Reversible folding energetics of Yersinia Ail barrel reveals a hyperfluorescent intermediate

Abstract

Deducing the molecular details of membrane protein folding has lately become an important area of research in biology. Using Ail, an outer membrane protein (OMP) from Yersina pestis as our model, we explore details of β-barrel folding, stability, and unfolding. Ail displays a simple transmembrane β-barrel topology. Here, we find that Ail follows a simple two-state mechanism in its folding and unfolding thermodynamics. Interestingly, Ail displays multi-step folding kinetics. The early kinetic intermediates in the folding pathway populate near the unfolded state (β _T ≈ 0.20), and do not display detectable changes in the local environment of the two interface indoles. Interestingly, tryptophans regulate the late events of barrel rearrangement, and Ail thermodynamic stability. We show that W¹⁴⁹ → Y/F/A substitution destabilizes Ail by ~0.13–1.7 kcal mol⁻¹, but retains path–independent thermodynamic equilibrium of Ail. In surprising contrast, substituting W⁴² and retaining W¹⁴⁹ shifts the thermodynamic equilibrium to an apparent kinetic retardation of only the unfolding process, which gives rise to an associated increase in scaffold stability by ~0.3–1.1 kcal mol⁻¹. This is accompanied by the formation of an unusual hyperfluorescent state in the unfolding pathway that is more structured, and represents a conformationally dynamic unfolding intermediate with the interface W¹⁴⁹ now lipid solvated. The defined role of each tryptophan and poorer folding efficiency of Trp mutants together presents compelling evidence for the importance of interface aromatics in the unique (un)folding pathway of Ail, and offers interesting insight on alternative pathways in generalized OMP assembly and unfolding mechanisms.

1. Introduction

The fundamental mechanism of membrane protein folding and assembly has been the focus of extensive recent studies [1–5]. The folding microenvironment and key residues in the primary sequence play vital roles in the folding process of both transmembrane α-helices and β-barrels. Despite several efforts, we still lack a complete understanding of how transmembrane β-barrels assemble. The energy landscape of β-barrel folding is dictated by the primary sequence of protein (intramolecular interactions in the polypeptide), and the surrounding lipidic milieu [1]. Under physiological conditions, outer membrane β-barrel proteins (OMPs) of Gram-negative bacteria assemble through folding intermediates, and this process is assisted by the barrel assembly machinery [3,6,7]. However, OMPs retain the ability to fold independently in vitro. This property of OMPs facilitates the biophysical analysis of folding intermediates in barrel assembly, and the influence of the protein sequence on the folding pathway [7–9]. For example, the outer membrane β-barrel proteins (OMPs) OmpA and OmpX from Escherichia coli sample several short-lived partially folded conformations during folding, and achieve complete folding by a multi-step concerted mechanism [9–11]. These transient intermediates direct the nascent polypeptide to its folded state within biologically relevant timescales, while avoiding misfolding and aggregation [8,11]. The characterization of such intermediates is crucial to understand the folding energy landscape. However, the partially folded conformations are transient and do not accumulate under equilibrium conditions [9,11]. A two-state folding mechanism is therefore sufficient to explain, and is consistent with the thermodynamic equilibrium observed during OMP folding [8,11–13]. Primarily, the rate-limiting step is studied, where the unfolded and native states are separated by a large energy barrier.

The folding of OMPs requires a particularly intricate interplay of thermodynamic and kinetic elements that together define the conversion of the unfolded polypeptide into a biologically active and stable β-barrel conformation. While OMPs do exhibit thermodynamic equilibrium in vitro, it has been proposed that OMPs in the bacterial outer membrane are kinetically stabilized, as they possess high activation energy barrier for unfolding, and show slow equilibration in thermodynamic measurements [4,10,14,15]. It is believed that this kinetic stability is essential for OMPs to function in a constantly changing external environment. Owing to this kinetic stability, in vitro (un)folding measurements of several OMPs are characterized by accumulation of trapped intermediates in the folding or unfolding pathway, leading to hysteresis [7,8,16]. The characterization of such intermediates can provide useful information on the microscopic structure of the transition state, and the macroscopic mechanism of OMP assembly. Our current knowledge of OMP assembly has been obtained from extensive studies of E. coli proteins. Interestingly, however, recent studies show that membrane proteins exhibit species-specific variations in their folding mechanism [17–19]. It is therefore important to assess the assembly of atypical OMPs of other bacteria.

The formation of stable or transient (un)folding intermediates in OMPs is strongly regulated by specific residues in the polypeptide. For example, conserved aromatic residues (Phe, Tyr, Trp) located at the solvent–membrane interface act as lipid anchors during OMP assembly. Indeed, previous studies on various E. coli OMPs including OmpA [20], OmpX [21], PagP [13], OmpLA [22], FomA [23], and the human mitochondrial VDAC [24] highlight the importance of interface aromatics, particularly tryptophan, for β-barrel folding. Recent studies of these OMPs in lipidic vesicles have revealed interesting insights on unassisted OMP folding [12]. However, a detailed study of folding intermediates in vesicles is compounded by the accumulation of a heterogeneous population of partly folded states. In interesting contrast, detergent micelles conveniently alleviate this problem by stabilizing fewer long-lived intermediates while also retaining path-independent folding [14,16]. For example, phosphocholine (PC) micelles, which possess a similar zwitterionic nature as diacyl PC lipids, promote β-barrel folding without aggregation, are suitable for structural studies, and facilitate the mapping membrane protein folding intermediates [1,9].

In this study, we investigate the thermodynamic and kinetic factors contributing to the folding and stability of an atypical OMP from the category A pathogen Yersinia pestis. The attachment invasion locus (Ail) protein is an 8-stranded OMP vital for Y. pestis serum resistance, adhesion to and internalization into host cells, Yersinia outer proteins (Yop) delivery, and inhibition of host inflammatory response [25,26]. The study of Ail is therefore of interest as a potent drug target against this category A bioterrorism agent. Ail possess two interface tryptophans, W⁴² and W¹⁴⁹ (of a total ~17% aromatic residues (2 Trp, 12 Tyr, and 12 Phe) in its polypeptide sequence of 157 amino acids) (Fig. 1A), which are important for Ail folding in detergent micelles [25]. Using fluorescence analysis of single tryptophan mutants, we show that Ail exhibits complex (un)folding kinetics, with a rapid assembly phase followed by slow rearrangement kinetics in PC micelles. A similar complex folding pathway is known so far only for E. coli OmpA [16]. An unexpected observation in our study is the disruption of Ail thermodynamic equilibrium upon mutation of W⁴². The Ail–PC system now shifts to kinetic control, leading to an apparent hysteresis. This hysteresis is accompanied by the formation of a hyperfluorescent intermediate that represents a lipid-solvated partially unfolded barrel. We find that the anchoring and stabilizing roles of W⁴² and W¹⁴⁹ together play distinct roles in deciding the folding landscape of Ail, thereby offering interesting insight on novel intermediates in the OMP folding and unfolding pathways. Our findings with Ail will also serve as an excellent starting point to study the complex assembly and unfolding pathways of OMPs from other pathogens.

Fig. 1. Mapping spectroscopic properties upon selective Trp mutation on the structure and sequence of Ail.

(A) Cartoon representation of Ail (PDB ID: 3QRA), highlighting W⁴² and W¹⁴⁹ (pink spheres). Distribution of the two other aromatic residues, Phe (blue) and Tyr (orange), are shown as sticks. (B) List of Trp mutants used in this study. Trp → Phe/Tyr/Ala substitutions at the 42nd or 149th position are highlighted (red). Numbering above the sequence denotes the position of each tryptophan in the full-length protein. (C) Far-UV CD spectra of folded Ail WT and its Trp mutants in DPC micelles. All the proteins exhibit a typical β-sheet structure with a negative maximum at ~215 nm (ME₂₁₅), and similar secondary structure content. (D) Fluorescence intensities of all the folded and unfolded proteins measured at λ_em = 336 nm, corresponding to the λ_em-max of folded Ail. The significant difference in fluorescence intensity of folded W42X and W149X mutants suggests that the local environment of W¹⁴⁹ and W⁴² are different [25] (see Fig. S4). In comparison, the unfolded proteins (half-filled circles) in 8.0 m GdnHCl show fluorescence intensities proportionate to their Trp content. The complete data is provided in Fig. S4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

2. Materials and methods

2.1. Protein preparation and folding

All the chemicals were purchased from Sigma-Aldrich Co. LLC., and DPC (n-dodecyl phosphocholine) from Avanti Polar Lipids, Alabaster. The ail gene (without signal sequence) was cloned into a pET3a vector, and all the single Trp mutants W⁴² → F/Y/A and W¹⁴⁹ → F/Y/A were generated using site directed mutagenesis. All the proteins were expressed as inclusion bodies in E. coli C41 cells, and purified using reported protocols [25,26]. Independent protein preparations were used to set up each replicate experiment.

Folding efficiency of Ail was screened in increasing concentrations of DPC (5 mM – 50 mM) from the unfolded Ail stock prepared in 8.0 m GdnHCl (guanidine hydrochloride) (data not shown). Above 10 mM DPC, complete folding of Ail was obtained without any detectable aggregation; thus, we selected 20 mM DPC concentration for further experiments. The micelle integrity in 6.4 m GdnHCl (highest concentration used in our experiments) was checked independently using SYPRO® orange (discussed later in Fig. S2), as described previously [25].

Folding of Ail was achieved by rapid 10-fold dilution of the unfolded protein (~1400 μM Ail in 8.0 m GdnHCl, 20 mM Tris-HCl pH 8.5), into the folding reaction containing 100 mM DPC prepared in 20 mM Tris-HCl pH 8.5, followed by incubation for 5 h at 25 °C. Trace amounts of aggregated protein (if any) was removed by high-speed centrifugation at 18,000 xg at 10 °C. This folded 5× stock was diluted 5-fold in all the experiments, to achieve a final protein and DPC concentration of 28 μM and 20 mM, respectively (detergent-to-protein ratio of ~700:1) in 20 mM Tris-HCl pH 8.5 containing 160 mM GdnHCl. Protein concentration was quantified using A ₂₈₀, with different extinction coefficients for the WT and Trp mutants (28,880 M⁻¹ cm⁻¹ for Ail WT; 23,380 M⁻¹ cm⁻¹ for W42F, W149F, W42A, and W149A; 24,870 M⁻¹ cm⁻¹ for W42Y and W149Y at 280 nm).

2.2. Assessment of folded Ail using mobility shift, far-UV CD, and NMR spectroscopy

Upon folding, Ail exhibits different electrophoretic mobility compared to the unfolded protein, on cold SDS-PAGE gels. All the folded protein samples and samples before and after each experiment were checked for sample integrity and folding efficiency on 15% SDS-PAGE gels using the electrophoretic mobility assay [25].

The secondary structure content of WT Ail and all the folded Trp variants was checked using far-UV and near-UV circular dichroism (CD). All the CD measurements were carried out on a J-815 CD spectropolarimeter (JASCO Inc.) equipped with a water-cooled peltier 6-cell system, at 25 °C. For all the CD measurements, we used 28 μM folded protein in 20 mM DPC and 20 mM Tris-HCl pH 8.5, as described above. We recorded the far-UV CD spectra for Ail-WT and all the Trp mutants from 205 to 260 nm at a constant scan speed of 100 nm/min, data integration time of 1 s, and data pitch of 0.5 nm. The data were averaged over three accumulations, and corrected for contributions from buffer and DPC micelles. The raw ellipticity value at 215 nm was then converted to molar ellipticity (ME ₂₁₅). All the samples were checked on cold SDS-PAGE gels before and after the experiment.

The folded 5× stocks of uniformly ¹⁵N-labeled WT, W42F, and W149F were dialyzed against 20 mM Tris buffer using stepwise dialysis to remove trace amounts of GdnHCl, as described previously [25]. HSQC-TROSY experiments were recorded at 45 °C on a 700 MHz NMR spectrometer equipped with a cryoprobe, as we were not able to observe well-dispersed HSQC-TROSY spectra of Ail at 25 °C due to sample viscosity (discussed in the results section). The final protein concentration of ~140 μM Ail in 100 mM DPC, corresponding to a DPR of ~700:1 was used for NMR measurements. The NMR samples were checked on 15% SDS-PAGE before recording HSQC-TROSY experiments. The data were processed and plotted using reported methods [27].

2.3. (Un)folding kinetics measurements and chevron analysis

The (un)folding kinetics measurements were carried out using the RX2000 rapid mixing accessory (Applied Photophysics Ltd.) that was coupled to a FluoroMax 4 fluorimeter (Horiba Jobin Yvon Ltd.), with a dead time of ~8 ms. The kinetics measurements were carried out at 25 °C by monitoring changes in Trp fluorescence, using λ_ex = 295 nm and λ_em-max = 340 nm. For folding kinetics, the 5× unfolded protein stock (140 μM unfolded Ail in 5.0 m GdnHCl, and 20 mM Tris pH 8.5), was diluted 5-fold in 25 mM DPC containing varying concentrations of GdnHCl. Similarly, unfolding kinetics was measured by rapid dilution of the 5× folded protein stock (140 μM folded Ail in 100 mM DPC and 800 mM GdnHCl, in 20 mM Tris pH 8.5) into varying concentration of GdnHCl, and change in the fluorescence intensity at 340 nm was monitored. The DPR of ~700:1 was maintained throughout the kinetics measurements. The change in Trp fluorescence was monitored till the reaction attained saturation and no further change in fluorescence intensity was observed. The kinetic traces were recorded for > 40 GdnHCl concentrations. All the kinetic measurements were carried out three times from the same stock sample. Each trace was fitted individually and the mean data obtained at each [GdnHCl] was used for the chevron analysis. In addition, at least 30 data points (~14 from folding kinetics and ~17 from unfolding kinetics measurements) were used for the chevron analysis of each protein to obtained reliable data.

The initial folding of Ail was completed within the dead time of the experiment. We fitted the slow rates of folding and unfolding for the final analyses (the fast rates were not measured due to experimental limitations). The data obtained from (un)folding measurements were processed as described previously [24]. Briefly, the kinetic traces were fitted to a single exponential function, and the rates k_f (folding rate) and k_u (unfolding rate) were obtained at various GdnHCl concentrations. Here, the initial ~50–100 s of the folding kinetic traces, and ~10–20 s of the unfolding kinetics traces were not used in the fitting (discussed later in Fig. S6). The chevron plots thus obtained from these rates were then fitted globally to a two-state equation [12] with a shared m _f and m _u (constants that represent the change in the accessible surface for a protein on folding or unfolding, respectively), to derive $k_{\text{f}}^0$ and $k_{\text{u}}^0$ (folding and unfolding rates, respectively, in buffer), and the kinetic free energy ${\Delta G_{\text{kin}}}^0$ was calculated for all the proteins. A shared m value was used, as the mutation of a single residue does not affect the overall change in accessible surface area of the protein (supported by the far-UV CD profiles). We obtain the standard deviation for ${\Delta G_{\text{kin}}}^0$ from propagation of uncertainty (propagation of error) obtained from $k_{\text{f}}^0$ and $k_{\text{u}}^0$ , as mentioned below [28]:

$$\delta \Delta G_{\text{kin}}^0=\sqrt{{\left(\delta k_{\text{f}}^0\right)}^2+{\left(\delta k_{\text{u}}^0\right)}^2}$$

here, ${\delta \Delta G_{\text{kin}}}^0$ = error for ${\Delta G_{\text{kin}}}^0;\delta k_{\text{f}}{}^0$ = error for $\text{r}{k}_{\text{f}}^0$ ; and $\delta k_{\text{u}}^0$ = error for $k_{\text{u}}^0$ .

3. Equilibrium (un)folding titrations

The equilibrium (un)folding experiments were carried out as described earlier with minor modifications [25]. The unfolded protein stock (140 μM Ail in 100 mM DPC and 5.0 m GdnHCl, in 20 mM Tris-HCl pH 8.5) and the folded protein stock (140 μM Ail in 100 mM DPC and 800 mM GdnHCl, in 20 mM Tris-HCl pH 8.5) were diluted 5-fold into a GdnHCl gradient, and a DPR of 700:1 was maintained throughout the experiments. The lowest GdnHCl concentration that can be achieved in the folding equilibrium is 1.0 M, while the lowest GdnHCl concentration for the unfolding equilibrium is 0.16 M. Therefore, the GdnHCl range for folding and unfolding titrations is from 1.0 M–5.0 m and 0.16 M–6.0 M, respectively. Reactions were incubated at 25 °C, and Trp fluorescence emission spectra were acquired between 320 and 380 nm, using a λ_ex = 295 nm. Equilibrium was achieved within 24 h, and the 48 h data was used for analysis. For all the equilibrium measurements, 33 data points (for both folding and unfolding equilibrium) were acquired from at least 3 independent datasets obtained from independent folding and unfolding stocks prepared from at least 3 independent purified protein preparations. The fraction unfolded (f_U ) at λ_em-max of 336 nm and the average wavelength (< λ >) values were plotted against the respective GdnHCl concentrations, as described previously [25]. The average wavelength data for folding and unfolding equilibrium of all the proteins were fitted globally to a two-state unfolding model with a shared m, as described previously [12]. We obtain a common m of −3.21 kcal mol⁻¹ M⁻¹ for all the proteins, except for the folding equilibrium data of W42X mutants, where the m value is different (−2.52 kcal mol⁻¹ M⁻¹). The calculated m values were used to derive the midpoint of the chemical denaturation (C _m), and absolute or apparent folding and unfolding free energy values (ΔG ^0,F or $\Delta G_{\text{app}}{}^{0,\text{F}}$ , and ΔG ^0,U or $\Delta G_{\text{app}}{}^{0,\text{U}}$ , respectively). Anisotropy (r) values of all postequilibrium (48 h) samples were recorded at 25 °C, as described previously [25]. The anisotropy values obtained for each sample was plotted against the respective GdnHCl concentrations, and the data were fitted to a sigmoidal function to derive the C _m. C _m-start is the intercept of the linear extrapolations of the pre-transition baseline and transition zone in the equilibrium profiles (see [29] for similar calculations from thermal measurements).

4. Acrylamide quenching and lifetime measurements

We studied the local environment of both tryptophans using its intrinsic fluorescence as the reporter. Post-equilibrium samples (48 h) were incubated with increasing concentrations of acrylamide (0.0–0.5 M) for 15 min, and Trp fluorescence (λ_ex = 295 nm and λ_em-max = 336 nm) was recorded. The λ_em-max for the sample without acrylamide was used for normalization, and the data obtained were fitted to a polynomial function. The slope of the polynomial fit corresponded to the Stern-Volmer quenching constant (K _SV) [25]. Average Trp lifetime (< τ >) was measured for all the samples on a DeltaFlex TCSPC system (Horiba Jobin Yvon Ltd.), as described earlier [25]. Briefly, samples with and without acrylamide were excited at 292 nm, and Trp lifetime was recorded at the λ_em-max for all the samples (336 nm for the folded, 340 nm for partially unfolded, 354 nm for partially folded and unfolded samples). Decay curves were fitted to a three exponential decay function to derive the < τ >. The χ² values were maintained at < 1.2. The bimolecular quenching rate constant (k _q) and apparent rotational correlation time (τ_c) were derived using the formulae k _q = K _SV/< τ > and τ_c =< τ > r/(r₀–r) [5]. Here, r ₀ = 0.3, and represents the anisotropy of free indole [5]. In all measurements, samples folded in DPC (containing 0.16 m GdnHCl), and protein in 8.0 m GdnHCl without DPC, served as the folded and unfolded controls, respectively.

4.1. Co-evolutionary analysis of W⁴² and W¹⁴⁹

Co-evolution analysis of the two intrinsic Trp residues of Ail was done using EVfold (http://evfold.org/) [30]. Here, we first generated the multiple sequence alignment of various OMPs using Pfam (pfam.sanger.ac.uk) database. The EVfold server (http://evfold.org/) [30] was then used to perform the co-evolution analysis to find the conservation pattern for W⁴² and W¹⁴⁹. For the EVFold analysis, Ail sequences from Y. pestis, and the crystal structure of Ail (PDB ID: 3QRA) was used as template. Coevolving residues are presented as a network, for comparison.

5. Results

5.1. Folding analysis of Ail and its mutants in zwitterionic DPC micelles

Ail possesses two tryptophans at positions 42 and 149. Here, W¹⁴⁹ is highly conserved across most OMPs, whereas W⁴² is less conserved (while Trp is more abundant, Leu and Phe are also observed) (Fig. S1). Both Trp residues of Ail are located at the membrane–water interface (Fig. 1A). To address the role of Trp in particular and aromaticity in general on Ail folding, we systematically substituted each Trp to Phe, Tyr, or Ala (Fig. 1B). In the first set of experiments, we folded Ail and its interface Trp mutants in the zwitterionic 12-C detergent n-dodecyl phosphocholine (DPC) micelles. The detergent-to-protein ratio (DPR) is critical for Ail folding and stability [25]. Ail remains largely monomeric and well-folded in DPC concentrations > 10 mM (data not shown). On the basis of the screening results, we selected 28 μM Ail and 20 mM DPC (DPR of ~700:1), in 20 mM Tris-HCl pH 8.5, and a temperature of 25 °C, for our experiments. As urea does not fully unfold Ail [25], for our thermodynamic measurements, we used guanidine hydrochloride (GdnHCl) as the denaturant. To ensure that Ail unfolding was not due to the dissociation of DPC micelles, we checked for micelle integrity in the presence of up to 6.4 m GdnHCl, using the fluorescence of SYPRO^® orange dye (CMC of DPC in buffer is ~1.25 mM, whereas the CMC increases to ~12.5 mM in 6.4 m GdnHCl) (Fig. S2). Therefore, we carried out all our studies in 20 mM DPC, which is at least 2-fold higher than the CMC of DPC in the highest [GdnHCl] used in this study. This ensured that the energetics we measure here stems primarily from the biophysical properties of the protein.

In DPC micelles, folded Ail has markedly different biophysical properties from the unfolded protein. For example, in cold SDS-PAGE gels, Ail exhibits a prominent electrophoretic migration upon folding (Fig. S3). Folded Ail is resistant to proteolysis (data not shown). The secondary structure content of Ail assessed using far-UV circular dichroism (CD) spectrometry shows prominent negative ellipticity centered at 215 nm, characteristic of a β-sheet rich structure (Fig. 1C). A similar β-sheet content seen for all the Trp mutants confirms that the mutation does not affect the overall scaffold of Ail. A blue-shifted fluorescence emission spectrum with maximum emission intensity (λ_em-max) at 336 nm is also seen for Ail WT and the mutants upon folding in DPC micelles (see Fig. S4). HSQC-TROSY spectra of uniformly ¹⁵N-la-beled Ail show well-dispersed resonances that superpose well for Ail WT and its Trp mutants (Fig. 2). Overall, these results confirm that under in vitro folding conditions, all the mutants fold to a similar extent, and Trp substitution does not impede the ability of Ail to form a well-folded barrel.

Fig. 2.

Heteronuclear NMR measurements confirm similar folded conformations of Ail and its Trp mutants. Overlay of ¹H-¹⁵N HSQC-TROSY spectra of WT (black) with (A) W42F (red, left) and (B) W149F (green, right), folded in 20mM DPC micelles (DPR of ~700:1), and recorded at 45 °C. We were not able to observe well-dispersed HSQC-TROSY spectra of Ail at 25 °C due to sample viscosity. NMR spectra were therefore recorded at higher temperatures (45 °C). Far-UV CD spectra confirm that the Ail scaffold is preserved and the protein shows identical secondary structure content at both 25 °C and 45 °C (Fig. S4B). In both spectra (A and B), peaks corresponding to each indole amide resonance are marked according to their position (in blue). A few other unambiguous resonances are also annotated [26]. The spectra of both mutants show a minor spectral shift for a few resonances (indicated by arrows) with the appearance or disappearance of a few other resonances (circled). Overall, the amide resonances for the mutants superpose well with the WT spectrum, suggesting that the overall barrel structure is largely unaltered by the mutation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Interestingly, the single Trp mutants (W42X and W149X) show distinctly different fluorescence intensities at 336 nm (Fig. 1D). This change in the spectral property suggests that W⁴² and W¹⁴⁹ reside in different local environments. The latter is further evident in HSQC-TROSY measurements of Ail WT and its mutants (Fig. 2). Here, a more downfield shifted W⁴² amide resonance indicates that this indole resides in a more polar environment than W¹⁴⁹. Our conclusion is also confirmed by anisotropy (r) measurements, with r ≈ 0.155 ± 0.004 for W⁴² (W149X mutants) and r ≈ 0.126 ± 0.008 for W¹⁴⁹ (W42X mutants) (see Table S1 for a comparison of r values), showing a difference of ~0.029. Differences in anisotropy between two distinct states of a membrane protein are small (Table S1), and such differences are indeed significant. The results from fluorescence, NMR, and anisotropy together reveal that W⁴² resides in a more polar and rigid environment, whereas W¹⁴⁹ is present in a comparatively more apolar and dynamic environment. Next, we carried out kinetics and equilibrium measurements in DPC micelles using Trp as the reporter to understand the biophysical characteristics that influence folding-unfolding pathways of Ail.

5.2. Ail core assembles independent of interface residues

To understand the early events in Ail (un)folding, we used the two interface tryptophans as reporters of β-barrel assembly. The two Trp residues in Ail are located at opposite faces of the barrel, and reside in dissimilar local environments, as discussed above. This provides a unique advantage to monitor distinct changes in Ail during barrel folding. We measured the conformational changes of DPC-folded Ail, spectroscopically using increasing concentrations of GdnHCl as the denaturant. Representative folding and unfolding kinetics of Ail WT are presented in Fig. 3A (see Fig. S5 for the magnified traces). The process is accompanied by a rapid change in the fluorescence intensity in both the folding and unfolding kinetics that occurs within the dead time of detection. Ail folding is, therefore, faster than other bacterial OMPs such as PagP [12], OmpX [21], and OmpA [16]. This “burst phase” intensity varies proportionately with increase in the denaturant concentration, and indicates the accumulation of an early intermediate in the (un) folding of Ail (see Fig. S5 for magnified traces highlighting the initial burst phase and lag phase). Although we could not capture this early intermediate, the kinetic traces are comparable for the mutants, allowing us to conclude that the early events in Ail (un)folding are Trp–independent and do not involve detectable changes in the local environment of both the indoles.

Fig. 3. Rapid burst phase in micelles precedes slow (un)folding kinetics of Ail.

(A) Folding (left panel) and unfolding (right panel) kinetics traces of Ail at various GdnHCl concentrations (low to high). The burst phase, which represents a rapid increase (in folding) or decrease (in unfolding) in Trp fluorescence when compared with the initial values in both the folding and unfolding profiles, is indicated by an arrow. Data from the slow kinetics was fitted to single exponential function (solid lines) to obtain the rates of (un)folding (see Table S2 for details). Data from the folded and unfolded controls are also shown. The magnified traces are shown in Fig. S5 and S6. All the kinetic analyses were carried out in triplicate with the same stock sample. (B) Chevron plots for different Phe (left) and Tyr (right) mutants; data for WT is included in both graphs for comparison. Each rate constant was derived by fitting the kinetic profile to a single exponential function; error bars represent the goodness of fit and are smaller than the data point. Fits to a two-state folding model are shown as solid lines. Slow folding rates between 3.0 and 4.0 m GdnHCl could not be accurately captured. (C) Free energy calculated from the global analysis of kinetic measurements for Ail WT and the mutants. The color profile from panel B is retained. The complete data are provided in Fig. S5-S7, and Table S2. We obtain the standard deviation for ${\Delta G_{\text{kin}}}^0$ from the propagation of uncertainty or propagation of error [28], calculated from the folding and unfolding rates. Significance levels of the ${\Delta G_{\text{kin}}}^0$ are therefore not calculated. We observe that un) folding kinetics of Ail is not significantly affected by Trp mutation and the differences observed in ${\Delta G_{\text{kin}}}^0$ are not significant.

Ail (un)folding kinetics shows the presence of at least two phases, where the rapid (un)folding event (in seconds) is followed by slow (un) folding transition that occurs over minutes (Fig. 3A). Here, the folding kinetics traces show three transitions: an early decay phase, followed by a fast folding phase and a slow folding phase (Fig. S6A). We fi tted all the transitions individually to single exponential functions to derive folding rates (k _d1, k _f1, and k _f2) (Fig. S6A). Interestingly, only the amplitude A _f2 associated with the rate k _f2 shows a significant dependence on the GdnHCl concentration, indicating that only k_f2 contributes significantly to the folding process that we are monitoring. Therefore, we used only the slowest rate for our analysis (Fig. S6A). Similarly, the unfolding kinetics traces can be fitted to both single as well as double exponential functions. As seen with the folding kinetics, the slow rate of unfolding (k _u) and the associated amplitude (A _u) obtained from a single exponential function was sufficient to explain the unfolding profiles (Fig. S6B).

Overall, Ail shows complex folding and unfolding profiles comprising of multiple transitions (Fig. S6C). Such complex folding kinetics is well-known for transmembrane β-barrels in membranes [10,14], but is rare in micelles and has been reported so far only for E. coli OmpA [16]. As the folding and unfolding amplitudes for the fast rates (A _d1 and A _u1) show no denaturant dependence (see Fig. S6), we analyzed only the slow rates of (un)folding (see Fig. S6 for details). Chevron plots generated from the slow rates (k _f2 and k _u) show linear folding and unfolding arms with no rollover for Ail WT and all the mutants (Fig. 3B, S6C). The data were fitted to a two-state model to derive the folding and unfolding rate constants, change in accessible surface area, and the stability. Comparison of the change in free energy ( ${\Delta G_{\text{kin}}}^0$ ) reveals that the (un)folding kinetics of Ail is not significantly affected by mutating the interface Trp residues. For example, Ail WT shows a ${\Delta G_{\text{kin}}}^0$ of ~7.9 kcal mol⁻¹, which is similar across the Trp mutants. A marginal destabilization of ~0.5 kcal mol⁻¹ is seen only for W149F (Fig. 3C, Table S2). Overall, we observe that Trp substitution does not significantly affect the ${\Delta G_{\text{kin}}}^0$ of Ail.

Next, we calculated the Tanford β (β _T) value, which is a measure of the compactness of the transition state. In other words, β _T measures the average degree of exposure of the protein in the transition state compared to the native or unfolded states [31]. High β _T values (up to ~0.95), indicates that the protein shows native-like compaction in the transition state. We obtain similar β _T values of ~0.20 for all the mutants (Table S2), which suggests that in the transition state, the environment of both tryptophans resembles the unfolded state. The values obtained for Ail are higher than bacteriorhodopsin (β _T = 0.13–0.14) [32], but marginally lower than the rhomboid protease GlpG (β _T = 0.23 ± 0.02) [33]. β _T values for bacteriorhodopsin and GlpG (unfolded using SDS) are underestimated, because the denatured state of helical proteins in SDS is more compact due to incomplete unfolding. On the other hand, the β _T of Ail is lower than other transmembrane β-barrel proteins; e.g., β _T is ~0.48 for human mitochondrial VDAC2 porin in GdnHCl [24], ~0.23 and ~0.29 for OmpA in urea and GdnHCl, respectively (here, β _T was calculated only using m _f and m _u) [16]. Hence, the transition state structure of Ail is less compact than other β-barrels.

Overall, we find that Ail folding in DPC micelles occurs through the formation of an early intermediate that subsequently undergoes slow rearrangement over several seconds–minutes to form the folded barrel. ${\Delta G_{\text{kin}}}^0$ and β _T values together support our conclusion that neither interface tryptophan contributes significantly to the early intermediates formed in the folding and unfolding processes. In other words, mutations of the two interface indoles bear little impact on Ail (un)folding kinetics. A similar folding mechanism is also observed in E. coli OmpA in vesicles [16], suggesting that the process of OMP folding in micelles and vesicles could indeed be similar. Further investigation of the transition state is presently underway to obtain a better understanding of Ail assembly.

5.3. Substitution of W⁴² shifts the thermodynamic equilibrium towards kinetic control

Ail exhibits path-independent reversible folding and attains thermodynamic equilibrium in DPC micelles in ~24 h (Fig. 4A, middle panel; and S8). Here, the folded protein shows a λ_em-max ≈ 336 nm, which shows a red shift to λ_em-max ≈ 360 nm upon barrel unfolding in high GdnHCl concentrations (Fig. S9). We fitted the data to a two-state model with a shared m of ± 3.21 kcal mol⁻¹ M⁻¹ to obtain the thermodynamic parameters listed in Table 1 (also see Fig. 4, Table S3). The W42X mutants showed poorer folding cooperativity and an m = −2.52 kcal mol ⁻¹ m ⁻¹ was used. The WT protein shows no hysteresis and we obtain a change in the equilibrium free energy derived from the unfolding transition (ΔG^{0, U}) of ~8.0 kcal mol⁻¹, which is similar to the ${\Delta G_{\text{kin}}}^0$ (see Fig. 3C).

Fig. 4. Substituting W⁴² drastically alters the thermodynamic stability of Ail.

(A) Representative unfolding (maroon) and folding (orange) equilibrium profiles of Ail WT and its single Trp mutants (lower panel). Cartoon representation of Ail WT (middle), W42F (left), and W149F (right) highlighting tryptophan residue(s) (pink spheres) are shown above the plot. Ail WT and W149X (shown here is W149F) exhibit thermodynamic equilibrium, while W42X (shown here is W42F) shows hysteresis. Fits of data to the two-state model is shown as solid lines. Notably, the range for folding and unfolding titrations is 1.0 m – 5.0 m and 0.16 m – 6.0 M, respectively (see methods for details). (B) Comparison of thermodynamic parameters (|ΔG⁰|, top; C_m, bottom) derived from the global analysis of equilibrium (un) folding data for the Ail Trp mutants. The error bars represent standard deviation (s. d.) derived from at least three independent experiments performed using independent protein preparations. S. d. for WT is also shown as a grey bar for comparison. Substitution of either Trp residue lowers Ail stability. However, substitution of W⁴² not only destabilizes Ail to a greater extent than substitution of W¹⁴⁹ (by ~2.3–3.5 kcal mol⁻¹ from Ail WT), but also introduces hysteresis in the equilibrium unfolding titrations. The complete data are provided in Fig. S8-S10. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1. Thermodynamic parameters calculated from equilibrium unfolding experiments.

Protein	ΔG0, U or $\Delta G_{\text{app}}{}^{0,\text{U}}$	Cm	ΔΔG0, U or $\Delta \Delta G_{\text{app}}{}^{0,\text{U}}$
	(kcal mol−1) a,b	(M)	(kcal mol−1) b
WT	8.12 ± 0.50	2.53 ± 0.16
W42F	8.46 ± 0.46	2.63 ± 0.14	−0.34 ± 0.68
W42Y	8.71 ± 0.38	2.71 ± 0.12	−0.59 ± 0.63
W42A	9.25 ± 0.33	2.88 ± 0.10	−1.13 ± 0.60
W149F	7.27 ± 0.41	2.26 ± 0.13	0.85 ± 0.65
W149Y	7.99 ± 0.46	2.49 ± 0.14	0.13 ± 0.68
W149A	6.43 ± 0.05	2.00 ± 0.02	1.69 ± 0.50

^am = 3.21 ± 0.04 kcal mol⁻¹ M⁻¹.

^b $\Delta G_{\text{app}}{}^{0,\text{U}}$ and $\Delta \Delta G_{\text{app}}{}^{0,\text{U}}$ (apparent values) are for the W42X mutants.

Similar to WT, the W¹⁴⁹ → F/Y/A mutants also show no hysteresis. Interestingly, Trp substitution lowers the thermodynamic stability of Ail (Table 1). Mutating W¹⁴⁹ to the other aromatic residues destabilizes Ail marginally (Fig. 4, and S8; Table 1 and S3). The extent of destabilization is highest in the Ala mutant, where there is a considerable decrease in free energy by ~1.7 kcal mol⁻¹. Further, the lowered C _m of W149A mutant suggests that at the 149th position, a preference is seen for aromatic side chains (Fig. 4B, Table 1 and S3), which is in line with the lipid-solvated environment for this indole. Our results show that W¹⁴⁹ helps in stabilizing the folded Ail barrel, and in resisting protein solvation by GdnHCl. Beyond a critical concentration of GdnHCl, native contacts are rapidly replaced by protein–GdnHCl interactions, and both Ail WT and W149X show cooperative unfolding (note the similar m_eq value in Fig. 4, Table 1 and S3).

In interesting contrast to the W149X mutants, the substitution of W⁴² (to F/Y/A) results in prominent hysteresis (Fig. 4A, left panel, and S8), and the stability of the mutant appears to now be under kinetic control. It must be noted here that in the W42X mutants, we monitor the fluorescence properties of W¹⁴⁹. From the folding titration for the W⁴² → F/Y/A mutants, we obtain $\Delta G_{\text{app}}{}^{\text{0, F}}$ values that are lower than Ail WT by ~2.3–3.5 kcal mol⁻¹, with the W42A mutant showing the highest destabilization (see Table S3). As with the W149A mutant, the substitution of W⁴² to Ala is more deleterious than Tyr and Phe. In all three W42X mutants, barrel formation is completed only when the GdnHCl concentration is lowered considerably and after prolonged incubation at 25 °C (for > 5 h, see Methods for details; complete folding is demonstrated in Fig. 1C, 2, S3 and S4). Comparison of the $\Delta G_{\text{app}}{}^{\text{0, F}}$ obtained from the folding titrations suggests that W⁴² is more important than W¹⁴⁹ for Ail stability. The absence of hysteresis in Ail WT and W149X mutants suggests that substitution of W⁴² gives rise to the observed hysteresis in the unfolding titrations, and that W⁴² is required for path-independent (un)folding of Ail.

Unlike the W149X mutants, all the folded W⁴² → F/Y/A mutants resist GdnHCl–mediated unfolding, and exhibit a substantial increase in thermodynamic parameters (Table 1 and S3). For instance, the ΔG_{app, W42→A} ^{0, U} is > 1.0 kcal mol⁻¹ higher than Ail WT (Fig. 4B, Table 1 and S3). To verify our observation, we monitored changes in fluorophore rigidity during barrel (un)folding by measuring Trp fluorescence anisotropy (r). In line with equilibrium measurements, the anisotropy (r) measurements also indicate that WT and W149X mutants show no hysteresis (Fig. S8 and S10). In contrast, the W42X mutants show hysteresis and a concomitant increase in the C _m of unfolding (Fig. S8 and S10). The observed hysteresis is consistent in our < λ >, f _U, and r measurements, which validates our inference that W⁴² maintains the path-independent folding of Ail and keeps the barrel under thermodynamic equilibrium, and mutation of W⁴² switches the unfolding pathway to kinetic control. Further, we also observe poor(er) folding efficiency of the W42X mutants and an associated lowering of the folding free energy. Previously, we have also shown an important role of W⁴² during Ail folding in LDAO micelles [25]. Putting together the rigid nature of W⁴², its position at the lipid interface, and the change in free energies (Tables S1, S2, and S3; Figs. 1, 3, and 4), we propose that W⁴² contributes to Ail barrel rearrangement by the likely formation of optimal anchoring interactions with the lipidic milieu.

To obtain molecular insight on how the local Trp environment influences our observation, we carried out a co-evolution analysis to map the interaction network of each indole. W⁴² shows the presence of three aromatic residues (F²⁸, F⁸⁰, Y³⁶) and several polar entities (E³⁷, N³⁹, D⁴⁰, D⁴¹, R⁸¹) in its 5 Å vicinity (Fig. S11). Abundance of polar residues in the vicinity of an aromatic residue such as W⁴² suggests a lipidanchoring role for the Trp, Phe, and Tyr residues at the periplasmic face of the barrel. On the other hand, W¹⁴⁹ is surrounded largely by aromatic (Y¹¹, Y¹³⁶) and hydrophobic residues (A¹², M¹⁵⁰, L¹⁵¹); polar entities (S¹⁴, Q¹⁴⁷, T¹⁴⁸) are less abundant than the vicinity of W⁴², and charged residues are absent (Fig. S11), indicating that W¹⁴⁹ is more lipid solvated than W⁴². Based on the structure of Ail, W¹⁴⁹ is likely to establish T-shaped aromatic interactions with Y¹¹ and Y¹³⁶. Such Trp-Tyr interactions are energetically favorable, and can contribute up to 1.0 kcal mol⁻¹ per interaction [20,27]. Interestingly, W⁴² coevolves with the three hydrophobic and apolar residues P⁷⁸, F⁸⁰, and L⁸⁸, whereas W¹⁴⁹ shows direct co-evolution with A¹³⁴, S¹³⁵, and Y¹³⁶ (Fig. S12 and Table S5). Both tryptophans form a strong evolutionary interaction network with both polar and apolar residues (Fig. S12 and Table S5), with W¹⁴⁹ being more conserved and forming a larger molecular interactome than W⁴². Hence, the evolutionary selection of W⁴² in Ail is a direct consequence of the role of this indole in Ail assembly and scaffold stability.

Overall, our results allow us to reach the following conclusions: (i) at the interface, the thermodynamic preference for residues 42 and 149 follows the order W > Y ≈ F > A; (ii) the mutation of W⁴² lowers Ail folding efficiency; thereby, W⁴² is more important than W¹⁴⁹ for Ail folding, (iii) W⁴² maintains the path-independent folding of Ail and retains the Ail barrel under thermodynamic equilibrium; (iv) once folded, W42X mutants show resistance to chemical denaturation, giving rise to the apparent hysteresis; (v) due to the apparent hysteresis, stability of folded Ail follows the order W42X > WT > W149X, where X = F/Y/A. Overall, we find that W⁴² is required for Ail folding, while W¹⁴⁹ contributes considerably to the stability of the folded Ail barrel. Additionally, we conclude that the observed hysteresis in Ail is driven primarily by the substitution of W⁴². In other words, W⁴² contributes to the activation energy barrier for Ail (un)folding. A recent study identified the role of negative charges in the occurrence of hysteresis in the E. coli β-barrel OmpLA [34]. Our results with Ail add to this finding by showing the importance of interfacial tryptophans for thermodynamic equilibrium and path-independent (un)folding of outer membrane proteins. The observed hysteresis upon substitution of W⁴² might arise from several factors (e.g., the formation of non-native contacts in the folded Ail W42X barrel, the population of unique unfolding intermediate(s), etc.). To further understand the unique molecular features of W¹⁴⁹ (W42X mutants), we carried out detailed spectroscopic measurements to check the local environments of both tryptophans.

5.4. Hyperfluorescent intermediate in the Ail unfolding pathway reveals the likely formation of a molten globule

The significant contribution of Trp residues to folding and stability of Ail directed us to investigate the local environment of W⁴² and W¹⁴⁹. Typically, an equilibrium unfolding profile shows transition of protein from the folded state (RF) to the unfolded state (UF). Here, Ail transitions cooperatively from the unfolded to the folded state between 2.0 M–4.0 m GdnHCl. We have categorized the [GdnHCl] at the start and end of this cooperative transition as partially unfolded (PAR-UF*) and partially folded (PAR-RF) conformations, respectively. We reasoned that at these intermediate [GdnHCl], the protein conformation resembles neither the fully folded nor the fully unfolded states of Ail. The schematic shown in Fig. 6 illustrates the folded (RF), partially unfolded (PAR-UF*), partially folded (PAR-RF), and unfolded (UF) states of the protein (we use this nomenclature merely to simplify our explanation of the complex protein folding process). The fluorescence properties of each state changes on the basis of extent of unfolding and solvent accessibility of the tryptophans.

Fig. 6. Increased rigidity and solvent occluded W¹⁴⁹ suggests the formation of molten globule–like state in Ail unfolding pathway.

(A) A schematic representation of the typical unfolding profile (f _U)of the W42X mutant, highlighting folded (RF), partially unfolded (PAR-UF*), partially folded (PAR-RF), and unfolded (UF) states of the protein. (B) Comparison of Stern-Volmer constants (K _SV) for the different states (RF, PAR-UF*, PAR-RF, UF) of the protein, calculated from acrylamide quenching measurements. The mean K _SV values for unfolding equilibrium data of WT (●, black), W42F (▾, red), and W149F (■, green) are plotted against different unfolded states of the protein. K _SV is largely invariant from RF to PAR-UF* samples of WT and W42F (black arrowhead), suggesting the formation of a solventinaccessible state of W¹⁴⁹ during Ail unfolding. In contrast, note how W⁴² is solvent accessible in the PAR-UF* state. (C) Comparison of the average lifetime (< τ >) for different states of the protein, from unfolding equilibrium samples. An increase (as against a considerable decrease) in < τ > in the PAR-UF* state compared to the RF state is seen only for W42F (black arrowhead), suggesting the presence of a rigid W¹⁴⁹ fluorophore. (D) Bimolecular quenching constant (k _q) for WT and W42F shows a similar pro file as for K _SV (black arrowhead), confirming the formation of a solvent-inaccessible state for W¹⁴⁹. Overall, the data reveal that in the presence of ~2.36 m GdnHCl, Ail undergoes a conformational change resulting in drastic alteration in the local vicinity of W¹⁴⁹. The change in K _SV, < λ >, and k _q suggest the formation of a molten globule–like state, with accompanied hyperfluorescence. The complete data are provided in Table S4. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

When exposed to a chemical denaturant such as GdnHCl, Ail shows unfolding as well as a proportional decrease in the fluorescence intensity (Fig. S9). By and large, folded Ail WT and the W149X mutants exhibit the anticipated decrease in Trp emission in the order RF → PAR-UF* → PAR-RF → UF (Fig. 5, middle and right panels). However, the W42X mutants show ~20% increase in the fluorescence intensity of the PAR-UF* state compared to RF state (Fig. 5A; left panel, 5B; right panel and S6B). The increase in fluorescence intensity of the W42X mutants in the PAR-UF* state suggests that at this intermediate concentration of ~2.4 m GdnHCl, W¹⁴⁹ experiences a change in its local environment where it occludes GdnHCl. In other words, the vicinity of W¹⁴⁹ transitions to a buried (lipid-solvated) hyperfluorescent state before barrel unfolding commences. Prolonged incubation of the W42X mutants in ~2.4 m GdnHCl does not alter the spectral properties, suggesting that this hyperfluorescent species is stable and under equilibrium. Absence of detectable hyperfluorescence during the unfolding of Ail WT and W149X suggests that formation of this species is modulated uniquely by W¹⁴⁹.

Fig. 5.

W42X mutants show the presence of hyper-fluorescent species in the unfolding pathway. Representative Trp fluorescence emission profiles from unfolding equilibrium of WT (center), W42Y (left) and W149Y (right) in specific GdnHCl concentrations (provided in the inset) are shown in the lower panel. Cartoon representation of Ail WT (middle), W42Y (left), and W149Y (right) is shown above the plot. Drop lines in the emission spectra mark the fluorescence emission maxima for each profile. WT and W149Y show progressive decrease in the fluorescence intensity and λ_em-max (downward arrow) upon protein unfolding, whereas W42Y shows an increase in intensity and λ_em-max at 2.36 m from the folded protein in 0.36 m GdnHCl (highlighted by an upward arrow). The observed increase in λ_em-max for W42Y indicates the accumulation of a hyperfluorescent intermediate in the course of unfolding. (B) Folding (left) and unfolding (right) equilibrium profiles (recorded at 48 h), for WT, W42Y, and W149Y. The folding titrations show the anticipated lowering of C _m-start as a consequence of Trp mutation (dotted drop lines). Interestingly, the C _m-start for W42X mutants (shown here is data for W42F) is shifted by ~1.0 m to higher GdnHCl concentrations, and is accompanied by ~20% decrease in the unfolded fraction (f _U) (red arrow), highlighting the accumulation of a hyperfluorescent intermediate in this mutant. The structural properties of this intermediate has been examined additionally using far-UV CD (see Fig. S14 for details). The complete data are provided in Fig. S8–S10. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Far-UV CD, fluorescence, and HSQC measurements confirm that the folded state of the W42X mutants is nearly identical to Ail WT (see Fig. 1C, 2, and S4). We further characterized this hyperfluorescent intermediate using CD. We find a possible variation in the rigidity of W⁴² and W¹⁴⁹, and different local environments for the indole rings in the folded proteins (Fig. S13). Additionally, the hyperfluorescent species (PAR-UF* state at ~2.4 m GdnHCl) in the W42X mutants shows an increased β-sheet content (Fig. S14). This suggests a possible change in the DPC-Ail interaction when only W¹⁴⁹ is retained, giving rise to an observable hyperfluorescent intermediate.

To obtain better insight into the spectral characteristics of W¹⁴⁹, we carried out an in-depth fluorescence analysis. First, we examined indole accessibility and quenching efficiency by measuring the Stern-Volmer constant (K _SV) and bimolecular quenching constant (k _q). In addition, we monitored indole rigidity using anisotropy (r), and the rotational correlation time (τ _c), and compared the results with Trp lifetimes (< τ >). A typical unfolding pro fi le of Ail W42F is presented in Fig. 6A, indicating different species in the unfolding titration, as the protein transitions from RF to PAR-UF*, PAR-RF, and UF states. We compared various parameters for all the mutants from both the unfolding and the folding profiles (Fig. 6B-D, Table S4).

As anticipated, we observe an increase in K _SV and k _q for WT and the W149X mutants upon unfolding, while the corresponding r,< τ >, and τ_c are lowered (Table S4). Remarkably, the partially unfolded hyper-fluorescent intermediate (PAR-UF* in Fig. 6 left panel) seen in the unfolding equilibrium for W42X mutants at ~2.4 m GdnHCl displays lowered K _SV, and k _q; the corresponding values of r,< τ >, and τ_c are anomalous and higher than WT in the mutants possessing only W¹⁴⁹ (Fig. 6B–D and S8, Table S4). The lowered accessibility of the W¹⁴⁹ fluorophore and the associated increase in W¹⁴⁹ lifetime values reaffirms our initial deduction that at GdnHCl concentrations ≈ 2.4 M, W¹⁴⁹ adopts a solvent-occluded lipid-solvated environment. This lipidsolvated state gives rise to the hyperfluorescence that we observe in the equilibrium titrations, and is coupled with an anomalous anisotropy value (also see Fig. S14 for changes in near-UV CD). Such a hyperfluorescent state is not the result of a spatially proximal quencher, as K _SV and k _q would be unaffected by static quenchers [29]. Further, we observe negligible differences in the data of W42F/Y/A mutants, suggesting that the nature of the amino acid at the 42nd position has no impact on the formation of the hyperfluorescent intermediate. Overall, on the basis of K _SV, k _q, r, τ_c, and < τ > measurements, we conclude that the removal of W⁴² and retention of W¹⁴⁹ is necessary and sufficient for the formation of this hyperfluorescent species.

The change in spectral properties of W¹⁴⁹ may arise due to (i) stronger detergent-indole interaction in the PAR-UF* state, or (ii) change in the local environment of W¹⁴⁹ to a buried state. In other words, the buried and solvent excluded conformation of W¹⁴⁹ could either be due to micelle expansion, protein compaction, or an increase in the secondary structure of the barrel. Based on our observations, we postulate that the quencher-inaccessible state of W¹⁴⁹ could resemble a molten globule–like conformation formed during Ail unfolding. Further, this lipid-anchoring property of W¹⁴⁹ may be crucial to establish strong protein-lipid interactions that influence Ail unfolding. Interestingly, the hyperfluorescent intermediate is W¹⁴⁹–dependent and occurs only in the W42X proteins, is absent in W149X wherein only W⁴² is retained, and is masked in WT Ail wherein both the tryptophans are present (see Fig. S14, S15). Therefore, we reason that the hysteresis in Ail W42X mutants arises due to a change in detergent-protein interactions in these proteins, and the intrinsic hyperfluorescence of W¹⁴⁹ is the consequence of the lipid-solvated state of the W¹⁴⁹ indole. We also conclude that the change in spectroscopic properties of W¹⁴⁹ is effected despite conserved substitutions at W⁴².

Finally, we conclude from our observations that Ail undergoes a conformational transition upon GdnHCl–mediated denaturation, and unfolds via a stable hyperfluorescent unfolding intermediate. Occurrence of this hyperfluorescent intermediate gives rise to the apparent hysteresis in the unfolding of Ail W42X mutants. The absence of detectable intermediate for WT and W149X suggests that the lipid anchoring properties of W¹⁴⁹ leads to a change in its fluorescence property by alterations in the detergent-protein interaction. The latter drives the local environment of W¹⁴⁹ to a solvent-occluded lipid solvated environment in the unfolding transition. Similar molten globule–like structures are rare for membrane proteins, and detailed characterization of their structure will help in understanding membrane protein folding, unfolding, and function.

6. Discussion

Understanding the mechanism of transmembrane β-barrel folding is important to deduce the significance of evolutionary selection of each unique polypeptide sequence in balancing protein structure, stability, and its function. Despite the several excellent reports on OMP folding, most of the studies use E. coli proteins (e.g., OmpA [10,11,20], OmpX [21.29], PagP [12], OmpLA [22], FomA [23]). Membrane proteins are known to exhibit species-dependent variations in their folding mechanism [17–19]. Hence, we reasoned that it is important to study the assembly of atypical OMPs. Here, we study the folding, stability, and unfolding pathway of the attachment invasion locus (Ail) protein β-barrel from the category A pathogen Y. pestis. Our study on the role of interfacial tryptophans on the folding of Ail has revealed the unexpected finding that, unlike bacterial and mitochondrial (VDAC2 [24]) OMPs that require tryptophans to critically aid their folding [10.20.21.24.29], interface tryptophans are neither important for the initial process of Ail assembly nor the early events of Ail unfolding. Additionally, our kinetics measurements show that Ail employs a multistep folding and assembly pathway, known to exist primarily in OMP folding in bilayer membranes [10,14], but is rare for OMPs in micelles [16].

Our study also reveals that specific interface tryptophans of Ail contribute to the final barrel rearrangement process, to the post-folding stability, and to the unfolding. In particular, W⁴² is more important than W¹⁴⁹ for the late events in Ail folding and barrel rearrangement. W¹⁴⁹ is crucial for post-folding stability of the Ail scaffold, as it forms strong protein-lipid interactions. Such distinction imparts differential importance to each tryptophan in the Ail scaffold, and demarcates this barrel from its E. coli analog OmpX [21]. We find that W⁴² is required for the path-independent folding and thermodynamic stability of Ail. This is evident from the hysteresis observed when W⁴² is substituted and only W¹⁴⁹ is retained. While the evolutionarily conserved W¹⁴⁹ establishes a strong intra-protein hydrophobic interaction network and contributes to barrel stability, W⁴² not only forms a weaker interaction network, but is also poorly conserved in other OMPs (see Fig. S1). Our study assigns a specific role for W⁴² in Ail, as vital for Ail folding kinetics and stability. The additive effect of both tryptophans on Ail stability correlates well with the evolutionary requirement of Ail for thermodynamic equilibrium, and to maintain plasticity in its structure [35,36]. OmpX (the structural analog of Ail in E. coli) possesses unusually high stability compared to Ail, but is functionally inactive. Therefore, we are tempted to suggest that the evolutionary selection of W⁴² in the Ail sequence for structural plasticity may additionally bear significance. Further studies in this direction are, however, needed.

Interestingly, hysteresis is unique only to the W⁴² mutants, indicating that the path-dependence of the unfolding event (and the associated apparent free energy values ( $\Delta G_{\text{app}}{}^{0,U}$ ) is imposed by the mutation (see Fig. 4A), and is not a consequence of the folding conditions. The occurrence of hysteresis in Ail unfolding provides a unique opportunity to map the hyperfluorescent state and to study the folding and unfolding pathways of Ail independently. Based on the results, we propose a plausible model both for Ail assembly and Ail unfolding. Here, the folding of Ail can be divided into (i) early folding phase that occurs within seconds, and (ii) final barrel rearrangement that occurs over several minutes-hours. The initial folding phase is Trp–independent and does not involve changes in the local environment of the two interface indoles. While the multi-step folding and unfolding kinetics we see for Ail, is reported thus far only for OmpA [16], our results are also in good agreement with the overall assembly mechanisms proposed for OMPs [8,10,16,37]. The anchoring property attributed to the indole side chain is relevant only after the initial barrel assembly is completed. The rearrangement process is now Trp–dependent and is mediated by the interplay of W⁴² and W¹⁴⁹ (Fig. 7).

Fig. 7.

Model for Trp-assisted pathway of Ail assembly. The folding pathway of Ail consists of rapid initial adsorption to the membrane (a) that occurs in milliseconds and manifests as the burst phase in our kinetics measurements. Slow rearrangement, facilitated by the anchoring role of W⁴² (b), gives rise to the completely folded barrel (c). The initial folding of Ail is Trp independent, but the interplay of W⁴² (lipid anchor) and W¹⁴⁹ (strand rearrangement) is crucial for rearrangement and formation of the complete barrel. W⁴² is also more important than W¹⁴⁹ for the stability of the folded barrel. Ail follows a complex unfolding pathway (d–f), which is nucleated in the vicinity of W⁴², and progresses towards W¹⁴⁹. Here, W¹⁴⁹ –mediated anchoring gives rise to a hyperfluorescent intermediate (d) with increased structural content of the barrel, during Ail unfolding. Overall, the folding and unfolding of Ail is influenced considerably by the uniquely positioned interface tryptophan residues. W⁴²–mediated anchoring of Ail during barrel folding, and W¹⁴⁹–mediated anchoring of Ail in post-folding stability suggests distinct roles for interface aromatics in this Yersinia OMP.

Notably, the folding process in vivo requires the barrel assembly machinery (BAM) complex, which acts as the chaperone. On the other hand, OMP unfolding is BAM-independent. Our observation of the novel hyperfluorescent intermediate provides the first experimental data that differentiates the unfolding process from folding. Thermodynamic studies of OMP folding are preferably carried out in conditions that facilitate path independence. Hence, molecular factors that selectively drive OMP unfolding are usually considered in similar lines as the folding events. The occurrence of hysteresis as a consequence of this hyperfluorescent intermediate in Ail provides us with the first unique opportunity to study the molecular steps involved in the unfolding of an OMP. We find that Ail unfolding is nucleated in the vicinity of W⁴², and progresses towards W¹⁴⁹ (the post-folding anchor). The anchoring ability of W¹⁴⁹ resist barrel unfolding and gives rise to the hyperfluorescent species before unfolding is complete. Hyperfluorescence is usually the signature of a folding intermediate that is thermodynamically stabilized [38]. A similar molten globule–like flexible state has been observed for ankyrin-repeat protein, ubiquitin, and yeast phosphoglycerate kinase [38–40], but has not yet been reported for any OMP.

The occurrence of hyperfluorescence in the unfolding pathway in vivo and the significance of this state is yet to be established. The latter will require novel sophisticated methods for direct detection of OMP folding in the Yersinia outer membrane. Nonetheless, we speculate based on our current observations, that W⁴² and W¹⁴⁹ are likely to play vital roles in the folding and anchoring, respectively, of Ail in the Yersinia outer membrane. A recent report on the folding pathway of the OmpA barrel in varying lipidic conditions highlights the relevance of results obtained from in vitro studies to the mechanism by which OMP assembly takes place in the cell [8,37]. Similar studies with other atypical OMPs can lead to the identification of other molten-globule like intermediate states, and provide novel insights into the unfolding pathway of kinetically stabilized OMPs. Our findings on Ail will serve as an excellent starting point to further understand the parallel (complex) folding and unfolding mechanisms of other biologically important OMPs from pathogenic bacteria, and mitochondria.

Abbreviations

Ail

attachment invasion locus protein

DPC

n-dodecyl phosphocholine

DPR

detergent-to-protein ratio

OMP

outer membrane protein