A single round of encapsulation within the closed chamber suffices for substrate folding
We initially examined whether the folding reaction is completed within the closed central chamber of group II chaperonins. In principle, folding of a polypeptide with a strict chaperonin requirement, i.e. a stringent substrate, could require several cycles of Cpn binding and release (). Alternatively, the substrate could fold in a single ATPase cycle event, without requiring multiple rounds of binding and release. To test these possibilities, we employed rhodanese, a stringent Cpn substrate (Martin et al., 1991
S-rhodanese binds to nucleotide-free Cpn in an unstructured, proteinase K (herein PK)-sensitive state (, left arrow; , lane 2 bottom panel and for Native gel analysis). Addition of ATP induces lid closure and encapsulates the substrate within the closed chamber (Meyer et al., 2003
; Reissmann et al., 2007
)(). Upon closure, the Cpn lid segments and the encapsulated 35
S-rhodanese are protected from proteolytic digestion (, lane 3). Importantly, ATP addition causes the time-dependent folding of rhodanese (, red symbols). Comparing the kinetics of rhodanese folding (t1/2
~12 min) with the estimated kinetics of a single round of ATP hydrolysis (Bigotti et al., 2006
; Reissmann et al., 2007
) indicates that completion of rhodanese folding involves several cycles of ATP binding and release. Importantly, addition of protease at any time following ATP addition interrupted the folding reaction (, PK, shown for t=0 and t=13 min). Since PK can only degrade the substrate if the lid is open, this result suggests that the Cpn-substrate complex undergoes repeated cycles of ATP-driven opening and closing during the folding reaction.
Role of substrate encapsulation and iterative cycling in Group II chaperonin action
We next examined whether such iterative cycling is required to achieve folding by exploiting the observation that addition of AlFx together with ATP locks group II chaperonins in a symmetrically closed state that fully encapsulates the substrate (Meyer et al., 2003
)(; right arrow). The ATP-AlFx-induced state of Cpn-rhodanese was locked closed, leading to full proteolytic protection of both Cpn and substrate (, lane 4) and a characteristic electrophoretic migration shift on native gels (). Under these conditions, ATPase cycling is interrupted (Fig. S1A
), and the substrate undergoes a single round of binding and encapsulation, allowing us to evaluate whether iterative cycling is required for group II chaperonin folding (). Strikingly, the rate and yield of rhodanese folding under these non-cycling conditions were identical to those observed for the actively cycling chaperonin (). Furthermore, addition of PK to the ATP-AlFx reaction did not interrupt folding, confirming that there was no reopening of the Cpn and no release of the non-native substrate under these conditions. We conclude that the closed chamber of group II chaperonins is the folding active compartment. Furthermore, a single round of encapsulation in this chamber can achieve maximum rhodanese folding, with similar kinetics and yield as observed under cycling conditions. Thus, although iterative cycling does occur, it is not strictly required for Cpn-dependent folding.
The closed, folding active, state of group II chaperonins requires ATP hydrolysis
To examine whether ATP binding suffices to promote the folding active state of group II chaperonins, we specifically impaired the ATPase active site by targeting Asp386, which is essential to coordinate the water molecule that participates as a nucleophile during the hydrolysis of the phosphate-anhydride bond (Cpn-D386A; ). Cpn-D386A cannot hydrolyze ATP but retains efficient ATP binding [data not shown and (Reissmann et al., 2007
)]. Importantly, unlike Cpn-WT, Cpn-D386A is unable to fold the stringent Cpn substrates rhodanese () and malate dehydrogenase (data not shown). This demonstrates that ATP binding is insufficient to induce the fully folding-active state observed upon ATP hydrolysis.
ATP hydrolysis is required for the closed folding active state
We next assessed the proposal that ATP binding leads to partial (Clare et al., 2008
) or full (Iizuka et al., 2003
; Llorca et al., 2001
) lid closure. To this end, the structure of ATP-bound Cpn-D386A was derived to 15 Å resolution by single particle Cryo-EM (, blue). Comparison of these structures with the ATP-free and ATP-bound states of Cpn-WT, derived to 10Å and 6Å respectively, revealed the conformational changes induced by ATP binding, distinguishing them from those induced by ATP-hydrolysis (; Fig. S2
). ATP incubation with Cpn-WT induces lid closure, yielding a symmetrically closed structure similar to that previously obtained for Cpn-WT with ATP-AlFx ( cyan and Fig. S5
)(Pereira et al., 2010
; Zhang et al., 2010
). In contrast, ATP binding to Cpn-D386A yielded an open structure that resembled the nucleotide free state ( for overlay and Fig. S2A, B
). Further addition of AlFx did not result in closure (not shown). Despite leaving the lid open, ATP binding induced a ~20Å constriction in the chaperonin opening ( and Fig. S2B
, 110 Å span versus 130 Å in the ATP-free state). Closer analysis of the conformational changes in a single subunit indicated that ATP binding induces an en masse
rigid body tilt of the entire intermediate and apical domains towards the ATP-binding equatorial domain (Fig. S2C
). We conclude that ATP-binding is insufficient to close the lid but triggers domain movements that lead, upon hydrolysis, to the closed state. These results are consistent with fluorescence experiments on the thermosome from Thermoplasma acidophilum,
indicating a rapid re-arrangement attributed to ATP binding, followed by a slower re-arrangement attributed to ATP hydrolysis and lid closure (Bigotti and Clarke, 2005
; Reissmann et al., 2007
The effects of ATP binding on the conformation of both the substrate and the lid were further examined using biochemical assays (). As described above, addition of ATP·AlFx to Cpn-WT stabilizes the closed state, locking the encapsulated substrate inside the chamber and leading to proteolytic protection of both the lids and the substrate (; and , lane 3; top panel for Cpn, bottom panel for 35
S-Rho for 35
S-Rho-Cpn-WT complex). Both ATP and ATP-AlFx induce a structurally similar closed state in Cpn-WT (e.g. , right panel) but the ATP-AlFx state displays a characteristic faster electrophoretic migration on native gels [; a similar effect is observed for TRiC/CCT; (Meyer et al 2003
)]. In contrast to Cpn-WT, incubation of Cpn-D386A with either ATP or ATP·AlFx failed to produce the signature mobility shift (). Furthermore, both the lid and the substrate remained in a largely unstructured, protease sensitive state upon ATP binding (, lanes 5, 6, 7), consistent with the result that ATP binding leaves Cpn in an open state (). Importantly, the lid also remains open under conditions where only one ring binds nucleotide [0.2 mM (Reissmann et al., 2007
)] or if Cpn-WT is incubated with the non-hydrolyzable ATP analogues AMPPNP or ATPγS (at either 0.2 mM or 1 mM; data not shown), further supporting the conclusion that ATP binding to either one ring or both does not suffice to close the lid.
ATP hydrolysis triggers substrate release from the chaperonin binding site
Lid closure and substrate encapsulation are essential for folding substrates such as actin for TRiC (Meyer et al., 2003
) and rhodanese (Reissmann et al., 2007
) and MDH for Cpn (Fig. S3A, B
). We next examined whether lid closure modulates the interaction of the substrate with the chamber. The “mechanical force” model proposes that the chaperonin does not release the substrate proteins into the closed cavity; in this scenario the chaperonin-substrate interaction persists in the closed state leading to the mechanical remodeling of the substrate conformation (, left) (Llorca et al., 2001
). Alternatively, ATP hydrolysis could promote substrate release into the closed chamber (, left). Since monitoring the substrate-chaperonin interaction inside the closed chamber is complicated by the presence of the lid, we exploited the previously characterized Cpn-Δlid variant that lacks the entire lid-forming segments (Pereira et al., 2010
; Reissmann et al., 2007
; Zhang et al., 2010
). Importantly, Cpn-Δlid achieves the same ATP-induced “closed” conformation as Cpn-WT (Zhang et al., 2010
) and its ATPase activity and substrate binding ability are unaffected (Reissmann et al., 2007
). These features of Cpn-Δlid allowed us to distinguish between the above models (, right panels). Thus, the model that proposes that the polypeptide remains associated with the chaperonin throughout the ATPase cycle predicts that the substrate will remain bound to Cpn-Δlid upon addition of ATP or ATP·AlFx (, ‘Δlid’ right). In contrast, if ATP weakens the chaperonin-substrate interaction, the absence of the lid will allow the polypeptide to diffuse away from the chaperonin (, ‘Δlid’ right). Of note, Cpn-Δlid cannot promote folding of substrates such as rhodanese and MDH (Fig. S3A, B
) and (Reissmann et al., 2007
)), thus substrate release from Cpn-Δlid cannot be ascribed to completion of folding.
ATP hydrolysis triggers substrate release from group II chaperonins
Purified 35S-rhodanese·Cpn complexes were incubated in the presence or absence of ATP for 10 min and analyzed using native gels followed by autoradiography (). Cpn-WT co-migrates with the substrate under both conditions (, WT), as expected given that 35S-rhodanese is encapsulated in the closed complex (See ). The small ATP-induced reduction in bound substrate is presumably due to loss through ATPase cycling and/or folding (see below, ). Strikingly, incubation of Cpn-Δlid with ATP led to a dramatic reduction in the amount of Cpn-bound rhodanese (, Δlid). This ATP-dependent loss of rhodanese required ATP hydrolysis, as it was not observed when the Cpn-Δlid also carried the D386A mutation (, Δlid/D386A). Similar results were obtained for other Cpn-bound polypeptides, including MDH (data not shown) and actin (see below ).
Structural basis of ATP-induced substrate release in group II chaperonins
The ATP-induced reduction in Cpn-substrate affinity was further evinced through the use of a “Trap”, a modified GroEL that scavenges non-native polypeptides () (Frydman and Hartl, 1996
). Trap will not bind to folded rhodanese but will bind to non-native polypeptides once they are released from the Cpn (Frydman and Hartl, 1996
) (, see Trap lane). For all Cpn variants tested, little or no 35
S-rhodanese was captured by the Trap in the absence of ATP, suggesting that rhodanese binds stably to all nucleotide free Cpn variants, and cannot be displaced by the Trap (, -ATP). Addition of ATP to Cpn-WT allowed a fraction of rhodanese to bind to the more rapidly migrating Trap (, WT+ATP). Comparing the WT incubations in the presence and absence of Trap (i.e. ) suggests that during normal ATP cycling a fraction of the substrate is released in a non-native form that rebinds to the chaperonin for another round of folding. This non-native polypeptide is captured by the Trap, which thus prevents Cpn rebinding and interrupts the cycle. Importantly, addition of ATP to Cpn-Δlid-35
S-rhodanese caused a near complete transfer of the bound polypeptide to the Trap (, Δlid), indicating that ATP induces substrate release from the chaperonin. Furthermore, no increase in substrate transfer to the Trap was observed upon ATP addition to Cpn-Δlid D386A (, Δlid/D386A) indicating that substrate dissociation from Cpn requires ATP hydrolysis.
The experiments above show that ATP hydrolysis has a function that is completely lid-independent, namely, to release the substrate from the chaperonin binding sites. We next employed ATP·AlFx, which mimics the trigonal-bipyramidal transition state of ATP hydrolysis (Meyer et al., 2003
) (). As with Cpn-WT (Fig. S1A
), the addition of AlFx to Cpn-Δlid immediately arrests its ATPase activity, suggesting that inhibition of ATP hydrolysis and trapping of the closed state occurs after a single cycle (Fig. S3C
). Whereas incubation of Cpn-WT-35
S-rhodanese with ATP·AlFx closes the chamber and encapsulates the substrate (; ), the substrate remains protease sensitive following incubation of Cpn-Δlid-35
S-rhodanese with ATP·AlFx (). Native gel analysis showed that Cpn-Δlid with ATP·AlFx undergoes the same signature shift as Cpn-WT, consistent with structural analyses showing both Cpns adopt the same closed conformation upon incubation with ATP·AlFx (Pereira et al., 2010
; Zhang et al., 2010
). ATP·AlFx induced a complete release of a broad panel of polypeptides ( for 35
S-rhodanese; Fig. S3D-G
for other substrates), indicating that ATP hydrolysis blocks general access to the substrate binding sites. The same conclusion was reached using size exclusion chromatography of purified Cpn-35
S-rhodanese complexes incubated in the presence or absence of ATP·AlFx and analyzed on a Bio-Sil SEC-400-5 column (Fig. S3H
). This experiment also indicated that ATP·AlFx induces full substrate release from Cpn-Δlid.
The effect of nucleotide hydrolysis on Cpn-substrate interactions was further examined using rhodanese carrying the environmentally sensitive fluorescent moiety Nile Red (Kim et al., 2005
) (herein NR-Rho; ). In free solution, NR-Rho exhibits a low fluorescence emission spectrum characteristic of an aqueous, polar environment, with a maximum at ~650nm (, gray trace). However, binding to Cpn caused a fluorescence intensity increase, as well as a blue shift of the maximal intensity to ~630nm (, red trace for Cpn-Δlid; similar results obtained for Cpn-WT, data not shown). This change in fluorescence upon Cpn-binding is diagnostic for rhodanese occupying a more hydrophobic environment (Kim et al., 2005
). We used the maximal fluorescence at 630 nm to monitor the effect of nucleotides on the substrate-chaperonin interaction. The Cpn-Δlid-NR-Rho fluorescence signal remained stable in the absence of nucleotide (; red traces). Addition of ATP produced a rapid decay in fluorescence intensity (, ‘+ATP’, blue trace). This supports our previous conclusion that ATP cycling by Cpn leads to substrate release. Addition of ATP·AlFx yielded similar results (, ‘+ATP·AlFx’, cyan trace), supporting the idea that the ATP hydrolysis transition state induces substrate release. We conclude that ATP hydrolysis has a dual function within the chaperonin cycle; it promotes lid closure () and also triggers substrate release from the chaperonin binding sites (). Strikingly, the latter function is not dependent on the presence of a lid.
The chaperonin substrate binding sites are unavailable in the closed state
A simple model explaining our results is that the ATP-induced Cpn conformation no longer exposes the substrate binding sites. We tested this model using an order of addition experiment (). In the control condition (, Ctrl), substrate was added to the open, apo-Cpn, which exposes the substrate binding sites. The second condition added the substrate first, prior to incubation with ATP·AlFx (); this condition measured the extent of ATP·AlFx-induced substrate release. In the third condition, we incubated with ATP·AlFx first, and then added substrate to the chaperonin (); this measured the ability of an ATP·AlFx-preincubated closed complex to bind substrate (). If the binding sites are still available in the closed state, we might expect some substrate binding for closed Cpn-Δlid in the A→S condition, which still retains a large opening allowing access to the central cavity (Pereira et al., 2010
; Zhang et al., 2010
). Since the pore size may restrict polypeptide entry to the cavity, and may sterically interfere with substrate binding we used both rhodanese () and a small 12 mer peptide substrate (herein PepB; Fig. S3G
and ). The small peptide substrate should be able to freely diffuse inside the closed chamber in the Cpn-Δlid.
The substrate binding sites are unavailable in the closed Cpn state
In the absence of nucleotide, both substrates bound to Cpn-WT and Cpn-Δlid (Ctrl; and Fig. S3G
). As expected, addition of ATP·AlFx to the Cpn-substrate complex () promoted substrate encapsulation for Cpn-WT (WT S→A; ) and substrate release for Cpn-Δlid (Δlid S→A; ). In the case of A→S, closing the Cpn-WT chamber with ATP·AlFx precluded substrate binding, thus the closed lid blocks access to the central cavity (WT A→S; ; scheme in ). For Cpn-Δlid, substrate should bind the chaperonin in the A→S condition provided that the binding sites are still available in the closed conformation. This was not the case; instead the ATP-AlFx preincubated Cpn-Δlid was unable to bind either 35
S-rhodanese or the small PepB (; Cpn-Δlid compare S→A and A→S). Thus, the ATP·AlFx state of Cpn-Δlid no longer exposes the substrate binding sites. Given that the ATP·AlFx conformations of Cpn-WT and Cpn-Δlid are virtually identical (Pereira et al., 2010
; Zhang et al., 2010
), these experiments show that the substrate binding sites are no longer available upon ATP-hydrolysis.
Mechanism of ATP-induced substrate release
What is the possible mechanism for substrate release in group II chaperonins? A structural analogy with the distantly related bacterial group I chaperonins, e.g. GroEL, is not possible, given that they use a detachable lid, GroES, which upon ATP binding, both caps the cavity and displaces the substrate. In contrast, we show that substrate release in group II chaperonins is lid independent and requires ATP-hydrolysis.
Closer examination of Cpn structures in the open and closed states led to a hypothesis for how ATP hydrolysis induces substrate eviction () (Pereira et al., 2010
; Zhang et al., 2010
). In the open state, the substrate binding region around helix 11 is well exposed (, pink in left panel) (Spiess et al., 2006
) leaving ample space to accommodate the bound substrate. In contrast, the closed state brings the apical domains from adjacent subunits into close proximity (). Closure causes helix 11 to form a tightly packed interface with a loop spanning residues 327-331 in its neighboring subunit (, cyan). Such lateral intra-ring contacts might displace the substrate from its binding site, causing the 327-331 region to acts as a ‘r
oop for the s
ubstrate’ (herein rls
loop). To disrupt this lateral interface, we made Ala substitutions in four loop residues making key contacts with helix 11 yielding the Cpn-rls variants (, T327A, N328A, K330A, and D331A). To better understand the role of the rls
loop within the chaperonin cycle, we used Cryo-EM to obtain a detailed structural characterization of the conformation of both Cpn-rls and Cpn-rls-Δlid in the presence or absence of ATP or ATP·AlFx (Fig. S4A
for Fourier Shell Correlation analysis of resolutions; S4B
for Cpn-rls-Δlid and S5
for Cpn-rls). The rls
chaperonins achieve essentially the same closed state as the wild type counterparts (; Fig. S4B
for Cpn-rls-Δlid; Fig. S5
for Cpn-rls). Consistent with their ability to reach a closed state, the Cpn-rls mutants were competent for ATP binding and hydrolysis (data not shown).
We initially focused on Cpn-rls-Δlid, since the absence of a lid simplifies analysis of substrate release (). Cpn-WT and Cpn-Δlid served as controls. In the absence of ATP, all chaperonins bound rhodanese and actin efficiently, as shown by native gel analysis (). Strikingly, Cpn-rls-Δlid was incapable of releasing either substrate in the presence of ATP, unlike Cpn-Δlid (, compare lane 6 to lane 4). This suggests that the lateral contacts between helix 11 and the rls loop 327-331 are indeed important for releasing the substrate upon ATP-hydrolysis.
We next examined the effect of the transition state mimic ATP·AlFx (). Native gel analysis showed that Cpn-rls-Δlid adopts the same fast migrating conformation observed for Cpn-Δlid and Cpn-WT (, Coomassie blue panel) consistent with the Cryo-EM analysis. Surprisingly, unlike ATP, incubation with ATP·AlFx caused Cpn-rls-Δlid to efficiently release all the substrates tested ( for rhodanese and actin). This observation was striking given the apparent similarity between the ATP and ATP·AlFx structures of Cpn-rls variants ( and Fig. S4B
). It thus appears that, in the rls
mutant the conformation promoting substrate release cannot be stably populated by ATP alone whereas ATP·AlFx can stabilize this state and evict the substrate.
Fluorescence spectroscopy provided independent support for the above conclusions. As for Cpn-Δlid, NR-Rho bound to Cpn-rls-Δlid had an emission spectrum characteristic of a hydrophobic environment (data not shown). In contrast to Cpn-Δlid (, blue trace), ATP incubation did not cause any appreciable change in the fluorescence of NR-Rho bound to Cpn-rls-Δlid (, blue trace), indicating that ATP alone cannot release the bound substrate. However, when AlFx was added to an ongoing incubation of NR-Rho·Cpn-rls-Δlid with ATP, the fluorescence rapidly dropped, indicating substrate release from the chaperonin (, cyan trace). A similar reduction in fluorescence intensity was observed if ATP and AlFx were added together, but was absent if only AlFx was added (not shown). We conclude that weakening the lateral contacts between helix 11 and its neighboring subunit prevents substrate release, even though Cpn-rls-Δlid can hydrolyze ATP and achieve the closed state. However, stabilizing the post-hydrolysis state by addition of AlFx populates the conformation that evicts the substrate.
Structural basis of substrate release and encapsulation
We next examined the effect of the rls
mutations in the Cpn with the intact lid (herein Cpn-rls, ). Detailed structural analyses of the ATP and ATP·AlFx induced states in both Cpn-WT and Cpn-rls revealed interesting differences between these chaperonins (; Fig. S5
). Single particle Cryo-EM reconstructions were obtained to 4–6Å for both chaperonins in the presence of either ATP or ATP·AlFx (Cpn-WT-ATP 6Å, Cpn-WT-ATP·AlFx 4.3Å, Cpn-rls-ATP 5Å, Cpn-rls- ATP·AlFx 6Å, and Fig. S5
). Models of these structures were then built by flexible fitting into the density map with Rosetta ( and Fig. S5A
; see Fig. S5B
for goodness of fit between model and density map) (DiMaio et al., 2009
). Cpn-rls achieved a closed state with either ATP or ATP·AlFx, similar to those obtained with Cpn-WT. Notably, superimposition of the structures of Cpn-WT and Cpn-rls in the different nucleotide states revealed variations in their structure, particularly in the region corresponding to the apical domains (; i.
top view of superimposed EM density maps). These differences were also evident when comparing the apical domain regions in the respective chaperonin models (; ii.
detail of apical domain and lid for a subunit within the complex). The ATP (magenta) and ATP·AlFx (blue) states of Cpn-WT were essentially identical (). Thus, ATP·AlFx generates the same closed state observed under ATP-cycling conditions (e.g. ). Importantly, we observed a shift in the apical domain regions between the closed Cpn-rls states induced by ATP (yellow) and ATP·AlFx (cyan) (). Cpn-rls-ATP also exhibited noticeable differences with both closed WT structures (e.g. ). The apical domain protrusions in Cpn-rls-ATP are shifted clockwise, and the apical domains including the lid, are tilted up compared to the ATP·AlFx state, exhibiting significant variations in helix 11 (ii.
red arrow) and the rls
blue arrow). In contrast, the ATP·AlFx states of Cpn-WT and Cpn-rls were nearly identical (). These structural analyses demonstrate that even though Cpn-rls can close with ATP, impairment of the helix 11/loop 327-331 contacts results in aberrant intra-ring interactions between the apical domains. This is consistent with the inability of Cpn-rls-Δlid to release the substrate in the presence of ATP (See .). Furthermore, ATP·AlFx induces a closed conformation in Cpn-rls that is indistinguishable from the closed state of Cpn-WT with either ATP or ATP-AlFx. This is consistent with, and explains, the finding that ATP·AlFx leads to substrate release in Cpn-rls-Δlid (See .)
Substrate release into the central chamber is required for group II chaperonin mediated folding
Substrate release and encapsulation are required for productive folding
The identification of a mechanism that evicts the bound polypeptide upon closure allowed us to test the relevance of substrate release for the folding cycle. First, the ability of Cpn-rls to encapsulate a bound substrate was examined by native gel analysis () and PK digestion () as shown above for Cpn-WT. Incubation of the Cpn-rls with rhodanese yielded a binary complex that behaved exactly as that of Cpn-WT (). Protease digestion analysis indicated that, in the absence of nucleotide, the substrate binds in an unstructured conformation (, lane 2). Importantly, incubation with either ATP or ATP·AlFx led to proteolytic protection of both the chaperonin lid segments (, lanes 3, 4, top panel) and the bound 35S-rhodanese (, lanes 3, 4, bottom panel). Thus, both ATP and ATP·AlFx induce stable lid closure and fully encapsulate the substrate within the central chamber of Cpn-rls.
Rhodanese-chaperonin complexes were prepared for Cpn-rls and Cpn-WT, which served as a control (). As expected, addition of ATP or ATP-AlFx to the Cpn-WT complex induced rhodanese folding (, black traces). Strikingly, addition of ATP to the Cpn-rls complex failed to promote rhodanese folding, even though the substrate was encapsulated within the closed chamber (, green trace). We hypothesized that failure to fold stems from the failure to release the bound substrate into the central chamber. Therefore, we tested the effect of ATP·AlFx, which should promote substrate release (See ). Addition of ATP-AlFx to the Cpn-rls reaction caused efficient rhodanese folding (). These experiments indicate that lid closure and substrate encapsulation are by themselves, unable to promote substrate folding. Importantly, they demonstrate that substrate release into the central closed chamber is essential for productive folding by group II chaperonins.