Engineering a Hyperstable Yersinia pestis Outer Membrane Protein Ail Using Thermodynamic Design

Abstract

Development of viable therapeutics to effectively combat tier I pneumopathogens such as Yersinia pestis requires a thorough understanding of proteins vital for pathogenicity. The host invasion protein Ail, although indispensable for Yersinia pathogenesis, has evaded detailed characterization, as it is an outer membrane protein with intrinsically low stability and high aggregation propensity. Here, we identify molecular elements of the metastable Ail structure that considerably alter protein−lipid and intraprotein thermodynamics. In addition, we find that four residues Q⁵⁰, L⁸⁸, L⁹², and A⁹⁴ contribute additively to the lowered stability of Ail, and their conserved substitution is sufficient to reengineer Ail to Out14, a thermodynamically hyperstable low-aggregation variant with a functional scaffold. Interestingly, Ail also shows two (parallel) folding pathways, which has not yet been reported for β-barrel membrane proteins. Additionally, we identify the molecular mechanism of enhanced thermodynamic stability of Out14. We show that this enhanced stability of Out14 is due to a favorable change in the nonpolar accessible surface, and the accumulation of a kinetically accelerated off-pathway folding intermediate, which is absent in wild-type Ail. Such engineered hyperstable Ail β-barrels can be harnessed for targeted drug screening and developing medical countermeasures against Yersiniae. Application of similar strategies will help design effective translational therapeutics to combat biopathogens.

Introduction

The successful development of therapeutic drugs for major microbial pathogens necessitates a systematic biochemical and structural analysis of cell surface proteins responsible for virulence phenotypes. The adherence and invasion locus (Ail; also referred to as attachment invasion locus) protein, found in the outer membrane of the deadly pathogen and category A bioterrorism agent Yersinia pestis, ^1,2 is one such excellent prototype to develop medical countermeasures for plague. ^1,3 Ail belongs to the Ail/OmpX/PagC/Lom family of outer membrane proteins (OMPs). Ail is responsible for host cell adhesion, invasion, and delivery of the Yersinia outer proteins to the host tissue, inhibiting the host inflammatory response as well as providing resistance to bactericidal properties of host complement-mediated killing. ^4–6 Ail preferentially binds extracellular matrix components, including laminin, fibronectin, and heparin, through residues found in the extramembranous loop regions. Mammalian cell invasion is linked directly with highly upregulated levels of Ail, which confers serum resistance and cell adhesion to Yersinia, and is vital for the survival of this pathogen in the mammalian host. ^7,8 Hence, our success in developing potent vaccines for plague ⁹ requires a detailed understanding of Ail biology.

Residues responsible for Ail function, namely adhesion to and invasion of host cells, are housed in the extramembranous loops. ¹⁰ The transmembrane β-barrel domain of this eight-stranded OMP comprises 67 residues that form a coevolved network of intraprotein and protein-lipid interactions ^5,11 and function primarily as the membrane-anchoring scaffold. The closest evolutionary ancestor and structural homologue of Ail is OmpX of Escherichia coli, with which it shares considerable sequence identity. ^5,10 Unlike OmpX, Ail is indispensable for the pathogenicity of its host. However, Ail displays physicochemical characteristics that distinguish this protein from other OMPs. For example, Ail is one of the energetically least stable OMPs characterized thus far. Additionally, Ail is less structured and a highly aggregation-prone membrane protein. ^10,11 Therefore, studies aimed at re-engineering this vital protein as a viable pharmacological target have remained challenging. ⁹

Identifying elements required to stabilize the biologically relevant structure is vital to engineer novel protein folds and develop targeted therapeutics or hyperstable scaffolds with improved chemical functions for industrial applications. ^12,13 Studies of evolutionarily related proteins have revealed that protein sequences accumulate modifications to accommodate microenvironment variations and cellular stress as well as to acquire newer or enhanced functions. ¹⁴ A common outcome of such sequence variations is lowered scaffold thermodynamics. Previous studies on OMPs including Ail indicate that the magnitude and strength of interactions formed in the transmembrane region can regulate OMP stability. ^13,15,16 We reasoned that in vitro directed mutagenesis of Ail based on the analysis of evolutionary sequence divergence in its transmembrane region can successfully identify molecular elements that stabilize the native β-barrel fold and destabilize non-native states.

While directed evolution ¹⁷ and rational design ¹⁸ are popular protein engineering strategies, consensus sequence design is preferred when one aims to increase the thermodynamic stability of the target protein. ^19,20 Here, using Ail as our prototypical OMP, we apply a semiconsensus directed approach to identify elements essential for the design of a hyperstable transmembrane β-barrel. We successfully couple residue-tailoring and thermodynamic re-engineering to enhance Ail stability while preserving Ail function such as heparin binding. This engineered Ail variant can serve as a potent biomolecule for drug screens and developing plague vaccines. Through this approach, we highlight the utility of minimal sequence modifications and conserved substitutions to engineer hyperstable OMPs with preserved function. Our findings can be broadened to design vaccine-based immunization approaches for other bacterial diseases, especially those caused by Gram-negative pathogens including Salmonella, Vibrio, Helicobacter, and Pseudomonas.

Experimental Methods

Mutant Design, Generation, and Folding

Residues variant in the transmembrane lipid-facing (Out-series) and core (In-series) regions of the Y. pestis Ail β-barrel were identified using E. coli OmpX as the template. ail (encoding the mature protein) cloned in pET3a was used to generate single, double, triple, and multimutants of the Inand Out-series using tPCR. Protein production as inclusion bodies was achieved using E. coli C41 cells. ^11,21 See SI for details.

Wild-type Ail (WT), Out-series, and In-series mutants were folded in bicelles prepared using DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine; PC), DMPE (1,2-dimyristoyl-sn-glycero-3-phos-phoethanolamine; PE), DMPG (1,2-dimyristoyl-sn-glycero-3-phos-pho-(1′-rac-glycerol); PG), and DPC. These conditions were identified from exhaustive screens for high folding efficiency, lowest aggregation propensity, and prolonged sample stability (see SI methods for details). Largely, folding was achieved by rapid 10-fold dilution of 840 μM protein (unfolded in 8 M GdnHCl and 20 mM Tris-HCl, pH 8.5) into the folding reaction containing q = 1 48:12:60 PCPE bicelles (48 mM DMPC/12 mM DMPE/60 mM DPC) or q = 0.5 30:24:6:120 PEPG bicelles (30 mM DMPC/24 mM DMPE/6 mM DMPG/120 mM DPC) prepared in 20 mM Tris-HCl, pH 8.5. A final lipid-protein ratio (LPR) of ~700:1 was achieved in both bicelles. Folding efficiency was assessed using the electrophoretic mobility shift assay, pulse proteolysis, fluorescence emission profiles, far-UV circular dichroism, and scattering measurements. ^11,22–24 See SI for details.

Equilibrium (Un)folding Titrations

Folded and unfolded protein stocks were diluted 6-fold in a premixed guanidine hydrochloride (GdnHCl) gradient to initiate unfolding and folding, respectively. The final reactions contained 14 μM Ail in 8:2:10 PCPE bicelles or 5:4:1:20 PEPG bicelles and 20 mM Tris-HCl, pH 8.5, in varying [GdnHCl] from 1.0−6.3 M, based on the Ail variant. The final LPR of ~700:1 was maintained throughout. All reactions were incubated at 25 °C. The progress of each reaction was monitored by measuring the change in fluorescence of the two intrinsic Trp residues between 320−380 nm, using a λ_ex = 295 nm on a SpectraMax-M5 microplate reader (Molecular Devices, LLC). Postequilibrium data (120 h; recorded up to 360 h) were used to calculate the intensityadjusted average wavelength (<λ>) profiles (eq 1). This was fitted to a two-state folding model (eq 2) to derive the thermodynamic parameters, ^22,25,26 namely the absolute or apparent equilibrium free energy ( $\text{Δ}{G}_{\text{F}}^0$ or $\text{Δ}{G}_{\text{app},\text{F}}^0$ , respectively) and chemical denaturation midpoint (C _m) from at least two independent experiments. A global equilibrium M _eq,F (folding cooperativity) of −5.870 kcal mol⁻¹ M⁻¹ in PCPE and −5.318 kcal mol⁻¹ M⁻¹ in PEPG was used in the analysis. The change in free energy upon mutation ( $\text{ΔΔ}{G}_{\text{F}}^0$ or $\text{ΔΔ}{G}_{\text{app},\text{F}}^0$ ) was calculated as $\text{Δ}{G}_{\text{F}}^{0,\text{WT}}$ − $\text{Δ}{G}_{\text{F}}^{0,\text{MUT}}$ .

$$<\lambda >=\frac{{{\displaystyle \sum }}^{\text{}}\left(F_{\text{i}}{\lambda }_{\text{i}}\right)}{{{\displaystyle \sum }}^{\text{}}{F}_{\text{i}}}$$ (1)

$$y_{\text{O}}=\frac{\left(y_{\text{U}}+m_{\text{U}}[D]\right)+\frac{1}{Q_{\text{R}}}\left[\left(y_{\text{F}}+m_{\text{F}}[D]\right)\exp \left[-\left(\text{Δ}{G}_{\text{F}}^0+M_{\text{F}}[D]\right)/RT\right]\right]}{1+\frac{1}{Q_{\text{R}}}\exp \left[-\left(\text{Δ}{G}_{\text{F}}^0+M_{\text{F}}[D]\right)/RT\right]}$$ (2)

Here, y _O is the observed intensity, y _F and m _F are the intercept and slope of the linear baseline of the folded region of the two-state transition, and y _U and m _U are the intercept and slope of the linear baseline of the unfolded region. [D] is the denaturant concentration. A Q _R (correction factor) of 1 was used. ²² R is the gas constant (1.987 × 10⁻³ kcal mol⁻¹ K⁻¹), and T is the temperature in kelvin.

Folding−Unfolding and Interrupted Unfolding Kinetics

A 84 μM concentration of protein in 48:12:60 PCPE bicelles and 20 mM Tris-HCl, pH 8.5, containing 0.8 M (folded stock) or 5.0 M GdnHCl (unfolded stock), was diluted 6-fold into defined [GdnHCl] at 25 °C to initiate the unfolding and folding reactions, respectively. Each final reaction contained 14 μM protein, 8:2:10 PCPE bicelles, 20 mM Tris-HCl, pH 8.5, an LPR of ~700:1, and [GdnHCl] from 0.83−6.8 M in 0.08−0.12 M increments. The folding kinetics was monitored using total fluorescence emission on an SX20 stoppedflow spectrometer (Applied Photophysics Ltd., UK) using a λ_ex of 295 nm, U-360 bandpass filter, and sampling period of 12.5 μs. The dead time was measured as 2.5 ms. The first 50 ms of data was excluded from the fits, as it carried artifacts due to convection effects from mixing of viscous solutions. Multiphasic profiles of the mean kinetics traces (from 6 to 10 technical repeats) were fitted to exponential functions to obtain the rates of decay (k _d) and rise (k _f1 and k _f2) and the corresponding amplitudes A _d, A _f1, and A _f2. Unfolding kinetics were slower and were recorded by manual mixing on a FluoroMax 4 spectrofluorimeter (Horiba Jobin Yvon) at 327 nm and 1.0 s intervals using a data integration time of 0.5 s. Traces from three technical replicates were fitted independently to calculate the unfolding rates (k _u1 and k _u2) and the corresponding amplitudes (A _u1 and A _u2) and averaged. Chevron plots generated from the folding and unfolding rates were fitted to a two-state (eq 3) ²⁷ or three-state (eq 4) ²⁸ model (to account for the rollover in the folding arm). Folding and unfolding rate constants extrapolated to zero denaturant ( $(k_{\text{F}}^0=k_{\text{D}}\text{or}{k}_{\text{F2};}{k}_{\text{U}}^0=k_{\text{U1}}{\text{or}k}_{\text{U2}}),$ and the corresponding m values (m _F = m _D or m _F2; m _U = m _U1 or m _U2) thus obtained were used to calculate the kinetic free energy of unfolding ( $\text{Δ}{G}_{\text{kin}}^0$ ) using eq 5 (two-state). Additional parameters (k _I and m _I) were used to account for the rollover, in eq 6 (three-state). ²⁹

$$\ln k_{\text{obs}}=\ln \left(\exp \left(\ln k_{\text{F}}^0+m_{\text{F}}[\text{D}]\right)+\exp \left(\ln k_{\text{U}}^0+m_{\text{U}}[\text{D}]\right)\right)$$ (3)

$$\ln k_{\text{obs}}=\ln \left(\frac{\exp \left(\ln k_{\text{F}}^0+m_{\text{F}}[D]\right)}{1+\exp \left(-\left(\ln k_{\text{I}}^0+m_{\text{I}}[D]\right)\right)}+\exp \left(\ln k_{\text{U}}^0+m_{\text{U}}[D]\right)\right)$$ (4)

$$\text{Δ}{G}_{\text{kin}}^0=-RT\left(\ln k_{\text{F}}^0-\ln k_{\text{U}}^0\right)$$ (5)

$$\text{Δ}{G}_{kin}^0=-RT\left(\ln k_{\text{F}}^0+\ln k_{\text{I}}^0-\ln k_{\text{U}}^0\right)$$ (6)

Interrupted unfolding kinetics were carried out using the SX20 sequential mixing setup. Here, 168 μM folded protein in PCPE bicelles (96:24:120), containing 1.6 M GdnHCl and 20 mM Tris-HCl, pH 8.5, was diluted 2-fold in a final [GdnHCl] of 4.8 M and incubated for different aging (delay) times (t _age) from 1 ms−999 s. A 30 μL solution from the delay loop was diluted 6-fold with buffer to achieve 14 μM protein, 8:2:10 PCPE bicelles, an LPR of 700:1, and a final [GdnHCl] of 0.8 or 1.2 M in the cuvette. The refolding traces were fitted to an exponential function to obtain the rates and corresponding amplitudes.

Assessment of Heparin Binding

The binding efficacy of Ail-WT and select Out mutants to heparin were assessed using 70 μM stock protein folded in PCPE bicelles. Stocks were diluted 5-fold to reduce the final GdnHCl to 0.16 M and promote protein binding on a HiTrap heparin HP column (GE Healthcare). Elution was achieved with FPLC using a linear NaCl gradient. Fractions were checked on 12% SDS-PAGE gels, and protein bands were visualized using silver staining.

Molecular Dynamics Simulations

The Ail structure was modeled using the SWISS-MODEL web tool. ³⁰ In order to match the experimental conditions, electrostatics corresponding to a pH of 8.5 was incorporated in the Ail structure using the PDB2PQR server. ³¹ The PQR file generated therein was oriented in the membrane using the PPM server available in CHARMM-GUI. ³² Ail-WT, Out14, and Out16 (Ail variants were generated in CHARMM-GUI by employing the mutation option) were assembled in a rectangular box in three different lipid bilayer membranes, as follows: (1) 50% DMPC, 40% DMPE, and 10% DMPG, with a total of 80 molecules per leaflet, and a box size of 75 × 75 × 105 Å; (2) 75% DMPE and 25% DMPG (total of 80 molecules) in the inner leaflet, 25 molecules of E. coli lipopolysaccharide (LPS) containing the R1 core oligosaccharide with 5 repeats of the O-antigen in the outer leaflet, and a box size of 75 × 75 × 185 Å; (3) 75% DMPE and 25% DMPG (total of 80 molecules) in the inner leaflet, 25 molecules of Y. pestis LPS with the type I core sequence in the outer leaflet, and a box size of 75 × 75 × 105 Å. Each bilayer was built using the membrane builder tool of CHARMM-GUI. ³³ A water layer of 20 Å thickness was maintained on either side of the bilayer to hydrate the system. The input protein-lipid assembly was generated with the CHARMM36m force field, NPT (constant particle number, pressure, and temperature) ensemble, and a temperature of 25 °C. ³⁴ The system was first equilibrated with six energy minimization steps, and the output file thus generated was used to run a final 500−550 ns all-atom molecular dynamics simulation a step size of 2 fs, using GROMACS v20.5. ^35–37 The final trajectory file generated after the simulation was analyzed using GROMACS v20.5 and rendered using PyMOL v2.1.0. ³⁸ In order to capture the effect of sequence variation on the barrel dynamics, regions with higher deviations in the root mean square deviation (RMSD) plot were excluded in the analysis of root mean square fluctuation (RMSF), radius of gyration (R _g), and free energy landscape (FEL). ³⁹

Results

Directed Design of Ail Transmembrane Domain

Y. pestis Ail shares homology with other pathogenic bacteria including Y. enterocolitica, Y. pseudotuberculosis, and S. typhimurium. Crystal and NMR structures of Ail reveal an eight-stranded β-barrel similar to other outer membrane proteins (OMPs) but with extended and highly dynamic extramembranous loops that harbor binding sites for heparin, laminin, and other ECM components. ^10,40 It is generally accepted that residues in the transmembrane domain (TMD) of an OMP guide its folding and stability. ^{13,25,41 42} To identify critical target residues in the TMD of Ail, we mapped consensus sites using OmpX/Lom family proteins that exhibit structural or functional homology with Ail (Figure S1), for both hydrophobic lipid-facing (protein−lipid interactions) and core (intramolecular packing interactions) residues. E. coli OmpX, despite evolutionary divergence between both proteins, showed the highest similarity with Ail, with ~65% sequence identity in the TMD (Figure 1A). However, OmpX differs significantly from Ail in its physicochemical characteristics including extremely high thermal stability (T_m ≈ 107 °C), ⁴³ resistance to unfolding, and high thermodynamic stability (e.g., ~5.69 kcal mol⁻¹ in low concentrations (10 mM) of lauryldimethylamine oxide (LDAO) micelles ⁴⁴ ). Ail, on the other hand, is sensitive to thermal (T _m ≈ 55 °C) and chemical perturbants (stability of ~3.67 kcal mol⁻¹ in 20 mM LDAO), ⁴⁵ is kinetically trapped, and is highly aggregation−prone. ¹¹ The secondary structure content in micelles is higher for OmpX (per-residue molar ellipticity at ~214 nm = ~1 × 10 ⁴ deg cm ² dmol⁻¹ res⁻¹ for OmpX and ~0.6 × 10 ⁴ deg cm ² dmol⁻¹ res⁻¹ for Ail). ⁴⁶ Hence, with OmpX TMD as our template, we probed for Ail residues that can augment the physicochemical properties of the Ail barrel.

Figure 1. Fingerprint Ail transmembrane residues identified for mutagenesis through consensus design.

(A) Multiple sequence alignment of Ail and OmpX with the parent mutants In1 for the barrel core-facing (top, purple) and Out1 for the lipid-facing (bottom, teal) residues. Important strand and residue numbers are indicated. (B) Positions of the 9 residues mutated in In1 (left) and 14 residues replaced in Out1 (right) are mapped on the structure of Ail (PDB ID: 3QRA) as spheres. (C) Electrophoretic mobility shift assay and pulse proteolysis (±PK; proteinase K) measurements in PCPE and PEPG bicelles establish folding efficiency of Ail. Unfolded (UF) Ail displays a gel mobility corresponding to its MW (~17.0 kDa) and is susceptible to PK. Folded Ail (F) shows retarded gel mobility and is resistant to proteolysis. Due to its protease susceptibility in PEPG, the folding of In1 was additionally confirmed using spectroscopic methods (Figures 1D and S1). (D) Far-UV circular dichroism profiles of folded Ail-WT, and the parent mutants In1 and Out1, in PCPE (left) and PEPG (right) bicelles, show comparable folding efficiency for all three proteins. ME: molar ellipticity. (E) Equilibrium folding (red) and unfolding (black) <λ> profiles highlighting the near two-state transitions for Ail-WT in PCPE (right) and PEPG (left) bicelles recorded at 120 h. Error bars represent s.d. from at least two independent experiments. (F) Representative folding equilibrium profiles of Ail-WT and the two parent variants Out1 and In1 in PCPE (left) and PEPG (right) at 120 h. Fits (solid lines) of the data to a two-state function used a shared M_eq,F (PCPE: − 5.870 kcal mol⁻¹ M⁻¹; PEPG: − 5.318 kcal mol⁻¹ M⁻¹) to deduce the thermodynamic parameters.(6)

The consensus analysis identifies 14 lipid-facing and 9 core residues as different between Ail and OmpX (Figures 1A and S1). As molecular energetics of protein−lipid interactions are distinct from intraprotein interaction networks, we generated two parent variants of Ail, namely Out1 and In1 (Figure 1A,B). While Out1 has 14 lipid-facing residues of Ail substituted with the corresponding OmpX residues, In1 bears similar substitutions in the 9 core-facing residues (Figures 1A,B and S1). Next, we folded Ail-WT (wild-type Ail protein), Out1, and In1 in PE-doped PC bicelles (PCPE) or PE- and PG doped PC bicelles (PEPG) (see Experimental Methods and SI methods) and verified the folded state of these proteins using electrophoretic mobility and pulse proteolysis assays, fluorescence spectroscopy, and far-UV circular dichroism measurements (Figures 1C,D and S2-S4).

Equilibrium thermodynamic measurements of Ail monitored using the change in fluorescence emission of its two intrinsic interface tryptophans (Figures S3 and S4) provide us with a reversible transition (path-independent folding) between its GdnHCl-denatured state and the lipid-stabilized folded state (Figure 1E). The transition is two-state (eq 2), with rearrangement of the unfolded form to the folded structure occurring with no detectable intermediates (Figure 1E,F). Furthermore, a vital contribution of the kinetic component in Ail stability is observed by comparing the equilibration times (monitored by the change in C_m with time), which is faster for the folding process (~72 h) than the unfolding process (≥120 h) particularly in PCPE bicelles (Figure S3).

We measure a similar end-state folding cooperativity in both lipidic environments (−5.87 kcal mol⁻¹ M⁻¹ in PCPE and −5.32 kcal mol⁻¹ M⁻¹ in PEPG). The calculated free energy of folding ( $\text{Δ}{G}_{\text{F}}^0$ ) for Ail-WT is −16.54 ± 0.32 kcal mol⁻¹ in PCPE and −14.78 ± 0.41 kcal mol⁻¹ in PEPG. While hysteresis is more evident in PCPE than PEPG bicelles (Figures 1E and S1), a lower stability in PEPG (~1.75 kcal mol⁻¹) could arise from changes in the curvature energy or lateral pressure due to phosphatidylglycerol. ⁴⁷ Unexpectedly, modification of the 9 core residues further lowers the stability of the In1 variant of Ail by ~1.5−1.75 kcal mol⁻¹ in both PCPE and PEPG (Figures 1F and S4). An incomplete electrophoretic mobility along with protease susceptibility (Figures 1C and S2), pronounced hysteresis in the unfolding titration in both PCPE and PEPG (Figure S4B), and a three-state unfolding transition in PEPG (Figure S4B), reveals that the mutations have a noticeable effect on the folded barrel’s energetics. In contrast, modifying the intrinsic lipid-facing residues of Ail increases barrel stability. In both lipids, Out1 shows path-independence and an increase in the $\text{Δ}{G}_{\text{F}}^0$ by >2.5−3.0 kcal mol⁻¹ (Figures 1F and S4B).

The close agreement of thermodynamic data from both lipidic membranes supports our conclusion that the physicochemical nature of the vicinal lipids and the addition of phosphatidylglycerol do not remarkably alter Ail’s intrinsic thermodynamic features. The secondary structure content remains unaltered between Ail-WT, In1, and Out1, despite the substitutions (Figure 1D), suggesting that intraprotein interactions in In1 and protein−lipid interactions in Out1 are altered without substantial structural changes. The magnitude of free energy change of In1 and Out1 correlates poorly with the number of residues that are substituted. Therefore, the measured $\text{ΔΔ}{G}_{\text{F}}^0$ between Ail-WT and its variants either stems from only a subset of the residues or indicates compensatory mutations.

Residues Unique to Ail Core Control Path-Independent Barrel Folding and Stability

Comparison of $\text{Δ}{G}_{\text{F}}^0$ of WT versus its In1 variant shows that substitutions in the Ail barrel core are destabilizing and affect the path independence of unfolding. Put simply, folded In1 is kinetically trapped in bicelles. To identify core residues that effect the switch from thermodynamic to kinetic control, we generated the In-series variants of Ail through selective residue replacements (Figures 2 and S5). Thermodynamic characterization of these variants reveals that restoring three of the nine OmpX residues to the intrinsic Ail residues is sufficient to augment β-barrel stability (Figures 2A, S6, and Table S1). As inferred from the $\text{ΔΔ}{G}_{\text{F}}^0$ , S⁷³ and V⁷⁹ of strand β4 and A¹¹⁷ of β6 intrinsic to Ail are vital and sufficient to regulate Ail stability and barrel core packing efficiency. Comparison of the $\text{ΔΔ}{G}_{\text{F}}^0$ values across the In-series shows that the contributions follow the order V⁷⁹ ≥ S⁷³ > A¹¹⁷ (Figure 2B), and the effect of each mutation is additive. Hence, the intrinsic energetics of Ail core residues is largely unaltered by modifications to vicinal interactions.

Figure 2. Intrabarrel assembly and stability require stereospecific β4−β6 interactions intrinsic to Ail.

The sequence of Ail-WT (gray fills) and residues mutated to generate the In-series (purple outlines) are shown in the top panels in both (A) and (B). The measured $\text{Δ}{G}_{\text{app},\text{F}}^0$ (bottom, left) and the corresponding $\text{ΔΔ}{G}_{\text{app},\text{F}}^0$ (bottom, right) in PCPE (open histograms) and PEPG (filled histograms) are shown for specific mutants. Error bars denote s.d. from two independent experiments. Hysteresis is prominent in nearly all In variants (see Figure S6). $\text{ΔΔ}{G}_{\text{app},\text{F}}^0=\text{Δ}{G}_{\text{F}}^{0,\text{WT}}-\text{Δ}{G}_{\text{F}}^{0,\text{MUT}}$ . The additive contributions of S⁷³, V⁷⁹, and A¹¹⁷, intrinsic to strands β4 and β6 of Ail, are evident from the increase in $\text{Δ}{G}_{\text{app},\text{F}}^0$ and $\text{ΔΔ}{G}_{\text{app},\text{F}}^0$ (A), when these residues are restored in the Ail sequence. This is verified further in (B) by the observed destabilization of the Ail structure by point mutations S⁷³ → G and V⁷⁹ → A in strand β4. Intrinsic Ail residues that are important for Ail’s stability are indicated above the histograms in (A) and (B). Also see Figures S6 and S9 and Table S1 for the complete data.

By and large, lipid-facing residues are key contributors for the assembly and stability of several β-barrels. ^26,48,49 The conserved substitution of residues with an aliphatic polar nature within the β-barrel core typically results in minor rearrangements in local interactions and is usually welltolerated thermodynamically. ^42,50,51 Interestingly, however, we find that the core Ail residues S⁷³, V⁷⁹, and A¹¹⁷ of β4 and β6 are specifically required for high thermodynamic stability. Additionally, while substituting residues of strands β1, β3, β5, and β8 do not considerably alter the measured $\text{Δ}{G}_{\text{F}}^0$ (In2 in Figure 2A), the thermodynamic reversibility is affected substantially. All In-variants remain kinetically trapped, as deduced from the pronounced hysteresis and three-state unfolding profiles (Figure S6). In both PCPE and PEPG, hysteresis is particularly prominent in In2, In4, In6, and In7 (Figure S6), which carry common substitutions in β1 (E⁴/S⁶/ S⁸ → T), β3 (A⁴⁹ → T), β5 (L⁹¹ → V), and β8 (M¹⁵⁰ → I) (sequences in Figure S5). Hysteresis in PCPE is most pronounced in In3, in which all strands (except β6 and β7) carry substitutions (Figure S6A). Hence, core residues of Ail are as important as lipid-anchoring residues, as they dictate path-independent folding and overall stability of the protein. We thus conclude from our findings that barrel core residues can considerably modify OMP stability by affecting the kinetic component of the free energy landscape. Since none of the core substitutions increases Ail stability, we next examined the utility of lipid-facing residues in designing a hyperstable Ail β-barrel.

Enhancing Ail Thermodynamic Stability through Engineered Protein−Lipid Interactions

Next, we carried out site-specific replacements of the 14 lipid-facing residues distributed across strands β1 and β3−β8 to generate the Out-series variants of Ail (Figures 3 and S5). Using equilibrium measurements (Figure S7 and Tables S2 and S3), we narrowed down residues altering the stability of the Ail scaffold to I ⁷ and I⁹ of β1, Q⁵⁰ of β3, L⁸⁸, L⁹², and A⁹⁴ of β5, and L¹⁵¹ and A¹⁵³ of β8. Strands β4 (V⁷⁴ → I, F⁸⁰ → Y) and β7 (I¹³² → L, A¹³⁴ → F, Y¹³⁸ → Q) do not alter the Ail stability significantly (compare Out3, Out6, Out8 in Figure S8 and Table S2). Comparison of $\text{ΔΔ}{G}_{\text{F}}^0$ across substituents of the central strands shows that residues inherent to Ail β3 (Q⁵⁰) and β5 (L⁸⁸, L⁹², and A⁹⁴) destabilize the β-barrel, whereas a substitution corresponding to the OmpX residue (Q⁵⁰ → Y, L⁸⁸ → I, L⁹² → V, and A⁹⁴ → V; Out14) increases Ail stability (Figure 3A). We further validated this by generating selective substituents in strands β3 and β5. Comparing the $\text{Δ}{G}_{\text{F}}^0$ and $\text{ΔΔ}{G}_{\text{F}}^0$ of Out9, Out11, Out13−15 (Table S2) reveals that Y ⁵⁰ as well as the I⁸⁸−V⁹²−V⁹⁴ triad show a near-equal contribution of ~2.5−2.7 kcal mol⁻¹ each to Ail stability in both PCPE and PEPG (Figure 3A). Single residue substitutions in β5 further reveal an additive effect of the I⁸⁸−V⁹²−V⁹⁴ triad on Ail stability (compare $\text{ΔΔ}{G}_{\text{F}}^0$ of Out11 with the single residue mutants in Table S2). Overall, we conclude that replacing four lipid-facing residues of Ail with the corresponding OmpX residues is sufficient to increase Ail stability by ~4.5−5.5 kcal mol⁻¹ in both PCPE and PEPG bicelles (see global comparison of all In- and Out-series variants in Figure S9). An additive contribution of the secondary structure preference (β-sheet preference ⁵² follows the order: I ≥ V > Y > Q > L > A), hydrophobicity (Y > Q, V > A), and membrane transfer free energies ⁴² (L > I > Y > V > A > Q) together result in the generation of a hyperstable Ail β-barrel scaffold.

Figure 3. Protein−lipid interactions effected by specific OmpX residues enhance Ail stability.

The sequence of Ail-WT and site-specific replacements carried out in strands β1 and β3−β8 of Ail-WT to generate specific mutants of the lipid-facing residues (Out-series) are shown in the top panels (teal outlines). The measured $\text{Δ}{G}_{\text{F}}^0$ (bottom, left) and the corresponding $\text{ΔΔ}{G}_{\text{F}}^0$ (bottom, right) derived in PCPE (open histograms) and PEPG (filled histograms) bicelles are also provided for select mutants (complete data in Figures S7-S9 and Tables S2-S3). Error bars denote s.d. from two independent experiments. $\text{ΔΔ}{G}_{\text{F}}^0=\text{Δ}{G}_{\text{F}}^{0,\text{WT}}-\text{Δ}{G}_{\text{F}}^{0,\text{MUT}}$ . The results of modifications to the central strands are highlighted in (A), and modifications to the terminal strands are highlighted in (B). (A) Residues intrinsic to β3 and β5 in Ail destabilize the scaffold. The highest $\text{Δ}{G}_{\text{F}}^0$ obtained for Out14 combined with the results of the three point mutations (L88I, L92V, A94V) reveal the additive stabilizing effect of Y⁵⁰, I⁸⁸, V⁹², and V⁹⁴ derived from OmpX on the structure of Ail. (B) Residues intrinsic to β1 and β8 stabilize Ail. The comparison of Out12, Out16, and Out17 highlights the importance of I⁹ along with I ⁷ of β1 as well as L¹⁵¹ and A¹⁵³ of β8 in stabilizing Ail. A destabilizing effect is evident when these residues are substituted to the corresponding amino acids of OmpX. Similar results observed in PCPE and PEPG in both (A) and (B) confirm that the measured thermodynamics is independent of the physicochemical nature of the lipid environment (also see Figure S9C).

In interesting contrast, substitution of residues intrinsic to strands β1 and β8 lowers Ail stability (Figure 3B). Here, comparison of the $\text{Δ}{G}_{\text{F}}^0$ measured for Out variants with selective substitution at the terminal strands allows us to infer that residues I⁷ and I⁹ (β1) and L¹⁵¹ and A¹⁵³ (β8) are vital for barrel stability (Out12, Out16, Out17), while mutations in β7 (see Out4) are well-tolerated by Ail. Additionally, these substitutions in β1 and β8 (in Out2, Out12, Out16, Out17) alter the cooperativity of unfolding, introduce hysteresis, and lead to the accumulation of an unfolding intermediate (Figure S7A). Hysteresis and three-state (un)folding is specifically prominent upon replacement of I⁹ → G in β1(Figure S7A,C). Hence, molecular interactions of residues intrinsic to Ail β1 and β8 are important for both barrel stability and folding, and their substitution is deleterious for barrel stability. Additional analysis of $\text{ΔΔ}{G}_{\text{F}}^0$ of the point mutants reveal that the scaffold destabilization is largely due to I⁹ → G, with an additive destabilization from I ⁷ → V (residues in β1 are more important; Figure 3B). I⁹ is located at the center of β1 and not only is a membrane-anchoring midplane residue but also plays a vital role in barrel closure through H-bonding with β8. Our findings reveal that the replacement of Ile, a residue with optimal β-sheet ϕ−ψ values and hydrophobicity, with Gly (highest ϕ−ψ conformational occupancy and poor hydrophobicity) is sufficient to lower Ail stability by ~2.5−3.0 kcal mol⁻¹ (Table S2). The observed effect is irrespective of the lipid composition (see Figure S9C for the correlation of $\text{Δ}{G}_{\text{F}}^0$ between PCPE and PEPG).

To assess if alterations in the Ail sequence affect its function, we examined heparin binding of WT and select Out-series variants folded in PCPE bicelles. As anticipated, all proteins exhibited comparable binding to heparin (Figure S10), confirming that mutations in the transmembrane region and the associated alterations in protein stability do not affect Ail function. Taken together, the results from the substitutions of lipid-facing Ail residues allow us to propose that I⁹ of β1is essential for path-independent two-state (un)folding of Ail and scaffold stability. In addition, molecular fine-tuning of Ail stability through Q⁵⁰ → Y, L⁸⁸ → I, L⁹² → V, and A⁹⁴ → V is regulated largely by the secondary structure propensity of these residues and their hydropathy. The substantial change we observe in the protein’s stability despite the conserved substitution of nonpolar residues is unprecedented for an OMP and provides a simplified approach for membrane protein re-engineering while also preserving the protein’s function.

Kinetically Accelerated Assembly of Hyperstable Ail through Parallel Pathways

A key mechanism by which lipid-facing residues of membrane proteins increase scaffold stability is by altering protein−lipid packing energetics. This influences the folding pathway, since a vital step in OMP assembly is the interaction of transmembrane residues with the lipid membrane. To map whether substitutions stabilizing the Ail barrel also influence barrel assembly, we studied the (un)folding kinetics of Ail-WT and the Out-series in PCPE bicelles. While equilibrium in the (un)folding transition is attained after prolonged incubation, Ail assembly in low [GdnHCl] is rapid. Early events in Ail folding are completed in ~100 s (Figures 4A and S11A,B), and the measured folding in PCPE bicelles is more rapid than in DPC micelles. ²³ Other bacterial OMPs such as PagP, ⁵³ OmpX, ⁵⁴ and OmpA ⁵⁵ exhibit monophasic folding rates in various lipidic systems, while Ail assembly is multiphasic.

Figure 4. Barrel stabilization alters parallel folding pathway of Ail to multistep scheme.

(A) Representative folding kinetic (FK; left panel; 1.15 M GdnHCl) and unfolding kinetic (UK; right panel; 5.36 M GdnHCl) traces of Ail-WT in PCPE bicelles (cyan open circles). We obtain three folding (k _f1, k _d, and k _f2) and two unfolding (k _u1 and k _u2) rates for Ail-WT (fits are shown as solid lines). (B) Chevron plots for WT (left) and Out1 (right) with fits shown as solid lines. Kinetics for the WT can be explained using a two-state model, but Out1 requires a three-state mechanism to account for the rollover (see Figures S12 and S13 for the complete data). The assembly kinetics of WT is slower than Out1, allowing for the capture of k _f1 in low [GdnHCl]. k _f1 could also represent a hyperfluorescent folding intermediate in the fast pathway. Rates from the single and double jump (interrupted unfolding with t _age = 600 s at 4.8 M GdnHCl, followed by refolding at 0.8 M GdnHCl; DJ_RF1 and DJ_RF2) experiments are similar, revealing the existence of a parallel folding pathway in Ail. (C) The measured $\text{ΔΔ}{G}_{\text{kin}}^0$ for the fast (N → I₁; top panel) and the slow (N → I₂; bottom panel) rates derived in PCPE bicelles for selective Out-series mutants (complete data in Figure S16 and Tables S4-S5). $\text{ΔΔ}{G}_{\text{kin}}^0=\text{Δ}{G}_{\text{kin}}^{0,\text{WT}}-\text{Δ}{G}_{\text{kin}}^{0,\text{MUT}}$ . (D) Proposed folding scheme for Ail-WT and Out1 proteins. Interrupted unfolding kinetics (Figure S15) suggests that the refolding rate is largely invariant with time, supporting the presence of parallel pathways in Ail folding and at least two intermediates (I₁ and I₂) linking the native (N) and unfolded (U) states of the protein. Out1 additionally accumulates an off-pathway intermediate I′ during assembly in both pathways, accounting for the rollover in the folding arms.

We obtain three distinct phases in Ail folding (Figures 4A and S11). This includes a rapid burst phase (k _f1; < 1s) observed only in low [GdnHCl], suggesting the accumulation of an early folding intermediate. k _f1 is linked with a second decay phase (k _d; <10 s), while a slow rate (k _f2; <100 s) is also measured. k _f1/k _d and k _f2 together lead to formation of the native structure (N). Unfolding involves two phases, namely a fast (k _u1; <10 s) and slow (k _u2; ~500 s) phase, giving rise to the unfolded (U) protein. At intermediate [GdnHCl] close to the C _m, the folding rates are too slow to be captured and are a characteristic of a slow equilibrium behavior of the system. ⁵⁶ We obtain two chevron profiles by combining k _d and k _u1 as well as k _f2 and k _u2 (Figures 4B and S11C). Similar chevron profiles for Ail-WT and all 14 Out-series variants establish the occurrence of at least two on-pathway intermediates (I₁ and I₂) in the early assembly process of Ail (Figures 4B and S12-S14). While membrane proteins are known to undergo stepwise assembly in the membrane, ^{23,41,53 57} we believe that the multiphasic behavior we observe for the early folding events of Ail has not been reported for any OMP. Here, I₁ and I₂ could accumulate in a sequential assembly mechanism (N ↔ I₁ ↔ I₂ ↔ U) or represent parallel folding pathways (N ↔ I₁ ↔ U and N ↔ I₂ ↔ U). ⁵⁸

To identify and validate the folding pathway of Ail, we carried out interrupted unfolding kinetics (double jump (DJ) measurements; Figures 4B and S15). Interrupted unfolding is achieved by subjecting the folded protein to unfolding for a defined delay time (t _age), following which it is allowed to refold. Measurement of a single rate generally supports sequential assembly, whereas the observation of both rates indicates parallel pathways. Interestingly, Ail exhibits two refolding rates that correspond well with the k _f1/k _d and k _f2 of the single jump experiment (Figure 4B). Further, these rates show no change with increase in t_age (Figure S15), establishing two parallel pathways involving the two independent intermediates I₁ and I₂ for Ail folding. Both I₁ and I₂ represent global folding intermediates that are independent of the residues substituted between Ail-WT and its Out variants (Figures S12-S14). Formation of these intermediates is both rapid (within seconds) and stable with time.

The $\text{Δ}{G}_{\text{kin},1}^0$ and $\text{Δ}{G}_{\text{kin},2}^0$ (folding free energy for N ↔ I₁ ↔ U and N ↔ I₂ ↔ U, respectively) we measured are similar, suggesting that intermediates of both pathways have comparable stability and near equal populations (see (un)-folding amplitudes compared in Figure S14). For Ail-WT, we obtain $\text{Δ}{G}_{\text{kin},1}^0$ of −5.78 ± 0.46 kcal mol⁻¹ and $\text{Δ}{G}_{\text{kin},2}^0$ of −5.95 ± 0.45 kcal mol⁻¹ (Table S4). The measured total change in $\text{Δ}{G}_{\text{kin}}^0$ is lower than the $\text{Δ}{G}_{\text{eq}}^0$ by ~4.80 kcal mol⁻¹, which could be attributed to slow barrel rearrangement steps of either pathway, in the membrane. Interestingly, however, the $\text{Δ}{G}_{\text{kin},1}^0$ and the $\text{Δ}{G}_{\text{kin},2}^0$ for several Out variants is higher than Ail-WT, which arises from the faster folding rates (higher k _f1/ k _d and k _f2) and increase in folding cooperativity (higher m _f and m _u values of both pathways) for these engineered barrels (Figures 4C, S12, and S13 and Tables S4−S5). Furthermore, these variants display a striking rollover in both the folding arms, with no alteration in the populations of either pathway (Figures 4B and S13). While W149 of Ail is known to exhibit hyperfluorescence ²³ during folding, the rollover is absent in Ail-WT folding (see Figure S11 for details). The end-state structure of these variants is also similar to Ail-WT (see Figure 1D). Therefore, the rollover suggests either the accumulation of an off-pathway collapsed state (I′) in one or both pathways or a moving transition state (Hammond behavior) as a consequence of the substitutions. ^59–62

The Tanford β (β _T) value, which is a measure of the compactness of the transition state, is ≥0.5 for both N → I₁ and N → I₂ for WT and the Out variants (Tables S4 and S5), indicating that the structured transition state is largely unaffected by the substitutions. Hence, the rapid assembly kinetics for the Out-series variants arises from stronger barrel−lipid interactions that additionally lead to a collapsed I′ state. Further rearrangement of I′ → I → N at lower [GdnHCl] completes the folding of these Out variants (Figures 4D and S16). Correlating the substitutions across the Out-series with the rollover suggest that residues L⁸⁸ → I and A⁹⁴ → V ofβ5 contribute to the accumulation of I₁′ and I₁′, respectively (Figures S12B-S13). An auxiliary influence of Q⁵⁰ → Y(β3) and L¹¹⁶ → F(β6) on I′ formation may also exist. Nine of the ten stabilizing Out-variants (except Out9; Figures S12B-S13 and Table S5) show kinetically accelerated folding despite the accumulation of I′ in one or both pathways, resulting in a 2−4-fold increase in the $\text{Δ}{G}_{\text{kin},1}^0$ and $\text{Δ}{G}_{\text{kin},2}^0$ . In Ail-WT, barrel folding is slow (primarily due to L⁸⁸ and A⁹⁴ in β5), thereby accounting for the lower k _f1/k _d and k _f2 and the corresponding $\text{ΔΔ}{G}_{\text{kin},1}^0$ and $\text{ΔΔ}{G}_{\text{kin},2}^0$ (Figure 4C). Finally, comparison of the free energy values obtained from kinetic and equilibrium measurements (Figure S17) further corroborates our conclusion that Y⁵⁰, I⁸⁸, V⁹², and V⁹⁴ accelerate both β-barrel assembly and end-state stability.

Lowered Barrel Dynamics Stabilizes Engineered Ail

To address the origin of the increase or decrease in β-barrel stability of select Out-series variants with respect to Ail-WT, we carried out all-atom molecular dynamics simulations (MDS) of Ail-WT, Out14 (stabilizing) (Figure 5A), and Out16 (destabilizing), in three lipidic membranes, namely, PEPG, LPS ^Ec , and LPS ^Yp . We calculated the per-residue dynamics (RMSF), global barrel dynamics (RMSD, R_g), and free energy landscape (FEL) in each case (summarized in Figures 5B, S18, and S19). The results reveal that (i) loop dynamics is regulated strongly by the membrane environment, with all proteins exhibiting the lowest dynamics in LPS ^Yp (Figure S18); (ii) for each membrane, both the fluctuations and energy landscape are lowest for Out14, followed by WT and Out16 (Figures 5B, S18, and S19).

Figure 5. Kinetic partitioning in Ail regulates barrel stability and plasticity.

The primary sequence is a key contributor to Ail stability. (A) Stabilizing (green) and destabilizing (red) residues of the TM domain of Ail-WT (top) and Out14 (bottom) are indicated as spheres on the ribbon diagram of the Ail barrel (PDB ID: 3QRA) and are zonally delineated with a dotted line. Residues substituted in Out14 are shown in light green. (B) Plots comparing FEL with total RMSD and R_g for Ail-WT, Out14, and Out16 in PCPE membranes. Narrowest landscape with lowest RMSD and R_g seen for Out14 is due to lowering of barrel plasticity as a consequence of the substitutions. See Figures S18 and S19 for the complete data and results in other membranes. (C) Linear correlation of the measured equilibrium thermodynamics ( $\text{Δ}{G}_{\text{eq}}^0$ ) in PCPE bicelles with the Moon-Fleming (M-F) free energy scale ⁴² and change in nonpolar ASA (ΔASA_NP) or total ASA (ΔASA_T). ^64,65 Values on the x-axis were calculated as the mathematical sum of the respective empirical parameters for the 14 residues (see Figure 1A, lower panel, and Figure S5B for the sequences) found in each Ail variant. The regression line (black), 95% confidence intervals (gray shaded area), and R² coefficient of determination are included in each graph. A poorer correlation with the M-F scale could arise from a nonadditive effect of the measured $\text{Δ}{G}_{\text{eq}}^0$ with the cumulative $\text{Δ}{G}_{\text{eq}}^0$ determined at the bilayer midplane for the 14 residues. Kinetic partitioning of the apolar side chains into the bilayer (best correlation with ASA_NP) contributes to the measured stability of Out9 (■), Out13 (•), and Out14 (▲).

Further analysis of the transmembrane dynamicity shows the lowest fluctuation for Out14 in all strands of the β-barrel (Figure S19). This is additionally supported by the relatively lower R_g and narrowest FEL for Out14 (Figures 5B and S18), which together suggest an increase in the rigidity and stability of the Out14 scaffold. Indeed, the lowered scaffold dynamicity of OmpX (when compared with Ail) is linked to the unusually high stability of this protein. ^10,63 We find that the targeted substitution of four Ail residues to the corresponding OmpX residues (Q⁵⁰ → Y ofβ3 and L⁸⁸ → I, L⁹² → V, A⁹⁴ → V ofβ5) (Figure 5A) increases the overall stability of Out14 by lowering scaffold plasticity across all strands of the transmembrane region. A similar qualitative outcome across all lipidic conditions (PEPG, LPS ^Ec , and LPS ^Yp ; Figure S19) suggests a primary sequence driven contribution to the measured thermodynamic variables.

To further examine the molecular basis of the observed stability, we correlated the $\text{Δ}{G}_{\text{eq}}^0$ in PCPE with other empirical parameters (Figure 5C) (note that the $\text{Δ}{G}_{\text{eq}}^0$ in PCPE correlates well with the $\text{Δ}{G}_{\text{eq}}^0$ in PEPG; see Figure S9C). The highest R² obtained for the change in nonpolar accessible surface area (ΔASA_NP) suggests the importance of solvent exclusion by hydrophobic side chains in stabilizing select Out-variants. Putting these observations together, we hypothesize that improved local ϕ−ψ values (β-sheet preference; discussed earlier), along with improved H-bond lengths and bond angles in membrana and local changes in hydropathy incurred by substitutions in strands β3 and β5, enhance the overall scaffold rigidity of select engineered Ail variants. Quantitative measures, however, are lipid-dependent (as seen both experimentally and in silico; Figures S9, S18, and S19; Tables S1−S5), and the protein is thermodynamically stabilized best under near-native environments (LPS ^Yp ; Figure S18G-I).

Conclusions

Membrane proteins are notoriously complex and intractable, deterring detailed investigations of their physicochemical characteristics. Difficulties in their direct characterization in native asymmetric membranes are compounded by the difficulty in identifying in vitro environments that support thermodynamic equilibria while suppressing off-pathway aggregation. ^{16,25,41,57,58} Yersinia pestis Ail, for example, is difficult to characterize, as it forms a structurally dynamic β-barrel with unfavorable folding stability compared to its evolutionary homologue OmpX. Ail is also intrinsically aggregation-prone. ^10,11 Yet, the biological significance of Ail as a promising vaccine target against plague, with potential for re-engineering and translational applications, outweighs the challenges inherent to its characterization.

Through this study, we identify signature molecular elements of Ail that biophysically modulate and energetically demarcate this β-barrel from other OMPs. Seven lipid-facing residues (I⁷ and I⁹ of β1, S⁷³ and V⁷⁹ of β4, A¹¹⁷ of β6, L¹⁵¹ and A¹⁵³ of β8) positively regulate Ail folding and stability, with four negative regulators (Q⁵⁰ of β3 and L⁸⁸, L⁹², and A⁹⁴ of β5) of Ail folding kinetics and barrel energetics. We believe that retaining Q⁵⁰, L⁸⁸, L⁹², and A⁹⁴ will allow Ail-WT to suppress the transient accumulation of off-pathway intermediates that might otherwise cause Ail misfolding in vivo. A weighted distribution of stabilizing elements in the lower phospholipid-anchoring domain and destabilizing residues in the upper lipopolysaccharide-docking site provides a metastable Ail structure that functions as a superior adhesion protein in vivo. We attribute a dynamic anchoring role to the lipid-facing Ail residues, which likely confer scaffold plasticity and in membrana Ail oligomerization for effective Yersinia outer protein (Yop) ⁶⁶ delivery. We also identify a markedly compact Ail core with a robust intraprotein network that is less tolerant to conserved substitutions. Such a stable β-barrel core may be essential for binding of the extramembranous loop to the host ECM components. Our findings from the energetics studies allow us to hypothesize that compensatory mutations coassimilated in the primary protein sequence during the evolutionary divergence of Yersiniae from Escherichiae and gave rise to a functionally superior Ail β-barrel that is deficient thermodynamically.

The primary sequence of Ail is conceivably a major determinant of Ail energetics, independent of its lipidic milieu. Using site-specific replacements identified with consensus design, we demonstrate how four targeted replacements (Q⁵⁰ → Y, L⁸⁸ → I, L⁹² → V, A⁹⁴ → V) are sufficient to enhance Ail stability sizably, without altering function. The mutations additively convert metastable Ail to a hyperstable scaffold that is both kinetically and thermodynamically stabilized. These mutations additionally destabilize non-native conformations, thereby lowering protein aggregation. Molecular fine-tuning strategies demonstrated previously for soluble proteins, ^19,20 and extended here to OMPs, now bear direct application in designing thermodynamically evolved chimeric proteins with superior functionality. Identification of similar factors in other functionally relevant OMPs in particular, and membrane proteins in general, can be applied successfully to sculpt novel proteins with improved foldability, thermodynamic stability, and specifically restructured functions for direct applications in biomedical and pharmaceutical industries.

Elucidating fundamental principles that regulate the folding, misfolding and aggregation, end-state stability, function, and interaction network of membrane proteins has remained incredibly challenging. ^13,25,57 Identifying molecular factors in the primary sequence that balance folding with in vivo function through biophysical methods, coupled with cryo-EM structures and large-scale computation, will open avenues for selective restructuring of membrane proteins while also advancing our fundamental understanding of these vital biomolecules. For example, β-barrel core residues could account for the occurrence of hysteresis in some OMPs. ²⁵ Our finding that Ail exhibits a parallel folding process is also unique and is speculated thus far only for an OMP ⁶⁷ from Fusobacterium. While the in vivo Ail folding mechanism has not yet been mapped, we speculate that periplasmic holdases and the barrel assembly machinery ⁶⁸ of Yersiniae could preferentially populate one folding pathway of nascent Ail. Further studies on early events in OMP folding in membrana will serve as excellent starting points in understanding their stepwise assembly mechanism(s).

Intrinsic difficulties in obtaining soluble and stable forms of membrane proteins for detailed characterization and large-scale inhibitor screens has substantially deterred the development of appropriate therapeutic interventions for multidrug resistant bacteria including the tier I pneumopathogen Y. pestis. We believe that with the availability of thermodynamically hyperstable Ail variants that are not aggregation-prone, targeted drug screening and vaccine development will now be possible. Fine-tuning of membrane protein targets to curtail other bioterrorism agents can also be done. We envision that similar findings with other membrane proteins will find widespread application in protein-based drug design and translational biopharmaceutical therapeutics.

Supplementary Material

SI File

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.1c05964.