Generating SucK- and GluK-modified recombinant proteins

At the outset of our studies, we decided to concentrate on site-specific incorporation of SucK and GluK via genetic code expansion, as MalK has been reported to be thermally labile due to decarboxylation25,26, and we expected the SucK and GluK derivatives bearing longer alkyl chains to be preferred substrates for PylRS and its evolved variants. To direct the site-specific incorporation of SucK and GluK via genetic code expansion, we synthesized a panel of SucK and GluK derivatives that mask the negative charge of the side chain carboxylate, thereby potentially improving cellular bio-availability and forming more favourable substrates for PylRS and its evolved variants. We focused on ester derivatives that would be stable under physiological conditions but cleavable post-translationally on protein by either enzymatic or mild chemical treatment. Different tested oxoester-masked SucK/GluK derivatives were either not substrates for tested PylRS variants or were resistant to hydrolysis after incorporation into POIs requiring very harsh acidic or alkaline treatment, neither of which is suitable for a wide variety of proteins. We therefore turned our attention to thioester-masked SucK/GluK derivatives.

We synthesized S-propyl thioesters of SucK/GluK by coupling succinic or glutaric anhydride to an appropriately protected lysine loaded on a solid support. Steglich esterification with propanethiol, followed by cleavage and deprotection, afforded multigram quantities of PrS-SucK and PrS-GluK (Supplementary Fig. 1b). Subjecting these ncAAs to a screening approach by co-expressing one of the approximately 200 distinct PylRS variants available in the laboratory, together with superfolder green fluorescent protein (sfGFP) bearing a premature amber (TAG) codon, identified a PylRS variant from Methanosarcina barkeri (Mb) that was able to charge PrS-SucK/PrS-GluK onto its respective tRNA (PylT). This MbPylRS variant bears mutations Y271A and C313V in its active site, as well as the previously reported IPYE mutations in its N-terminus27 and was dubbed ThioRS. We showed that ThioRS is efficient in site-specifically incorporating PrS-SucK/PrS-GluK into sfGFP bearing an amber codon at position 150 (Fig. 1c). Transferring the mutations from ThioRS to the PylRS variant from Methanomethylophilus alvus28,29 (Ma) and introducing further mutations that have been found beneficial for incorporation of ncAAs in Escherichia coli30 (MaThioRS with mutations Y126A, H227I and Y228P) showed that also this MaPylRS variant is efficient in encoding PrS-SucK/PrS-GluK (Supplementary Fig. 2).

Interestingly, analysis by mass spectrometry (MS) of purified sfGFP that was expressed in the presence of PrS-GluK revealed an approximately 1:1 mixture of GluK-modified sfGFP and the corresponding sfGFP variant with intact S-propyl thioester modification, indicating that the thioester was partially hydrolysed during protein expression and purification. This ratio shifted in favour of GluK-bearing sfGFP upon longer incubation in aqueous buffer at pH 7, and overnight incubation at elevated pH (pH 9) led to quantitative hydrolysis to yield pure sfGFP-N150GluK (Supplementary Fig. 3a–c). To accelerate hydrolysis and inspired by literature on hydrolysing peptide-α-thioesters, we incubated purified sfGFP at pH 7 with 10–100 mM β-mercaptoethanol (BME)31. We hypothesized that also on folded proteins, BME may transthioesterify the PrS-GluK thioester, followed by an (S,O)-acyl shift yielding the corresponding 2-mercaptoethanol oxoester of GluK, which should spontaneously decompose through intramolecular displacement of ethylene sulfide to give GluK-modified sfGFP (Supplementary Fig. 3d). Indeed, incubation with 100 mM BME at pH 7 led to quantitatively hydrolysed GluK/SucK-bearing sfGFP within 3–5 h (Fig. 1d and Supplementary Fig. 3e). In the presence of 5 mM tris(2-carboxyethyl)phosphine (TCEP), concentrations as low as 10 mM BME at pH 7 were effective for quantitative on-protein hydrolysis within a few hours (Supplementary Fig. 3f)32.

Recombinantly installed SucK/GluK are deacylase substrates

Studies on histones have revealed lysine succinylation and glutarylation in both the linker histone H1 and the four core histones H2A, H2B, H3 and H42,33. Site-specifically modified histone proteins are typically accessed by a combination of solid-phase peptide synthesis and native chemical ligation and have elucidated that glutarylation of K91 in H4, as well as succinylation of K77 in H4 and K122 in H3, lead to destabilization of nucleosomes34,35,36. The methods to access site-specifically succinylated or glutarylated histones and nucleosomes remain, however, elaborate and require expert chemistry techniques and are often inefficient. We therefore tested if we could leverage our approach to express H3 in E. coli with site-specifically encoded SucK and/or GluK at K122. This residue is situated in the so-called lateral surface region at the dyad axis of the nucleosome, where its side chain is in close contact to the negatively charged DNA wrapped around the nucleosome core particle (Fig. 2a and Supplementary Fig. 4a). Reversing the charge of this lysine residue from +1 to −1 by succinylation destabilizes nucleosomes, as shown by fluorescence resonance energy transfer experiments, utilizing reconstituted nucleosomes with chemically synthesized H3-K122SucK34.

Fig. 2: Generation of site-specifically succinylated and glutarylated histone H3.

a, Structure of the nucleosome core particle, with histone H3 highlighted in blue (Protein Data Bank, 1kx5)67. The modified lysine (K122, red) is located at the dyad axis of the histone–DNA interaction. b, SDS–PAGE analysis of purified histone H3 variants (WT H3, H3-K122SucK and H3-K122GluK). c, LC–MS analysis confirmed the integrity of purified histone variants H3-K122SucK and H3-K122GluK. Comprehensive LC–MS analysis can be found in Supplementary Figs. 4 and 16. d, SDS–PAGE and western blot analysis of deacylation assays with H3-K122SucK and H3-K122GluK using SIRT5. The full gels can be found in Supplementary Fig. 13. Consistent results were obtained over three distinct replicate experiments.

Source data

We recombinantly expressed H3 bearing PrS-SucK/PrS-GluK at K122 and confirmed specific incorporation and quantitative BME-induced hydrolysis to H3-K122SucK/GluK via MS (Fig. 2b,c and Supplementary Fig. 4b,c). To prove that genetically encoded SucK is indistinguishable from its endogenously installed counterpart, we incubated H3-K122SucK with SIRT5, a deacylase that has previously been shown to desuccinylate H3-K122SucK peptides. Both western blot analysis using a SucK-specific antibody and liquid chromatography (LC)–MS analysis confirmed that fully folded H3-K122SucK is a substrate for SIRT5 (Fig. 2d and Supplementary Fig. 4d)12,34. By incubation of glutarylated histone H3 with SIRT5, we furthermore identified SIRT5 as efficient deglutarylase for H3-K122GluK (Fig. 2d and Supplementary Fig. S4d).

SIRT5 has been reported as a promiscuous desuccinylase12. The enzyme has an atypical acyl-binding domain that prefers negatively charged residues, such as SucK, over Nε-acetyl-l-lysine (AcK)8,12. To test whether SIRT5 is able to desuccinylate also bacterial target proteins, we produced the bacterial oxidoreductase AzoR (flavin mononucleotide-dependent NADH:quinone oxidoreductase) bearing SucK at position 133 (ref. 6) and incubated it with SIRT5 (Supplementary Fig. 5a–c), showing that, in our in vitro assay, SIRT5 acts on bacterial succinylated proteins. The only identified deacylase in E. coli, the NAD+-dependent class III sirtuin, CobB shares sequence similarity with SIRT5 in its acyl-binding domain12, suggesting that CobB may be able to catalyse lysine desuccinylation8. Proteomic quantification of succinylation sites in wild type (WT) E. coli cells and CobB knockout cells indicated, however, that succinylation was not globally altered by loss of CobB6. Interestingly, in our in vitro assay, CobB was able to partially hydrolyse succinyl-lysine within AzoR-K133SucK (Supplementary Fig. 5d).

Glutarylation modulates the activity of metabolic enzymes

Having established a general tool to site-specifically succinylate and glutarylate various target proteins, we next set out to study the effects of these PTMs on enzyme activity and protein–protein interactions. We screened proteomic databases37 and chose interesting target proteins with reported succinylation/glutarylation sites either close to their respective active sites or in interactions sites where we assumed introduction of negative charges may have an impact on enzymatic activity4,5,6,7.

We first concentrated on the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH), a ubiquitous and essential enzyme that catalyses the conversion from d-glyceraldehyde-3-phosphate (d-GAP) to 1,3-bisphosphoglycerate in the presence of inorganic phosphate and the cofactor NAD+ (Supplementary Fig. 6a). In addition to this metabolic function, GAPDH has been implicated in several unrelated non-metabolic processes, such as control of gene expression and apoptosis38. There is evidence that this functional versatility may be regulated, at least in part, by PTMs that alter its catalytic activity and influence the subcellular localization of the enzyme38. Among phosphorylation, acetylation, ubiquitylation and redox-PTMs, GAPDH has been reported to be succinylated at various surface lysines6 and glutarylated at K1945, a residue close to the active site. In the cytosol, GAPDH forms a homotetramer that is stabilized by several hydrogen bonds between the individual subunits (Fig. 3a)39. Each monomer consists of a C-terminal cofactor-binding domain and an N-terminal catalytic domain harbouring the catalytic residues C152 and H179. An important structural feature of the catalytic domain is a long ordered loop called the S-loop (amino acids 181–209) that sits on top of the cofactor and is important for closing of the NAD+-binding site (Fig. 3a). This loop extends towards the adjacent monomer stabilizing interactions between subunits positioned across from each other in the tetramer. The amino acids in the S-loop are highly conserved, and we assumed that glutarylation of K194 situated in the middle of this loop might be important for GAPDH activity, although there are no direct interactions with cofactor or active site residues.

Fig. 3: Glutarylation of GAPDH at K194 regulates its enzymatic activity.

a, Left: Structure of the GAPDH tetramer with the four monomers shown in different colours (blue, black, teal and grey) in two different orientations. Interacting S-loop regions of two adjacent monomers (blue and black) are coloured in yellow and dark grey, respectively. K194 is shown in red, and NAD+ in pink (Protein Data Bank, 1znq)39. Right: Structure of the GAPDH monomer highlighting the cofactor-binding domain, the catalytic domain and S-loop. The modified residue (K194, red) is located in the S-loop (residues 181–209, yellow) in proximity to the NAD+-binding site. The catalytic residues H179 and C152 are shown in orange. b, SDS–PAGE analysis of the site-specific incorporation of PrS-GluK into GAPDH at position 194 (left, trnc. denotes truncated GAPDH) and of purified GAPDH variants (right). Consistent results were obtained over three distinct replicate experiments. The full gels can be found in Supplementary Fig. 14. c, LC–MS analysis confirmed the integrity of the purified GAPDH variants. The comprehensive LC–MS analysis can be found in Supplementary Figs. 6 and 16. d, Analysis of the enzymatic activity and kinetics of WT GAPDH and GAPDH variants according to Michaelis–Menten. Glutarylation of GAPDH at K194 or the GAPDH-K194E mutant as a charge-mimic reduced both Km and Vmax. The initial reaction velocity (v0) was plotted against d-GAP concentration and fitted with a Michaelis–Menten model to determine Km and Vmax values. Average values and errors (±s.e.m.) were calculated from three biologically independent experiments (n = 3). All data processing was performed using GraphPad Prism 10 (GraphPad software).

Source data

We expressed and purified human WT GAPDH and a GAPDH mutant bearing either glutamate (GAPDH-K194E) or GluK at position 194 (GAPDH-K194GluK) heterologously in E. coli. (Fig. 3b and Supplementary Fig. 6b,c). In addition, we also prepared a GAPDH variant bearing AcK at position 194 (GAPDH-K194AcK)40, as K194 was also found to be acetylated (Supplementary Fig. 6b,c)6,37. To determine enzymatic activity of the diverse GAPDH variants, we set up an assay to follow consumption of d-GAP by monitoring spectrophotometrically the fluorescence of NADH at 450 nm upon incubation of GAPDH with NAD+, arsenate and increasing concentrations of d-GAP to calculate maximal velocity (Vmax) and the Michaelis–Menten constant (Km) (Fig. 3d and Supplementary Figs. 6a,d, 7 and 8)41.

Importantly, thioester hydrolysis (incubation with BME or at pH 9) did not impact the enzyme kinetics of WT GAPDH (Supplementary Fig. 6d). Interestingly, the GAPDH-K194GluK variant showed a four- to five-fold reduction in both Vmax and Km, indicating that, indeed, glutarylation of this lysine residue impacts enzymatic activity. The glutamate mutant as a mimic for the negatively charged GluK showed similar results (Fig. 3d and Supplementary Fig. 6d). For the less bulky and uncharged AcK-GAPDH variant, on the contrary, we observed only 30–50% lower Km and Vmax values compared with WT GAPDH (Supplementary Fig. 6d). Introduction of a bulky group with negative charge into the highly conserved S-loop could either occlude access to the active site by hindering NAD+ binding or also impair interactions between the subunits thereby weakening the enzymatically active tetrameric structure of GAPDH.

Succinylation controls protein–protein interactions

We expected that acidic lysine acylations may be especially important for regulating protein–protein interactions, as they represent quick and reversible means to revert the lysine side chain charge from +1 to −1, apart from introducing also steric bulk. Given recent evidence that PTMs such as phosphorylation and acetylation on ubiquitin (Ub) enhance the complexity of the Ub code42,43, we wondered whether the lysine residues in Ub are also targets for succinylation/glutarylation. With the potential exception of K29, all seven lysine residues of Ub can be acetylated43, and it has been shown that acetylation at individual Ub lysines impedes E2/E3-mediated assembly of polyUb chains, as neutralization of the positive charges seems to affect non-covalent interactions of Ub with E2/E3-enzymes44. From system-wide proteomic approaches it is known that there is an extensive overlap of acetylation with succinylation on multiple prokaryotic and eukaryotic proteins6, and indeed, screening PTM databases revealed that Ub lysine residues are also modified by succinylation37,45,46. SucK has been detected on K6, K11, K27, K33, K48 and K63. We were especially interested in how these site-specific modifications might govern interaction with specific deubiquitylases (DUBs), as it has been shown that Ub phosphorylations affect DUB activity and specificity47.

Examining structures of linkage-specific DUBs with their respective diUb substrates revealed that an unusual surface of the proximal Ub around K33 makes close contacts with the S1′ site in the K11-specific DUB Cezanne (Fig. 4a)48. The direct interaction between residue E157 in Cezanne and K33 in the proximal Ub of K11-linked diUb impacts catalytic turnover of the enzyme and a K11-linked diUb with a K33E mutation in the proximal Ub showed reduced hydrolysis rates compared with WT K11-diUb48. As K33 has been reported to be succinylated as well as acetylated37,45,46, we expressed and purified Ub-K33SucK and Ub-K33AcK and assembled K11-linked diUbs with WT Ub in the presence of the respective E1 and E2 enzymes (UBE1 and UBE2S), as described previously49 (Fig. 4b and Supplementary Fig. 9). We prepared and purified WT K11-diUb (WT K11-diUb-H6), as well as a K11-linked diUb bearing SucK or AcK at K33 of the proximal Ub (K11-diUb(K33SucK)-H6 or K11-diUb(K33AcK)-H6, Fig. 4c and Supplementary Fig. 9). Incubation of SucK-modified K11-linked diUb with Cezanne led indeed to substantially reduced isopeptide bond hydrolysis compared with unmodified K11-diUb (Fig. 4d and Supplementary Fig. 10). Similarly, the K11-linked diUb variant bearing an acetylated lysine residue at position 33 showed impaired Cezanne-induced hydrolysis rates, although less pronounced than the succinylated variant. These data confirm that the direct salt-bridge interaction between E157 in Cezanne and K33 in proximal Ub is important for enzymatic activity of Cezanne and can be modulated by PTMs. (Fig. 4d,e and Supplementary Fig. 10). This reveals a regulatory potential of succinylation in repressing or potentially also increasing DUB activity, as has been shown for phosphorylation47.

Fig. 4: Succinylation of K33 in Ub modulates its interaction with the DUB Cezanne.

a, Structural insights into the interaction between Cezanne and a K11-linked diUb (Protein Data Bank, 5lrv)48. A salt bridge between K33 (red) of the proximal Ub and E157 of Cezanne stabilizes the interaction. b, A schematic representation of the generation of K11-diUb(K33SucK)-H6. c, LC–MS analysis confirms the integrity of the generated succinylated diUb. LC–MS analysis of WT and acetylated diUb can be found in Supplementary Fig. 9, and further comprehensive LC–MS analysis can be found in Supplementary Fig. 17. d, SDS–PAGE analysis of a DUB assay time course. K11-diUb(K33SucK)-H6, WT K11-diUb-H6 and K11-diUb(K33AcK)-H6 (all 5 µM) were individually incubated for the denoted time points with Cezanne (7.5 nM). Cezanne-mediated diUb cleavage is most severely impaired by succinylation of K33 in the proximal Ub. The full gels and quantification of bands can be found in Supplementary Figs. 10 and 15. Consistent results were obtained over three biologically independent replicate experiments. e, Model of Cezanne K11-diUb interaction displaying the E157-K33 salt bridge that may be modulated by post-translational modification of K33.

Source data

Succinylation regulates DNA–protein interactions

Arguably, acidic lysine acylations may exert the biggest effect on nucleic acid–protein interactions, as a charge reversion in a DNA/RNA-interacting protein from +1 to −1 may severely destabilize the corresponding complexes, as has been shown for chemically prepared SucK/GluK-modified histones and their destabilizing effect on nucleosomes16,17. Our approach of genetically encoding SucK/GluK has the advantage of being applicable to complex and non-refoldable proteins, including multidomain proteins. We were therefore interested in examining the effect of the reported succinylation sites within the DNA clamp proliferating cell nuclear antigen (PCNA) on DNA loading and replication.

PCNA is a ring-shaped, homotrimeric protein, essential to all living organisms, that plays a crucial role in DNA replication and in maintaining genome integrity50. As a molecular sliding clamp, it encircles DNA and assists in coordinating the interactions and activities of various factors involved in DNA replication and repair. PCNA typically exists as a closed ring and does not load onto DNA readily. Clamp opening and its loading onto DNA is facilitated by the clamp loader complex, replication factor C (RFC), a hetero-pentameric AAA+ ATPase complex51. The RFC complex binds in a slightly tilted way on top of a closed PCNA ring (Fig. 5a) and ATP binding by the RFC subunits promotes disruption of one of the PCNA subunit interfaces and thereby opening of the PCNA ring. The PCNA:RFC:ATP complex binds specifically to primed DNA and recognition of the double-stranded/single-stranded junction stimulates ATP hydrolysis coupled to conformational changes in the RFC subunits. These changes, in turn, decrease the binding affinity of the RFC subunits for the outer surface of PCNA and lead to expulsion of the clamp loader and eventual reclosing of the clamp around DNA (Fig. 5b). The inner surface of the sliding clamp is lined with positively charged α-helices that mediate contacts with the negatively charged DNA phosphate backbone. Recent findings have shown that these positively charged residues are important for proper RFC binding in yeast systems. Saccharomyces cerevisiae PCNA variants with lysine/arginine residues mutated to alanine within this inner surface showed reduced RFC ATPase activity in the presence of DNA52. PCNA’s function and interactions are heavily regulated and fine tuned by PTMs, such as ubiquitylation, SUMOylation, methylation and acetylation53. Interestingly, proteomics studies have also found two succinylation sites within PCNA: one on the outer surface of the ring (K164), a position that is ubiquitylated upon DNA-damage and triggers recruitment of specialized translesion synthesis polymerases, and one position (K13) within the inner ring surface that is packed with positively charged residues and may be important for proper RFC-mediated DNA loading6,37,46.

Fig. 5: Succinylation regulates hRFC-mediated clamp loading onto DNA.

a, Left: Structural insights into the PCNA:hRFC interaction. Right: Top and side view of the PCNA:DNA complex. Known succinylation sites (K13 (red) and K164 (yellow)) are highlighted (Protein Data Bank, 1sxj ref. 51 and 6gis ref. 68). b, Schematic representation of the hRFC-mediated charging of PCNA onto DNA. hRFC binds on top of the PCNA ring, and ATP binding triggers opening of PCNA, allowing loading of primer-templated DNA. Upon ATP hydrolysis, PCNA is released bound to DNA. c, SDS–PAGE analysis of purified PCNA variants. Consistent results were obtained over three distinct replicate experiments. d, Native PAGE analysis of WT PCNA and PCNA variants indicates intact trimerization for the PCNA-SucK variants. G178S is a previously reported PCNA mutant that is deficient in forming stable trimers and was used as control. Consistent results were obtained over three distinct replicate experiments. The full gels can be found in Supplementary Fig. 11. LC–MS analysis of the d