Sortase A Ligation: Reversibility, Enzyme Loading and the LPXTG Scar

Sortase A from Staphylococcus aureus forms one bond in one place, and both the appeal and the difficulty of the method follow from how it does it. Purified recombinant sortase cleaves peptides bearing an LPXTG motif at the peptide bond between threonine and glycine; in the presence of NH₂-Gly₃ it catalyses exclusively a transpeptidation, linking the carboxyl group of threonine to the amino group of the triglycine. Both the hydrolysis and the transpeptidation require the sulfhydryl of cysteine 184, which is what pointed to a thioester acyl-enzyme intermediate, and the presence of amino group donors increases the rate of cleavage at the motif ^[1].

The awkward part is visible in the same description. The ligated product regenerates an LPXT-Gly junction, so the product is itself a substrate; and the fragment displaced during cleavage begins with a glycine, so it is itself a competent nucleophile. Nothing in the chemistry stops when you have the species you wanted.

The catalytic cycle, and where the material can go

The NMR structure of sortase showed a distinctive β-barrel fold in which the active-site sulfhydryl of Cys184 is poised for ionisation by His120, presumably enabling the resulting thiolate to attack the LPXTG peptide; the structure was read as suggesting a mechanistic relationship to the papain and cathepsin proteases. Calcium binding near the active site stimulates catalysis, possibly by altering the conformation of a surface loop that recognises newly translocated polypeptides ^[2].

Steady-state analysis of both reaction types supports a ping-pong mechanism in which a common acyl-enzyme intermediate is formed in transpeptidation and in hydrolysis, but with distinct rate-limiting steps: formation of the acyl-enzyme limits transpeptidation, whereas hydrolysis of that same acyl-enzyme limits the hydrolytic reaction. When a nucleophile is not available, sortase slowly hydrolyses the LPETG peptide at the same site ^[3]. That is the practical shape of the problem. Starving the reaction of nucleophile does not pause it; the acyl-enzyme still forms and is resolved by water instead, and that material is gone to a hydrolysed dead end rather than merely delayed.

The nucleophile site is fussier than most protocols imply. In the same kinetic study the nucleophile binding site was found to be specific for diglycine: the S1′ and S2′ subsites both prefer a glycine residue, the S1′ subsite is exclusively selective for glycine, and lengthening the polyglycine acceptor beyond diglycine did not further enhance binding and catalysis ^[3]. Longer glycine tracts remain common at the bench for steric and solubility reasons, but that measurement gives you no grounds to expect GGGGG to bind or turn over better than GG.

What the kinetic constants say about enzyme loading

An HPLC-based assay of wild-type SrtA reported k_cat = 1.1 ± 0.1 s⁻¹, a K_m for the LPETG acyl donor of 7.1 ± 0.9 mM, a catalytic efficiency of 159 ± 25 M⁻¹ s⁻¹, and a K_m for the GGG nucleophile of 71 ± 8 µM ^[15]. A separate HPLC measurement, summarised in a structural review of the family, gives k_cat = 0.28 ± 0.02 s⁻¹ with K_m values of 7.33 ± 1.01 mM for the LPXTG substrate and 196 ± 64 µM for the Gly₅ substrate ^[16]. The two assays disagree about k_cat and agree about the number that will actually govern your experiment: the K_m for the sorting signal is millimolar. A protein substrate at 20–100 µM sits two orders of magnitude below it, so you are working on the linear part of the curve, where rate scales with acyl-donor concentration and with enzyme concentration and with nothing else you can easily change.

A second reason for heavy enzyme loading sits in the protonation state. pH dependencies of k_cat and k_cat/K_m for the substrate Abz-LPETG-Dap(Dnp)-NH₂ were bell-shaped, with pK_a values of 6.3 ± 0.2 and 9.4 ± 0.2 for k_cat and 6.2 ± 0.2 and 9.4 ± 0.2 for k_cat/K_m; solvent isotope effect measurements were inverse, and the combined evidence was read as a reverse protonation mechanism in which only about 0.06% of SrtA is competent for catalysis at physiological pH, that small competent fraction being highly active with an estimated k_cat/K_m greater than 10⁵ M⁻¹ s^{−1 [4]}. Most of the enzyme you weigh out is, on this reading, spectating.

Two further constants are worth carrying. Removal of Ca²⁺ reduces SaSrtA activity fivefold; the ion binds a pocket between the β3/β4 and β6/β7 loops, and NMR dynamics indicate that binding alters the mobility and structure of the β6/β7 active-site loop, thereby allosterically regulating affinity for the sorting signal ^[16]. And the enzyme you have chosen is already the fast one: native SaSrtA catalyses transpeptidation roughly 20–500-fold faster than the other sortases characterised in vitro ^[16]. Do not assume another sortase class will drop into a protocol tuned for SrtA.

Driving the equilibrium

Because the sortase reaction reaches equilibrium, a large excess of polyglycine nucleophile is often employed to drive the reaction forward and suppress sortase-mediated side reactions ^[12]; most labelling protocols using this enzyme have required a large excess of both substrate and sortase to give high yields of ligation product ^[11]. That default is expensive when the nucleophile carries a synthetically demanding payload, and it is the constraint that the following approaches were built to remove. They are not interchangeable, and they differ in which half of the reaction they modify.

Depsipeptide acyl donors. Substituting an ester linkage between the threonine and a glycolic acid residue in place of the Thr–Gly amide effectively renders the reaction irreversible, allowing complete labelling of proteins carrying a single N-terminal glycine residue in 4–6 h with only a small excess of substrate and catalytic quantities of sortase; synthesis of the depsipeptide itself typically takes 2–3 d ^[11]. The displaced species is a hydroxy acid rather than a glycine-initiated peptide, so it cannot re-enter as a nucleophile. Note the format this was demonstrated in: the protein supplies the N-terminal glycine and the depsipeptide carries the label, which is the N-terminal ligation geometry, not the ordinary C-terminal one.

Self-destructing leaving groups. A related tactic keeps the amide but makes the excised fragment consume itself. LPETGG-isoacyl-Ser and LPETGG-isoacyl-Hse substrates can be irreversibly ligated to N-terminal glycine-containing moieties through deactivation of the sortase-excised fragment by diketopiperazine formation, and both substrates were reported as stable in the ligation buffer ^[10].

Chelation of the leaving group. A modified recognition motif containing a masked Ni²⁺-binding peptide has been used to the same end: deactivating the Ni²⁺-binding peptide with a Ni²⁺ additive improved reaction performance at low to equimolar ratios of glycine amine nucleophile to sortase substrate, with both peptide and protein substrates ^[14].

Enzymatic destruction of the by-product. Coupling a transpeptidase to a d-aminopeptidase allows sequence-specific degradation of the peptide by-product, shifting the equilibrium to favour product formation; after optimisation this permitted N-terminal labelling using essentially equimolar ratios of peptide label to protein, and C-terminal labelling with only a small excess, using unactivated commercially sourced peptides rather than the activated substrates the depsipeptide and isoacyl routes require ^[18].

Physical separation. Immobilising sortase A within a flow microreactor permits efficient sortagging at low nucleophile concentrations, and was used to generate a protein bioconjugate that was not accessible by solution-phase batch sortagging ^[12]. This is the option to reach for when the nucleophile is the scarce component and modifying the acyl donor is not open to you.

The LPXTG scar, and how far it can be negotiated

The motif does not disappear on ligation. When sortase-mediated ligation is used for protein semisynthesis, up to four mutations resulting from the strict requirement of the LPXTG sorting motif are introduced into the target protein ^[7]. Whether that matters is a question about your construct, not about the enzyme: at a solvent-exposed terminus five extra residues are usually inert, whereas an internal or interface-proximal site is a different proposition.

Two routes narrow the scar. The first is engineering the enzyme's recognition. Directed evolution of a sortase A library randomised in the substrate recognition loop, screened by phage display, yielded a mutant with broad substrate selectivity, and that gain-of-function was exploited for the traceless semisynthesis of histone H3 ^[7]. Nine rounds of yeast display screening with negative selection produced two orthogonal variants recognising LAXTG and LPXSG, with specificity changes of up to 51,000-fold relative to the starting sortase and without substantial loss of catalytic activity, and up to 24-fold specificity for their target substrates relative to their next most active peptide substrate — orthogonal enough to conjugate multiple peptide substrates to their respective targets simultaneously in a single solution ^[13].

The second route is choosing where the scar goes. The standard protocol engineers the LPXTG motif at the place where modification is desired, which allows functionalisation of any given protein at its C terminus or, for select proteins, at an internal site ^[9]. The phrase "for select proteins" is doing work: an internal loop must be exposed, tolerant of five inserted residues, and tolerant of being transiently cut.

N-terminal, C-terminal, and the fusions genetics cannot make

In the C-terminal geometry the protein carries the LPXTG motif and the probe carries the oligoglycine: sortase cleaves between threonine and glycine and attaches an exogenously added oligoglycine peptide bearing the functional group of choice, which may be a fluorophore, biotin, a protein or a lipid ^[9]. Reversing which partner carries which half of the motif gives the N-terminal geometry, where the protein presents an N-terminal glycine and the probe supplies the acyl donor. Selective N-terminal labelling of proteins containing N-terminal glycine residues has been achieved with S. aureus SrtA and LPXT derivatives, with near-quantitative labelling of multiple protein substrates and excellent site specificity ^[5].

The two geometries can be run on one polypeptide because the enzymes differ. In the same work, SrtA from Streptococcus pyogenes was used to label the C terminus at an LPXTG site with a fluorescently labelled dialanine nucleophile, while S. aureus SrtA handled the N terminus, giving a single chain with different labels at each end ^[5]. The orthogonality here is in the nucleophile preference, and it is what makes double labelling possible without protecting one terminus.

The N-terminal route has a construct requirement that is easy to underestimate: the glycine must actually be N-terminal, with a free α-amino group, which usually means a protease-cleavable leader or an expression strategy that exposes it, and the exposure has to be confirmed rather than assumed.

Sortase also reaches arrangements that no genetic fusion can produce. Installing, on the N or C terminus, the modifications required to execute a strain-promoted copper-free cycloaddition allowed the assembly of N-to-N and C-to-C dimers with full retention of the biological activity of both fusion partners and without inflicting chemical damage on the proteins to be joined ^[8]. That construction is the clearest illustration of the division of labour between this method and purely chemical conjugation: the enzyme decides where, the cycloaddition decides how the two halves meet.

Choosing the enzyme

Evolved variants change the economics rather than the chemistry. A yeast-display screen integrated with enzyme-mediated bioconjugation and cell sorting produced variants of sortase A with up to a 140-fold increase in LPETG-coupling activity compared with the starting wild-type enzyme, and an evolved variant labelled LPETG-tagged human CD154 on the surface of HeLa cells much more efficiently than wild type ^[6]. A later screen of libraries at residues Y187 and E189 on the pentamutant background, combined with a D124G mutation identified separately, reported k_cat/K_m for LPETG of 11,425 ± 1,315 M⁻¹ s⁻¹ for 5M/Y187L/E189R and 16,722 ± 2,329 M⁻¹ s⁻¹ for 5M/D124G/Y187L/E189R, against 159 ± 25 M⁻¹ s⁻¹ for wild type in the same assay ^[15].

Read the second column before adopting one. In that table the K_m for GGG rises from 71 ± 8 µM for wild type to 2,367 ± 610 µM for 5M/Y187L/E189R and 1,700 ± 420 µM for 5M/D124G/Y187L/E189R ^[15]. The efficiency gain is on the acyl-donor side; the nucleophile is bound more than an order of magnitude more weakly. An evolved enzyme can therefore let you cut enzyme loading while requiring you to hold nucleophile concentration up, which is exactly the wrong trade if the nucleophile is your expensive component.

The loading numbers are worth having in front of you. For labelling at the C terminus of an antibody, 5M/Y187L/E189R achieved comparable levels of labelling to wild-type sortase at 24–40-fold lower enzyme concentration — 0.25 µM versus 10 µM for the heavy chain and 2.5 µM versus 60 µM for the light chain — and attachment of an azide to that C terminus went to completion with 1 µM of the variant, whereas comparable conjugation with wild-type sortase required enzyme concentrations 60-fold higher ^[15].

Calcium dependence is the other axis. In vivo and cell-based applications of sortase have been somewhat limited by the large range of calcium concentrations, as well as by the often transient nature of protein-protein interactions in living systems; combining multiple mutations produced a variant that was both calcium-independent and highly active, with enhanced activity for N-terminal and C-terminal labelling and for cell surface modification under physiological conditions ^[17]. If your buffer must contain EDTA or EGTA, or the reaction is to run in a compartment where free Ca²⁺ is low, this is the variant class to start from rather than an attempt to rescue wild type.

Against the chemical alternatives

Set beside the copper-catalysed and strain-promoted cycloadditions treated in the click chemistry guide, and the payload-release chemistries treated in the linker guide, sortase trades one problem for another. It gives you site specificity from a five-residue genetically encoded motif and forms an ordinary amide bond, and the reaction runs in aqueous buffer with no copper catalyst and no strained or activated small-molecule reagent, though the wild-type enzyme still wants Ca²⁺ for efficient catalysis ^{[1, 9, 16]}. What it costs is stoichiometric awkwardness: a millimolar K_m for the acyl donor ^{[15, 16]}, micromolar enzyme, an equilibrium that must be pushed ^{[11, 12]}, and a sequence scar that only a re-engineered recognition motif narrows ^[7]. A cycloaddition, by contrast, is effectively irreversible once formed but tells you nothing about where the handle went; you still have to install the azide or alkyne site-specifically, which is precisely the problem sortase solves.

The two are therefore better read as complementary than as competing, and the N-to-N and C-to-C fusion work is the proof of that reading: sortase installed the cycloaddition handles at defined termini, and the strain-promoted reaction did the joining ^[8]. Where a payload cannot survive being appended to a synthetic oligoglycine peptide, the same sequencing applies — ligate a small bioorthogonal handle enzymatically, then attach the payload chemically.

What a completed reaction licenses you to conclude

An intact-mass shift of the expected size tells you that something of the right mass was added; it does not tell you where. Site assignment requires fragmentation, and MS/MS was used to confirm the modification site in the antibody conjugates described above ^[15]. On a substrate carrying a single engineered motif the inference is short, but the site claim should still rest on fragment ions rather than on design intent.

Three species should be resolved and quantified rather than assumed absent: unreacted acyl donor, the hydrolysed product terminating in LPXT-OH, and the ligated product. The hydrolysis product is a distinct mass, and its abundance tracks how nucleophile-starved the reaction ran, alongside incubation time and enzyme loading ^[3]. If it dominates, adding enzyme will make it worse, not better; the correction is more nucleophile, or one of the equilibrium-driving substrates.

Reaction time is a variable, not a constant to be copied. Sortase-mediated reactions can take only a few minutes, but reaction times can be extended to increase yields, and expression and purification of the enzyme itself takes about 3 d ^[9]. Because the product remains a substrate, an extended incubation is not monotonically better; the yield curve can turn over, so time courses should be run on a new pairing rather than a single endpoint taken on faith.