|Home | About | Journals | Submit | Contact Us | Français|
Intrinsic conformational transitions contribute to the catalytic action of many enzymes. Here we use a single-molecule approach to demonstrate how such transitions are linked to the catalytic sites of the eukaryotic proteasome, an essential protease of the ubiquitin pathway. The active sites of the cylindrical proteasomal core particle (CP) are located in a central chamber accessible through gated entry channels. By atomic force microscopy (AFM), we found continual alternation between open and closed gate conformations. We analyzed the relative abundance of these conformers in wild-type and mutated yeast CPs upon exposure to substrates or inhibitors. Our data indicate that the dynamic gate can be opened by allosteric coupling to a tetrahedral transition state at any of the working active centers. The results point to the Nα-amine of the N-terminal active site threonyl residue as the major effector group responsible for triggering the essential conformational switch.
Local and global conformational fluctuations are an intrinsic property of proteins (Henzler-Wildman et al., 2007). Conformational diversity in a single protein is defined by the ligand-independent presence of more than one conformational state (Bahar et al., 2007). Such intrinsic dynamics are postulated to determine catalytic functions and allosteric behavior, broadly understood as a coupling of conformational changes between two widely separated sites (Gunasekaran et al., 2004; Henzler-Wildman et al., 2007). Methods such as NMR, hydrogen/deuterium exchange and molecular modeling based on crystal structures have helped to advance our knowledge of the role of enzyme dynamics in catalysis (Dodson et al., 2008; Henzler-Wildman et al., 2007; Liu and Konermann, 2008). For large multi-component complexes, detailed analysis of the relationship between structural dynamics and activity is far from trivial. In such cases cryo-electron microscopy, atomic force microscopy (AFM) or fluorescence resonance energy transfer have allowed the detection of allosteric transitions (da Fonseca and Morris, 2008; Osmulski and Gaczynska, 2002; Tang et al., 2007). Such methods are also capable of determining whether a ligand-induced change in enzyme activity is best described by conformational selectivity and population-shift, in which the relative fractions of pre-existing conformational isomers are redistributed, or by induced fit, where an effector molecule directly induces a conformational change in the bound protein (Nevo et al., 2004).
Here we analyze the link between conformational dynamics and enzyme activity in the large hetero-oligomeric 20S proteasome. Most of the regulated degradation of intracellular proteins in eukaryotes occurs through the ubiquitin-proteasome system. Substrates are polyubiquitinated, and the tagged proteins are then degraded by the 26S proteasome, which is composed of a 20S proteasome catalytic core particle (CP) capped at each end by a 19S regulatory particle (RP) (Glickman and Ciechanover, 2002). The latter confers energy- and ubiquitin-dependence on substrate proteolysis. The CP alone is able to degrade small peptides and unfolded proteins or disordered loops in native proteins (Liu et al., 2003).
The 20S proteasome (700 kDa) is made of four stacked heptameric rings, which in eukaryotes assemble from 14 different but related α and β subunits. The rings have an α-β-β-α arrangement, with the outer a rings providing attachment sites on their outer surface (the α face) for the RPs or other regulatory modules; the α-ring forms a gate across the entrance to a central channel leading to the inner catalytic chamber (Groll et al., 1997, 2000a) (Figure 1A). This chamber is formed by the β rings and conceals three pairs of active sites: of the seven β subunits in eukaryotes, only three bear proteolytic active sites: β1, β2 and β5. The proteasome has been characterized as having chymotrypsin-like (ChT-L), trypsin-like (T-L) and postglutamyl peptide-hydrolyzing (PGPH; post-acidic, caspase-like) activities toward small peptide substrates. By mutational analysis in the yeast Saccharomyces cerevisiae, the β5 (Doa3/Pre2/Prg1), β2 (Pup1) and β1 (Pre3) subunits, respectively, are responsible for these peptidase activities (Arendt and Hochstrasser, 1997; Chen and Hochstrasser, 1996; Heinemeyer et al., 1997). All active β subunits are synthesized with an N-terminal propeptide, which is autoprocessed during proteasome assembly to free an N-terminal amine, ready to serve as a base for a nucleophilic hydroxyl of the catalytic Thr1 residue (Chen and Hochstrasser, 1996; Schmidtke et al., 1996; Seemuller et al., 1996). Mutation of this Thr residue blocks subunit processing; the propeptide fragment that remains in the mutant prevents both substrate binding and catalysis at the affected active center.
Mounting evidence suggests that the proteasome is an allosteric enzyme. The examples of presumed allostery include interactions between catalytic centers and putative distal substrate-binding sites and routes from the α face to active sites upon binding distinct activator complexes (Rechsteiner and Hill, 2005; Schmidtke et al., 2000). The most intriguing cases describe apparent allosteric coupling between the catalytic chamber and the interface between the CP and RP. In a recent study, binding of a competitive inhibitor to an active site was correlated with increased stability of the CP-RP complex (Kleijnen et al., 2007). Coupling between degradation and the dynamic structure of the a ring was also apparent in our earlier AFM–based analysis of S. pombe proteasomes, where we reported shifts in the partitioning between open and closed conformations of the α-ring upon addition of substrates (Osmulski and Gaczynska, 2000, 2002).
The structures of open and closed conformations are known in great detail from X-ray crystallography (Groll et al. 2000a, Whitby et al., 2000) (Figure 1A). The gate responsible for opening and closing the outer pore is formed by N-terminal tails of a subset of a subunits, which either lock together (“closed gate”) or move upward in a concerted fashion (“open gate”) (Forster et al., 2003). The gate is closed in crystal structures of wild-type yeast CP. Such proteasomes in solution are “latent”: they have relatively low, albeit measurable, peptidase activity (Bajorek et al., 2003). Activity is elevated many-fold by opening the gate by, for example, binding regulatory protein modules such as the RP to the α ring or by mutational removal of parts of the gate (Groll et al., 2000a; Smith et al., 2007; Forster et al., 2003). Gate opening may also result from interactions of the CP with substrates – either at the active sites or elsewhere (Kisselev et al., 2002; Liu et al., 2003; Osmulski and Gaczynska, 2000, 2002). The dynamics of the area surrounding the entrance to the central channel in archaeal proteasomes was recently shown by NMR (Sprangers and Kay, 2007). What remains obscure, however, is the nature of the initiating triggers for these structural changes and the allosteric pathways by which they are transmitted.
In this study, we examined the relationship between proteolytic active-site engagement and the conformational state of the α-ring gate in 20S proteasomes purified from S. cerevisiae. Atomic force microscopy (AFM) is a type of single-molecule analysis that can be performed in aqueous media that mimic physiological conditions. Using AFM, we observed a strict correlation between specific states of the β-subunit active centers and gate opening or closing in the α-ring. Our data suggest that each active center in the catalytic chamber is allosterically coupled to the α-ring gate, with a tetrahedral transition state at an active center serving as the likely trigger for gate opening (or for inhibiting its closing). Here we propose a model of a dynamic α-ring gate, in which active centers influence the gate by biasing a pre-existing equilibrium of conformational states. Such a dynamic gate allows access of substrate to the catalytic chamber even without the attachment of the RP or other proteasomal activators. This is consistent with the observation that free 20S proteasomes exhibit physiologically relevant catalytic activity in vivo (Liu et al., 2003).
We examined the topography of latent 20S yeast proteasomes in aqueous buffer under nondestructive conditions with tapping mode AFM, using experimental conditions described previously (Osmulski and Gaczynska, 2000, 2002) and in Supplemental Data. The topography of the particles was consistent with the CP cylinders either standing on their α–ring ends (top-view orientation; majority of particles; Figure 1B, Figure S1A–C) or lying on their sides (side view; Figure S1D). For both orientations, the particles showed a cylinder diameter of about 8–12 nm and a cylinder length of 13–17 nm, which agreed well with crystallographic (Groll et al., 1997) and electron microscopic (da Fonseca and Morris, 2008) data for the 20S proteasome and with our previous AFM data (Gaczynska and Osmulski, 2008) .
Figure 1A–C shows AFM images of S. cerevisiae CPs standing on their α-ring ends (top view) in two distinct conformations: either they presented a smooth, cone-shaped apical end (“closed”) or a crater-like depression in the middle (“open”). As in our previous studies, we classified the molecules on the basis of the shape of the median sections of the α-rings (Figure 1B). Individual proteasomes were observed to alternate between the open and closed states on the mica surface (see Supplemental Data). We confirmed the presence of two conformers with multiple species (Gaczynska and Osmulski, 2008) and determined that approximately 25% of wild-type S. cerevisiae proteasomes imaged without addition of a ligand were present in the open state, and 75% in the closed state (Table 1).
The distribution of open and closed conformers did not always match the 1:3 ratio noted above. As in our previous studies with S. pombe proteasomes (Osmulski and Gaczynska, 2000), in S. cerevisiae particles this ratio inverted to a ~3:1 ratio of open-to-closed conformers when we added a model peptide substrate for any of the three active centers (Figure 1C). Specifically, the percent of open particles in the presence of a substrate for the ChT-L site (β5) was 73.9% ± 3.1%, for the T-L site (β2) 74.5% ± 2.8%, and for the PGPH site (β1) 75.7% ± 2.3% (200–500 top-view particles each, from 8–20 fields). The similar observation with S. pombe proteasomes had prompted us to hypothesize that the catalytic act promoted channel opening (Osmulski and Gaczynska, 2000).
To confirm the coupling between catalytic sites and α-ring conformational changes, we tested if a functional active center was necessary to change the distribution of conformers upon addition of substrate. For this purpose, we initially analyzed the effect of mutating each active site Thr1 to Ala. Strikingly, disruption of the β1, β2 or β5 active sites only blocked α-ring channel opening when a model peptide substrate specific for the disabled site was used (Figure 2A). Substrates of the remaining intact sites continued to stimulate high levels of the open conformers, and the basal distributions of conformers were the same as that of the wild-type proteasome. Therefore, inactivation of each active site specifically disrupts the shift toward the open channel conformation normally induced by a peptide substrate specific to that site.
Next we attempted to determine which step(s) of the proteolytic catalytic cycle triggers α-ring channel opening. In particular, we considered substrate binding at active sites or other sites, threonine acylation and formation of a tetrahedral intermediate (Figure 3). Although all three analyzed active site mutants lacked the essential Thr1 residue in their respective active sites, a potentially important difference existed between them. For the β1 and β2 T1A substitutions, the mutations were introduced into the full-length precursor forms of these subunits, which causes a precursor processing defect (Arendt and Hochstrasser, 1997). As a result, a residual propeptide (leader sequence, LS) is left in the substrate-binding pocket, as was shown in crystal structure of the β1T1A proteasome (Arendt and Hochstrasser, 1997; Groll et al., 1997). The analogous mutation in the β5 precursor is lethal, but if the Ala1 mutation is introduced into β5ΔLS (propeptide deleted) in cells with the propeptide expressed in trans, proteasome assembly can take place and cells survive (Chen and Hochstrasser, 1996). Hence, a β5ΔLS-T1A mutant was analyzed here, resulting in a β5 substrate-binding pocket that should be fully accessible. Despite these differences in binding-pocket occupancy in the three active-site mutant proteasomes, all the particles behaved analogously in response to substrate peptides (Figure 2A). These data suggest that some step associated with substrate cleavage, rather than peptide binding per se (at either the active site or elsewhere in the proteasome), is necessary for channel opening.
Based on this preliminary inference, we began to dissect how altering components of a proteasome active center or catalytic events at that site affect the distribution of α-ring conformers. Each active center includes a Thr1-based catalytic dyad with its hydroxyl side chain acting as the attacking nucleophile and α-amino group as general base (Figure 3). A substrate binding pocket and water molecules are also essential components of the active site (Groll et al., 1997). Analogous to serine proteases, the proteasomal catalytic cycle most probably consists of the following steps: (1) nucleophilic attack by the Thr1 hydroxyl on the carbonyl of the peptide bond with transfer of the hydroxyl proton through a bridging water to the Nα-amino group; (2) rearrangement of a tetrahedral intermediate to form a covalent acyl intermediate and release of the first product; (3) nucleophilic attack of a water molecule on the carbonyl group esterified to Thr1, leading to formation of a second tetrahedral complex; and (4) rearrangement of the complex and release of the final product (Figure 3) (Osmulski et al., 2009). We addressed the state of the catalytic dyad and the formation of a tetrahedral complex in relation to channel opening using mechanistically distinct active-site inhibitors and yeast mutants.
Peptide epoxyketones constitute a family of highly specific and irreversible inhibitors of the proteasome, forming a covalent morpholino adduct with the active site Thr1 (Supplemental Data, Figure S2). Chemically, they engage both the hydroxyl and α-amino group of the threonyl residue (Elofsson et al., 1999; Groll et al., 2000b). To test their effects on proteasome conformational switching, the epoxyketone compounds epoxomicin (directed primarily toward β5), YU101 (specific for β5) and YU102 (directed primarily toward β1) were used at concentrations sufficient to block the single favored active center (Table 1, experiments 1–4) (Elofsson et al., 1999; Myung et al., 2001). Strikingly, inhibitor treatments shifted the fraction of open conformers detected in AFM images to as high as 96%. The targeted active center, however, had to be intact in order to support the conformational switch (Table 1, experiments 5–7). Mutant proteasomes with Thr1 of β1 replaced by Ala failed to shift to a high fraction of open conformers upon their treatment with the β1-specific inhibitor YU102. In contrast, the conformational shift was preserved in b1T1A mutant proteasomes treated with an inhibitor targeting the β5 active centers (YU101). Since no product is released by proteasomes blocked with epoxyketones, the conformational shift cannot be due to the release of peptide products. These results supported the possibility that an interaction of a substrate with at least some component of the active center shifts proteasomes to the open state.
Following this lead, we compared the effects of other proteasome inhibitors that form chemically distinct adducts with the hydroxyl of Thr1. First, we tested Z-LLL-VS, a vinyl-sulfone peptide derivative that irreversibly inhibits the β5 active center as its primary target. Peptide vinyl sulfones attack the hydroxyl of Thr1 residue, forming an ether adduct, but do not react with the α-amino group (Figure S2) (Bogyo et al., 1997). In contrast to epoxyketone addition, treatment of proteasomes with Z-LLL-VS did not lead to α-ring opening (Table 1, experiment 8). A similar result was obtained with clasto-lactacystin β-lactone, which also reacts with the hydroxyl group but not with the α-amino group (Table 1, experiment 9; Figure S2). The β-lactone forms an ester adduct analogous to the acyl intermediate in the proteasome reaction cycle (Corey et al., 1999;). Therefore, the AFM data argue against an acyl intermediate as a key allosteric effector for α-ring opening.
Next, we inspected proteasomes treated with the boronic acid peptide derivatives bortezomib [pyrazylcarbonyl-FL-B(OH)2] and MG262 [Z-LLL-B(OH)2]. Boronic acid derivatives directly block the hydroxyl group, occupy the binding pocket and increase basicity of the catalytic amine in serine or threonine proteases. The stable, anionic covalent adduct formed by boronic acids closely resembles the transition-state tetrahedral complex created during peptide bond hydrolysis (Figure 3, Figure S2) (Groll et al., 2006; Transue et al., 2004). Binding of bortezomib or MG262 caused proteasomes to switch their conformations from mostly closed to mostly open in a dose-dependent manner (Table 1, experiment 10 and Figure 1B,C; Figure 2C). These results suggest that formation of a tetrahedral transition state during peptide bond cleavage, which is mimicked by the boronic acid adduct, is linked to the conformational shift from closed to open proteasomes.
In summary, these results point to a tetrahedral reaction intermediate as a key allosteric factor promoting α-ring opening. Since occupancy of the substrate-binding pocket or blocking the Thr1 hydroxyl group was insufficient for stimulating the conformational shift, we turned our attention to the N-terminal amino group.
None of the available proteasome inhibitors exclusively targets the α-amino group of any catalytic N-terminal threonine. However, we could take advantage of the previous observation that the propeptides of active β subunits are needed to protect the Thr1 catalytic dyad from Nα-acetylation during protein synthesis. The propeptides are cleaved off during proteasome assembly, when the exposed threonines are concealed within the newly formed catalytic chamber and thus no longer accessible to acetyltransferases (Arendt and Hochstrasser, 1999). As was established before, in yeast strains with a mutation that deleted a particular β-subunit propeptide sequence, the Nα-acetyltransferase Nat1-Ard1 specifically modified the α-amino group of the exposed Thr1 (Figure S2; Arendt and Hochstrasser, 1999). Acetylation of Thr1 abolishes catalytic activity of the affected active center. For example, deletion of the β1 propeptide in the β1ΔLS mutant strain MHY1377 resulted in almost complete (as proved with N-terminal peptide sequencing [Arendt and Hochstrasser, 1999] and mass spectrometry; data not shown) Nα-acetylation of β1-Thr1 and loss of all detectable PGPH activity (Arendt and Hochstrasser, 1999). As reported before, PGPH activity was fully restored by elimination of the Nat1 acetyltransferase subunit (β1ΔLS nat1Δ; MHY1373; Arendt and Hochstrasser, 1999).
Remarkably, when assayed by AFM, β1ΔLS mutant proteasomes, in which β1-Thr1 is Nα-acetylated, were nearly all in the open conformation (MHY1377, Figure 1B,C, Figure 2C–a). In contrast, β1ΔLS proteasomes isolated from cells lacking Nat1 (MHY1373) again showed the wild type ratio of conformers (mostly closed; Figure 2C–b). Deletion of Nat1 in cells with wild-type proteasomes did not influence the partitioning of conformers (MHY1372; Figure 2C–e). If the α-amino group, but not the hydroxyl of Thr1, was pivotal in the allosteric transition, proteasomes with a substitution of the β1ΔLS Thr1 with Ala (β1ΔLS-T1A MHY2267) should behave the same as β1ΔLS proteasomes because an N-terminal alanine is also a favored target for Nat1-Ard1 (Polevoda et al., 1999). Indeed, we found that these proteasomes were always open (Figure 2C–c), and the shift was again reversed by deletion of Nat1 (MHY2271; Figure 2C–d). Interestingly, the relative abundance of open conformers correlated closely with the degree of acetylation of the N-terminal residue of different catalytic β subunits (see Supplemental Data). From these results, we conclude that the irreversible acetylation of the α-amino group of an active site residue induces a permanent opening of the proteasome α ring.
In order to test whether the observed opening correlates with functional parameters of the particles, we tested the ability of wild-type (WT) and mutant 20S proteasomes (CPs) to be functionally activated, and compared the results with the partitioning between open and closed conformers. As a measure of functional activation we used the rate of degradation of SucLLVY-MCA (ChT-L activity). As an activating agent we chose the human 11S REG/PA28αβ (“PA28”) activator complex. A homolog of PA28 was proved to open the gate of the yeast CP, and to increase the rate of degradation of SucLLVY-MCA (Whitby et al., 2000). Interestingly, the “open-gate” mutants (“morphologically activated”) were refractory to functional activation with PA28 (Figure 2C). In contrast, SucLLVY-MCA cleavage by WT CPs or mutant CPs that had mostly closed-gate conformers was activated upon addition of PA28 (Figure 2C). This result links our structural observations with the functional properties of the CP. It is worth noting that active site-affecting treatments or mutations that shift the CP toward the open state do not generally cause significant activation of the remaining unaltered sites (Figure 2C and data not shown). As shown previously (Myung et al., 2001), peptide substrates for one catalytic site can inhibit activity at the other catalytic centers by a still unknown allosteric mechanism. Thus, a complex set of allosteric interactions occurs in the proteasome, making it difficult to predict the exact functional consequences of altering a particular active center. This does not amend the conclusion of specific allosteric coupling between active sites and the α-ring gate.
In apparent contradiction to our AFM findings, crystal structure-based models of eukaryotic 20S proteasomes with an intact α-ring gate display exclusively closed particles even when they had been treated with epoxomicin (Groll et al., 2000b), bortezomib (Groll et al., 2006) or had the Thr1 amine acetylated (Loidl et al., 1999), all treatments that led to preferred α-ring opening in the AFM analyses. This difference might reflect the influence of physical constraints on proteasome conformation imposed by crystal packing forces. In an attempt to mimic the confinement of particles in the crystal lattice, we examined the relative abundance of different proteasome conformers deposited on mica at higher densities. Whenever proteasomes were deposited at a density ensuring minimal contacts among the particles, the abundance of the open conformer stayed close to the expected 25% for the control (Table 1) and was almost 100% for proteasomes with Nα-acetylated β1-Thr1 (MHY1377, Figure 1B,C; Figure 2C–a). Increasing the density to a tightly packed monolayer of proteasomes reduced these values to 8% ± 2.9% for the control (n=30 fields with 1084 top-view particles) and to only 16.0% ± 5.7% for MHY1377 proteasomes (n=35 fields, 1081 particles). These results strongly suggested that tight packing of proteasomes hampers the opening of the α-ring gate. This was further supported by the results of Thess et al. (2002), where AFM analysis of 20S proteasomes close-packed in a two-dimensional crystal monolayer revealed that the molecules were exclusively in the closed conformation.
Collectively, our AFM analyses suggest that the presence of an uncharged primary amino group on the N-terminal residue in the active center favors a closed conformation in the α-ring. This is the state of the Nα-amino group in “idle” proteasomes and in proteasomes blocked with lactacystin or a peptide vinyl sulfone (Table 1). In contrast, formation of a secondary amine, either by its acetylation or reaction with epoxyketone (but not by blocking with a propeptide; see Discussion), correlates with a shift toward the open conformation (Table 1). A similar preference is observed when the threonyl amine turns strongly basic during tetrahedral intermediate formation (Figure 1B,C, Figure 2B, Figure 3). We propose that the departure from the neutral primary amine status leads to structural rearrangement of the active center or its immediate surroundings, triggering an allosteric opening of the α-ring channel (Figure 3).
The present work has revealed a striking coupling between the proteasome β-subunit active centers and the opening and closing of the α-ring channel. Substrate engagement at any of the three different pairs of active sites results in a much higher proportion of particles with an open α-ring. Systematic studies of inhibitor-proteasome complexes and yeast active-site mutants indicate that the allosteric signal is linked to the α-amino group of the catalytic Thr1 residue. We infer that a tetrahedral transition state, which occurs during the proteasome reaction cycle, initiates local conformational changes that eventually propagate to the α-ring gate.
The characteristic cylindrical shape of core particles facilitated their identification in AFM images, either as particles standing on their α-ring end or lying on their side. By AFM, S. cerevisiae proteasomes were indistinguishable from S. pombe or human complexes (Gaczynska et al., 2003; Osmulski and Gaczynska, 2000). The proteasomes were attached to the mica surface at a concentration sufficiently low to allow complete separation of the individual particles. This contrasts with other studies in which a dense layer of particles was created when His-epitope tagged proteasomes were arrayed on a derivatized surface (Dorn et al., 1999). Although such dense immobilization tends to enforce a favored orientation of the packed molecules, it limits the ability to analyze single molecules (Dorn et al., 1999; Furuike et al., 2003). By contrast, our approach allowed straightforward particle recognition and dimensioning without the need for modification or enforced placement on the surface. Our measurements of proteasome dimensions agree very well with those from published crystallographic data (Groll et al., 1997). Therefore, we conclude that AFM imaging of untagged proteasomes imaged at low density accurately depicts the topography of the yeast proteasome.
Latent yeast proteasomes, although attached to a mica surface, still switched continually between states in which the central pore in the α-ring was either open or closed in end-on views. The cycles of opening and closing resemble anharmonic fluctuations since they did not follow any detectable pattern when individual proteasomes were monitored for an extended period. There also was no observable tendency among the proteasomes to synchronize their gate cycling. However, on average, ~75% of the wild-type complexes were in the closed conformation, similar to other eukaryotic proteasomes (Gaczynska and Osmulski, 2008). The less abundant open conformer is presumably less stable and as a result, shortlived in comparison to the more prevalent closed conformer. This hypothesis is strengthened by our recent data suggesting that the millisecond-scale lifetime of the closed form is three times longer than the lifetime of the open form (Osmulski and Gaczynska, unpublished observations).
It was initially puzzling that the open proteasome conformation clearly distinguished in AFM images had not been detected in crystal structures of wild-type proteasomes (Groll et al., 1997). One plausible explanation is that the closed state, which is already more stable in solution than the open state, is further stabilized by the crystal lattice. Indeed, our previous results suggest that the transition from the open to closed state is not restricted to a small section of the α-ring, but rather, is accompanied by a change in the overall shape of the particle (Osmulski and Gaczynska, 2002). Very recently, a pronounced widening of the α-ring in the 20S proteasome was detected by electron microscopy upon binding of the 19S RP, which opens the central gate (da Fonseca and Morris, 2008). Proteasomes immobilized in a crystal lattice may not have enough energy to achieve such a structural transition. In agreement with this idea, more densely packed proteasomes deposited on mica exhibit a greatly decreased tendency to switch out of the closed conformation. Moreover, proteasomes with an acetylated catalytic N-terminus, which were classified in AFM images as being almost entirely open when deposited at low concentration, apparently closed the gate in more than 80% of particles when analyzed as a densely packed layer. These data are also consistent with the possibility that α-ring gate opening is coupled with a global conformational change of the β rings. Interestingly, conformational fluctuations of proteins were found to be impeded by molecular crowding based on a combination of X-ray scattering and computational modeling (Makowski et al., 2008).
The proteasome is by no means unique in its structural dynamics. Conformational diversity is likely to be a common if not universal property of proteins (Bahar et al., 2007). Such intrinsic, thermodynamically driven motions can be harnessed to achieve catalytic proficiency in enzymes (Bahar et al., 2007; Henzler-Wildman et al., 2007). Closed and open conformations are evident in many multisubunit assemblies with cylindrical or ring structures, such as chaperones, membrane channels, and receptors (Muller and Engel, 1999; Schoehn et al., 2000; Stolz et al., 2000). With up to 25% of particles showing an alternative conformer, proteasomal open and closed conformers must have a rather small free energy difference. The relatively high probability of latent proteasomes switching to a state allowing access to the catalytic chamber explains the well-documented ability of latent CP to hydrolyze substrates. A dynamic gate in the α-ring would allow an unfolded protein or protein domain to enter the CP without involvement of the 19S RP or other components of the ubiquitin system. This has in fact been documented with intrinsically unfolded proteins (Liu et al., 2003). The low, but measurable rates of degradation of model peptide substrates or denatured polypeptides commonly observed with latent eukaryotic CPs can also be explained by a stochastic opening of the α-ring gate (Bajorek et al., 2003; Rosenzweig et al., 2008). Previously, the activity of latent CP against certain hydrophobic peptides or unfolded domains was proposed to derive from their direct interaction with the α-ring, promoting gate opening “from the outside” (Kisselev et al., 2002; Liu et al., 2003). Although plausible, this model applies to only a subset of substrates and requires the existence of specific receptor sites in the α ring. The conformational diversity hypothesis explains activity toward all CP substrates without evoking additional structural features.
Proteasomes treated with substrates, both short peptides and denatured proteins, retain the ability to switch between open and closed states, and the respective conformers are indistinguishable from those observed in the absence of ligands. However, exposure to a degradable polypeptide shifts the conformational equilibrium toward the open conformer (Osmulski and Gaczynska, 2000). The shift is fully reversible by washing out the substrate from the reaction mixture (Osmulski and Gaczynska, 2002). Here we confirm that the cognate active center has to be fully functional for this effect. Any mutational disruption or chemical modification with active-site inhibitors prevents the equilibrium shift. Moreover, our preliminary findings show that with substrate concentrations below the steady-state conditions, the shift was reversed with time: a strong suggestion of substrate depletion (Osmulski and Gaczynska, unpublished observations). Finally, we note that our model of active site-regulated proteasome dynamics is not in conflict with previously postulated noncatalytic allosteric sites (Myung et al., 2001; Schmidtke et al., 2000). Indeed, such interactions can be invoked to explain the strong activating effect of certain hydrophobic peptides on substrate hydrolysis without the need of gate opening “from the outside” by these peptides (Kisselev et al., 2002).
Previously, we had hypothesized that the relative abundance of the two detectable proteasome conformations might be related to proteasomal catalytic activity (Osmulski and Gaczynska, 2000, 2002). Our new AFM data reveal a clear coupling between the CP active centers and α-ring conformation. Such coupling is consistent with recent observations suggesting that changes in the stability of CP-RP interactions correlate with changes in catalytic chamber engagement (Kleijnen et al., 2007). Our analysis of proteasomes exposed to reagents modifying specific elements of the active centers and examination of yeast active-site mutants point to the tetrahedral transition state as the most probable initiator of the conformational switch. Transition state-related structural changes in the active center are proposed to trigger a cascade of conformational alterations leading to the opening of the remote α-ring gate. Slow, millisecond-scale conformational transitions connected to catalytic cycles have been reported for a variety of enzymes (Hammes-Schiffer and Benkovic, 2006). In general, the observed activity-related motions on a microsecond-to-millisecond time scale fit the concept of conformational sampling performed by enzymes to optimize catalysis (Bahar et al., 2007).
A search for specific molecular elements in the active center that were potentially involved in conformational signaling fingered the Thr1 α-amino group as the strongest candidate for the allosteric trigger (Table 1, Figure 4). We observed a dramatic shift of the conformational equilibria toward the open form when the α-amine was engaged by a substrate, a transition state analog, epoxyketone inhibitors, or an acetyl group. All of these compounds react with Thr1 in a way that either creates a secondary amine or substantially increases the basicity of the primary amine (Figure 4). By analogy to serine proteases, whose catalytic mechanism is closely related to that of the proteasome, these changes should cause subtle but significant structural remodeling of the active center (Groll et al., 2006; Radisky et al., 2006). We therefore propose that alteration of the a amine of Thr1 is the proximate trigger for the conformational switch that culminates many nanometers away at the proteasome gate.
Studies on allosteric signaling have suggested that the transmission of signals over long distances is based on reorganization of hydrogen bonds or, less often, salt bridges (for example: Begg et al., 2006; Guallar et al., 2006; Yun et al., 2007). Sometimes breaking just a single hydrogen bond is sufficient to initiate a cascade of allosteric changes, as in the E. coli purine repressor (Lu et al., 1998). In another example, coupling between local conformational fluctuations, proton binding, and global structural motions was described for Staphylococcal nuclease (Whitten et al., 2005).
By analogy, we propose that changes to the proteasome Thr1 α-amine are translated into a set of adaptations within the active site hydrogen-bonding network, which is eventually transmitted into a structural remodeling of the proteasome gate. Thermally induced rearrangements of hydrogen bonds in the active center of “idle” proteasome may mimic the transition state-like configuration and thereby support a significantly populated open gate state (25%). This model can accommodate the one known discrepancy in the correlation between Thr1 a amine modification and gate opening. Specifically, in the β1T1A proteasome, which still retains a β1 propeptide remnant in the substrate binding pocket and therefore has a modified amine at Ala1, the α-ring gate is nevertheless closed both in the crystal structure (Groll et al., 1997) and in the majority of particles analyzed by AFM (Figure 2C, Table 1). The distinct geometry of the propeptide extending beyond the active center, in combination with the unique organization of surrounding hydrogen bonds, disables the expected shift toward the predominantly open conformer (Osmulski, Hochstrasser, Gaczynska, in preparation).
Studies are underway to test the proposed role of remodeling of the hydrogen-bond network in the active site vicinity as a source of the observed gate opening. We are also exploring the allosteric pathways leading from the active centers to the α-ring gate. Ultimately, a combination of AFM, high-resolution crystallography, mutational, and biochemical studies will be needed to derive a complete picture of the dynamic processes linking the proteasome active sites and its gated entry channels.
Latent 20S proteasomes were isolated from exponentially growing cultures of Saccharomyces cerevisiae in YPD medium (see Supplemental Data). Activity measurements were performed with one of three commonly used fluorogenic peptide substrates, as outlined in Supplemental Data.
The imaging was performed in tapping mode in liquid, as described previously (Osmulski and Gaczynska, 2000) and in Supplemental Data. Standard plane fits provided by the NanoScope IIIa software (v. 5.1) were the only processing procedures applied to the images of the whole 1 µm2 fields in the height mode.
In our previous studies, we found that 20S proteasomes imaged with AFM had two conformations, representing an open or closed entrance to the central channel (summarized in Gaczynska and Osmulski, 2008). They were distinguished by the qualitative shape of the median sections of particles in top-view orientation (Osmulski and Gaczynska, 2000; Supplemental Data). In short, median sections (density profiles) of the particles were analyzed with section-analysis software (Veeco). Only if all the sections were crater-shaped, with a depression in the middle of the α-ring, the particle was counted as open (Figure 1B).
This work was supported by NIH grants R01 GM069819 (M.G.) and R01 GM083050 (M.H.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.