Native β-barrel substrates pass through two shared intermediates during folding on the BAM complex

Abstract

The assembly of β-barrel proteins into membranes is mediated by the evolutionarily conserved β-barrel assembly machine (BAM) complex. In Escherichia coli, BAM folds numerous substrates which vary considerably in size and shape. How BAM is able to efficiently fold such a diverse array of β-barrel substrates is not clear. Here, we develop a disulfide crosslinking method to trap native substrates in vivo as they fold on BAM. By placing a cysteine within the luminal wall of the BamA barrel as well as in the substrate β-strands, we can compare the residence time of each substrate strand within the BamA lumen. We validated this method using two defective, slow-folding substrates. We used this method to characterize stable intermediates which occur during folding of two structurally different native substrates. Strikingly, these intermediates occur during identical stages of folding for both substrates: soon after folding has begun and just before folding is completed. We suggest that these intermediates arise due to barriers to folding that are common between β-barrel substrates, and that the BAM catalyst is able to fold so many different substrates because it addresses these common challenges.

Keywords: disulfide crosslinking; folding intermediates; gram-negative bacteria; outer membrane proteins; β-barrel assembly machine.

Fig. 1.

Deletion of a single aspartate in LptD (ΔD330) causes assembly defects on BAM. (A) BAM catalyzes the folding of outer membrane β-barrels. E470 lies on the interior surface of the BamA lumen. (B) Structure of the β-barrel domain of LptD with the lipoprotein plug LptE (PDB: 4RHB). Deletion of residues D330 to D352 (highlighted in red) produces a slow-folding variant of LptD named LptD4213. D330 (Inset, dark red) is the N-terminal-most residue within the deleted region. (C) Two-plasmid system used to probe cell permeability defects caused by deletions within LptD. (D) Deleted regions tested (red) span the LptD4213 23-amino acid deletion. (E) LptD4213 displays dominant phenotypes with respect to vancomycin sensitivity. Cells containing plasmid A, encoding LptD with deletions, and plasmid B were plated on LB agar containing 0.2% arabinose and/or 75 μg/mL vancomycin. LptD-ΔD330 recapitulates the permeability defects caused by LptD4213. (F) vancomycin sensitivity assay of E. coli containing LptD-ΔD330 and LptD4213 in cells with or without a chromosomal copy of BamA-E470G. Cells were plated on LB agar containing varying concentrations of vancomycin. The permeability defects conferred by LptD-ΔD330 and LptD4213 are each partially rescued by BamA-E470G. Plating assays are representative of at least two biological replicates.

Fig. 2.

LptD-ΔD330 and LptD4213 accumulate in the BamA lumen near E470 during folding. (A) In vivo photocrosslinking assay used to assess residence time of LptD variants within the BamA lumen. The unnatural amino acid pBpa forms covalent crosslinks to nearby moieties upon irradiation with UV light. (B) Positions within BamA β-barrel which form in vivo crosslinks to LptD4213 (light blue, yellow, PDB: 6V05). (C) BamA crosslinks to LptD-ΔD330 at the lateral gate and luminal wall. S439 and Y468 (yellow) crosslink to both LptD4213 and LptD-ΔD330. Positions which crosslink to LptD4213 but not LptD-ΔD330 are shown in light blue. (D) Additional positions within β-strand 4 of BamA where crosslinking to LptD-ΔD330 or LptD4213 was assessed (light blue, green) (PDB: 6V05). (E) The BamA luminal wall crosslinks to LptD-ΔD330, albeit significantly less strongly than to LptD4213. Weaker crosslinks are denoted with red arrowheads. BamA×LptD adducts were detected by His pulldown, followed by α-His and α-LptD immunoblot analysis. Immunoblots are representative of at least two biological replicates.

Fig. 3.

Development of a disulfide crosslinking assay to determine what regions of LptD variants accumulate at the BamA luminal wall proximal to E470. (A) Disulfide crosslinking assay used to assess the residence time of LptD variant positions within the BamA lumen during folding. Regions of substrate with a substituted cysteine which spend a long time in the BamA lumen are more likely to crosslink to a substituted cysteine on the BamA luminal wall. (B) Positions Y468 or D512 near E470 in BamA were substituted with cysteine to enable disulfide crosslinking (PDB: 6V05). (C) Diagram of the LptD β-barrel showing the N-terminal region and C-terminal region tested in our crosslinking experiments. Residues deleted in LptD4213 are colored red. (D) Detail of the N-terminal region of LptD. Residues deleted in LptD4213 are colored red. Residue N274 is colored green. The positions in each strand tested for disulfide crosslinking to the BamA luminal wall are shown in yellow. The list of residues replaced with cysteine are listed in SI Appendix, Table S4. (E) BamA-D512C crosslinks LptD4213 β-strands 5 to 3 more strongly than BamA-Y468C. (F) N274 faces the membrane on strand 4 of LptD, proximal to the residues on β-strands 5 to 3 (Inset) used for disulfide crosslinking experiments (PDB: 4RHB). (G) LptD4213-N274I reduces residence time of the N-terminal region of LptD4213 within the BamA lumen. Crosslinks between the BamA luminal wall and β-strands 5 to 3 of LptD4213 are virtually eliminated upon addition of the N274I mutation. BamA×LptD adducts were detected by His pulldown, followed by α-His and α-FLAG immunoblot analysis. Cellular levels of LptD and LptE were assessed by α-FLAG and α-LptE immunoblot analysis. Immunoblots are representative of at least two biological replicates.

Fig. 4.

The folding pathway of wild-type LptD passes through a stable N-terminal intermediate and a stable C-terminal intermediate. (A) β-strands 4 to 2 in the N-terminal region of both wild-type LptD and LptD-ΔD330 have a long residence in the lumen of BamA. LptD-ΔD330 displays stronger crosslinks in this region than wild-type LptD. Immunoblots assaying the residence time of the full set of wild-type LptD β-strands 12 to 1 can be found in SI Appendix, Fig. S3. (B) A set of C-terminal extracellular loops and a set of N-terminal extracellular loops of both wild-type LptD and LptD-ΔD330 have long residence time in the BamA lumen. (C) β-strand 19 of LptD crosslinks to the BamA luminal wall regardless of N-terminal deletions that slow-folding. β-strands 26 to 13 of wild-type LptD, LptD-ΔD330, or LptD4213 were all assessed for disulfide crosslinking to BamA-D512C. The intensity of the wild-type LptD β-strand 19 crosslink is comparable to that of the two LptD mutants. The list of residues replaced with cysteine are listed in SI Appendix, Table S4. BamA × LptD adducts were detected by His pulldown, followed by α-His and α-FLAG immunoblot analysis. Immunoblots are representative of at least two biological replicates.

Fig. 5.

BamA^S exhibits the same C- and N-terminal folding intermediates as LptD. (A) Disulfide crosslinking assay used to assess residence time of BamA^S β-strands or extracellular loops within the lumen of BamA^M during folding. (B) Diagram of the 16-stranded BamA^S β-barrel. (C) Disulfide crosslinking between each β-strand 16 to 1 of BamA^S and the BamA^M luminal wall with or without D512C in BamA^M. Two sets of BamA^S β-strands crosslink to the BamA^M luminal wall: β-strands 12 to 8 (red arrowhead) and β-strand 3 (green arrowhead). These crosslinks depend on the presence of BamA^M-D512C. (D) Disulfide crosslinking between positions within each extracellular loop of BamA^S and the BamA^M luminal wall. Six different positions within extracellular loop 6 were assessed. Crosslinking positions fall within the C-terminal extracellular loops 7 to 5 of BamA^S. The list of residues replaced with cysteine are listed in SI Appendix, Table S4. BamA^M × BamA^S adducts were detected by His pulldown, followed by α-His and α-FLAG immunoblot analysis. Immunoblots are representative of at least two biological replicates.

Fig. 6.

Proposed mechanism for how LptD-N274I and BamA-E470G/K suppress folding defects. (A) BamA-E470G does not decrease the stability of the N-terminal folding intermediate while folding LptD-ΔD330. Disulfide crosslinks between β-strands 4 and 3 of LptD-ΔD330 are not decreased in strength after chromosomal substitution of BamA-E470G. BamA×LptD adducts were detected by His pulldown, followed by α-His and α-FLAG immunoblot analysis. (B) E470G and E470K partially rescue LptD-Y721D from degradation by DegP. Levels of wild-type LptD and LptD-Y721D variants in cells were assessed in the presence of chromosomal copies of BamA-E470G/K variants or DegP deletion. Cellular levels of LptD were assessed by α-FLAG and α-LptE immunoblot analysis. (C) BamA-E470G and BamA-E470K have similar mechanisms of suppression as judged by vancomycin sensitivity. E. coli containing LptD-ΔD330 and a chromosomal copy of wild-type BamA, BamA-E470G, or BamA-E470K were plated on varying concentrations of vancomycin. BamA-E470G and BamA-E470K similarly rescue the membrane permeability defect imparted by LptD-ΔD330. (D) Model for the differential activity of the suppressors LptD-N274I and BamA-E470G/K. LptD-N274I is proposed to alleviate the deep energy well in which LptD4213 becomes trapped late in folding, thereby increasing the folding rate by easing catalysis. Conversely, BamA-E470G/K is proposed to increase productive binding for substrates by providing an additional path to bind directly to BamA. Immunoblots are representative of at least two biological replicates.