Mutations in penicillin-binding protein 2 from cephalosporin-resistant Neisseria gonorrhoeae hinder ceftriaxone acylation by restricting protein dynamics

The global incidence of the sexually transmitted disease gonorrhea is expected to rise due to the spread of Neisseria gonorrhoeae strains with decreased susceptibility to extended-spectrum cephalosporins (ESCs). ESC resistance is conferred by mosaic variants of penicillin-binding protein 2 (PBP2) that have diminished capacity to form acylated adducts with cephalosporins. To elucidate the molecular mechanisms of ESC resistance, we conducted a biochemical and high-resolution structural analysis of PBP2 variants derived from the decreased-susceptibility N. gonorrhoeae strain 35/02 and ESC-resistant strain H041. Our data reveal that mutations both lower affinity of PBP2 for ceftriaxone and restrict conformational changes that normally accompany acylation. Specifically, we observe that a G545S substitution hinders rotation of the β3 strand necessary to form the oxyanion hole for acylation and also traps ceftriaxone in a noncanonical configuration. In addition, F504L and N512Y substitutions appear to prevent bending of the β3–β4 loop that is required to contact the R1 group of ceftriaxone in the active site. Other mutations also appear to act by reducing flexibility in the protein. Overall, our findings reveal that restriction of protein dynamics in PBP2 underpins the ESC resistance of N. gonorrhoeae.

The global incidence of the sexually transmitted disease gonorrhea is expected to rise due to the spread of Neisseria gonorrhoeae strains with decreased susceptibility to extended-spectrum cephalosporins (ESCs). ESC resistance is conferred by mosaic variants of penicillin-binding protein 2 (PBP2) that have diminished capacity to form acylated adducts with cephalosporins. To elucidate the molecular mechanisms of ESC resistance, we conducted a biochemical and high-resolution structural analysis of PBP2 variants derived from the decreased-susceptibility N. gonorrhoeae strain 35/02 and ESC-resistant strain H041. Our data reveal that mutations both lower affinity of PBP2 for ceftriaxone and restrict conformational changes that normally accompany acylation. Specifically, we observe that a G545S substitution hinders rotation of the ␤3 strand necessary to form the oxyanion hole for acylation and also traps ceftriaxone in a noncanonical configuration. In addition, F504L and N512Y substitutions appear to prevent bending of the ␤3-␤4 loop that is required to contact the R1 group of ceftriaxone in the active site. Other mutations also appear to act by reducing flexibility in the protein. Overall, our findings reveal that restriction of protein dynamics in PBP2 underpins the ESC resistance of N. gonorrhoeae.

Gonorrhea, a sexually-transmitted infection caused by
Neisseria gonorrhoeae, is increasingly difficult to treat due to emergence of strains exhibiting decreased susceptibility or resistance to the extended-spectrum cephalosporins (ESCs) 3 ceftri-axone and cefixime, which are the last remaining options for monotherapy of gonorrhea. As a result, the Centers for the Disease Control and Prevention in the United States and many other public health agencies worldwide have revised their treatment guidelines for gonorrhea from monotherapy with ceftriaxone or cefixime to dual therapy with ceftriaxone and azithromycin (1)(2)(3)(4). An important event that led to this change was the isolation in Japan of the first N. gonorrhoeae strain (H041) exhibiting high-level resistance to cefixime and ceftriaxone (5). The minimum inhibitory concentrations (MICs) of ceftriaxone and cefixime for H041 are 2 and 8 g/ml, respectively, and well-above the EUCAST breakpoints (Ͼ0.125 g/ml) for these antibiotics (6). A second ESC-resistant (ESC R ) strain (F89) was subsequently isolated in France and Spain (7,8). Consistent with the spread of ESC R strains, there have now been several reports of treatment failures with ESCs (9 -12), and the first international spread of a ceftriaxone-resistant strain has been verified (2). In 2018, the first strain with ceftriaxone resistance in combination with high-level azithromycin resistance was identified (13,14).
A number of resistance determinants contribute to ␤-lactam resistance to N. gonorrhoeae, including mtrR, penB, and ponA (15)(16)(17)(18)(19), but the primary determinant in the emergence of ESC R strains in N. gonorrhoeae is acquisition of so-called mosaic alleles of the penA gene through allelic replacement (20). The penA gene encodes penicillin-binding protein 2 (PBP2), an essential transpeptidase (TPase) that catalyzes the formation of cross-links between peptides emanating from parallel glycan strands during the latter stages of peptidoglycan synthesis (21). PBP2 is a two-domain protein composed of a TPase catalytic domain and an N-terminal domain of unknown function (22). This latter domain may act as a pedestal to project the active site toward peptidoglycan or could mediate protein-protein interactions with other components of the peptidoglycan machinery. Although evidence has uncovered an allosteric site in the N-terminal domain of Staphylococcus aureus PBP2a (23), there is no indication of a similar site in PBP2.
PBP2 recognizes and reacts with the D-Ala-D-Ala C terminus of the peptide chain to form an acyl-enzyme complex with the penultimate D-Ala, and then it catalyzes the transfer of the acylated peptide to the amino group of meso-diaminopimelic acid from an adjacent peptide strand to form an amide cross-link. All ␤-lactam antibiotics, including ceftriaxone and cefixime, are mimics of the peptide D-Ala-D-Ala C terminus and react with PBPs to form a long-lived acylated complex that inhibits the enzyme and leads to loss of TPase activity required for proper peptidoglycan synthesis (for PBP reviews, see Refs. 24 -26).
The active sites of PBPs contain three signature motifs: SXXK, SXN, and KTG, where X is a variable amino acid. In PBP2, Ser-310 of the SXXK motif is the site of acylation by both peptide substrate and ␤-lactams. PBP2 variants from ESC R strains contain upwards of 60 amino acid mutations compared with PBP2 from the penicillin-and cephalosporin-susceptible strain FA19 (7,20,27). These variants have arisen through homologous recombination of DNA segments from commensal Neisseria species and are therefore referred to as mosaic penA alleles. Prior investigation of PBP2 from the ESC-decreased susceptibility strain 35/02 showed that out of the 57 mutations compared with FA19, three (I312M, V316T, and G545S) were responsible for the majority of resistance to ESCs and that mutations present in the N-terminal domain do not contribute (28). Strikingly, the penA gene from H041 (penA41) can transform FA19 to full resistance for cefixime and ceftriaxone in the absence of other resistance determinants (29). Its PBP2 contains 61 amino acid mutations compared with FA19 (20), and 13 of these are new or different compared with 35/02. Three of the mutations (A311V, V316P, and T483S) significantly increase the MICs of ceftriaxone and cefixime when introduced into PBP2 35/02 (29). Two additional mutations, N512Y and F504L, are present on the ␤3-␤4 loop after the KTG motif (28). 4 Only the A311V and I312M mutations are within a conserved sequence motif; the rest generally surround the active site ( Fig. 1). An alignment of the relevant PBP2 sequences is shown in Fig. S1. Together, the mutations lower the secondorder rates of acylation of PBP2 by ESCs by up to 12,000-fold (29).
Recently, the molecular mechanism underpinning acylation of N. gonorrhoeae PBP2 by ESCs has begun to be understood (30). Acylation of a PBP2 construct comprising the TPase domain by ceftriaxone or cefixime is associated with structural changes within and near the active site (Fig. S2). One is twisting of the ␤3 strand, which contains the KTG motif, to form the oxyanion hole that stabilizes the tetrahedral intermediate/ transition state of acylation, and the other is movement of the ␤3-␤4 loop toward the antibiotic to form a cluster of interactions around the aminothiazole group of the R1 side chain of the cephalosporins. Rotation of the side chain of Thr-498 on ␤3 to form a hydrogen bond with the ␤-lactam carboxylate appears to be the trigger for twisting of the ␤3 strand, because in a crystal structure of PBP2 in complex with a phosphate occupying the same position as the ␤-lactam carboxylate, a similar hydrogen bond has formed, and ␤3 has also twisted.
This recent new understanding of how PBP2 is acylated by ESCs sets the stage to investigate how this process is impaired in the presence of mutations associated with ESC resistance. Elucidation of the molecular mechanisms of these mutations will both inform treatment and enable design of new antimicrobials that circumvent such mechanisms. Toward this goal, we present crystal structures of the TPase domain of PBP2 35/02 , derived from the ESC-decreased susceptibility strain 35/02 in apoform, and of PBP2 H041 , derived from the ESC R strain H041, in both apo-form and acylated by ceftriaxone. The structures strongly suggest that mutations present in PBP2 from resistant strains of N. gonorrhoeae, notably the G545S mutation on the ␤5-␣11 loop and F504L and N512Y mutations on the ␤3-␤4 loop, hinder acylation by restricting conformational flexibility of the protein.

ESC R mutations lower affinity of PBP2
For our studies of PBP2, we used a construct comprising only the TPase domain of the protein, referred to here as tPBP2 (where "t" denotes truncated). Such constructs retain identical acylation rates as the full-length protein for a range of antibiotics (31) and are therefore valid systems to examine the cephalosporin resistance of PBP2.
The kinetics by which PBPs react with ␤-lactam antibiotics or peptide substrate are as shown in Equation 1, Figure 1. Mutations in PBP2 identified as contributing to ESC resistance of N. gonorrhoeae mapped onto the crystal structure of tPBP2 WT acylated by ceftriaxone (30). Only the active-site region of PBP2 is shown, in which the protein is in ribbon form and colored gray, and sites of mutation are shown in yellow and labeled in red. Ceftriaxone (CRO) is shown with green bonds, and potential hydrogen bonds made by the antibiotic are indicated by dashed lines.

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
where formation of the noncovalent Michaelis-Menten complex, E⅐S, is described by the dissociation constant K s (i.e. k Ϫ1 / k ϩ1 ); k 2 is the rate of formation of the covalent acyl-enzyme complex, E-SЈ; and k 3 is the rate constant for deacylation that releases the product P (32,33). Mutations in PBP2 that contribute to ESC resistance could in principle alter kinetics by the following: (a) reducing affinity for the antibiotic (K s ); (b) lowering the rate of acylation (k 2 ); (c) increasing the rate of deacylation (k 3 ); or (d) by any combination of these. The covalent nature of the PBP reaction means that K s cannot be measured independently from k 2 ; therefore, the second-order rate of acylation (k 2 /K s ) is widely used as a measure of inhibition by ␤-lactam antibiotics (34 -37). Only by separating K s and k 2 can it be determined whether mutations impact affinity, the rate of acylation, or both. To address this, we mutated Ser-310 to alanine in tPBP2 WT , tPBP2 35/02 , and tPBP2 H041 and used isothermal titration calorimetry (ITC) to measure the dissociation constants (K d ) for ceftriaxone (traditionally, K s has been used to denote the binding affinity of ␤-lactam antibiotics, but it is identical to K d ). A crystal structure of an S310A mutant in the background of tPBP2 WT was solved at 1.7 Å resolution and shows that mutation of Ser-310 does not alter the architecture of the protein, including in the active site ( Fig. S3 and supplemental Table 1).
The three constructs exhibited dramatic differences in their affinities for ceftriaxone ( Fig. 2A and Table 1). tPBP2 WT has a K d of 2.3 Ϯ 0.6 M, whereas for tPBP2 35/02 and tPBP2 H041 , the binding isotherms were essentially flat, making it impossible to derive K d values. The curve for tPBP2 WT is unimodal, indicative of 1:1 stoichiometry. These data demonstrate that PBP2 mutations associated with ESC resistance have a significant impact on affinity for ceftriaxone. Using the K d value of 2.3 M determined from ITC, we estimate that ceftriaxone acylates tPBP2 WT at a rate of 3.9 s Ϫ1 , based on the previously measured second-order acylation constant of ceftriaxone for WT PBP2 of 1,710,000 M Ϫ1 s Ϫ1 (29). Without corresponding K d values, k 2 values could not be derived for tPBP2 35/02 or tPBP2 H041 .

ESC R mutations increase the rate of deacylation of Bocillin-FL-tPBP2 complex
We next investigated whether mutations present in PBP2 from H041 alter the rate of deacylation using the fluorescent penicillin, Bocillin-FL, as a reporter molecule (38). Acylated complexes of tPBP2 WT and tPBP2 H041 were formed by incubation with equimolar amounts of Bocillin-FL, and deacylation was measured after dilution into 300-fold molar excess of ceftriaxone by quantifying the decrease in fluorescence of enzyme-bound Bocillin-FL over A, isothermal titration calorimetry data for the interaction between ceftriaxone and S310A mutants of tPBP2. Upper panels show the binding isotherms for the wildtype (WT), 35/02, and H041 variants of tPBP2, and the lower panel shows the fit of the data for tPBP2 WT by nonlinear regression with a single-site isotherm using NanoAnalyze software (TA Instruments). Three technical replicates were performed under identical conditions using 100 M protein and 1000 M ceftriaxone. B, deacylation rates (k 3 ) for tPBP2 WT -and tPBP2 H041 -acylated complexes. WT and H041 tPBP2 constructs were incubated with Bocillin-FL for an hour, after which ceftriaxone was added to a concentration of 3 mM (t ϭ 0). Aliquots were removed at time intervals and separated by SDS-PAGE, and the amount of Bocillin-FL remaining bound was detected by UV illumination and quantified by densitometric scanning. The data were normalized for protein levels using Coomassie staining. Data were fit to first-order kinetics to derive the Bocillin-FL deacylation constant, k 3 . Experiments were performed in triplicate using protein samples from the same purification. For both experiments, the errors reported are the standard error of the mean.

Structures of tPBP2 35/02 and tPBP2 H041 are essentially identical to one another but differ from tPBP2 WT by the location of the ␤3-␤4 loop
The structures of tPBP2 35/02 and tPBP2 H041 were solved at resolutions of 1.93 and 1.55 Å, respectively, from crystals obtained in space group P2 1 2 1 2 1 with one molecule in the asymmetric unit (Table 2). For both structures, the majority of the protein is visible in the electron density map, including the ␤3-␤4 loop, and only residues Tyr-543 and Tyr-544 occupy relatively weak density. The two structures are highly similar to each other, with a root mean square deviation (RMSD) of 0.21 Å for 313 C␣ atoms, and the superimposition shows essentially no differences in the architecture of the proteins other than a very slight alteration in the ␣10 -␤3 loop (Fig. 3). Having established that the structures of tPBP2 35/02 and tPBP2 H041 are highly similar, we have mostly limited our comparisons below to tPBP2 H041 and tPBP2 WT for clarity.
The single molecule in the asymmetric unit of tPBP2 H041 superimposes onto the two molecules of tPBP2 WT (30) with RMSDs of 1.0 Å for molecule A (303 common C␣ atoms) and 1.5 Å for molecule B (310 common C␣ atoms). The higher RMSD for molecule B of tPBP2 WT reflects the different positions of the ␤3-␤4 loop, whereas this loop could not be modeled in molecule A. For this reason, molecule B of the tPBP2 WT structure was used for comparison with that of tPBP2 H041 .
Overall, the tPBP2 H041 structure overlaps closely with the tPBP2 WT structure, but there is a major difference in the ␤3-␤4 loop and minor changes in the ␤2b-␤2c hairpin (Fig. 4A). In particular, the ␤3-␤4 loop in tPBP2 H041 occupies a conformation in which the loop has bent away ("outbent") from the active site compared with its extended conformation in tPBP2 WT . The difference in the position of the ␤3-␤4 loop in tPBP2 H041 compared with that of tPBP2 WT is important because when tPBP2 WT is acylated by ceftriaxone or cefixime, this loop transitions from an extended conformation toward the active site to form a cluster of interactions around the R1 aminothiazole group (Fig. S2) (30). Its displaced conformation in tPBP2 H041 would appear to make the distance that must be traversed to reach the R1 group considerably larger.
Another difference between the structures of tPBP2 WT and tPBP2 H041 occurs in the active site. Although the positions of residues in the active site are to a large extent unchanged in tPBP2 H041 compared with the WT structure, the side-chain rotamer of Ser-362 (the S of the SXN motif) has altered such that its hydroxyl side chain is no longer hydrogen-bonded with either Ser-310 or Lys-313 of the SXXK motif and now only contacts Lys-497 of the KTG motif (Fig. 4B). Interestingly, this alteration is not observed in tPBP2 35/02 , and this may be related to a T483S mutation present in tPBP2 H041 but not tPBP2 35/02 (discussed below).

Most Ceph R mutations have minimal impact on apo structures
The structures of tPBP2 35/02 and tPBP2 H041 in apo-form allow an examination of the seven mutations implicated in ESC resistance in the pre-covalent states (Fig. 5A). Both PBP2 35/02 and PBP2 H041 have mutations located on helix ␣2: I312M and V316T for tPBP2 35/02 , and A311V, I312M, and V316P for tPBP2 H041 . This helix is an important part of the structure because the serine nucleophile, Ser-310, is located at its N terminus. Helix ␣2 is mostly buried within the protein and surrounded by hydrophobic residues. The mutations do not significantly alter the conformation of the helix in either tPBP2 H041 (Fig. 5B) or tPBP2 35/02 (Fig. S4). It is nevertheless interesting to note that the A311V and I312M mutations involve hydrophobic residues and increase the size of the side chain. These mutations would therefore be expected to increase the degree of hydrophobic packing around ␣2: the valine side chain at position 311 in tPBP2 H041 packs against Ile-519 of ␤3, and the methionine at position 312 packs against hydrophobic residues on ␣6, as well as Phe-315 on ␣2. The situation at position 316 is slightly different because this residue is relatively less engaged in hydrophobic packing. In tPBP2 35/02 , this is a threonine, which could form a weak hydrogen bond (3.4 Å) with the hydroxyl group of Tyr-383 from helix ␣6, and in tPBP2 H041 it is replaced by a proline.
The G545S mutation is present in both tPBP2 H041 and tPBP2 35/02 and is located on the ␤5-␣11 loop (Fig. 5C). This loop is immediately adjacent to the KTG motif on ␤3 and projects toward the active site. An obvious consequence of the G545S mutation is that the serine side chain forms a hydrogen bond with the side chain of Thr-498, thus forming a new connection to ␤3. This is expected to have a direct impact on the rate of acylation, as discussed later.
One mutation that is found in tPBP2 H041 but not tPBP2 35/02 is the T483S mutation. We have shown previously that T483S is

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
a major contributor to the increased resistance conferred by the penA allele from H041 compared with that conferred by the penA allele from 35/02 (29). Thr-483 is present on the ␣10 -␤3 loop, above the KTG motif as viewed in Fig. 5C. Two other mutations present in ESC-decreased susceptibility or ESC R variants of PBP2, P480A and T485I, also occur on this loop, but these do not appear to contribute to ESC resistance (29). The conformation of the loop is similar in both structures, and the impact of the T483S mutation is difficult to ascertain. The most obvious difference is that in tPBP2 H041 , Ser-483 is slightly shifted with respect to Thr-483 in tPBP2 WT , and it is interesting to note that the side-chain rotamer of Ser-362 in the SXN motif has altered to point toward the serine. This change may be elicited by the absence of the methyl group in Ser-483 (tPBP2 H041 ) compared with Thr-483 (tPBP2 WT ). Finally, the F504L and N512Y mutations, which are present in both tPBP2 35/02 and tPBP2 H041 , are located on the ␤3-␤4 loop (Fig. 5D). When PBP2 is acylated by ceftriaxone or cefixime, this loop moves toward the active site to form a cluster of interactions around the R1 side chain of the antibiotic (30). As noted above, this loop is bent away from the active site in tPBP2 H041 and tPBP2 35/02 , whereas it is largely extended in tPBP2 WT . As discussed below, these mutations may impede the conformational change in the ␤3-␤4 loop that occurs when ESCs bind and acylate PBP2.

Acylated structure of tPBP2 H041
The "outbent" conformation observed for the ␤3-␤4 loop in the structures of tPBP2 35/02 and tPBP2 H041 poses the question whether this loop can move as a result of acylation by ESCs to form the cluster of interactions around the R1 aminothiazole group as occurs in tPBP2 WT (Fig. S2) (30). To address this, the crystal structure of tPBP2 H041 acylated by ceftriaxone was determined at 1.8 Å resolution, obtained by soaking crystals of tPBP2 H041 with the antibiotic (Table 2). An unbiased ͉F o ͉ Ϫ ͉F c ͉ difference electron density map identifies ceftriaxone covalently bound to Ser-310 (Fig. 6). Electron density is observed for the majority of the ceftriaxone molecule with the exception of the aminothiazole ring and the C3 position on the dihydrothia-  Shown is an overlay of the two backbones where tPBP2 35/02 is orange, and tPBP2 H041 is green. The secondary structure nomenclature is the same as in Powell et al. (22).

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
zine ring. Weaker density of the aminothiazole is consistent with flexibility of this group, and only the more dominant occupancy is modeled. This contrasts with ceftriaxone binding to tPBP2 WT , where the R1 group is ordered (30). No density was observed for the C3 leaving group, although this group was unexpectedly observed in one of molecules of the asymmetric unit in the structure of tPBP2 WT acylated by ceftriaxone. The weakness in density at C3 is consistent with a mixture of C3 endo and exo forms of the dihydrothiazine ring. Ceftriaxone forms a number of hydrogen-bonding interactions with tPBP2 H041 (Fig. 6). These include bonds between the ester carbonyl oxygen and the main-chain amides of Ser-310 and Thr-500, which together form the oxyanion hole, and between the carbonyl oxygen preceding the aminothiazole ring (R1 group) and the side chain of Asn-364 from the SXN motif. In addition, a water molecule bridges the ␤-lactam carboxylate and amide of Gly-546. The ␤-lactam carboxylate forms hydrogen bonds with the main-chain amides and side-chain hydroxyls of both Thr-500 on ␤3 and Ser-545 from the ␤5-␣11 loop. As noted later, these latter interactions are different compared to when ceftriaxone binds to tPBP2 WT .

Acylation of tPBP2 H041 elicits only minimal changes to the protein
The structure of acylated tPBP2 H041 superimposes onto that of apo-tPBP2 H041 with an RMSD of 0.27 Å (319 C␣ atoms), indicative of the closely overlapping structures (Fig. 7A). This contrasts sharply with the significant changes observed when PBP2 WT is acylated by ceftriaxone (30). Hence, rather than moving toward the active site, the ␤3-␤4 loop remains in the same "outbent" conformation present in the structure of apo-tPBP2 H041 (Fig. 7B), and ␤3 does not twist, suggesting it may be more difficult to form the oxyanion hole in PBP2 from H041. Although the apo and acylated structures are generally highly similar, there are some small-scale differences in the active site. One is alteration of the Ser-362 side chain to adopt a position where it forms hydrogen bonds with Ser-310 and Lys-313, as well as Lys-497 (Fig. 7C). There are also slight differences in the positions of the side chains of lysines 497 and 313.

Acylated structures of tPBP2 WT and tPBP2 H041 are different
Given the similarity in the apo tPBP2 WT and tPBP2 H041 structures, and the large changes between apo and acylated tPBP2 WT structures, differences between the acylated forms of tPBP2 WT and tPBP2 H041 were expected. Indeed, the structures superimpose with an RMSD of 1.7 Å, which is considerably higher than apo versus acylated tPBP2 H041 (Fig. 8A). The most obvious difference is in the position of the ␤3-␤4 loop. As noted earlier, this loop moves when tPBP2 WT is acylated by ceftriaxone to form a cluster of interactions around the R1 group of the antibiotic, but in acylated PBP2 H041 , the loop remains in the same "outbent" conformation observed in the apo structure. In tPBP2 H041 , this loop contains two mutations (N512Y and F504L) that are located at the approximate righthand angle of the bend (Fig. 8B). The ␤-hairpin region (␤2b-␤2c) is also shifted between the two structures, although the significance of this difference is unclear. Finally, the ␤5-␣11 loop containing the G545S mutation is slightly closer to the active site in acylated tPBP2 WT compared with tPBP2 H041 .
Although the folds are generally similar, the positions occupied by ceftriaxone are remarkably different in the active sites of tPBP2 WT and tPBP2 H041 , particularly with respect to the positioning of the ␤-lactam carboxylate (Fig. 9A). This group points in different directions in the two structures, and the hydrogen-

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
bonding interactions it makes are also different. In tPBP2 WT , the ␤-lactam carboxylate contacts only Thr-498, but in tPBP2 H041 this bond is absent, and instead the carboxylate forms hydrogen bonds with Thr-500 and Ser-545. The G545S mutation may be a major determinant of the altered binding of ceftriaxone in tPBP2 H041 because it seems to act as a buttress to prevent direct binding between Thr-498 and the ceftriaxone carboxylate group (Fig. 9B), and the hydrogen bond made by

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
Ser-545 with the carboxylate may be responsible for the relative shift of ceftriaxone toward ␤3 and ␣11.

Discussion
The goal of this work was to understand how mutations in PBP2 contribute to ESC resistance of N. gonorrhoeae. We find that structures of PBP2 derived from the reduced susceptibility strain 35/02 and ESC-resistant strain H041 are nearly identical to each other, but differ from WT PBP2 (derived from the susceptible FA19 strain) in the ␤3-␤4 loop. We also find that the ␤3-␤4 loop occupies the same position when tPBP2 H041 is acylated by ceftriaxone, whereas this loop moves a considerable distance when tPBP2 WT is acylated by ESCs. We also show that ceftriaxone binds to a nonacylating S310A mutant of tPBP2 WT with micromolar affinity, but binding affinities for the S310A variants of tPBP2 H041 and tPBP2 35/02 are below detection. Finally, we observe that ceftriaxone occupies a markedly different position in the active site of tPBP2 H041 compared with tPBP2 WT .
The first question we addressed is what component of the catalytic reaction is impacted by mutations: the noncovalent binding affinity (K S ), the rate of acylation (k 2 ), or the rate of deacylation (k 3 )? We demonstrate that all three are altered in the tPBP2 H041 variant relative to tPBP2 WT . ITC measurements of the interaction of ceftriaxone with S310A mutants of tPBP2 variants showed that mutations in tPBP2 3502 and tPBP2 H041 have a marked effect on affinity, such that noncovalent binding of ceftriaxone was undetectable at the same concentrations used to demonstrate binding to tPBP2 WT . An unavoidable caveat to our experimental approach is that we used S310A mutants to measure binding (in order to prevent acylation, which confounds the analysis), but aside from the absence of the Ser-310 nucleophile, the mutation did not significantly alter the architecture of tPBP2 WT (Fig. S3).
One of the more surprising findings from our investigation is that mutations conferring resistance to ESCs in N. gonorrhoeae also decrease the half-life (t1 ⁄ 2 ) of the ceftriaxone-acylated complex by 28-fold compared with tPBP2 WT . Typically, t1 ⁄ 2 values for ␤-lactam-acylated complexes of high-molecular mass PBPs are on the order of hours-to-days (42)(43)(44) and are considered too low to have any impact on resistance. The t1 ⁄ 2 of 69 h for the acylated ceftriaxone complex of tPBP2 WT is on the same scale as these other PBPs, but the t1 ⁄ 2 of 2.4 h for the deacylation of tPBP2 H041 is markedly faster, raising the question of whether increased hydrolysis of the acylated complex could contribute to resistance. The doubling time for N. gonorrhoeae in culture of ϳ1 h would appear to make this decrease outside the range of physiological relevance, but it remains possible that within the genital tract or other host niches where N. gonorrhoeae colonizes, growth rates may be considerably slower. The molecular basis for the increased rate of deacylation is unclear, but it may result from the noncanonical binding mode for ceftriaxone in tPBP2 H041 compared with tPBP2 WT .
Of the seven mutations known to contribute directly to the ESC resistance of N. gonorrhoeae H041, a clear mechanism of action can be inferred for some, and less so for others. The three mutations occurring on helix ␣2 (A311V, I312M, and V316T/ V316P) fall into the latter category because they do not visibly alter the structure of ␣2. However, the most likely effect of these mutations is to increase rigidity, with the A311V and I312M mutations increasing hydrophobic packing around ␣2, and the V316T/V316P mutations introducing either a hydrogen bond (35/02) or a conformationally-restricted side chain (H041). It has been suggested previously that the N terminus of helix ␣2 undergoes conformational change to promote acylation in PBP2a of methicillin-resistant S. aureus (45). Although this helix does not shift in the crystal structures when either WT or mutated PBP2 is acylated by ESCs, it is tempting to speculate that flexibility in ␣2 undetected by crystallography promotes acylation, and this is reduced in the presence of the mutations in tPBP2 35/02 and tPBP2 H041 .
Another mutation whose mechanism is not immediately clear is T483S. This is an important mutation for conferring elevated ESC resistance in H041 compared with 35/02 (29). It lies on the ␣10 -␤3 loop, which moves only very slightly when comparing the apo and acylated structures of both tPBP2 WT and tPBP2 H041 . In the structure of tPBP2 WT acylated by ESCs, Thr-483 appears important for acylation because it forms a hydrogen bond with Thr-498, which in turn contacts the cephalosporin carboxylate. Even though Ser-483 in tPBP2 H041 also possesses a hydroxyl group, the hydrogen bond with Thr-498 is absent in the structure, and its loss may impede rotation of ␤3. In addition, the loss of the methyl group caused by the T483S mutation could alter packing interactions with the SXN motif, thus accounting for the altered side-chain rotamer for Ser-362 seen in the structure (see Fig. 4B). The resulting loss of polar interactions with the SXXK motif due to the change in rotamer may affect the pK a values of Ser-310 and Lys-313 in a way that impairs acylation by ESCs.

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
By contrast, the mechanisms by which the F504L and N512Y mutations in tPBP2 35/02 and tPBP2 H041 reduce the rate of acylation are more evident. Whereas the F504L and N512Y mutations initially appear too distant from the active site to have an obvious effect on reactivity with ESCs, their roles can be understood in the context of the movement of the ␤3-␤4 loop toward the active site observed in the acylated structures of PBP2 WT (30). The N512Y mutation resides on ␤4 at the precise location where the ␤3-␤4 loop in PBP2 H041 bends away from the active site, thereby acting as a hinge (Fig. 8B). Likewise, the F504L mutation, which is found in tPBP2 35/02 , tPBP2 H041 , and the penicillin-resistant strain FA6140 (22,46), occupies the equivalent "hinge" position on the ␤3 side of the loop. Together, both mutations appear to impede bending of ␤3-␤4, thus preventing formation of the cluster of residues around the R1 group of the antibiotic with subsequent impact on the reaction with ESCs. In support of this, the ␤3-␤4 loop occupies the same "outbent" conformation in both apo and acylated structures tPBP2 H041

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
and therefore appears incapable of movement toward the active site.
A potential caveat here is the possibility that the conformation of ␤3-␤4 loop is the product of crystal packing interactions, especially because the structures of tPBP2 35/02 and tPBP2 H041 were solved in a different crystal system compared with tPBP2 WT . As shown in Fig. S5, there are some symmetry contacts involving the ␤3-␤4 loop. Although it is possible these contacts have trapped the ␤3-␤4 into the characteristic "outbent" conformation observed, they could equally result from the loop adopting this conformation in solution prior to crystallization. In this regard, it is important to note that the space between the ␤3-␤4 loop and where ceftriaxone binds in the active site is unoccupied by symmetry-related molecules, suggesting there is sufficient space for the loop to adopt a conformation closer to the active site in the crystal when ceftriaxone binds.
Among the mutations that decrease the acylation rate by ESCs, the impact of the G545S mutation present on the ␤5-␣11 loop is perhaps the easiest to understand. New hydrogen bonds introduced by the replacement of glycine with serine have two potential outcomes that may affect both binding and the rate of acylation by ceftriaxone. One is a "locking" mechanism whereby the hydrogen bond formed between Ser-545 and Thr-498 prevents side-chain rotation of Thr-498 and subsequent rotation of ␤3, thus hindering formation of the oxyanion hole required for acylation. In this context, Ser-545 also acts as a wedge to prevent contact between Thr-498 and the ␤-lactam carboxylate. The other outcome may result from the hydrogen bond between Ser-545 and the ␤-lactam carboxylate. With Thr-498 locked into position by Ser-545, the hydrogen bonding requirement of the ␤-lactam carboxylate is now satisfied by a direct hydrogen bond with Ser-545, and this may be responsible for ceftriaxone occupying a position that is relatively closer to the ␤5-␣11 loop in tPBP2 H041 compared with the structure of acylated tPBP2 WT . This could be regarded as an induced-fit mechanism where ceftriaxone moves after initial binding, but to a conformation that is considerably less favorable for acyla-

Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae
tion. This binding mode may also explain why tPBP2 H041 exhibits a higher rate of deacylation of ceftriaxone compared with tPBP2 WT .
Mutations in several PBPs have been implicated in resistance to ␤-lactams, and it is important to consider whether mechanisms are shared or distinct across species. The seven mutations in PBP2 known to contribute to ESC resistance occur in three loop regions (␣10 -␤3, ␤3-␤4, and ␤5-␣11) and helix ␣2, and in general, mutations are also observed in these regions in PBPs from other bacterial species (47). A known hot spot is the ␤3-␤4 loop because mutations in this loop associated with ␤-lactam resistance are observed in a number of PBPs, including Streptococcus pneumoniae PBP2x, PBP1a, and PBP2b and Haemophilus influenzae PBP3 (44,(47)(48)(49). How such mutations work has been unclear, but given the movement of the ␤3-␤4 loop observed in N. gonorrhoeae PBP2 acylated by ESCs (30), it is possible they function in a similar way as the F504L and N512Y mutations in PBP2. Similarly, mutations in the ␤5-␣11 loop are observed in Pseudomonas aeruginosa PBP3 (47), and of these, G352D/G352E mutations could affect interactions with the KTG motif. Finally, helix ␣2 harbors mutations in several PBPs, including those close to the serine nucleophile similar to A311V in N. gonorrhoeae PBP2 (48,(50)(51)(52)(53), although again, their contribution to resistance remains unclear.
The overall picture emerging from our studies of PBP2 is that mutations implicated in resistance both lower affinity and restrict motions important for acylation by ESCs. By locking the side-chain rotamer of Thr-498, the G545S mutation restricts rotation of the ␤3-strand needed to form the oxyanion hole, and by forming a new hydrogen bond with the ␤-lactam carboxylate, it prevents optimal binding of ceftriaxone. Through loss of the hydrogen bond between Thr-483 and Thr-498, the T483S mutation may also restrict rotation of ␤3. In addition, the L504F and N512Y mutations appear to prevent movement of the ␤3-␤4 loop that occurs during acylation. Finally, the increased hydrophobicity conferred by mutations in ␣2 may hinder flexibility in the vicinity of the Ser-310 nucleophile.
Taken together, our data suggest a mechanism of dynamic restriction, whereby mutations hinder flexibility in the protein and therefore create an energetic barrier against binding and acylation by cephalosporins. Whereas mechanisms of antibiotic resistance are often attributed to steric exclusion, especially where mutations sterically impede binding of antibiotic outside the so-called substrate envelope (54,55), those involving differences in protein dynamics have also been reported in a variety of systems, including HIV1 protease, reverse transcriptase and integrase, and eukaryotic protein kinases (56 -59). For example, by introducing a new requirement for conformational changes in the protein to accommodate the drug, mutations in HIV-1 integrase associated with resistance to raltegravir (58) may act in a similar way as in PBP2.
Because ␤-lactam antibiotics are analogs of the D-Ala-D-Ala terminus of the peptidoglycan substrate, mutations that lower binding or acylation for ESCs would also be expected to affect the binding and cross-linking of peptidoglycan substrate.
Hence, a key question remaining is how the essential TPase function of PBP2 is preserved in the presence of mutations that lower the rate of inactivation by ESCs and thus maintain strain fitness. In the absence of structural information for the interaction of PBP2 with its peptidoglycan substrate, how discrimination is successfully achieved is unclear, but we believe the answer lies in the larger size of the peptidoglycan substrate compared with ␤-lactams. Compared with ESCs, PBP2 would be expected to make many more contacts with peptidoglycan, not only with the peptide C terminus but potentially also with the N-terminal region of the peptide and parts of the glycan strand. We hypothesize that the larger number of contacts with the substrate helps overcome the energetic barrier created by the ESC resistance mutations and allows TPase activity to proceed. That said, while it is not possible to measure TPase activity directly without a suitable biochemical assay, it is clear that the activity of both tPBP2 35/02 and tPBP2 H041 is lower than WT, which manifests in a slower growth rate for strains harboring the mosaic penA alleles (60,61). These data suggest that the energetic barrier is not entirely overcome, but nevertheless sufficient activity remains to support cell growth. Testing this hypothesis must await a detailed understanding of PBP2-peptidoglycan interactions and examination of the dynamic behavior of PBP2 from ESC-resistant strains of N. gonorrhoeae.

Cloning and expression of transpeptidase domain constructs
We previously reported the cloning of a TPase domain construct for WT PBP2 derived from FA19 (referred to here as tPBP2 WT ) (31) that retained identical acylation rates as the fulllength protein for a range of antibiotics. For this study, we made equivalent constructs for PBP2 from 35/02 and H041. Briefly, nucleotides 709 -1746 of penA35 from 35/02 (tPBP2 35/02 ) or penA41 from H041 (tPBP2 H041 ), both encoding amino acids 237-582, were cloned into pMALC2KV, which expresses tPBP2 as an N-terminal hexa-histidine-tagged fusion with maltose-binding protein, separated by a tobacco etch virus (TEV) protease site. Nucleotides encoding amino acids 283-297 were not included in the construct to promote the chances of crystallization, as these form a loop that projects from the TPase domain (22,31). A non-native glycine was included to connect residues Pro-282 and Arg-298.
The constructs were transformed into Escherichia coli BL21 (DE3) cells for expression. Two liters of culture were grown at 37°C, and protein expression was induced by the addition of 0.3 mM isopropyl ␤-D-thiogalactoside, followed by overnight growth at 20°C. The harvested cells were lysed in 20 mM Tris⅐HCl, pH 8.0, 500 mM NaCl, and 10% glycerol (TNG), and the solution was passed over a 5-ml HiTrap TM FF Ni 2ϩ -nitrilotriacetic acid affinity column (GE Healthcare) equilibrated in TNG buffer. After elution of the protein by a TNG/15-500 mM imidazole gradient, His 6 -tagged TEV protease was used to digest the fusion protein at a molar ratio of 50:1 (fusion protein/ TEV), followed by overnight dialysis at 4°C against buffer containing 20 mM Tris, pH 7.0, 10% glycerol (TG). The digested protein was then passed over a 5-ml HiTrap TM SP FF ion-ex-Crystal structures of PBP2 from ESC-resistant N. gonorrhoeae change column (GE Healthcare) pre-equilibrated with TG buffer, and the protein was eluted by a 0 -500 mM NaCl linear gradient in the same buffer, and fractions were pooled.

Site-directed mutagenesis
The S310A mutants for tPBP2 WT , tPBP2 35/02 , and tPBP2 H041 were generated using the QuikChange Lightning kit (Agilent, Santa Clara, CA). The DpnI-treated amplified PCR product was used to transform E. coli BL21 (DE3) cells. All mutants were confirmed by sequencing.

ITC
Measurements of heat exchange to quantify the interaction between ceftriaxone and the S310A mutants of tPBP2 WT , tPBP2 35/02 , and tPBP2 H041 were conducted on a Nano ITC instrument (TA instruments, New Castle, DE). Purified proteins were dialyzed against 50 mM sodium phosphate, 180 mM NaCl, 10% glycerol, pH 7.1, and concentrated to 100 M. Measurements were carried out at 25°C, with a 300-rpm stirring rate, and a total of 18 injections of 2.7 l of ceftriaxone (1 mM stock) was added to the protein solution (initial volume 300 l). Three technical replicates were performed, using a fresh sample of protein derived from the same purification. The binding parameters ⌬H, K d , and n (stoichiometry) were calculated using the Nano analyzer software using nonlinear regression for a single-site isotherm (62). Heats of reaction between the ligand and buffer only were subtracted from the measured values.

Derivation of k 2
Using K d values obtained from ITC measurements and previously measured values of the second-order rate constant (k 2 / K S ) of ceftriaxone for full-length WT PBP2 (29), estimated values for the rate of acylation (k 2 ) of tPBP2 WT for ceftriaxone were calculated using the formula k 2 ϭ K S (k 2 /K S ), where K S is equivalent to K d (32,33).

Measurement of deacylation rate
The deacylation rate constant (k 3 ) values were determined using Bocillin-FL as a reporter molecule. Purified tPBP2 protein at a concentration of 20 M was incubated with 20 M Bocillin-FL for 1 h at 22°C in 50 mM phosphate, 10% glycerol, 180 mM NaCl, pH 7.1, at which time the final concentration of protein and Bocillin-FL was adjusted to 10 M, and ceftriaxone was added at a final concentration of 3 mM. The amount of Bocillin-FL covalently bound to PBP2 was determined by removing aliquots at various times, adding to SDS sample buffer, and boiling at 95°C for 3 min. After separation by SDS-PAGE and scanning using an electrophoresis analysis system (ESAS 290, Kodak Inc., Rochester, NY) , the relative amounts of Bocillin-FL-labeled tPBP2 remaining over time were quantified by densitometric analysis using ImageJ software (National Institutes of Health). The deacylation constants (k 3 ) were calculated by plotting the ln of % Bocillin-FL bound (PBP t /PBP o ) versus time, with the slope of the line equal to Ϫk 3 , using Equation 2, where PBP t is the Bocillin-FL-PBP concentration at time t, and PBP o is the initial concentration of the Bocillin-FL-PBP complex. The half-life (t1 ⁄ 2 ) of the acylated complex was then calculated using the formula t1 ⁄ 2 ϭ ln2/k 3 .

Crystallization
tPBP2 variants were concentrated to 13 mg/ml, and crystallization trials were performed using a Gryphon liquid-dispensing system (Art Robbins, Sunnyvale, CA) in a 96-well sittingdrop format in which 200 nl of protein solution was mixed with 200 nl of well solution. The proteins were screened against a number of sparse matrix screens, including JCSG Cores I-IV and Suiteϩ (Qiagen, Germantown, MD), with incubation at 18°C. A number of hits were obtained for tPBP2 35/02 and tPBP2 H041 in solutions containing PEG 600. After optimization, the best crystals were obtained at 18°C over wells containing 37-40% PEG 600, buffered with 0.1 M CHES at pH 9.1-9.3 for tPBP2 H041 and pH 9.6 for tPBP2 35/02 . Crystals appeared after 3-4 days and exhibited a plate-like morphology.

X-ray data collection and model refinement
Crystals of tPBP2 35/02 and tPBP2 H041 were flash-frozen without adding cryo-protectant, and diffraction data were collected at a wavelength of 1.00 Å on an Eiger 16M detector at the SER-CAT 22-ID beamline at the Advanced Photon Source in Argonne, IL (Table 2). For crystals of tPBP2 H041 , 360°of data in 0.25°oscillations were collected, with an exposure time of 0.2 s/frame and a crystal-to-plate distance of 240 mm. For crystals of tPBP2 35/02 , 360°of data in 0.2°oscillations were collected, with an exposure time of 0.25 s/frame and a crystal-to-plate distance of 200 mm. Both datasets were processed and scaled with HKL2000 (63). Structures were solved by molecular replacement using PHASER (64). To generate the tPBP2 35/02 structure, the tPBP2 WT structure (30) was used as a search model, whereas the structure of tPBP2 H041 was solved by molecular replacement using an earlier structure of tPBP2 H041 determined at low pH. 4 All models were refined with iterative rounds of model building and automated refinement, using the graphics programs O (65) or COOT (66) and REFMAC (67). The stereochemistry of models was analyzed with PROCHECK (68).
To generate an acylated complex of tPBP2 H041 , crystals were soaked in a super-saturated solution of ceftriaxone (Sigma) for 2-3 h at room temperature, followed by flash-freezing without additional cryo-protectant. Diffraction data were collected at a wavelength of 1.00 Å on a MX300-HS detector at the SER-CAT 22-BM beamline. 360°of data were collected in 1°oscillations with an exposure time of 6 s/frame and a crystal-to-plate distance of 240 mm and were processed using HKL2000. The structure was solved by refinement of the tPBP2 H041 structure. Ceftriaxone was modeled using the ͉F o ͉ Ϫ ͉F c ͉ difference electron density map, followed by iterative cycles of model building and refinement.

Data availability
Coordinates and structure factors have been deposited with the Protein Data Bank with PDB codes as follows: 6VBL,