Introduction

Transmembrane (TM) proteins are, by and large, ‘kinetically trapped’ in biological systems1,2,3,4. Early evidences that Anfinsen's equilibrium thermodynamic model for refolding soluble proteins cannot be extended to membrane proteins arose in the ‘90s, from studies of multi-pass hydrophobic helices by Popot, Engelman and others5,6,7,8,9,10,11,12. The folding process of TM helices fit well to a two-stage model which described each helix as an autonomous folding unit that assembled in the apolar interior of the lipid bilayer by an intricate, largely irreversible, mechanism following alternative pathways for protein assembly (folding) and dissociation (unfolding)3,5,8,13. TM β-barrels are also known to adopt a multi-step refolding event which oftentimes involves rapid protein collapse and adsorption onto the lipid bilayer, followed by a much slower insertion event that culminates in barrel assembly1,12,14,15.

Speculation on whether Anfinsen's hypothesis would be extendable to membrane proteins, owing to their unique membrane environment, was resolved when more recent studies demonstrated that in vitro refolded TM β-barrels of bacterial and human origin are in thermodynamic equilibrium12,14,15,16,17,18. Such systems follow superimposable reversible pathways for refolding and unfolding in vitro, in the carefully screened lipid and buffer conditions that permit such equilibrium10,15,19. The native protein conformation in such cases is a consequence of the Gibbs free energy minimum achieved by the system and the folding event itself is entropically driven and enthalpically balanced. Not surprisingly, however, as Popot and Engelman surmised5,8, most membrane proteins do not exist in thermodynamic equilibrium and are plagued by hysteresis2,4,10,11,12,20. A ‘hysteresis loop’, or ‘open loop’, is a condition wherein microscopic differences in a system's configuration perturb thermodynamic equilibrium, sacrificing the free energy minimum for a presumably enhanced functionality11.

In a large population of TM proteins, the kinetics of interconversion between the folded state and unfolded form, through metastable intermediates, gives rise to hysteresis in an otherwise equilibrium system4,10,11,12. Such systems are trapped by kinetic stabilization, a process that is influenced by residue-specific protein-protein and protein-lipid interactions, lipid or detergent morphology and the inherent tendency of hydrophobic proteins to form stable aggregates outside their lipid environment. Hence, despite the reversibility observed in a handful of TM β-barrels thus far, the more pressing question is the cause of hysteresis in these proteins. The wide repertoire of refolding methodologies and membrane mimetics used for the 8–24 stranded TM β-barrels in vitro21, as well as the influence of parameters such as temperature and pH, employed in the reaction, on the physical properties of the lipid or detergent10,11,12,22, point towards diversity in the metastable states populating the folding and unfolding pathways. Identification of key elements in the protein which contribute to hysteresis, would allow us to understand β-barrel refolding in vitro and shed light on membrane protein recycling in vivo.

We probed the factors contributing to the kinetic component of membrane protein stability using the 17.6 kDa 8-stranded TM β-barrel Ail from Yersinia pestis, the causative agent of bubonic, pneumonic and septicemic plague, as our model system. The Attachment invasion locus (Ail) protein belongs to the Ail/Lom family of outer membrane proteins coded by the y1324 locus in both KIM and CO92 strains of Y. pestis and is present only in pathogenic Yersiniae23,24. Ail expression levels coincide with the pathogenicity of the host bacterium, owing to its associated role in virulence25,26 and it functions in adhesion and internalization of the bacterium into host cells through Yop delivery and ‘attack complex’ formation in vivo25,27,28. Ail also confers serum resistance to complement-mediated bacteriolysis, blocking phagocytosis and survival within macrophages26,27,28 and is therefore one of the prime targets for drug development against this category A bioterrorism agent.

Demarcation of the transmembrane and loop regions of Ail and other TM β-barrel is achieved by the aryl groups populating the ‘aromatic girdles’, which defines barrel depth and the lipid or micelle interaction interface29,30,31,32,33,34,35,36. Aryl residues mediate structural stability and associated ‘invasin’ property of Ail23, exposing binding pockets for protein function. Stereospecifically positioned aromatics therefore play a key role in barrel formation (folding), stability (susceptibility to unfolding), depth definition (TM span) and activity29,31,32,33,35,36,37,38,39. Biophysical characterization of how Ail interacts with its folding environment and the contribution of aromatics, particularly Trp residues, would therefore throw light on protein behaviour.

Herein, we address the impact of micelle-protein interaction and the aryl moiety tryptophan, on the thermodynamic versus kinetic stability of Ail. We demonstrate that, of the two indoles, the Trp required for Ail refolding in lauryldimethylamine oxide (LDAO) micelles is different from the Trp that confers subsequent barrel stability. Furthermore, DPR (detergent-to-protein ratio) holds significant bearing on protein unfolding, while the effect on the refolding event is less significant. Such differential contribution of Trp residues has not been observed thus far for other known 8-stranded TM β-barrels. We discuss the implications of hysteresis and the associated contribution of aryl groups, on Ail function and barrel stability in Yersiniae.

Results

Tryptophans reside in unique local environments but do not affect Ail refolding ability

Yersinia Ail possesses two Trp residues25; the crystal structure maps W42 to the membrane interface whereas W149 is slightly closer to the lipid core (Fig. 1a). To assess the influence of Trp mutations on β-barrel formation, we carried out gel mobility shift assays24,40,41,42, of unboiled Ail samples refolded in 20 mM, 50 mM and 100 mM LDAO. Wild type Ail (Ail-WT) and the single Trp mutants (Ail-W42F and Ail-W149F) exhibit retarded gel mobility and migrate at ~24 kDa on SDS-PAGE (Fig. 1b), indicating protein folding and barrel formation in the three DPRs used in this study. Ail, thus refolded, retains the ability to bind heparin25 (not shown) and exhibits resistance to proteolysis by proteinase K (Fig. 1b), suggesting that a Trp→Phe substitution does not affect the ability of Ail to refold in various LDAO concentrations. Comparable secondary structure content for all three proteins from circular dichroism measurements (not shown), substantiates this observation.

Figure 1
figure 1

Response of Trp residues of Ail to DPR and its influence on Ail refolding.

(a) Ribbon diagram of Ail obtained from the crystal structure (PDBID: 3QRA25) highlighting the two Trp residues as red spheres. The putative transmembrane segment is also indicated. Segments of two extracellular loops, absent in the structure, are denoted as dotted lines. (b) SDS-PAGE gel mobility shift assay of unboiled Ail-WT, W42F and W149F samples refolded in 20 mM (top), 50 mM (middle) and 100 mM (bottom) LDAO. Refolded samples also subjected to pulse proteolysis using proteinase K (PK) are indicated as (+) and undigested samples as (-). Bands corresponding to the folded dimer (FD) and higher order oligomers are most prominent in the 20 mM LDAO sample. Refolded monomer (FM), migrating at ~24 kDa and the unfolded monomer (UM) migrating at ~17.0 kDa, are also marked. Unfolded Ail in 8.0 M urea served as the control (C). The percentage of observed Ail oligomers (%O) in each condition was quantified by densitometry42 and is provided below the respective sample lanes. Band positions of molecular weight standards are indicated beside each gel. The unfolded control from the bottom gel has been shown in the top and middle gels and is marked (*). Dotted lines separate different sections of gels that have been presented together. Note that the various DPRs underwent different dilutions before SDS-PAGE (to minimize interference from LDAO on band migration) and therefore protein concentrations across the three gels are not similar. (c) Plot of Trp fluorescence quenching of Ail-WT (circles), W42F (inverted triangles) and W149F (squares) with increasing acrylamide concentrations, fitted to a linear equation (fits are shown as lines), to obtain the Stern-Volmer constant (KSV) (shown as inset). Shown here are data obtained in 20 mM (black), 50 mM (red) and 100 mM (green) LDAO. Note that the W149F construct displays an increase in KSV upon addition of LDAO, whereas corresponding values for the WT and W42F proteins are largely DPR insensitive (also see Supplementary Fig. 1). KSV values for unfolded Ail (in 8.0 M GdnHCl; all three constructs) was ~7.4 M−1 and aggregated Ail was ~3.1 M−1.

Full size image

Interestingly, Ail exhibits oligomeric forms on SDS-PAGE gels of unboiled samples (Fig. 1b). Ail multimerization has not been observed earlier and no known function has been associated with Ail multimers. It is therefore likely that the higher order oligomers are an artefact of in vitro conditions. In high DPR, the oligomeric forms are lowered, supporting a non-specific barrel association that may be diffusion driven. A similar behaviour has been observed earlier for OmpA43 and OmpX44, both 8-stranded barrels structurally similar to Ail. Notably, oligomerization is highest in W149F, which may perhaps bear implications on barrel behaviour.

Using acrylamide quenching measurements, we examined the accessibility of the indole ring, to probe for possible variations in the local Trp environment of the three proteins. Our data (Fig. 1c) indicates that W42 (W149F mutant) undergoes an increase in indole solvent accessibility with increasing LDAO (and DPR), reflected in the increase in the KSV value. This may arise from contributions from static and dynamic quenching of W42 fluorescence by acrylamide or complex protein-micelle-quencher interactions (see Supplementary Fig. 1).W42F and WT constructs show lower KSV values that do not change significantly with LDAO concentration (Fig. 1c), suggesting that tryptophans in these proteins are occluded from acrylamide.

A better insight to the Trp environment can be obtained from measurements of average lifetime (<τ>) and calculation of the bimolecular quenching constant (kq) (Table 1). The former provides information on the rigidity (longer lifetimes indicate the presence of a rigid fluorophore), while the latter specifies the degree of solvent exposure of the fluorophore45. Interestingly, W149 (W42F mutant) shows the lowest lifetime, despite being buried. In contrast, W42 (W149F mutant) displays higher lifetimes that show a strong DPR dependence, coupled with higher Trp anisotropy. This together indicates that the W149 side chain is conformationally mobile, despite residing in a micellar environment (low KSV). W42, however, becomes increasingly immobile in high DPR, due to its location at the interface, thereby displaying higher lifetime and anisotropy, as a consequence of poor non-radiative losses in fluorescence. The WT displays an intermediate behaviour with values closer to W42F, possibly balancing favourable solvent-indole interactions for both W42 and W149.

Table 1 Summary of Trp fluorescence parameters measured for Ail-WT and its Trp mutants in various DPRs
Full size table

Reported kq values for moderately exposed Trp residues are typically <1.5 × 109 M−1s−1, while solvent (and thereby quencher) accessible indoles display kq >1.5 × 109 M−1s−145,46. In low DPRs, W149F particularly possesses the most solvent-accessible Trp (kq = 1.85 × 109 M−1s−1; Table 1), while in high DPRs, the kq is lowered in all three proteins, with WT and W42F now possessing relatively less exposed Trp residues compared to W149F. However, since all the kq values are within the accepted range for moderately exposed indoles (kq for buried Trp lies in the 0.6–1.0 × 109 M−1s−1 range45,46), we propose that all three proteins show the presence of interfacial Trp residue(s); differences arising from their respective interaction pockets give rise to the observed variation in KSV and <τ>. Put together, these properties provide the order WT ≥ W42F > W149F for increase in solvent exposure of the indole. Both Trp residues in Ail therefore reside in different local environments and possess features that are uniquely modulated by their surroundings; however, this does not influence barrel refolding.

Ail refolding is largely DPR independent but displays an LDAO-stabilized intermediate

Dilution of guanidine hydrochloride (GdnHCl)-denatured Ail drives rapid barrel formation within minutes, in LDAO. We therefore monitored the refolding process of Ail-WT, W42F and W149F using a GdnHCl gradient and Trp fluorescence as the probe. A typical emission profile acquired at 48 h is shown in Fig. 2a. The Cm values (chemical denaturation mid-point) derived from folded fractions (fF) calculated at the λem-max for the folded protein (330 nm) or the observed anisotropy values (Fig. 2b and Supplementary Figs. 2–4), are provided in Fig. 2c. In 50 mM and 100 mM LDAO, the anisotropy Cm shows lowering and fFCm increases in all three proteins; the global change in Cm is, however, marginal. We observe considerable differences in the Cm values derived from Trp fluorescence and anisotropy measurements between the various proteins only in low DPRs. Furthermore, in 20 mM LDAO, WT shows a high anisotropy Cm that could arise from association of unfolded Ail with LDAO moieties at the vicinity of both Trp residues, thereby nucleating refolding. The W42F mutant, which lacks the interface Trp residue, has the lowest Cm and requires sufficient lowering of GdnHCl for refolding.

Figure 2
figure 2

Refolding of Ail in LDAO micelles is largely DPR independent.

(a) Representative Trp emission profiles of Ail-WT and Trp mutants in 20 mM LDAO, highlighting the observed increase in fluorescence intensity and prominent blue shift in the λem-max (depicted in red to blue color scheme), as the protein refolds when GdnHCl is lowered from 8.0 M to 1.6 M (red to blue). The fluorescence intensity of Trp 149 in the W42F mutant is lower than Trp 42 in the W149F mutant. This is likely to arise from non-radiative losses in Trp 149 fluorescence, although the presence of a proximal quencher cannot be eliminated. (b) Fraction folded (fF) derived from Trp emission intensity at 330 nm (top) and corresponding anisotropy (r) values (bottom) for WT (), W42F () and W149F () in 20 mM (black), 50 mM (red) and 100 mM (green) LDAO. Data were collected at every 0.1 M increment of GdnHCl and the mean of two – three experiments were calculated. However, representative mean data points are shown in fF plots without error bars, for clarity. All samples (lower GdnHCl concentrations) were checked using gel mobility shift on SDS-PAGE for completion of the refolding process. While ~100% refolding was observed, the presence of trace amounts of unfolded protein cannot be completely excluded. Representative data are shown for the anisotropy experiments. Fits are shown as solid lines and corresponding LDAO concentrations are indicated in each plot. Note the appearance of an intermediate (indicated by an arrow) at ~4.0 M GdnHCl only in the fF plots of W42F in all LDAO concentrations and additionally in WT in 100 mM LDAO. (c) Cm values obtained from fF (blue, solid fill) and anisotropy (r) (blue, diagonal stripes) data are compared for the three Ail constructs across various LDAO concentrations. Note that the Cm values are different only in 20 mM LDAO (DPR 700:1) and are largely analogous in 50 mM (DPR 1750:1) and 100 mM (DPR 3500:1) LDAO. Error bars represent goodness of fits. Also see Supplementary Figs. 2–4.

Full size image

Marginal differences in the Cm (fF) and Cm (anisotropy) suggest that barrel refolding conceivably proceeds by sequestering LDAO in micelle-binding hydrophobic pockets, following which the barrel assembles only at the threshold GdnHCl concentration, in a DPR-independent manner. Interestingly, we noticed the population of a possible folding intermediate in high DPR, which appeared in the order W42F > WT > W149F (arrow in Fig. 2bfF plots). To probe this further, we examined Ail refolding in 200 mM LDAO (DPR 7000:1) (Fig. 3a). The data not only clearly demonstrates the presence of a refolding intermediate in all three Ail constructs (three-state unfolding) but also its surprising absence in the corresponding anisotropy measurements (two-state transition). The Cm for the transition from intermediate to folded protein obtained from fF data compares well with the anisotropy Cm (Fig. 3b), suggesting that W42 and/or W149 in the folding intermediate experiences a change in the polarity of its local environment but retains side chain mobility. Such behaviour mimics the transition of the amphipathic indole from a polar (solvent-exposed) to apolar (lipid- or micelle-exposed) environment before completion of the folding process and is characteristic of a vesicle-adsorbed barrel intermediate1. In our micellar systems, aromatic residues at the two girdles could sequester LDAO to nucleate barrel formation, which we observe as a blue-shifted λem-max for the mobile fluorophore.

Figure 3
figure 3

A refolding intermediate of Ail is observed in high DPR.

(a) Comparison of the folded fraction (fF) and anisotropy values of Ail-WT (, purple) and the Trp mutants W42F (, blue) and W149F (, grey) refolded in 200 mM LDAO (DPR 7000:1) indicates the presence of a stable intermediate species in the fF (indicated by an arrow), which is absent in the anisotropy (r) data. Fits are indicated by solid lines that are color-coded to match the corresponding symbols. (b) Plot of the Cm values calculated from the fF (solid fill) and anisotropy (r) (diagonal stripes) data using linear extrapolation. Color codes are matched to Fig. 3a. For the fF data, Cm reflects the mid-point of the transition of the ‘adsorbed’ protein intermediate to the fully refolded β-barrel form. The lowest Cm is seen in the case of W42F, suggesting destabilization of this protein in very high DPR. Also see Supplementary Figs. 2–4.

Full size image

Surprisingly, a uniform lowering of the Cm is observed in all three proteins in high DPR (7000:1; compare Fig. 2c and Fig. 3b), with the values now bearing semblance to the data from 20 mM LDAO, suggesting that the stability of folded Ail is marginally lowered. Barrel stability therefore exhibits a non-linear dependence on micelle concentrations. Tight micelle packing in 200 mM LDAO is less tolerant to perturbations of the micellar forms by the protein, thereby lowering the folding efficiency of Ail and promoting the formation of a stabilized (adsorbed) intermediate. Such acute DPR dependence has been observed previously for the human VDAC-2 19-stranded barrel20 and may be a property of the LDAO micelles. This effect is most dramatic in Ail-W42F, which also shows lower unfolding cooperativity (mapp values in Supplementary Table 1), with changes in DPR. The conspicuous DPR dependence of this mutant indicates that W42 plays a crucial role in driving proper refolding of Ail.

The and mapp values calculated for the refolding process are summarized in Supplementary Table 1 (also see Supplementary Fig. 4). As the Cm values are largely similar up to 100 mM LDAO, marginal variations in the therefore arise from changes in the cooperativity of the refolding process. This is affected both by the LDAO concentration as well as the Trp mutation. The values are low in 20 mM LDAO, as anticipated. With increase in LDAO, increases due to an increase in the refolding cooperativity (Supplementary Fig. 4). Very high LDAO (200 mM), however, lowers these thermodynamic parameters for all three proteins. Furthermore, in Ail-W42F, the barrel folding cooperativity is lowered, thereby affecting the mapp and accounts for the lowest for this protein (Supplementary Fig. 4). Overall, our refolding experiments therefore provide us with the order W149F > WT > W42F, for cooperativity in Ail barrel formation when the denaturant concentration is lowered, in all the three DPRs. Based on our results, it also seems likely that W42 assists the barrel in overcoming the kinetic barrier stabilizing a trapped folding intermediate.

W149 determines stability of refolded Ail in a DPR-dependent manner

In GdnHCl, although (un)folding can be achieved in various LDAO concentrations, the equilibrium unfolding experiments demonstrate hysteresis within experimental time frames, which is not uncommon for TM β-barrels10,11,15 and suggests the existence of alternate folding and unfolding pathways and/or dissimilar activation energy values in the two processes. Using Trp fluorescence (Fig. 4a and Supplementary Figs. 5–9) and anisotropy (Fig. 4c), we evaluated the contribution of tryptophan and DPR to the stability of refolded Ail and derived the apparent thermodynamic parameters (Supplementary Table 1). Contrary to folding, Ail unfolding follows a two-state transition and we do not detect intermediates in this event. Ail refolding does not depend substantially on the LDAO amounts used in the reaction. However, unfolding of all three proteins is strongly modulated by the DPR and the presence of higher LDAO amounts delays the onset of the unfolding event (Fig. 4b). This is reflected in an increase in the Cm with increasing LDAO concentrations (Fig. 4c). Unfolding, is however, nearly completed at the highest GdnHCl concentrations (Supplementary Figs. 5–8), even with sufficient micellar forms in solution (Supplementary Figs. 10–12).

Figure 4
figure 4

Ail unfolds in a DPR-dependent manner.

(a) Representative Trp emission profiles of Ail-WT and Trp mutants in 20 mM LDAO, highlighting the observed reduction in fluorescence intensity accompanied by a prominent red shift in the λem-max (depicted in red to blue color scheme), as the refolded protein unfolds when GdnHCl is increased from 0 M to 6.4 M (blue to red). (b) GdnHCl-dependent unfolding of Ail monitored through Trp fluorescence, plotted as the variation in unfolded fractions (fU) (top graphs) and the corresponding anisotropy (r) change (bottom graphs), with increasing denaturant concentration. Symbol and color codes for Ail-WT (), W42F () and W149F () are black (20 mM), red (50 mM) and green (100 mM) LDAO. Data collected at increments of 0.1 M GdnHCl from two – three independent experiments were used to calculate the mean fU data set. Shown here are representative points from the mean data without the error bars, for clarity. A representative dataset is provided for anisotropy plots. Fits are shown as solid lines and respective LDAO concentrations are provided in each plot. (c) Plots of Cm values obtained from fU (blue, solid fill) and anisotropy (r) (blue, diagonal stripes) data for the three Ail constructs across various LDAO concentrations. Note the increase in Cm with increasing LDAO (and DPR) in the case of WT and W42F, while W149F is less affected by changes in DPR. Error bars represent goodness of fits. Also see Supplementary Figs. 5–9.

Full size image

A remarkable difference between WT Ail and the single-Trp barrels is observed in the unfolding experiments. The apparent thermodynamic parameters derived for the refolding process provided us with the order W149F > WT > W42F. Cm values obtained from fU calculations of the unfolding process (Fig. 4c and Supplementary Figs. 8–9), provides us with the order W42F ≥ WT > W149F, which is the exact reverse of the Ail refolding process. Furthermore, resistance of Ail to denaturation by GdnHCl increases proportionately with the LDAO concentration. Indeed, in 100 mM LDAO, WT and W42F barrels undergo complete unfolding only in the highest GdnHCl concentrations (~6.4 M) used in the experiment. This is in contrast to the lowering of the (folding free energy) observed for Ail in the refolding reaction in very high LDAO concentrations (Supplementary Fig. 4). Hence, our experiments suggest that Ail responds differently in the refolding and unfolding processes, when the DPR is varied (the source of hysteresis; described later).

Calculated values (unfolding free energy), derived from the unfolding curves are significantly lower than the calculated for the refolding process (Supplementary Table 1). This arises from the lower mapp values for the unfolding process. Further the only shows a marginal increase for all three proteins, with increase in DPR. Hence, despite the high Cm, we speculate that this lower arises from incomplete dissociation of LDAO from the protein, even after barrel unfolding. This causes the unfolding to be less a cooperative process, unlike the highly cooperative refolding event.

Most interestingly, the anisotropy Cm values not only follow the order WT > W42F > W149F but also escalate significantly with change in LDAO (and DPR) amounts. Despite the Trp fluorescence measurements (Fig. 4b, top panel) and SDS-PAGE analysis (not shown) indicating near-complete barrel unfolding even in the presence of intact LDAO micelles (see Supplementary Figs. 10–12), higher anisotropy values under similar conditions suggest a preponderance of LDAO-bound protein. While the anisotropy Cm is lowered with increase in LDAO for the refolding process (see Fig. 2c), we observe that the unfolding profiles show higher anisotropy with increasing LDAO. Efficient sequestering of tightly-bound LDAO, even after barrel unfolding, likely results in a complex protein-LDAO association event, leading to the higher anisotropy Cm. Such tightly bound complexes in unfolded membrane proteins have been proposed earlier47. Moreover, this increase is predominant in Ail-WT and Ail-W42F, pointing to a critical role of W149 in anchoring the refolded protein to its micellar environment.

Discussion

Several studies, which address the contribution of aromatic residues to the stability of β-barrel membrane proteins prefer model systems including OmpA, PagP, FomA and, to some extent, VDAC1,14,15,18,19,32,33,36,38,40,48,49,50,51. Mutational analyses in OmpA and PagP have attempted the Trp→Ala as well as Trp→Phe substitutions32,33,38; such studies have shown that the more dramatic replacement of aryl groups with aliphatic side chains in the water-bilayer interface considerably lowers the free energy of the refolded barrel. Surprisingly, however, even conserved Trp→Phe substitutions that retain the aromatic character at the mutation site, have indicated an average loss in free energy of between 1.0–2.0 kcal/mol33. To our knowledge, none of the studies so far have examined the role of aromatic residues in the refolding process itself. Here, we specifically addressed the requirement of the two Trp residues of Y. pestis Ail, on barrel folding and stability and by selective Trp→Phe mutations, we have been able to deduce the contribution of the amphipathic Trp residue to the protein scaffold.

Cumulative analysis of the apparent thermodynamic parameters derived for the three proteins from GdnHCl-mediated (un/re)folding experiments, in the three DPRs examined (700:1, 20 mM LDAO; 1750:1, 50 mM LDAO; 3500:1, 100 mM LDAO) provides us with an overall variance in the order 20 < 50 ≈ 100 > 200 (mM) LDAO. The difference between Cm values derived from fits to the fF data and anisotropy measurements are most prominent in low DPRs (Fig. 2c), suggesting that unfolded Ail has the ability to sequester LDAO molecules even prior to the barrel collapse event. This process occurs in the tryptophan vicinity, leading to the higher anisotropy Cm and may either be independent of barrel formation or may serve as a nucleating factor for refolding. Furthermore, in the W42F mutant, accumulation of an intermediate in the refolding reaction and based on the low mapp values, we propose that W42 facilitates the Ail barrel to better overcome trapped refolding intermediates, mediating a cooperative refolding process in various LDAO concentrations (Fig. 5; also see Supplementary Fig. 13). Additionally, we find that Ail refolding requires pre-formed micelles, as observed for OmpA48, but detergent adsorption seems to precede barrel formation.

Figure 5
figure 5

Hysteresis in Ail arises from variations in barrel-micelle interaction efficacy.

(Top) Comparison of the GdnHCl-mediated refolding (filled circles) and unfolding (open circles) curves, shown here for Ail-WT. Data were obtained from unfolded fractions (fU) derived from Trp fluorescence, in 20 mM (black), 50 mM (red) and 100 mM (green) LDAO. The mean values from independent experiments are provided here, without error bars. Notice the cooperativity of the refolding reaction and the comparatively low cooperativity of the unfolding reaction. (Bottom) Representative Trp anisotropy values obtained for the GdnHCl-mediated refolding (filled circles) and unfolding (open circles) of WT Ail. Note that while the fU data for the unfolding experiment tends to 1.0 in 100 mM LDAO (top panel), the corresponding anisotropy values do not reach ~0.06 (for a fully unfolded protein), suggesting that unfolded Ail probably binds substantially to LDAO in the vicinity of the aromatics. A similar color scheme is retained in all datasets. Also see Supplementary Fig. 13.

Full size image

The low Cm of W42F in the refolding experiments in 20 mM LDAO merits comment (see Fig. 2c and Supplementary Fig. 13 and Supplementary Table 1). Hydrophobic interactions mediated by the ‘aromatic girdle’ of TM β-barrels act as critical determinants between favourable protein-lipid/detergent interactions that result in protein folding and unfavourable protein-protein interactions that cause aggregation. Indeed, Trp residues in the aromatic girdle act as ‘stop transfer’ signals during the coupled membrane protein synthesis – folding process30,52, define the ‘span’ of the TM region of β-barrels52, explaining their occurrence at the bilayer interface in several membrane proteins (Table 2). Our results contradict the anticipated destabilization of W149F mutant, which lacks the transmembrane Trp. We speculate that in Ail, it is W42 (and not W149), which defines the ‘barrel edge’; replacement with the more hydrophobic Phe results in pronounced stabilization of a partly folded intermediate and induces ‘roughness’ in the folding energy landscape of the W42F mutant, allowing trapped on-pathway intermediates to be experimentally detected (Fig. 6).

Table 2 Position and distribution of Trp residues in structures of known transmembrane barrels
Full size table
Figure 6
figure 6

Schematic highlighting the contribution of W42 and W149 in the folding and stability of Ail in LDAO micelles.

WT Ail (blue/cyan) is shown in the center, W42F (brown) is shown to the left and W149F (green/yellow) to the right of the folding funnel. Dotted lines depict possible folding pathways on the folding funnel (generated using a Mathematica® notebook available at www.oaslab.com). W42 of Ail plays a key role in driving barrel refolding and therefore mutation of this residue populates refolding intermediates during barrel formation, as depicted here for the W42F mutant (brown). Furthermore, presence of only W42 (in the W149F mutant) lowers the incidence of stable intermediates in the refolding pathway. In all three proteins, increasing the LDAO concentration affects the folding efficiency by stabilizing the on-pathway intermediates. Once folded, barrel stability is determined by W149 and the three proteins now follow the order W42F > WT > W149F. The most favorable free energy minimum in LDAO micelles is achieved in the Ail barrel possessing a single W149 in the transmembrane region.

Full size image

The Ail unfolding process shows a distinct variance of 20 < 50 < 100 (mM) LDAO with a DPR- and LDAO-dependent increment of the Cm. This results in a marginal increase in the unfolding free energy; however, the cooperativity of the unfolding process is not affected (Supplementary Fig. 9). It is noteworthy that while the is in the approximate range of -6.0 to -11.0 kcal/mol, the values are significantly lowered (~4.0 kcal/mol) and have correspondingly low mapp values (< -1.0 kcal/mol.M). Despite the higher Cm values, the obvious loss in unfolding cooperativity (compared to the refolding cooperativity; see Fig. 5. Also see Supplementary Fig. 4 and Supplementary Fig. 9) suggests that LDAO does not stabilize Ail to the extent observed for OmpX42 or VDAC18. The highest stability is observed in the W42F mutant, which retains W149, suggesting that W149 anchors the protein to the micelle environment and delays the onset of unfolding. Hence, it is likely that W42 has a destabilizing effect on the refolded protein. Furthermore, in the case of Ail, once folded, the barrel is likely to retain bound LDAO molecules, which gives rise to the difference in Cm values for anisotropy and fU calculations. Hysteresis thereby ensues in Ail (discussed later) and W149 may contribute to this process. We do not, however, rule out the possibility that our observations on role of indoles may differ in lipid systems versus detergent micelles employed in this study. For instance, earlier studies using other β-barrels have observed substantial variations in barrel (un/re)folding kinetics in lipid vesicles versus micelles16,19,33,49,53,54.

It is tempting to propose that W149 has little role during folding, but establishes stereospecific contacts after barrel formation is completed. Such interactions are critical for subsequent Ail stability and are not mimicked by Phe in that position39,52. On the other hand, presence of W42 seems to be necessary only for the refolding process and does not contribute to the subsequent stability of the folded protein. Comparable data for the WT and W149F (Fig. 2 and Fig. 3) and several instances of similar behaviour of refolded WT and W42F Ail (see Fig. 4bfUCm) support the importance of W42 during barrel formation and W149 for barrel stabilization (Fig. 6). Such positional contribution of indoles could evolutionarily demarcate various 8-stranded β-barrels. For instance, the close homologue of Yersinia Ail in E. coli, namely, OmpX, shows unusual thermal stability and is resistant to aggregation in LDAO micelles42. Despite > 40% identity in sequence and nearly superposable structures25, the distribution of aromatics, particularly the positioning of W42 (of Ail) versus W76 (of OmpX), is different in both proteins. Likewise, a thermodynamic study on Trp mutants of another 8-stranded TM barrel from E. coli, outer membrane protein A, shows a global reduction in barrel stability upon mutation of several Trp residues32,33,38,55. Put together, this suggests that a contextual evaluation of aromatic contributions to β-barrels is necessary30,31,32,33,39,52.

The elaborate assembly of transmembrane β-barrels necessitates that the folding equilibrium is sizably favoured towards a stable refolded protein. Such complex folding pathways oftentimes result in kinetically trapped proteins that display hysteresis and/or irreversibility in their barrel formation pathway. While thermodynamic equilibrium can be achieved for similar 8-stranded TM β-barrels such as OmpA and PagP14,15,19,32,33, other studies have recognized the different responses elicited by the folding and unfolding kinetics of these barrels to factors such as lipid concentrations14,18,36,56. Residue-wise contributions, based on stereospecific positioning and protein-lipid interaction mechanisms31,32,57,58, may together heavily impact (un)folding rates and pathways, giving rise to the observed hysteresis. Such kinetic stability may be critical, for outer membrane protein function in hostile environments and mandated by the low protein recycling observed for TM barrels.

In LDAO micelles, the Ail protein of pathogenic Yersinia is kinetically stabilized, which we deduce from the observed hysteresis in the folding/unfolding pathways and the differential responses of both mechanisms to the presence of chemical denaturants (Fig. 5). Such hysteresis in Ail is likely to arise from differences in the activation energy barrier for the folding/unfolding events, giving rise to different apparent free energy values (Supplementary Table 1). In GdnHCl, while we obtain reversibility, hysteresis may be introduced in either the refolding or the unfolding process15. Indeed, a comparison of the Cm values (Supplementary Fig. 13a, bottom panel) of both processes derived from fU plots (Supplementary Fig. 13a, top panel), clearly demarcates the DPR dependence of the unfolding process, suggesting that in Ail, this event falls outside experimentally quantifiable timescales. While considerable unfolding can still be achieved in Ail in ~24 h, studies on OmpA have demonstrated this process to be sizeably slow, attaining completion in days (GdnHCl-denatured) to weeks (urea-denatured), or may require higher temperatures16,19,54. A similar pattern from the anisotropy Cm supports our conclusions that the presence of excess LDAO (up to 100 mM) stabilizes the refolded protein and slows the onset of unfolding. Moreover, we can achieve Ail refolding within minutes, under ambient conditions, indicating that barrel formation is spontaneous when denaturant levels are sufficiently lowered.

By and large, the prolonged unfolding timescales of Ail barrel in high DPR is akin to the slow unfolding of green fluorescent protein, wherein chromophore formation by the covalent modification, leads to a ‘dual-basin’ energy landscape in equilibrium (un)folding experiments of this protein59. In Ail, such ‘locked’ barrel structures are facilitated by stabilizing interactions of the transmembrane segment with its surrounding lipid or detergent environment. These interactions are directed by the transmembrane W149, which contributes significantly to the high even in micellar systems (see Supplementary Table 1) and dictate barrel stability against both chemical (Fig. 6) and thermal denaturants (not shown). While the probe position may influence our conclusions, parameters derived for the WT versus Trp mutants are not additive, suggesting that the Trp residue contribution to Ail behaviour is substantial and masks positional contributions of the fluorophore to the data.

Owing to its amphipathic nature, the indole is largely observed at the solvent-lipid interface of membrane proteins and serves as a stabilizing anchor for the refolded TM domain onto the surrounding lipid membrane and possesses favourable partition energy between the polar solvent exterior and hydrophobic lipid interior33. A recent study that estimated the influence of lipid solvation of Trp residues of E. coli OmpA found greater β-barrel stabilization when the indole is buried, i.e., is highly lipid solvated (buried indole)36. A complementary report investigating multi-pass TM helices provides contrasting evidence, wherein preferential interfacial Trp localization and a noticeable affinity of Phe for the lipid core were obtained52. Another elaborate study on the Trp residues of E. coli OmpA (outer membrane protein A) has identified a contribution of -2.0 kcal/mol and -1.0 kcal/mol for isolated interfacial Trp and Phe, respectively32,33. Our study re-affirms the concurrent existence of both indole preferences, which stems from the differential functions of both Ail indoles. W42 possesses characteristics of a rigid indole that is favourably localized at the solvent-micelle interface, while W149 is a dynamic micelle-embedded tryptophan. This disparity in indole contribution influences barrel stability in various DPRs, serving as the major deciding element for hysteresis in Ail.

It has not escaped our notice that the W42F mutant, in addition to retaining W149, also has a F42, which may together contribute to the high (kinetic) stability (Fig. 6)39. However, a Trp-less Ail displays only marginally superior barrel stability over the W149F mutant (data not shown), indicating that it is indeed W149 that is necessary for the observed characteristics of the Ail barrel. It is therefore intriguing why a single Trp mutant was not evolutionarily selected over the WT Ail with two tryptophans. Analysis of Trp distribution across known TM β-barrels (72 non-redundant structures) reveals only two proteins having fewer than two Trp residues (Table 2). Mapping the location of Trp residues in those proteins (with upto 4 Trp residues are listed in Table 2), clearly indicates that the indole is preferentially located at the interface. Further, mutational studies on PagP have shown that extra-membrane interface indoles are also important for barrel stability60. Our data also points to the necessity of W42 to assist protein folding, while W149 warrants subsequent β-barrel stability. We conjecture that Ail and possibly other homologous barrels, have evolutionarily attained a balance between the folding and unfolding events by favouring Trp residues (in place of Phe) at the solvent-lipid interface. It would be interesting to address whether the indole ring has additional roles or is merely an evolutionary accident.

Our current study provides a unique insight on the selective contributions of W42 to Ail folding and W149 to kinetic stability of the refolded Ail β-barrel (and the associated hysteresis). This is the first report, to our knowledge, of differential Trp contributions to the refolding pathway and unfolding event of a β-barrel membrane protein, as well as increased barrel stability upon mutation of a tryptophan (in Ail-W42F). Our study contradicts the several reports that attribute TM α-helical or β-barrel stability to the presence of interface Trp residues in the sequence31,32,33,34,35,36,37,39,57,61. At this juncture, one may expect that kinetic stability modulated by the surrounding lipid amounts provides the necessary malleability to steer efficient Ail functioning in Yop delivery during Yersinia infection. Indeed Ail behaviour is vastly different from the thermodynamic properties of its close homologue OmpX. Detailed analysis of such biophysical characteristics of Ail will also expedite the development of structure-based drugs and vaccines against plague.

Methods

Protein preparation

Gene corresponding to the mature Ail protein (without the signal sequence) was cloned in pET-3b vector, expressed as inclusion bodies in E. coli C41 cells24 and purified on a cation exchange column, using minor modifications of reported protocols24,42. Single Trp mutants (W42F and W149F) were generated using site directed mutagenesis and purified using similar methods. The mutation was characterized by DNA sequencing of the genes and mass spectrometric fingerprinting of the expressed proteins (not shown). Refolding was achieved by rapid 10-fold dilution of 25 μg/μL Ail from 8.0 M GdnHCl, into the refolding reaction, consisting of 100 mM, 250 mM or 500 mM LDAO prepared in 20 mM Tris-HCl pH 8.5, at 25°C. Protein quantitation was achieved by A280, using an extinction coefficient of 28880 M−1 cm−1 and 23380 M−1 cm−1 for WT and Trp mutants, respectively.

Refolded samples were dialyzed to remove excessive GdnHCl, followed by extensive centrifugation, to obtain the 5X refolding stocks. These stocks contained 2.5 μg/μL Ail (~140 μM) and 100 mM, 250 mM or 500 mM LDAO in 20 mM Tris-HCl pH 8.5 and trace amounts of GdnHCl. The final DPRs achieved in chemical and thermal denaturation experiments were ~700:1 (for 20 mM LDAO), ~1750:1 (for 50 mM LDAO) and ~3500:1 (for 100 mM LDAO), in 20 mM Tris-HCl pH 8.5, unless stated otherwise. At these LDAO concentrations, micellar forms of the detergent were persistent even in high denaturant concentrations(Supplementary Figs. 10–12) and thereby did not give rise to artefacts in our unfolding experiments due to the dissolution of LDAO micelles by high GdnHCl employed in our equilibrium (un/re)folding experiments.

Gel quantification and proteolysis

Quantification of the refolding reaction was achieved by densitometry, as described earlier42. Proteolysis was carried out as reported earlier42, with minor modifications. Briefly, proteinase K was added to the refolding reaction to a final concentration of 2.0 µg/µl and incubated at 25°C for ~15 min, prior to loading on the gel. At these proteinase K concentrations, we could not completely arrest the protease activity using frequently used inhibitors. Reactions were therefore diluted to a final LDAO concentration of ≤20 mM using gel loading dye (as high LDAO interferes with band migration) and analyzed on 15% SDS-PAGE.

Fluorescence measurements

Trp fluorescence emission spectra, acrylamide quenching (without inner filter correction) and calculation of the Stern-Volmer constant (KSV), as well as measurement of average lifetimes (<τ>) were carried out in various DPRs for the refolded proteins and denatured controls, using reported protocols18,62. Bimolecular quenching constant (kq) was derived using the formula kq = KSV/<τ>45.

Equilibrium unfolding and refolding

(Un/re)folding experiments were carried out as described earlier44. Briefly, (un)folded 5X protein stocks in LDAO with or without 8.0 M GdnHCl, respectively, were diluted into a GdnHCl gradient, while the protein and LDAO concentrations were maintained at 0.5 μg/μl (~28 μM) and 20, 50 or 100 mM, respectively. Progress of all reactions was monitored for 48 h, using Trp fluorescence between 320–380 nm, using a λex-max = 295 nm. Equilibrium was achieved in ~24 h (Supplementary Fig. 14) and no further change in the fluorescence intensity was observed upon prolonged incubation. Based on comparison of the fluorescence intensities observed for the unfolded protein prepared under identical sample/buffer conditions, it was estimated that ~95% of refolded Ail could be successfully unfolded by ~24 h in 6.4 M GdnHCl (see Supplementary Figs. 5–7). Since no further significant change in the sample fluorescence was observed, the apparent free energy values were derived for the 48 h data (described below). Anisotropy (r) values were recorded at the end-points (48 h).

For fluorescence measurements, plots of (un)folded fractions (fU or fF) versus GdnHCl were generated using the observed intensity at 330 nm, as described15,63 and analyzed by linear extrapolation19,36,44,63, to derive the mid-point(s) of chemical (de/re)naturation (Cm), and (apparent folding and unfolding free energy at 48 h) and mapp (apparent (un)folding cooperativity at 48 h). In refolding experiments that showed an intermediate species, transition from the intermediate to the folded form was used to derive all the thermodynamic parameters by linear extrapolation. Anisotropy data were fit to two- or three-state equations17,63 and only Cm values were considered.