Peptide tag forming a rapid covalent bond to aprotein, through engineering a bacterial adhesinBijan Zakeria,1, Jacob O. Fierera,1, Emrah Celikb, Emily C. Chittocka, Ulrich Schwarz-Linekc, Vincent T. Moyb, andMark Howartha,2

aDepartment of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, United Kingdom; bDepartment of Physiology and Biophysics,University of Miami, Miller School of Medicine, P.O. Box 016430, Miami, FL 33101-6430; and cBiomedical Sciences Research Complex, University ofSt. Andrews, North Haugh, St. Andrews, Fife KY16 9ST, United Kingdom

Edited by James A. Wells, University of California, San Francisco, CA, and approved January 17, 2012 (received for review September 21, 2011)

Protein interactions with peptides generally have low thermo-dynamic and mechanical stability. Streptococcus pyogenes fibro-nectin-binding protein FbaB contains a domain with a spontaneousisopeptide bond between Lys and Asp. By splitting this domainand rational engineering of the fragments, we obtained a peptide(SpyTag) which formed an amide bond to its protein partner (Spy-Catcher) in minutes. Reaction occurred in high yield simply uponmixing and amidst diverse conditions of pH, temperature, andbuffer. SpyTag could be fused at either terminus or internally andreacted specifically at the mammalian cell surface. Peptide bindingwas not reversed by boiling or competing peptide. Single-moleculedynamic force spectroscopy showed that SpyTag did not separatefrom SpyCatcher until the force exceeded 1 nN, where covalentbonds snap. The robust reaction conditions and irreversible linkageof SpyTag shed light on spontaneous isopeptide bond formationand should provide a targetable lock in cells and a stable modulefor new protein architectures.

bacterial attachment ∣ chemical biology ∣ microbiology ∣ proteinengineering ∣ single molecule biophysics

Tagging with peptides (e.g., HA, myc, FLAG, His6) is one of themost common ways to detect, purify, or immobilize proteins

(1–4). Peptides are very useful minimally disruptive probes (5)but they are also “slippery”- antibodies or other proteins typicallybind peptides with low affinity and poor mechanical strength(6–9). We sought to form a rapid covalent bond to a peptide tagwithout the use of chemical modification, artificial amino acids,or cysteines (disulfide bond formation is reversible and restrictedto particular cellular locations).

It has recently been found that Streptococcus pyogenes, likemanyother Gram-positive bacteria, contains extracellular proteins stabi-lized by spontaneous intramolecular isopeptide bonds (10). Herewe explored the second immunoglobulin-like collagen adhesindomain (CnaB2) from the fibronectin binding protein FbaB, foundin invasive strains of S. pyogenes (11,12) and essential for phago-cytosis-like uptake of the bacteria by endothelial cells (13). CnaB2contains a single isopeptide bond conferring exceptional stability:CnaB2 remains folded even at pH 2 or up to 100 °C (12). By split-ting CnaB2 into peptide and protein fragments, followed byrational modification of the parts, we developed a peptide tag of13 amino acids that rapidly formed a covalent bond with its proteinpartner (138 amino acids, 15 kDa) and characterized the condi-tions for reaction, cellular specificity of bond formation, and resi-lience of the reacted product.

ResultsDesign of SpyTag for Rapid Covalent Bond Formation. Crystallogra-phy and NMR have shown that CnaB2 forms a spontaneousintramolecular isopeptide bond (11, 12); quantum mechanical/molecular mechanical calculations (12) indicate that the unpro-tonated amine of Lys31 nucleophilically attacks the carbonylcarbon of Asp117 (Fig. 1A), catalyzed by the neighboring Glu77

(Fig. 1B). We hypothesized that splitting the CnaB2 domain into

a peptide representing the C-terminal β-strand containing the re-active Asp and a protein partner derived from the rest of the pro-tein would allow the two partners to reconstitute and undergocovalent reaction (Fig. 1C). This reaction occurred initially overhours for each species present at 10 μM, but by rational optimi-zation of the S. pyogenes (Spy) protein partner (termed SpyCatch-er) (SI Appendix, Fig. S1) and the peptide tag (termed SpyTag) (SIAppendix, Fig. S2), reaction now occurred in minutes (vide infra).Mixing SpyTag fused to Maltose Binding Protein (SpyTag-MBP)with SpyCatcher led to high yield of a product resistant to boilingin SDS (Fig. 1D). Mutation of the catalytic Glu77 in SpyCatcher(EQ mutant) or the reactive Asp117 in SpyTag (DA mutant) abol-ished covalent bond formation (Fig. 1D). Isothermal titration ca-lorimetry showed that SpyTag DA-MBP and SpyCatcher formeda noncovalent complex with a Kd of 0.2 μM (SI Appendix, Fig. S3),a high affinity for a peptide-protein interaction (6–8).

Electrospray ionization mass spectrometry confirmed thatreaction of the peptide with SpyCatcher led to covalent bondformation, with the combined mass 18 Da less than the sum ofthe individual masses, from loss of water (Fig. 2A).

We determined the efficiency of reaction between SpyTag-MBP and SpyCatcher and saw high yielding and rapid reconstitu-tion, with more than 40% forming a covalent bond in the firstminute (Fig. 2B). The second-order rate constant was 1.4 × 103 �43 M−1 s−1 (SD, n ¼ 3) and the reaction half-time was 74 s(Fig. 2C). Reaction was still rapid with each partner present attwofold or 10-fold lower concentration (SI Appendix, Fig. S4).

SpyTag Reaction was Robust to Diverse Conditions.We characterizedhow sensitive the reaction between SpyTag-MBP and SpyCatcherwas to the sample conditions. The reaction proceeded similarlyat 25 °C as at 37 °C (Fig. 3A). Reaction was also efficient at 4 °C,although significantly slower than at 25 °C (for the 1 min time-point: P < 0.0001, n ¼ 3) (Fig. 3A).

We were also interested in the dependence on pH, to seewhether SpyTag could be used in low pH compartments of thecell, such as endosomes. Reaction was efficient at all pH valuestested from 5 to 8 (Fig. 3B). In fact, the reaction was slightly faster

Author contributions: B.Z., J.O.F., E.C., U.S.-L., V.T.M., and M.H. designed research; B.Z.,J.O.F., E.C., E.C.C., U.S.-L., and M.H. performed research; B.Z., J.O.F., E.C., U.S.-L., V.T.M.,and M.H. analyzed data; and M.H. wrote the paper.

Conflict of interest statement: M.H. and B.Z. are authors on a patent application regardingpeptide targeting via spontaneous amide bond formation (United Kingdom PatentApplication No. 1002362.0).

This article is a PNAS Direct Submission.

Data deposition: The sequence reported in this paper has been deposited in the GenBankdatabase, www.ncbi.nlm.nih.gov/genbank (accession no. JQ478411).1B.Z. and J.O.F. contributed equally to this work.2To whom correspondence should be addressed. E-mail: mark.howarth@bioch.ox.ac.uk.

See Author Summary on page 4347 (volume 109, number 12).

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1115485109/-/DCSupplemental.

E690–E697 ∣ PNAS ∣ Published online February 24, 2012 www.pnas.org/cgi/doi/10.1073/pnas.1115485109

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

at pH 5 and 6 than at pH 7 (comparing pH 6 with 7 for the 1 mintime-point: P < 0.0001, n ¼ 3) (Fig. 3B).

Reaction efficiency was compared in the presence of a range ofbuffers (PBS, phosphate-citrate, Hepes, Tris), but these had littleeffect, indicating that the reaction was robust to buffer composi-tion (Fig. 3C). In particular there was no need for Ca2þ or Mg2þ,even though these divalent cations are important for the functionof many extracellular proteins, such as integrins or sortases.Because there are no cysteines in SpyTag or SpyCatcher, as ex-pected there was no effect of reducing agents on the reaction(SI Appendix, Fig. S5), so that reaction should be efficient inthe cytosol/nucleus, secretory pathway, or outside the cell.

We also tested the effect on the reaction of adding detergent,to investigate if SpyTag could be used in cell lysates or in condi-tions used to stabilize membrane proteins. There was no substan-tial effect on the reaction in the presence of high concentrationsof nonionic detergents (Fig. 3D). We have not come across bufferconditions that impair SpyTag reaction, apart from the presenceof sodium dodecyl sulfate from the loading buffer for SDS-PAGE.

SpyTag was Reactive at N-Terminal, Internal, and C-Terminal Sites.Antibodies often preferentially recognize peptide tags at specifictermini (1), while split intein/sortase tags must be at particularsites in the protein (14, 15). To establish the generality of SpyTagfunction, we tested whether SpyTag could react at different loca-tions in proteins. SpyTag-MBP has the SpyTag internally: SpyTagis preceded by a His6-tag and a thrombin cleavage site at the Nterminus and is followed by MBP. N-SpyTag-MBP has SpyTagdirectly after the initiating formyl-methionine: reaction of N-Spy-Tag-MBP with SpyCatcher was still efficient (SI Appendix, Fig. S6A).We also generated a construct with two SpyTags: the first SpyTag

was between MBP and the three zinc fingers of Zif268, while thesecond SpyTag was right at the C terminus, following Zif268 (termedMBP-SpyTag-Zif-SpyTag). Both SpyTags were reactive, generating adoubly-branched protein with two SpyCatcher moieties covalentlyattached (SI Appendix, Fig. S6B).

SpyTag Reactionwas Not Reversible over a Day. Spontaneous isopep-tide bond formation is most common between Lys and Asn (10),where there is loss of NH3, which may diffuse away and so helpto promote irreversibility. For spontaneous isopeptide bond for-mation between Lys and Asp, there is loss of H2O but there isapproximately 55 M H2O present (at least at the surface of theprotein), which could make the SpyTag:SpyCatcher reaction re-versible (16) (SI Appendix, Fig. S7A). To test whether the SpyTagreaction would reverse, we allowed SpyCatcher to react withSpyTag-MBP and then, to see if there was any backward reaction,we added a 20-fold excess of Cna peptide to capture any freeSpyCatcher (SI Appendix, Fig. S7A). However, overnight incuba-tion with competing free peptide did not lead to a decrease in theSpyTag-MBP:SpyCatcher covalent complex and there was no ap-pearance of the Cna peptide:SpyCatcher complex (SI Appendix,Fig. S7B). Cna peptide gave complete conversion of unreactedSpyCatcher (SI Appendix, Fig. S7B, lane 8). Therefore, the Spy-Tag reaction did not reverse under these conditions. Note thatonce the protein is unfolded the isopeptide bond between SpyTagand SpyCatcher should be as chemically stable as a typical amidebond, resisting prolonged boiling.

SpyTag Reacted Efficiently Inside Cells. Because SpyTag and Spy-Catcher are genetically encodable, we tested whether SpyTag-MBPand SpyCatcher could react together inside the cytosol of Escher-

Fig. 1. Spontaneous intermolecular amide bond formation by SpyTag. (A) Amide bond formation between Lys and Asp side chains. (B) Key residues for amidebond formation in CnaB2 shown in stick format, based on PDB 2X5P. (C) Cartoon of SpyTag construction. Streptococcus pyogenes (Spy) CnaB2was dissected intoa large N-terminal fragment (SpyCatcher, left) and a small C-terminal fragment (SpyTag, right). Reactive residues are highlighted in red. (D) SpyTag and Spy-Catcher associated covalently. SpyTag-MBP and SpyCatcher were mixed each at 10 μM for 3 h and analyzed after boiling by SDS-PAGE with Coomassie staining,alongside unreactive controls, SpyCatcher E77Q (EQ) and D117A (DA) SpyTag-MBP.

Zakeri et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ E691

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

Fig. 2. Characterization of isopeptide bond formation. (A) Mass spectrometry of reconstitution between Cna peptide and SpyCatcher. The minor peak resultsfrom the His6-tag on SpyCatcher leading to some gluconylation (see SI Appendix). (B) Time course of SpyTag-MBP:SpyCatcher covalent complex formation, witheach partner at 10 μM at 25 °C, pH 7.0 determined by SDS-PAGE. (C) Rate constant for SpyTag reaction, calculated from triplicate measurements (each pointshown) of SpyCatcher depletion under conditions as in (B). The equation for the trend-line and the correlation coefficient are shown.

Fig. 3. Sensitivity of SpyTag reaction to conditions. (A) Temperature dependence of SpyTag-MBP reconstitution with SpyCatcher at 10 μM at pH 7.0 after 1 or10 min. (B) pH-dependence of reaction as in (A) at 25 °C. (C) SpyTag-MBP and SpyCatcher were incubated each at 10 μM at 25 °C for 1 or 10 min at pH 7.0 inphosphate buffered saline (PBS), phosphate-citrate (PC), Tris, or Hepes buffers. (D) SpyTag-MBP and SpyCatcher were incubated each at 10 μMat 25 °C for 3 h atpH 7.0 in phosphate-citrate buffer containing no detergent (None), 1% Triton X-100 (Tx100), 1% Tween 20, or 0.5% Nonidet P-40 (NP40). All reactions wereanalyzed by SDS-PAGE and Coomassie staining. (All graphs mean of triplicate �1 s:d:; some error bars are too small to be visible.)

E692 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115485109 Zakeri et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

ichia coli. Cells were transformed with a plasmid encoding theproteins individually or bicistronically, induced with IPTG, andsubsequently boiled with SDS loading buffer to prevent any ex vivoreaction. Lysates were then analyzed by SDS-PAGE. A covalentcomplex corresponding to the expected molecular weight ofSpyCatcher:SpyTag-MBP could be clearly observed and little un-reacted SpyTag-MBP remained (Fig. 4A). This covalent complexwas not present when SpyTag-MBP was coexpressed with the con-trol SpyCatcher EQ (Fig. 4A). Mixing lysate or cells from bacteriaexpressing SpyCatcher or SpyTag-MBP individually did not lead tocovalent complex formation (Fig. 4A), indicating that reaction wasnot occurring ex vivo and consistent with SpyTag being able to reactefficiently in the cytosolic environment.

To address whether SpyCatcher had reacted with other cyto-solic proteins, His6-tag containing proteins were pulled downwith Ni-NTA. SpyTag-MBP, SpyCatcher, and SpyCatcher EQ allcontain an N-terminal His6-tag, and so we would expect to pull-down these proteins and any covalent complexes they formed. Asexpected SpyTag-MBP and SpyCatcher had efficiently reacted,depleting almost all SpyTag-MBP, but there was minimal pull-

down of proteins of other molecular weights as assessed by Coo-massie staining, indicating specificity of the SpyTag reaction inthe cytosol. Using silver staining and anti-His immunoblottingas more sensitive probes for lower abundance interactions de-tected a small amount of a second protein pulled down in Spy-Catcher-expressing bacteria (SI Appendix, Fig. S8).

SpyTag Reaction was Specific at the Mammalian Cell Surface. Stableand specific targeting of biophysical probes at the mammaliancell surface is an important challenge in the study of receptorfunction (1, 15). SpyTag was targeted to the surface of HeLa cellsby genetic fusion to green fluorescent protein-labeled Intercellu-lar Adhesion Molecule-1 (ICAM1), to give SpyTag-ICAM1-GFP(Fig. 4C). SpyTag-ICAM1-GFP trafficked efficiently to the plas-ma membrane (Fig. 4D). Alexa Fluor 555-labeled SpyCatcheradded to the medium labeled GFP-positive cells, even those withlow expression levels, but did not label neighboring nonexpres-sing cells (Fig. 4D), showing that there was little nonspecificbinding. In addition, cells expressing SpyTag-ICAM1-GFP werenot labeled by the negative control dye-labeled SpyCatcher EQ

Fig. 4. SpyCatcher reacted specifically. (A) SpyTag was reactive inside the E. coli cytosol. SpyTag-MBP and SpyCatcher were expressed in isolation or in the samecells. Cells were lysed and denatured by boiling with SDS buffer, before SDS-PAGE and Coomassie staining. SpyCatcher EQ was a negative control. To show thatreconstitution happened within cells, lysates (lane 6) or cells (lane 7) expressing SpyCatcher alone (lane 1) or SpyTag-MBP alone (lane 3) were mixed. (B) SpyTagspecificity inside the E. coli cytosol. Proteins were expressed as in (A) and the His6-tags, present on SpyTag-MBP, SpyCatcher (EQ) and anything with which theyhave reacted, were pulled down with Ni-NTA, before SDS-PAGE and Coomassie staining. (C) Cartoon of targeting SpyTag to the mammalian cell surface, bygenetic fusion to ICAM1 linked to GFP, with an HA tag after the signal sequence. (D) HeLa SpyTag-ICAM1-GFP were incubated with dye-labeled SpyCatcher for15 min and analyzed by fluorescence microscopy. The GFP image (green, left) highlights cells expressing SpyTag-ICAM1-GFP, the 555 image (red, middle) showsthe cells to which SpyCatcher bound, and the bright-field image (grayscale, right) shows all the cells. SpyCatcher EQ-555 (lower boxes) is a negative control.(Scale bar, 20 μm).

Zakeri et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ E693

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

(Fig. 4D). SpyCatcher attachment to cells was resistant to acidwash (SI Appendix, Fig. S9 A and B), suggesting that this cellularlabeling was highly stable.

The nature of the labeling was further explored by immuno-blotting against the HA tag on SpyTag-ICAM1-GFP, showing anadduct of SpyTag-ICAM1-GFP consistent with interaction withSpyCatcher and stable to boiling in SDS, indicating that the re-action was covalent (SI Appendix, Fig. S9C). To see if SpyCatcherhad reacted with other cellular proteins, we blotted against theHis6-tag on SpyCatcher. The anti-His blot showed a new band,corresponding to the molecular weight of SpyCatcher:SpyTag-ICAM1-GFP, while there was no detectable reaction on cells notexpressing SpyTag-ICAM1-GFP (SI Appendix, Fig. S9C). Theseblots indicate that SpyCatcher showed high specificity to detectSpyTag amidst the diverse other surface proteins (17) and that thereaction was efficient at cellular expression levels.

SpyTag Formed Mechanically Stable Bonds with SpyCatcher. Becausethe environment of the SpyTag:SpyCatcher reactive residues low-ers the activation energy for amide bond formation, it is possiblethat the activation energy for amide bond hydrolysis could also bedecreased. We showed that under mild conditions, in the pre-sence of excess competitor, there was no sign of breakage of thebond to SpyTag (SI Appendix, Fig. S7). However, it was possiblethat pulling on SpyTag would distort the energy profile and leadto efficient bond breakage (18), making the SpyTag system me-chanically labile. We assessed the mechanical stability of SpyTagat the single-molecule level, using dynamic force spectroscopy withan atomic force microscope (AFM). We fused SpyCatcher to twoI27 domains, which provide a fingerprint from their characteristicunfolding force, to validate that one is observing specific cantileverbending from pulling on a single molecule (19) (Fig. 5A).

For the interaction between SpyTag and SpyCatcher we ob-served traces with a breakage force >1 nN (Fig. 5 B and C), con-sistent with mechanochemistry taking place as the cantileverbreaks a covalent bond when it retracts (18). The median breakageforce for the covalent interaction was 1.9 nN, which is >20 timesstronger than streptavidin-biotin or antibody-antigen interactions(20, 21). C-N and C-C bonds are not predicted to break until 5 nN(18), so it is likely that was not the isopeptide bond, nor anotherbond in the protein which broke, but rather a weaker link in thechain, such as a C-S bond linking the proteins to the cantilever/bead. The nonreactive SpyCatcher EQ, after 1,940 tests with Spy-Tag, formed zero interactions above 500 pN (Fig. 5D, comparingSpyCatcher and SpyCatcher EQ: Fisher’s exact test P < 0.0001).As further negative controls, with no SpyCatcher on the cantileveror with SpyTag on the bead preblocked with free SpyCatcher, therewere also zero force traces above 500 pN (Fig. 5D), consistent withthe specificity of the >500 pN interactions between SpyTag andSpyCatcher seen by AFM.

The cantilever was lowered for 20 s, in which time we wouldexpect only a fraction of covalent bonds to form (Fig. 2B), sothat we could probe the early stages of the SpyTag:SpyCatcherinteraction. The unusually long contact times of this experimentmade it challenging to avoid nonspecific interaction between thecantilever and the bead surface (22). Hence there was a low back-ground of tests where the polyethylene glycol (PEG) and agaroselinkers were extended for the negative controls, using no Spy-Catcher or using preblocked SpyTag (Fig. 5D), giving force traces<500 pN. The frequency of PEG extensions for SpyCatcher EQinteracting with SpyTag was comparable to these two controls(Fig. 5D). Therefore there was no detectable interaction betweenSpyCatcher EQ and SpyTag in this experiment. However, the fre-quency giving force traces <500 pN for SpyCatcher interactingwith SpyTag was significantly higher than for the other three con-ditions (chi-squared P < 0.0001) (Fig. 5D), supporting that the100–200 pN peak for SpyTag and SpyCatcher (Fig. 5C) was a spe-cific interaction. This peak was likely the force from a noncova-

lent interaction between SpyTag and SpyCatcher, as SpyTag docksand forms β-sheet hydrogen bonds. 100–200 pN is strong for a pro-tein-protein interaction (23,24) and, in concert with the isothermaltitration calorimetry data (SI Appendix, Fig. S3), the AFM resultssuggest that the SpyTag:SpyCatcher interaction has unusual me-chanical strength even before the covalent bond forms.

DiscussionFormation of an amide bond from reaction of a carboxylate withan amine is thermodynamically unfavorable under standard bio-chemical conditions (16); the equilibrium constant for two aminoacids reacting to form a dipeptide plus water is approximately10−3 (16). Hence proteases reach equilibrium with almost com-plete hydrolysis of their peptide substrates, while protein synthesisdepends on ATP-dependent amino acid activation by amino-acyltRNA synthetases. Studies attempting to use proteases to catalyzeamide bond formation have established that factors moving theequilibrium towards bond formation are a low dielectric constantto facilitate deprotonation (removing the energetic cost from de-hydration of the charges) and trapping the product, either by trans-fer to an organic phase, precipitation, or an exergonic bindinginteraction (16). Therefore, for intermolecular amide bond forma-tion by SpyTag, the orientation of the reactive groups, the hydro-phobic environment, and proton-shuffling by Glu77 all acceleratethe reaction kinetics, but the environment created in the SpyTag:SpyCatcher complex is also likely to determine that the position of

Fig. 5. Dynamic force spectroscopy of the SpyTag:SpyCatcher interaction.(A) Cartoon of the constructs for testing the SpyTag:SpyCatcher interactionby AFM. (B) Representative force extension curve for Cys-I272-SpyCatcherforming a strong interaction with SpyTag-MBP-Cys, showing regions corre-sponding to PEG/agarose stretching, unfolding of two I27 domains (enlargedin the inset), and covalent bond breakage. (C) Histogram of different bondbreakage forces for Cys-I272-SpyCatcher interactions with SpyTag-MBP-Cys.(D) Table of statistics of breakage forces. SpyTag:SpyCatcher was comparedwith the control SpyTag:SpyCatcher EQ. SpyTag was also tested against a can-tilever only coated with PEG (SpyTag:PEG), or SpyTag was preblocked withfree SpyCatcher before testing with a SpyCatcher-coated cantilever.

E694 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115485109 Zakeri et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

equilibrium lies firmly on the side of bond formation; we found thatSpyTag formed the covalent complex with high yield both in vitro(Figs. 1 and 2) and inside E. coli (Fig. 4).

Isopeptide bond formation is a spontaneous posttranslationalmodification found in several extracellular proteins from Gram-positive bacteria, including human pathogens such as Staphylo-coccus aureus, Corynebacterium diphtheriae, and Streptococcuspneumoniae (10). There has been little indication of how fast iso-peptide bond formation occurs, because the protein recoveredfrom recombinant expression had reacted to completion (12, 25).With this split system we provide the clearest evidence that reac-tion can occur in minutes, consistent with the biological time scalefor export of these proteins (13, 26). Regarding the evolution ofthis posttranslational modification, our search for a fast reactingpeptide tag showed that every change we tried to the residues inthe C-terminal β-strand, even conservative changes pointing awayfrom the protein partner, greatly reduced the speed of reaction(SI Appendix, Fig. S2B), suggesting that the reactive aspartic acidmust be precisely aligned for reaction and that there is strongevolutionary selection for the stability the bond provides tothe protein. Because residues in the C-terminal β-strand couldnot be changed, we managed to increase the efficiency of reactionby including an extra 5 C-terminal amino acids (SI Appendix,Fig. S2), which were not seen in the crystal structure (11) butmay have an interaction with the main domain, facilitating dock-ing of the peptide with SpyCatcher.

Our results also provide insight into the mechanical stability ofspontaneous isopeptide bonds. The pilin Spy0128, has been ana-lyzed by dynamic force spectroscopy and shown to be resistant tostretching because of its isopeptide bond, but forces >1;000 pN(sufficient to break covalent bonds) were not explored (27). Also,in Spy0128 the reaction was between Lys and Asn (25, 28), ratherthan Lys and Asp for CnaB2 (16). Like Spy0128, in CnaB2 theisopeptide is between the N-terminal and C-terminal β-strands,so that, if the isopeptide does not break, then the force on theprotein will not cause any extension of the protein chain. Inter-molecular reconstitution can be achieved with split Spy0128 butthe partners react orders of magnitude slower, undergo side re-actions, and the protein is undesirably large (29). The stability ofthe SpyTag interaction before the isopeptide bond forms may re-late to the known mechanical strength of long parallel β-strands,where multiple hydrogen bonds have to break simultaneously(23). The resilience of SpyTag:SpyCatcher to pulling forces afterisopeptide bond formation suggests that the CnaB2 domain maybe exposed to high force in bridging S. pyogenes to fibronectinduring cell invasion (13). Precise attachment of proteins to AFMtips, to understand protein folding and resistance to extension, isusually achieved by introducing cysteines. However, carbon-sul-fur or sulfur-sulfur bonds break at lower forces than amide bonds(18) and introducing free cysteines often disrupts folding of pro-teins containing disulfides; instead fusing a protein of interest toSpyTag may prove a useful handle to allow AFM tips to pull onspecific proteins, even on the cell surface.

Peptide tags are a central tool in molecular biology (1). How-ever, the reversibility of peptide binding often compromises sen-sitivity of immunoassays, stability of arrays/nano-assemblies, andpurity of protein purification (1, 30, 31). SpyTag is a short peptidewith fundamentally different binding characteristics to peptidetags in current use, forming a covalent isopeptide bond to itsprotein partner. Reaction requires only mixing and is rapid, highyielding, robust to experimental conditions, and shows goodspecificity. The reactivity of SpyTag at 37 °C means that SpyTagcan be used in live cell imaging of mammalian cells and also po-tentially in transgenic mammalian systems. The reactivity at 4 °C,combined with the resistance to commonly used detergents,should allow the use of SpyTag for pull-downs in cell lysate, wherethe low temperature minimizes proteolysis. The range of pHvalues over which SpyTag can react and the absence of Cys in

SpyTag or SpyCatcher should facilitate its use in the cytosol,nucleus, outside the cell, but also in the low pH endosomes andlysosomes. This robustness of spontaneous isopeptide formationto various conditions may be because docking of the peptide andprotein partner could create a microenvironment buried from thesurface. Note that derivatizing SpyCatcher with an N-hydroxysuccinimide (NHS) activated dye did not block reaction withSpyTag (Fig. 4D), suggesting a reduced accessibility of Lys31 toexogenous labels, which will facilitate surface attachment of Spy-Catcher or conjugation with other biophysical probes.

There are a number of ways to engineer peptides to react withsmall molecules, most commonly SNAP-tag (an engineered pro-tein tag for multiprotein labeling in living cells) and HaloTag (2,4, 32, 33), but unlike SpyTag, these methods are not suitable fordirecting interactions with other proteins. Coiled coils can drivespecific protein association in living cells but have limited stability(34). Native chemical ligation is an elegant way to link proteins(35, 36) but specificity and yield appear to be limited in cells (36).Sortase can be used to link together two proteins in vitro or to linka peptide with a small molecule on the surface of living cells, butrequires high [Ca2þ] (15), which is damaging in the cytosol ornucleus. Photoreactive amino acids are powerful for identifyingunknown binding partners (37,38,39) but (i) require UV which isoften damaging in living systems, (ii) do not react with a definedpartner, and (iii) typically give low yields (38). A protein contain-ing an artificial alkyne amino acid can react with another proteincontaining an artificial azido amino acid (40) but the reactionmust be catalyzed by toxic CuI . Furthermore, the reaction wouldnot just be between the tagged proteins but also between the freeamino acids. Split inteins can covalently splice proteins togetherand have the great advantage of leaving no trace after reaction(14, 41, 42), but must be located at defined termini and may un-dergo side reactions (43). Thus SpyTag has distinctive features interms of flexibility of reaction conditions, specificity in cells, andhigh yield of reaction, in comparison to existing chemical biologyapproaches for irreversible protein targeting.

Limitations of SpyTag are that it is not traceless, in contrast tosplit intein-mediated ligation, and that the reaction rate is stillfar from the diffusion limit, although future work with phage orribosome display (44) should be able to improve the on-rate. Therapid on-rate makes the femtomolar affinity streptavidin-biotininteraction so useful for isolation of biotin-conjugates at traceconcentrations (31, 45). However, the strongest noncovalent in-teractions, such as streptavidin-biotin, can be broken in secondsby molecular motors (46, 47) or in milliseconds by shear forces(48), so it has been very difficult to provide barriers or locks incellular systems using current tools. Molecular motors will not beable to break the covalent bond formed by SpyTag, which shouldbe targetable to specific proteins, compartments or cell types, ex-pressed from inducible promoters. Overall, SpyTag should enablenew possibilities for cell biologists to probe the effect of force incells (49) and for biochemists to create new protein architec-tures (34).

Materials and MethodsIn Vitro Reconstitution Reactions. Cloning, protein expression, statistical ana-lysis, isothermal titration calorimetry, AFM, microscopy, in vivo reconstitu-tion, mass spectrometry, and immunoblotting are described fully in the SIAppendix. Briefly, SpyTag-MBP, SpyCatcher, and variants of these proteinswere expressed in E. coli and purified using Ni-NTA resin.

Amide bond formation between the protein and peptide binding part-ners was monitored by SDS-PAGE. To demonstrate covalent reconstitution(Fig. 1D), proteins were mixed at 10 μM in PBS pH 7.4 at 25 °C for 3 h. Allquantified reactions were performed in triplicate. To stop reactions, sampleswere heated in SDS loading buffer on a Bio-Rad C1000 thermal cycler at 95 °Cfor 7 min. SDS-PAGE was performed on 14% polyacrylamide gels, using anX-cel SureLock (Life Technologies) at 200 V for approximately 1 h. Gels werestained with Instant Blue Coomassie stain (Triple Red Ltd.) and band inten-sities were quantified using a Gel Doc XR imager and Image Lab 3.0 software(Bio-Rad).

Zakeri et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ E695

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

Reactions for analyzing speed, pH-dependence, temperature-depen-dence, and the reversibility of amide bond formation were performed bymixing 10 μM of each protein in 40 mM Na2HPO4 with 20 mM citric acidpH 7.0 (phosphate-citrate) for the indicated time at 25 °C, or at the indicatedtemperature and pH. (PBS alone would not enable proper buffering over thepH range explored.) For determining temperature dependence all reactionswere incubated in a Bio-Rad C1000 thermal cycler at 4, 25, and 37 °C with aheated lid to prevent evaporation.

To calculate the rate constant, SpyTag-MBP and SpyCatcher at 10 μMweremixed in triplicate in phosphate-citrate and incubated at 25 °C for 1, 3, or5 min, in the linear part of the reaction. Samples were then heated to95 °C for 7 min in SDS loading buffer and analyzed on 14% SDS-PAGE withCoomassie staining. Unreacted SpyCatcher concentration was quantifiedfrom band intensity as above. 1∕½unreacted Spy Catcher�was plotted againsttime and a straight line, whose gradient corresponds to the second order rateconstant, was fitted using the “LINEST” linear least squares curve-fittingroutine in Excel. The units were converted from μM−1 min−1 to M−1 s−1.The correlation coefficient and error bars to the gradient were calculatedusing the LINEST function in Excel. The reaction half-time was calculated fromt12¼ 1∕k½SpyCatcher�0.To test reversibility, 10 μM SpyTag-MBP and 10 μM SpyCatcher in phos-

phate-citrate were incubated for 3 h at 25 °C to allow reaction to take place.100 μM Cna peptide (RSGAHIVMVDAGSR, made by solid-phase synthesis at98% purity by Insight Biotechnology Ltd.) was then added and incubated at25 °C for 16 h, before SDS-PAGE. In control experiments, 10 μM SpyCatcher orSpyCatcher EQ was incubated with 100 μM Cna peptide for 3 h at 25 °C in thesame buffer. Also, 10 μM SpyCatcher or SpyCatcher EQ was incubated with10 μM SpyTag-MBP for 3 h at 25 °C in the same buffer.

The survey of SpyCatcher reactivity against the series of peptide-MBPvariants was performed by mixing each protein at 10 μM in PBS pH 7.4and incubating at 25 °C for 30 min.

For analyzing the effect of various buffers on the covalent reconstitutionbetween SpyCatcher and SpyTag-MBP, each protein was mixed at 10 μM ineither phosphate buffered saline (PBS) pH 7.4, phosphate-citrate, 50 mM Tris(tris-hydroxymethyl aminomethane) pH 7.0, or 50 mM Hepes [4-(2-hydro-xyethyl)-1-piperazine ethanesulfonate] pH 7.0 and incubated at 25 °C for 1or 10 min.

For analyzing the effect of detergent on amide bond formation, Spy-Catcher and SpyTag-MBP were mixed at 10 μM for 3 h at 25 °C in either

PBS pH 7.4, or PBS pH 7.4 containing 1% Triton X-100, 1% Tween 20, or0.5% Nonidet P-40.

Concentration-dependence of amide bond formation was analyzed byincubating SpyCatcher and SpyTag-MBP both at 1, 5, or 10 μM in phosphate-citrate pH 7.0 at 25 °C for the indicated times.

For analyzing the effect of reducing agent on amide bond formation,SpyCatcher and SpyTag-MBP were mixed at 10 μM in phosphate-citratepH 7.0 with or without 10 mM final concentration of freshly prepared dithio-threitol (DTT) for 10 min at 25 °C.

To determine how SpyTag reacted at terminal or internal locations, N-Spy-Tag-MBP or MBP-SpyTag-Zif-SpyTag was incubated with SpyCatcher orSpyCatcher EQ for 30 min at pH 7.0 in phosphate-citrate buffer at 25 °C andanalyzed by SDS-PAGE and Coomassie staining; proteins were present at10 μM, except for those marked 3× which were present at 30 μM.

Calculation of percent reconstitution was performed by dividing the in-tensity of the band for the covalent complex by the intensity of all the bandsin the lane, then multiplying by 100. To analyze the speed of SpyCatcher:Spy-Tag-MBP covalent complex formation, the reduction in intensity of the bandfor SpyCatcher was monitored relative to a control not incubated withSpyTag-MBP.

Cell Culture and Labeling.HeLa cells were grown in DMEMwith 10% Fetal CalfSerum, 50 U∕mL penicillin and 50 μg∕mL streptomycin. HeLa cells weretransfected with a plasmid encoding SpyTag-ICAM1-GFP using TurboFect(Fermentas) following manufacturer’s instructions. After 48 h, transfectantswere selected and maintained with 5 μg∕mL puromycin. HeLa cells stably ex-pressing SpyTag-ICAM1-GFP were washed with PBS containing 5 mM MgCl2(PBS-Mg) and incubated in PBS-Mg containing 1% BSA and 5 μM SpyCatcher-Alexa Fluor 555 or 5 μM SpyCatcher EQ-Alexa Fluor 555 for 15 min at 25 °C.Cells were then washed twice in PBS-Mg at 4 °C and imaged live.

ACKNOWLEDGMENTS. Funding was provided by the Clarendon Fund (B.Z.,J.O.F.), the National Institutes of Health (V.T.M., E.C.), the National ScienceFoundation (V.T.M.), the James and Esther King Biomedical Research Program(V.T.M), the School of Biology of the University of St. Andrews (U.S-L.), OxfordUniversity Department of Biochemistry (B.Z., E.C.C. and M.H.), St. Peter’s Col-lege (B.Z.), New College (J.O.F.), and Worcester College (E.C.C. and M.H.).

1. Jarvik JW, Telmer CA (1998) Epitope tagging. Annu Rev Genet 32:601–618.2. Griffin BA, Adams SR, Tsien RY (1998) Specific covalent labeling of recombinant

protein molecules inside live cells. Science 281:269–272.3. Allen KN, Imperiali B (2010) Lanthanide-tagged proteins–an illuminating partnership.

Curr Opin Chem Biol 14:247–254.4. Halo TL, Appelbaum J, Hobert EM, Balkin DM, Schepartz A (2009) Selective recognition

of protein tetraserine motifs with a cell-permeable, pro-fluorescent bis-boronic acid.J Am Chem Soc 131:438–439.

5. Huh WK, et al. (2003) Global analysis of protein localization in budding yeast. Nature425:686–691.

6. Stanfield RL, Wilson IA (1995) Protein-peptide interactions. Curr Opin Struct Biol5:103–113.

7. Korndorfer IP, Skerra A (2002) Improved affinity of engineered streptavidin for theStrep-tag II peptide is due to a fixed open conformation of the lid-like loop at thebinding site. Protein Sci 11:883–893.

8. Houk KN, Leach AG, Kim SP, Zhang X (2003) Binding affinities of host-guest, protein-ligand, and protein-transition-state complexes. Angew Chem Int Ed Engl 42:4872–4897.

9. Horne WS, Gellman SH (2008) Foldamers with heterogeneous backbones. Acc ChemRes 41:1399–1408.

10. Kang HJ, Baker EN (2010) Intramolecular isopeptide bonds: protein crosslinks built forstress? Trends Biochem Sci 36:229–237.

11. Oke M, et al. (2010) The scottish structural proteomics facility: targets, methods andoutputs. Journal of Structural and Functional Genomics 11:167–180.

12. Hagan RM, et al. (2010) NMR spectroscopic and theoretical analysis of a spontaneouslyformed Lys-Asp isopeptide bond. Angew Chem Int Ed Engl 49:8421–8425.

13. Amelung S, et al. (2011) The FbaB-type fibronectin-binding protein of Streptococcuspyogenes promotes specific invasion into endothelial cells. Cell Microbiol 13:1200–1211.

14. Zettler J, Schutz V, Mootz HD (2009) The naturally split Npu DnaE intein exhibits anextraordinarily high rate in the protein trans-splicing reaction. FEBS Lett 583:909–914.

15. Popp MW, Antos JM, Grotenbreg GM, Spooner E, Ploegh HL (2007) Sortagging: aversatile method for protein labeling. Nat Chem Biol 3:707–708.

16. KullmannW (1985) Proteases as catalytic agents in peptide synthetic chemistry—shift-ing the extent of peptide-bond synthesis from a quantite negligible to a quantiteconsiderable. J Protein Chem 4:1–22.

17. McDonald CA, Yang JY, Marathe V, Yen TY, Macher BA (2009) Combining results fromlectin affinity chromatography and glycocapture approaches substantially improvesthe coverage of the glycoproteome. Mol Cell Proteomics 8:287–301.

18. BeyerMK, Clausen-SchaumannH (2005)Mechanochemistry: themechanical activationof covalent bonds. Chem Rev 105:2921–2948.

19. Marszalek PE, et al. (1999) Mechanical unfolding intermediates in titin modules. Nat-ure 402:100–103.

20. Moy VT, Florin EL, Gaub HE (1994) Intermolecular forces and energies between ligandsand receptors. Science 266:257–259.

21. Neuert G, Albrecht C, Pamir E, Gaub HE (2006) Dynamic force spectroscopy of thedigoxigenin-antibody complex. FEBS Lett 580:505–509.

22. Celik E, Moy VT (2012) Nonspecific interactions in AFM force spectroscopy measure-ments. J Mol Recognit 25:53–56.

23. Crampton N, Brockwell DJ (2010) Unravelling the design principles for single proteinmechanical strength. Curr Opin Struct Biol 20:508–517.

24. Huang J, Nagy SS, Koide A, Rock RS, Koide S (2009) A peptide tag system for facilepurification and single-molecule immobilization. Biochemistry 48:11834–11836.

25. Kang HJ, Coulibaly F, Clow F, Proft T, Baker EN (2007) Stabilizing isopeptide bondsrevealed in gram-positive bacterial pilus structure. Science 318:1625–1628.

26. Ton-That H, Marraffini LA, Schneewind O (2004) Protein sorting to the cell wall envel-ope of Gram-positive bacteria. Biochim Biophys Acta 1694:269–278.

27. Alegre-Cebollada J, Badilla CL, Fernandez JM (2010) Isopeptide bonds block themechanical extension of pili in pathogenic Streptococcus pyogenes. J Biol Chem285:11235–11242.

28. Hu XQ, et al. (2011) Autocatalytic intramolecular isopeptide bond formation in gram-positive bacterial pili: a QM/MM simulation. J Am Chem Soc 133:478–485.

29. Zakeri B, Howarth M (2010) Spontaneous intermolecular amide bond formation be-tween side chains for irreversible peptide targeting. J Am Chem Soc 132:4526–4527.

30. Huang J, Makabe K, Biancalana M, Koide A, Koide S (2009) Structural basis forexquisite specificity of affinity clamps, synthetic binding proteins generated throughdirected domain-interface evolution. J Mol Biol 392:1221–1231.

31. Rodriguez P, et al. (2006) Isolation of transcription factor complexes by in vivo biotinyla-tion tagging and direct binding to streptavidin beads. Methods Mol Biol 338:305–323.

32. Nonaka H, Tsukiji S, Ojida A, Hamachi I (2007) Non-enzymatic covalent protein labelingusing a reactive tag. J Am Chem Soc 129:15777–15779.

33. Uttamapinant C, et al. (2010) A fluorophore ligase for site-specific protein labelinginside living cells. Proc Natl Acad Sci USA 107:10914–10919.

34. Woolfson DN (2005) The design of coiled-coil structures and assemblies. Adv ProteinChem 70:79–112.

35. Kent SB (2009) Total chemical synthesis of proteins. Chem Soc Rev 38:338–351.36. Chattopadhaya S, Abu Bakar FB, Srinivasan R, Yao SQ (2008) In vivo imaging of a

bacterial cell division protein using a protease-assisted small-molecule labelingapproach. Chembiochem 9:677–680.

37. Chin JW, et al. (2002) Addition of p-azido-L-phenylalanine to the genetic code ofEscherichia coli. J Am Chem Soc 124:9026–9027.

E696 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1115485109 Zakeri et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020

38. Lee HS, Dimla RD, Schultz PG (2009) Protein-DNA photo-crosslinking with a geneticallyencoded benzophenone-containing amino acid. BioorgMed Chem Lett 19:5222–5224.

39. Suchanek M, Radzikowska A, Thiele C (2005) Photo-leucine and photo-methionineallow identification of protein-protein interactions in living cells. Nat Methods2:261–267.

40. Hancock SM, Uprety R, Deiters A, Chin JW (2010) Expanding the genetic code of yeastfor incorporation of diverse unnatural amino acids via a pyrrolysyl-tRNA synthetase/tRNA pair. J Am Chem Soc 132:14819–14824.

41. Shi J, Muir TW (2005) Development of a tandem protein trans-splicing system based onnative and engineered split inteins. J Am Chem Soc 127:6198–6206.

42. Tavassoli A, Benkovic SJ (2007) Split-intein mediated circular ligation used in thesynthesis of cyclic peptide libraries in E. coli. Nat Protoc 2:1126–1133.

43. Siebold C, Erni B (2002) Intein-mediated cyclization of a soluble and a membraneprotein in vivo: function and stability. Biophys Chem 96:163–171.

44. Griffiths AD, Duncan AR (1998) Strategies for selection of antibodies by phage display.

Curr Opin Biotechnol 9:102–108.

45. Laitinen OH, Hytonen VP, Nordlund HR, Kulomaa MS (2006) Genetically engineered

avidins and streptavidins. Cell Mol Life Sci 63:2992–3017.

46. Chivers CE, et al. (2010) A streptavidin variant with slower biotin dissociation and

increased mechanostability. Nat Methods 7:391–393.

47. Perumal SK, Raney KD, Benkovic SJ (2010) Analysis of the DNA translocation and

unwinding activities of T4 phage helicases. Methods 51:277–288.

48. Pierres A, Touchard D, Benoliel AM, Bongrand P (2002) Dissecting streptavidin-biotin

interaction with a laminar flow chamber. Biophys J 82:3214–3223.

49. Vogel V, Sheetz MP (2009) Cell fate regulation by coupling mechanical cycles to

biochemical signaling pathways. Curr Opin Cell Biol 21:38–46.

Zakeri et al. PNAS ∣ March 20, 2012 ∣ vol. 109 ∣ no. 12 ∣ E697

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

6, 2

020