The mechanism of the phage encoded protein antibiotic from ΦX174

Abstract

The historically important phage ΦX174 kills its host bacteria by encoding a 91-residue protein antibiotic called protein E. Using single particle electron cryo-microscopy, we demonstrate that protein E bridges two bacterial proteins to form the transmembrane YES complex (MraY, protein E, SlyD). Protein E inhibits peptidoglycan biosynthesis by obstructing the MraY active site leading to loss of lipid I production. We experimentally validate this result for two different viral species, providing a clear model for bacterial lysis and unifying previous experimental data. Additionally, we characterize the first MraY structure from a human pathogen, revealing new features of this essential enzyme, and the first structure of SlyD bound to a protein. These results provide a route towards new antibacterial therapies.

One-Sentence Summary:

The YES structure reveals the mechanism of bacterial cell death utilized by phage ΦX174 to escape its host.

The full realization of phage therapy as a solution to the anti-microbial resistance problem requires a fundamental understanding of the mechanisms viruses use to kill their host (1–4). Nearly 100 years ago, the first medical application of phage to treat infections used a cocktail that included the historically important phage ΦX174 (5). A rich source of critical discoveries in molecular biology (6–9), ΦX174 is a member of the Bullavirinae subfamily within Microviridae. It is found broadly in environments that contain coliform bacteria, such as the human gut (10), with E. coli as its primary host. The 1977 publication of the ΦX174 single-stranded 5.4-kilobase DNA genome was a milestone for genomics revealing 11 open reading frames (ORF) (11). Most notable was the gene for lysis, E, that is embedded within the ORF of the scaffolding gene D (12). Expression of E alone is sufficient to kill bacteria (13). While a variety of lysis mechanisms have been proposed (14–16), the most likely is that the product of E, protein E, disrupts peptidoglycan (PG) biosynthesis, culminating in a breach in the cell wall. In particular, protein E was demonstrated to directly inhibit the integral membrane enzyme MraY (16) dependent on the cytoplasmic chaperone SlyD (sensitivity to lysis D) (17, 18).

Phage-derived single gene lysis (SGL) proteins, such as protein E, trigger cell lysis by inhibiting PG biosynthesis (reviewed in (19)) providing novel routes for killing bacteria. Protein E includes a conserved N-terminal transmembrane domain (TMD), a cytoplasmic C-terminal domain (CTD) with conserved positively charged residues in a predicted amphipathic helix, and an unstructured tail (Fig. 1A and fig. S1). Fusions of globular proteins to the C-terminus of the TMD have demonstrated that the TMD is sufficient for lysis (17, 20, 21). Extensive mutational analysis identified key residues in protein E including the essential proline 21 (P21) (17, 20–23). The mechanism through which SlyD and MraY work with protein E to facilitate rupture of the cell wall remains unknown. MraY, an important target for antimicrobial drug discovery (reviewed in (24), catalyzes the transfer of a phospho-MurNAc-pentapeptide from ‘Park’s nucleotide’, UDP-MurNAc-pentapeptide, onto the phosphate group of the 55 carbon polyisoprenyl-phosphate (C₅₅P) generating lipid I. X-ray crystal structures of MraY in detergent from the thermophilic Gram-negative Aquifex aeolicus (AaMraY) or the Gram-positive Enterocloster bolteae (EbMraY) revealed a conserved homodimeric structure, with ten transmembrane domains (TMDs) per subunit. The structures localized the active site in a vestibule on the cytoplasmic side of a membrane-exposed cleft formed by TMDs 4, 5, and 9, which is the predicted access site for the lipid substrate (28). The constitutively expressed but non-essential (29) metallochaperone (30) SlyD contains two conserved core domains: an FK506-binding protein (FKBP) prolyl-isomerase (PPI) domain and an insertion-in-flap (IF) domain that binds unfolded peptides by β-augmentation (18, 31–33). The disordered SlyD C-terminus (31), while important for metal binding, can be deleted without affecting chaperone activity (34).

Fig. 1. The structure of the YES complex.

(A) An alignment of a representative subset of protein E isoforms. Residue coloring is based on ClustalW. Secondary structure elements are shown above the sequence. Sequences are ordered as in fig. S1. Residues highlighted with a (*) at the bottom of the alignment are discussed in the text. (B) SDS-PAGE gel of the purified YES_ID21 complex. (C) A lysis assay for protein E expression. Cells containing either empty vector (black) or the protein E genes for ID21 (pink) or ΦX174 (green) were induced at time 0 and the absorbance at 600 nm was monitored over time. Error bars represent the standard deviation derived from N=3. (D) Overview density maps of the YES_ID21 complex viewed in the plane of the membrane. The map in grey highlights the detergent micelle. The higher contoured map shows the six components of the two-fold complex with density for E. coli MraY (cyan), protein E (yellow), and E. coli SlyD (purple) highlighted. The pairs of each protein are distinguished by the B-subunit colored lighter and the general coloring scheme is maintained throughout the figures. Density that is likely lipid is shown in transparent orange. (E) Cartoon representation of the YES_ID21 complex oriented and colored as in (D). Foreground TMDs for the MraYs are numbered and the bilayer is represented by lines. N-terminus and C-terminus of protein E_A are labeled.

In the present study, we resolve the mechanism of peptidoglycan biosynthesis inhibition by protein E. We determined the structure of the dimeric heterotrimer YES complex (EcMraY, viral protein E, EcSlyD) revealing that protein E physically blocks access of the lipid substrate to the MraY active site. This work provides mechanistic insight into all three proteins and suggests a path towards development of antibacterial agents.

The structure of the YES complex

We performed the following studies using the protein E sequence from either the original phage ΦX174 or the shorter protein E isoform from phage ID21 (91 and 76 residues respectively), both from within Bullavirinae (Fig. 1A and fig. S1). We first established that protein E and MraY formed a stable complex by co-expressing an affinity tagged protein E_ΦX174 and wild-type EcMraY. After purification, we identified a stable complex that showed a single peak in size-exclusion chromatography (SEC), but resolved into four bands on a gel, two of which corresponded to the endogenous EcSlyD (fig. S2). A SlyD variant with the disordered C-terminus removed rescued lysis activity in an E. coli ΔslyD strain (fig. S2) and ran as a single band on a gel (Fig. 1B)(18, 30). This truncation, SlyD₁₅₄, was used for subsequent work, except where noted.

The two protein E isoforms induced lysis at similar efficiencies when expressed in a wild-type E. coli strain (Fig. 1C). For purification, all three genes in the YES complex (wild-type EcMraY, protein E, and truncated EcSlyD) were recombinantly expressed together in the ΔslyD strain with a C-terminal affinity tag on protein E. The complex was extracted in detergent and ran as a single peak by SEC with all three proteins in an apparent stoichiometric complex (Fig. 1B). Structures for both the YES_ΦX174 and YES_ID21 complexes were solved using single particle electron cryo-microscopy (Fig. 1D and figs. S3 and S4). The final density maps were obtained following several rounds of data processing with heterogeneous and homogeneous refinements (figs. S3 and S4). The final overall masked resolution was 3.5 Å for the YES_ID21 complex and 3.6 Å for the YES_ΦX174 complex (fig. S5). Statistics are provided in table S2. For both structures, the resolution was higher for regions in and adjacent to the membrane (figs. S3 and S4). We unambiguously built 90% of the protein residues in the complex. Sequence differences between the protein E variants are visible in the density (fig. S5), generally exposed on the surface of the complex. The YES_ID21 complex is used as the reference structure, except where noted.

Within the density map, we could clearly distinguish two copies of each member of the YES complex (six separate proteins). When contoured to remove the detergent micelle, densities for 22 TMDs are clear, 20 of which are accounted for by the EcMraY dimer (Fig. 1D,E). The majority of the cytoplasmic density can be accounted for by two SlyD molecules, which is the most flexible region of the complex (fig. S6D). The remaining protein density corresponds to two protein E molecules that each contain a kinked TMD and a soluble domain that bridges between MraY and SlyD. Densities for lipids are visible around the membrane exposed surface of the MraY dimer (Fig. 1D).

For protein E, starting at the N-terminus in the periplasm, the TMD binds in the groove formed between TM5 and TM9 of MraY ending on the cytoplasmic side where it makes a sharp turn into the active-site pocket (Figs. 1E and 2A). The TMD is followed by an amphipathic helix that crosses the active-site, parallel to the membrane, presenting a positive face toward MraY and a hydrophobic face towards a SlyD IF-domain. The C-terminal residues adopt an extended conformation that primarily interacts with the second SlyD. This results in a cross over point between the two protein Es with each contacting both SlyDs. Overall, the dimeric complex (two of each of the heterotrimers) has a near two-fold symmetry perpendicular to the plane of the membrane. While the membrane and periplasmic facing regions overlay perfectly, the symmetry is broken at the cytoplasmic face where the C-termini of the two protein E molecules cross each other at different residues and the SlyDs adopt slightly different orientations (Fig. 1E and fig. S6). Following towards the end of protein E, we can see continuous backbone density that positions proline 65 in the active site of the FKBP domain. Beyond that, the density is insufficient to resolve the sequence and we see little difference between ΦX174 and ID21 (fig. S5).

Fig. 2. The interaction of protein E with EcMraY.

(A) Cartoon of the YES_ID21 complex as in Fig. 1E with a 90° rotation except that MraY_A is color-ramped using the Viridis palette that goes from purple to yellow for N- to C-terminus. Prolines are shown in red. The bilayer is modeled in white. The B-subunits and SlyDs are faded. Boxes indicate regions highlighted in panels (B-D). Side chains for protein E residues that contact MraY are shown as sticks. (B) As in (A) with positions of protein E resistance mutants in MraY highlighted (dark cyan). (C) As in (A) with residues at the interface highlighted as sticks. (D) The region where the two protein E molecules cross highlighting the asymmetry colored as in (A) with interacting residues shown as sticks. Density for the amphipathic helix of protein E shown as a blue mesh. SlyDs are removed for clarity. (E) Lysis assay of expressed protein E_ΦX174 variants. Error bars represent the standard deviation derived from N=3. (F) Similar to (D) for protein E_ΦX174.

The interaction of protein E with MraY

Functional studies have consistently revealed the requirement for a proline at position 21 (21, 22, 35). Our structure allows an elegant explanation for this requirement. The protein E TMD binds in the cleft of MraY that is defined by the angled TM9b (Fig. 2A). The proline at position 21 breaks the hydrogen bonding introducing a kink that allows bending of the TMD around TM9b following the groove to the active site. Mutation of P21 would result in loss of the kink favoring a straight TMD that could not bind in the groove. Residue P29, also completely conserved (Fig. 1A and fig. S1), creates a second bend that completes the wrap around TM9b and a mutation at this position results in delayed lysis onset (21, 22). Additional alanine mutations in the TMD identified residues that result in delayed lysis onset, postulated to decrease the binding affinity of protein E to MraY (21). In the structure, most of these residues (L19, L20, L23, & M26) (Fig. 2C) make direct contact with MraY. The exception, F27, appears to sterically position M26 into a tight interaction with Y134 in MraY, which is conserved in most Gram-negative bacteria (fig. S7).

At the cytoplasmic interface, protein E residues A36 through M50 form the amphipathic helix spanning the width of an MraY subunit. The hydrophilic face of this helix orients toward the membrane in the MraY active site. The helix contains conserved positively charged residues that interact with conserved negatively charged residues in MraY (Fig. 2D). An example is the K46 salt bridge where, in our lysis assay, a K46A mutation resulted in delayed lysis onset (Fig. 2E). Here, isoform differences can be visualized, for example position 42 is a leucine in ΦX174 while in ID21 it is an arginine that forms a salt bridge with E335 in MraY (Fig. 1). The helix ends at the crossover between the two protein Es at residue V54 in protein E_A and A51 in protein E_B resulting in differing interactions with residues in MraY, such as the essential H326 (Fig. 2D).

The Epos (plaques on ΔslyD) mutations, R3H and L19F, allow phage propagation in a ΔslyD background (23). The R3H mutant of protein E in ΦX174 results in a silent mutation in protein D and is found native to other species such as ID21 (fig. S8). Previous work reported that this variant resulted in higher levels of protein E in the membrane (23). In the structure (fig. S5G,H), this residue does not make specific contacts to MraY. It is likely that the loss of a positive charge in the periplasm favors a higher percentage of protein E molecules that are correctly inserted in the membrane due to the positive-inside rule. L19F does not result in higher levels of protein E (21) and it likely hastens lysis onset by having a higher affinity to MraY. Another phenylalanine mutation at the interface, L23F, also hastens lysis onset likely by higher affinity; although this is not a general rule as other leucine to phenylalanine mutants did not affect lysis (21). Both L19 and L23 are near the conserved F182 in MraY and may add additional stability through aromatic π interactions (Fig. 2C). The opposite mutation, phenylalanine to leucine, can show loss of binding. For example, the F288L mutation in MraY (16) is located at the interface with protein E and results in a loss of lysis, most likely due to lower affinity.

All the mutations in MraY that allow resistance to protein E mediated lysis (P170L, ΔL172, G186S, F288L, V291M) (16, 35) are at the interface with protein E (Fig. 2B). The mutant F288L, as noted above, lowers affinity to protein E. Residue G186 is at the nearest approach between the two proteins and a mutation to serine would prevent protein E binding. The mutant V291M lies directly at the interface near L19 in protein E; although a specific effect for this mutation is not clear. Finally, P170L and ΔL172 are located within the periplasmic loop 4–5 that interacts with the conserved N-terminus of protein E. Although more resistant to lysis, these two mutants are predicted to still bind protein E, albeit with lower affinity (35).

The mechanism of inhibition of MraY by protein E

The YES complex structure allows us to propose a simple mechanism for inhibition of MraY by protein E. Superposition with previous MraY inhibitor complexes (26, 27) on the YES complex (Fig. 3) shows the predicted path of the polyisoprenyl chain of C₅₅P is the groove formed between TM5 and TM9, which is occluded by protein E. Therefore, one mechanism of inhibition is that protein E prevents access of the lipid substrate to the active site (Fig. 3A). Protein E is a noncompetitive inhibitor of Park’s nucleotide (37) and, consistent with this, the pocket that binds the nucleoside is fully accessible in the structure (Fig. 3B). Loop 9–10 in MraY contains catalytic histidines that must move toward the binding pocket to facilitate catalysis by completing the active site (28). The cytoplasmic helix of protein E separates loop 9–10 from the rest of the active site, blocking this transition and providing a second mechanism of inhibition. Overall, protein E blocks access of the lipid substrate and prevents formation of the active site upon substrate binding.

Fig. 3. The mechanism of inhibition by protein E.

(A) Accessible surface of MraY (colored as in Fig. 2A) viewed from the cytoplasm looking towards the active site cleft. Protein E is shown in cartoon. The structures for the inhibitors tunicamycin (gray) and carbacaprazamycin (pink) are shown as sticks based on a structural alignment to EcMraY_A with their respective complex structures (PDBID: 6OYH (26) & 5JNQ (27)). (B) Similar to (A) from a slightly different angle highlighting the catalytic pocket. MraY (dark cyan) is shown in cartoon. The two substrate binding sites are highlighted by dashed boxes. Predicted catalytic residues in MraY are shown as sticks.

Key structural features of MraY

The YES complex contains the first E. coli MraY structure (fig. S9) which is also the first from either an important human pathogen or model organism. Overall, comparing MraY from the YES complex to those solved by crystallography, there is significant agreement when comparing monomers, with backbone RMSDs around 1Å (Fig. 4A and fig. S10). Our cryo-EM structure of EcMraY reveals additional regions of MraY previously disordered in the crystal structures. Most notably, we resolve all the cytoplasmic loops that enclose the active site. We model loop 1–2, which likely adopts distinct conformations during the catalytic cycle and indeed has two slightly different conformations in our dimer (fig. S6C). In this structure, loop 9–10 adopts a conformation that is similar to the reported small-molecule inhibitor co-crystal structures of MraY (26, 27, 38), but distinct from the apoprotein structure (Fig. 4 and fig. S10). This conformation of loop 9–10 may be a general feature of MraY complexed with inhibitors.

Fig. 4. New features observed in the EM structure of E. coli MraY.

(A) Cytoplasmic view of a structural alignment of EcMraY (dark cyan) against uninhibited AaMraY (green, PDBID:4J72 (25)) and carbacaprazamycin inhibited AaMraY (pink, PDBID:6OYH (26)). RMSDs to monomer A of EcMraY are shown. The color scheme for the various MraY crystal structures is used throughout the figures. (B) A view in the plane of the membrane showing the region that includes TM1 and TM2 in backbone ribbons. Left, the EcMraY structure colored from the N-terminus through TM2 in the Viridis color scheme. Right, the inhibited AaMraY structure from (A) and the EbMraY structure (gray, PDBID:5JNQ (27)). Each structure is aligned to EcMraY. The location of each N-terminus is indicated by an asterisk. (C) Accessible surface of EcMraY (cyan) and unmodeled densities (orange/purple) that are likely lipids or detergent. The inset is a view of the periplasmic cavity viewed from a removed monomer.

An unexpected structural feature in EcMraY occurs at the N-terminus. In the crystal structures, the N-terminus begins at either TM1 or is a helix that projects away from the structure in an orientation incompatible with the bilayer (Fig. 4B). In the YES complex, the N-terminal end of the first helix (NTH) hydrogen-bonds to the C-terminal end of TM2, effectively forming a helical stacking structure (Fig. 4B and fig. S9C). This is a unique structural feature that, to our knowledge, has not been observed in a protein structure before. A multiple sequence alignment across bacteria shows that this feature of MraY is conserved across Gram-negative bacteria but missing in Gram-positives (figs. S7 and S11). While this feature is not found in the crystal structures, AlphaFold (39) predicts the NTH stacking for E. coli and other Gram-negative bacteria, although the hydrogen bonding and orientation is slightly different from the cryo-EM structure (fig. S11). For the AaMraY structures (25, 26), the positioning of the N-terminus is likely a product of crystallization, as AlphaFold predicts the NTH stacks for this and the related Hydrogenivirga species (fig. S11). For Gram-positive bacteria, both the EbMraY and predicted structures lack the NTH (27)(Fig. 4B and fig. S11).

Protein E as a general antibacterial protein

While phage ΦX174 is restricted to E. coli, evidence suggests that protein E has broad spectrum activity against Gram-negative bacteria. Expression of protein E leads to lysis that results in a subset of killed bacteria becoming ‘ghosts’, essentially empty cell walls (15). Ghosts have been used in a variety of contexts including vaccine development (reviewed in (40)). For ghosts to be made, protein E must inhibit the native MraY of the target bacteria. Towards this end, many bacteria have been probed for ghost formation by expression of protein E. All Gram-negative bacteria tested (fig. S11A) resulted in the formation of ghosts (reviewed in (41)). On the other hand, protein E was unable to cause lysis in the Gram-positive Staphylococcus carnosus (42). Expression of the Bacillus subtilis MraY in E. coli prevented cell lysis, suggesting that it has a low affinity for protein E (35). Comparing the residues in EcMraY that contact protein E to those in Gram-positive species, there are no single residue differences that easily explain the inability to inhibit MraY from Gram-positive species. There are a few residues that may play a role, e.g. P170 in EcMraY, which confers resistance when mutated to leucine, is absent in Gram-positive MraYs (fig. S7). EcMraY residue Y134, which forms a bridge with protein E M26 in our model (Fig. 2C), is also missing in Gram-positive species (fig. S7). Altogether, these results suggest broad activity against Gram-negative bacteria and sequence differences in Gram-positive MraYs that prevent protein E binding.

Lipids bound to the YES complex

The YES complex was solubilized from its native environment and, for the YES_ID21 complex, the final detergent solution was supplemented with E. coli lipids. We observe lipid densities around the membrane surface of MraY (Fig. 1D and fig. S12A,B). Evidence supports a functional role for anionic phospholipids with MraY including stabilizing the dimer (43–45). In support of this, we observe substantial lipid density near the dimer interface most with features consistent with glycerophospholipids, yet there are a few exceptions (fig. S12). As seen in the previously reported structures (25, 27), there is a hydrophobic periplasmic cavity within the MraY dimer that contains unexplained density. Here, while the density has clear structure, we are unable to fit typical E. coli phospholipids (Fig. 4C).

Recently, a study using native mass spectrometry and molecular dynamics identified a binding site for C₅₅P at the dimer interface with the phosphate headgroup predicted to form a salt-bridge to R341 in EcMraY (45). In our structure, we observed a long tube of density at the dimer interface (fig. S12A, purple) that ends at R341 and is consistent with C₅₅P (fig. S12C). While we cannot confidently identify this lipid, it supports that C₅₅P binds in this position, although the role for this binding site remains to be determined. Overall, the structure provides further context for the functional importance of various lipids.

The role of SlyD in protein E mediated lysis

The amphipathic helix of protein E bridges the two E. coli proteins, MraY and SlyD, which make no specific contacts to each other (Fig. 2A). The IF domain sits on the hydrophobic face of the protein E helix (Fig. 5A) and contacts the extended C-terminus of the opposing protein E which then extends to bind the FKBP domain (Fig. 5B). This results in a bowtie like interaction with each SlyD binding both protein E soluble domains (Fig. 1E). There are no contacts between the two SlyD molecules (fig. S6D). These protein E interfaces validate structural studies of SlyD where peptides bind to each of the interfaces seen here including β-augmentation in a groove in the IF domain (Fig. 5C,D) (32). This is the first evidence of the IF domain binding to an α-helix, which may be an important chaperoning interaction. The FKBP domain is well-ordered but adopts a range of orientations in our particles (fig. S13) and, consequently, has the lowest resolution (fig. S3). This flexibility relative to the IF domain is consistent with the NMR structure of SlyD where the orientations of the two-domains are not constrained by each other (31). The only contact of the FKBP domain to the rest of the YES complex is the flexible linker to the IF domain and binding to the extended C-terminus of protein E. We observe continuous backbone density that allows us to place P65 at the FKBP active site. Interestingly, as seen before for some SlyD bound peptides (32), protein E binding to the FKBP domain adopts a non-canonical orientation. Protein E P65 is not completely conserved (fig. S1), although other species have a proline at residue 63 which could reach the active site with additional tilting of the FKBP domain. A proline near this position may help to localize the chaperone to the complex and facilitate assembly.

Fig. 5. Interactions between protein E and SlyD.

(A) The amphipathic helix of protein E (yellow) as a cartoon with side chains contacting SlyD as sticks. SlyD (purple) is shown as transparent accessible surface and cartoon. (B) Full view of SlyD bound to two protein E molecules shown in different shades of yellow. (C) As in (A) highlighting the β-augmentation of the extended C-terminal protein E. (D) Structures of SlyD from the YES_ID21 complex and Thermus thermophilus SlyD (PDBID:7OXI (33)) (green) aligned to the IF domains. The two S2 peptides bound to the TtSlyD are shown in pink.

The lack of contacts between SlyD and MraY suggest the soluble domain of protein E alone could form a complex with SlyD. We co-expressed EcSlyD with N-terminal truncations of protein E from either ID21 (residues 33–76) or the shortest isoform α3 (residues 33–75). Both form complexes that could be purified by an affinity tag on the soluble domain (fig. S2E). These observations point to a high-affinity interaction between SlyD and the C-terminal domain of protein E.

Protein E is unstable in the absence of SlyD (18) and is rapidly degraded (23). While it has been speculated that prolyl-isomerization was central to the lysis mechanism, evidence does not support this. Non-proline mutations in protein E can rescue lysis in a ΔslyD background (18, 46), as well as complete replacement of the cytoplasmic domain of protein E with an unrelated globular protein (20, 21, 46). To explore this, we performed our lysis assay with several SlyD variants in the ΔslyD strain (fig. S2A). As before, protein E_ΦX174 was unable to promote lysis in the absence of SlyD, however lysis could be rescued by expression of either EcSlyD or EcSlyD₁₅₄. TtSlyD has high structural homology to EcSlyD (Fig. 5D) and also rescued lysis in the ΔslyD strain (fig. S2A). We generated a EcSlyD Y68K mutant that dramatically reduces prolyl-isomerase activity (47) and this too could rescue lysis. Finally, we purified a complex of protein E using the Epos rescue mutants (R3H & L19F) (46) in the ΔslyD strain. This YE complex was very unstable and showed significant aggregation by SEC (fig. S2C). This all supports that the primary role of SlyD is not prolyl-isomerization, but to protect protein E and stabilize the YES complex. This does not obviate a role for proline binding in complex assembly. Supporting this, the two prolines nearest the FKBP domain have a slower-lysis phenotype (22), which may indicate that binding of these residues by SlyD is key to assembling the YES complex.

Protein E is evolutionary constrained

Protein E arrived late in the evolution of ΦX174 and was overprinted into a +1 reading frame in the ORF for gene D (48). The structure supports that this embedding constrains the evolution of protein E (12). Considering the sequence changes across protein E isoforms, we note that from the N-terminus through residue 70, with few exceptions, each position that is not completely conserved is either silent or a slight change in protein D (fig. S8A). The silent changes vary the codon’s second position, which is the wobble position in the overlapping codon for protein D. For example, residue W7 in ΦX174 is replaced with a Ser or Leu in other species. While seemingly dramatic changes, each is coded by the sequence UXG and all three variations at this position are sampled. These variable positions generally do not contact MraY. An exception is found in the G4 isoform where a phenyalanine occurs at position 19, as in the Epos mutant, which results in a change from alanine to serine at position 79 in protein D. This places a polar residue in the hydrophobic core of protein D (fig. S8). This single nucleotide change, while increasing protein E affinity to MraY, likely lowers the stability of protein D and the overall fitness of the virus. This is demonstrated experimentally where introduction of this mutation results in smaller plaque formation relative to wild-type (23). The C-terminus of protein D and the extensions in the longer isoforms of protein E are likely disordered and less constrained, hence the increased sequence variability (fig. S8).

Discussion

The mechanism for ΦX174 inhibition of peptidoglycan biosynthesis is remarkably simple. Protein E binds in the active site cleft preventing access of the lipid substrate and blocking conformational changes needed for catalysis (fig. S14). The YES complex structure provides a template for interpreting all the previous results for lysis by ΦX174. Notable among these are that the extensive functional mutations in protein E and MraY map to the interface between the two proteins. SlyD binds to the cytoplasmic domain of protein E to stabilize the MraY/protein E complex. As we demonstrated, the requirement for SlyD is based on the protein binding properties, as distantly related SlyD homologs can rescue lysis in a ΔslyD strain. SlyD in this case does not serve directly in the mechanism of inhibition of MraY and can be functionally replaced.

The EcMraY structure by cryo-EM reveals features previously disordered. An example is the NTH helical-stacking, that merits further study. Visualization of loop 1–2 supports an active role in the catalytic mechanism. The position of loop 9–10 provides evidence that this is a general inhibited conformation. The lipid densities in our map support the importance of these in the function of MraY, including evidence for an additional C₅₅P binding site near the dimer interface (45). While providing no further resolution to the identity of the molecule in the periplasmic cavity, the density indicates an ordered hydrophobic molecule that may have functional implications.

One of the most remarkable features of protein E is its simplicity. During the evolution of a phage, a lysis mechanism likely evolves late, as phage that have their lysis genes removed can still propagate, albeit much less efficiently (19). This can be seen in action in ssRNA phages whose hosts vary due to their receptor being plasmid-borne pili, which can transfer across species and, therefore, require new lysis genes in new hosts (reviewed in (49)). For these phages, new lysis genes rapidly develop and can be found overprinted throughout the genome (49). For protein E, it clearly evolved late (50) likely due to the introduction of a ribosome binding site. This new peptide would be constrained by being embedded in an essential gene. Protein E is rapidly degraded, likely due to its disordered C-terminus, and requires SlyD for stabilization. We know that simple mutations, such as L19F, can hasten lysis onset. The timing of lysis is directly related to the total amount of protein E in the membrane and its affinity to MraY. All of this supports that protein E can be dramatically improved to become a more efficient inhibitor of MraY.

The YES complex structure sheds light on the distinct ΦX174 lysis mechanism for the first discovered SGL. There are several ways SGL proteins, such as protein E, can be developed for therapies. First, protein E binding to EcMraY identifies a distinct mechanism from current inhibitors. The blocking of the lipid substrate access to the active site in MraY can be exploited by small molecules and provides a new template for structure-based drug design. Expanding protein E efficacy to Gram-positive bacteria and, potentially, other related polyisoprenyl-phosphotransferases such as bacterial WecA or the mammalian DPAGT1 could further increase applicability. Second, the genetically efficient SGL proteins can be potent programmable tools for killing engineered cells. Finally, SGL genes can be used to improve the antibacterial potency of bacteriophage. The use of phage for medical therapies, while known for a hundred years, has only recently become possible in the West (1). The desperate need to combat the dangerous rise of anti-microbial resistance will include phages. SGL genes, such as protein E from ΦX174, will play an important role.

Materials and Methods

Co-expression of EcMraY, protein E and EcSlyD

ΔslyD BL21(DE3) competent cells were co-transformed with pET22b-SlyD₁₅₄ and either pRSFDuetEcMraY-E_ID21 or pRSFDuet-EcMraY-E_ΦX174 and plated in LB-agar containing 35 μg/ml Kanamycin and 100 μg/mL Ampicillin. Our pET22b-SlyD₁₅₄ construct expresses E. coli SlyD, modified by the removal of the flexible C-terminus. The pRSFDuet-EcMraY-E_ID21 plasmid contains the ID21 isoform of protein E, along with a wild-type EcMraY to prevent cell lysis from the overexpression of protein E. Cells were grown in 2xYT media at 37°C, 225 r.p.m., and induced at an OD₆₀₀ of 0.9 with 0.4mM IPTG at 18°C overnight. The culture was harvested by centrifugation for 10 minutes at 9,000xg, 4°C then frozen or used immediately for purification.

Purification of the YES complex

The cells were resuspended in lysis buffer (20mM Tris-HCl pH 7.5, 300 mM NaCl, 10% Glycerol, 5mM βME, 0.1mM PMSF, 0.1mM Benzamidine) and homogenized using a M-110L microfluidizer (Microfluidics). The lysate was cleared by a 20-minute centrifugation at a speed of 22,000xg. The supernatant was then centrifugated at 167,424xg and the resulting membrane pellet was then solubilized in the extraction buffer (10 mM HEPES pH 7.5, 300 mM NaCl, 5% Glycerol, 5mM β-mercaptoethanol (BME), 0.1mM phenylmethylsulfonyl fluoride (PMSF), 0.1mM benzamidine, 10 mM imidazole and 1% dodecyl 4-O-α-D-glucopyranosyl-β-D-glucopyranoside (DDM)) After allowing for extraction for 1.5 hours at 4°C, the solution was centrifuged at 167,424xg for 30 minutes and the remaining lysate was mixed with 1mL NiNTA resin (Qiagen, Alameda, CA) then nutated at 4°C for two hours. This solution was loaded onto a gravity column and then washed with five column volumes of wash buffer (10 mM HEPES pH 7.5, 150 mM NaCl, 5% glycerol, 5mM BME, & 0.03% DDM) with 10mM imidazole followed by five column volumes of wash buffer with 30 mM imidazole. The YES complex was eluted in 20mL of wash buffer containing 200 mM imidazole. The final purification step was SEC (Superdex 200 5/150 GL, Millipore Sigma) in 10mM HEPES pH 7.5, 75 mM NaCl, 5% Glycerol, 5mM BME and 0.03% DDM. Fractions were assessed by SDS-PAGE and directly used for cryo-EM sample preparation.

Co-expression of EcMraY and protein E in various SlyD backgrounds

The pRSFDuet-E-EcMraY and pRSFDuet-Epos-EcMraY expression vectors were transformed into BL21-Star cells (Novagen). Similarly, the pRSFDuet-E(C-term)-SlyD₁₅₄ was transformed into SlyD-knockout cells. The cultures were grown at 37 °C to an OD₆₀₀ 0.8 and induced with 1 mM IPTG. Induced cultures were grown for 3 hours followed by harvesting by centrifugation at 9,000xg for 20 min. Cell pellets were resuspended in lysis buffer and lysed by sonication. The lysate was then cleared by centrifugation at 22,000xg, followed by a second centrifugation at 234,78xg for 1 hour to isolate the membrane fraction. The complex was extracted in 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% Glycerol, 10 mM Imidazole, and 1% n-Decyl-β-Maltoside (DM) and incubated at 4 °C for 1.5 hours. The debris was cleared by centrifugation at 234,788xg for 30 min. The sample was incubated with 1 mL NiNTA resin for 1 hour, followed by a wash with 50 column volumes lysis buffer with 30mM Imidazole. The protein E complexes were similarly eluted in 300mM Imidazole. The elutions were concentrated and further purified by size exclusion chromatography (Superdex 200 5/150 GL, Millipore Sigma).

Lysis assays of WT protein E ΦX174 and ID21

LEMO DE3 competent cells were transformed with a pRSF-Duet vector either empty, with protein E_ΦX174, or protein E_ID21. Cultures were grown to an OD₆₀₀ of 0.2 and inoculated into a Corning 96-well Clear Flat Bottom plates in 100μL triplicate aliquots and induced as described previously. Cultures were incubated at 37C with orbital shaking at 220rpm using an Infinite M Nano+ (Tecan, Switzerland). Readings were taken in 5-minute intervals for 90 minutes.

Lysis assay for protein E constructs

LEMO DE3 competent cells (New England Biolabs, MA, USA) were transformed with a pRSFDuet vector either empty, with C-terminally FLAG tagged protein E_ΦX174 variants (WT, P21A, K46A). The lysis assays were performed in three biological replicates as previously described (21). Absorbance readings were recorded in 5-minute intervals for 1 hour and 30 minutes. Manual readings were taken using a Biowave Cell Density Meter CO8000. The values were plotted using GraphPad Prism version 9.1.1 for macOS.

Lysis assays based on SlyD variants

ΔslyD (18) cells were transformed with either a control empty pRSF-Duet vector or pRSFDuet-Protein_ΦX174 and either pET22b-EcSlyD, pET22b-SlyD₁₅₄, pET22b-EcSlyD Y68K, or pET22b-Thermus thermophilus SlyD. Cultures were grown in 2xYT media at 37°C and induced with 0.4mM IPTG once at an OD₆₀₀ of 0.2. Absorbance measurements were manually recorded in 5-minute intervals for 70 minutes. Similarly, ΔslyD cells were transformed with either a control empty pRSF-Duet vector or pRSF-Duet-Protein_ID21 either alone, with pET22bEcSlyD, or with pET22b-EcSlyD₁₅₄ and induced with 0.4mM IPTG. Readings were recorded using an Infinite M Nano+ plate reader as described above.

Sample preparation for cryoEM

The YES complex was diluted to 5.0 mg/mL in 10 mM HEPES ph 7.5, 75 mM NaCl, 5% Glycerol, 5mM βME and 0.03% DDM. Additionally, the YES_ID21 sample was supplemented with 1mM E. coli total lipid extract (Avanti Polar Lipids, 100600P). Quantifoil holey carbon films R1.2/1.3 300 Mesh, Copper (Quantifoil, Micro Tools GmbH) grids were glow discharged with a 2-minute 20Å plasma current using a Pelco easiGlow, Emitech K100X. Grids were prepared using a Vitrobot (FEI Vitrobot Mark v4 ×2, Mark v3) by applying 3μL of sample onto the grid followed by a 3.5 second blot using a +8-blot force and plunge frozen into liquid ethane.

Data acquisition and analysis

The grids were imaged in a 300 kV cryo-TEM microscope equipped with a Gatan K3 6k x 4k direct electron detector and a Gatan Energy Filter (slit width 20eV) in super-resolution mode using Serial EM. Datasets were collected at a 105k magnification with a pixel size of 0.416 Å/pixel. Movies with 40 frames were recorded with a total exposure dose of 60 e⁻/Ų and a defocus range of −1.0 to −2.5 μm. For the YES_ID21 complex, a total of 12,070 movies were recorded.

Movies were normalized by gain reference and motion corrected using the patch motion correction built in function in cryosparc (v3.3.2) with a two-fold bin that resulted in a pixel size of 0.832 Å/pixel (51). The contrast transfer function (CTF) was estimated using CTFFIND4 (52). Micrographs were manually curated, and low-quality images were removed for further analyses. A total of 2,462,335 particles were obtained followed by the generation of 6 ab-initio models. Out of the 6 models, two models are selected for classification into “good” and “trash” volumes. All particles were then sorted in these two volumes through heterogeneous refinement using particles extracted with a 4x bin, which produced 6,589,696 good particles. Heterogeneous refinement was used in an iterative manner to sort the particles into the 5 volumes (4 good and 1 trash). The 1,151,777 good particles were used for non-uniform homogeneous refinement to generate a higher resolution volume. The particles were then extracted with a 3x bin and sorted into 4 iterations of the higher resolution volume and 1 trash volume. Iterative rounds of heterogeneous refinement at 3x bin produced 935,754 particles. Particles were then extracted in a 2x bin and heterogeneously refined into either high- or low-resolution volumes. At this point, discerning features in the soluble region of the model were used to select the most complete volumes. The volumes were individually refined through non-uniform refinement and the particles that composed the volumes with most complete and highest resolution were used. A total of 122,452 particles were used for the most complete model obtained upon non-uniform refinement. The FSC-masked resolution was 3.5Å, while the unmasked resolution was 3.9Å. The half-maps were then used for post-processing through DeepEMhancer (53) with the high-resolution model selected for our most complete density map. Post-processing through DeepEMhancer removed the micelle and improved the features on the soluble portions of the map, however the lipid densities were also removed. The lipid densities described in this work are those of the YES_ID21 map before post-processing. Notably, the dimer-interface lipid density was also present in the YES_ΦX174 density map without the supplementation of E. coli lipid extract. Figure 1D uses the densities before post-processing for the MraY dimer, micelle and lipid densities, and the DeepEMhancer post-processed map for protein E and SlyD. Supplemental figures S9 and S12 were made with the map before DeepEMhancer sharpening. The YES_ΦX174 complex dataset was processed in this same manner. A total of 10,798 movies were recorded. The model for the YES_ID21 complex was then used as a template for template picking, from which 1,516,368 particles were picked and curated. Following gradual un-binning and sorting into good and trash volumes, 155,270 particles were used for the final iteration of non-uniform refinement. The local resolution of both maps was performed on cryosparc (v3.3.2). The half-maps were then post-processed using DeepEMhancer as described previously.

Model building

For starting models we used the Aquifex apo structure (PDBID:4J72(25)) for EcMraY and the E. coli NMR structure (PDBID:2K8I(31)) for EcSlyD which were fit using phenix.dock. SlyD was then split into its two domains, IF and FKBP, at residues Y68 and G127. Protein E was modeled de novo up to residue P65 using Coot 0.8.9.2. The structures of the EcMraY, protein E, and EcSlyD-IF domain were refined using phenix.real space refinement and ISOLDE 1.6, and validated with PHENIX-1.19.2. After the refinements of EcMraY, protein E, and the EcSlyD-IF domain were completed, the FKBP domains were docked into density using ChimeraX. The complete YES complex structure was then refined using PHENIX-1.19.2 and ISOLDE 1.6. RMSDs were calculated using ChimeraX Matchmaker chain alignment. Structure figures were made using ChimeraX and sequence alignments using Jalview and ClustalW (54–56).

Data and materials availability:

All experimental data are available in the main text or the supplementary materials. Coordinates with experimental maps have been deposited to the RCSB or the EMDB for both the YES_ID21 and YES_ΦX174 complexes with the respective accession numbers PDBIDs 8G01 & 8G02 and EMDB-29641 & EMDB-29642.