(Received for publication, July 31, 1996, and in revised form, January 6, 1997)
From the Universität Regensburg, Institut für Biophysik und Physikalische Biochemie, D-93040 Regensburg, Germany
Firefly luciferase has been used as a model
protein to study cotranslational and chaperone-assisted protein
folding. We found conditions for reversible unfolding of luciferase in
the absence of cellular factors, and we characterized
denaturant-induced equilibrium unfolding transitions and refolding
kinetics of the enzyme. Luciferase unfolding induced by guanidinium
chloride at 10 °C can be described as a four-state equilibrium with
two inactive intermediates highly populated around 1 and 3 M denaturant. The transitions occur around 0.3, 1.7, and
3.8 M denaturant. The free energy of denaturation to the
first inactive intermediate
(G0N
I1 = 15 ± 3 kJ·mol
1) is small for a protein of 60 kDa. Fluorescence
and circular dichroism spectra of the intermediates indicate that
I1 has a compact conformation, whereas aromatic side chains
are highly exposed in the second intermediate, I2, despite
its high content of secondary structure. In the presence of a
hydrophilic detergent, significant reactivation of luciferase is
observed up to temperatures at which the native protein is unstable.
Reactivation kinetics of luciferase are exceedingly slow and probably
not limited by proline isomerization, as suggested by their
independence from the time spent in the unfolded state.
The observation that some organisms are able to convert chemical energy into visible light has been fascinating researchers for many years. In Photinus pyralis, a beetle in the superfamily of Cantharoidae, the 61-kDa monomeric protein firefly luciferase is responsible for the production of light. It is located in specialized peroxisomes in the abdominal lantern organ of the insect and was one of the first enzymes to be investigated in biochemical detail (1). Although subsequent research has led to a more detailed understanding of the light reaction and although the crystal structure of the enzyme has recently been solved, some aspects of its enzymatic mechanism still remain mysterious (2-4).
Luciferase catalyzes the Mg2+-ATP-dependent oxygenation of the heterocyclic component firefly luciferin, yielding ultimately an electronically excited oxyluciferin species. This excited state product then returns to the ground state emitting yellow-green light with an emission quantum yield close to unity (5, 6). The resulting flashes of light are employed by the firefly in its reproductive behavior.
Because the enzymatic luminescence assay is highly sensitive,
luciferase is widely used to rapidly determine small concentrations of
ATP (7). The cloning of the cDNA coding for firefly luciferase has
opened a wide field of applications in molecular biology in which
luc is used as a genetic marker or reporter gene (8). These
applications are facilitated by the fact that the luciferase assay,
with a detection limit of about 1018 mol of the enzyme,
can be performed in crude cell extracts or even in whole cells. For the
same reason, firefly luciferase has become a popular model in studies
of protein folding in vivo or upon in vitro
translation in cell extracts (9-13).
Upon in vitro translation of the luc gene, luciferase activity could be detected almost immediately after release of the polypeptide from the ribosome (11). Moreover, when chemically denatured or heat-inactivated luciferase was diluted into reticulocyte lysate or wheat germ extract, the enzyme reactivated in minutes and with high yields of 60-80% when ATP was provided by a regenerating system (10-14). In contrast, renaturation of luciferase in buffer lacking cellular components was inefficient, and the enzyme was found to aggregate during refolding (9, 10). Reconstitution experiments with purified molecular chaperones identified DnaK, DnaJ, and GrpE as candidates for the chaperoning system facilitating luciferase folding upon expression in Escherichia coli (9, 15).
On the other hand, it is generally held that the information required for correct folding is entirely encoded in the amino acid sequence of a protein (16), and proteins found to depend on molecular chaperones for efficient folding in vivo generally can be efficiently renatured if refolding conditions are carefully optimized (17). Such refolding experiments are a prerequisite for understanding critical steps in protein-folding pathways and to identify folding intermediates as candidate targets of chaperone action. The purpose of the present work was to find conditions for efficient refolding of P. pyralis luciferase in the absence of accessory proteins and to characterize the folding pathway of the enzyme. We found that firefly luciferase does indeed refold spontaneously and with high yield at low temperature and low protein concentration, conditions chosen to minimize aggregation as a competing side reaction. This allowed us to determine the thermodynamic stability of the enzyme from equilibrium unfolding transitions, to physically characterize two highly populated folding intermediates, and to follow the kinetics of luciferase refolding.
Purified, lyophilized P. pyralis
luciferase was obtained from Boehringer Mannheim, dissolved in storage
buffer (0.5 M Tris acetate buffer, pH 7.5), and stored at
70 °C at protein concentrations of 1 or 5 mg/ml. The enzyme was
found to be homogeneous by SDS gel electrophoresis and silver staining.
Peptidylprolyl cis-trans-isomerases from the periplasm of
E. coli (rotamase) and from rabbit cytoplasm (FKBP-52) were
kindly provided by Dr. Elke Prohaska and Dr. Sushira Bose
(Universität Regensburg). For reactivation experiments and equilibrium transitions, ultrapure Tween 20 from Sigma was used; for
gel filtration experiments, it was protein-grade Tween 20 from
Calbiochem. Dithioerythritol was purchased from Roth (Karlsruhe, Germany) and GdmCl1 from ICN/Schwarz-Mann
(Cleveland, OH). Other chemicals were analysis-grade from Merck, and
solutions were made up from quartz double distilled water.
Unless indicated otherwise, luciferase activity was assayed using the GenGlow 100 kit from Bio-Orbit (Turku, Finland). A fluorescence microcuvette (Hellma, Type 105.251-QS) was filled with 25 µl of luciferin reagent and 25 µl of ATP reagent, both preincubated at 25 °C as recommended by the supplier. For samples from equilibrium transitions induced by GdmCl, 2 µl of a corresponding GdmCl buffer solution were mixed into the assay to keep the overall GdmCl concentration constant at 55 mM. Addition of 3 µl of luciferase solution (0.2-18 µg/ml enzyme) started the light reaction. Light emission was followed at 550 nm at a spectral bandwidth of 25-40 nm for 2 min using a Spex FluoroMax or a Perkin-Elmer MPF-2A spectrofluorometer with switched-off light source. Because autoinactivation causes a slight decrease in emission intensity with time (less than 8% over 2 min), the signals were corrected by extrapolation to zero time. The test was linear up to at least 18 µg/ml luciferase in the sample.
SpectroscopyFluorescence emission spectra were recorded in quartz cells (Hellma, Type 119.004F-QS) in a Spex FluoroMax spectrofluorometer equipped with a thermostatted cell holder. The excitation wavelength was set to 280 nm, and the spectral bandwidths were 4 and 8 nm for excitation and emission, respectively. With an increment of 1 nm, the integration time for each data point was 0.5 s. All spectra were corrected for fluorescence of buffers. Fluorescence of samples from folding equilibrium transitions was measured in 1-cm semimicro cells (Hellma, Type 104-QS) at the emission maximum of the native protein (334 nm) or at the maximum of the difference spectrum between native and unfolded luciferase (331 nm).
Circular dichroism was measured in calibrated fused silica semimicro fluorescence cells (path length, 4.2 mm; Hellma, Type 104F-QS) or in standard 1-mm silica cells (Hellma, Type 100-QS) in an Aviv 62A-DS spectropolarimeter at 10 °C at a spectral bandwidth of 1 nm unless indicated otherwise. For CD spectra, data points were taken at 0.2-nm intervals with integration over 1 s. Spectra were accumulated over four repeated scans, corrected for solvent background, and smoothed using polynomial interpolation. To follow conformational transitions, CD was measured at 220 nm with integration over 180 s. Residue ellipticities were calculated using a mean residue weight of 110.5 calculated from the amino acid composition.
Reactivation of GdmCl-denatured LuciferaseLuciferase was
denatured in buffer A (100 mM potassium phosphate, pH 7.8, 1 mM EDTA, 1 mM dithioerythritol) or buffer B
(100 mM potassium phosphate, pH 7.8, 1 mM EDTA,
1 mM dithioerythritol, 0.2% (v/v) Tween 20) containing 5 M GdmCl for 30 min at 10 °C. These conditions are
sufficient to unfold the protein completely, as demonstrated by
fluorescence and circular dichroism spectroscopy (see Fig. 2).
Renaturation was initiated by 100-fold dilution into thermostatted
buffer A or buffer B (10-40 °C, as indicated). To follow the
reactivation course, the renaturation samples were further incubated at
the given temperature by taking 3-µl aliquots at the indicated times
and assaying bioluminescence activity. Reactivation experiments in the
presence of peptidylprolyl cis-trans-isomerases were done at
10 and 25 °C at a luciferase concentration of 60 nM with
120 nM FKBP-52 or 1 µM rotamase,
respectively.
Analytical Chromatography
Native luciferase and the folding intermediate were separated on a Superdex 200 HR 10/30 column (Pharmacia Biotech Inc.) at a flow rate of 30 ml/h and a temperature of 5 °C using buffer B as the running buffer. The enzyme was denatured, and renaturation was initiated at 10 °C in buffer B as described above. After varied times of refolding, 200-µl samples were taken and subjected to gel filtration chromatography. The fluorescence emission of the eluted protein was detected by a Merck/Hitachi F-1050 spectrofluorometer. Fractions were collected to be assayed for luciferase activity immediately after chromatography or after further incubation for 1 or 3 days at 10 °C. For comparison of peak areas with the time course of luciferase reactivation, elution profiles were fit to the sum of two Voigt peaks using PeakFit (Jandel Scientific).
Analytical UltracentrifugationSedimentation velocity and sedimentation equilibrium runs were performed at 2 °C in a Beckman Instruments Model E analytical ultracentrifuge in an AN-D rotor at 60,000 and 12,000 rpm, respectively. The sample contained 0.2 mg/ml luciferase in storage buffer, and the sedimentation data were corrected to water and 20 °C. For the determination of the molecular mass from sedimentation equilibrium profiles, a partial specific volume of 0.748 ml/g, as calculated from the amino acid composition, was used (18).
In an initial series of experiments, we tested whether efficient reactivation of unfolded firefly luciferase is possible in the absence of molecular chaperones. P. pyralis luciferase was diluted into denaturant solution to a final concentration of 5 M GdmCl and kept for 30 min at 10 °C. Similar conditions have been employed previously (10) and are sufficient to completely unfold the protein (cf. Fig. 2 below). To prevent off pathway aggregation, refolding was attempted at low protein concentration (2-5 µg/ml) and low temperature (10 °C). Samples of unfolded luciferase were rapidly diluted 100-fold into cold phosphate buffer without additional ingredients and kept at 10 °C for several days. Within the first 24 h, refolding samples exhibited 50 ± 5% of the luminescence activity of native controls, for which no change in activity with time was observed under these conditions. Upon further incubation for 2 days, the activity of the refolding samples increased to 65 ± 5% of the controls. Addition of 1 mM dithioerythritol did not affect these results.
A further increase in reactivation yield was achieved by addition of a hydrophilic detergent. In the presence of 0.2% (v/v) Tween 20 and 1 mM dithioerythritol (to suppress peroxide formation; see Ref. 19) samples refolded for 24 h or 3 days exhibited luminescence activities of 65 ± 5 and 80 ± 5% of those observed for native controls, respectively. The activities of the controls were also higher after incubation in the presence of the additives (20-50% compared with samples incubated in buffer alone), indicating that the detergent may act by preventing adsorption of the enzyme to vessel walls, as observed with other proteins (20). In our hands, the addition of Tween 20 into the enzyme assay did not have a stimulatory effect, as reported previously (21).
Equilibrium Unfolding of Firefly LuciferaseEquilibrium
transitions induced by GdmCl and measured as changes in enzymatic
activity and fluorescence in the absence of detergent at low protein
concentration are depicted in Fig. 1, A and
B. Because GdmCl has an inhibitory effect on the enzymatic activity of P. pyralis luciferase (data not shown),
corresponding amounts of GdmCl were added to the luminescence assay
buffer in all experiments, keeping the overall GdmCl concentration in
the luminescence assay constant at 55 mM.
After 3 days at 10 °C no further change in luminescence activity (Fig. 1A) or intrinsic protein fluorescence (Fig. 1B) were observed, and the midpoints of unfolding and refolding transitions were essentially coincident. Although the enzymatic activity was lost in a single cooperative transition with a midpoint around 350 mM GdmCl, the denaturant-induced fluorescence change was clearly biphasic. In a first transition, the fluorescence intensity at the emission maximum of the native protein (334 nm) was reduced by ~35% with little change in the position of the emission maximum. This transition coincided with the inactivation of the enzyme. Beyond 1.2 M GdmCl, a further reduction in fluorescence intensity was observed, with a transition midpoint around 1.7 M GdmCl, and the emission maximum shifted to 351 nm. The presence of a pronounced plateau region around 1 M GdmCl indicated a highly populated equilibrium unfolding intermediate.
The unfolding transition, as observed by fluorescence, was not affected by changes in protein concentration (Fig. 1C). The transition measured at 3.7 µg/ml enzyme was superimposable onto that determined at 50 µg/ml. In contrast, renaturation yields were clearly dependent on luciferase concentration (Fig. 1D). Reactivation yields, measured relative to controls of native enzyme diluted to and incubated at the same protein concentration, approached 100% below 2 µg/ml but decreased to less then 40% at 10-fold higher protein concentration. To a first approximation, the concentration dependence could be described by a second-order aggregation reaction competing with a unimolecular folding process (22), as represented by the solid line in Fig. 1D.
The loss of secondary structure during unfolding at 50 µg/ml was followed by circular dichroism (Fig. 1E). Close to half of the CD signal amplitude at 220 nm was lost in a cooperative transition coincident with inactivation between 0 and 1 M GdmCl. Surprisingly, only a minor CD change paralleled the second phase of unfolding around 1.7 M GdmCl, whereas more than one-third of the CD change occurred in a highly cooperative transition with a midpoint at 3.8 M denaturant. The change in secondary structure around 3.8 M GdmCl was fully reversible (Fig. 1F) and rapid. When samples of native and completely unfolded luciferase were diluted to denaturant concentrations between 3 and 5 M, equilibrium ellipticities were attained in the dead time of manual mixing (data not shown). Thus, two equilibrium intermediates with varying contents of secondary and tertiary structure are populated during the reversible unfolding of luciferase.
Characterization of the Equilibrium IntermediatesFluorescence emission and CD spectra of the
intermediates in comparison with native and completely unfolded
luciferase are presented in Fig. 2. Both types of
spectra indicated that the enzyme was completely unfolded at 5.0 M GdmCl. The CD signal around 220 nm was 180 ± 600 degrees·cm2·dmol1, proving that the
protein had lost its secondary structure. Maximum fluorescence emission
occurred at 351 nm, near the emission maximum of solvent-exposed
tryptophan, and a distinct peak around 305 nm indicated spatial
separation of tyrosine and tryptophan fluorophores. Very similar
fluorescence spectra were observed in the plateau region of the CD
transition around 3 M GdmCl. In contrast, the fluorescence
spectrum of the first inactive intermediate was similar to that of
native luciferase, with an emission maximum at 334 nm, differing mainly
in fluorescence intensity. The data indicate that the intermediate
populated around 1 M GdmCl is compact enough to largely
shield its aromatic side chains from the solvent. In contrast, the
intermediate observed around 3 M GdmCl, although retaining
a large part of the secondary structure, does not significantly protect
the fluorophores.
It is apparent from the results presented above that luciferase unfolding is not completely reversible at elevated protein concentrations, i.e. under conditions in which a full set of spectroscopic and activity data can be acquired that might be subjected to global fitting procedures. Under certain conditions, however, individual inactivation and unfolding transitions could be sufficiently well separated to allow a quantitative analysis.
As pointed out above, the reversibility of the inactivation transition
could be improved by addition of Tween 20. Unfolding and refolding
transitions measured by fluorescence and bioluminescence after
incubation in the presence of Tween 20 were very similar to those
observed in buffer alone (Fig. 3, A and
B), although the reactivation yields were higher and the
midpoints of unfolding and refolding transitions coincided more
precisely. In the presence of Tween 20, the inactivation transition, as
well as the first phase of the fluorescence transition, were somewhat
more cooperative than in the absence of the detergent. This suggests an
interaction of the unfolding intermediate with the detergent molecules.
Since Tween 20 concentrations far above the critical micelle
concentration were used, Tween 20 micelles might be the interacting
species. The reactivation was superimposable with the inactivation
transition when the data were normalized to the final yield obtained at
very low denaturant concentrations. It was well separated from the second phase of the unfolding transition by a pronounced plateau in the
fluorescence signal around 1 M GdmCl (Fig. 3B).
On the basis of these observations, a thermodynamic analysis of the
inactivation transition appeared justified.
The inactivation data were fit to a N I two-state model (Fig.
3A, solid lines) because sedimentation velocity
and sedimentation equilibrium ultracentrifugation experiments indicated
the native enzyme to be monomeric (s20,w = 3.9 S, molecular mass = 60,600 Da). The analysis yielded a
denaturation free energy of
G0N
I1 = 15 ± 3 kJ·mol
1 in the absence of denaturant with a
denaturant dependence of m = 50 ± 6 kJ·mol
1·M
1. These results
did not change significantly when we used a three-state model with
inactive intermediate (I) and unfolded (U) species and a midpoint for
the I
U transition around 1.7 M, as observed in the
fluorescence transition (data not shown).
Like the inactivation transition measured under this specific set of
conditions, the unfolding of the second intermediate observed by
circular dichroism between 3 and 5 M GdmCl was highly cooperative, reversible, and well separated from other phases of the
unfolding transition. Gel filtration experiments at 4 M GdmCl resulted in a single peak consistent with a rapid equilibrium between two monomeric conformations. Thus, the unfolding free energy of
the second intermediate could be estimated. Assuming the CD of the
intermediate I2 and of the unfolded protein U to be
independent of the solvent composition, the data depicted in Fig.
1F can be described by a two-state equilibrium with an
unfolding free energy of
G0I2
U = 17 ± 3 kJ·mol
1 in the absence of denaturant and a denaturant
dependence of m = 4.5 ± 1 kJ·mol
1·M
1. Because no
reliable CD data could be obtained under conditions in which the first
two phases of the transition (N
I1
I2)
were found to be highly reversible, i.e. at nanomolar
protein concentrations and/or in the presence of Tween 20, a rigorous
quantitative description of the complete four-state unfolding
transition is not possible at present.
The equilibrium
experiments described had already indicated that luciferase
reactivation occurs very slowly at 10 °C. Therefore, we undertook a
more detailed analysis of reactivation kinetics at various
temperatures. Results are illustrated in Fig. 4. Because the time course of reactivation did not follow monophasic first-order kinetics, no attempt at calculating rate constants was made. The time
of half-maximal reactivation at 10 °C was 4 h. It was
independent of protein concentration between 0.2 and 4 µg/ml. With
increasing temperature, it decreased to 2.5 h at 15 °C and to
1.5 h at 20 °C. Above 20 °C, the activity of the native
control decreased significantly during the experiment, and the activity
of the refolding sample went through a maximum before decreasing at the
rate observed for the control. When reactivation was calculated
relative to the activity of the control sample at equivalent times, the
time of half-maximal reactivation was close to 50 min at 30 °C. At temperatures above 15 °C, the effect of Tween 20 on refolding yields
was more pronounced; however, omitting the detergent did not accelerate
the reactivation reactions.
Slow folding of monomeric proteins is frequently caused by the rate-limiting isomerization of peptide bonds preceding proline residues. Although such peptide bonds are in a defined conformation, either cis or trans, in native proteins, an equilibrium of 10-40% cis and 60-90% trans conformers is attained for any given Xaa-Pro peptide bond in the unfolded state. Because this equilibrium is reached slowly, a test for proline isomerization as a rate-limiting folding reaction comprises rapid unfolding of a protein at high denaturant concentration in a time that is short compared with the half-time of proline isomerization in the unfolded state (around 5 min at 10 °C; compare Ref. 23) followed by immediate dilution into refolding buffer (24). Thus, firefly luciferase was diluted into buffer containing 5 M GdmCl at 10 °C. Under these conditions, complete unfolding of the enzyme occurs in the dead time of mixing (data not shown). Renaturation, initiated within 10 s by a second dilution into refolding buffer, occurred at the same rate as it did in a sample diluted after 5 h of incubation in unfolding buffer (Fig. 4A). Accordingly, the presence of stoichiometric amounts of proline isomerases from the E. coli periplasm or from the eucaryotic cytosol did not accelerate luciferase reactivation after a long unfolding time, nor did the enzymes significantly affect luciferase reactivation yields (data not shown; see "Experimental Procedures" for details). The results prove that the rate of luciferase reactivation is not limited by proline isomerization.
The Kinetic Intermediate of Luciferase RefoldingAs pointed
out above, fully unfolded luciferase was observed to be in rapid
equilibrium with the I2 conformation at 3-5 M GdmCl. Similarly, a rapid fluorescence change in the dead time of
manual mixing was observed when the unfolded enzyme was diluted to
denaturant concentrations between 1 and 3 M and the final
amplitudes corresponded to those observed in the equilibrium
transitions. Both the I2 and the I1
intermediates thus appear to rapidly equilibrate with the fully
unfolded state. In contrast, slow reactivation with half-times of many
hours was observed when luciferase reactivation was initiated after
denaturing the enzyme overnight at 1 M GdmCl. Furthermore,
a fluorescence increase of around 20% of the final signal was observed
to coincide with the slow reactivation of unfolded luciferase. Taken
together, the observations suggest that the conformation of luciferase
preceding the rate-limiting step in the reactivation pathway may be
related to the I1 equilibrium intermediate. For a direct
comparison with the equilibrium unfolding intermediates, the putative
kinetic intermediate accumulating during refolding at low denaturant
concentration would have to be characterized further by fluorescence
and CD spectroscopy. This was precluded by the low protein
concentrations necessary for efficient refolding and by the strong
tendency of luciferase to adsorb to vessel walls during refolding in
fused silica cells. However, it was possible to characterize some
properties of the kinetic intermediate by analytical gel filtration
chromatography (Fig. 5).
In a series of experiments, equal amounts of luciferase, which had been allowed to renature for different time periods at 10 °C as described before, were applied to a high-resolution gel filtration column. Fluorescence emission was used to detect the eluted protein. Native luciferase eluted from the column after 31.3 min, corresponding to an apparent molecular mass of 53 kDa when the column was calibrated with globular proteins. As is apparent from the elution profiles depicted in Fig. 5A, two forms of luciferase were present in the refolding samples in varying proportions. In the absence of detergent, only the native protein fraction was eluted as a distinct peak, increasing with longer renaturation time (data not shown). The remainder of the protein was lost by adsorption to the column material. Addition of Tween 20 to the running buffer efficiently suppressed this interaction with the column matrix. Two protein peaks were now resolved in the elution profile, one eluting at 28.9 min and a second peak at 31.3 min coeluting with native luciferase. The peak at 28.9 min continually decreased and finally disappeared with progressive renaturation, whereas the peak corresponding to the native enzyme increased simultaneously. During refolding, the time course of formation of the native peak coincided closely with the appearance of enzymatic activity (compare Figs. 5A and 4A). To verify the precursor-product relationship, fractions of the two peak regions observed with the sample renatured for 10 min were collected and assayed for luciferase activity (Fig. 5B). When the fractions were assayed immediately after chromatography, only a very low amount of enzymatically active luciferase could be observed. Its elution position around 31 min indicated that it corresponded to the fraction of native protein formed from the intermediate during chromatography. However, when the fractions were assayed after further incubation for 1 or 3 days at 10 °C, the earlier eluting protein species became enzymatically active, proving that it represented a precursor of the native protein (Fig. 5B). The elution position of the intermediate corresponds to a globular protein of molecular mass 104 kDa. The 25% enlarged Stokes radius of the intermediate might indicate that it is a partially folded form, highly expanded compared with the native enzyme, or that it may be caused by an interaction with Tween 20 micelles.
In vitro renaturation of firefly luciferase in the absence of chaperone proteins has repeatedly been described as a very inefficient process blocked by the rapid aggregation of non-native protein chains (10, 12, 15). Our results, however, demonstrate that the enzyme is able to efficiently renature in the absence of cellular components. The conditions of efficient luciferase refolding, i.e. low protein concentration and low temperature, are those found to favor the refolding of other popular chaperone substrates such as citrate synthase, ribulose-bisphosphate carboxylase/oxygenase, malate dehydrogenase, and rhodanese (25, 26). Luciferase reactivation yields approach unity at protein concentrations below 2 µg/ml (~30 nM polypeptides), and the concentration dependence is that expected for kinetic competition between unimolecular folding and bimolecular aggregation reactions, as observed for lactate dehydrogenase (22). On the other hand, the midpoints of inactivation transitions induced by GdmCl are independent of protein concentration in the concentration range investigated (up to 50 µg/ml). This may suggest that the fraction of luciferase not reactivated is lost due to the much higher local protein concentrations transiently present during dilution from the denaturant solution.
Compared with other proteins, luciferase reactivation is exceedingly slow. At 10 °C, the time of half-maximal reactivation is greater than 4 h, and it takes 3 days to approach the refolding equilibrium. Although the renaturation reaction speeds up with increasing temperature, it clearly does not reach the rate observed for luciferase reactivation in cell extracts. In wheat germ extract at 25 °C or in reticulocyte lysate at 30 °C, reactivation half-times of 14 and 8 min, respectively, were observed, and the activation of de novo translated luciferase chains occurs even more rapidly (10-12). The extremely slow rate of refolding explains in part why firefly luciferase refolding in the absence of chaperones has not become apparent in earlier experiments.
The time course of luciferase reactivation does not follow first-order kinetics. At temperatures above 20 °C, renaturation is superimposed upon thermal inactivation processes in which the native protein as well as folding intermediates could be involved. Even at low temperatures, at least two exponentials are necessary to describe the reactivation kinetics, a faster first phase in the minutes time range and a very slow second phase taking hours. Complex kinetics in the formation of native molecules during protein refolding are often caused by the co-existence of fast folding and slow folding species in the unfolded state. Slow equilibrium reactions in unfolded protein chains, such as the cis-trans isomerization of Xaa-Pro peptide bonds, create such a mixture of species (24). Proline isomerization is most often found to be the rate-limiting step in the in vitro refolding of small proteins lacking disulfide bonds. Luciferase contains 29 proline residues and is therefore proline-rich (27). Nevertheless, we found no evidence for a role of proline isomerization in its slow reactivation process using a double jump procedure, in which the protein is rapidly unfolded in a time that is short compared with the half-life of proline isomerization and then is returned to native conditions after keeping it in the unfolded state for varied time spans. Similar observations were made by Kolb et al. (11) for the much faster reactivation of luciferase in wheat germ extract, which, however, does contain enzymes catalyzing proline isomerization. Accordingly, we did not observe acceleration of luciferase reactivation by added proline isomerases. Thus, the slow reactivation of luciferase under in vitro conditions and the heterogeneity in the reactivation kinetics need further investigation.
Upon prolonged incubation at low protein concentration, the midpoints
of unfolding and refolding transitions induced by GdmCl almost coincide
even in the absence of detergent. However, the reversibility of
luciferase unfolding is significantly improved and the loss of native
enzyme during prolonged incubation in dilute solution is essentially
prevented by addition of Tween 20, a relatively hydrophilic detergent
used previously to prevent adsorption of protein to vessel walls (20,
28). The unfolding/refolding equilibrium transitions observed by
fluorescence and CD spectroscopy and as changes in enzymatic activity
are compatible with a four-state model according to N I1
I2
U, where N is native,
enzymatically active luciferase, I1 and I2 are
partially unfolded, inactive intermediates, and U is the completely
unfolded polypeptide. The free energy of denaturation
(
G0N
I1) of around
15 kJ·mol
1 at 10 °C is small for a protein of 61 kDa. A significant fraction (around 0.2%) of the enzyme is denatured
at equilibrium even at 10 °C and in the presence of the stabilizing
phosphate ion. This is compatible with the observation that the enzyme
is inactivated when it is incubated with the chaperonin GroEL, which
tightly binds denatured proteins in the absence of
ATP.2 Although a rigorous thermodynamic
characterization of the complete four-state transition has not been
possible so far, it is obvious that the three unfolding steps add up to
a thermodynamic stability relative to the completely unfolded state of
roughly 50 kJ·mol
1. Thus, firefly luciferase is not an
unusually unstable structure; rather, the presence of stable folding
intermediates destabilizes the native state.
Surfactants have been used in a number of cases to increase refolding
yields of proteins (29-31). In the case of carbonic anhydrase II, the
addition of different surfactants to the renaturation buffer not only
affects refolding yields at high protein concentration but also
accelerates or slows down renaturation (32). Tween 20 is a relatively
hydrophilic detergent with a large, bulky polyoxyethylenesorbitan head
group. It is widely used as a nondenaturing detergent, and this, as
well as related detergents, were used previously as additives in the
enzymatic assay for firefly luciferase (21, 33). It clearly increases
luciferase renaturation yields, but it does not significantly affect
the rate of luciferase refolding. Although it hardly shifts the
midpoints of the two phases of the guanidine-induced equilibrium
transition, the cooperativity of the N I1 transition, measured as inactivation or by fluorescence, is somewhat higher in its
presence. This indicates a weak stabilizing interaction of Tween 20 with the equilibrium unfolding intermediate.
For a further analysis of the kinetic intermediate of refolding by gel filtration chromatography, the addition of Tween 20 to the elution buffer was inevitable because the detergent efficiently suppressed the adsorption of the non-native protein species to the column material. In its presence, an enzymatically inactive kinetic folding intermediate was quantitatively regained from the column. It eluted notably earlier than native luciferase at a position typical for a globular protein of about twice its molecular mass. A number of observations indicate that the kinetic intermediate identified by gel filtration chromatography is related to the equilibrium unfolding intermediate I1: (i), reactivation from the equilibrium intermediate and from the completely unfolded protein are equally slow; (ii), the conversion of the kinetic intermediate to the native protein coincides with a fluorescence increase; and (iii), CD and fluorescence changes across the second and third transitions at 1.7 and 3.8 M GdmCl occur rapidly. Since the equilibrium transitions indicated an interaction of I1 with the detergent, a possible explanation of its elution position is complex formation between the intermediate and a Tween 20 micelle. However, an enlarged Stokes radius should be expected for the intermediate on the basis of the reduced far-ultraviolet CD and fluorescence signals. The spectra suggest that about half the native secondary structure is retained in the intermediate I1 and that the aromatic side chains are still in a moderately hydrophobic environment. In contrast, the aromatic fluorophores are highly exposed in the intermediate I2 populated around 3 M GdmCl. Because both intermediates unfold and fold rapidly and their unfolding transitions are only moderately cooperative, they would be compatible with the definitions of collapsed, "molten globule" and "pre-molten globule" conformations (34).
Alternative explanations may be suggested by the crystal structure of P. pyralis luciferase that has recently been reported (3). Firefly luciferase folds into two compact domains connected by a short linker: a large N-terminal domain comprising residues 4-436 and a smaller C-terminal domain formed from amino acids 440-544. Since the active site of the enzyme is presumed to be located between the two domains, an intermediate with one of the two domains unfolded, although retaining large parts of the secondary and tertiary structure, would be enzymatically inactive. Moreover, the two tryptophan residues are located close to the linker region. The biphasic change in the fluorescence signal could therefore represent successive denaturation of the two domains, and the large CD amplitude of I2 might correspond to a subdomain of the large and highly complex N-terminal domain. We are currently trying to distinguish between the two possibilities using a fragmentation approach.
In any case, the observed equilibrium and kinetic intermediates in the folding of firefly luciferase constitute possible targets for the action of molecular chaperones. Because luciferase folding in the cell and in cellular extracts occurs orders of magnitude faster than in our refolding experiments, cellular folding helpers must either prevent the formation of the kinetically trapped species or catalyze its conversion to the native enzyme.
We thank Rainer Jaenicke for performing the ultracentrifugation experiments and Tatjana Tannenberg for gel filtration experiments with partially unfolded luciferase.