Folding Determinants of Transmembrane β-Barrels Using Engineered OMP Chimeras

Abstract

Transmembrane β-barrel proteins (OMPs) are highly robust structures for engineering and development of nanopore channels, surface biosensors, and display libraries. Expanding the applications of designed OMPs requires the identification of elements essential for β-barrel scaffold formation and stability. Here, we have designed chimeric 8-stranded OMPs composed of strand hybrids of Escherichia coli OmpX and Yersinia pestis Ail, and identified molecular motifs essential for β-barrel scaffold formation. For the OmpX/Ail chimeras, we find that the central hairpin strands β4–β5 in tandem are vital for β-barrel folding. We also show that the central hairpin can facilitate OMP assembly even when present as the N- or C-terminal strands. Further, the C-terminal β-signal and strand length are important but neither sufficient nor mutually exclusive for β-barrel assembly. Our results point to a nonstochastic model for assembly of chimeric β-barrels in lipidic micelles. The assembly likely follows a predefined nucleation at the central hairpin only when presented in tandem, with some influence from its absolute position in the barrel. Our findings can lead to the design of engineered barrels that retain the OMP assembly elements necessary to attain well-folded, stable, yet malleable scaffolds, for bionanotechnology applications.

Protein engineering has application in industry, where there is an immense need to increase yields and stability of proteins, particularly enzymes. The strategy of engineering enzymes has been exploited to change substrate specificity, turnover number, catalytic efficiency, enantioselectivity, and thermal stability by incorporating mutations. ^1,2 Among structural proteins, engineered transmembrane β-barrels have been exploited for bacterial display libraries, ³ and the generation of synthetic pores of desired size and geometry. ⁴ For example, the Escherichia coli outer membrane protein FhuA, which is a 22-stranded transmembrane β-barrel, has been engineered to obtain a 24-stranded β-barrel by duplicating the N-terminal strands of the parent protein at its C-terminus. ⁴ The increased channel diameter now allows for FhuA-24 to act as a bigger pore. Such synthetic nanopores have found industrial applications as drug transporters, multicomponent sensors, and enzymes. ⁵ In another interesting study, the E. coli maltoporin was engineered by addition, deletion, and β-strand replacement with the Salmonella typhimurium maltoporin to obtain engineered β-barrels with interesting gain or loss of function. ⁶

The high stability of the E. coli outer membrane protein X (OmpX) β-barrel has been exploited to generate a bacterial display library by engineering the extracellular loop regions and circular permutation of the barrel. ³ OmpX exhibits high structural plasticity, which is evident from its high tolerance to mutations. ^3,7 This feature of OmpX has also been established experimentally in a study where this 8-stranded β-barrel has been engineered to generate 8–16-stranded scaffolds through β-hairpin duplication. ⁷ However, extensive analysis is required to obtain successful barrel re-engineering for specific functions. For example, generating functionally unique transmembrane β-barrel scaffolds requires the identification and characterization of molecular motifs that determine the folding nucleation, assembly, and stability of these scaffolds. ⁸ Detailed studies toward identifying the molecular factors for barrel folding which can be used to design and engineer transmembrane β-barrels, are limited. ^7,9–12 The central hairpin, composed of strands β4–β5, is believed to be essential for OMP assembly. ⁷ The β-signal motif that is thought to be important for OMP assembly in vivo ^8,13–16 is dispensable in vitro ¹⁷ and in vivo for _{some OMPs.} ^14,18

OmpX exhibits high structural plasticity, which is evident from the tolerance of this protein to extensive mutagenesis, and its ability to attain a stably folded β-barrel scaffold in various lipidic and micellar systems. ^3,7,19–21 Owing to this structural plasticity, OmpX serves as an excellent model protein to understand the biophysical determinants of transmembrane β-barrel folding. Anfinsen’s hypothesis ²² that the primary protein sequence carries sufficient information for scaffold formation has been verified experimentally for several transmembrane β-barrels. However, bacterial OMPs exhibit remarkable diversity in their primary sequence, although they share very similar structural scaffolds. ²³ This is best illustrated by comparing the crystal structures of OmpX from E. coli and Ail (attachment and invasion locus protein ²⁴ ) from that of Yersinia pestis, both of which belong to the OMP family involved in virulence. ²⁵ While both proteins share a high structural identity with a root-mean-square deviation of 1.7 Å for C^α atoms, ²⁴ the overall sequence identity is <45%, and ~60% in the transmembrane region (Figure 1A). Both OMPs also differ in their secondary structure content, because Ail contains larger disordered extracellular loops, ^24,26 whereas OmpX has a protruding β-sheet structure termed “fishing rod” (Figure 1B). ²⁷

Figure 1. Sequence and structural alignment of Ail and OmpX.

(A) Pairwise alignment of Ail and OmpX sequences. Regions in yellow are the transmembrane strands. The proteins show ~60% sequence identity in the transmembrane region. (B) Overlay of the crystal structures of Ail (3QRA; green) and OmpX (1QJ8; gold) showing significant structural overlap.

Despite belonging to the same family of 8-stranded OMPs, and sharing structural homology, Ail and OmpX show contrasting biophysical characteristics, including differences in their stability and interactome. ^28,29 Ail is known to bind with extracellular matrix components including heparin as well as laminin and fibronectin, but not collagen. ^24,26 The interaction of Ail with laminin triggers the delivery of Yops (Yersinia outer proteins) into mammalian cells. ²⁴ OmpX is not known to interact with any of these components. Hence, the two proteins serve as ideal candidates to map the underlying commonality in folding motifs in their primary sequences. Here, we have generated chimeric β-barrels of OmpX and Ail to identify the molecular elements for β-barrel scaffold formation. Using biophysical and spectroscopic characterization of these barrel chimeras, we have identified the importance of the central β-hairpin and the C-terminal β-signal for the in vitro folding of OmpX/Ail chimeras. Our observations are in excellent agreement with previous findings on the importance of the central hairpin and the β-signal motifs. ^7,12 Additionally, we find that OMP assembly is likely to be nonstochastic in micelles. We discuss the implications of chimeric strands and barrel nucleation motifs on the observed frustration in OMP folding pathways as well as the scaffold stability. Our approach using segmented β-barrels to identify signature elements for scaffold assembly has allowed us to narrow down molecular differences between two structurally identical OMPs.

Experimental Methods

Chimera Generation Using Overlapping PCR

All the constructs used in this study were cloned in pET3a vector between NdeI and BamHI sites, and were expressed without any tag using reported methods. ²⁰ For the construction of chimeric barrels of Ail and OmpX, gene fusions were created using overlapping PCR ³⁰ in both permutations, i.e., ail-ompX (protein product: Ail-OmpX, AO) and ompX-ail (protein product: OmpX-Ail, OA). These fusions were further used to generate the six permutants of 8-stranded chimeric β-barrels. These six constructs do not retain the strand positions of the parent barrels. In addition, two strand-conserved chimeras were also generated by using the deletion strategy of overlapping PCR. Here, the protein products retained the position of the strands as that of the parent protein. The design strategy and mutants generated are described in Figure 2. The chimeras were named based on whether OmpX or Ail was at the N-terminus of the chimera, and the number of strands from each parent protein. Additional point mutations were generated using conventional site-directed mutagenesis to drive protein folding.

Figure 2. Strategy for chimera design.

(A) ompX (black) and ail (gray) genes were fused using overlapping PCR as ompX-ail, and ail-ompX. Using ompX-ail fusion, three chimeras o6a2, o4a4, and o2a6 were generated such that each protein product that was encoded had 8 strands. In the same way, three chimeric barrels were produced from ail-ompX fusion using specific primers. The chimeric barrels thus generated had strand positions distinct from both the parent barrels. For example, O6A2 (coded by o6a2) possesses the last six strands (third to eighth strand) from OmpX and first two strands (first and second strand) from Ail in the order N-ter-β3-β4-β5-β6-β7-β8-β1-β2-C-ter. (B) Cloning strategy for generating the strand conserved mutant O4A4^SC. The central undesired gene segment was deleted and the desired segments were amplified and ligated with two rounds of PCR. A4O4^SC was constructed in a similar manner using A8O8 as the template. All gene products were cloned into pET3a vector between NdeI and BamHI sites. The colored dots in the figure denote the position of tryptophans from Ail and OmpX. Here red and brown dots correspond to W76 and W140 from OmpX whereas green and blue dots correspond to W42 and W149 from Ail, respectively. The central hairpin is shown in pink and olive for OmpX and Ail, respectively, in both (A) and (B).

E. coli C41 cells were transformed with the respective constructs and the protein was overproduced as inclusion bodies. Inclusion bodies were purified in the denatured form using ion exchange chromatography, ^20,29 depending on the pI of the protein construct. Purified protein samples were dialyzed against water to remove urea. The precipitated protein was further lyophilized to obtain a protein powder and stored at −80 °C until use.

Protein Folding

The folding of Ail and OmpX was carried out as described previously ^20,29 by rapid dilution of the unfolded protein in n-dodecylphosphocholine (DPC) micelles. All the new constructs were folded by rapid 5-fold or 10-fold dilution of the denatured protein in 8 M urea (OmpX and all chimeric constructs except O2A6^F) or 8 M guanidine hydrochloride (GdnHCl) (Ail, O2A6^F) into the folding reaction containing 250 mM DPC in 50 mM Tris-HCl pH 9.5. All samples were incubated overnight at 25 °C. The folded protein stock was further diluted 5-fold for use in all experiments. The final concentrations achieved in the experiments were 15 μM protein in 50 mM DPC (corresponding to a detergent to protein ratio (DPR) of ~3300:1), in 50 mM Tris pH 9.5, 0.32 M urea, and 0.16 M GdnHCl.

Gel Mobility Shift Assay and Pulse Proteolysis

Membrane proteins show anomalous electrophoretic mobility upon folding. ²⁰ Gel mobility shift assay of all the folded samples was carried out on cold 15% SDS-PAGE of unboiled samples. Boiled samples served as controls. All the folded samples were further checked for protection from proteinase K (PK) digestion by pulse proteolysis of unboiled samples. The folding reaction was incubated at 37 °C for 1 min with 0.7 μM PK, following which SDS-PAGE gel loading dye containing 5.0 mM PMSF was added to arrest the protease activity, and the samples were analyzed on 15% SDS-PAGE. As controls, folded proteins were unfolded by boiling for 3 min at 100 °C, and were then treated with PK under similar conditions.

Secondary Structure Estimation Using far-UV CD

The secondary structure content of all the proteins was determined by recording far-UV circular dichroism (CD) spectra on a JASCO J815 CD spectropolarimeter (JASCO Inc., Japan). CD wavelength scans were recorded at 25 °C, from 205 to 260 nm at increments of 1 nm, and the data were averaged over 3 accumulations. The bandwidth was set at 1 nm and the data integration time for all the accumulations was 1.0 s. The scan speed was maintained at 100 nm/min. Each scan was corrected for buffer contribution, and converted to molar residue ellipticity (MRE) using reported methods. ²⁰

HSQC Measurements

¹H–¹⁵N HSQC-TROSY spectra of uniformly ¹⁵N-labeled proteins were obtained at 40 °C (OmpX) and 50 °C (A2O6 and A6O2) on a 700 MHz FT-NMR spectrometer. These temperatures were chosen to obtain well-dispersed spectra for each sample. The NMR samples contained 30 μM protein, 100 mM DPC, 50 mM Tris-HCl pH 9.5, 1.6 M urea, and 10% D₂O. The DPR used in CD and fluorescence measurements was retained here. The data were processed using NMRPipe ³¹ and plots were rendered using Sparky. ³²

Equilibrium (Un)folding Measurements

Equilibrium unfolding measurements were carried out by subjecting the folded samples to chemical denaturation using increasing GdnHCl (0.16 to 6.5 M) concentration. The folded protein stock (75 μM protein in 250 mM DPC) was diluted 5-fold into different GdnHCl concentrations. For folding measurements, unfolded protein prepared in 5 M GdnHCl was diluted 5-fold in various GdnHCl concentrations (5.0 to 1.0 M); further dilution of the sample could not be achieved. The progress of the reaction was monitored at 25 °C by measuring the change in intrinsic Trp fluorescence using an excitation wavelength of 280 nm, and emission scans were collected between 320 and 400 nm. The measurements were taken at 24 h intervals for ~6–12 days (measurements for OmpX were recorded for 12 days). All the thermodynamic parameters were derived from the equilibrium unfolding profiles achieved at different time-points for each protein. Each spectrum was corrected for buffer contributions. The data was processed using reported methods ³³ and fitted to a two-state or three-state function ¹⁷ to obtain the unfolding free energy ( $\text{Δ}{G}^0{}_{\text{U}}$ or $\text{Δ}{G}_{\text{U}1}^0$ and $\text{Δ}{G}_{\text{U}2}^0$ ), change in accessible surface (m value or m ₁ and m ₂) and midpoint of chemical denaturation (C_m or C_m1 and C_m2). Errors represent the goodness of fit.

Results

C-Terminal β-Signal Residues Can Contextually Influence OMP Folding

OMP sequences are considerably diverse to allow for direct mapping of folding motifs. To address whether, and to what extent, the folding is influenced by the sequence, we generated a library of chimeric OMPs using β-strand pairs derived from OmpX and Ail (Figure 2). To ensure that the strand positions are not conserved in the chimeras, we derived the chimeric barrels from Ail-OmpX or OmpX-Ail fusions (see Figure 2A for the chimera generation strategy). We also ensured that apart from the positioning of strands, certain basic characteristics, such as the β-signal motif, that are usually evolutionarily conserved in the 8-stranded β-barrels were also compromised. This allowed us to deduce the prerequisites that were indispensable for the folding of β-barrels.

We have previously shown that the folding of both OmpX and Ail in vitro from its urea- or GdnHCl-denatured state into lipidic micelles is accompanied by a prominent gel mobility shift in cold SDS-PAGE gels. ^20,29 The folded protein exhibits a retarded mobility compared to its unfolded counterpart in Laemmli gels, while the reverse effect is obtained in tricine gels. ^20,34 Proteins thus folded are also resistant to proteolysis by robust proteases with broad specificity, such as proteinase K (PK). ²⁰ We used gel mobility shift assays coupled with pulse proteolysis to assess the folding efficiency of OmpX-Ail chimeras.

Figure 3 summarizes the results of gel shift and proteolysis studies of the OmpX-Ail chimeras. Three out of the six chimeric constructs, namely, O6A2, A6O2, and A2O6, displayed >85% folding in DPC micelles. The nature of residues in the C-terminal β-signal motif ¹³ is crucial for determining the in vitro folding and stability of OMPs. ¹⁷ Both OmpX and Ail retain a conserved Tyr-Arg-Phe as a part of this β-signal (see Figure 1A). However, none of the chimeric barrels retain the β-signal at the C-terminus (Figure 3A). This is evident from a comparison of the C-terminal residues of the three chimeras that fold. While O6A2 possesses a Tyr-Glu-Phe that follows a similar aromatic-polar-aromatic motif seen in β-signals, A6O2 and A2O6 show comparable folding levels despite possessing Glu-Glu-Asp and Asn-Pro-Met, respectively, as their C-terminal residues. Our observation that three of the six chimeras fold efficiently and show resistance to proteolysis (Figure 3B) suggests that the β-signal motif could be dispensable for in vitro folding.

Figure 3. Monitoring folding efficiency for the OmpX-Ail chimeric barrels.

(A) Last ten amino acids (^#C-terminal β-signal motif is in bold) of the six hybrids and the two parent barrels. ^@Folding was deduced from gel mobility shift and protection to proteolysis. (B) Representative images showing gel mobility shift and protease susceptibility of engineered chimeric barrels separated on cold 15% SDS-PAGE. Dotted vertical lines demarcate independent gels that are presented together. M: marker; UC: unfolded control in 8 M urea; F: folded protein band; UF: unfolded protein band; PK: proteinase K. “−” and “+” sign above each lane corresponds to the absence and presence of PK, respectively.

The three barrels that did not fold were O2A6, O4A4, and A4O4. They lack the gel shift and are susceptible to proteolysis (see Figure 3B). In an attempt to promote folding of these chimeras, we first questioned whether the β-signal would influence barrel folding. We chose O2A6 as our test construct, which ends with the residues Gln-Phe-Asn. A library of O2A6 mutants generated by systematically adding, deleting, or substituting residues in the last (eighth) strand is listed in Figure 4A. Interestingly, adding a single Ala to the C-terminus to obtain the chimera O2A6^A promoted >30% folding of this construct (Figure 4B). Further, we found that the hydrophobicity of this additional residue also contributed considerably to the folding of O2A6, with phenylalanine (O2A6^F; FN → FNF) increasing the folding efficiency to >70%. Strand length and the hydrophobicity of the terminal residue might therefore be conditional elements in promoting unassisted OMP folding in vitro.

Figure 4. Folding screens of O2A6 mutants.

(A) Alignment of the last ten residues of O2A6 with its mutant library. The O2A6 chimeric barrel did not fold (see Figure 3B); hence, five mutant barrels that were generated are listed with changes (residue additions and deletions) incorporated in the parent O2A6 construct. (B) Gel mobility shift and protease susceptibility of O2A6 mutants checked using cold SDS-PAGE. Dotted vertical lines demarcate independent gels that are presented together. M: molecular weight marker; UC: unfolded control in 8 M urea; F: folded protein; UF: unfolded protein; PK: proteinase K. “−” and “+” sign above each lane corresponds to the absence and presence of PK, respectively. Also see Figure S1 for complete gel images.

Intact Central β-Hairpin in Tandem Is Required for Optimal OMP Folding.

We extended the strategy of adding a C-terminal aromatic residue to the O4A4 and A4O4 chimeras. However, strand extension did not provide us with the folded protein (Figure 5A,B), confirming that multiple factors contribute to OMP folding. The chimeras also possessed the mortise-and-tenon (Gly-Tyr) motif, ^12,35 which is believed to be evolutionarily conserved across β-barrels. One aspect that these chimeras lack is an intact central hairpin. ⁷ The central β-hairpin is composed of strands β4, β5 and the periplasmic turn T2. All our chimeric constructs retain the central hairpin strands as consecutive entities in the primary sequence, irrespective of their absolute position in the folded barrel, while the central hairpin is disrupted only in the O4A4 and A4O4 chimeras. In both chimeras, strands β4 and β5 now occur at positions β8 and β1, respectively. Proximal positioning of these two strands will therefore require barrel assembly. Hence, the inability of O4A4 and A4O4 chimeras to fold suggests that β4–β5 assembly precedes barrel scaffold formation.

Figure 5. Comparison of folding efficiency of the parent barrels with mutant strand-conserved chimeras.

(A) O4A4^F and A4O4^F represent the mutant barrels obtained by adding phenylalanine as the terminal residue to the original O4A4 and A4O4 chimeric barrels, respectively. (B) Gel mobility shift and protease susceptibility of O4A4^F and A4O4^F; the mutant chimeras did not fold and remained protease susceptible. (C) Evidence that the chimeric barrels (O4A4 and A4O4) fold after barrel redesign is seen from gel mobility shift and resistance to proteolysis. Also included are gels of the parent OmpX and Ail barrels. M: molecular weight marker; F: folded protein; UF: unfolded protein; UB: unboiled sample; B: sample boiled for 3 min at 100 °C in gel loading dye; PK: proteinase K. “−” and “+” sign above each lane corresponds to the absence and presence of PK, respectively. Dotted vertical lines demarcate independent gels that are presented together. Complete gel images are in Figure S1. Boiled samples were proteolyzed by PK; note that Ail stays folded despite boiling, when SDS is present.

To check if an intact central hairpin is essential for OMP folding, we generated strand-conserved chimeras O4A4^SC and A4O4^SC (SC: strand conserved; see Figure 2B for the chimera design). We designed the chimeras such that strands β4 and β5 were from the respective parent proteins, but are placed successively in the primary sequence. The strands also retain their native positions as the parent barrels. For example, the O4A4^SC chimera possesses β4 of OmpX at its fourth position and β5 of Ail at its fifth position. Additionally, these two proteins now end with the signature β-signal motif (Tyr-Arg-Phe). Interestingly both the strand-conserved chimeras are now well-folded and resistant to proteases (Figures 5C and S1). We conclude that the positioning of the central β-hairpin in tandem is important to fold transmembrane β-barrels. Mutants lacking this motif posed difficulties in rescuing the folding using other strategies.

It must be noted that the O2A6 mutant possesses the central hairpin as strands 6 and 7. Unlike its counterpart A2O6, this chimera could not be folded despite extensive screening of folding conditions. However, we could readily rescue the folding of this mutant by extending the last strand by one amino acid. Further, the nature of the amino acid (aliphatic in O2A6^A and aromatic in O2A6^F) determined the final folding efficiency (30% and 70%, respectively; see Figure 4B). Hence, the central hairpins β4–β5 in tandem is sufficient for OMP assembly, with minor re-engineering required additionally in specific circumstances.

OmpX–Ail Chimeras Retain Structural Signature of Parent Barrels

Next, we asked if the chimeric barrels show additive structural properties of their parent barrels. We first evaluated the secondary structure content of the chimeras using far-UV circular dichroism, and compared it with the parent barrels (Figure 6). OmpX, owing to the extra-membranous structured strands and shorter connecting loops, possesses ~1.5-fold higher secondary structure content than Ail, which shows longer connecting loops. Interestingly, we find a correlation and dependence of the secondary structure on the number of strands contributed by each parent protein (Figure 6). With the exception of O2A6^F, chimeras with a greater number of strands from OmpX are linked to higher β-sheet content. It also appears that in the strand-conserved chimeras (O4A4^SC and A4O4^SC) the parent protein that contributes to the fourth β-strand also decides the secondary structure content. We do not have a convincing explanation for this observation, and the limited data available precludes reasonable speculation at this point.

Figure 6. Secondary structure of folded chimeric barrels.

(A) Representative far-UV circular dichroism wavelength scans of OmpX, Ail, and the chimeric barrels shown as per-residue molar ellipticity (MRE) values. The data are corrected for buffer contribution and smoothened using the Means-Movement method after averaging over three accumulations. (B) Comparison of the secondary structure content of the six final chimeras with the parent barrels, OmpX and Ail. The molar residue ellipticity (MRE) averaged between 213 and 215 nm is shown for each chimera folded in DPC. Errors represent the standard deviation from 2 to 3 independent experiments.

Along with far-UV CD, the chimeras also showed well-dispersed HSQC spectra that resemble the parent protein (Figures 7, 8, and S2-S3). In support of this observation, we present the superposition of the ¹H–¹⁵N HSQC spectra of two chimeras with both their parent proteins. We find that the backbone resonances of both A6O2 (Figure 7) and A2O6 (Figure 8) superpose reasonably well with Ail and OmpX resonances, respectively. Conservation of the secondary structure suggests that the chimeras exist as β-barrels even after randomization of the strand positions. The well-dispersed HSQC spectra of the chimeras also support barrel closure and the formation of a stable scaffold in these sequences. Further, the superposed resonances show that local characteristics of the individual strands are retained to a reasonable extent, even when placed at different sites and interactions with non-native strand neighbors are established. Further quantitative analysis of the spectra can be achieved after we obtain the complete resonance assignment of the chimeric barrels. Overall, our far-UV CD and HSQC measurements establish the existence of high structural plasticity in transmembrane β-barrels, provided the signature motifs for β-barrel folding, namely the central β-hairpin strands in tandem and sufficient strand length, are available.

Figure 7.

Overlay of HSQC spectra of A2O6 and A6O2 with Ail. ¹H–¹⁵N HSQC spectra of uniformly ¹⁵N-labeled A2O6 (left) and A6O2 (right) folded in DPC micelles, with Ail (gray). Spectra were recorded on a 700 MHz NMR spectrometer at 50 °C for the chimeric barrels and 45 °C on a 600 MHz spectrometer for Ail. As A6O2 has six strands from Ail, there is a reasonable overlap in the chemical shift dispersion between the two proteins. See Figure S2 for the HSQC spectrum of A6O2. Ail spectrum is reprinted with permission from ref 26. Copyright (2011), with permission from Elsevier.

Figure 8.

Overlay of HSQC spectra of A2O6 and A6O2 with OmpX. ¹H–¹⁵N HSQC spectra of uniformly ¹⁵N-labeled A2O6 (left) and A6O2 (right) with OmpX (gray), folded in DPC micelles. Spectra were acquired on a 700 MHz NMR spectrometer at 50 °C for the chimeric barrels and 40 °C for OmpX. As A2O6 has six strands from OmpX, there is a reasonable overlap in the chemical shift dispersion between the two proteins See Figure S3 for the HSQC spectrum of A2O6.

Chimeric Constructs Show Folding and Stability Distinct from Parent Scaffolds

The folding of OMPs in membranous environments is believed to be concerted, ^10,11 while a more stochastic mechanism appears to predominate in micellar systems. ²¹ To address if the folding pathway and stability of the OMP are affected by scrambling the parent OmpX and Ail barrel sequences, we studied the equilibrium folding of the two parent proteins and the six chimeric constructs that are well folded (O6A2, O2A6^F, A6O2, A2O6, O4A4^SC, A4O4^SC). All constructs contain two interface tryptophans (see Figure 2). The change in tryptophan fluorescence serves as an excellent reporter of the folded protein population at different denaturant concentrations (see Figure S4 for fluorescence emission spectra). We used guanidine hydrochloride as the denaturant in our measurements, and monitored the progress of the reaction for 6–12 days at 25 °C.

All barrel constructs exhibited path independence in our experimental conditions, as evident from the superposition of the folding and unfolding profiles (Figure S5). We observed a two-state (un)folding profile in Ail, while the equilibrium profiles of OmpX could be explained by a two-state model in the initial 48 h, which slowly progressed to a three-state profile beyond ~8 days, indicating the slow accumulation of a distinct intermediate as the protein folds. A comparison of the thermodynamic parameters derived from fits of the profiles to a two-state or three-state model is provided in Table 1. Interestingly, all the chimeras except O4A4^SC and O2A6^F exhibit three-state (un)folding with the distinct accumulation of an intermediate. Although we can only derive limited conclusions from the thermodynamic parameters, we observe no influence of the thermodynamic stability of the parent barrels on the chimeras. Further, the change in free energy and folding cooperativity of the strand-conserved mutants do not match either OmpX or Ail. The folding efficiency of O2A6^F was up to 70% (as discussed earlier), and thus the thermodynamic parameters obtained from the two-state fit are partial estimates of the overall process.

Table 1. Free Energy Parameters Derived from Equilibrium Denaturation Measurements.

Protein a	$\Delta G^0\text{U}$ (kcal/mol) d		m value (kcal/mol/M)		C m (M)
OmpX (288 h)	8.06 ± 1.53	7.79 ± 3.81	−3.11 ± 0.71	−2.19 ± 0.74	2.59	3.55
O6A2 (96 h)	8.14 ± 1.05	1.57 ± 0.76	−5.70 ± 0.73	−0.55 ± 0.28	1.42	2.85
O4A4SC (144 h)	6.60 ± 0.60		−2.80 ± 0.30		2.36
O2A6F (96 h) b	6.21 ± 0.30		−3.10 ± 0.10		2.0
A2O6 (96 h)	11.68 ± 3.80	1.80 ± 1.30	−8.69 ± 2.95	−0.73 ± 0.64	1.34	2.46
A4O4SC (48 h)	6.21 ± 0.87	2.66 ± 2.60	−2.93 ± 0.72	−0.90 ± 0.73	2.11	2.95
A6O2 (24 h) c	4.99 ± 1.30	4.71 ± 1.25	−4.43 ± 1.05	−1.32 ± 0.30	1.12	3.56
Ail (72 h)	10.67 ± 0.5		−4.47 ± 0.2		2.38

^aTime point at which the thermodynamic parameters were derived is provided in parentheses.

^bFolding efficiency was ~70% in O2A6F.

^cPretransition baseline was not well-defined post 24 h.

^dErrors were high for the free energy parameters derived from three-state fits.

Overall, the chimeric barrels show no dependence or correlation of their thermodynamic stability with the parent barrels. They behave as independent barrels with unique biophysical attributes. Therefore, we conclude that the energetics and folding pathway of OMPs are regulated by their primary sequence, providing useful insight on evolutionary diversification of OMPs through mutagenesis.

Discussion

Transmembrane β-barrels evolved from simple genetic events wherein extreme sequence diversification followed an earlier gene duplication process. ²³ Prototypical β-barrels with larger strand numbers generated artificially using the 8-stranded OmpX barrel ^7,23 suggests that the primary sequence of the protein is not essential for barrel assembly. Consequently, the sequence divergence of OMPs is a consequence of evolutionary pressure on the host organism. Yet, we and others ^7,8,12,13,15 hypothesized that sequences or motifs that are vital for forming and assembling the β-barrel scaffold would still remain conserved. Through this study, we find that specific features of OMPs are indeed conserved for barrel assembly, which include an intact central hairpin, optimal strand length, and the β-signal sequence. These findings are in excellent agreement with previous observations. ^7,12,13,15 We find that the position of the central hairpin can be changed without significantly affecting the assembly of OMPs. Additionally, hybrid central hairpins (derived from two different proteins) are equally effective as conserved hairpins (derived from the same protein) for OMP assembly. Further, our approach using hybrid strand segments has revealed molecular factors that demarcate two structurally similar β-barrel OMPs, supporting a previous hypothesis that structural similarity does not necessarily translate to sequence homology. ²³

We summarize our findings from the OmpX-Ail chimeras in Table 2, where we compare parameters that are likely to affect the outcome of OMP folding. We obtain a consensus with the need for strands β4-β5 of the parent protein to be retained as consecutive entities. In other words, disrupting the central hairpin abolishes OMP folding. The absolute position of the central hairpin strands appears to be less important. For example, in A2O6 and A6O2, the central hairpin is located toward the C-terminus and N-terminus, respectively. Further, neither construct retains a β-signal motif at its C-terminus. Yet, both chimeras retain the ability to fold in DPC micelles. The O2A6^F mutant additionally emphasizes the importance of strand length and the presence of aromatic residues at the C-terminal of OMPs for enhanced stability. Experimental evidence suggests that the C-terminal phenylalanine plays a significant role in assisted folding of OMPs, ^13,14 while the extent of its energetic contribution to the unassisted assembly of OMPs is likely to be OMP-specific. ^17,36 Our results with the strand-conserved mutants also support this conclusion that the C-terminal hydrophobic residues are necessary but not sufficient to fold OMPs. We find that the C-terminal β-signal is not always essential for in vitro assembly of OMPs. Other recognition elements that assist the assembly process are additionally required for efficient β-barrel folding.

Table 2. Key Elements Regulating Folding Efficiency of an 8-Stranded Transmembrane β-Barrel.

Protein	Central hairpin	Central hairpin position	C-terminal residues	Last residue	Mortise and tenon motif	Trp position	Folding efficiency
OmpX	Present	4–5	YRF	F	Present	β4, β8	100%
O6A2	Present	2–3	YEF	F	Present	β2, β6	100%
O4A4	Absent	1, 8 a	RIN	N	Present	β4, β7	0%
O4A4F	Absent	1, 8 a	INF	F	Present	β4, β7	0%
O4A4SC	Present	4–5	YRF	F	Present	β4, β8	100%
O2A6	Present	6–7	QFN	N	Present	β2, β5	100%
O2A6A	Present	6–7	FNA	A	Present	β2, β5	30%
O2A6F	Present	6–7	FNF	F	Present	β2, β5	70%
A6O2	Present	2–3	EED	D	Present	β1, β6	100%
A4O4	Absent	1, 8 a	RIN	N	Present	β4, β8	0%
A4O4F	Absent	1, 8 a	INF	F	Present	β4, β8	0%
A4O4SC	Present	4–5	YRF	F	Present	β3, β8	100%
A2O6	Present	6–7	NPM	M	Present	β2, β6	100%
Ail	Present	4–5	YRF	F	Present	β3, β8	100%

^aCentral hairpin sequence not in tandem, and assembles upon barrel folding.

Interface tryptophans play an anchoring role in transmembrane proteins, and are vital as boundary-defining entities. ^28,29,37,38 We have previously observed that both OmpX ^28,39 and Ail ²⁹ require interface tryptophans for folding and stability. In OmpX, Trp76 present in strand β4 nucleates β-barrel folding in micelles and Trp140 of strand β8 serves as the postfolding anchor. ²⁸ In contrast, Trp42 of strand β3 and Trp149 of β8 together drive the assembly of Ail in micellar environments. ²⁹ With the exception of A4O4^SC and O4A4^SC, the parent strand positions, and thereby the positions of the tryptophans, are not conserved in the chimeras. The three-state unfolding we observe in a few mutants may arise due to the change in tryptophan position. For example, single tryptophan mutants of Ail exhibit three-state unfolding profiles in high DPRs of ~7000:1. ²⁹ Whether the absolute position of the interface tryptophan influences the (un)folding pathway we observe requires further investigation. Nevertheless, we find that the presence of two interface tryptophans is sufficient for assembly of the barrel chimeras, and the strand it is located on is less important.

Another interesting observation of our results is the overlapping HSQC resonances of the chimeras with both their parent proteins. The ¹H and ¹⁵N chemical shifts are sensitive to the backbone ϕ values. In contrast to the seemingly altered behavior of the tryptophan fluorescence, the backbone conformation of the parent constructs are retained in the hybrid chimeric barrels. We surmise from the overlapping HSQCs that irrespective of strand hybridization, each residue in an OMP can retain the backbone conformation of its parent protein. If transmembrane β-barrels arose from gene duplication, our observation raises the possibility that the ancestral barrel structure had considerable local conformational symmetry. Whether this facilitated evolutionary divergence of the OMP polypeptide sequence while retaining the overall scaffold structure is presently unclear. Further work correlating OMP ancestral sequences with the process of scaffold assembly ^11,40,41 may provide better insight on OMP evolution.

In the bacterial outer membrane, the BAM (barrel assembly machinery) complex assembles OMPs. Various reports suggest that the C-terminal β-signal motif plays an important role in the recognition of substrate OMPs by BAM, facilitating barrel insertion. ^8,13–15,42 The C-terminal residues also contribute to the overall stability of the OMP. ¹⁷ However, there is presently limited evidence that describes the importance of the central hairpin in OMP folding in vitro, ^7,19 and there is no evidence linking the central hairpin with BAM-assisted folding. ⁸ Based on our results and other work, ^7,19 we speculate that the central hairpin might help in the β-augmentation process ^42–45 in vivo, by stabilizing an intermediate in the folding pathway. It would be noteworthy to monitor the efficiency and rate of assisted folding of the chimeric OMPs in the presence of BAM.

In conclusion, our results suggest that the driving factor in OMP folding in vitro is the retention of strands corresponding to the central hairpin in tandem. Our conclusion is supported by previous observations that the central β-hairpins retained residual structure when OmpX was unfolded in urea. ^7,19 Hence, one could envision that the sequence corresponding to the central hairpin can constitute the folding nucleus and must be important for OMP assembly. While the sequence of the central hairpin is conserved across several bacterial OMPs, ^7,29 whether this sequence is important for in vivo assembly ^42–45 remains to be established. Other auxiliary factors such as the C-terminal β-signal, interface aromatic residues, and strand length, additionally facilitate folding. Differences in the latter can alter the folding pathway and assembly time for individual OMP sequences. ^{17,36,46–48} A proposed model for OmpX assembly in lipidic micelles is believed to follow a stochastic pathway, ²¹ while a concerted multistep assembly has been proposed in lipid membranes for OmpA. ^9–11,41 In contrast to the stochastic model, ²¹ our results point to a nonstochastic assembly of OmpX and Ail in lipidic micelles, as seen in lipidic vesicles, ^9,11,41 with a likely nucleation of folding at the conserved sequences and motifs. Therefore, we hypothesize that the folding of transmembrane β-barrels is not necessarily stochastic, but could follow directed folding in micellar systems that is predefined by the assembly of the central hairpin strands. Overall, our findings are in excellent agreement with previous studies that demonstrate both the versatility of the OMP sequence and the need for core motifs. The tolerance of OMPs to sequence heterogeneity is a useful tool for engineering novel OMP barrels. The retention of the central hairpins in tandem is sufficient to drive OMP folding, providing a handle for the de novo design and modeling of novel OMP sequences for application in bionanotechnology.

Abbreviations

Ail

Attachment and invasion locus

CD

circular dichroism

C_m

midpoint of chemical denaturation

DPC

n-dodecylphospho-choline

DPR

detergent-to-protein ratio

ΔG⁰ _U

Gibbs free energy of protein unfolding

GdnHCl

guanidine hydrochloride

HSQC-TROSY

heteronuclear single quantum coherence–transverse relaxation optimized spectroscopy

m value

change in accessible surface area upon protein unfolding

MRE

molar residue ellipticity

OMP

outer membrane protein

OmpX

outer membrane protein X

PK

proteinase K