQC of the ACS reaction demonstrated transfer of ¹³C-acetate to generate ¹³C-acetyl-CoA (Figure 1D). Using synthetically or enzymatically derived ¹³C-acetyl CoA as a substrate, the [¹H, ¹³C]—HSQC of ¹²C H3 yielded a new resonance at 1.87 ppm ¹H, 22.0 ppm ¹³C, which is distinct from the resonances from free ¹³C acetate and ¹³C-acetyl CoA (Figures 1E,F). Thus, traditional ¹H,¹³C-HSQC spectroscopy demonstrates the ease of detecting acetylated lysine residues by NMR.

While the [¹H, ¹³C]—HSQC spectrum has the advantage of a short collection time, under three minutes for the 350 μM H3 (1-20) samples reported here, ¹³C direct-detect NMR spectroscopy offers advantages for investigations of IDPs. For example, these spectra generally tolerate a wide range of solution conditions, including basic pH and high salt, that are often required to keep IDP samples soluble and stable, and are generally free from artifacts created in ¹H-detected spectra by stabilizing co-solutes (Bastidas et al., 2015). Therefore, we set out to optimize acetyllysine-selective ¹³C direct-detect 2D spectra analogous to those employed with wide success in backbone-centric spectroscopy. For all spectra, the¹³C-acetyl group provides the source of ¹³C nuclei, while isotopic enrichment to produce ¹⁵N labeled H3 (1-44) peptide expands the suite of available experiments to include ¹³C-¹⁵N correlation spectroscopy. Following pulse program optimization, the [¹³C,¹³C] CaliCO-Kac (Figure 2A) and [¹³C,¹⁵N] CON-Kac (Figure 2B) both yield excellent spectral quality, with only a single major peak present, corresponding to the selected N_ε acetyllysine. Note that because mass spectrometry had already confirmed the presence of both K14 and K18 acetylation on H3 (1-20), this suggests that the chemical shifts of the H3 acetyllysine sites are degenerate.

¹³C direct-detect spectroscopy is well known to require higher sample concentrations than ¹H detect methods, creating a burden on investigators who wish to leverage the other inherent advantages of this technique. For example, the data presented here were collected on samples of 800–1,000 μM H3 (1-44). Therefore, having demonstrated the utility of the [¹³C,¹³C] CaliCO-Kac and [¹³C,¹⁵N] CON-Kac experiments, we next aimed to increase their sensitivity by starting the magnetization transfer pathway on either the methyl or amide protons, instead of using the “protonless” ¹³C-start strategy of the foundational pulse sequences. For the [¹³C,¹³C] CaliCO-Kac, starting the magnetization transfer pathway from the methyl protons, followed by transfer to the methyl ¹³C, provided a 4-fold sensitivity enhancement for otherwise identical experimental conditions (Figure 3A). For samples that have been uniformly ¹⁵N-enriched, beginning the magnetization transfer pathway on the amide proton, followed by transfer into the carbonyl carbon for indirect chemical shift labeling is also an option. While this technique did yield a 2.5-fold sensitivity increase relative to an identically collected ¹³C-start spectrum, the amide-start version perhaps unsurprisingly underperformed the enhancement observed in the methyl-start experiment. In summary, the improved sensitivity of the ¹H-methyl-start [¹³C,¹³C] CaliCO-Kac should be advantageous for applications to future proteins with more strict solubility limits.

Having demonstrated improved [¹³C,¹³C] CaliCO-Kac in the proton-start formats tested, we next set out to optimize the [¹³C,¹⁵N] CON-Kac. Surprisingly, no increase in the signal-to-noise per scan was noted in the ¹H-amide-start [¹³C,¹⁵N] CON-Kac experiment, while the sensitivity of the ¹H-methyl-start experiment was diminished 2-fold relative to ¹³C-start (Figure 3B). While the peak intensity is mostly unchanged in the ¹H-amide-start spectrum, inspection of Figure 3B also clearly shows enhanced line broadening relative to the ¹³C-start experiment. Thus, while the integrated peak intensity is enhanced, relaxation losses during the additional spin echoes (especially for the ¹H-methyl-start spectrum), and those resulting from excitation of the ¹H-nucleus, offset the intended gains. This observation is consistent with our experiences using ¹H-start strategies for backbone detection, where the expected signal enhancement is on occasion lost to efficient relaxation. For other protein systems, it remains possible that ¹H-start CON-Kac spectroscopy will yield benefits. Despite the lack of sensitivity improvement realized here, these ¹H-start experiments may also prove advantageous to future methods development as the starting points for 3D spectroscopy that aims to characterize the structure and dynamics of acetyllysine residues more thoroughly.

Maximal spectral resolution, generally limited by the number of points collected in the indirect dimension of a 2D spectrum, is imperative because lysine- Cε chemical shifts fall in a relatively narrow range (Ulrich et al., 2008). Non-uniform sampling (NUS) can theoretically be used to improve spectral resolution in this case, as one can reconstruct a higher number of points in the indirect dimension in an experiment using NUS compared with a traditionally sampled experiment collected in the same amount of time. Although we do not report NUS data here, the utility of this strategy has been reported for other ¹³C-detect experiments, including those utilizing the IPAP virtual decoupling scheme (Bermel et al., 2009; Bastidas et al., 2015). Coupled with the inherent narrow linewidth in the ¹³C direct dimension, this may present an opportunity to resolve resonances that are partially overlapped. However, care must be taken if NUS is implemented for real-time kinetics or spin relaxation measurements, as it is known that NUS can lead to intensity artifacts dependent on the sampling scheme chosen (Maciejewski et al., 2011).

While the experiments described to this point require little to no isotopic enrichment of the target protein, it remains likely that, in many cases, investigators will want or need to produce uniformly ¹³C-, ¹⁵N-enriched protein samples, which may be valuable to acetylate. Interestingly, backbone-optimized [¹³C,¹⁵N] CON spectra do also yield resonances for the N_ε acetyllysine moiety, thus complicating spectral analysis (Figure 4A). Most importantly for the present application, the acetamide functional group differs in covalent structure from the polypeptide backbone in that the carbonyl is adjacent to a methyl carbon, as opposed to the α-carbon encountered in the backbone. Methyl groups can be selected for during the preparation period of a pulse sequence based on their spin topology granting access to a unique triple-quantum state, at a known cost of reduced sensitivity (Schubert et al., 1999). This property has previously been utilized to generate alanine-selective ¹³C direct-detect CON spectroscopy (Bermel et al., 2012; Sahu et al., 2014). Application of the alanine-selective pulse sequence to ¹³C-acetyl-labled H3 results in retention of both the alanine backbone resonances and the aceteyllysine (Figure 4B), likely because the transfer step from the methyl carbon to the carbonyl carbon refocuses both a one-bond and a long-range coupling. Indeed, further optimization to yield the methyl-selective [¹³C,¹³C] CaliCO-Kac suppresses the two-bond coupling from the alanine methyl to the backbone carbonyl. This results in a clean spectrum, presenting only the acetyllysine resonance, even when a uniformly ¹⁵N, ¹³C-enriched protein is used (Figure 4C). This strategy supports straightforward assignment of the acetamide resonance in a double-labelled sample generated for backbone chemical shift assignment and/or structure characterization, without the need to generate an additional NMR sample.

To demonstrate the biological utility of the methods described here, we have conducted NMR experiments that monitor Gcn5 bromodomain binding to acetylated histone H3. While bromodomains have a highly conserved structure, their acetyllysine binding pockets are lined with divergent amino acids across homologs, which may contribute to binding specificity. Additionally, bromodomain and extraterminal-domain (BET) family proteins are of interest for their application as pharmaceutical targets (Mujtaba et al., 2007; Cheung et al., 2021). Our results, presented in Figure 5, demonstrate the general potential for the methods described here to monitor binding and bound-state structure for bromodomain-acetyllysine interactions.

For these experiments, ¹H-methyl-start [¹³C,¹³C] CaliCO-Kac spectra were collected 1 mM ¹³C-H3 (1–44W), titrated with either .5 mM or 1.0 mM natural isotope abundance Gcn5 bromodomain. While the chemical shift of the methyl carbon remains unchanged within the resolution of the experiment, a clear chemical shift perturbation in the carbonyl chemical shift is observed upon overlay of the spectra (Figure 5). In the presence of sub-stoichiometric Gcn5, two resonances are observed, neither of which directly overlay with the unbound resonance position. Additional chemical shift perturbations are observed at equimolar concentrations. These data suggest that the binding interaction is in fast exchange on the NMR timescale. Intriguingly, the presence of two resonances in the Gcn5-bound spectra suggests that K14Ac and K18Ac may have distinct carbonyl chemical shifts in the bound state, although this hypothesis must be followed up with detailed future studies. The clear conclusion from these data is that ¹³C direct-detect methods yield chemical shift perturbations upon change of biophysical state and are well-suited to studying acetyllsine interactions with binding partners.

Lysine N_ε acetylation is a prevalent protein post-translational modification, but the lack of a convenient and label-free method for its biochemical and structural characterization has hindered development of molecular scale models for its function. Here, we demonstrate that the synthesis of ¹³C-acetyl CoA for enzymatic acetylation of lysine residues in proteins is a facile approach to generating samples suitable for NMR spectroscopy. Further, we report the optimization of seven pulse programs that directly detect the acetyllysine mark, yielding easily interpreted spectra. Most importantly, the samples required are virtually unmodified, with the addition of only 2–3 Da of mass and no extrinsic atoms, which supports application of this method to investigation of modified protein structure, interactions, and enzymatic deacetylation in a variety of biologically relevant contexts.

Synthesis of ¹³C-acetyl CoA can be accomplished through a synthetic or enzymatic route. The synthetic route performs best in a two-step workflow where the acetyl-CoA is fully synthesized prior to enzymatic acetylation of a target protein substrate. The synthetic route activates all available CoASH to acetyl-CoA but is only able to utilize half of the ¹³C-acetate functionalities present in the anhydride starting material, discarding the other half as ¹³C acetate (although this can be recycled). In contrast, ACS is able to scavenge ¹³C-acetate to activate CoASH and produce ¹³C-acetyl CoA. Further, this enzyme functions best when coupled in a one-step workflow where ¹³C-acetyl CoA synthesis and enzymatic acetylation of the target protein occur simultaneously. To assay diverse lysine acetyltransferases (other than Ada2/Gcn5), some optimization of solution conditions may be required if, for example, the acetyltransferase of interest displays poor solubility or activity under solution conditions that favor ACS activity. ¹³C-acetate is a significantly more cost-effective starting material than ¹³C-acetic anhydride, and the enzymatic route uses a larger fraction of available ¹³C atoms in the ¹³C-acetyl CoA product. Thus, the fully enzymatic route to protein ¹³C-acetylation can be highly efficient.

Traditional [¹H,¹³C]-HSQC spectroscopy is suitable to detect ¹³C-enriched acetyllysine but does have drawbacks compared to the new ¹³C direct-detect methods reported here. Specifically, disambiguation of the acetyllysine resonance from other peaks in the spectrum can be complicated. For example, the resonance from unincorporated acetate will remain in the spectrum, if it is not removed through buffer exchange prior to spectroscopy. Such a buffer exchange step would not be possible in the context of real time NMR monitoring of catalysis, which we have previously demonstrated to yield significant mechanistic insights for both serine phosphorylation (Gibbs et al., 2017) and lysine methylation (Usher et al., 2021). Further, many IDPs require organic co-solutes for solubility, stability, and to maintain a monomeric assembly state; in ¹H-detected spectroscopy, these co-solvents will often introduce additional resonances that may create baseline distortions if present in sufficiently high concentration. ¹³C direct-detect methods are highly resilient to co-solute addition.

The methods described here rely on preparation of a suitable acetyltransferase to generate the NMR sample, but they do not oblige the investigator to recombinantly prepare target proteins with isotopic enrichment. So long as ¹³C-enriched acetate is transferred, either [¹H,¹³C]-HSQC or [¹³C,¹³C] CaliCO-Kac spectroscopy can be employed on target protein from nearly any desired source. When isotopic enrichment is a possibility, [¹³C,¹⁵N] CON-Kac spectroscopy becomes available, yielding additional advantages. Specifically, because the detected carbon-nitrogen bond only exists in the acetylated lysine residue, and not in any precursor or biproduct of the reaction, its detection provides unambiguous evidence toward the installation of the acetyl group. Furthermore, the complexity of the selected spin system confers extremely high background suppression, which should support investigation of acetyllysine in the native nucleosome context, or even in the presence of cellular extracts.

In summary, we envision the biosynthetic and NMR spectroscopic techniques described here providing a simple, cost-effective, and versatile probe to study lysine acetylation in vitro in a broad range of contexts. Notably, these methods generalize well to any pairing of acetyltransferase enzyme and protein substrate; the selection of histone H3 and Ada2/Gcn5 was not required. If non-enzymatic means to install ¹³C-enriched acetate are employed, the NMR spectra described here will still provide an excellent means to characterize the resulting system. Alternatively, the enzymatic approaches described here also lend themselves well to preparation of isotopically enriched substrates for readout by mass spectrometry. Beyond monitoring the installation of acetyllysine modifications, we envision these techniques being used to monitor changes in the chemical environment of acetyllysine upon interaction with reader proteins, as demonstrated here through binding to the Gcn5 bromodomain, and for (potentially real-time) tracking of deacetylase activity. In the future, these methods can also be generalized to yield 3D spectroscopy to support the functional assays described here. Ultimately, wide use of these techniques should facilitate closure of the knowledge gap that currently exists between the list of known acetylated proteins and mechanistic insights into their functions.