Antibiotics are a vital part of modern healthcare systems and one of the most significant advances in modern medicine 1. Beta-lactams are among the most widely prescribed and effective antibiotics, with a broad range of activity against most pathogenic bacteria 2. Beta-lactams function by inhibiting the synthesis of the peptidoglycan layer in the bacterial cell wall 3. Structurally, the beta-lactam antibiotics are currently separated into four classes: penicillins, cephalosporins, carbapenems and monobactams, which differ in either the R groups attached to either the four-membered beta-lactam ring or in the case of the penicillin, carbapenem and cephalosporins, the attached five or six-membered rings 4. Due to the wide range of derivatives based on these core scaffolds, the uptake and retention of beta-lactam antibiotics within the human body vary partly due to their specific interactions with different families of solute carrier (SLC) transporters 5–7. A particular focus in current antibiotic drug development is improving the oral bioavailability of carbapenems, which are highly effective against Gram-negative, Gram-positive, and anaerobic bacteria. However, current formulations of carbapenems show increased breakdown in the gut and poor transport across the gut epithelia, restricting the options available to clinicians in delivering these drugs to patients 8. Understanding the interactions between small molecule drugs and solute carriers is a promising route to improving drug pharmacokinetics and efficacy 9.
The proton-coupled peptide transporters PepT1 (SLC15A1) and PepT2 (SLC15A2) have been extensively studied due to their influence on the oral bioavailability and renal clearance of beta-lactam antibiotics in the body 7,10–21. The primary physiological role of the SLC15 family is the absorption and retention of dietary nitrogen in the form of di- and tripeptides 22,23. Ingested protein is broken down and transported across the intestinal brush border membrane via the plasma membrane peptide transporter, PepT1 24–26. In contrast, circulating peptides are retained in the body through reabsorption via PepT2, which selectively retains peptides in the kidney 27,28, as well as regulates peptide transport across the blood-brain barrier 7,29 (Fig. 1A). PepT1 and PepT2 are unusual solute carriers in being highly promiscuous 30, able to recognize a large and diverse range of chemically distinct substrates 31. The ability to recognize different chemical groups underpins the role played by these proteins in beta-lactam antibiotic uptake, as these drugs display both steric and chemical similarity to tripeptides 32.
PepT1 and PepT2 are members of the Proton-coupled Oligopeptide Transporter or POT family, which is widely distributed within pro- and eukaryotic genomes 30. POT family transporters belong to the Major Facilitator Superfamily (MFS). They are all proton (H+) driven symporters, using the inwardly direct proton electrochemical gradient (ΔµH+) to drive the concentrative uptake of peptides and drugs into the cell 32. Recently, the structures of PepT1 and PepT2 have been reported from both human and mammalian species, providing valuable insights into the mechanisms underpinning peptide and prodrug recognition 33–35. These studies build on previous structural and biochemical studies on the prokaryotic POT family members to understand the structural basis for ligand promiscuity 36–55. However, the molecular basis for the selective recognition and uptake of beta-lactam antibiotics remains unclear, hampering efforts to improve the pharmacokinetic profiles of new antibiotic drugs.
To address this important aspect of SLC15 function, we determined the structure of PepT2 from Rattus norvegicus in complex with three different beta-lactam antibiotics. Using in vitro assays, we demonstrate that one of the beta-lactam antibiotics, cloxacillin, functions as a competitive inhibitor. Combining our cryo-EM structures with molecular dynamics to probe the role of protonation in substrate recognition, we identify a crucial role for the primary amine group in orientating beta-lactam antibiotics in the biding site and proton binding in locking the drug via the carboxylate group. Our results establish a working pharmacophore model for beta-lactam recognition and explain the differences between substrate and inhibition mechanisms.
Characterization of beta-lactam transport
The proton-coupled peptide transporters are known to transport several families of beta-lactam antibiotics, including cephalosporins and penicillins. Although the substrate range recognized is extensive, encompassing many different drug classes, substrate specificity exists within the beta-lactam families, with some antibiotics exhibiting higher affinity, such as cefadroxil with a Ki 3 µM, medium affinity, such as amoxicillin with a Ki 0.2–0.43 mM and low affinity, such as Cloxacillin Ki 1 mM 10,19 (Fig. 1B). However, to date, the transport has been primarily characterized using studies reporting the inhibition of a radioactive reporter peptide in cell-based assays 10,56. To verify the Ki values reported for these cell-based studies, we reconstituted rat PepT2 into liposomes and calculated IC50 values using inhibition of the uptake of radioactive di-alanine peptide (Fig. 1C). While cefadroxil displayed an IC50 of 19 µM ± 3, both amoxicillin and cloxacillin displayed significantly weaker values of 270 µM ± 39 and 203 µM ± 21, respectively. In comparison, the di-alanine peptide had an IC50 of 32 µM ± 3. However, a drawback of IC50 studies is the inability to discriminate between transported ligands and non-transported inhibitors. Therefore, we tested the ability of these antibiotics to drive transport via a counterflow experiment compared to the di-alanine peptide (Fig. 1D). As expected, cefadroxil was able to drive transport more effectively than di-alanine, consistent with the calculated IC50 values.
Similarly, amoxicillin was less effective, showing ~ 35% uptake. However, cloxacillin was unable to drive transport, even at high concentrations (10 mM). To verify the inhibitory property of this antibiotic, we used the negative control of Lys[Z(NO2)]-Pro (LZNP), a known high-affinity inhibitor of PepT2 57. Recently, the use of solid support membrane devices has enabled measurements of changes in membrane capacitance following charge movement via secondary active transporters 58. We analyzed the activity of the reconstituted PepT2 using the SURFR2 platform, which also confirmed the radioactive uptake and counterflow assays (Fig. 1E). Whilst the addition of di-alanine, cefadroxil and amoxicillin all generated a significant capacitance change, the addition of cloxacillin did not. Thus, we conclude that cloxacillin is not transported by PepT2 but likely acts as a competitive inhibitor. We considered the question of why certain antibiotics function as ligands while others are inhibitors, as this behaviour has implications for inhibitor design more broadly within the SLC15 family, which includes members linked to inflammatory regulation 59. We, therefore, sought to understand the nature of the binding interactions that determine good compared to weak substrates and verify the inhibitory mechanism of cloxacillin.
Cryo-EM structure of PepT2 in complex with Cefadroxil and Amoxicillin
To gain further insight into the mechanism of antibiotic recognition, we determined the structure of PepT2 in complex with both cefadroxil and amoxicillin at 3.1 Å and 3.2 Å, respectively, using cryo-EM (Fig. 2A & B, Table 1, Fig. S1 and S2). The structures were obtained using an inhibitory nanobody identified in our previous study reporting the apo structure 33. In both structures, the transporter adopts an outward open, extracellular facing state, with a large solvent-accessible cavity extending from the exterior of the transporter down towards the intracellular gate, which is constructed from the packing together of TMs 4–5 and TMs10-11 (Fig. S3). The overall arrangement of TM helices in both structures is essentially identical to the previously obtained apo structure (PDB:7NQK), except the coulomb density map around the extracellular domain (ECD) is much lower resolution, resulting in our decision not to build the IgG domains in these structures. However, clear density for the nanobody was obtained, which binds to the extracellular regions of TMs 1 and 2 33 (Fig. S1 and S2). The structures have root mean square deviations (r.m.s.d) of 0.62 Å and 0.45 Å for the cefadroxil and amoxicillin structures, respectively, when aligned against the TM helices (562 Cα atoms) of the apo state of the rat PepT2 protein (PDB:7NQK).
Cefadroxil can be clearly observed in the map and adopts an elongated 'L-shaped' conformation (Fig. 3A & B). The hydroxyphenyl group interacts with Asp317 on the extracellular gating helix, TM7, which also forms part of a negatively charged pocket within the binding site. The amino group sits adjacent to Glu622 on the intracellular gating helix, TM10, and interacts with Asn192 on TM4, which forms part of the intracellular gating helices in the N-terminal bundle. The carbonyl group of cefadroxil also interacts with Asn192, anchoring this region of the drug within the transporter. The remainder of the drug molecule makes fewer direct interactions with the binding site, with only the carbonyl group in the beta-lactam ring hydrogen bonding to Tyr61 on the extracellular gating helix, TM1. The methyl group at the C3 position on the dihydrothiazine ring sits within a hydrophobic pocket formed by Leu650, Val653 and Trp649 on gating helix TM11 and observed in bacterial homologues of the POT family 37. The carboxylate group sits within the positively charged pocket formed by Arg57 on TM1 and Lys161 on TM4. Previous structures of peptide-bound prokaryotic POT family homologues and human PepT2 have identified the critical role electrostatics play in correctly orientating peptides within the binding site 34,37,44. Specifically, the conserved glutamate, Glu622 on TM10 and Arg57 on TM1, function to clamp the amino and carboxy termini of peptides, respectively.
Table 1
Cryo-EM data collection, refinement, and validation statistics
| RnPepT2 cefadroxil bound state (EMDB-44599) (PDB 9BIR) | RnPepT2 amoxicillin bound state (EMDB-44600) (PDB 9BIS) | RnPepT2 cloxacillin (pose 1) bound state (EMDB-44601) (PDB 9BIT) | RnPepT2 cloxacillin (pose 2) bound state (EMDB-44602) (PDB 9BIU) |
Data collection and processing | | | | |
Magnification | 165,000 | 165,000 | 165,000 |
Voltage (kV) | 300 | 300 | 300 |
Electron exposure (e–/Å2) | 54.8 | 55.3 | 57.6 |
Defocus range (µm) | -2.5 to -0.8 | -2.0 to -0.5 | -2.0 to -0.5 |
Pixel size (Å) | 0.693 | 0.698 | 0.732 |
Symmetry imposed | C1 | C1 | C1 |
Initial particle images (no.) | 8,932,246 | 4,358,411 | 4,006,243 |
Final particle images (no.) | 93,759 | 87,246 | 106,684 | 201,206 |
Map resolution (Å) FSC threshold | 3.1 0.143 | 3.2 0.143 | 3.1 0.143 | 2.9 0.143 |
Refinement | | | | |
Initial model used (PDB code) | RnPepT2 (7NQK) | | | |
Model composition in the asymmetric unit Non-hydrogen atoms Protein residues Ligands | 4667 590 25 | 4710 595 25 | 4714 595 29 | 4714 595 29 |
Average B factors (Å2) Protein Ligand | 34.55 42.62 | 77.51 87.05 | 41.84 39.31 | 49.03 52.96 |
R.m.s. deviations Bond lengths (Å) Bond angles (°) | 0.004 0.49 | 0.004 0.62 | 0.003 0.57 | 0.002 0.57 |
Validation MolProbity score Clashscore Poor rotamers (%) | 1.91 7.15 0.20 | 1.90 7.09 0.40 | 1.92 5.72 1.60 | 1.93 4.86 2.00 |
Amoxicillin adopts a similar binding position to that observed for cefadroxil, with the hydroxyphenyl, amino and carboxylate groups sitting in equivalent positions in the binding site (Fig. 3C & D). However, the beta-lactam ring is rotated ~ 90º relative to cefadroxil, with the carbonyl group pointing towards the extracellular entrance to the binding pocket. The two methyl groups on the 5-membered thiazolidine ring point towards Trp313, Val653 and Trp649, which constitute a hydrophobic pocket near TM11. Consistent with the lower IC50 values, amoxicillin makes far fewer direct interactions with the transporter, which appear to result from the altered orientation of the beta-lactam ring, presumably to accommodate the geometry of the thiazolidine ring. As observed in the cefadroxil structure, the amino group sits in the negatively charged pocket and interacts with Asn192 via a hydrogen bond but now sits a little further from Glu622 (~ 3.5 Å). Tryosine 61 now interacts with the carbonyl group of the peptide bond, as opposed to the beta-lactam ring observed in cefadroxil. As noted above, repositioning the beta-lactam ring substantially changes the location of the carbonyl group of amoxicillin, which now interacts with Tyr94 on TM2. Together, these structures establish the fundamental recognition mechanism for beta-lactam antibiotics within PepT2 and enable the comparison with physiological peptide binding, as discussed below.
Structural basis for inhibition by Cloxacillin
Having established how PepT2 discriminates between a cephalosporin and an aminopenicillin, we next sought to uncover the mechanism of inhibition observed for cloxacillin, a semisynthetic penicillin carrying a 3-(2-chlorophenyl)-5-methylisoxazole-4-carboxamido group at position 6 of the beta-lactam ring. From the cryo-EM data, we could generate two maps derived from independent 3D classification schemes of a consensus particle set, which show two different positions for the drug within the binding site (Fig. 4A, Figs. S4-S5). Pose one was obtained from 106,684 particles and generated a map at 3.1 Å resolution (map 1), whereas pose two was obtained from 201,206 particles, which generated a map at 2.9 Å resolution (map 2) (Table 1).
Similar to the cefadroxil and amoxicillin complex structures, the protein backbone showed no obvious differences to those obtained for the apo PepT2 (r.m.s.d of 0.35 Å for 531 Cα atoms). In pose 1, cloxacillin adopts a vertical orientation with the chlorophenyl group sitting within 3Å of the conserved E53xxER motif on TM1 and adjacent to the intracellular gate occlusion formed by TMs 4, 5 and 10, 11 packing against one another (Fig. 4B & Fig. S5). The methylisoxazole group sits close to Trp313, Trp649 and Tyr188, with the carbonyl group of the peptide bond sitting close to Lys161. The beta-lactam ring makes no direct contact; however, the sulphur atom in the thiazolidine ring sits close to Tyr61. At the end of the molecule, the carboxylate group makes the only observed hydrogen bond interaction with Tyr94, which presumably stabilizes the vertical orientation in the binding site. In pose 2, the chlorophenyl group adopts a similar location to pose 1, at the base of the binding site and sitting close to Glu53 in the E53xxER motif, but now sits closer to Ile191 and Arg57 (Fig. 4C & Fig. S5). Similarly, the methyl group in the methylisoxazole ring extends towards Trp313 and Val653. However, the two poses differ more substantially in the location of the beta-lactam and thiazolidine rings. Whereas in pose 1, the drug molecule sits in roughly the centre, in pose 2, cloxacillin makes several direct interactions with the binding site. Specifically, the peptide carbonyl group interacts with Tyr188 (TM5) and Ser626 (TM10). Tyrosine 188 also interacts with the carbonyl group in the beta-lactam ring, while the sulphur atom interacts with Tyr61, similar to that observed in pose 1. Of note is the beta-lactam ring carbonyl, which interacts with Glu622 in this pose. However, we can model two rotamer positions for Glu622 in the maps, suggesting this interaction is not stable, supporting our prediction that a free amino group is necessary to lock transported ligands into the binding site. The carboxyl group makes further hydrogen bond interactions with Asn192, which also interacts with the nitrogen in the beta-lactam ring. The two poses suggest that cloxacillin binding is dominated by the positioning of the chlorophenyl group at the base of the transporter, with no strict specificity for the location of the beta-lactam backbone. Taken together, these structures reveal that while cloxacillin interacts with many of the same side chains observed in the cefadroxil and amoxicillin complexes, it cannot adopt a stable binding pose, which likely results in this antibiotic's observed inhibitory properties. As discussed below, the inability to adopt a stable binding is likely to result from the absence of a free amino group, which would orientate the drug to engage Glu622 and position the carboxyl to engage Arg57.
Interplay of protonation and ligand recognition
PepT2 is a proton-coupled transporter, and therefore, understanding how drugs interact with the binding site requires consideration of the protonation states of key protonatable side chains 33,36,40,60. To gain further insights into substrate discrimination, proton coupling and inhibition in PepT2, we undertook unbiased molecular dynamics (MD) simulations of the drug molecules using the cryo-EM structures as starting poses (Fig. 5A & Fig. S6). For both cefadroxil and amoxicillin, three sets of simulations were run with the protonation state of Glu53 and Glu56 modulated (see methods). When both Glu53 and Glu56 are deprotonated, i.e. in their standard protonation states at pH 7.5, the amino group of cefadroxil remains stably bound to Glu622, and with some flexibility also to Asn192 and Asn348 (Fig. S7). In contrast, the carboxylate group is unstable and sits ~ 6–10 Å away from Arg57 (Fig. 5A). These results are consistent with the interaction network observed in the cryo-EM structure (Fig. 3) and previous MD simulations of peptide ligands, which highlighted the role of the amino terminus in the initial capture of ligands 33. However, following the protonation of Glu56 and, to a lesser extent, Glu53, the carboxylate group of cefadroxil moves to interact with Arg57 (Fig. 5B). The same pattern of interactions occurs in the amoxicillin simulations, with the protonation of Glu56 releasing Arg57 to clamp the carboxylate group. However, when Glu56 is protonated, and to a lesser extent with Glu53 protonated, the amino group of amoxicillin moves away from Glu622 and detaches entirely from Asn192 (Fig. S7). This change to the orientation can be visualized by projecting the pooled trajectories onto a 2D plane, representing a slice through the 3D volume of the binding site (Fig. 5C). These 2D plots illustrate the extent to which protonation of Glu56 moves the centre of mass for the carboxylate group of both cefadroxil and amoxicillin closer to Arg57, which establishes the necessary interactions between the ligand and N- and C-terminal bundles of the transporter. The observation that amoxicillin cannot interact stably with the Glu622, Asn192 and Asn348 triad at the amino group and Arg57 at the carboxyl group simultaneously provides a reasonable explanation for its reduced experimental affinity compared to cefadroxil (Fig. 1C).
We next validated the significance of our results from the unbiased MD by running absolute binding free energy (ABFE) simulations of cefadroxil and amoxicillin for deprotonated and protonated Glu56 (see methods). The results indicate that cefadroxil and amoxicillin have similar binding free energies to the transporter in their cryo-EM poses (Table 2). Once Glu56 is protonated, however, cefadroxil experiences a substantial gain in affinity of ~ 6 kcal mol− 1, while amoxicillin affinity is reduced by ~ 1.7 kcal mol− 1, supporting the conclusions drawn from the unbiased MD runs above.
Table 2
Impact of protonation on the binding free energies of selected antibiotics in PepT2. Absolute Binding Free Energy values were obtained (see methods for details) with either Glu56 protonated or deprotonated for the antibiotic complexes.
Ligand | Glu56 | ΔG / kcal mol− 1 |
Cefadroxil | not protonated | -12.0 ± 1.5 |
protonated | -18.0 ± 0.5 |
Amoxicillin | not protonated | -10.6 ± 1.7 |
protonated | -8.9 ± 1.1 |
Cloxacillin | not protonated | ~ -6.8 |
protonated | ~ -5.5 |
The cryo-EM structures of cloxacillin revealed a more dynamic binding mode compared to either cefadroxil or amoxicillin (Fig. 4). Indeed, our MD simulations support these observations, as cloxacillin fails to adopt a stable binding pose within the 6 x1 µs trajectories for each of the three cryo-EM models (Fig. S8). The absence of an amino group results in no significant interaction with Glu622 by any functional group in the drug molecule. Similarly, the carboxylate group also fails to interact with Arg57 with increased frequency following Glu56 protonation. The failure of cloxacillin to engage in the specific binding pocket interactions formed by cefadroxil and amoxicillin is also quantitatively reflected in ABFE affinities (Table 2). When Glu56 is deprotonated, the affinity is ~ 4 kcal mol− 1 lower than either cefadroxil or amoxicillin, and unlike the two transported antibiotics, there is no stabilization of binding upon Glu56 protonation. As discussed below, the likely mechanism for inhibition appears to be simple steric occlusion and failure to trigger the necessary interactions to Glu622 and Arg57 required for transport.
Structural discrimination between substrates and inhibitors within the beta-lactam family.
The inhibitory property of cloxacillin was unexpected but presented an opportunity to decode further the structural differences between substrates and inhibitors within the beta-lactam family 61. For a comparative analysis, we tested five additional antibiotics from the penicillin and cephalosporin families. Using our counterflow assay, we first determined that PepT2 cannot transport either moxalactam, ceftibuten or benzylpenicillin (Fig. 6A).
Cefaclor, however, acts as a good substrate, achieving 76% activity relative to the physiological di-alanine peptide in the counterflow assay, and an IC50 of 60 µM (Fig. 6B). Ampicillin is also transported, albeit at a lower level than cefaclor, achieving 13% activity in the counterflow assay and an IC50 of 750 µM. This compares to IC50 value for moxalactam of 500 µM (Table S1). Together with our previous analysis of cefadroxil, amoxicillin and cloxacillin, we can now separate these beta-lactam drugs into substrates and inhibitors (Fig. 6C). Of note is the hydroxyl on the hydroxyphenyl group of cefadroxil and amoxicillin, which is absent in cefaclor and ampicillin respectively. In both cases, cefadroxil and amoxicillin are better substrates for PepT2 than either cefaclor or ampicillin, considering their IC50 values (Table S1). A plausible explanation for the increased affinity comes from the cryo-EM structure of cefadroxil, which reveals the hydroxyl is positioned close to a key gating helix, TM7 and interacts directly with Asp317 (Fig. 3A & B). Aspartate 317 forms part of an interaction network that controls the extracellular gate dynamics in response to proton binding to a conserved histidine on TM2, His87 in PepT2 36,62. Thus, the interaction between the ring hydroxyl in cefadroxil and amoxicillin with Asp317 positively affects both the recognition and transport of these drugs via PepT2.
Conversely, when assessing the inhibitory potency of the beta-lactam drugs, our data shows that cloxacillin, with an IC50 of 203 µM, is more effective than moxalactam (500µM), benzylpenicillin (1mM) or ceftibuten (1mM). Our MD analysis of cloxacillin shows the drug is unable to adopt a stable binding orientation with respect to Glu622 or Arg57 (Fig.S8), which is consistent with the three binding poses we observe in the cryo-EM structures (Fig. 4). A notable difference between cloxacillin and amoxicillin is the replacement of the primary amine group in the former compound for a methylisoxazole group in the latter. The absence of the primary amine removes a positive charge from cloxacillin and, therefore, the ability of the antibiotic to engage the Glu622, Asn192 and Asn348 triad on TMs 10, 4 and 9, respectively. Without this anchoring interaction, the binding of the drug is dominated by the chlorophenyl moiety, which in the cryo-EM structures binds in a hydrophobic pocket below Arg57 (Fig. 4). Interestingly, the chlorophenyl moiety of cloxacillin sits in a similar position to that modelled previously for the prodrug valacyclovir 33, indicating that the occupation of this pocket is not the reason for the inability of cloxacillin to trigger transport. More likely, the failure to adopt a stable binding pose and engage Glu622 explains the inhibitory properties of this drug. It is also consistent with the requirement of a free amino terminus for peptide ligands 27,63. Neither moxalactam nor benzylpenicillin have primary amine groups, so the most likely explanation for their ability to inhibit PepT2 is their inability to engage the Glu622 pocket, as observed for cloxacillin. Our results from ceftibuten, which does contain a primary amine, also highlight the importance of distance between the free amino and carboxyl groups in beta-lactam substrates. The addition of the amino-thiazol group extends the primary amine at a distance equivalent to a tetrapeptide, which is similar to moxalactam and likely makes these drugs too large to transport.