2.1 Sortases and Sortase-Mediated Ligation

Sortases are a family of cysteine transpeptidases that play crucial roles in anchoring proteins on the cell surface of Gram-positive bacteria.^{13, 21} Sortases specifically recognize secreted proteins bearing the respective sorting motifs and covalently attach them to positioned amino groups on the cell surface. As a result, sortases are involved in various cellular processes, such as peptidoglycan synthesis, pili formation, the assembly of multi-subunit hair-like fibers, iron acquisition, and spore formation.²² The transpeptidation function of sortase was first reported in 1999 through the study of a housekeeping sortase A from Staphylococcus aureus (SaSrtA, Figure 1a).²³ SaSrtA cleaves the amide bond between threonine and glycine within one polypeptide (acyl donor), generating a sortase-peptide intermediate. This generated intermediate is then intercepted by a glycine nucleophile (acyl acceptor), resulting in fusion of the acyl donor and acceptor in a high site-specific manner (Scheme 1).²³ To date, a superfamily of sortases are found in Gram-positive bacteria and are partitioned into six distinct classes (sortase A to F) based on their functions.²¹ The discussions of biological functions of sortases in Gram-positive bacteria exceed the scope of this review and we recommend other reviews for a more complete overview.^{13, 21, 22, 24}

Early structural studies revealed that SaSrtA harbors an eight-stranded β-barrel structure housing three active site residues: His120, Cys184, and Arg197 (Figure 1a).²⁵ Subsequent studies have demonstrated the conservation of this eight-stranded β-barrel core across all sortase classes (Figure1).²⁶ A wealth of evidence from mutagenesis and structural investigations has shown that the flexible loops that connect strands β6 and β7 (β6/β7 loop) as well as strands β7 and β8 (β7/β8 loop) play pivotal roles in modulating the activity and sorting specificity of sortases.²⁷ For instance, upon binding to the LPXTG substrate, SaSrtA undergoes adaptive recognition by closing and immobilizing the β6/β7 loop, while the β7/β8 loop undergoes conformational changes, unmasking the binding site for the tri-glycine nucleophile and thereby facilitating catalysis towards product formation.^25a Unlike the conserved eight-stranded β-barrel backbones, the length and configuration of loops of sortases, even within the same class, exhibit significant diversity (Figure 1). This diversity may serve as a primary reason for the variations in activity and sequence specificities among sortases.

SaSrtA requires Ca²⁺ for efficient ligation.²⁸ Ca²⁺ binds to a pocket situated between the β3/β4 and β6/β7 loops (Figure 1a). Analysis of NMR dynamics indicates alterations in mobility and a restructured β6/β7 loop.^{25a, 28} Both of these changes are believed to facilitate the accurate positioning of the LPXTG motif within the binding groove. Interestingly, this Ca²⁺-dependent regulatory mechanism has thus far only been identified in SaSrtA and not in other sortases.

The first proof of principle of protein ligation through sortase-mediated ligation was reported in 2004.³ Since then, SML has emerged as a powerful tool for protein modifications. Figure 2 provides an overview of the typical applications of SML, including protein/protein ligations,^{2a, 29} functionalization of protein termini,³⁰ head-to-tail cyclization of proteins,³¹ surface labeling of phages or cells,³² and immobilization of proteins onto solid or soft carriers.³³ Notably, SML has shown its general compatibility with various types of conjugations, extending beyond protein-protein ligations. Additional examples include the site-specific peptide-sugar,³⁴ peptide-nucleotide,³⁵ protein-lipid,³⁶ protein-amine,³⁷ protein-hydrazine,³⁸ protein-polymer,³⁹ protein-phage,^{32b, 40} protein-cell,^32a polymer-polymer⁴¹ conjugations and more.^{29b, 29c, 42} In this review, the primary focus is to present a protein engineer‘s perspective on the successfully employed approaches utilized for customizing the characteristics of sortases. These approaches are aimed at exploring new conjugation possibilities and/or meet the demands of further applications. Table 1 summarizes the protein engineering campaigns of sortases. Specific examples are highlighted in the following sections to illustrate their novelty and application impact.

2.2 Engineering for Enhanced Transpeptidation Activity

Prior to 2011, almost all the SMLs were conducted using SaSrtA wild-type (SaSrtA WT). A notable drawback of SaSrtA WT is the relatively low activity (k_cat ≤ 1.5 ⋅ s⁻¹).^{27, 64} Consequently, either extended reaction time or large quantities of the enzyme were often used to achieve substantial yields of conjugates. Engineering of SaSrtA for enhanced activity was highly envisaged to promote an efficient SML process. In 2011, Liu et al. reported the first directed evolution campaign for enhancing SaSrtA activity.⁴³ They developed an ultra-high screening strategy via a yeast surface display to evolve sortases, or in principle, any bond-forming enzymes. Notably, after nine rounds of evolution, variants were identified with up to 45-fold improvement in LPETG affinity (K_m reduces from 7.6 to 0.17 mM) and a 3.6-fold increase in turnover number (k_cat increased from 1.5 to 5.4 s⁻¹). Subsequently, the authors demonstrated that the significant improvement in LPETG affinity of evolved variants enabled efficient SML on the cell surface of living Hela cells.⁴³ Consequently, the penta-mutant (P94R/D160 N/D165 A/K190E/K196T, hereafter referred to as SaSrtA 5 M, Figure 3) has been used in many reports of SMLs.⁶⁵

The kinetics of single mutational variants in SaSrtA 5 M (e.g. D160 N, D165 A, and K196T) revealed moderate beneficial effects on turnover (k_cat) and more significant beneficial effects on LPETG motif recognition.⁴³ Structurally, the identified residues are all located around the LPAT-binding groove (Figure 3). P94 lies at the N terminus of H1 helix, while D160/D165, K190/K196 lie in the β6/β7 and β7/β8 loops, respectively. Furthermore, residues that are closely located at the N-terminus of the LPETG motif tend to be substituted to less changed or hydrophobic amino acid (e.g. D160 N, D165 A, K196T), suggesting the hydrophobic interactions with the Leu and Pro within the LPETG motif are favored. Taken together, the localizations and substitutions of residues in 5 M indicate the conformational changes in the important β6/β7 and β7/β8 loops may be improving both substrate binding and activity.

In 2016, Chen et.al. conducted a directed evolution campaign on SaSrtA 5 M with the aim of further improving its transpeptidation activity.⁴⁴ Through this campaign, several beneficial substitutions, including D124G, Y187L, and E189R, were identified and subsequently integrated into SaSrtA 5 M, resulting in a variant called SaSrtA 8 M (Figure 3). The k_cat of SaSrtA 8 M was further significantly improved to 22.2 s⁻¹, enabling more efficient ligations (versus WT or 5 M) at termini or endogenous lysine residue of proteins/antibodies.⁴⁴ D124 lies in the β4/H2 loop, and the k_cat of 5M-D124G is 4.5-fold higher than that of 5 M (16.5 s⁻¹ versus 3.7 s⁻¹).⁴⁴ It is hypothesized that the D124G mutation may alter the orientation of the β4/H2 loop, affecting intermediate formation and leading to enhanced enzymatic activity. Like the K190, the Y187 and E189 lie in the middle region of the β7/β8 loop (Figure 3). Kinetic results demonstrated that the recombination of Y187L/E189R to 5M-D124G enhanced the binding affinity both of the LPETG motif and the tri-glycine nucleophile. In a later study, the same group evolved SaSrtA to enhance its activity towards the N-terminal monoglycine proteins.⁴⁶ The resulting variant, named mgSrtA (Figure 3), has been successfully employed for intercellular labeling to monitor cell-cell interactions.⁴⁶

More successful evolution campaigns have been reported to enhance the activity of SaSrtA.⁴⁵ Notably, most of these substitutions are located within the loop regions, such as the β3/β4 loop;⁴⁵ the β6/β7 loop;⁴³ and the β7/β8 loop (Figure 3).^{43, 44} Considering the conservation of these loops among various sortases, applying the insights gained from engineering SaSrtA activity, particularly in these loop regions, is a reasonable approach to improving the transpeptidase activity of other sortases.

As previously mentioned, the low binding affinity for SaSrtA WT to LPETG motif (K_m > 5mM) is a primary factor that resulted in poor reaction kinetics. Rather than embarking on a directed evolution campaign to enhance binding affinity, Tsourkas and colleagues ingeniously addressed this challenge by devising a proximity-based sortase-mediated ligation (PBSL) approach.⁶³ In PBSL, SaSrtA is fused and expressed with the SpyCatcher protein (12.3 kDa) and subsequently immobilized on a cobalt resin. The protein of interest is expressed by appending an LPXTG motif and a short SpyTag (thirteen amino acids) at its C-terminus. SpyTag spontaneously reacts with the SpyCatcher, forming an irreversible isopeptide bond.⁶⁶ The SpyCatcher-SaSrtA resin is used to initially isolate the protein of interest from cell lysate via isopeptide bond formation. Subsequently, PBSL on the column is initiated by the addition of the desired peptide nucleophile and Ca²⁺. Due to the increased local concentration of the protein of interest in proximity to the tethered SaSrtA, PBSL proceeds rapidly and efficiently (> 95 % efficiency). Following the ligation reaction, only the product is released into the solution, while both SpyCatcher-SaSrtA and SpyTag remain bound to the resin. Notably, PBSL streamlines the SML by achieving protein purification, labeling, and tag removal in a single step, thus expanding SML applications.⁶⁷

2.3 Engineering of Altered Substrate Specificities

SML of proteins with broad substrate specificity, rather than relying on the canonical oligoglycine/LPXTG pair, is one of high interest in protein modification. The ability to join different substrate specificities offers a way to achieve multiple and orthogonal protein modifications. However, despite diverse motifs present in various sortases (Table 2), their activities are in general very low (k_cat in general ≤ 0.02 s⁻¹ except for SaSrtA)^{57, 68} and insufficient for applicable SMLs (low yield; long conversion times). Among the different sortases, SaSrtA stands out for its remarkably high activity level.⁶⁸ Instead of attempting to enhance the activity of a sortase with a different sorting motif starting from low k_cat values, numerous studies have attempted to alter the sequence-specificity of SaSrtA.

The specificity of SaSrtA at each position of the five-amino-acid sorting motif was thoroughly analyzed by McCafferty et. al.⁶⁹ The results demonstrated that the LPXTG motif is the most favored recognition motif of SaSrtA, but it is not completely stringent. Activity toward different motifs such as MPXTG, LAXTG, LPXAG, LPXSG, LPXVG, or LPXLG (Table 2), has been observed. The latter finding offers promising opportunities for protein engineers to alter or expand the substrate specificity/scope of SaSrtA, particularly through the engineering of loop regions such as the β6/β7 and β7/β8 loops.

McCafferty et al. reported the first specificity engineering campaign, wherein they swapped the β6/β7 loop from SaSrtB (recognizing an orthogonal NPQTN motif) into SaSrtA.⁵⁰ The resulting chimeric enzyme (SrtLS, Table 2) conferred the ability to recognize NPQTN-containing substrates and cleave the amide bond between threonine and asparagine, although it was unable to perform the transpeptidation stage of the reaction.⁵⁰

Another directed evolution campaign targeting the β6/β7 loop of SaSrtA for altered recognition motif specificity was reported in 2011.⁵¹ A phage display-based “panning” method was used to screen a simultaneously randomized library containing six amino acid residues on the β6/β7 loop. Finally, the variant F40 was identified (Table 2), which exhibited a comparable transpeptidase activity towards an orthogonal APKTG motif.⁵¹ Altered recognition motif specificity or substrate specificity represents a gain-of-function that was exploited for the traceless semi-synthesis of histone H3.⁵¹ Building on this work, a second library containing nine amino acid residues on the β6/β7 loop was generated and screened. This led to the generation of a variety of SaSrtA variants, such as SaSrtA A1-22 and SaSrtA F1-21, which displayed significantly enhanced activity for SML of APKTG/FPKTG and oligoglycine.⁵²

In parallel, Liu and colleagues reprogrammed the substrate specificity of SaSrtA 5 M through directed evolution using the developed yeast display screening platform.⁴³ They generated two families of SaSrtA variants that recognized orthogonal LAXTG and LPXSG motifs with comparable high catalytic efficiencies to the ‘standard’ LPXTG motif.¹⁶ The versatility of these evolved SaSrtA was demonstrated in site-specific modifications of endogenous fetuin A in human plasma, N- and C-termini dual functionalization of two therapeutically relevant fibroblast growth factor proteins (FGF1 and FGF2), and the orthogonal conjugation of fluorescent peptides onto different surfaces.¹⁶ More recently, the Liu lab used the yeast display screening platform once again to reprogram the specificity of SaSrtA to selectively recognize an endogenous LMVGG motif in Alzheimer's disease (AD)-associated amyloid-β (Aβ) protein.⁴⁸ They identified a sortase variant, SrtAβ, with a > 1400-fold change in substrate preference for LMVGG over LPESG motif. SrtAβ efficiently conjugates a hydrophilic peptide to Aβ, significantly delaying the initiation of detectable aggregation, which could contribute to the development of new AD treatments.

In another study, Amacher et. al reported another loop-swapping study for chimeric sortases.⁷⁰ They designed eight chimeric SrtAs by replacing the β7/β8 loop of Streptococcus pneumonia sortase A (SpnSrtA) with loops from other SrtAs from different species. They found some of these chimeric SrtAs retained the activity and substrate specificity of the wild-type protein from which the loop sequence was derived (e.g., SaSrtA), while other chimeric SrtAs (e.g., SpnSrtA_faecalis and SpnSrtA_lactis) can recognize a range of residues in the final position of the substrate motif such as LPATA, LPATF, LPATG, LPATL, LPATS, and LPATY (Table 2).

The screening of the acyl acceptor revealed that SaSrtA stringently recognizes protein with N-terminal glycine residue.⁶⁸ Interestingly, specificity engineering of acyl acceptor has barely been reported for SaSrtA. Sortase A homologs from other bacteria show a broad scope of acyl acceptor (Table 2).⁶⁸ A typical example is the sortase A from Streptococcus pyogenes (SpSrtA), which exhibits a comparable activity towards peptide with N-terminal alanine and serine in comparison to the canonical N-terminal glycine. Several studies have demonstrated dual modifications of proteins using SaSrtA and SpSrtA with orthogonal motifs.^{40, 71} However, the low activity of SpSrtA often requires large SpSrtA quantities. Enhancing the activity of SpSrtA for different acyl acceptors through engineering would therefore expand applications of SML.

Recently, we reported the first rational design campaign of SpSrtA.⁵⁷ After conducting a sequence and structure-based analysis, substitutions were introduced in positions near the active sites or within the β6/β7 and the β7/β8 loops. One variant, SpSrtA M3 (E189H/V206I/E215A) was identified and exhibited up to 6.6-fold improved activity towards N-terminal alanine or serine compared with wild-type (Table 2). Furthermore, the specificity of SpSrtA M3 for amine compounds was investigated. Unlike SaSrtA, which only recognizes unbranched primary amine at α-carbon,^37a SpSrtA M3 accepts both unbranched and branched primary amine as an acyl acceptor (Table 2).⁵⁷ This expanded specificity could facilitate peptide-amine biorthogonal conjugations, for instance in vivo protein labeling and cell imaging.

2.4 Engineering of Calcium Independency

Specific modifications of intracellular proteins hold significant potential in applications such as tumor therapies, tissue engineering, cell image, and biosensing.⁷⁷ The efficacy of SML lies in its high specificity and accessibility to the incorporated LPXTG or oligoglycine sequences, making it a broadly applicable tool for protein modifications. Importantly, its applicability extends beyond in vitro settings, as evidenced by emerging in vivo applications.^37a However, the most widely used sortase, SaSrtA, is dependent on calcium for its activity and typically requires a high Ca²⁺ concentration (≥ 3 mM) to be highly functional.^{14, 28, 78} This poses a significant challenge for in vivo SML since the cytoplasmic concentration of calcium ions is approximately 100 nM, significantly lower than the optimal required concentration.⁴⁷ Therefore, the development of SaSrtA variants with calcium-independent activity is crucial to broaden in vivo applications of SML.

The first study of Ca²⁺-independent SaSrtAs was reported in 2012.⁵³ Structural analysis revealed that four negatively charged residues, E105, E108, D112, and E171, are responsible for Ca²⁺ binding in SaSrtA (Figure 4a). Sequence alignment of SaSrtA with six other Ca²⁺-independent sortase A in the β3/β4 loop region demonstrated that K105 and Q108 (in the order of SaSrtA sequence) are highly conserved in all sortases except in SaSrtA (Figure 4b). Consequently, site-directed mutagenesis was performed for these two positions. Two variants, E105K/E108A and E105K/E108Q, were identified and exhibited approximately 35 % catalytic efficiencies compared to the SaSrtA WT (in the presence of 10 mM Ca²⁺) in absence of Ca²⁺ (Table 3).⁵³ Later, the identified substitutions at positions 105 and 108 were further recombined to SaSrtA 5 M and 8 M (Figure 3) and generated Ca²⁺ independent 7 M⁷⁹ and 7^+[80] variants. SaSrtA 7 M efficiently catalyzed the fusion of tyrosine kinases and maltose-binding protein in the cytoplasm of E. coli.⁷⁹ SaSrtA 7⁺ exhibited comparable activity in the presence or absence of Ca²⁺, surpassing the functionality of SaSrtA 7 M in both scenarios (Table 3).⁸⁰

In 2016, a compartmentalization-based ultra-high screening method was developed to evolve SaSrtA with high activity in absence of Ca²⁺ ions.⁴⁷ After 15 rounds of evolution, a variant R15-68 was identified, showing a 114-fold enhancement in k_cat/K_m (vs. SaSrtA WT) in the absence of Ca²⁺ (Table 3).⁴⁷ Interestingly, the substitutions E105V and E108A were identified in the variant R15-68, aligning with the previous study for calcium independency.⁵³ Additionally, substitutions P94R, D124N, D160V, E189K, and K196N were also obtained in the variant R15-68, which are in agreement with the previous studies for improved catalytic efficiency.⁴⁴ The intracellular activity of the variant R15-68 was further validated by catalyzing a head-to-tail cyclization of green fluorescent protein (GFP) in cytoplasm.

More recently, Wu et. al conducted an engineering campaign of SaSrtA for Ca²⁺-independent ligation activity specifically towards histone H3.⁵⁴ By deleting positive charges and introducing negative charges around the substrate binding pocket of SaSrtA 5 M variant, specifically R94S and A165D (which was reverted back to the wild-type), a hexa-mutant 6 M was generated. This 6 M variant exhibited a 13.2-fold improved catalytic efficiency for ligation of the histone H3L (26–39) peptide in comparison to the 7 M variant.⁸¹ Notably, the 6 M was successfully utilized in editing the H3L tail on native chromatin in isolated nuclei and living cells.

2.5 Engineering of Enhanced Robustness

SML has been studied for the labeling of proteins in batch and continuous-flow systems.¹⁵ Flow-based systems offer several advantages over one-pot SML, including the elimination of purification steps for LPXTG substrates, a low glycine nucleophile concentration to minimize the back reaction, and immediate product release. However, the storage and process robustness of SaSrtA remains a significant challenge for cost-effective applications in flow systems. Therefore, there is a need for engineered SaSrtA with enhanced robustness in respect to process conditions, such as elevated temperatures, protease resistance, and tolerance toward denaturing agents.

In 2014, the first study aiming to improve the resistance of SaSrtA against denaturing agents was reported.⁶² The researchers implemented a head-to-tail backbone cyclization of SaSrtA using an intein-mediated posttranslational modification. The cyclized SaSrtA, referred to as CsrtA_ΔN59, exhibited high resistance and activity in the ligation of peptides and proteins in the presence of up to 2.5 M urea.⁶²

In 2018, Grossmann et al. introduced an approach called in situ cyclization of proteins (INCYPRO) to stabilize SaSrtA through intramolecular crosslinks.⁶⁰ In detail, they introduced three substitutions D111C, E149C, and K177C to the sortase enzyme (referred to as SaSrtA S7) and achieved bicyclic stabilization upon reaction with a triselectrophile t1. The resulting bicyclic enzyme, named SaSrtA S7-t1, exhibited significantly improved thermal stability (melting temperature T_m=70.6 °C vs. SaSrtA WT (T_m=59.4 °C)) and increased enzymatic activity (8.7-fold) at 65 °C.⁶⁰ SaSrtA S7-t1 also demonstrated remarkable resistance to hydrolytic conditions, retaining 40 % activity in the presence of 1 M GdnHCl, whereas the SaSrtA WT was completely inactivated.⁶⁰ More recently, the authors further utilized INCYPRO to engineer the SaSrtA 8 M (Figure 3) and generate a bicyclized variant xS11, which showed a 12 °C improvement in T_m and high resistance under denaturing conditions.⁶¹

In 2020, we conducted an engineering campaign to enhance the robustness of a highly active variant SaSrtA rM4 (P94S/D160N/D165A/K196T).⁴³ The variant rM4 exhibited a 140-fold improved activity but showed a 10.8 °C reduced T_m compared to the SaSrtA WT (48.6 °C (rM4) vs. 59.4 °C (WT)).⁵⁶ The latter was attributed to the β6/β7 loop of SaSrtA, which is reported to have high mobility and be involved in substrate binding and ligation.^{25a, 49} To address the lowered T_m value, which often correlates to process robustness, we screened eleven single site-saturation mutagenesis libraries (SSM: 159, 161, 162, 163, 164, 166, 167, 168, 169, 170, and 172) lie in the C-terminus of β6 strand and the β6/β7 loop of SaSrtA rM4. Subsequent recombination of beneficial mutations resulted in a variant called M6 (rM4-R159N/K162P, Figure 5a). The M6 variant has a 6.0 °C increase in melting temperature (54.6 °C). Furthermore, by subjecting M6 to head-to-tail backbone cyclization, we generated a cyclized variant CyM6 with a melting temperature at 56.1 °C. In comparison to rM4, Cy6 harbored up to a 4.6-fold increase in resistance and gains up to a 2.6-fold increase in yield of peptide and primary amine conjugates under denaturing conditions.⁵⁶

As aforementioned, the relatively high K_m value (> 5 mM)^{55, 82} of SaSrtA often requires high concentrations of peptide nucleophiles employed in SML to achieve high bioconjugation yield. However, the latter can be limited by the solubility of peptides. To overcome this limitation, organic solvents such as dimethyl sulfoxide (DMSO) and dimethylformamide (DMF) are commonly added as co-solvents to improve the solubility of hydrophobic molecules. It has been previously reported that the activity of SaSrtA is significantly reduced at co-solvent concentrations ≥ 30 % v/v.¹⁴ In 2018, we carried out a directed evolution campaign to gain the resistance and activity of SaSrtA in high concentrations of DMSO (≥ 45 % v/v). Two SaSrtA variants M1 (R159G) and M3 (D165Q/D186G/K196V) with increased resistance (2.2-fold) and activity (6.3-fold) in 45 % (v/v) DMSO were identified (Figure 5b). We also tested the activity of these sortase variants in other co-solvents. Interestingly, M1 retained the high resistance profiles in DMF and methanol, while M3 showed remarkable enhanced specific activity (vs. WT) in 30 % (v/v) DMF, 30 % (v/v) ethanol, and up to 50 % (v/v) methanol.⁴⁹ Notably, the position R159, which lies in the C-terminus of the β6 strand (Figure 5a), has also been reported for contributing thermal stability in the previous study.⁵⁶ This finding suggests that mutations at position 159 may affect the flexibility/mobility of the β6/β7 loop. The D165Q is located at the N-terminus of the 3₁₀ helices (within the β6/β7), which forms upon the substrate binding. Additionally, D186G and K196 V are situated at the N and C-termini of the β7/β8 loop, respectively (Figure 5a). Collectively, these results suggest that alterations in these loops not only impact the activity but also the stability of SaSrtA. Proof of concepts was successfully reported for M3 by ligating a hydrophobic peptide and biohybrid conjugation of primary amines with up to 94 % conversion in the presence of DMSO (45 %, (v/v)) as a co-solvent.⁴⁹ These findings highlight the promising potential of SML for versatile bioconjugation in synthetic applications beyond natural aqueous solvents.

2.6 Engineering for Enhanced Activity in Lysine-Isopeptide Bond Formation

SML is usually restricted to: (1) modifications that take place at termini of the protein of interest, and (2) only a single molecule can be grafted to the protein harboring the recognition motif. In nature, Gram-positive bacteria such as Corynebacterium diphtheriae and Actinomyces naeslundii, assemble pilin monomers to pilin polymer by isopeptide bond formation catalyzed by sortase A.⁸³ Given that the sortase A-protein thioester undergoes a nucleophilic attack by the ϵ-amino group of a lysine residue in a pilin domain (WXXXVXVYPKH; where X can be any amino acid residue) (Scheme 2a).^83b

In light of these observations, Roy and his co-workers investigated the capability of SaSrtA in isopeptide SML.⁸⁴ By using a synthesized model peptide within the referred sequences, they showed that SaSrtA attached LPXTG peptide substrates to the side chain of lysine residues in the pilin domain and form cyclized and branched oligomers.⁸⁴ As a step forward for more efficient isopeptide ligation, engineering campaigns of Corynebacterium diphtheria sortase A (CdSrtA) were reported in 2018.^{58, 85} After obtaining a high-resolution crystal structure of the CdSrtA,⁸⁵ substitutions were introduced to unmask the lid of the active site. A variant ^CdSrtA^3M with triple substitutions (D81G/W83G/N85A) was finally identified and showed up to 10.5-fold improved yield (vs. WT) and 95 % conjugation efficiency in SpaA isopeptide modification (Figure 5b).⁵⁸ Noteworthy, the ϵ-amino group in lysine is a distinct acyl acceptor in comparison to oligoglycine. Therefore, ^CdSrtA^3M and SaSrtA were used to orthogonally label the ubiquitin-like modifier (SUMO) protein at its N- and C-termini in a one-pot reaction.⁵⁸ In further studies, the inhibitory polypeptide appendage was overcome by removing a lid that normally masks the active site. This resulted in a more active variant, ^CdSrtA^Δ, that showed more than a 3.6-fold increase in catalytic efficiency compared to the variant of ^CdSrtA^3M.⁵⁹

2.7 Reducing Reversible Reaction

SML is susceptible to a reversible reaction due to competition between the released product, which acts as a substrate-competing nucleophile due to the free primary amine. To address this challenge, Nagamune and colleagues introduced a rigid and stable β-hairpin structure around the LPXTG recognition motif, resulting in a product unrecognized by Sortase A (SrtA) (Scheme 2b).⁸⁶ They demonstrated that a fusion protein consisting of thioredoxin (Trx) and maltose binding protein (MBP), incorporating a GSKKWTWTWLPATGGWTWTWQESS-based β-hairpin, exhibited significantly enhanced resistance towards sortase A-mediated cleavage in the presence of an excess amount of triglycine. Meanwhile, the ligation of Trx-GSKKWTWTWLPATGG and GGWTWTWQESS-MBP resulted in a 70 % conversion to the fusion product, whereas the control only achieved a conversion of less than 50 %. In 2015, Antos et al. reported a metal-assisted sortase-mediated ligation (MA-SML) strategy to improve the ligation efficiency of SML.^87a This approach involved a minor modification of the substrate motif, replacing the last glycine with histidine (LPXTGGH). During the SML process, the resulting product GGH, which includes a Ni²⁺-binding peptide and an active nucleophile, was inactivated by the addition of Ni²⁺, thereby reducing competitive nucleophilic attacks and minimizing reversibility (Scheme 2c). By employing MA-SML, ligation conversions of up to 91 % were achieved using equimolar ratios of the glycine nucleophile. Recent developments in MA-SML have further expanded its versatile applications for the ligation of diverse cargos with high yields.^87b

In addition to the previously mentioned strategies for reducing reversibility, several other methods have been explored to inhibit or minimize reversibility in SML. However, these methods often require the use of non-natural substrates (see next chapter).

2.8 Development of Unnatural Sortase-Mediated Ligations

The initial investigation into unnatural SMLs employing unnatural substrates was conducted in 2012.^88a The authors synthesized an array of depsipeptides bearing an LPET motif and used them as substrates for N-terminus labeling of proteins through SML. The resulting side-product, alcohols, are of poor nucleophilicity for sortase and therefore the reverse reaction in SML is minimized (Scheme 3a). As demonstrated, the SML reaction is of high efficiency and its widespread utility for the N-terminal modification of proteins was reported.^88b In addition to the remarkable ligation efficiency, this research also provided evidence that sortases can recognize unnatural motifs beyond the typical amide bond backbone structure. This observation highlights the potential of SML for a novel field of SML with noncanonical ligations purposes.

In 2014, an irreversible ligation of modified substrates (LPETGG-isoacyl-Ser and LPETGG-isoacyl-Hse, Scheme 3b) to amine nucleophiles of interest with SML was reported.⁸⁹ The irreversibility was achieved due to the concurrent inactivation of the SrtA-excised peptide fragment through diketopiperazine

Liu et al. introduced a novel and irreversible SML process through hydrazinolysis of proteins (Scheme 3c).³⁸ The finally generated protein hydrazide is no longer recognizable for sortases and therefore results in an irreversible reaction with minimized hydrolysis. Under optimized conditions, yields > 95 % with ‘protein hydrazides’ were achieved within 1 h. The utility of this process was exemplified in the semi-synthesis of deoxy-D-ribose 5-phosphate aldolase (DERA), the functionalization of ubiquitin, and the fluorescent labeling of anti-EGFR (epidermal growth factor receptor) nanobodies. Of particular interest, SaSrtA WT exhibited comparable activity in hydrazinolysis of LPXAG compared to the LPXTG motif while previous studies indicated only partial activity in conjugation to LPXAG using triglycine as the nucleophile.⁶⁸

More recently, Liu et al. reported another irreversible SML method called thioester-assisted SrtA-mediated ligation (Scheme 3d).⁹⁰ This technique efficiently enables bioconjugation between a protein thioester and a N-terminal Gly protein. Notably, the method demonstrates tolerance towards a wide variety of LPXT-derived sequences as substrates, thereby overcoming the two primary limitations of the SrtA technique: irreversibility and strong preference for LPXTG motif.⁹⁰ The availability of peptide/protein thioesters through many recombinant methods or chemical synthesis further facilitates the thioester-assisted SrtA-mediated ligation, enabling the efficient functionalization of proteins from both termini with expanded recognition motifs beyond canonical LPXTG. These advancements solved the yield/back reaction challenges in SML.

3.1 Combination of Sortase-Mediated Ligation and Click-Chemistry

Click chemistry reactions, such as Huisgen cycloaddition and Staudinger ligation, which incorporate small abiotic functional groups, are well recognized for their remarkable reaction efficiency, unparalleled chemoselectivity, and biocompatibility with mild reaction conditions. These features make click chemistry a valuable tool for biomolecule functionalization, particularly for proteins, both in vitro and in vivo.⁹⁷ The emergence of click chemistry led to the development of biorthogonal chemistry which was jointly awarded the Nobel Prize in Chemistry in 2022. The combination of bioorthogonality in click chemistry with the sequence specificity of SML provides exceptionally powerful tools for protein bioconjugation. In 2011, Ploegh et al. first demonstrated the application of SML for grafting click handles onto the N- or C-terminus of proteins. Subsequently, these clickable proteins were conjugated via a strain-promoted cycloaddition in an aqueous environment at neutral pH, resulting in the formation of noncanonical N-to-N and C-to-C fused dimers.⁹⁸ The same sortase-click strategy was then further used to generate a C-to-C-fused antibody with bispecific anti-influenza virus activity and many other applications.⁹⁹

Another notable application of the sortase-click strategy was carried out by Kondo et.al.¹⁰⁰ They developed a hydrogel by combining streptavidin and branched poly(ethyleneglycol) to efficiently immobilize biomolecules. In their approach, tetrameric streptavidin molecules containing four LPETG tags were linked to azido-GGG peptides using sortase A. Subsequently, by introducing dibenzylcyclooctyne-modified branched poly(ethyleneglycol) and azido-modified branched poly(ethyleneglycol) as flexible spacers, a streptavidin-based hydrogel was formed through copper-free click chemistry. By leveraging the strong interaction between streptavidin and biotin, the hydrogel functioned as a carrier for immobilizing biotinylated proteins. To showcase its practicality, the researchers immobilized glucose dehydrogenase and applied a coating of this gel onto a glassy carbon electrode for glucose fuel cells.¹⁰⁰

Recently, we have developed a versatile toolbox for surface functionalization of a broad range of materials, including polymers, metals, and silicon, using the sortase-click strategy.¹⁰¹ In this study, an evolved sortase variant M3 was used to ligate the hydrophobic click reagent dibenzocyclooctyne (DBCO) to the C-terminus of a universal material binding peptide called LCI. Subsequently, LCI-DBCO adhered to different materials such as polypropylene, polystyrene, gold, stainless steel, and silicon. By conducting the click reaction between surface-bound LCI-DBCO and a fluorescent Cy-3 azide, the samples coated with LCI-DBCO exhibited up to 3.9-fold higher fluorescence intensity compared to the corresponding controls.¹⁰¹ These results highlight the versatility of the surface functionalization approach, paving the way for novel coatings with unique properties. The combination of material-specific binding peptides and sortase conjugation opens up a new field for sortase-mediated functionalization of materials/surfaces decorated with material binding peptides.

3.2 Combination of Sortase-Mediated Ligation and Unnatural Amino Acid Incorporation

Incorporation of UAA-bearing functional groups at specific positions in a target protein offers a powerful tool for deciphering biorthogonal chemistry. Leveraging the capabilities of UAAs, significant advancements have been made in site-specific protein labeling and functionalization, both in vitro and in vivo systems.¹⁰³ The combination of UAAs and SML techniques for site-specific protein engineering holds significant potential, but its full power has not been fully realized. In 2019, the Lang group reported the combination of UAA, bioorthogonal Staudinger reduction, and SML to develop a versatile tool termed Sortylation for the controlled ubiquitylation of proteins (Figure 6b).¹⁰² Specifically, they performed site-specifical incorporation of a UAA, AzGGK, into target proteins using genetic-code expansion methodology. Subsequently, the AzGGK side chain, carrying a terminal azide moiety, was converted in vivo to GGK through Staudinger reduction. The resulting GGK, bearing a sortase-accepted primary amine moiety was then ligated to an LPLTG- or LALTG-tagged ubiquitin using the evolved SaSrtA 5 M variant (Figure 3).⁴³ The versatility of Sortylation was shown in ubiquitylation/SUMOylation of different proteins in living cells such as E. coli and HEK293T cells. Based on Sortylation, the Lang lab further developed a modular toolbox, namely Ubl-tools, to stepwise generate polymeric ubiquitin architectures using evolved SaSrtAs variants that recognize orthogonal motifs.¹⁰⁴

3.3 Combination of Sortase-Mediated Ligation and CRISPR-Based Genome Editing

In the past decade, we have witnessed the impact of CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) in biological research and medicinal applications.¹⁰⁵ A prerequisite to use SML is the presence of a sortase recognition motif such as LPXTG within the target protein. For this reason, in most cases, mutagenesis methods are required for the incorporation of the motif at gene level. CRISPR technology offers several significant advantages over traditional PCR-based mutagenesis methods such as simplicity in handling (no requirement of isolated template, PCR-based in vitro cloning), allowing simultaneous targeting of multiple sites within the genome, offering high flexibility for gene knockouts, insertions, and other complex modifications.¹⁰⁶ All these features make CRISPR an extremely powerful tool for in vivo applications in general.¹⁰⁷

The first proof of principle of such a CRISPR-Sortase strategy was carried out in 2014 by Lodish et. al, in which LPETG motif was inserted at the C terminus of the murine endogenous Kell, a type-II membrane protein of red blood cell, by CRISPR-Cas9. The generated Kell-LPETG exhibited no inhibition of erythroid differentiation and was retained on the plasma membrane of mature red blood cells. Further studies demonstrated the LPXTG-tagged red blood cell can be performed with SML such as biotin probes with an efficiency of 81±28 % as determined by flow cytometry.¹⁰⁸ Based on this work, the authors further covalently attach disease-associated autoantigens to red blood cells as a means of inducing antigen-specific tolerance. Results proved that transfusion of red blood cells bearing self-antigen epitopes alleviated and prevented signs of disease in experimental autoimmune encephalomyelitis, as well as maintaining normoglycemia of type 1 diabetes in a mouse model.¹⁰⁹ All these results highlight the immense potential of applying the CRISPR-Sortase strategy for prophylaxis and therapy of autoimmune diseases.

In 2018, Muzykantov and his colleagues developed a straightforward approach to produce site-specifically modified antibodies using the CRISPR-Sortase strategy. They utilized CRISPR-Cas technology to genetically incorporate an LPETGG motif into the C-terminus of the third immunoglobulin heavy chain constant region (CH3) within a hybridoma cell line. This resulted in the production of antibodies that were readily capable of site-specific bioconjugation by SML. By performing SML, fluorescent and radioactive cargoes were conjugated to the antibodies. Notably, these immunoconjugates exhibited almost double the specific targeting in the lung compared to chemically conjugated maternal mAb. Furthermore, there was a concurrent reduction in uptake in the liver and spleen. The approach described in this study offers a straightforward method for cost-effective production of homogeneous, effective, and scalable antibody conjugates for a broad range of therapeutic and diagnostic applications.¹¹⁰

More recently, Scheeren et. al developed a versatile CRISPR/HDR methodology to efficiently engineer the constant immunoglobulin domains to generate recombinant hybridomas, which secrete antibodies such as secreting antigen-binding fragments, isotype-switched chimeric antibodies, and Fc-silent mutants in preferred formats, species, and isotypes. Furthermore, by incorporating the sortase recognized motif in the antibodies, site-specifically attachment of cargo to these antibody products via chemoenzymatic modification.¹¹¹ Based on the work, an expanded genomic engineering toolbox to enable duel modification of antibodies on the HC and LC loci of the mouse IgG1 (mIgG1) hybridoma was reported. In brief, two orthogonally sortase A recognition motifs were incorporated in HC and LC of Fab, which resulted in a dual-tagged Fab using CRISPR/Cas9. Subsequently, two distinct cargos were sequentially conjugated to the dual-tagged Fab with two evolved sortase variants with corresponding specificities in a site-specific manner. The latter case shows a versatile use of multifunctional conjugates for therapeutic applications.¹¹²