Revealing and exploiting the biochemistry of O-GlcNAc through protein semisynthesis

Abstract

O-GlcNAcylation is a dynamic post-translational modification regulated by the enzymes O-GlcNAc transferase (OGT) and O-GlcNAc hydrolase (OGA). It involves the attachment of N-acetylglucosamine to serine or threonine residues of proteins in the cytosol, nucleus, and mitochondria. As a dynamic and abundant modification, O-GlcNAcylation functions as a sensor of the cell’s metabolic state. Fluctuations in O-GlcNAc levels signal cellular stress or metabolic changes and have been implicated in various human diseases. The overall impact of this modification is protein-dependent, underscoring the importance of studying its biochemical consequences in a protein- and site-specific manner. To achieve this, enzymatic and chemical strategies have been developed to incorporate O-GlcNAc into peptides and proteins. These synthetic glycopeptides and glycoproteins have been instrumental in elucidating how O-GlcNAcylation influences protein structure, function, and diverse biochemical pathways. Recently, O-GlcNAcylation has also emerged as a tool for glycosylation-assisted folding of proteins and as a solubility tag for the chemical synthesis of glycopeptides and proteins. Here, we overview the current methods enabling the preparation of specific O-GlcNAc modified proteins and highlight recent developments.

Introduction

Protein glycosylation is an important post-translational modification that plays a crucial role in many cellular processes. O-GlcNAcylation is a special type of glycosylation that involves the addition of the monomer β-N-acetylglucosamine (GlcNAc) to serine and threonine side chains in intracellular proteins (Figure 1a)^1–4. This modification is very similar to phosphorylation in that they compete for the same residues on proteins and can also be installed and removed multiple times during a protein’s lifetime. The addition and removal of O-GlcNAc is performed by GlcNAc transferase (OGT) and O-GlcNAc hydrolase (OGA), respectively^5,6. Structurally, OGT is a 116.9 kDa protein that consists of the N-terminal and C-terminal domains. The N-terminal domain contains 34-residue tetracopeptide repeats (TPRs) necessary for mediating protein-protein interactions between enzyme and substrate. The C-terminal domain, which consists of the N-terminal (N-Cat) and C-terminal (C-Cat) regions, is responsible for the transferase activity. OGA is a member of the glycoside hydrolase 84 (GH84) family of glycosidases. OGA, a 102 kDa enzyme, is made up of an N-terminal with glycoside hydrolase activity, a stalk domain that contributes to protein-protein interactions, and a C-terminal pseudo histone acetyltransferase (HAT) domain. Through the combination of carbohydrate, amino acid, and fatty acid metabolism, UDP-GlcNAc, the donor sugar for OGT is produced biosynthetically through the hexosamine biosynthetic pathway. This ultimately links the global O-GlcNAc levels to the metabolic state of the cell and has been implicated in multiple diseases including cancer and neurodegeneration^7,8. The addition of this modification has been shown to impact the structure and biochemistry of a diverse range of proteins^9–11. Furthermore, O-GlcNAc has been exploited recently as a tool for glycosylation-assisted folding, a technique that draws inspiration from the cell’s protein folding machinery in the endoplasmic reticulum, and also as a tag during the chemical synthesis of glycoproteins that helps improve the synthetic efficiency and solubility of peptides while reducing the aggregation of the synthesized glycopeptides and glycoproteins.

Figure 1. O-GlcNAc modifications.

O-GlcNAc is the addition of the monosaccharide N-acetylglucosamine to serine and threonine residues of intracellular proteins. O-GlcNAc transferase (OGT) integrates a variety of metabolic inputs through utilization of its co-substrate UDP-GlcNAc, while O-GlcNAcase (OGA) removes O-GlcNAc and renders the modification dynamic.

Chemical tools have had large impact on the study of O-GlcNAc^12–16. Herein, we review the recent literature concerning the consequences of site-specific O-GlcNAcylation of proteins. First, we briefly describe the methods available for incorporating the modification into peptides and proteins. Next, we review recent studies utilizing these modified peptides and proteins to probe the biochemistry of OGT and OGA. Furthermore, we discuss recent works from our lab and others on the structure and function of disease-relevant O-GlcNAc modified proteins. Lastly, we review works focused on the use of this modification as a synthetic tag and protecting group in the synthesis of glycosylated peptides and proteins and as a strategy to aid in the folding of otherwise difficult-to-fold proteins.

Approaches for the preparation of O-GlcNAc modified peptides

A straightforward means of installing O-GlcNAc on peptides and proteins is by incubation with a recombinantly expressed and purified OGT (Figure 2). Many groups have utilized this strategy since Lubas and coworkers first described the recombinant expression of human OGT in E. Coli and established the kinetic parameters to be very similar to that of the native enzyme¹⁷. This strategy has been used to map O-GlcNAc sites on proteins and to provide mechanistic insights into OGT’s substrate specificity^18,19. An alternative mammalian OGT-substate coexpression system was developed in E. coli to overcome the difficulty associated with the expression and refolding of the enzyme (Figure 2). In this strategy, the authors showed that the coexpression of OGT with p53, sp1, and p62 leads to the O-GlcNAcylation of the substrates in cellulo²⁰. To optimize this system, Han and coworkers used a dual-plasmid strategy to improve the expression and modification efficiency²¹. Subsequently, they optimized this system by incorporating the GlmM and GlmU enzymes into the coexpression system to elevate the UDP-GlcNAc levels in the cells thereby improving O-GlcNAcylation efficiency of targets substrates²². As useful as the enzymatic approach has proven, its major disadvantage is the resulting heterogeneity of the modified substrates¹. The addition of O-GlcNAc is often substoichiometric in nature, leading to the presence of unglycosylated substrates in the reaction mixture. This can somewhat be circumvented with longer incubation periods, but at the risk of protein degradation or reduced expression. Additionally, OGT displays very little preference for any primary amino acid sequence, and multiple sites of O-GlcNAc on the same protein substrate are not uncommon. In coexpression systems, the installed modification can also be removed by endogenous glycosidases²³. Depending on the protein substrate, different numbers of O-GlcNAc modifications and/or sites of glycosylation can be difficult if not impossible to separate.

Figure 2. Enzymatic methods for O-GlcNAc modification.

Recombinant proteins incubated in vitro with OGT and UDP-GlcNAc can result in O-GlcNAc modifications, but these are often heterogeneous and difficult to separate. This process can be simplified by co-expression of OGT in bacteria; however, similar limitations exist.

To investigate site-specific effects of O-GlcNAcylation, homogeneously modified substrates are ideal. To this end, a synthetic GlcNAc-mimic developed by the Davis group has proven particularly useful (Figure 3a). In this strategy, a cysteine residue, incorporated in the target protein through mutagenesis, is converted to dihydroalanine (Dha) by treating with 2,5-dibromohexanediamide. The resulting Dha-containing protein is reacted with GlcNAc-thiol to give yield an S-GlcNAc modified protein²⁴. They showed that this modification on H2A histone affects its oligomeric association, nucleosome stability, and interactome. Our lab also previously established that S-GlcNAc is hydrolytically stable and can recapitulate the biophysical properties of O-GlcNAcylated α-Synuclein²⁵. The drawback to this approach is that it does not accurately capture the native state of the sugar modification as the linkage is through a thiol cysteine instead of hydroxyl serine or threonine. Additionally, the addition of the GlcNAc thiol to Dha is not inherently diastereoselective, often leading to a mixture of L- and D -products at the site of modification. To date, the only way to achieve site-specific O-GlcNAcylation while achieving homogeneity and capturing the native state of the modification is through native chemical ligation (NCL) or expressed protein ligation (EPL)^25,27. The traditional versions of these reactions take advantage of a selective reaction between peptide or protein thioesters and peptides or proteins bearing N-terminal cysteine residues through a transthioesterification followed by a rearrangement to yield a native amide bond (Figure 3b). NCL is limited to synthetic peptide thioesters, while EPL is an extension of NCL that enables the generation of protein thioesters through recombinant expression^28,29. In these ligation strategies, O-GlcNAcylated serine or threonine residues prepared through synthetic routes, including our facile one-step synthesis²⁶, are incorporated into a desired protein fragment through solid phase peptide synthesis (SPPS). NCL and EPL reactions, including an increasing range of ligation patterns beyond thioesters and cysteines, allows the modified peptide to be joined with other fragments to afford a modified full-length protein³⁰. For example, selenocysteine mediated ligations display improved reaction kinetics^104,31–33, particularly when paired with seleno- instead of thio-esters, enabling ligation reactions to be performed at lower concentrations. Additionally, selenocysteine can be readily transformed to alanine after ligation reactions, even in the presence of other cysteine residues. Advances such as this and others such as cysteine protection by maleimide during desulfurization reactions³⁴, continue to expand the range of protein targets accessible by ligation chemistries.

Figure 3. Methods for the site-selective incorporation of O-GlcNAc.

a) Posttranslational mutagenesis involves the conversion of a surface-exposed cysteine residue into a dehydroalanine (Dha) followed by addition of nucleophiles, including GlcNAc-thiol to create an O-GlcNAc analog. b) Native and expressed protein ligation chemistries take advantage of the unique reaction between a C-terminal thioester and N-terminal cysteine to combine fragments into a full-length protein. Solid-phase peptide synthesis can be used to install a site-selective O-GlcNAc modification for ligation.

Probing the molecular mechanisms of OGT and OGA using synthetic peptides

In mammalian cells, O-GlcNAcylation of proteins is achieved by the pair of enzymes OGT and OGA. To put in perspective, phosphorylation which is a modification that shares similarities with O-GlcNAcylation, is regulated by hundreds of kinases and dozens of phosphatases. This raises a very fundamental question in the field of O-GlcNAc biology of how the O-GlcNAc cycling enzymes select their substrates. As a fundamental approach to answering this, many structural studies have focused on the binding modes of glycopeptide substrates in attempts to establish a general substrate recognition sequence.

The first structure of a truncated version of human OGT, consisting of 4.5 TPRs and the catalytic domains, was determined in 2011 by Lazarus and coworkers³⁵. The authors reported OGT-UDP and OGT-UDP in complex with a CKII peptide. The peptide-bound structure revealed that substrates bind in an extended conformation within a cleft between the TPR and catalytic domains, explaining the enzyme’s preference for flexible region proteins. Attempts to solve an OGT-UDP-GlcNAc complex were rendered impossible due to the hydrolysis of the substrate. However, using computational docking, they established that GlcNAc moiety oriented itself in a way that places the anomeric carbon near the reactive serine. Finally, using structural and kinetic techniques, they established that OGT follows an ordered sequential bi-bi mechanism where the nucleotide sugar binds first followed by the substrate and identified His498 as the catalytic base. In a similar work, the structure of OGT in complex with UDP-5S-GlcNAc and a TAB1-derived peptide was solved³⁶. Contrary to Lazarus and coworkers, the authors concluded that both H498 and H558 were too distant from the active site to act as a catalytic base but rather the nucleotide sugar pyrophosphate acts as the base prior to acceptor substrate modification. This model also provided molecular basis for why UDP-5S-GlcNAc is a potent and specific OGT inhibitor. Although synthetic glycopeptide and structural biology have contributed immensely towards elucidating OGT’s mechanism of glycosylation, the basis for substrate selectivity remains elusive. Given that the enzyme is solely responsible for modifying thousands of proteins, it is not surprising that it lacks a rigid consensus recognition sequence. However, using libraries of synthetic peptides and in vitro modification, efforts have been made to characterize the general features of the modified peptides^37,38. These studies demonstrated that the efficiently modified peptides (>30% efficiency) bear the sequon [TS][PT][VT]S/T[RLV][ASY] at the −3 to +2 sites^39,40. Recently, the Int-D domain, along with aspartate and asparagine ladders located within the TPR lumen, has been shown to contribute to OGT substrate selection^41–43.

Concerning structural studies on OGA, the Jiang group reported the crystal structures of the truncated (OGA^cryst) and full-length enzyme (hOGA)⁴⁴. In this work, OGA in its apo form and in complexes with the inhibitor thiamet-G and a p53 glycopeptide were studied. The structures reveal that OGA forms an unusual arm-in-arm homodimer in which the stalk domain of one monomer sits on top of the catalytic domain of the other, creating a potential a substrate-binding cleft that provide spatial restrictions for larger substrates. Using a catalytically dead variant (D175N) to prevent sugar hydrolysis, the authors showed that the catalytic core engages in extensive hydrogen bonding interactions with the GlcNAc moiety, suggesting that the binding of GlcNAc is essential for substrate recognition. Furthermore, the hydrophobic residues lining the inner surface of the substrate-binding cleft forms sequence-dependent side chain interactions and sequence-independent backbone interactions with the glycopeptide. In a follow up study, OGA in complex with glycopeptides derived from αB-crystallin, TAB1, ELK1, and Lamin B1 were studied⁴⁵. Like their previous study, the substrates bind within the homodimeric cleft. Analyses of substrate binding revealed that the substrates bind in a bidirectional yet conserved manner. Extensive interactions were observed between the inner surface residues and the side chains and backbones, corroborating the previous findings that this network of interactions confer enzyme adaptability and specificity. Li and coworkers systematically explored the tolerance of OGA towards a panel of UDP-GlcNAc analogs⁴⁶. By generating glycopeptides bearing modifications at C2, C4, and C6 of the GlcNAc via OGT prior to subsequent hydrolysis by OGA, the authors showed that OGA displays stringent specificity for the native sugar, in contrast to OGT’s broad donor tolerance.

Elucidating O-GlcNAc biochemistry using protein (semi)synthesis.

In the past few decades, advances in mass spectrometry-based proteomics have led to the identification of thousands of O-GlcNAcylated proteins¹. Consequently, this has fueled efforts to elucidate the site-specific biochemical consequences of O-GlcNAcylation, most commonly using SPPS and EPL. Because the addition of GlcNAc to serine or threonine introduces steric bulkiness, one major area of investigation has been how this modification influences a protein’s capacity to engage with its binding partners. Recently, Lv and coworkers studied the effects of O-GlcNAcylation on the formation of liquid-liquid phase separation (LLPS) between SynGAP and PSD-95, two interacting proteins abundant in the pre-synaptic density of neurons⁴⁷. Using mass spectrometry and Y289L GalT1 chemoenzymatic labeling, the authors first confirmed multiple O-GlcNAc sites on SynGAP. Two of these sites, S1159 and T1306, were situated in the C-terminal PDZ-binding domain, a region crucial for interaction with PSD-95. They semi-synthetically constructed O-GlcNAcylated SynGAP at either residue and used biophysical assays to reveal that O-GlcNAcylation at T1306 significantly suppresses LLPS formation with PSD-95 relative to S1159 (Figure 4a). To elucidate the underlying mechanism, they used ITC and structural modeling to show that O-GlcNAcylation disrupts hydrogen bonding interaction between T1306 and His 369 of PDS-95. Lastly, they showed that modification can be reversibly cycled by OGA and OGT, where removal and addition restores and LLPS formation and inhibition, respectively.

Figure 4. Recent examples of O-GlcNAc biology uncove red using protein semisynthesis.

a) O-GlcNAc modification of the low-complexity domain of SynGAP reduces its ability to liquid-liquid phase separate. b) O-GlcNAc modification of the small heat shock protein HSP27 disrupts a regulatory interaction, promotes protein-refolding through BAG3/HSP70, and also inhibits caspase activation in apoptosis. c) O-GlcNAc modification of α-synuclein at S87 forces the formation of a different amyloid fiber structure that induces much less pathology in neurons and in vivo.

Our lab previously showed that heat shock protein 27 (HSP27) modified with O-GlcNAc at T184 (gT184) had superior chaperone activity compared to the unmodified form, mainly due to diminished interaction between the IXI and ACD domains⁴⁸. Notably, both domains also can mediate interactions with other proteins. To further elucidate how the modification remodels the HSP27 interactome in a cellular context, N-terminally biotinylated gT184 and unmodified HSP27 were incubated in HeLa cells. Analysis of binding partners revealed that gT184 had significantly enhanced interaction with co-chaperone BAG3, further emphasizing how this modification promotes HSP27’s interactions by leaving it in an “on” state (Figure 4b). BAG3 had been previously shown by Rauch et al. to promote ATP turnover and thus protein folding by HSP70 chaperones⁴⁹. In a follow up study, this same lab demonstrated that BAG3 can link sHSPs to HSP70 for “hand-off” and refolding of protein clients⁵⁰. In a refolding assay, we showed that HSP70 showed improved refolding of luciferase in the presence of HSP27 gT184 and BAG3 compared to the same mixture with unmodified HSP27⁵¹. Furthermore, label-free quantitative proteomics revealed that BAG3 showed preference towards HSP27 gT184 whereas unmodified HSP27 was biased towards αBc. This demonstrates how HSP27 O-GlcNAcylation not only directly enhance its chaperone activity, but also promotes functional integration with the BAG3/HSP70 machinery and modulates its interactions with other sHSPs. Existing evidence draws strong correlations between O-GlcNAc levels and various disease states⁷. O-GlcNAc levels are elevated in all cancer types examined, and we previously demonstrated that inhibiting UDP-GlcNAc biosynthesis improves the susceptibility of breast and lung cancer cells to oxidative stress-induced apoptosis⁵². Similarly, HSP27 a known inhibitor of apoptosis^53,54, also has elevated levels in cancers, raising the question of whether HSP27 O-GlcNAcylation has implications on this pathway. A recent work from our group compared the ability of HSP27 gT184 and unmodified HSP27 to inhibit the processing of caspase-3 by caspase-9⁵⁵. Caspases are cysteine proteases that initiate and carry out proteolysis events that lead to programmed cel death or apoptosis. The intrinsic apoptosis pathway begins with the release of cytochrome c from mitochondria, leading to the oligomerization and self-cleavage and activation of caspase-9. This protease then activates the executioner caspases, including caspase-3, which then cleave downstream targets. We found that HSP27 gT184 significantly delayed the cleavage and activation of caspase-3 by caspase-9 to a greater extent than the unmodified chaperone, possibly through a previously identified interaction with the pro-domain of caspase-3⁵³ (Figure 4b), emphasizing how the modification enhances some chaperone-client interactions.

O-GlcNAcylation has major implications in neurodegenerative diseases with many proteins involved in neurodegeneration being O-GlcNAcylated. Typical examples of such proteins are Tau in Alzheimer’s disease and α-Synuclein in Parkinson’s disease. We and others have found that O-GlcNAc generally inhibits the kinetics of amyloid protein aggregation of these proteins^56–60, a hallmark of neurodegenerative diseases. In more recent work, we showed that the modification alters the fibrillar strain, toxicity, seeding capability, and pathology of α-synuclein (Figure 4c)⁶¹. After building the protein using EPL, we used a variety of biochemical assays to show that α-synuclein gS87 forms fibrils with a distinct core structure compared to the unmodified protein. Although gS87 showed similar aggregation and seeding potential as the unmodified variant in vitro, gS87 preformed fibrils (PFFs) exhibited diminished seeding activity, propagation, and reduced Lewy body formation, in primary neurons and mouse models compared to unmodified PFFs. Interestingly, gS87 showed no decreased degradation, altered uptake, or reduced stability, highlighting that the difference in pathology is due cellular differences in seeded aggregation. Cryo-EM analysis of gS87 fibrils revealed new amyloid strains with unique inter-protofilament salt bridges.

Similar to O-GlcNAcylation, phosphorylation is also a dynamic modification that can be added and removed multiple times during a protein’s lifetime by kinases and phosphates. The presence of O-GlcNAc on the same or nearby residues as phosphorylation raises the possibility of cross-talk between these modifications to affect protein structure, function, and signaling pathways^62,63. Delineating the crosstalk between O-GlcNAcylation and phosphorylation can be done in mammalian cell-based assays, in which global O-GlcNAcylation or phosphorylation levels are perturbed^64–68, but these experiments can be complicated if mutation of one modification site (e.g., the same serine/threonine) will directly impact the occupancy of both O-GlcNAc and phosphate. This sets up cross-talk events as excellent problems to be solved using protein semisynthesis, which would enable the selective installation of either modification in a controlled fashion. This synthesis of reciprocally-modified proteins is less common in the literature; however, a few examples exist. In a first-of-its-kind study from the Cole lab, the roles of phosphorylation and O-GlcNAcylation in regulating casein kinase II were investigated⁶⁹. Using EPL, the authors generated CK2α proteoforms bearing phosphorylated T344 (pT344), a phosphonodifluoromethylene alanine phosphomimic (Pfa), and S-GlcNAc at S347. The authors chose Pfa and cysteine-linked S-GlcNAc as surrogates that would not be removed by endogenous phosphates and OGA, respectively. After subsequent microinjection of the prepared constructs into cells, they showed that pT344 stabilizes CK2α through binding with Pin1. Conversely, S347 O-GlcNAcylation antagonizes the effects of pT344 and promotes proteasomal degradation, highlighting how adjacent phosphorylation and O-GlcNAcylation can encode reciprocal regulatory information. Using a human protein microarray experiment, it was also observed that both modifications reprogrammed CK2α substrate selectivity. Finally, they demonstrated that the commonly used T-to-E phosphomimetic failed to recapitulate the observed phosphorylation-dependent phenotypes, underscoring the importance of semisynthesis in elucidating the roles of PTMs. Our lab subsequently collaborated with the Cole lab to use a similar strategy to examine phosphorylation and O-GlcNAcylation on the kinase Akt1⁷⁰, where both modifications have been identified on. Previously, the Cole lab showed that phosphorylation of S473 (pS473) by mTORC2 within the C-terminal tail activates Akt1 through interaction with the pleckstrin homology (PH)-kinase domain linker, thereby relieving PH-domain mediated Akt1 autoinhibition⁷¹. Additionally, Akt1 is also phosphorylated at Tyr474 (pY474)⁷² and O-GlcNAcylated at S473 (gS473)⁷³. Because S473 phosphorylation is known to be activating, its O-GlcNAcylation were thought to be inhibitory^73,74, although this possibility was not systemically explored. Using EPL, the effects of pY474 or gS473 on Akt1 activity were investigated. Contrary to literature expectations, pY474 and gS473 partially activate Akt1 towards a GSK3 peptide and promoted PRAS40 phoshorylation. In both cases, substrate phosphorylation was greater than the non-C-terminally modified Akt1 but lower than that achieved by the pS473 proteoform. Using a human protein microarray, PPM1H and NEDD4L were identified and validated as substrates for both pS473 and gS473 Akt1, with phosphorylation occurring in a concentration-dependent manner. Together, this study demonstrates a unique PTM cross-talk scenario that could not have been identified without protein semisynthesis.

Exploring O-GlcNAc as a structural support for in vitro folding and enhancing the solubility of glycoproteins and peptides

A significant challenge in the field of chemical protein synthesis is the aggregation and misfolding of intermediates under the aqueous ligation conditions. In vitro folding conditions lack the chaperones and PTMs that can stabilize folding intermediates and prevent them from going down misfolding pathways. Secreted or cell surface proteins are also transported to the Golgi apparatus and endoplasmic reticulum where they are glycosylated. This glycosylation event helps stabilize the folding intermediates and leads to the correct folding structures^4,75,76. Taking inspiration from this and the increased solubility of O-GlcNAc modified tau and α-Synuclein, groups have explored the use of O- and N-linked glycosylation to aid the folding of proteins otherwise difficult to fold and to probe the biological functions of these modifications (Figure 5)^77–80. The Liu lab recently developed the removable glycosylation modification strategy (RGM) which involves incorporating GlcNAc-modified S/T into a protein of interest through SPPS and NCL, prior to refolding. After refolding, the sugar moiety is removed by treatment with OGA⁸¹. This strategy is particularly useful for the synthesis of disulfide rich proteins, where the correct formation of the intra/intermolecular disulfides poses a significant challenge. Using this strategy, the authors successfully synthesized correctly folded hepcidin with improved overall yields. Impressively, they also reported the total chemical synthesis of Interleukin-5 homodimer with correctly formed intermolecular disulfide for the first time. Separate but similar works out of the Zheng and Li labs used chemical synthesis, NCL, and RGM to synthesize a panel of correctly folded and functional D-proteins; tumor necrosis factor α (D-TNFα), SARS-CoV-2 Omicron spike protein receptor binding domain (D-RBD), and the immunoglobin-like domain of tropomyosin receptor kinase A (D-IgC^TrkA)^82,83. These works expanded the applicability of the RGM strategy and showed that OGA retains hydrolase activity towards D-proteins, although it does so less efficiently relative to its activity towards the L-counterparts.

Figure 5. Non-native O-GlcNAc can be used to promote protein folding.

Synthesis of O-GlcNAc modified variants of difficult-to-fold proteins can increase their solubility and promote proper refolding after synthesis. Recombinant OGA can be used in a final step to remove the O-GlcNAc moieties.

The chemical synthesis and purification of peptide sequences rich in hydrophobic residues and many peptides relevant to neurodegeneration remain a significant challenge mainly due to their increased propensity to aggregate. This phenomenon not only exponentially diminish coupling efficiency during synthesis, but significantly slow down or abolish chemical ligation reactions by masking the reaction sites under common ligation conditions. Over the years, efforts towards solving this problem include peptide backbone and amino acid side chain modifications^84–86 and incorporating of solubility tags at peptide C-terminus^87,88. Solubility tags are typically short chains of either charged amino acids, like Gln, or one capable of hydrogen bonding like Arg that generally work by enhancing the hydrophilicity of the synthesized peptide. However, due to the irreversible nature of the incorporation methods, these tags can sometimes end up as part of the peptide, leading to a “scar” in the sequence that could potentially impact function⁸⁷. A growing class of reversible solubility tags that are installed on amino acid side chains and can be removed by either metal catalysis^89,90, photocleavage^89,91,92, or enzyme catalysis⁹³ has become more attractive.

Glycosylation has increasingly become an attractive reversible solubility tag due to its ability to be installed on multiple types of amino acid residues⁷⁶. In a recent work by Chen and co-workers, they developed the lysine-masked GlcNAc (LMG) strategy to improve the solubility of aggregation-prone glycoproteins and glycopeptides during chemical synthesis⁹⁴. The lysyl ester attached to the 6-OH of GlcNAc was added on S/N residues through a multistep synthetic route and incorporated into the desired peptide through SPPS. They showed that the tag is compatible with Fmoc-based SPPS and its incorporation greatly enhances the solubility of the glycopeptides compared to the non-glycosylated forms. In a final step the O-ester between the monosaccharide and the lysine can be selectively hydrolyzed to reveal the underlying native glycosylation. Using this strategy, the authors prepared multiply N-GlcNAcylated TMEM160B (residues 120–254). Their method also outperformed existing methods such as RBM⁸⁶ in terms of solubility enhancement. Common NCL conditions were sufficient to remove the lysyl group and reveal the native GlcNAc, although the rate of this hydrolysis is greatly influenced by pH. Furthermore, upon lysyl group removal, the revealed GlcNAc can be further elaborated by varying glycans to yield different glycoforms of PD-1 IgV domain (81–122). This strategy is highly compatible with existing SPPS and NCL conditions, which highlights its broad applicability.

Conclusions

Advances in analytical techniques have dramatically increased the number of known O-GlcNAcylated proteins, spurring extensive research into the cellular roles of this modification. Over the years, several groundbreaking methods have allowed the incorporation of the modification into relevant proteins to study its impact on protein structure, biochemistry, and functions. Synthetic protein methodologies such as SPPS, EPL, and NCL, have dramatically changed how PTMs are studied by enabling site-specific incorporation. In addition, using O-GlcNAcylated polypeptides to probe OGT and OGA has revealed mechanistic insights into the writer and eraser of this important cellular modification. Furthermore, exploration of O-GlcNAc beyond its biochemical effects has revealed its usefulness as a temporary support structure for chemical synthesis and refolding of difficult glycoproteins and glycopeptides. Together, these works highlight the impact and utility of protein semisynthesis to investigate protein PTMs, positioning O-GlcNAcylation as an important, dynamic, and versatile PTM. We believe there are also translational opportunities from this research. For example, several OGA inhibitors have been developed and are in different stages of clinical development for Alzheimer’s disease and related tauopathies. Our work on α-synuclein has laid the foundation for the exploration of these compounds in Parkinson’s disease, as evidenced by recent studies in mouse models, where OGA inhibition was found to be protective^95,96. Understanding the biochemistry of O-GlcNAc on relevant proteins should continue to expand the breadth of pathways and human diseases that may benefit from these types of inhibitors that have already passed safety trials in humans. The exploitation of O-GlcNAc as a protein-synthesis surrogate has the potential to expand the types of biological therapeutics that can be prepared by chemistry, opening up the possibility of total chemical control of peptide and small-protein drugs. We are excited to see how biochemical and technological advances enabled by the selective installation of O-GlcNAc into proteins expands the application of small molecule inhibitors and peptide/protein drugs moving forward.

However, there are also avenues for improvement in the field of using protein semisynthesis to dissect the roles of O-GlcNAcylation and other PTMs. For example, access to of O-GlcNAc modified amino acids remains limited, and the few that are commercially available are costly. Although synthetic approaches, such as the improved method developed by our lab²⁶, can help circumvent this limitation, they require substantial synthetic expertise and remain inaccessible to many researchers. Synthetic routes with commercial scalability will greatly improve the availability of these building blocks and drive down cost. Secondly, decreasing coupling efficiency as the peptide chain length increases limits the practical size of peptides that can be made by SPPS to approximately 50 amino acids⁹⁷. Additionally, so-called difficult peptides, which are rich in hydrophobic residues, further reduces the maximum achievable peptide length. The maximum achievable peptide peptide length can be greatly improved by flow synthesis protocols developed by the Pentelute lab^98,99. However, the higher temperatures involved in said protocols makes it unsuitable for O-GlcNAc modified amino acids due to the sensitivity of the glycosidic bond. Furthermore, flow synthesizers are not widely distributed and are expensive to build and maintain. Also, expression of fragments for EPL can sometimes be challenging, and when combined with the difficulties of glycopeptide synthesis, can impede optimization. Lastly, traditional protein ligation techniques require a cysteine at the ligation junction. However, with cysteines being one of the least abundant residues¹⁰⁰, methods that exploit alternative residues have been developed¹⁰¹. In these approaches, thiol-functionalized amino acids are incorporated at the ligation site and subsequently desulfurized to regenerate the native residue. Over the years, many desulfurization methods have been developed^102–105. However, these methods suffer from major drawbacks including poor yield, lack of robustness, and limited applicability. More recently, selenocysteine has emerged as an attractive alternative for protein ligation due to it enhanced nucleophilicity and ease of deselenization¹⁰⁶. Nevertheless, selenocysteine is not widely commercially available, can be difficult to handle, and again requires synthetic expertise. Development of more robust, improved, and widely applicable methods will significantly improve protein semisynthesis.

Table 1.

O-GlcNAc modified proteins prepared by (semi)synthesis

O-GlcNAcylated protein/peptide	Modification site(s)	Methods	Biochemical Effects	References
	O-GlcNAc Biology
α-synuclein	T72, T75, T81, S87,	EPL and desulfurization	Inhibits/alters amyloid aggregation	58, 59, 60, 61, 107–109
HSP27	T174, S176, T184, S187	EPL and desulfurization	Enhances chaperone acitivity, promotes luciferase refolding through BAG3/HSP70, inhibits caspase-3 cleavage by caspase-9, rescues some harmful CMT-2 mutations	48, 51, 55, 110
αA-crystallin	S162	EPL and desulfurization	Enhances chaperone acitivity	48
αB-crystallin	T162, T170	EPL and desulfurization	Enhances chaperone acitivity	48
SynGAP	S1159, T1306	EPL and desulfurization	Prevents protein-protein interaction with PSD-95, inhibits SynGAP LLPS	47
CK2α	S347	EPL	Antagonizes pY344 effects and promotes proteosomal degradation	69
Akt1	S473	EPL	Partially activates kinase acitivity	70
	Synthesis Enhancement
Hepcidin	S17	SPPS	Correctly formed intramolecular disulfides	81
IL-5 homodimer	T14, T23, T54, S73	NCL, OGA treatment	Correctly folded protein with correctly formed intramolecular disulfides	81
D-TNFα	D-S133, D-S147	NCL, OGA treatment	Correctly folded protein with correctly formed intramolecular disulfides	82
D-RBD (SARS-CoV-2)	D-T13, D-S34	NCL, OGA treatment	Correctly folded protein with correctly formed intramolecular disulfides	82
D-IgC (TrkA)	D-S312, D-S320	NCL, OGA treatment	Correctly folded protein with correctly formed intramolecular disulfides	83
TMEM106B (145–166)	N145, N151, N164	SPPS	Improved solubility	94
IL-2 (105–133)	S127	SPPS	Improved solubility	94
PD-1 IgV domain (35–145)	N49, N58, N74, N116, S137	NCL	Improved solubility of glycopeptide fragments, PD-L1 recognition	94