VNRX-5133 (Taniborbactam), a broad-spectrum inhibitor of serine- and metallo-β-lactamases, restores activity of cefepime in Enterobacterales and Pseudomonas aeruginosa.

As shifts in the epidemiology of β-lactamase-mediated resistance continue, carbapenem-resistant Enterobacterales (CRE) and carbapenem-resistant Pseudomonas aeruginosa (CRPA) are the most urgent threats.

( Fig. 1) relative to those of avibactam and vaborbactam, including the rate of covalent bond formation (k 2 /K i ), off rates (k off ), and the half-lives of active-site occupancy (t 1/2 ) with CTX-M-15 (class A), KPC-2 (class A), and P99 AmpC (class C) are presented in Table  1. The inhibition behavior of all three BLIs with SBLs fits a two-step inhibition model, in which a noncovalent complex forms, followed by the formation of a reversible covalent bond with the active-site serine residue (equation 1).
where E is enzyme, I is inhibitor, and k 1 is the rate of association of the noncovalent Michaelis-Menten complex, k 2 is the rate of formation of the covalent bond, k Ϫ2 is the off rate, and EI* is the covalent enzyme-inhibitor complex. The second-order rate constants (k 2 /K i , where K i is the inhibitor constant) of covalent bond formation to the active-site serine of the three ␤-lactamases examined (CTX-M-15, P99 AmpC, and KPC-2) were on the order of 10 4 to 10 5 M Ϫ1 s Ϫ1 for taniborbactam, whereas they were 10 3 to 10 5 for avibactam and 10 3 for vaborbactam (Table 1). Due to slow inhibitor off rates (k off value range, 1.1 ϫ 10 Ϫ4 to 3.8 ϫ 10 Ϫ4 s Ϫ1 ), taniborbactam exhibited a significant residence time within the active site, with the t 1/2 values ranging from 30 to 105 min, whereas the t 1/2 values ranged from 29 to 249 min for avibactam, consistent with published data (33), and from 5 to 32 min for vaborbactam (Table 1). A 1.1-Å resolution X-ray cocrystal structure of taniborbactam and CTX-M-15 (PDB accession number 6SP6) (32) previously confirmed that the inhibitor binds covalently to the catalytic Ser70 with a bond distance of 1.53 Å and that the boron atom adopts a tetrahedral conformation (Fig. 2). The location of the boron hydroxyl group in the oxyanion hole suggests that the inhibitor acts as a mimetic of the tetrahedral intermediate formed during the acylation step. Additionally, the binding of taniborbactam displaces the deacylation water molecule (Wd) by 1.4 Å (Fig. 2C) and exploits substratelike interactions with conserved active-site residues in serine-␤-lactamases (Asn104, Ser130, Asn132, Asn170, and Thr235) (32).
Taniborbactam is distinguished from avibactam and vaborbactam by the ability to inhibit the most clinically relevant subclass B1 MBLs (VIM-and NDM-type enzymes). As observed in the cocrystal structure of taniborbactam and VIM-2 (PDB accession number 6SP7) (Fig. 3) (32), the boron atom adopts an sp 3 hybridization state due to geminal diol formation after reacting with the active-site hydroxide anion (the so-called bridging water coordinated to the Zn 2ϩ active-site cations). The boron hydroxyl interacts with Zn-1 and both the conserved Asn233 and Asp120 residues. The carboxylate and oxygen atom of the cyclic oxaborinane interacts with Zn-2, thus behaving as a mimetic of the tetrahedral intermediate in subclass B1 enzymes. The substituted amino group of the inhibitor side chain interacts with Glu149, which is conserved in NDM-1. Interestingly, the inhibitor carboxylate does not interact with Arg228 but instead interacts with the backbone of conserved Asn233. By comparing the bound structure with that for apoVIM-2 (PDB accession number 1KO3) (34), the binding of taniborbactam induces a narrowing of the active-site cleft due to the approach of conserved Asn233 and Phe61 (Fig. 3C). Finally, a surface rendering of VIM-2 bound by taniborbactam shows the presence of an electronegative pocket stabilizing the inhibitor side chain (Fig. 3D), also providing a structural basis for inhibition of both VIM-and NDM-type MBLs, as residues constituting the pocket (Glu149 and Asp236) are conserved in both enzyme subgroups. The inhibition of VIM-2 and NDM-1 by taniborbactam was reversible, as enzymatic activity was fully recovered after a rapid jump dilution. Using steady-state kinetic analysis, taniborbactam was confirmed to be a competitive inhibitor of VIM-2 and NDM-1 with K i values of 0.019 and 0.081 M respectively, whereas inhibition of IMP-1 was outside of the spectrum of inhibition, with K i being greater than 30 M (Table 2). Regarding the serine ␤-lactamases, taniborbactam had potent inhibitory activity against class A and C enzymes, with K i values ranging from 0.002 to 0.017 M for SHV-5, KPC-2, CTX-M-15, and P99 AmpC, similar to the values for avibactam. Against the class D OXA-48 enzyme, taniborbactam had a K i of 0.35 M, similar to the values for avibactam and vaborbactam ( Table 2), and that level of potency was sufficient to protect cefepime from this ␤-lactamase subtype, which exhibits weak cefepimase activity and, consequently, which contributes minimally to cefepime resistance (Table  3) (35,36).
Potency and spectrum of activity defined in engineered Escherichia coli strains overproducing individual ␤-lactamases. The spectrum of antibacterial activity of cefepime-taniborbactam and the breadth of inhibitory activity of this next-generation cyclic boronate BLI were assessed in 34 engineered E. coli strains each overproducing an individual ␤-lactamase and directly compared to those of the clinically approved agent ceftazidime-avibactam. In this manner, the antibacterial activity of the BL-BLI combinations relative to that of the partnered cephalosporins alone provided a quantitative measure of ␤-lactam potentiation that directly reflects the levels of inhibition of ␤-lactamase function achieved by addition of BLI. The BLI concentration was fixed at 4 g/ml, as described in Materials and Methods.
In strains overproducing Ambler class A ␤-lactamases, addition of taniborbactam potentiated cefepime activity from 8-to 1,024-fold, which is comparable to the level of potentiation of ceftazidime activity by avibactam (Table 3). In 16/17 class A ␤-lactamase-overproducing strains, cefepime activity was potentiated by taniborbactam to within 4-fold of the activity of cefepime-taniborbactam against the vectorcontrol strain (cefepime-taniborbactam MIC, 0.12 g/ml), whereas the potentiation of  (34) and its covalent complex with taniborbactam (green; PDB accession number 6SP7) (32). (B) Active-site close-up and omit map of the taniborbactam-bound VIM-2 complex. (C) Mode of binding of taniborbactam in the active site of the VIM-2 metallo-␤-lactamase, showing the main interactions between the enzyme and VNRX-5133 (magenta). Residues were numbered according to the BBL consensus numbering scheme (57). (D) Surface rendering of the VIM-2 active site in the VNRX-5133inhibited complex, showing an inhibitor-induced narrowing of the active-site cleft resulting from a closer contact between the side chains Phe61 and Asn233. The presence of the electronegative pocket interacting with the inhibitor side chain and contributing to the stability of the inhibitor-enzyme complex is also shown. ceftazidime activity by avibactam was found in only 4/17 strains (ceftazidime MIC for the vector-control strain, 0.25 g/ml). Cefepime-taniborbactam provided potent coverage against clinically derived serine ␤-lactamases associated with elevated MICs of ceftazidime-avibactam, including strains producing KPC-3 ⍀ loop variants (D179Y, V240G, A177E/D179Y, and D179Y/T243M) (25,26), TEM-24, VEB-9, and PER-1 and -2. This side-byside comparison demonstrated the differentiation of cefepime-taniborbactam from ceftazidime-avibactam through improved class A ESBL and KPC variant coverage by cefepime-taniborbactam.
In strains producing Ambler class B MBLs, the addition of taniborbactam potentiated the antibacterial activity of cefepime by 16-to 256-fold in all class B ␤-lactamaseproducing strains, with the notable exception of strains producing IMP-1, against which taniborbactam had insufficient inhibitory activity (K i Ն 30 M; Table 2) to potentiate the antibacterial activity of cefepime (Table 3). The addition of avibactam failed to restore ceftazidime activity in these strains, as would be expected from its lack of inhibitory activity against MBLs. In contrast, the activity of cefepime against strains overproducing clinically important NDM variants (NDM-1, -5, and -7) was highly potentiated (64-to 128-fold) by the addition of taniborbactam. This analysis therefore established the spectrum of MBL enzymes inhibited by taniborbactam to be SPM-1, GIM-1, and clinically important variants of NDM (NDM-1, -5, and -7) and VIM (VIM-1, -2, and -4).
Among strains producing selected class C enzymes, cefepime-taniborbactam MICs ranged from 0.12 to 0.5 g/ml, reflecting an 8-to 32-fold potentiation of cefepime activity to within 4-fold of the MIC for the vector-control strain. In contrast, ceftazidimeavibactam MICs were 0.5, 2, and 8 g/ml for the ACT-17-, ACT-C189-, and CMY-2overproducing strains, respectively. Although the MIC of 0.5 g/ml was within 2-fold of that for the vector-control strain, the last two values were 8-and 32-fold higher than the MIC for the vector-control strain for ACT-C189 (P99 AmpC) and CMY-2, respectively, despite avibactam potentiating ceftazidime activity by 64-to 128-fold in these strains.
In strains overproducing class D ␤-lactamases (OXA-48, OXA-162, OXA-163, OXA-181, and OXA-232), cefepime activity was potentiated 8-to 512-fold by addition of taniborbactam, resulting in MICs of 0.12 to 0.25 g/ml, which were within 2-fold of the MIC for the vector-control strain. By comparison, addition of avibactam to ceftazidime had a 2-fold potentiation of ceftazidime activity against four of five strains, resulting in MICs of 0.5 to 1 g/ml, which is in a range similar to the MICs of cefepimetaniborbactam. In the remaining class D ␤-lactamase-producing strain, OXA-163 exhibited high ceftazidimase activity (ceftazidime MIC, 256 g/ml), and although ceftazidime activity was potentiated 64-fold by addition of avibactam to an MIC of 4 g/ml (16-fold higher than the MIC for the vector-control strain), cefepime activity was potentiated 512-fold by the addition of taniborbactam, providing a significantly improved MIC of 0.25 g/ml, within 2-fold of the MIC for the vector-control strain. From this side-by-side comparison with class D ␤-lactamase overproducers, cefepime-taniborbactam and ceftazidime-avibactam appear to be equivalently active, though cefepime-taniborbactam exhibited better potency than ceftazidime-avibactam against OXA-163. Overall, within this panel of 34 distinct ␤-lactamase-overproducing strains, the potentiation of the cefepime MIC by taniborbactam ranged from 8-to 1,024-fold in 33/34 strains, with a modal potentiation of 512-fold and MIC 50 /MIC 90 values of 0.25/ 4 g/ml, respectively. Thirty-three of 34 (97%) strains had cefepime-taniborbactam MICs of Յ4 g/ml. The exception was the IMP-1 overproducer (MIC of cefepime alone, 64 g/ml, with no potentiation by taniborbactam). In contrast, the potentiation of ceftazidime MICs by avibactam ranged from 2-to 1,024-fold in 25/34 strains, with a modal potentiation of 64-fold and MIC 50 /MIC 90 values of 8/1,024 g/ml. Nine of 9 (100%) MBL-producing strains and 8 of 25 (32%) SBL-producing strains fell outside of the ceftazidime-avibactam inhibitory spectrum, with MICs of Ն16 g/ml, providing a measure of differentiation for cefepime-taniborbactam against this challenge set of SBL-overproducing strains encompassing Ambler class A, C, and D ␤-lactamases. Antibacterial activity of cefepime-taniborbactam in reference isolates. The antibacterial activity of cefepime, cefepime-taniborbactam, and taniborbactam alone against seven publicly available reference type isolates from the CDC, NCTC, and ATCC was assessed (Table 4). This panel encompassed 3 E. coli, 2 K. pneumoniae, and 2 P. aeruginosa isolates and included four quality control (QC) isolates for broth microdilution (37). Of particular note, taniborbactam demonstrated no clinically relevant antibacterial activity (MIC Ն 512 g/ml) against the isolates within this panel. Among the ␤-lactamase-producing Enterobacterales isolates, E. coli NCTC 13353 producing CTX-M-15 and K. pneumoniae BAA-1705 producing KPC-2 are routine QC isolates used for the testing of cefepime-taniborbactam (37). Cefepime antibacterial activity was selectively potentiated by addition of 4 g/ml taniborbactam against isolates producing ␤-lactamases, with the exception of P. aeruginosa ATCC 27853 (Pseudomonas-derived cephalosporinase 5 [PDC-5]), against which cefepime alone was active ( Table 4).
Potentiation of cefepime activity by taniborbactam in clinical isolates of Enterobacterales and P. aeruginosa. The antibacterial activity of cefepime-taniborbactam was compared to that of ceftazidime-avibactam, ceftolozane-tazobactam, and cefepimetazobactam against a diverse panel of Enterobacterales and P. aeruginosa clinical isolates with defined ␤-lactamase subtypes (see Table S1 in the supplemental material). IMP-producing strains were excluded, as none of the BLI-protected cephalosporins inhibited IMP sufficiently to provide a clinically relevant rescue of the partner ␤-lactam, a characteristic also shared by approved BLI-protected carbapenems (meropenemvaborbactam and imipenem-relebactam). The panel does not reflect current epidemiological trends but instead highlights the differences in coverage provided by cefepime-taniborbactam from that provided by the comparators. The antibacterial activity of the cephalosporins alone was included to ascertain the level of potentiation by the partnered BLI. BLIs were tested at a fixed concentration of 4 g/ml in all combinations except cefepime-tazobactam, where tazobactam was fixed at a concentration of 8 g/ml (38). The collection comprised 143 clinical isolates collected from 2005 and 2018. Isolates were subdivided by phenotypic profile, with some level of molecular characterization into Enterobacterales isolates producing (i) mixed class A and class C ␤-lactamases and ESBLs (subdivision 1) or (ii) serine-and metallocarbapenemases (OXA-48/48-like, KPC, and NDM/VIM-type MBLs) (subdivision 2) and P. aeruginosa isolates producing (iii) basal levels of PDCs and downregulated OprD combined with upregulated RND drug efflux systems (subdivision 3), (iv) ceftolozane-tazobactamresistant PDC variants (subdivision 4), or (v) serine and metallocarbapenemases (GES, KPC, VIM) (subdivision 5) (Table S1).
The overall activity of the tested agents against 102 Enterobacterales isolates is summarized in Fig. 4C and Table S1. The distributions of the MICs of cefepime alone along with those of the four cephalosporin-BLI combinations are presented. Addition of taniborbactam reduced the MIC 90 from Ն256 g/ml for cefepime alone to 1 g/ml (Table S1). Cefepime-taniborbactam was the most active combination, followed by ceftazidime-avibactam, with MIC 50 /MIC 90 values of 0.12/1 and 1/Ն64 g/ml, respectively ( Fig. 4C; Table S1).
Among 10 isolates of P. aeruginosa (subdivision 4) resistant to ceftolozanetazobactam (MIC, Ͼ8 g/ml) through the production of eight different PDC variants (Table S1), a single isolate remained susceptible to ceftazidime-avibactam with an MIC of 8 g/ml, while all others had MIC values ranging from 16 to Ն64 g/ml. Cefepime was more stable against these PDC variants than ceftazidime was, and the combination of cefepime-taniborbactam provided good antibacterial activity with MICs ranging from 2 to 16 g/ml (Table S1).
Among 17 isolates of P. aeruginosa (subdivision 5) producing KPC (n ϭ 3), GES (n ϭ 9), or VIM (n ϭ 5) carbapenemases, cefepime-taniborbactam was more active than the comparator cephalosporin-BLI combinations. The MIC values of cefepimetaniborbactam for all 17 isolates ranged from 1 to 8 g/ml, with an MIC 50 of 4 g/ml and an MIC 90 of 8 g/ml (Table S1). Ceftolozane-tazobactam activity was weak, with MIC values ranging from 16 to Ն64 g/ml. Ceftazidime-avibactam was active with an MIC of Յ8 g/ml for 5/17 isolates, including all 3 KPC producers and 2 GES-6 producers, whereas all other isolates exhibited MICs ranging from 16 to Ն64 g/ml. The data highlight a gap in coverage of the ceftazidime-avibactam combination against GESproducing P. aeruginosa isolates, with 6/9 isolates having MICs of Ն16 g/ml (Table S1).
Overall, among the 41 P. aeruginosa isolates tested (subdivisions 3 to 5), addition of taniborbactam to cefepime dramatically shifted the distribution of MICs to a lower range, with an MIC 90 of 8 g/ml for cefepime-taniborbactam compared to an MIC 90 of Ն256 g/ml for cefepime alone (Fig. 5B; Table S1). Ceftazidime-avibactam and ceftolozane-tazobactam had elevated MIC 90 values of Ն64 g/ml due to the production of VIM and GES variants along with cross-resistance to ceftazidime-avibactam among the ceftolozane-tazobactam-resistant isolates, consistent with previously published findings (27,28).
Addition of taniborbactam restores the bactericidal activity of cefepime in NDM-1-producing K. pneumoniae and VIM-2-producing P. aeruginosa clinical isolates. Restoration of the bactericidal activity to cefepime by addition of taniborbactam was further confirmed by time-kill assays with K. pneumoniae CDC-0049 producing NDM-1 and P. aeruginosa Ps-12 producing VIM-2 (Fig. 6). Both isolates were resistant to cefepime, with MICs of 256 g/ml and 32 g/ml, respectively. Addition of taniborbactam at a fixed concentration of 4 g/ml shifted the cefepime MIC to 4 g/ml in both cases. In both isolates, cefepime-taniborbactam achieved a 3-log 10 reduction in the number of CFU per milliliter relative to the starting inoculum by 6 h, without regrowth thereafter. In contrast, ceftazidime-avibactam tested at 32 g/ml (4-fold above the CLSI susceptibility breakpoint of 8 g/ml) had little to no impact on either isolate due to the high MICs originating from a lack of coverage of VIM-and NDM-type MBLs.
Resistance to cefepime-taniborbactam. Frequency-of-resistance (FoR) studies were performed at 4ϫ MIC of cefepime with taniborbactam fixed at 4 g/ml against 8 Enterobacterales and P. aeruginosa strains expressing Ambler class A, B, C, and D ␤-lactamases. The FoR ranged from 1.6 ϫ 10 Ϫ9 to 8.9 ϫ 10 Ϫ11 (Table 5), indicating a low potential for the spontaneous development of resistance to cefepimetaniborbactam among the target pathogens. Moreover, no single-step taniborbactamresistant ␤-lactamase variants were obtained from this standard first-pass investigation.
High selectivity for ␤-lactamases. The selectivity and specificity of taniborbactam were evaluated in the DrugMatrixScreen panel of pharmacological targets at Eurofins/ Panlabs. Binding, enzymatic, and uptake assays representing a wide range of cellular and subcellular target classes were performed. At the screening concentration of 100 M, no notable off-target findings (Յ23% inhibition) were reported in 128 of 129 in vitro assays. The one exception in which significant inhibition (Ͼ50%) was observed was with a ␤-lactamase.

DISCUSSION
The goal of a ␤-lactamase inhibitor in a BL-BLI combination is to rescue the ␤-lactam from degradation by the complement of ␤-lactamases present, thereby restoring its activity to that seen in the absence of ␤-lactamase enzymes and essentially restoring the MICs to a wild-type distribution. The BL-BLI approach was effectively introduced with the first-generation BLIs tazobactam, sulbactam, and clavulanic acid paired with piperacillin, ampicillin, and amoxicillin, respectively. This strategy has been extended more recently to protected cephalosporins (ceftazidime-avibactam and ceftolozanetazobactam) and carbapenems (meropenem-vaborbactam and imipenem-relebactam). The focus of these recently approved BL-BLI combinations had been to address the growing concerns regarding KPC-type enzymes. With the advent of protected cephalosporins and carbapenems, there is an opportunity to reestablish the stratification of ␤-lactams into front-line protected cephalosporins (ceftazidime-avibactam, ceftolozanetazobactam) and to reserve carbapenems (e.g., meropenem-vaborbactam, imipenem-

FIG 6
Time-kill curves for cefepime-taniborbactam relative to those for ceftazidime-avibactam in two metallo-␤-lactamase-producing clinical isolates. The log 10 value of the number of viable CFU per milliliter is displayed on the y axis versus time (in hours) on the x axis. Curves for 1ϫ, 2ϫ, and 4ϫ MIC for cefepime (FEP) with taniborbactam (TAN) fixed at 4 g/ml are shown for each strain, while the curves for ceftazidime (CAZ) at 32 g/ml with avibactam (AVI) fixed at 4 g/ml, representing 0.015ϫ MIC (K. pneumoniae CDC-0049) and 0.5ϫ MIC (P. aeruginosa PS-12) of ceftazidime, which are well above the CLSI susceptibility breakpoint for ceftazidime-avibactam in Enterobacterales and P. aeruginosa of 8 g/ml, are shown for each strain.  (42), and NDM-1, -5, and -7 MBL-producing Enterobacterales in India and China (43)(44)(45). None of the recently approved BL-BLI combinations is sufficiently active against Enterobacterales or P. aeruginosa isolates producing MBLs, and meropenem-vaborbactam also lacks coverage against OXA-48 producers (46). Taniborbactam, which is being developed in combination with cefepime for the treatment of complicated urinary tract infections (cUTI) and hospital-acquired or ventilator-associated bacterial pneumonia (HABP/VABP), represents the next stage of this approach. We have now demonstrated the uniquely potent activity of cefepime-taniborbactam against both SBLs and MBLs and especially against these emerging MBLs and OXA-48/OXA-48-like SBLs compared to the activities of ceftazidime-avibactam, ceftolozane-tazobactam, and cefepime-tazobactam. In addition to emerging MBLs and OXA-48/OXA-48-like SBLs, reports of antibiotic resistance evolving during the course of clinical therapy with BL-BLIs, including resistance to ceftazidime-avibactam in KPC-producing Enterobacterales and to ceftolozanetazobactam in P. aeruginosa strains producing altered PDC variants, are increasing. Most concerning with the PDC variants is the fact that cross-resistance to ceftazidimeavibactam is often associated with resistance to ceftolozane-tazobactam (27,28). The activity of cefepime-taniborbactam against these resistant variants of PDC and KPC would, importantly, provide a future alternative therapeutic option to address evolving resistance. Cefepime-taniborbactam thus has the potential, once it is approved for clinical use, to provide the broadest coverage of ␤-lactamases to complement the existing repertoire of agents active against Gram-negative bacteria in the clinic and help address these problematic ␤-lactamase expansions by providing additional options for physicians to effectively treat these infections.

MATERIALS AND METHODS
Expression plasmid construction. Plasmid DNA, PCR product purification, and gel extractions were performed using Wizard Plus SV miniprep and SV gel and PCR extraction kits (Promega). The NdeI, BamHI, and XhoI restriction enzymes, T4 DNA ligase, and E. coli BL21(DE3) competent cells were purchased from New England Biolabs. All oligonucleotide primers for PCR amplification were purchased from Integrated DNA Technologies. All PCRs were performed with Phusion high-fidelity DNA polymerase, and cloning was performed in subcloning-efficient E. coli DH5␣ chemically competent cells (Thermo Fisher). The pET9a (Agilent) and pET24a (MilliporeSigma) expression clones were made using PCR amplification products from molecularly characterized clinical isolates carrying the desired ␤-lactamase gene inserted into pET9a using NdeI and BamHI in all cases except for the bla NDM-1 gene, for which NdeI and XhoI were used in pET24a. All ␤-lactamases except NDM-1 were cloned with signal peptide-encoding sequences. NDM-1 lacked the coding sequence for the first 28 amino acids, encompassing the signal peptide and the lipobox (LSGC) peptide sequence. All transformants were verified by PCR amplification, restriction endonuclease mapping, and DNA sequencing. Confirmed expression plasmids were isolated by use of a plasmid miniprep kit and used to transform the expression cell line E. coli BL21(DE3) or E. coli JM109(DE3).
Construction of isogenic strains of E. coli producing individual ␤-lactamases. The isogenic strains were used to establish a spectrum of inhibitory activity against 34 ␤-lactamases that were engineered into E. coli DH5␣ cells carrying chromosomal AmpC (ESC-1), a non-ESBL enzyme unable to hydrolyze these third-or fourth-generation cephalosporins. Each ␤-lactamase gene region encoding the periplasmic protein was placed under the control of the bla TEM-1 promoter and a signal peptide sequence to drive the expression and localization of the ␤-lactamases to the periplasm. The DNA fragments containing the promoter, the signal sequence, and each ␤-lactamase-coding gene were synthesized and cloned into pTwist Chlor High Copy in DH5␣ at Twist Bioscience. The individual expression plasmids along with their GenBank accession numbers are listed in Table S2 in the supplemental material and below. The protein sequences of ␤-lactamases were obtained from the Beta-Lactamase DataBase (www.bldb.eu) (47). Expression of the ␤-lactamases was confirmed by verifying decreased susceptibility to example test antibiotics that are known substrates of the enzymes. E. coli DH5␣ carrying pTU501, which expressed only the TEM-1 signal sequence from the bla promoter, was used as a control. The plasmid sequences are available through GenBank (Table S2).
␤-Lactamase purification. For all ␤-lactamases except P99 AmpC, a 50-ml preculture of E. coli BL21(DE3) cells containing the pET-based expression vector (pET24a for NDM-1 and pET9a otherwise) for the individual ␤-lactamases was grown overnight in lysogeny broth (LB) medium at 37°C in the presence of kanamycin selection at 50 g/ml. Common to all ␤-lactamase purifications, bacterial cells were lysed by three consecutive passes through a chilled French pressure cell at 18,500 lb/in 2 and clarified by centrifugation at 10,000 ϫ g for 30 min at 4°C. The ␤-lactamase activity was monitored using nitrocefin at 100 M, and purity was examined by 10% SDS-PAGE with Coomassie brilliant blue staining. Purified proteins exhibiting Ͼ95% purity by SDS-PAGE were quantified by use of a Pierce bicinchoninic acid protein assay kit and bovine serum albumin (BSA) as the standard (Thermo Fisher), concentrated to a working range of 1 to 5 mg/ml, and frozen at Ϫ80°C in buffer containing 10% glycerol. Column purifications were performed using an Åkta fast-performance liquid chromatograph (GE Healthcare). The purification schemes were generally similar, with the enzyme-specific differences being described below.
CTX-M-15 was obtained from E. coli BL21(DE3) carrying plasmid pET-CTX-M-15, grown in 2 liters of MagicMedia autoinduction medium (Invitrogen) containing 50 g/ml kanamycin (Sigma) for 24 h at 23°C. Cells were harvested at an A 600 of 2.2 by centrifugation at 7,500 ϫ g at 4°C and resuspended in 60 ml of 10 mM HEPES, pH 7, supplemented with 0.5 mM EDTA. The lysate was diluted 5-fold with cold 50 mM sodium acetate, pH 4.8, and incubated overnight at 4°C. The extract was clarified by centrifugation at 14,500 ϫ g at 4°C, filtered through an Amicon nitrogen concentrator with a 10-kDa-cutoff filter to a volume of 50 ml, and loaded onto a HiTrap Capto S column that had been preequilibrated in 50 mM sodium acetate, pH 4.8. Protein was eluted with a linear gradient of 50 mM sodium acetate, pH 4.8, supplemented with 500 mM NaCl. Fractions containing active CTX-M-15 were pooled, concentrated, and buffer exchanged in 20 mM HEPES, pH 7.2, 150 mM NaCl, and 10% glycerol using Amicon Ultra-15 centrifugal concentrators. CTX-M-15 was further separated by use of a Superdex 200 gel filtration column.
OXA-48 was obtained from E. coli BL21(DE3) carrying plasmid pET-OXA-48, grown in 2 liters of MagicMedia autoinduction medium (Invitrogen) with kanamycin as described above for 24 h at 23°C. Cells were harvested at an A 600 of 3 by centrifugation at 7,500 ϫ g at 4°C, resuspended in 60 ml of 20 mM triethanolamine, pH 5.5, and then purified as described above for CTX-M-15, with the exception that the final buffer contained 10 mM NaHCO 3 to maintain the critical active-site lysine residue carbamylated (48).
KPC-2 was obtained from E. coli BL21(DE3) carrying plasmid pET-KPC-2, grown in 3 liters of MagicMedia autoinduction medium (Invitrogen) with kanamycin as described above for 24 h at 23°C. Cells were harvested at an A 600 of 3.3 by centrifugation at 7,500 ϫ g at 4°C and resuspended in 70 ml of 20 mM MES (morpholineethanesulfonic acid; pH 5.5). The extract was clarified by centrifugation at 14,500 ϫ g at 4°C, filtered through an Amicon nitrogen concentrator by use of 10-kDa-cutoff filters to a volume of 50 ml, and loaded onto a HiTrap Capto S column that had been preequilibrated in 20 mM MES, pH 5.5. Protein was eluted with a linear gradient of 20 mM MES, pH 5.5, supplemented with 500 mM NaCl. KPC-2 active fractions were pooled, concentrated, and buffer exchanged in 20 mM HEPES, pH 7.3, 150 mM NaCl using Amicon Ultra-15 centrifugal concentrators. KPC-2 was further separated by gel filtration chromatography with a Superdex 200 column.
P99 AmpC was purified directly from the Enterobacter cloacae SIP9925 P99ϩ clinical isolate after the sequence of the ␤-lactamase-encoding gene had been verified by DNA sequencing of the PCR-amplified product. To produce P99 AmpC, Enterobacter cloacae P99ϩ cells were grown in LB in the presence of a sub-MIC (0.015 g/ml) of imipenem to induce maximal expression of the enzyme. Cells were harvested at an A 600 of 2.4 by centrifugation at 7,500 ϫ g at 4°C, resuspended in 50 ml of 20 mM MES, pH 5.0, and otherwise purified in a manner similar to that for KPC-2 described above.
SHV-5 was obtained from E. coli BL21(DE3)/pLysS carrying plasmid pET-SHV-5, grown in 2 liters of Super broth at 25°C to an A 600 of 0.5, when IPTG (isopropyl-␤-D-thiogalactopyranoside) was added to a final concentration of 0.05 mM, and induction proceeded for 6 h. Cells were harvested by centrifugation at 5,500 ϫ g for 15 min at 4°C. The cell pellet was resuspended in 60 ml of 10 mM HEPES buffer, pH 7.5, and the cells were lysed with a French press. The lysate was diluted 5-fold in 50 mM sodium acetate buffer, pH 5, and kept at 4°C overnight. The extract was clarified by centrifugation at 14,500 ϫ g at 4°C, filtered through an Amicon nitrogen concentrator by the use of 10-kDa-cutoff filters to a volume of 50 ml, and loaded onto a HiTrap Capto S column equilibrated with 50 mM sodium acetate buffer, pH 5. SHV-5 was eluted with a linear gradient of 50 mM sodium acetate, pH 5, supplemented with 500 mM NaCl. Active fractions were pooled, concentrated, and buffer exchanged in 20 mM HEPES, pH 7.3, 150 mM NaCl, and 10% glycerol using Amicon Ultra-15 centrifugal concentrators. Finally, the SHV-5 sample was further purified by chromatography over a Superdex 200 gel filtration column.
VIM-2 was purified from E. coli BL21(DE3) carrying pET-VIM-2 as previously described (49) with the following changes. MagicMedia autoinduction medium was used instead of Luria-Bertani (LB) broth with IPTG induction, and the cells were harvested at an A 600 of 2.1 by centrifugation at 7,500 ϫ g for 15 min at 4°C. The 50 to 80% ammonium sulfate precipitate was collected by centrifugation at 13,000 ϫ g for 1 h at 4°C and solubilized in 20 mM HEPES, pH 7.2, 50 M ZnSO 4 at 1/20 of the original volume, and the solution was loaded onto a 30-ml Q Sepharose anion exchange column (GE Healthcare) that had been equilibrated with the same buffer. Elution of VIM-2 was achieved with a linear NaCl gradient to 1 M. Fractions containing VIM-2 were pooled and concentrated with Amicon Ultra-15 centrifugal concentrators with a 10-kDa cutoff. VIM-2 was dialyzed in 50 mM Tris, pH 9, 50 M ZnSO 4 overnight and loaded onto a MonoQ column that had been preequilibrated in the same buffer, and elution was achieved with a linear gradient of NaCl up to 500 mM. Active fractions containing VIM-2 were pooled, concentrated, buffer exchanged in 50 mM HEPES, pH 7.5, 50 M ZnSO 4 , 200 mM NaCl, and further separated by gel filtration with a Superdex 200 column (GE Healthcare) that had been equilibrated in the same buffer.
IMP-1 was purified from E. coli BL21(DE3) carrying plasmid pET-IMP-1, grown in 3 liters of LB at 37°C with 50 g/ml kanamycin selection as previously described with minor modifications (50). At an A 600 of 0.7, IPTG was added to a final concentration of 0.5 mM, and induction of IMP-1 expression was allowed to proceed for an additional 5 h at 30°C. Cells were centrifuged at 7,500 ϫ g for 15 min at 4°C. The pellet was resuspended in 50 mM HEPES, pH 7, containing 50 M ZnSO 4 . The lysate was clarified by centrifu-gation at 12,500 ϫ g for 30 min at 4°C and loaded onto a HiTrap Capto S column equilibrated in 50 mM HEPES, pH 7, 50 M ZnSO 4 . The column was washed with the same buffer, and the enzyme was eluted with a linear gradient of NaCl to 500 mM. The active fractions were pooled and concentrated by ultrafiltration with Amicon centrifugal concentrators with a 10-kDa cutoff. The protein solution was dialyzed in 50 mM HEPES, pH 7, 50 M ZnSO 4 and loaded onto a MonoS cation exchange column that had been preequilibrated in the same buffer. The enzyme was eluted with a linear gradient of NaCl to 500 mM. Fractions containing active IMP-1 were collected, pooled, and concentrated to 1 mg/ml. NDM-1 lacking the signal peptide (first 28 amino acids) and lipobox sequence was purified as a soluble protein from the cytoplasm of Escherichia coli JM109(DE3) carrying plasmid pET-NDM-SP, grown in 3 liters of MagicMedia medium with kanamycin selection as described above. Protein expression was induced by the addition of 0.5 mM IPTG at an A 600 of 0.5, followed by incubation at 23°C for 18 h. Cells were harvested by centrifugation at 7,500 ϫ g for 15 min at 4°C and washed twice with 25 mM Tris-HCl, pH 7, 50 M ZnSO 4 . Cell lysates obtained with a French press were clarified by centrifugation at 12,500 ϫ g for 30 min at 4°C. The supernatant was loaded onto a Q Sepharose column that had been . Similar concentration ranges were tested for both avibactam and vaborbactam. A reversible two-step inhibition model was fit to the data.
Pseudo-first-order rate constants of enzyme inactivation (k obs ) were determined in the presence of various inhibitor concentrations by fitting equation 2 to the time courses.
where P is product, V s is rate at steady state, t is time, V 0 is initial velocity, and k is rate of inactivation. A plot of k obs versus the inhibitor concentrations generated a linear plot with a slope of (k 2 /K i ) and a y intercept of k Ϫ2 (equation 3). The reported values were corrected for the substrate concentration and the K m for each substrate-enzyme combination, as defined by the 1 ϩ ([S]/K m ) term in equation 4, where S is the substrate concentration, used to obtain the second-order rate constant k 2 /K i and the off rate (k Ϫ2 ).
The error values reported are standard errors from the fit.
A plot of the fractional steady-state velocities showed a linear relationship with the inhibitor concentrations tested, from which an equilibrium dissociation constant (K i * ; equation 4) was derived (equation 3).
Imipenem and cefotaxime were used as substrates for determining the K i * of taniborbactam with OXA-48 and SHV-5, respectively. Taniborbactam was tested in triplicate at concentrations ranging from 0.050 to 5 M with OXA-48 (2.5 nM) and from 0.4 to 100 nM with SHV-5 (5 nM) in 50 mM sodium phosphate (pH 7.0). Similar concentration ranges were tested with avibactam and vaborbactam. Off rates (k Ϫ2 ) were assessed by a jump dilution method in triplicate reactions monitored continuously for the recovery of nitrocefinase enzymatic activity (Δ, Ϫ20,500 M Ϫ1 cm Ϫ1 at 486 nm). CTX-M-15 (2 M), KPC-2 (4 M), and P99 AmpC (1 M) were inactivated by taniborbactam at 10, 20, and 5 M, respectively. Similar concentration ranges were tested with avibactam and vaborbactam. A 4,000-fold jump dilution was performed for the ␤-lactamase-taniborbactam reaction mixtures. Enzyme activity was monitored continuously by determination of the absorbance at 486 nm after the final 10-fold dilution, consisting of the addition of 20 l of an inactivation reaction mixture to 180 l of 1ϫ phosphate-buffered saline, pH 7.4, 0.1 mg/ml BSA at 37°C and nitrocefin at a 150 M final concentration. The percentage of enzyme activity at each assay point was determined by comparison to the activity of the enzyme in the absence of inhibitor, and the recovery of enzymatic activity was fit to a single exponential with the associated standard deviation.
Inhibition of VIM-2 and NDM-1 MBLs. The inhibition modality of taniborbactam with both the VIM-2 and NDM-1 metallo-␤-lactamases was assessed by monitoring the impact of various concentrations of taniborbactam on the Michaelis-Menten kinetic parameters K m and V max . The time dependence of the inhibitory activity was examined by preincubation of the enzyme with the inhibitor for up to 30 min prior to the initiation of the reactions, and the recovery of enzymatic activity was assessed by the jump dilution method, as previously described (33). For VIM-2, triplicate reactions (150 l) were initiated by addition of the substrate nitrocefin, which was present at concentrations ranging from 11.5 to 200 M. The reaction mixture contained 141 ng (4.5 pmol) of VIM-2 tested against five concentrations of taniborbactam (5,10,20,40, and 80 nM), and substrate hydrolysis was monitored at 486 nm using a BioTek PowerWave XS2 UV/visible spectrophotometric plate reader. In contrast, for NDM-1, triplicate reactions (300 l) were initiated by addition of the substrate cefotaxime, which was present at concentrations ranging from 13.1 to 150 M. The reaction mixture contained 622.4 ng (24 pmol) of NDM-1 tested against seven concentrations of taniborbactam (6,12,24,48,72,96, and 120 nM), and substrate hydrolysis was monitored at 260 nm. Avibactam and vaborbactam showed no significant inhibition up to 50 M. Nitrocefin was used as the substrate for assays with IMP-1. No preincubation of the enzyme and inhibitor was required for either VIM-2 or NDM-1. The inhibition modality was ascertained by global fitting of the Michaelis-Menten equation for competitive inhibition (equation 5) to all the data, from which the K i values obtained with VIM-2 and NDM-1 were calculated.
where [I] is the inhibitor concentration. Antibacterial activity. The in vitro antibacterial activity of the ␤-lactams alone or in combination with the BLIs was determined in cation-adjusted Mueller-Hinton broth (CAMHB) microdilution assays according to CLSI recommendations (51). The potentiation of antibacterial activity by taniborbactam, avibactam, or tazobactam was assayed by fixing the concentration of the BLI at 4 g/ml for cefepime, ceftazidime, and ceftolozane, while tazobactam was added at a fixed concentration of 8 g/ml to cefepime, as described by Wockhardt Pharmaceuticals. The inocula for the broth microdilution assays were prepared by the broth culture method (51). The ␤-lactams were 2-fold serially diluted in CAMHB, with the final concentrations generally ranging from 0.06 to 128 g/ml when the agents were tested alone and from 0.016 to 32 g/ml when they were tested in combination with their respective BLI. The MICs reported are modal values from 5 independent replicates performed on separate days either in engineered E. coli DH5␣ producing individual ␤-lactamases or in clinical isolates of Enterobacterales and P. aeruginosa.
Justification for use of a fixed concentration of 4 g/ml of taniborbactam for in vitro testing. The in vitro activity of cefepime-taniborbactam was measured as the activity of cefepime in the presence of taniborbactam at a fixed concentration of 4 g/ml. The rationale for choosing 4 g/ml of taniborbactam for in vitro testing was based on in vitro assessments by broth microdilution and time-kill assays, the correlation of the results obtained with nonclinical models of infection, and pharmacokineticpharmacodynamic modeling.
Taniborbactam alone lacks antibacterial activity. At a fixed concentration of 4 g/ml taniborbactam, the MIC 90 of cefepime was reduced from Ͼ128 to 2 g/ml for cefepime-nonsusceptible Enterobacterales isolates, with 97.6% of the isolates being inhibited by 8 g/ml of cefepime (52). This value corresponds to the susceptible, dose-dependent (SDD) breakpoint for cefepime (37). The MIC distribution of cefepime-taniborbactam with taniborbactam fixed at 4 g/ml against cefepime-nonsusceptible Enterobacterales from surveillance similarly resembles the MIC distribution of cefepime against cefepimesusceptible isolates from the same study (52). The overlapping MIC distributions for cefepime against cefepime-susceptible strains and cefepime-taniborbactam against cefepime-nonsusceptible strains (with taniborbactam fixed at 4 g/ml) reflect the complete or nearly complete rescue of the activity of cefepime by taniborbactam against target organisms. Importantly, humanized dosing of cefepime and taniborbactam in the neutropenic murine thigh infection model reduced the bacterial burden of all isolates of Enterobacterales and P. aeruginosa with cefepime-taniborbactam MIC values of 8 g/ml or below by at least 1 log 10 (53). These results are consistent with the MIC values for cefepime derived in the presence of 4 g/ml of taniborbactam.
Isolation of avibactam and ceftolozane from clinical preparations. Avibactam was isolated from commercially available ceftazidime-avibactam (Avycaz). A single bottle of Avycaz (containing 2 g of ceftazidime and 0.5 g of avibactam) was dissolved in approximately 6 to 8 ml of water and purified using a Biotage 120-g C 18 reverse-phase column. A gradient utilizing a 99:1 mixture of H 2 O (with 0.1% trifluoroacetic acid [TFA])-acetonitrile (ACN) (with 0.1% TFA) for four column volumes (CV), followed by a ramp to 1:100 H 2 O-ACN over 1 CV, was employed. All fractions containing avibactam were combined, frozen, and lyophilized to provide pure avibactam. Electrospray ionization mass spectrometry (ESI-MS) confirmed an m/z of 266.1 (M ϩ H) ϩ .
Ceftolozane was isolated from commercially available ceftolozane-tazobactam (Zerbaxa). A single bottle of Zerbaxa (containing 1 g of ceftolozane and 0.5 g tazobactam) was dissolved in water (ϳ6 to 8 ml) and purified using a Biotage Snap Ultra 120-g C 18 reverse-phase column. A gradient beginning with 5 CVs of 95:5 H 2 O (with 0.1% TFA)-ACN (with 0.1% TFA) was followed by a ramp to 80:20 H 2 O-ACN over 6 CVs. The collected fractions were combined, frozen, and lyophilized. The isolated solid was triturated with 4.0 N HCl in diethyl ether and concentrated three times to provide ceftolozane as the hydrochloride salt. ESI-MS confirmed an m/z of 667.1 (M ϩ H) ϩ .
In vitro antibacterial kill assay. Bactericidal activity was assessed by a time-kill assay according to standard CLSI methods (54). The cefepime concentrations to be assayed in combination with tanibor-bactam were selected based on antibacterial activity (MIC), while ceftazidime-avibactam was tested at 32 g/ml (which is 4 times the CLSI susceptibility breakpoint of 8 g/ml) (37). Time-kill assays were performed in 14-ml glass tubes with a bacterial inoculum in CAMHB of 5 ϫ 10 5 CFU/ml. The tubes were incubated at 37°C with shaking at 200 rpm, and aliquots were drawn at six time points (0, 2, 4, 6, 8, and 24 h) to prepare 0.5-log 10 dilutions in CAMHB in 96-well plates with activated charcoal. From these dilution plates, 10 l was spotted onto OmniTrays using a replicator. Both the OmniTrays and the dilution plates were incubated overnight at 37°C and used to ascertain the viable bacterial counts. The smallest quantifiable amount obtained by this method was 2 log 10 CFU.
Spontaneous frequency of resistance. Five colonies were picked from agar plates with an inoculating loop and transferred aseptically to a glass Erlenmeyer flask containing 20 ml of CAMHB. The culture was grown for Ն5 h at 37°C with shaking at 200 rpm. CAMHB agar (15 g/liter; Difco agar; Becton, Dickinson) was prepared according to the manufacturer's directions. Cefepime was added at 4ϫ the MIC along with taniborbactam at a fixed concentration of 4 g/ml. The agar was dispensed into 100-mm petri dishes (ϳ20 ml per dish), with 100 plates being used per test organism. Agar plates without drug were used to obtain colony counts. The optical density of the growing cultures used for inoculation was assessed spectrophotometrically at 600 nm. The inocula were adjusted to a final suspension of 1 ϫ 10 8 CFU/ml to 1 ϫ 10 9 CFU/ml. The inoculum (100 l) was spread onto each plate with an inoculating loop. The plates were incubated at 37°C, and colony counts were assessed at 24 and 48 h.
Data availability. The GenBank accession numbers for the individual expression plasmids are MN307371, MN401186, MN401152 to MN401166, and MN401168 to MN401185.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.9 MB.