Cryo-EM structures of PamB2 on the ribosome

To investigate how paenilamicins inhibit translation, we generated PamB2-stalled ribosome complexes (PamB2-SRC) for single-particle cryo-EM analysis. Rather than forming complexes on vacant ribosomes, or with predefined functional states, we instead aimed to use more physiological complexes where translating ribosomes become stalled by PamB2. To achieve this, we performed in vitro translation reactions with E. coli ribosomes using the Met-Leu-Ile-Phe-stop-mRNA (MLIFstop-mRNA), a model template that we had previously used to successfully determine structures of drosocin-stalled ribosomal complexes12. Toeprinting was used to monitor the position of ribosomes on the MLIFstop-mRNA in the presence of increasing concentrations of synthetic PamB2 (Extended Data Fig. 2). As a positive control, we used thiostrepton that traps ribosomes on the AUG initiation codon in cell-free systems13,14,15, and as a negative control, we included the inactive N-acetylated form of PamB2 (N-Ac-PamB2)4 (Extended Data Fig. 2). In the absence of drug, bands are evident for ribosomes on the AUG start codon and the adjacent UUG (Leu) codon, suggesting that initiation and/or the first two elongation steps are slow on this messenger RNA (mRNA) or that the mRNA contains secondary structure in this region. In the presence of thiostrepton, a single strong band is observed that corresponds to ribosomes trapped on the AUG start codon (Extended Data Fig. 2), as expected13,14,15. By contrast, increasing concentrations of PamB2 led to a gradual loss of ribosomes at the AUG codon and an increase in ribosomes stalled one codon further with the UUG (Leu) codon in the P-site. This shift in ribosome positioning was not observed for the N-Ac-PamB2, where the pattern looks similar to the no-drug control, consistent with the inactivity of this compound4 (Extended Data Fig. 2). We also tested PamB2_2 with the nonnative (αS)-configuration that, like N-Ac-PamB2, appeared to have little inhibitory activity in this assay (Extended Data Fig. 2).

PamB2-SRCs were generated as above using 50 μM PamB2 and subjected to single-particle cryo-EM analysis. In silico sorting of the cryo-EM data revealed three main populations of ribosomal states, namely nonrotated 70S ribosomes with P-site tRNA only (15%), or with A- and P-site tRNAs (31%), as well as a population containing rotated 70S ribosomes with A/P- and P/E-hybrid site tRNAs (17 %) (Supplementary Fig. 1), which after refinement yielded final reconstructions with average resolutions of 2.4, 2.2 and 2.3 Å, respectively (Fig. 1b–e, Extended Data Fig. 3a–I, Supplementary Video 1 and Supplementary Table 1). In both reconstructions containing two tRNAs, we observed additional density located between the A- and P-site tRNAs that could be unambiguously assigned to PamB2 (Fig. 1f–i). The density of PamB2 was well-resolved, enabling the orientation of the inhibitor to be determined, and the N-terminal Agm and Hpa as well as central d-Ala, d-Orn, mDap1 and mDap2 and Gla moieties to be modeled (Fig. 1h,i and Supplementary Fig. 2). The exception was the C-terminal Spd moiety that was poorly ordered in both maps, with density observed only at low thresholds (Extended Data Fig. 3j–m). No density for PamB2 was evident in the cryo-EM reconstruction where only one tRNA (the initiator tRNA in the P-site) was present, suggesting that PamB2 may require an A-site tRNA to bind stably to the ribosome.

Interaction of PamB2 with the ribosomal P-site

The PamB2 binding site is located predominantly on the 30S subunit of the 70S ribosome, where it inserts into the cleft between the A- and P-site tRNAs (Fig. 2a, Supplementary Fig. 2 and Supplementary Video 1). Although we describe the interactions of PamB2 for the nonrotated A- and P-site tRNA state, we note that within the limits of the resolution of the reconstructions, the binding mode of PamB2 is similar, if not identical, for the rotated A/P- and P/E-hybrid state (Fig. 2b). In both states, PamB2 is oriented with the Agm side chain extending toward h24, while the central region of PamB2 runs parallel to the mRNA as well as one strand of nucleotides in h44 (Fig. 2a). The central mDap1 region of PamB2 interacts with H69 of the 23S rRNA, and then kinks such that the C-terminal (Gla-mDap2-Gly-Spd) region passes between the A- and P-site tRNAs, with the Spd moiety extending toward h31 (Fig. 2a). The kinked conformation of PamB2 is likely to be stabilized by three intramolecular hydrogen bonds (Fig. 2c), as well as two water-mediated interactions (Fig. 2d). The structural similarity with PamB2 (Extended Data Fig. 1d–i) suggests other paenilamicins (PamB1, PamA1 and PamA2) and also galantin I are likely to interact with the ribosome in the same manner.

Fig. 2: Interaction of PamB2 with the ribosomal P- and A-sites.

a, PamB2 (light blue) binding pocket located on the 30S subunit of the nonrotated PamB2–70S complex, with A-site tRNA (purple), P-site tRNA (light green), 16S rRNA (gray), 23S rRNA (yellow) and mRNA (cyan). b, Superimposition of the PamB2 binding pocket of the nonrotated (gray, with PamB2 in light blue) and rotated (light purple with PamB2 in purple) PamB2–70S complexes. c–f, Direct and water-mediated interactions (dashed yellow lines) between PamB2 and the ribosome, colored as in a. c, Direct and intramolecular interactions of PamB2 with 16S rRNA of h44 and mRNA of the P-site codon. d, Water-mediated interactions of PamB2 with h44 of the 16S rRNA, mRNA of the P-site codon and P-site tRNA. e, Direct and intramolecular interactions of PamB2 with H69 of the 23S rRNA, mRNA of the A-site codon and A-site tRNA. f, Water-mediated interactions of PamB2 with h44 of the 16S rRNA, H69 of the 23S rRNA, mRNA of the A-site codon and A-site tRNA.

In the P-site, most of the interactions of PamB2 are with 16S rRNA nucleotides (G1494-m3U1498) in h44, on one side and with the P-site codon of the mRNA on the other (Fig. 2c,d). Together with U1495, C1496 and G1497, the N-terminal amino group of Agm coordinates an ion, which we assign to a K+ ion based on the coordination distances and the presence of a K+ ion in a similar position of a previous E. coli 70S–hygromycin B structure16. We note that acetylation of the N-terminal amino group of Agm by the N-acetyltransferase PamZ4,7, or modification with N-acyl-d-Asn8, inactivates PamB2. Modeling these modified forms of PamB2 into the binding site indicates that they would clash with the surrounding 16S rRNA (Extended Data Fig. 4a–d), suggesting that these modifications would prevent PamB2 from binding to the ribosome. The binding mode of PamB2 also explains the reduction in activity of PamB2_2 since the (αS)-configuration of the N-terminal amino group of Agm would lead to loss of direct contact with the N7 of G1497, as well as the K+ ion-mediated interaction with 16S rRNA nucleotides in h44 (Extended Data Fig. 4e,f).

With regard to the P-site codon of the mRNA, there are two main points of contact, namely, the O2 of U1 (first position of the P-site codon) with the η2- and ε-nitrogens of Agm (Fig. 2c), and second, involving G3, located in the third position of the codon, where the ribose O2′ and N2 can form hydrogen bonds with γ-nitrogen and carbonyl–oxygen of mDap2 of PamB2 (Fig. 2c). In addition, water molecules (W2 and W8) mediate interactions between the backbone of U2 and Hpa of PamB2, as well as the O4′ (ribose) of G3 with the carbonyl–oxygen of Gla of PamB2 (Fig. 2d). Although PamB2 approaches the P-site tRNA, there is relatively little direct interaction, with the closest point of contact being 3.6 Å between the η2-nitrogen of Agm and the ribose O2′ of A37 of the P-site tRNA. However, we do observe a water-mediated (W9) interaction between the carbonyl–oxygen of Gla of PamB2 and the O2′ and N3 of A35 of the P-site tRNA (Fig. 2d).

Interaction of PamB2 with the ribosomal A-site

In the A-site, PamB2 contacts not only with 16S rRNA nucleotides in h44, but also A1913 from H69 of the 23S rRNA, and, in contrast to the P-site, the compound makes extensive interactions with the A-site tRNA, albeit less with the mRNA codon (Fig. 2c–f). Interactions of PamB2 with the A-site tRNA revolve around nucleotides ct6A37 and A38, which are located in the anticodon-stem loop, directly adjacent to the anticodon (34CAU36) (Fig. 2e,f). Specifically, three direct hydrogen bonds are possible with A38 (Fig. 2e). Interactions with ct6A37 are indirect, being mediated by water W6, which is coordinated by the carbonyl–oxygen of D-Orn as well as the O2′ and N3 of ct6A37 (Fig. 2f). Interaction of PamB2 with the A-site codon of the mRNA is restricted to a direct interaction of backbone amide of Gly and a backbone oxygen of A4, which is located in the first position of the A-site codon, and a water-mediated interaction from the backbone amide of mDap2 via water W10 with N7 (3.4 Å) of A4 (Fig. 2f). The interactions between PamB2 and the A-site tRNA are likely to be critical for binding of PamB2 to the ribosome, since we observe no density for PamB2 in the P-tRNA-only state. We note that in structures of 70S ribosomes lacking A-tRNA17, the conformation of A1913 in H69 differs from that when the A-site tRNA is present, such that it would be incompatible with the interactions observed for PamB2 on the elongating ribosome (Extended Data Fig. 4g–j). The A1913 conformational shift induced by A-tRNA binding may therefore contribute to preventing stable binding of PamB2. Although a shift in the position of A1913 occurs during decoding when the A-site tRNA is still bound to EF-Tu, the A-tRNA itself is still suboptimally placed to interact with PamB2 in such a state18 (Extended Data Fig. 4k,l), suggesting that full accommodation of A-tRNA is required for stable interaction of PamB2 with the ribosome. We note that the PamB2 binding site is conserved on eukaryotic ribosomes (Extended Data Fig. 5a), and could accordingly demonstrate that PamB2 efficiently inhibits eukaryotic in vitro translation (Extended Data Fig. 5b), consistent with its antifungal activity4. However, PamB2 displays no cytotoxicity against eukaryotic cell lines (Extended Data Fig. 5c), suggesting that the compound is not internalized.

PamB2 inhibits tRNA2–mRNA translocation

Careful examination of the tRNAs in the PamB2-bound elongation complexes revealed the presence of additional density attached to the CCA-end of the A-site tRNA in the nonrotated elongation state and to the A/P-tRNA in the rotated hybrid state, indicating that peptide bond formation has already occurred in these complexes (Supplementary Fig. 3a,b). This suggests that PamB2 does not interfere with the decoding and accommodation by the A-tRNA, nor peptide bond formation, and also allows the ribosome to oscillate between the canonical and hybrid pretranslocation states (Fig. 3a). During normal translation, elongation factor EF-G binds and translocates the tRNA2–mRNA complex into the P- and E-sites, forming a posttranslocational state of the ribosome19,20,21. The accumulation of pretranslocational states in the presence of PamB2, as well as the absence of posttranslocation states (Fig. 1b–e and Extended Data Fig. 6a–c), suggests that PamB2 may interfere with the process of translocation. To directly assess this, we analyzed the effect of PamB2 on EF-G-dependent translocation using the toeprinting assay. Ribosome complexes were formed with tRNAfMet in the P-site and N-AcPhe-tRNAPhe in the A-site, with toeprinting revealing a band corresponding to the expected pretranslocation state in the absence of EF-G (Fig. 3b). On addition of EF-G, but in the absence of PamB2, the toeprinting band shifted by three nucleotides, indicating that the A- and P-site tRNAs were translocated to the P- and E-sites (Fig. 3b). In contrast, little to no shift in the toeprint band was observed when the same reactions were performed in the presence of PamB2 or the control antibiotic negamycin, which was reported previously to interfere with translocation22. Hence, we conclude that binding of PamB2 to ribosome inhibits the process of translocation (Fig. 3b). Comparison with recent structures of EF-G-bound translocation intermediates provides a structural rationale for the PamB2-mediated translocation inhibition19,20,21. While the initial binding of EF-G to the ribosome may be possible in the presence of PamB2 (Extended Data Fig. 6d–f)20,21, the subsequent steps where EF-G accommodates, releases the decoding center23 and promotes a shift of the anticodon stem of the A/P-site tRNA that would lead to severe clashes with PamB2 (Fig. 3c–e and Extended Data Fig. 6g–i)19,20,21. Moreover, in the early translocation intermediate with EF-G, A1913 rotates away from its position in the hybrid states19,20,21, which would require disruption of interactions between A1913 and PamB2 (Extended Data Fig. 6j–l). Collectively, this leads us to suggest that the interactions of PamB2 with the anticodon-stem loop of the A-site tRNA, as well as with the extended conformation of A1913, would interfere with productive translocation and thereby inhibit protein synthesis.

Fig. 3: PamB2 inhibits tRNA2–mRNA translocation.

a, Superimposition of PamB2 and tRNAs of the nonrotated (light blue) and rotated (purple) PamB2–70S complexes. b, Toeprinting assay monitoring the effect of PamB2 on EF-G dependent translocation, with initiator tRNAfMet and N-AcPhe-tRNAPhe in the absence of drugs and in the presence of the translocation inhibitor negamycin22. Toeprinting assays were performed in duplicate, with the duplicate gel shown in the source data. c, Superimposition of PamB2 and hybrid tRNAs of the rotated (purple) PamB2–70S complex with hybrid tRNAs and EF-G bound to the E. coli 70S ribosome in the Int1 state (salmon, PDB ID 7N2V)19. d,e, Sphere representation of the hybrid A/P-tRNA anticodon-stem loop of the rotated PamB2 complex sterically clashing with EF-G (salmon, PDB ID 7N2V)19 (d) and of the hybrid A/P- and P/E-tRNAs of the Int1 state (PDB ID 7N2V)19 (e) clashing with PamB2 (purple). Steric clashes are highlighted with red lines.

Source data

Influence of A-site mRNA context on PamB2 inhibition

While the translocation assay (Fig. 3b) and structures of PamB2 bound to pretranslocation complexes (Fig. 1b–e) support the conclusion that PamB2 interferes with the EF-G mediated translocation process, our initial toeprinting assay indicated that it was not the first translocation step that was inhibited, but rather the second (Extended Data Fig. 2). If the first translocation reaction was inhibited, then ribosomes would be trapped with the AUG start codon in the P-site being decoded by the initiator tRNAfMet and with a UUG codon in the A-site. While the density indicates that the initiator tRNAfMet is present in the P-site of the P-tRNA-only volume (Supplementary Fig. 4a), the density for the mRNA codons and tRNAs in the A- and P-sites in structures of the PamB2-bound pretranslocation states indicated that one round of translocation had occurred before stalling of the complex (Supplementary Fig. 4b–e), that is UUG and AUA codons are in the P- and A-sites being decoded by tRNALeu (anticodon 5′-35CAA37-3′) and tRNAIle (anticodon 5′-35CAU37-3′), respectively (Supplementary Fig. 4d,e). Moreover, we observe extra density for the 2-methylthio-N6-isopentenyladenine (ms2i6A) at position 37 of the P-site tRNALeu as well as the cyclic N6-threonylcarbamoyladenosine (ct6A) at position 37 of the A-site tRNAIle (Supplementary Fig. 4f–i). Collectively, these findings suggest that PamB2 allowed the first translocation on the MLIFstop-mRNA, but prevented the second translocation reaction from taking place.

To investigate whether it is the initiation context that interferes with the action of PamB2 or whether PamB2 acts in a sequence context-specific manner, we generated a series of model mRNA templates encoding a short ErmBL protein24,25 with 1–5 repeats of the UUG codon directly following the AUG start codon (Fig. 4a). In the absence of antibiotic, ribosomes initiate on the AUG start codon of the wildtype ErmBL mRNA (with a single UUG codon), and translate uninterrupted to the twelfth codon (AUC encoding Ile), where they become trapped due to the presence of the Ile-tRNA synthetase inhibitor mupirocin that was added to all reactions (Fig. 4a). Addition to the control reaction of the antibiotic retapamulin traps ribosomes on the AUG start codon26, whereas the macrolide erythromycin leads to the accumulation of ribosomes stalled with the tenth CAU codon (encoding Asp) in the P-site (Fig. 4a), as we observed previously on the ErmBL mRNA24,25. Unlike retapamulin, the presence of PamB2 did not lead to a strong accumulation of ribosomes on the AUG start codon of the wildtype (1×UUG) ErmBL mRNA template, but rather ribosomes became stalled only once the second ermBL codon (UUG) moved into the P-site (Fig. 4a), as we observed for the MLIFstop-mRNA (Extended Data Fig. 2). While the insertion of UUG codons into the ErmBL mRNA shifted the band for initiating ribosomes upward in the gel as expected, the first main stalling bands remained constant (Fig. 4a), indicating that in the presence of PamB2, ribosomes can translate through stretches of up to five UUG codons unhindered. We conclude therefore that the lack of effect of PamB2 on the first translocation event in the wildtype ErmBL mRNA is not related to the initiation context, but rather the presence of the UUG codon in the A-site. We also note that unlike for the short MLIFstop-mRNA, additional bands were observed on the longer ErmBL mRNA indicating that a subset of ribosomes also become stalled at subsequent sites in the mRNA, for example, with the fourth UUC (encoding Phe) in the P-site, but not the third GUA codon in the P-site (Fig. 4a).

Fig. 4: Influence of A-site mRNA context on PamB2 inhibition.

a,b, Toeprinting assays monitoring the position of ribosomes on the wildtype ErmBL mRNA in the presence of water, 50 or 100 µM retapamulin (Ret), 50 µM erythromycin (Ery) and an ErmBL mRNA with an increasing number of UUG repetitions in the presence of 100 µM PamB2 (a) and with an ErmBL mRNA with the second codon mutated to UUG, AUG, CUG, GUG (orange) in the presence of 50 µM PamB2 (b). Arrows indicate the stalling sites on the isoleucine catch codon in the presence of mupirocin (pink), at initiation (green), on the erythromycin-ErmBL stalling site (purple) and stalling induced by PamB2 (blue). Toeprinting assays were performed in duplicate, with the duplicate gel shown in the Source Data. c–f, Water (red) mediated interaction (dashed line) of PamB2 (blue) and the first nucleotide of the mRNA of the A-site codon (cyan), and superimposed with in silico mutated first position of the A-site codon (orange) to guanine (d), cytosine (e) or uracil (f). The loss of the water-mediated interaction is indicated by a red cross.

Source data

An initial examination of the nonstalling contexts revealed that that the presence of U in the first position of the A-site codon antagonizes PamB2 action. Therefore, to test whether the nature of the A-site codon can influence the efficiency of PamB2-mediated translocation inhibition, we mutated the U in the first position of the A-site codon of the ErmBL mRNA to A, C and G (Fig. 4b). In contrast to U in the first position where little to no inhibition of the first translocation event was observed (Fig. 4b), clear toeprint bands were observed with each of the other nucleotides, indicating that ribosomes accumulate with the AUG start codon in the P-site when the A-site codon was changed from UUG to AUG, CUG or GUG (Fig. 4b). Although the inhibition by PamB2 with C in the first position appeared to be stronger than with U, it was reproducibly weaker than with A and G (Fig. 4b). The inhibition with G in the first position of the A-site codon appeared to be the strongest (Fig. 4b). Although PamB2 does not directly interact with the first position of the A-site codon, we observed that a water-mediated interaction is present between the backbone amide of mDap2 via water W10 with the N7 of A in the first position of the A-site codon in the ribosome stalled on the MLIFstop-mRNA (Fig. 4c). A similar interaction would be maintained with a G in the first position (Fig. 4d) where we observe strong inhibition (Fig. 4b), but would not be possible with C or U (Fig. 4e,f) where inhibition was weaker (Fig. 4b).

Influence of A37 modification of A-tRNA on PamB2 inhibition

While the water-mediated interaction between PamB2 and the first position of the A-site codon may contribute to the specificity of stalling of PamB2, it does not rationalize the difference in efficiency of inhibition of PamB2 that we observed between U and C in the first codon position (Fig. 4b). Therefore, we considered whether the nature of the tRNA in the A-site may also contribute, especially given that we noticed interactions between PamB2 and nucleotides A37 and A38 of the A-site tRNA (Fig. 2e,f). Since we observe no inhibition by PamB2 when Phe-tRNA decodes UUC, we superimposed a ribosome structure with Phe-tRNA in the A-site27 and immediately noticed that tRNAPhe bears a 2-methylthio-N6-isopentenyladenine (ms2i6A) at position 37, with the 2-methylthio moiety encroaching on the PamB2 binding site (Fig. 5a,b). In fact, with one exception (see later), all tRNAs that decode mRNA codons beginning with U have ms2i6A37, which is proposed to help stabilize the weaker U–C codon–anticodon interaction between the mRNA and the tRNA28. Consistently, the ribosome is not inhibited by PamB2 when tRNALeu (with ms2i6A37) decodes UUG in the A-site (Fig. 4a), and similar results would be expected for tRNASer decoding UCU/UCA/UCG, tRNATyr decoding UAU/UAC, tRNACys decoding UGU/UGC and tRNATrp decoding UGG. The one exception is tRNASer that decodes UCU and UCC but has A37 unmodified28. To directly test this, we generated a series of mRNA templates based on the ErmBL-(UUG)4 mRNA where we changed the seventh GUA (Val) codon to each of the four serine codons UCC, UCU, UCA and UCG and performed the toeprinting assay in the presence of PamB2 (Fig. 5c). As hypothesized, strong stalling was observed at the UCC and UCU codons, which are decoded by the tRNASer isoacceptor lacking any modification at position A37, whereas only weak stalling was observed at the UCA and UCG codons, which are decoded by the tRNASer isoacceptor bearing ms2i6A37 (Fig. 5c). Thus, we conclude that PamB2 is a poor inhibitor of translocation when it has to compete with the A-site tRNA containing ms2i6A37.

Fig. 5: Influence of A37 modification of A-tRNA on PamB2 inhibition.

a, PamB2 (light blue) and the modified A-site tRNA residue cyclic N6-threonylcarbamoyladenine (ct6) in position 37 (purple) from the nonrotated PamB2 complex superimposed with an in silico model of an unmodified A37 (yellow). b, Superimposition of PamB2 from a with the 2-methylthio-N6-isopentenyladenine (ms2i6, light orange) at position 37 of the A-site tRNAPhe from the T. thermophilus 70S ribosome preattack state (PDB ID 1VY5)27 shown as sphere representation with clashes indicated by red lines. c, Toeprinting assay monitoring the position of ribosomes on the (UUG)4-ErmBL mRNA in the presence of water (–), 50 µM retapamulin (Ret, green) and 100 µM PamB2 (light blue). The seventh codon was modified to different serine codons (orange). Arrows indicate stalling for the isoleucine catch codon in the presence of mupirocin (pink), the initiation (green) and PamB2-induced stalling (light blue). Toeprinting assays were performed in duplicate, with the duplicate gel present in the Source Data. d,e, Superimposition of PamB2 from a with 1-methyl-guanine (m1G, dark orange) at position 37 of the A-site tRNAPro on the T. thermophilus 70S ribosome (PDB ID 6NUO)41 (d) and an in silico modified 2-methyl-adenine (m2A, yellow) shown as sphere representation with steric clashes indicated by red lines (e).

Source data

Although PamB2 inhibited translation when tRNALeu decoded CUG in the A-site, the extent of inhibition was relatively weak (Fig. 4a). Therefore, we superimposed a ribosome structure with tRNALeu in the A-site17 and recognized that the m1G at position 37 of tRNALeu would clash with PamB2 due to a steric hindrance between the N2 group of m1G37 of the A-tRNA and the d-Orn of PamB2 (Fig. 5d). In fact, most tRNAs decoding CNN codons, including CUN by tRNALeu, CCN by tRNAPro as well as CGG by tRNAArg contain m1G37 (Supplementary Fig. 5)28. The exceptions are tRNAHis and tRNAGln that decode CAU/C and CAA/G, as well as tRNAArg that decodes CGU/C/A, however, all of these tRNAs have m2A37 (ref. 28) (Supplementary Fig. 5) that would also be predicted to clash with the d-Orn of PamB2, similar to m1G37 (Fig. 5d,e and Supplementary Fig. 5). Collectively, we conclude that the efficiency of translocation inhibition by PamB2 is directly influenced by the nature of the A-site tRNA and in particular by modifications at position A37, such as m1G37, possibly m2A37 and especially ms2i6A37 (Supplementary Fig. 5) where the steric overlap with the drug is largest.