Reversible Unfolding of Rhomboid Intramembrane Proteases

Abstract

Denaturant-induced unfolding of helical membrane proteins provides insights into their mechanism of folding and domain organization, which take place in the chemically heterogeneous, anisotropic environment of a lipid membrane. Rhomboid proteases are intramembrane proteases that play key roles in various diseases. Crystal structures have revealed a compact helical bundle with a buried active site, which requires conformational changes for the cleavage of transmembrane substrates. A dimeric form of the rhomboid protease has been shown to be important for activity. In this study, we examine the mechanism of refolding for two distinct rhomboids to gain insight into their secondary structure-activity relationships. Although helicity is largely abolished in the unfolded states of both proteins, unfolding is completely reversible for HiGlpG but only partially reversible for PsAarA. Refolding of both proteins results in reassociation of the dimer, with a 90% regain of catalytic activity for HiGlpG but only a 70% regain for PsAarA. For both proteins, a broad, gradual transition from the native, folded state to the denatured, partly unfolded state was revealed with the aid of circular dichroism spectroscopy as a function of denaturant concentration, thus arguing against a classical two-state model as found for many globular soluble proteins. Thermal denaturation has irreversible destabilizing effects on both proteins, yet reveals important functional details regarding substrate accessibility to the buried active site. This concerted biophysical and functional analysis demonstrates that HiGlpG, with a simple six-transmembrane-segment organization, is more robust than PsAarA, which has seven predicted transmembrane segments, thus rendering HiGlpG amenable to in vitro studies of membrane-protein folding.

Introduction

In vitro analysis of the macromolecular organization and folding of membrane proteins is essential to gain insight into the challenges a cell overcomes for proper protein folding. Pioneering investigations on the chemical denaturation and unfolding of α-helical-bundle membrane proteins have focused on bacteriorhodopsin (bR) (1, 2), the copper ion transporting ATPase (CopA) (3), the G-protein-coupled receptor rhodopsin (4), disulphide binding protein (5), small multidrug transporter (6), ATP-binding Cassette Transporter (7) and the galactose transporter major facilitator (8). These studies have relied on a variety of either chemical denaturants, such as urea and guanidinium chloride (GdmCl), or harsh detergents such as sodium dodecyl sulfate (SDS). Complete regain of native protein structure and function after exposure to a chemical denaturant was possible only if denaturation involved only partial unfolding (4, 9). The exceptions are bR, which can be fully unfolded in harsh solvents and successfully refolded to its native form (10, 11), and CopA, whose unfolding in the presence of high GdmCl concentrations is reversible (3). Recently, such a negative inverse correlation between the extent of unfolding and successful refolding has also been observed for four different rhomboid intramembrane proteases (12). For all of these proteins, SDS titration suggests a two-state denaturation process, which is reversible, as judged from recapitulation by recovery of enzymatic activity. However, SDS suppresses activity although residual helicity is still present for these rhomboids (12, 13), thus allowing recovery of native structure and activity upon removal of the denaturant. Thermal denaturation, by contrast, results in irreversible loss of secondary structure and activity (12).

To shed light on the relationship between the degree of structural perturbation and the reversibility of this process in terms of both structure and function, we embarked on an in-depth study to analyze the extent of unfolding and refolding of distinct rhomboid proteases. Rhomboids are an evolutionally conserved family of intramembrane serine proteases with roles in human diseases such as cancer (14, 15) and Parkinson’s disease (16). They hydrolyze peptide bonds in membrane-protein substrates to release soluble signaling domains from the membrane environment (17, 18, 19). Distinct topological forms, consisting of either six or seven transmembrane (TM) segments, have been predicted for this family (20). Crystal structures of Escherichia coli GlpG (EcGlpG) and Haemophilus influenza GlpG (HiGlpG) rhomboid proteases have revealed a compact 6TM α-helical bundle (21). Catalytic residues are buried 10–12 Å from the extracellular face and form contacts between two distinct helices that are held in close association by a GxxxG dimerization motif (19, 21, 22). Rhomboids are dynamic enzymes, known to have regions that accommodate substrate access to this buried active site (23, 24, 25). Furthermore, the oligomeric state of rhomboids is known to influence their protease activity (26).

Here, we demonstrate that two prokaryotic rhomboids, HiGlpG and PsAarA from Providencia stuartii, can be chemically unfolded and successfully refolded. Both rhomboids were suspended in the nonionic detergent n-dodecyl-β-D-maltoside (DDM) and substantially unfolded by the chaotrope GdmCl, and conditions were identified that allow maximal refolding. Regain of helical secondary structure was validated using circular dichroism (CD) and intrinsic fluorescence spectroscopy. Reformation of quaternary structure, that is, the dimeric form of the protease, was demonstrated by size-exclusion chromatography (SEC). Native-like functional properties of refolded HiGlpG and PsAarA were determined using the model substrate fluorescence-labeled (FL) casein and the TM substrate PsTatA. The distinct substrates allowed resolution of different functional and refolding characteristics between HiGlpG 6TM and PsAarA 6+1TM topologies. Further thermal denaturation studies were performed to explore temperature-induced structural and functional changes. Our studies demonstrate that HiGlpG is an α-helical membrane protein that is able to refold and regain native activity after suffering a major loss in secondary structure through exposure to a chemical denaturant.

Materials and Methods

Protein production and purification

Protein purification for both HiGlpG and PsAarA was carried out in buffer containing 1.95 mM DDM, as previously described (26). In brief, expression clones of PsAarA and HiGlpG were constructed using cDNA encoding the respective genes inserted into a pBAD-Myc/HisA expression vector (Invitrogen, Burlington, Canada). Top10 cells were used for protein production. We used 1 μg of plasmid DNA to transform 50 μL of competent cells. Single colonies were grown in Luria broth medium supplemented with 100 μg/μL of ampicillin (Amresco, Solon, OH) at 37°C. Protein production was induced by addition of 0.002% (w/v) of arabinose followed by growth for a period of 8 h at 24°C. Cells were then harvested by centrifugation at 6000 × g for 20 min. The pellet was resuspended in a buffer containing 50 mM Tris, pH 8.0, 300 mM NaCl, 10 μg/mL DNase, and EDTA-free protein inhibitor tablets (Roche, Mississauga, Canada). Cells were lysed using Emulsiflex (Avestin, Ottawa, Canada) at a pressure of 206.8 MPa, and the lysate was clarified by centrifugation at 17,000 × g for 30 min. The membrane fraction was isolated by ultracentrifugation at 100,000 × g for 2 h, and membrane pellets were solubilized in buffer containing 50 mM Tris, pH 8.0, 300 mM NaCl, 10 mM imidazole, 20% (v/v) glycerol, 1.95 mM DDM, and EDTA-free protease inhibitor tablets. Solubilized fractions were stirred for 30 min at 4°C, followed by centrifugation at 12,000 × g for 30 min to remove insoluble particles. The supernatant was then incubated with Ni-NTA resins (Qiagen, Toronto, Canada) for 2 h at 4°C. The protein was eluted with a step gradient (0.1–1 M of imidazole in 50 mM Tris, pH 8.0, 300 mM NaCl, 20% (v/v) glycerol, and 1.95 mM DDM). Before biophysical studies, proteins were concentrated 20-fold and further purified by SEC on a Superdex 200 (16/60) column in 25 mM Tris, pH 8.0, 100 mM NaCl, 10% (v/v) glycerol, and 0.97 mM DDM. Fractions corresponding to nonaggregated protein (i.e., not present in the void volume of the column) were pooled and concentrated using Amicon concentrators according to the requirements needed for subsequent biophysical experiments. Oligomeric-state analysis was conducted on a Superdex 200 (13/30) column in the above buffer. Molar masses of the peaks obtained were estimated using a logarithmic interpolation of elution volumes (V_e) using a gel filtration LMW calibration kit (GE Healthcare, Pittsburgh, PA) comprising 1) blue dextran 2000 (V₀), 2) thyroglobulin (670 kDa), 3) γ-globulin (158 kDa), 4) ovalbumin (44 kDa), 5) myoglobulin (17 kDa), and 6) vitamin B₁₂ (1.3 kDa).

Proof-of-principle unfolding and refolding studies

Both HiGlpG and PsAarA were unfolded by addition of 6 M GdmCl in powder form and incubation at 15°C for 20 min. Refolding was initiated by 10-fold dilution of chemically unfolded protein with refolding buffer containing 50 mM Tris, pH 8.0, 200 mM NaCl, 20% (w/v) glycerol, and 0.97 mM DDM, but no GdmCl. The concentration of GdmCl was further reduced by dialysis against a buffer containing 50 mM Tris, pH 8.0, 200 mM NaCl, and 20% (w/v) glycerol.

CD spectroscopy and thermal melting

Far-UV CD spectra were recorded from 185 to 260 nm at 15°C using a Jasco-810 (Easton, MD) spectropolarimeter. Spectra were recorded for 10 μM of protein in 10 mM sodium phosphate buffer, pH 8.0, in a quartz SUPRASIL cuvette (Hellma Analytics, Markham, Canada) with a path length of 1 mm. Measurements were made with a wavelength increment of 0.5 nm, an integration time of 4 s/step, and a spectral bandwidth of 2 nm. The signal corresponding to buffer alone was subtracted from that of the protein. In the case of refolded proteins, the absorbance (high-tension values exceeded 600 V at ≤187 nm; hence, spectra were not analyzed below this wavelength. All experiments were performed in triplicate. Secondary-structure contents were estimated from the ellipticity [Θ] values between 190 and 240 nm using the DichroWeb server (27, 28) and the CDSSTR algorithm. For thermal denaturation studies, the change in molar ellipticity was monitored at 222 nm during a temperature ramp from 15 to 85°C at a rate of 1°C/min before the thermally denatured sample was cooled back to 15°C at the same rate.

Fluorescence spectroscopy

Fluorescence measurements were performed with a SynergyMx (Bio Tek, Winooski, VT) fluorimeter using 100 μL of protein at 0.15 mg/mL in 384-well plates. Fluorescence emission upon excitation at 290 nm was collected between 300 and 400 nm, typically using 1–5 nm excitation and emission bandwidths and a step size of 5 nm.

Automated CD spectroscopy

Automated CD (ACD) experiments were performed as described previously (29, 30). In brief, a stock solution containing 2–3 mg/mL protein was diluted into buffer A (25 mM Tris, pH 8.0, 25 mM NaCl, 5% (w/v) glycerol, and 7.5 mM DDM) and buffer B (7 M GdmCl, 25 mM Tris, pH 8.0, 25 mM NaCl, 5% (w/v) glycerol, and 7.5 mM DDM) to a final protein concentration of 0.25 mg/mL each. GdmCl concentration was checked by refractometry. Protein solutions A and B were mixed to yield 48 different GdmCl concentrations equally distributed between 0 M and 6 M or higher. Data were collected on an automated Chirascan-plus spectropolarimeter (Applied Photophysics, Leatherhead, UK). Data acquisition was performed from 260 nm to 210 nm with a wavelength increment of 1 nm, a spectral bandwidth of 1 nm, and a digital integration time of 1 s. Each sample was scanned 10 times to obtain the average and the associated standard deviation (29, 30). Experiments were performed in triplicate. Baseline-corrected data were fitted globally according to the two-state linear extrapolation model, as described elsewhere (29).

Effect of temperature on enzyme stability and thermodynamic analysis

Thermal inactivation studies of rhomboids were performed in the temperature range 25–75°C at 10° intervals. We thermostatted 500-μL samples of 5 μM enzyme in buffer containing 50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 20% (v/v) glycerol, and 1.95 mM DDM in a water bath, and 20-μL aliquots were taken out at specified times, placed on ice, and stored until activity measurements were performed as previously described (26) for 2 h at 37°C in buffer containing 50 mM MES, pH 6.0, 150 mM NaCl, 20% (v/v) glycerol, and 1.95 mM DDM. The concentration of PsAarA was fixed at 0.25 μM, and the concentration of PsTatA-FRET was 3 μM. FRET-based AarA cleavage assay was conducted by measuring the emission intensity at 530 nm upon excitation at 414 nm in a fluorescence multiwell plate reader (SynergyMx, BioTek). The activity assay with FL-casein as substrate was performed at the same enzyme and substrate concentrations. Data obtained from thermal inactivation studies were used to calculate kinetic parameters associated with PsAarA and HiGlpG activity.

Enzyme inactivation over time was described by a first-order equation of the form

$$\text{ln}A/A_0=-kt,$$ (1)

where A represents enzyme activity at time t, A₀ is the initial activity at time zero, k is the rate constant at the tested temperature (min⁻¹), and t is time (min). Inactivation rate constants were obtained from the slopes of the semilogarithmic plots of residual activity versus incubation time at each temperature. Calculated rate constants were replotted in Arrhenius plots as natural logarithms of k versus the reciprocal of absolute temperature. Arrhenius law describes the temperature dependence of rate constants as

$$\text{ln}\left(k\right)=-E_{\text{a}}/RT+c,$$ (2)

where E_a is the activation energy, R is the universal gas constant (8.31 J mol⁻¹ K⁻¹), and T is the absolute temperature. E_a values were calculated from the slopes of the Arrhenius plots.

The half-life of rhomboids (t_1/2 in min), defined as the time after which activity was reduced to one-half the initial value (31), was determined as

$$t_{1/2}=\text{ln}\left(\mathit{2}\right)/k.$$ (3)

In some cases, the inactivation rate is given as the D value, which is the time required to reduce the enzyme activity to 10% of its original value. The D value is inversely related to the inactivation rate constant, k, according to

$$D=\text{ln}\left(\mathit{10}\right)/k.$$ (4)

The activation free energy (ΔG° in kJ mol⁻¹) and the associated enthalpy (ΔH° in kJ mol⁻¹) and entropy (ΔS° in J mol⁻¹ K⁻¹) values for the thermal denaturation of rhomboid enzymes were determined as

$$\Delta G°=-RT\text{ln}\left(kh/k_{\text{B}}T\right)$$ (5)

$$\Delta H°=E_{\text{a}}-RT$$ (6)

$$\Delta S°=\left(\Delta H°-\Delta G°\right)/T,$$ (7)

where h is the Planck constant (6.626 × 10⁻³⁴ Js) and k_B is the Boltzmann constant (1.38 × 10⁻²³ J K⁻¹). Experiments were performed in duplicate.

Results

We sought to understand the capacity of two distinct forms of rhomboid proteases, HiGlpG and PsAarA, to undergo reversible structural transition in the presence of the chemical denaturant GdmCl. These proteins were purified as homogenous dimers, as demonstrated by SEC and SDS-polyacrylamide gel electrophoresis (Fig. S1, A and B, available online). As had been reported, we observed autoproteolysis with PsAarA (32), which was abolished by mutating the catalytic serine (S150) residue to alanine. This inactive mutant did not affect PsAarA secondary structure (Fig. S2) or its ability to bind substrates (unpublished data). The wild-type protein was used for functional studies.

Effect of denaturant on secondary structure and activity

CD spectra of both rhomboid proteases, HiGlpG and PsAarA, suspended in the nonionic detergent DDM were characteristic of proteins with a large percentage of α-helical secondary structure, showing a large positive band at ∼192 nm and two smaller negative bands centered at ∼208 and 222 nm (Fig. 1) (33). With HiGlpG, the secondary structure content was estimated to be 86% α-helical, which was in excellent agreement with the crystal structure (PDB: 2NR9). In both wild-type PsAarA and its inactive mutant, the helical content of the native protein was ∼80%. In both cases, addition of 6 M GdmCl resulted in the formation of an unfolded state obtained by progressive destabilization of the folded state. For HiGlpG, this phenomenon was reflected by a substantial decrease in intensity of the negative peaks at 208 nm and 222 nm (Fig. 1 A). Upon a 10-fold dilution of the unfolded protein in protein-folding buffer, refolding went to completion, as demonstrated by the complete recovery of secondary structure. With the PsAarA inactive mutant, an unfolded conformation was observed on addition of 6 M GdmCl, similar to the case for HiGlpG (Fig. 1 C). Upon refolding, a decrease in the positive ellipticity at 192 nm was accompanied by a slight decrease in the negative ellipticities at 208 and 222 nm with respect to the native form. Thus, ∼92% of the helical conformation was regained upon refolding in PsAarA. The presence of three and four tryptophan residues in HiGlpG and PsAarA, respectively, allowed us to further validate refolding using intrinsic fluorescence spectroscopy. Refolded proteins displayed native-like fluorescence spectra (Fig. S3).

Figure 1.

Secondary-structure analysis and recovery of rhomboid protease activity upon refolding. (A) Far-UV CD spectra of native, unfolded, and refolded HiGlpG. (B) Recovery of HiGlpG activity upon refolding. The proteolytic activity of native HiGlpG (blue) is compared to the activity of the refolded form (green) using a casein substrate. (C) Far-UV CD spectra of native, unfolded, and refolded PsAarA inactive mutant. (D) Recovery of PsAarA activity upon refolding. The proteolytic activity of native PsAarA (light green) is compared to the refolded form using casein (gray) and its physiological substrate, PsTatA (dark green). To see this figure in color, go online.

In studying rhomboid stability, analysis of activity is crucial for validating that refolding returns the protein to a functional and native-like tertiary state. Activities of HiGlpG and wild-type PsAarA, were assessed using casein as a model substrate, as described previously (26). PsAarA activity was further validated using its physiological TM substrate PsTatA. Our fluorescence-based activity assay revealed a 90% recovery of activity for HiGlpG, 75% for PsAarA with casein, and 65% for PsAarA with PsTatA (Fig. 1, B and D). Since rhomboid proteases are known to form an α-helical bundle with catalytic residues on two different helices (21, 34), our experiments indicated that refolding was successful.

The lower efficiency in refolding and regain of activity with the PsAarA, having an additional predicted TM helix compared to HiGlpG, suggests possible folding differences between the two topological forms. This led us to question whether AarA with removal of the seventh predicted TM region to generate a 6TM topological form would be able to unfold and refold properly. A construct of AarA lacking the C-terminal predicted TM helix (residues 237–281), however, did not generate any folded protein, suggesting that this region plays a structural role during folding.

Dimerization upon refolding

The oligomeric states of the refolded proteins were validated using SEC. Refolded HiGlpG displayed an SEC elution behavior similar to that of native HiGlpG, where a major peak was observed at an elution volume of ∼12.8 mL, corresponding to the Stokes radius observed for the dimeric form (Fig. 2 A). SDS-polyacrylamide gel electrophoresis analysis of the peak fractions confirmed the purity of refolded HiGlpG, suggesting that GdmCl treatment did not cause proteolysis (data not shown). With the PsAarA inactive mutant, elution volume peaks were observed at ∼12.9 mL (native) and ∼13.5 mL (refolded) (Fig. 2 B). Both peaks are consistent with the hydrodynamic radius of the dimeric form of PsAarA. The slight shift of the hydrodynamic radius of the refolded protein agrees with the modest change in secondary structure observed by CD (Fig. 1 C).

Figure 2.

Oligomeric state upon refolding as obtained from SEC. (A) Overlay of chromatograms for native and refolded (RF) HiGlpG. (B) Overlay of chromatograms for native and refolded (RF) PsAarA (inactive mutant). To see this figure in color, go online.

Conformational changes at intermediate denaturant concentrations

ACD experiments were performed to follow the details underlying conformational changes of HiGlpG and PsAarA induced by GdmCl (Figs. 3 A and 4 A). For both proteins, incremental addition of GdmCl led to a gradual reduction in helicity (∼25% and ∼20%, respectively) and a concomitant increase in unstructured conformation (∼50% and ∼40%, respectively). Across the entire wavelength range studied, isotherms obtained for HiGlpG revealed a weakly sigmoidal transition at >4 M that was not clearly separated from a steeply sloping pretransition baseline (Figs. 3 B and S4 A). The transition midpoint was at 4.8 M GdmCl. By contrast, PsAarA isotherms also revealed a broad sigmoidal transition, but with a much shorter pretransition range and a transition midpoint at a GdmCl concentration of 2.6 M (Figs. 4 B and S4 B), which was substantially lower than that for HiGlpG. For both rhomboids, however, such isotherm shapes impeded reliable determination of thermodynamic parameters quantifying conformational stability.

Figure 3.

Automated CD studies on HiGlpG. (A) Influence of GdmCl on secondary structure of HiGlpG. CD spectra of 0.25 mg/mL HiGlpG were recorded in the presence of 0–7 M GdmCl. (B) GdmCl denaturation isotherms of HiGlpG. The mean residual ellipticity at 222 nm is plotted against GdmCl concentration for three independent series of experiments. The solid line represents a global fit to all three 222-nm isotherms. Dashed lines indicate best-fit pre- and posttransition baselines and the transition midpoint at 4.8 M GdmCl. To see this figure in color, go online.

Figure 4.

Automated CD studies on the inactive mutant of PsAarA. (A) Influence of GdmCl on secondary structure. CD spectra of 0.25 mg/mL of the protein were recorded in the presence of 0–7 M GdmCl. (B) GdmCl denaturation isotherms. The mean residual ellipticity at 222 nm is plotted against GdmCl concentration for three independent series of experiments. The solid line represents a global fit to all three 222-nm isotherms. Dashed lines indicate best-fit pre- and posttransition baselines and the transition midpoint at 2.6 M GdmCl. To see this figure in color, go online.

Temperature-induced conformational changes

With an aim to further explore differences between the native and refolded forms of HiGlpG and PsAarA, thermal stability studies were undertaken. Native folded and refolded versions of these two proteins were exposed to a range of temperatures (15–85°C). CD spectra of HiGlpG remained essentially unchanged up to 55°C, beyond which a slight decrease in negative ellipticity at 208 and 222 nm and a comparatively marked decrease in positive ellipticity at 192 nm were observed (Fig. 5 A). Similar spectra were observed with HiGlpG that had been refolded after GdmCl-induced unfolding (Fig. 5 B). Thermal denaturation profiles of native HiGlpG revealed a sigmoidal shape with a transition midpoint temperature, T_m, of ∼65°C. By contrast, the GdmCl-refolded form of HiGlpG showed a rather featureless, nonsigmoidal change in the CD signal (Fig. S5 A). Upon cooling from 85°C to 15°C, no change was observed in either native or GdmCl-refolded forms of HiGlpG (Fig. S6 A), demonstrating irreversible thermal denaturation.

Figure 5.

Effect of temperature on secondary structure of HiGlpG. (A) Far-UV CD spectra of HiGlpG at different temperatures. (B) Far-UV CD spectra of refolded HiGlpG at different temperatures. To see this figure in color, go online.

Similar to HiGlpG, in the natively folded and refolded versions of the inactive mutant of PsAarA, there was a slight change in the CD spectrum at 55°C followed by further loss of secondary structure upon an increase in temperature. Again, CD scans at different temperatures for the native and refolded forms of the PsAarA inactive mutant were different (Fig. 6, A and B). As with HiGlpG, the thermal denaturation profiles of PsAarA were sigmoidal and nonsigmoidal in natively folded and refolded forms, respectively (Fig. S5 B). As observed in the case of HiGlpG, both forms of PsAarA demonstrated irreversible denaturation with temperature (Fig. S6 B).

Figure 6.

Effect of temperature on secondary structure of PsAarA inactive mutant. (A) Far-UV CD spectra of the native form of the protein at different temperatures. (B) Far-UV CD spectra of the refolded form of the protein. To see this figure in color, go online.

Enzyme kinetics of native and refolded states

Thermal inactivation of both native and refolded forms was investigated in a time- and temperature-dependent manner. Thermal inactivation curves, demonstrating the change in residual activity at different temperatures over time, are represented in Fig. S7, A and B. The inactivation kinetics of the native form of HiGlpG demonstrated an interesting effect: at 45°C and 55°C, the protein activity remained unaltered for the initial 60 min, with further increase to 120–150% within the next 60 min. (Fig. 7 A). This phenomenon was not noted at the other tested temperatures. At 25°C and 35°C, the enzyme was stable for 2 h, whereas at 65°C, a gradual loss of activity was observed after 20 min. Refolded HiGlpG showed the same phenomenon of enhanced activity at 45°C, but at 55°C, the protein lost activity after 90 min (Fig. 7 B). PsAarA was stable at 25°C and 35°C; however, loss of activity was observed at 45°C. With PsTatA as substrate, PsAarA completely lost activity within 2 h at 45°C, whereas with casein the residual activity was still observed. In contrast to refolded HiGlpG, refolded PsAarA started to lose activity with both substrates at 35°C and became completely inactive at 55°C. Thus, at any temperature, native PsAarA showed a higher percentage of residual activity than the refolded form.

Figure 7.

Time course of residual activity (%) of HiGlpG at different temperatures. (A) Native HiGlpG with casein as substrate. (B) Refolded (RF) HiGlpG with casein as substrate. Insets show the respective Arrhenius plots. To see this figure in color, go online.

The semilogarithmic plots of residual activity versus incubation time for both proteins were linear at all temperatures, indicative of a simple first-order monophasic process. The exception was HiGlpG inactivation at 45°C and 55°C, where the straight line could not be fitted, for obvious reasons. All lines extrapolate back to the origin, suggesting that inactivation of the sole isoenzyme is being measured. From the slopes of semilogarithmic plots, the inactivation rate constants were calculated (Table 1).

Table 1.

Kinetic Parameters Characterizing the Thermal Inactivation of PsAarA and HiGlpG in Their Native and Refolded Forms

Enzyme (Substrate)	T (°C)	kd (min−1)	t1/2 (min)	D (min)	ΔH° (kJ mol−1)	ΔG° (kJ mol−1)	ΔS° (J mol−1 K−1)
PsAarA native (casein)	25	0.0011	620.7	2062	202.85	88.04	0.408
35	0.0014	467.2	1552.3	202.44	85.09	0.37
45	0.01304	53.1	176.5	202.35	80.16	0.36
55	0.198	3.5	11.6	202.27	80.35	0.37
PsAarA native (PstatA)	25	0.0013	533.1	1771.2	194.52	85.5	0.365
35	0.0032	216.6	719.5	194.44	86.04	0.351
45	0.032	21.6	71.9	194.35	82.73	0.351
55	0.36	1.9	6.3	194.27	78.77	0.352
PsAarA refolded (casein)	25	0.0006545	1059	3518	216.52	87.19	0.433
35	0.0084	82.5	274.1	216.44	83.56	0.431
45	0.173	4	13.3	216.35	78.4	0.433
PsAarA refolded (PstatA)	25	0.001	693.1	2302.5	156.65	86.1	0.236
35	0.005	138.6	460.5	156.57	84.89	0.232
45	0.061	11.3	37.7	156.48	81.02	0.237
HiGlpG native (casein)	35	0.00021	3300.7	10,964.6	150.92	93.08	0.194
65	0.024	28.8	95.9	150.59	88.86	0.182
75	0.277	2.5	8.3	150.5	84.4	0.189
HiGlpG refolded (casein)	35	0.00046	1506.8	5005.6	200.44	91.01	0.355
55	0.037	18.7	62.2	200.27	84.95	0.351
65	0.59	1.1	3.9	200.19	79.76	0.356

T, temperature in °C; k_d, inactivation rate constant; t_1/2, half-life of rhomboids (i.e., the time after which activity is reduced to one-half of the initial value), D, the time required to reduce the enzymatic activity to 10% of its original value; ΔG°, activation free-energy barrier; ΔH°, activation enthalpy; ΔS°, activation entropy of thermal denaturation.

The dependence of the inactivation rate constants on temperature was fitted by the Arrhenius equation, and Arrhenius plots were used to calculate apparent activation energies (E_a) of thermal inactivation. For native and refolded HiGlpG, Arrhenius plots were linear, yielding E_a values of 153 kJ mol⁻¹ and 203 kJ mol⁻¹, respectively. By contrast, Arrhenius plots of native PsAarA with both substrates showed upward curvature, suggesting two distinct denaturation processes with different activation energies (Fig. 8, A and B, insets): high at the temperatures above 35°C, demonstrating a strong temperature-dependent rate; and low at temperatures below 35°C, where the inactivation rate reflects a distinct process (Fig. S7, C–F). The activation energies for native PsAarA were calculated to be 205 kJ mol⁻¹ when measured with casein and 197 kJ mol⁻¹ when measured with PsTatA. The Arrhenius plot of the refolded form of PsAarA demonstrated simple linear fits for both substrates. The apparent activation energies were 219 kJ mol⁻¹ for casein and 159 kJ mol⁻¹ for PsTatA. Inactivation rates (k) of both native and refolded PsAarA progressively increased with temperature (∼10-fold per 10°C) (Table 1), demonstrating a high degree of irreversible denaturation. As was the case for HiGlpG, the refolded PsAarA protein was less stable compared to its native counterpart. The reason for this decrease in thermal stability remains to be explored.

Figure 8.

Time course of residual activity (%) of PsAarA at different temperatures. (A) Native PsAarA with casein as substrate. (B) Refolded (RF) PsAarA with casein as substrate. (C) Native PsAarA with PsTatA as substrate. (D) Refolded (RF) PsAarA with PsTatA as substrate. Insets show the respective Arrhenius plots. To see this figure in color, go online.

Calculation of the kinetic parameters of inactivation is essential for understanding enzyme thermal stability and revealing any secondary destabilization effects. The comparison of half-life (t_1/2) and the decimal reduction time (D value) for HiGlpG and PsAarA at the same temperature allowed us to conclude that the former was relatively more stable (Table 1). The free-energy barrier for enzyme inactivation, ΔG°, and its enthalpic and entropic components, ΔH° and ΔS°, respectively, were calculated from the experimental data using Eqs. 3–7 (Table 1). With elevated incubation temperatures and concomitant destabilization, a significant reduction in ΔG° was observed for both HiGlpG and PsAarA. Furthermore, positive values of activation entropy (ΔS°) were found for both rhomboids (Table 1).

Discussion

The rhomboid proteases HiGlpG and PsAarA can be refolded in DDM after substantial unfolding in the presence of high concentrations of the chemical denaturant GdmCl. For both enzymes, this process leads to a substantial loss of secondary structure and is largely reversible. However, the rather gradual transition from the folded to the unfolded population with increasing GdmCl concentration prevents thermodynamic analysis for a two-state equilibrium model. This is in contrast both with reversible SDS-mediated denaturation, during which the same rhomboids undergo only marginal changes in secondary-structure content (12, 13), and with thermal denaturation, which is irreversible. Although both rhomboid proteases were active upon refolding, differences were observed between HiGlpG and PsAarA with respect to the extent of refolding and thermal inactivation of enzymatic activity. Notably, HiGlpG adopts a core 6TM architecture (21), whereas PsAarA is predicted to have a seventh TM helix at the C-terminus (20).

Recovery of structure and activity after GdmCl unfolding

Our results provide detailed insights into secondary-structural and functional changes occurring when different topological forms of the intramembrane rhomboid proteases HiGlpG and PsAarA are exposed to the chemical denaturant GdmCl. Unfolding in the form of loss of secondary structure upon GdmCl treatment was observed for both proteins. Dilution of the chemically unfolded protein into buffer containing DDM without denaturant followed by further removal of the denaturant by dialysis enabled recovery of the protein’s secondary and tertiary structure, since we observed a regain of activity to near-native values, in particular for HiGlpG. This implies that not only did the catalytic core reform, but the flexible regions were properly positioned for substrate gating as well as dimerization. This adds rhomboid proteases to the short list of α-helical proteins that have been functionally refolded in virtually quantitative yield after a significant loss of regular structure (4, 9). Although it is widely accepted that all information required for proper folding is contained in the primary structure (i.e., sequence) of the polypeptide chain (35), refolding of α-helical membrane proteins has proven challenging. Thus far, CopA (3) and bR (10, 11) are the only other α-helical membrane proteins successfully refolded after substantial unfolding by GdmCl. In all other cases, a negative correlation has been observed between the extent of unfolding and the recovery of native-like refolded protein. For instance, detergent-solubilized DAGK displays a native-like CD spectrum upon addition of urea, from which it can be refolded to ∼50% (36). When the harsher denaturant GdmCl is used, no indications of regular structure can be found in the unfolded state, and this correlates with a drastic decrease in refolding yield (36). Similar observations have been made for light-harvesting complex II, for which the yield of refolded protein from a partially structured denatured state in harsh detergent (37, 38) are considerably higher than those from a largely unfolded state in GdmCl (39).

Absence of two-state folding

ACD experiments are useful for monitoring unfolding and reversibility at very high resolution (29, 30). For proteins folding by a typical two-state model, such experiments reveal a pretransition region, which reflects the dependence of the CD signal of the native protein on denaturant concentration, a transition region, and a posttransition region, which reflects the dependence of the CD signal of the unfolded protein on denaturant concentration. Apparent two-state behavior has previously been observed for CopA folding in GdmCl (3), but the same authors also pointed out that this transition most likely includes several intermediates, at least one of which is enzymatically inactive but spectroscopically indistinguishable from the native protein (35). A comprehensive study combining equilibrium and kinetic experiments has recently provided compelling evidence for fully reversible two-state folding of 69 EcGlpG mutants during SDS denaturation (13). As noted by the authors, this surprising simplicity in the behavior of a large TM protein can be explained by conformational restrictions in the largely helical SDS-denatured state, so that renaturation involves only rearrangement, but no de novo formation of helical segments.

In stark contrast with this, both rhomboids studied here revealed a more complex unfolding behavior, with a gradual transition from the folded to a predominantly unfolded state that is not accessible by SDS denaturation. Notwithstanding these challenges to more detailed quantitative analysis, our data clearly demonstrate that the transition midpoint for GdmCl-induced unfolding of HiGlpG is substantially higher than that of PsAarA, indicating that the former is more resistant to chemical denaturation. Thermodynamic stabilities of HiGlpG and PsAarA have been previously estimated using SDS-mediated denaturation, and protease activity has been validated (12). In these experiments, two-state behavior, but no major loss in secondary structure, was observed, again suggesting that SDS-mediated denaturation is a very different process compared to GdmCl-induced unfolding. More recently, mechanical unfolding studies with the E. coli rhomboid GlpG (EcGlpG) have revealed cooperative unfolding (40) compared with the gradual transition we observe in the presence of chemical denaturants, further highlighting differences between these distinct methods of membrane protein (un)folding.

Tertiary structure and activity upon refolding

Recapitulation of activity for both HiGlpG and PsAarA after denaturation in GdmCl suggests that rhomboids have a robust structural core. For the minimal-core rhomboid, HiGlpG, we observe virtually complete recovery of activity. Although many membrane proteins have been successfully refolded after incubation in mild denaturants such as SDS (41), it is less common for α-helical membrane proteins to be successfully refolded to an active state after substantial unfolding (42). In some cases, native-like secondary structure can be restored, although reassociation of different subunits seems to be a limiting factor, as observed with the multidrug transporter, AcrB (43). Both HiGlpG and PsAarA exist as dimers in DDM (44); indeed, we observe a dimeric state for both HiGlpG and PsAarA upon refolding.

Refolded rhomboid proteases HiGlpG and PsAarA have different activities in detergent, which may be a consequence of their topology. PsAarA differs from HiGlpG in having an extra TM helix appended at its C-terminus. Modeling of PsAarA also reveals extended loops compared to the HiGlpG structure, possibly important for function. Furthermore, PsAarA has an allosteric site (exosite) for recognition of its physiological substrate, PsTatA (26). Thus, compared to HiGlpG, more structural elements may need to be reformed for PsAarA upon refolding to recapitulate a native-like activity.

Thermal inactivation to probe rhomboid protease stability

We studied the effect of temperature on the biophysical and functional properties of the two rhomboids to account for differences in activity after refolding. On the basis of isothermal experiments in the temperature range 25–65°C and the Arrhenius equation, thermal inactivation of the native and refolded forms of the two rhomboids can be explained as first-order monophasic process. It was observed that activation enthalpy was independent of temperature. High enthalpy values and first-order monophasic kinetics have been previously reported for cytosolic enzymes such as polyphenol oxidase (45). Our observations suggest that rhomboids in their native form are stable enzymes. A significant decrease in ΔG° value with temperature indicated destabilization and a dominant contribution of activation entropy to the thermal inactivation process, with positive ΔS° values suggesting increased thermal motion within the proteins with increasing temperature. Thermodynamic parameters have been determined only for a few α-helical membrane proteins, for example cation-transporting Na,K-ATPase enzymes (46).

Dynamics in structural elements of rhomboid proteases have also been observed using both x-ray crystallography and molecular dynamics simulations (18, 34, 47, 48, 49). The TM domains of rhomboid proteases are evolutionarily constrained because of extensive packing interactions necessary for stabilization (12, 50); however, conformational plasticity has been shown to be essential for substrate entry to the buried active site (51). Loop 5 and TM5 are the mobile elements, which occlude the active-site in the absence of substrate (23, 24, 25). A GxxxG dimerization motif between TM4 and TM6 brings together the catalytic serine and histidine residues located on these respective helices (17, 18). The increase in activity observed with HiGlpG at 45°C and 55°C suggests that the GxxxG dimerization motif is not disrupted, yet mobility in the flexible loop 5 and TM5 is increased. Incubation at higher temperatures likely disrupts these interhelical packing interactions, resulting in decreased activity. At 85°C, although the tertiary structure is lost, ∼50% of residual helicity is observed (Fig. 5), indicating that some TM helices, which are stabilized by intrahelical hydrogen bonds as well as hydrophobic interactions, are still intact. The refolded form of this protein shows similar secondary-structural characteristics during thermal denaturation. However, in contrast to the case for native HiGlpG, a drastic decrease in activity at 55°C is observed. This suggests that packing in the refolded protein is not identical to that in the native form. Reduction of ΔG° with temperature, as seen in thermal inactivation experiments, further supports the above explanation of destabilization of structure for these rhomboids (Table 1).

HiGlpG is more stable than PsAarA. In contrast to HiGlpG, we observe a gradual loss of PsAarA activity at temperatures below 45°C, although secondary structure is preserved. This suggests that the molecular scaffold of PsAarA, consisting of the catalytic dyad and the residues forming the substrate-binding pocket, is more easily disrupted compared with HiGlpG. With time, this disruption affects cleavage of the physiological substrate (PsTatA) to a greater extent than that of the model substrate, casein. The binding of PsAarA to its physiological substrate PsTatA is exosite-mediated and occurs with positive cooperativity (26). This exosite might be destabilized with an increase in temperature. The refolded enzyme is relatively less stable and loses its activity when incubated at 45°C. Similar to the case for HiGlpG, ∼50% of secondary structure is retained in the native and refolded forms. Thus, refolded rhomboid proteases appear less stable than their native counterparts, although no differences in secondary structure were observed for HiGlpG. Moreover, the crystal structure of HiGlpG (PDB: 2NR9) contains three bound lipid molecules per monomer (21), which are taken up during recombinant protein isolation. It is tempting to hypothesize that the decreased thermal stability of the refolded enzyme could be due to the absence of bound lipids after refolding, although folding studies on bR with and without lipids suggest that their presence is not necessary for functional folding (52). Refolding studies in the presence of lipids could further aid in understanding their effects on folding of these rhomboid proteases.

Conclusions

The fact that HiGlpG regains activity upon refolding after being treated with a harsh chemical denaturant such as GdmCl demonstrates that this is a robust form of rhomboid intramembrane protease. Kinetic parameters such as t_1/2 support this conclusion. The activity of PsAarA with its physiological substrate PsTatA is not completely regained upon refolding, indicating that the allosteric site of the enzyme may not be completely structured in the refolded protein. Similar observations for HiGlpG are currently not possible, because its physiological substrate is not known. Our studies suggest that, in contrast with other α-helical membrane proteins that reversibly denature without undergoing major changes in secondary-structure content, the two-state refolding model could not be applied to these rhomboids. However, in our case, refolding and association events result in regain of the native dimeric states. Furthermore, our thermal inactivation studies provide evidence for the importance of conformational changes in rhomboid proteases during catalysis.

Author Contributions

P.P. determined refolding conditions. R.P. and P.P. performed CD experiments. R.P. performed SEC and P.P. intrinsic fluorescence studies. R.P. and E.A. designed thermal denaturation kinetics experiments. E.A. performed and analyzed biochemical assays. K.G. and S.K. performed and analyzed ACD experiments. R.P., M.J.L., K.G., and S.K. wrote the manuscript with input from all authors.

Editor: Francesca Marassi.