Site‐Specific Immobilization of the Peptidoglycan Synthase PBP1B on a Surface Plasmon Resonance Chip Surface

Abstract Surface plasmon resonance (SPR) is one of the most powerful label‐free methods to determine the kinetic parameters of molecular interactions in real time and in a highly sensitive way. Penicillin‐binding proteins (PBPs) are peptidoglycan synthesis enzymes present in most bacteria. Established protocols to analyze interactions of PBPs by SPR involve immobilization to an ampicillin‐coated chip surface (a β‐lactam antibiotic mimicking its substrate), thereby forming a covalent complex with the PBPs transpeptidase (TP) active site. However, PBP interactions measured with a substrate‐bound TP domain potentially affect interactions near the TPase active site. Furthermore, in vivo PBPs are anchored in the inner membrane by an N‐terminal transmembrane helix, and hence immobilization at the C‐terminal TPase domain gives an orientation contrary to the in vivo situation. We designed a new procedure: immobilization of PBP by copper‐free click chemistry at an azide incorporated in the N terminus. In a proof‐of‐principle study, we immobilized Escherichia coli PBP1B on an SPR chip surface and used this for the analysis of the well‐characterized interaction of PBP1B with LpoB. The site‐specific incorporation of the azide affords control over protein orientation, thereby resulting in a homogeneous immobilization on the chip surface. This method can be used to study topology‐dependent interactions of any (membrane) protein.


Introduction
Surfacep lasmon resonance (SPR) is ap owerful technique for the kinetic characterization of biomolecular interactions. This technique requires the immobilization of one part of an interacting pair to as ensor-chip surface, over which the second part is passed in solution. Bindingo ft he soluble analyte to the immobilized ligand generates an SPR sensorgram,w hich is a plot of arbitrary responseo rr esonance units against time. The resonance units result from the change in refractivei ndex at the chip surfaceu pon analyte binding, as measured by sensitive opticala pparatus. The generated data can be used to calculate the kinetic parameters of an interaction, such as associa-tion and dissociation rate constants and hence affinity,a sw ell as the equilibrium constant of the interaction. [1][2][3] Common methods for protein immobilizationi nvolvec ovalent coupling on an SPR chip surface to naturally occurring amine or thiol groups within the protein. Immobilization at amines occurs after reaction with N-hydroxysuccinimide (NHS) esters coupled to the chip surface during manufacture. [4] For thiol coupling, reactive disulfide or maleimide groups can be introduced on the chip surface. Other functional groups, such as aldehydes, can also be chemically introduced in the protein to allow specific coupling to the chip surface; [5] affinity fusion tags, such as glutathione S-transferase, maltose bindingp rotein, poly-histidine tags in combinationw ith glutathione,a mylose and nickel-NTA-functionalized surfaces have also been used. [1,3] SPR has been used to study numerousi nteractions in biological systems, including those of penicillin-bindingp roteins (PBPs).
PBPs are peptidoglycan( PG) synthesis enzymes responsible for the final steps in the productiono ft he major component of the bacterial cell wall. [6] PG is am esh-like heteropolymer composed of glycan strands interconnectedb ys hort peptides, and is synthesizeda tt he outer leaflet of the cytoplasmic membrane.I ti ss ynthesized from lipid II by two enzymatic reactions:p olymerization of glycan strandsb yg lycosyltransferase (GTase) reactions, and cross-linkage of peptides by transpeptidase (TPase) reactions. [7][8][9] PBPs form af amily of enzymes with members capable of either TPase activity or both GTase and TPase;t hey are so namedb ecause they readily form covalent complexes with penicillin and other b-lactam antibiotics at their TPase domains. [10][11][12] All PG synthases are anchored to the cytoplasmic membraneb yasingle transmembrane helix, with Surfacep lasmon resonance (SPR)i so ne of the most powerful label-free methodst od etermine the kinetic parameters of molecular interactions in real time and in ah ighly sensitivew ay. Penicillin-binding proteins (PBPs) are peptidoglycan synthesis enzymesp resent in most bacteria. Established protocols to analyze interactions of PBPs by SPR involvei mmobilizationt oa n ampicillin-coated chip surface( ab-lactam antibiotic mimicking its substrate), thereby forming ac ovalentc omplex with the PBPs transpeptidase (TP) active site. However,P BP interactions measured with as ubstrate-bound TP domain potentially affect interactions near the TPase active site. Furthermore, in vivo PBPs are anchored in the inner membrane by an N-terminal transmembrane helix, and hence immobilization at the C-terminal TPase domain gives an orientation contrary to the in vivo situation. We designed an ew procedure:i mmobilization of PBP by copper-free click chemistry at an azide incorporated in the Nterminus. In ap roof-of-principle study,w ei mmobilized Escherichia coli PBP1B on an SPR chip surface and used this for the analysiso ft he well-characterized interaction of PBP1B with LpoB. The site-specific incorporationo ft he azide affords control over protein orientation,t hereby resulting in a homogeneous immobilization on the chip surface. This methodcan be used to study topology-dependenti nteractions of any (membrane) protein.
the catalytic site on the outside. Escherichia coli is the beststudied modelo rganism for the interactions of PG enzymes. PBP1B, am ajor E. coli PG synthase, hasb oth GTase and TPase activities. [13][14][15] Severali nteractions of PBP1B have been characterized by SPR, including with other PG synthesis enzymes (PBP3, MltA-MipA) and regulatory proteins (LpoB, CpoB, FtsN). [16][17][18][19][20] SPR was also used to demonstrate dimerization of PBP1B. [21] In all these examples, PBP1B (or PBP3) was immobilized onto the chip surfacea ti ts TPase domain.T his was achieved by coupling ampicillin to the chip surface ands ubsequently applying the PBP,t hereby resulting in ac ovalenti nteraction of ampicillin with the active site of the TPase domain of the PBP.
The oriented coupling of PBPs to the chip surface via ampicillin is suboptimal in some cases. As the TPase active site is occupied in this immobilization strategy,a ny interaction interfaces proximal to this positionc an be occluded and/or altered compared to the apo state. Furthermore, in the cell PBPs are anchored in the cytoplasmic membrane by an N-terminal transmembrane helix, with the majority of the protein oriented outwards andt he TPase domain typically furthestf rom the point of anchoring ( Figure 1). Thus, immobilization of aP BP by its TPase domain gives ac ontrary orientation to that in vivo, thus exposing the GTase domain and membrane anchor and potentially occluding interaction sites or hindering access of analyte molecules.
To address this, we have designed an ew immobilization method, based on site-specific labeling, that can be generally appliedt oa ny (membrane) protein. Many different bioorthog-onal chemical reactions have been described to site-specifically label proteins for surfacei mmobilization. [22,23] We chose to anchorP BP1B to ac hip surfaceb ys ite-specific incorporation of an azide-containing unnatural amino acid in the N-terminal sequence of the protein (Figure 1), followed by covalent attachment to immobilized dibenzylcyclooctyne by copper-free click chemistry,a st his reactiono ccurs spontaneously under physiological conditions and does not need metal catalysts, which can have undesirable effects on protein activity. [23] Immobilization by this method yields the correct topological orientation of the protein with an accessible and unaltered TPase domain,t hereby allowing characterization of interactions with this domain.

Results and Discussion
Confirmation of the presence of the azide in the PBP1B mutantsb yc oupling to acyclooctyne-containing fluorescent dye An azide-containing unnatural amino acid was incorporated in the N-terminal tail of the protein by using nonsense suppression mutagenesis.T his azide was used to covalently attach the protein to the dibenzylcyclooctyne-coated chip surface by copper-free click chemistry.B ecause this is an ew method and there is no information about the efficiency of this immobilization method and the dependency on the position of the azide, we substituted three adjacent amino acids in the N-terminal region of PBP1B for the unnatural amino acid p-azidophenylalanine. By site-directed mutagenesis, the codon for Gly53, Lys54, or Gly55 of PBP1B was mutated to an amber (TAG) codon.W hen each mutated PBP1B variant was expressed with an orthogonal tRNA/aminoacyl-tRNA synthase pair that recognizes TAGa nd is specific for the incorporation of the unnatural amino acid p-azidophenylalanine, three mutantp roteins were produced:a zidophenylalanine in place of either Gly53,L ys54, or Gly55.I no rdert ov erify azide incorporation,w ei ncubated purifiedprotein with afluorescent dye containing acyclooctyne group, which spontaneously reacts with the azide (Figure 2A). In this way,t he azide-containing proteins are fluorescently tagged. This reaction mixture was separated by SDS-PAGE, and the gel was scanned with af lorescence scanner to visualize labeled protein,t hen stained with Coomassie Brilliant Blue to assesst he total protein content loaded. Thea zide was indeed incorporated into all three mutant proteins, according to the fluorescences ignals (Figure2B). An excesso fc yclooctyne-containingd ye was neededf or an efficient reactionu nder these conditions (1:1 vs 10:1). Incubation with wild-type protein did not result in fluorescence, thus showingt hat the reactionw as specific.

Azide-containing PBP1Bproteins showb oth GTase and TPase activity in an in vitro peptidoglycan synthesis assay
For the implementation of our SPR method, we used fully active PBP1B proteins (both GTase and TPase activities). We performeda ni nv itro PG synthesis assay to verify that the  [34] showing the previously used immobilization site (serine residue in the actives ite of the TPase domain)a nd the site used in our immobilization strategy (cytoplasmic tail). azide-containing proteins retained both activities. Protein was incubated in ab uffer containing all the ingredients needed for activity,s upplemented with fluorescently labeled lipid II for the detection of the produced polymers. This labeled lipid II cannotb eu sed as as ubstrate for the crosslink-formingT Pase reaction, as the positiono ft he attached fluorophore is the positionu sed in crosslink formation. Furthermore, E. coli PBP1B needs al ipid II version with a meso-diaminopimalic acid at this positiono ft he donorp eptidef or crosslink formation, and the labeled version originated from al ysine version.T herefore, unlabeled meso-diaminopimalic acid lipid II was included in the mixturesf or the TPase reaction.
In order to analyze solely GTase activity,p enicillin Gw as added to some of the reaction mixtures (to inhibitTPase activity). As ar esult of GTase activity,s ugar moieties of lipid II were polymerized into glycan strands. These glycans trands were separatedb yT ris/Tricine SDS-PAGE. [24][25][26][27][28][29]    These results show that the three mutant proteins are fully active (both GTase and TPase) in this in vitro PG synthesis assay.
Use of azide-incorporatedPBP1B for site-specific immobilization on an SPR chip Optimization of immobilization conditions: For immobilization of the azide-containing PBP1B variants, we used an aminefunctionalized chip surface to perform the SPR experiments. First, it was functionalized by using the amine-reactive sulfo-dibenzylcyclooctyne-NHS ester.A st he efficiencies of functionalization and the subsequentc lick-reaction with the azide in the proteins were not known,w ev aried the concentration of sulfo-dibenzylcyclooctyne-NHS ester from 0.25 to 1mm and the protein concentration from 0.04 to 0.5 mm.I no rder to identify the optimal conditions for PBP1B immobilization and interaction measurement, we used the well-characterizedi nteraction between PBP1B and LpoBa satest system,a st he kinetic parameters of this interaction have been well established. [18,24,30] The amount of protein bound to ac hip surface is represented by the response of local ligand (RLL) value, expressed in resonanceu nits (RUs). 1RUc orresponds to approximately 1pgmm À2 ,a nd the binding capacity (R max )d ependso nt he amount of protein immobilized on the chip surface according to R max = (analyte MW/ligand MW) RL Sm (stoichiometric ratio). At ypical RLL values for our type of measurement is 1000 RU.
All three azido-protein variants were well immobilized on the chip surface, thus suggesting that the position of the azide is not crucial for immobilization efficiency in this case ( Figure 4). The highest protein concentrationt ested (0.5 mm) resultedi nt he highest amount of immobilized protein on the chip without causing protein aggregation,w hich would render the protein inactive. Sulfo-dibenzylcyclooctyne-NHSe ster concentration did not affect immobilization efficiency or the SPR signals (datan ot shown). Therefore, we used 0.5 mm protein and 1mm sulfo-dibenzylcyclooctyne-NHS ester with variant Gly55 in furthere xperiments. This variant was slightly more active in the in vitro PG synthesis assayt han PBP1B-Gly53, and on average produced SPR curvesw ith ah ighers ignal than PBP1B-Lys54 upon injection of LpoB (PBP1B-bindinga nalyte). We decided to include an azidoethanol blockings tep because this resulted in slightly higher responses under the above conditions and, more importantly,i no rder to block possible hydrophobic interactions between injected protein and free cyclooctyneg roups on the chip. Ab locking step (with ethanolamine) is included in the ampicillin immobilization method, so including ab lockings tep in our methoda lso allowed ab etter comparison between the two methods.T he results of the all optimization experiment are showni nT able S2 in the Supporting Information.

Do the immobilized PBP1Bvariants still interact with LpoB in as imilar way?
Next, we immobilizedP BP1B on every spot of the chip (except for some controls pots) with the optimized conditions.I njection of LpoB over the PBP1B-immobilizedS PR surfacer esulted as an increase in RU;s topping injection resulted in release of the interacting molecules, and thus ad ecreasei nR U.
The sensorgram for the injection of LpoB over immobilized PBP1B ( Figure 5, left) shows that LpoB has avery quick association with PBP1B, as published before. [18] The immediate rise in RU to the equilibrium made it impossible to determine the association rate constant.T he same holds for the dissociation of LpoB from PBP1B when injection ceased. The maximum reso- nance unit (maxRU)v alues for the different analyte concentrations were plotted by non-linear regressionw ith the formula y = B max x/(K D + x)( one site saturation in the simple ligand-binding tool of SigmaPlot), in order to determine the equilibrium constant. This resulted in calculated K D values of 0.71-0.97 mm with as tandard deviation of AE 0.052, which is close to the 0.81 AE 0.08 mm found by Egane tal. [18] The small differences in equilibrium constant could have arisen from slight differences in buffer composition, pH or temperature at which the measurements were performed.
These resultsshow that this new PBP1B-immobilization technique, with an azide incorporated in the protein cytoplasmic tail, is ag ood alternative to the ampicillin-immobilization methodf or SPR experiments. We show that it produces similar results when analyzing the interaction of PBP1B with LpoB. We have not yet identified the specific interactions (of the TPase domain of PBP1B) that would be preferably analyzed by our new immobilization strategy. Incubation of PBP1B with ampicillin prior to immobilization did not alter the binding of LpoB to PBP1B (data not shown), thus suggesting that the interaction is independent on the state of the TPase domain. This also implies that the activation of the TPase by LpoB does not depend on the availability of aT Pase substrate, consistentwith primary activation of the GTase by LpoB. [18,31] This new immobilization methodc an be used for the immobilizationo fa ny desired protein, and createst he possibility to control the orientation of the protein by the site specific incorporation of the azide. Replacing different surface amino acids and homogeneously orientating the protein on the chip surface opens the possibility to study the topology-dependence of interactions of membrane proteins.

Experimental Section
Bacterial strains and plasmids: Escherichia coli DH5a cells were used for DNA amplification. E. coli BL21(DE3) cells were used for protein expression. Plasmid pDML924 carrying the mrcB gene, which encodes the N-terminal His 6 -tagged variant PBP1Bg (a gift from Mohammed Te rrak, University of Liege, Belgium), [15] was used for overexpression of PBP1B and as at emplate for the generation of PBP1B mutants. Plasmid pEvol-pAzF encoding the orthogonal aminoacyl tRNA synthase-tRNA CUA pair was used for incorporation of p-azidophenylalanine at the site of an amber mutation. Plasmid pET28LpoB (signal sequence and lipid anchor of LpoB replaced by an oligohistidine tag, LpoB(sol)) was used for the overexpression of LpoB (sol). [18] Site-directed mutagenesis: The amber mutants were created by mutagenesis PCR (primers in Ta ble S1). The reaction mixture contained fwd and rev primer (125 ng), dNTPs (10 mm each, 1 mL), template DNA (DNA at an end concentration of 1.23 ng mL À1 ,1mL of a6 1.5 ng mL À1 )a nd Phusion DNA polymerase (1 U, 0.5 mL; Thermo Fisher Scientific) in at otal volume of 50 mLi n[ 1 Phusion buffer].W ep erformed 17 cycles of 30 sa t98 8C, 1min at T m (depending on the primers), and 5min at 72 8C. PCR products were digested with DpnI (10 U; Fermentas) and amplified in E. coli DH5a. Sequencing confirmed the intended mutations.
SPR studies: An IBIS-MX96 (IBIS Technologies, Enschede, The Netherlands) was used. PBP1B variants were immobilized on the surface of aSensEye P-NH2 sensor (IBIS) coated with sulfo-dibenzylcyclooctyne-NHS ester (Jena Bioscience, Jena, Germany). After activation of the chip, the spots were coated for 60 min with sulfo-dibenzylcyclooctyne-NHS ester (1, 0.5, or 0.25 mm in HEPES (20 mm, pH 7.5)). After ar inse with PBS, PBP1B (0.5, 0.2, or 0.04 mm) in running buffer (Tris/maleate (10 mm, pH 7.5), NaCl (150 mm), Triton X 100 (0.05 %)) was spotted for 30 min. After ar inse with running buffer,e xcess sulfo-dibenzylcyclooctyne-NHS ester was blocked with azidoethanol (0.5 m in running buffer) for 10 min. The amount of immobilized protein ranged from 1000 to 3000 RU (1 RU corresponds to approximately 1pgo fp rotein per mm 2 ). Some spots were controls with different treatments (Table S2). Analytes in running buffer were injected at different concentrations:2min baseline, 10 min association, 5min dissociation, and two times 30 so f regeneration with running buffer (containing NaCl (1 m)) were recorded. The sensorgrams were evaluated with SPRintX (IBIS), and the parameters of the interaction were calculated in SigmaPlot (Systat Software, San Jose, CA) by using the simple ligand-binding tool and one-site saturation.