VHL as a model substrate for GroEL/GroES
VHL normally interacts with the chaperones TRiC and Hsp70 in eukaryotic cells in order to form its active folded complex with two additional accessory proteins Elongin B and C 22–24, 24
. These interactions are mediated by two hydrophobic regions on VHL, named the Box1 (B1) and Box2 (B2) regions, which are necessary and essential for chaperonin binding 23, 25
. These regions are distant in the sequence but are located in adjacent regions in folded VHL () 26
. VHL can also interact with the prokaryotic chaperonin GroEL and can reach its properly folded state in the presence of Elongin B/C via ATP-driven interaction with GroEL/ES (see Fig. S1, Supplemental Material
). However, since the fully folded state of VHL is only formed in the presence of Elongin B/C, study of the complete folding process of VHL would require the assembly of multiple components and is not of primary interest here. Rather, we view VHL as a model poly(peptide) containing two hydrophobic regions B1 and B2 known to drive binding to chaperonins 25
, and concentrate on how the conformation of VHL changes during VHL-GroEL binding and during the formation of the VHL-GroEL/ES complex in the presence of nucleotides. In this way, specific information can be obtained about the conformational changes produced by the GroEL/ES system on a non-native substrate. This study thus provides information regarding the nonspecific, ubiquitous ability of the chaperonin to assist many substrates.
Figure 1 (a) The crystal structure of VHL in the VBC complex and the locations of Cys mutations and the inter-residue distances in this study. Hydrophobic binding sites are highlighted in red (Box1 or B1) and blue (Box2 or B2). (b) The table shows the sequence (more ...)
The structure of VHL after binding to GroEL is more compact
In order to characterize the conformational changes of VHL, we applied the well-known Förster Resonance Energy Transfer (FRET) method 27, 28
to probe changes in pairs of amino acids of VHL during interactions with GroEL, first in ensemble measurements and subsequently in the single-molecule regime. The fluorophores Alexa488 (AX488) and Texas Red (TR) were chosen as a FRET pair to label double-cysteine variants of VHL. This pair of dyes has a relatively small R0
value (~46.5 Å) 29
, which is necessary for monitoring a small protein like VHL. Six different double Cys variants were prepared and labeled with both AX488 and TR maleimide. The labeling positions and the β-carbon distances between each amino acid pair for VHL in its native complex with ElonginBC and the sequence separations are shown in . Importantly, the N and C positions report on the relatively polar termini of the protein, whereas the B1 and B2 positions report on the two hydrophobic core regions 25
shows the relative bulk solution FRET efficiencies (expressed as a proximity ratio F=IA
)) for six different variants of VHL in three different conditions; 6M GdnHCl, native buffer, and bound to GroEL in native buffer. As described in Materials and Methods, any unbound VHL that could lower the FRET efficiency was removed using a gel filtration column before each measurement. In the unfolding conditions of 6M GdnHCl, FRET efficiencies for all variants were the lowest compared to all the other conditions. As may be expected, the FRET value was roughly inversely proportional to the sequence separation between the labeling sites in the polypeptide chain (data not shown) 30, 31
. For example, the F value for the B1-B2 variant (36 aa separation) in 6M GdnHCl was highest (~0.50), while that of the N-C variant (122 aa separation) had the lowest value (~0.12) in the unfolded state. For VHL in native buffer (light gray columns in ), the degree of FRET was larger than for VHL in 6M GdnHCl for all six variants, suggestive of the collapse observed by others 31
. Finally, the FRET proximity ratio for all VHL variants was significantly increased when VHL was complexed with GroEL (dark gray columns in ). The N-B1 variant showed more than ~10 Å contraction after binding to GroEL (roughly extracted from F value increase from 0.39 to 0.72, see Materials and Methods for controls and equations used in distance estimates). These results suggest that simple binding of VHL to the chaperonin passively induces significant conformational changes in VHL, specifically compaction with respect to a random coil extended conformation.
Single-molecule FRET reveals position-dependent conformational changes in the substrate upon lid closure
Next, we measured FRET efficiencies for six VHL double variants with GroEL/ES at the single-molecule level. In order to immobilize individual molecules for long-term observation, we prepared biotin-labeled GroEL (C473Bio-EL) and bound it to a glass coverslip by biotin-Neutravidin interactions. A carefully prepared PEG-coated glass coverslip with a small amount of biotin bound to the surface was used to avoid any non-specific GroEL binding to the surface (see ). Several control experiments demonstrated that we observed the functional GroEL-VHL complex, rather than non-specifically bound forms of VHL protein on the surface. First, in the absence of Neutravidin, no Cy3-labeled C473Bio-EL test molecules, no fluorophore-labeled VHL, and no biotin-labeled GroEL in complex with VHL were observed stuck to the PEG-coated surface. Both the amount of Neutravidin and the incubation time for biotin-Neutravidin interaction were critical to observe any fluorescence from the VHL, and therefore non-specific binding of VHL to the surface was not occurring. Secondly, we measured the single-molecule emission polarization of the VHL-GroEL complex after immobilization on the glass coverslip (see ). Since we used a biotin molecule with a long linker between the biotin and the maleimide to label C473EL (Biotin-PEG-maleimide with MW 3810), this complex should be free to rotate after immobilization if there is no strong protein-surface interaction. AX488 dye molecules spun on an unpassivated glass coverslip showed a broad, heterogeneous polarization (P) distribution due to non-specific binding, while both AX488- and the TR-labeled VHL(N terminal labeled)-GroEL complexes bound to the prepared surface had narrow P distributions centered at P=0. All the other single mutants (B1, B2, and C) with either of the two different fluorophores (total 6 samples) also showed similar narrow distributions, demonstrating that GroEL-VHL complexes are free to rotate without any specific interaction with the surface on the 0.1 sec time scale of our measurements. (For fluorescence anisotropy measurements on the scale of the excited state lifetime, see Materials and Methods.) Single-molecule FRET measurements of VHL only in the absence of GroEL were attempted by immobilizing VHL in 1.5% agarose gel. However, due to the the additional heating/cooling process, the conformation of VHL in agarose gel was disturbed. In fact, we observed increased average FRET in these perturbed cases (data not shown).
Figure 2 (a) PEG-coated glass coverslips were prepared as described by Joo, et al. 43 (Bottom) A specific binding test was performed using biotin- and Cy3-labeled GroEL (C473Bio-EL). In the presence of Neutravidin (left panel), Cy3-labeled C473Bio-EL could be (more ...)
shows the time-averaged single-molecule FRET efficiency distributions for four of the double variants denoted by the two labeling positions (N-B1, B1-B2, B2-C, and N-C). The upper row shows the distributions for unfolded VHL complexed with GroEL only, the lid-open state. The dashed lines locate the average FRET values for each case. In the lower row of , the lid-closed complex has been formed by incubation of labeled VHL with GroEL, GroES, and ATP/AlFx 21, 32, 33
, a transition state mimic of ATP hydrolysis. It has been shown that addition of ATP or ADP with AlFx stopped the ATPase cycle of GroEL, producing a symmetric complex (bullet form) of GroEL/ES. In this case, both GroEL cavities are closed by GroES. For two variants, N-B1 and B2-C, the mean of the distributions did not shift appreciably (as was the case for two other variants, N-B2 and B1-C, not shown), although the distribution showed a greater width extending to lower FRET values for the N-B1 variant. However, significant shifts appeared for the B1-B2 and N-C variants. While the degree of FRET shifted to higher values for the B1-B2 case suggesting a decrease in the B1-B2 distance, the distribution shifted to lower values for the N-C variant suggesting an increase in the N-C distance. This suggests that confinement of a non-native polypeptide within the GroEL/ES cavity leads to compaction of the hydrophobic core but expansion of the overall conformation of the polypeptide.
Figure 3 Single-molecule FRET efficiency distributions measured with individual VHL-GroEL molecules. After preparation of the biotin-PEG coated surface, complexes were prepared by diluting unfolded VHL into buffer A with C473Biotin-EL for 20 min, incubated on (more ...)
One of the advantages of single-molecule measurements is that conformational heterogeneity can be estimated by the width of the distribution 28
. . Importantly, the distributions for all cases were much wider than can be explained by experimental noise arising from the number of photons detected. The standard deviation of the shot noise distribution, calculated from the method in Gopich et al. 34
(see Materials and Methods) was about 0.005 for all cases in , which is ~20 times smaller than the observed widths. Therefore, it is likely that a large degree of conformational heterogeneity is present in VHL bound to GroEL in the open and closed states of the chaperonin. All six variants of lid open state (upper row in ) showed similarly shaped efficiency distributions with similar widths (standard deviation of about 0.09, i.e. 3.0 Å in distance, on average). In the case of the N-C variant and the N-B1 variant, a further increase in the distribution width could be observed in the lid-closed state
Dynamic single-molecule measurements upon ATP addition reveal the stochastic or probabilistic nature of GroEL/ES substrate interactions
Dynamic (time-dependent) single-molecule FRET measurements were performed to follow the interaction between VHL and GroEL upon ATP and GroES addition in real time. Only the B1-B2 and N-C variants were considered here, because these two pairs sample the two extremes: the overall protein structure (N-C), and the highly hydrophobic regions providing the likely chaperonin binding region (B1-B2). Before interpreting any ATP dependent changes, the effect of photobleaching of the fluorophores must be addressed. As usual, photobleaching is determined on the single-molecule level by the disappearance of a single-molecule spot; however, here we must distinguish between photobleaching and spot disappearance due to release of the labeled VHL from the surface-attached chaperonin. First, we measured the survival distribution of AX488 bound to the actual protein by recording continuous images of singly-labeled AX488-VHL bound to surface-attached GroEL with 0.1 sec integration time and found a characteristic τbleach
~4.1 sec (). To follow conformational changes induced by ATP addition over longer times before fluorophore photobleaching, time-lapse imaging was then implemented using a 0.1 sec exposure and a 0.9 sec dark interval. This had the effect of extending the photobleaching time to 41 sec (41 cycles of time-lapse), much longer than the estimated ATP hydrolysis time of GroEL (~15 sec 10
). GroEL has two distinct rings, and the GroEL capped with GroES is called the cis
ring, while the non-capped ring is called the trans
ring. As has been shown, upon ATP hydrolysis to ADP and binding of an additional ATP to the trans
ring, the GroES-GroEL complex is destabilized, leading to both GroES and substrate release from the cis
. Following substrate release, the fluorescent spot would then disappear as the fluorescent substrate diffuses away from the surface-attached chaperonin. Using this experimental strategy, the fluorescence from VHL (complexed to GroEL) on the surface was monitored for t = 15 sec (15 cycles of time-lapse) prior to nucleotide addition, and then the subsequent effects of ATP-induced interaction changes were investigated. This regimen allowed us to have enough data points both before and after ATP addition to observe time-dependent changes with 1 sec time resolution.
Figure 4 (a) The number of surviving molecules distribution for single molecules of AX488 VHL (N) with GroEL. Continuous illumination with 0.1 sec integration time was used to measure the number of molecules remaining after specific illumination times to assess (more ...)
When ATP was added to the AX488-VHL-GroEL complex on the surface, roughly 10% of the AX488-VHL single-molecule spots quickly disappeared within 5 sec, compared to the case with no ATP addition. A further decrease (~40%) in the number of bound single molecules was observed when GroES was present in the solution before ATP addition ( inset), and this was also observed with all of the other mutants. The fast disappearance of a fraction of the single-molecule spots upon ATP addition was likely due to release of VHL into the solution during cis
ATP/GroES complex formation 10, 35
, rather than encapsulation into the cis
cavity. A similar partial “loss” of substrate during encapsulation was also reported by Weissman et al. where they observed that only 30–40% of rhodanese substrates were encapsulated after lid closure 36
. Under time-lapse conditions (where the effective photobleaching/disappearance time is 41 sec on average), the addition of ATP in the presence of GroES caused the disappearance time constant to decrease to 30 sec (inset of ), which reflects the ATP hydrolysis and complex disassembly time scale.
The fraction of single-molecule spots which did not disappear after ATP/ES addition remained associated with the chaperonin and likely encapsulated, and the FRET signal for these complexes was analyzed. shows example traces for a B1-B2 variant (left side) and a N-C variant (right side). Note that in each case, the donor and acceptor signals (upper panels) and the FRET signal (lower panels) were monitored before and after the moment of ATP addition (solid arrow). For the B1-B2 case (left side), the FRET value increased upon closure of the cavity, followed by photobleaching at 40 sec total elapsed time. This molecule is therefore an example of the average behavior of : closure of the cavity induced a compaction between the B1 and B2 hydrophobic regions. For the N-C case (right side) the FRET value decreased upon ATP-induced closure of the cavity, followed by photobleaching at 37 sec total elapsed time. This molecule is also an example of the average behavior of : closure of the cavity induced an expansion between the N and C hydrophobic regions.
Importantly, because the actual FRET signal is followed for each single molecule before and after ATP addition, the conformation of VHL simply bound to GroEL can be directly correlated with its conformation after complex formation. shows this by displaying the distance (extracted from properly corrected efficiency values) between two probes before and after ATP addition in a scatter plot for many single-molecule traces like (101 molecules for B1-B2, 80 molecules for N-C mutant were observed). The data fall into two regions separated by the line representing no distance change. Due to the large fluctuation of the FRET signals, we applied Welch’s t-test to each molecule and present only those transitions where the difference in mean efficiency before/after ATP addition was significant using a 95% confidence interval; this criterion selected about 40% of all the molecules studied. As expected from the average efficiency changes in , 84% of the B1-B2 molecules showed a distance-decrease transition, and 69% of the N-C molecules showed the opposite change, a distance increase. Of note, a minority of the molecules showed the opposite behavior. For B1-B2, ATP addition was followed by a distance increase for 16% of the molecules (or drop in FRET), and for N-C, ATP addition was followed by a distance decrease for 31% of the molecules (or increase in FRET). Therefore, the structural changes upon ATP binding appear be heterogeneous and probabilistic.
Figure 5 Characterization of individual FRET transitions of (a) single B1-B2 and (b) single N-C molecules upon ATP addition. The calculated distances extracted from true FRET efficiency E before and after ATP addition were used as x- and y- coordinates. Only transitions (more ...)
We also considered the possibility that the conformation of the VHL still bound to the GroEL/ES complex might change with time after the moment of complex assembly induced by ATP addition (data not shown). Taking into account the noise in the FRET signals, no evidence was found for further significant conformational changes after cavity closure.
A picture of the action of GroEL/ES on a bound polypeptide
Despite intense study, it is still unclear in detail how GroEL affects substrate conformation to facilitate folding. Several previous studies examined the conformational changes-either compaction or expansion of GroEL substrates before and after GroES lid closure at the bulk level 37–39
. Our work provides complementary information on how GroEL binding affects the conformation of a bound polypeptide both at the bulk and single-molecule level, as summarized in . Upon binding to GroEL, the protein becomes more compact relative to its conformation in denaturant, with a shortening of distances between all regions in the protein. For example, the estimated distance between N-C termini is about 45 Å. Following substrate encapsulation upon ATP and GroES addition, the hydrophobic core becomes more compact on average, whereas the N-C distance increases on average. This is not due simply to folding, since VHL can only reach the final native state following release from the cavity and binding to its oligomeric partners ElonginBC.
This work finds significant heterogeneity, both in the average FRET distributions () and in the time-dependent changes in FRET induced by EL/ES cavity assembly induced by ATP ( and ). Recent single-molecule studies by both Hillger et al. 19
and Sharma et al. 31
also described heterogeneous populations of rhodanese and mutant MBP upon binding to GroEL as well as further conformational changes upon ATP addition. Both rhodanese and MBP are present with GroEL/ES in prokaryotic cells, while VHL did not co-evolve with the bacterial proteostasis network. This might explain some fraction of the heterogeneity we observe, but it is unlikely to explain all of the heterogeneity. For the case of VHL, the majority behavior shows an ATP- and GroES induced compaction for B1-B2 and expansion for N-C, but there is a significant minority of the single molecules with the opposite behavior. These results may be rationalized with a simple physical picture. In the open, unliganded state, GroEL exposes hydrophobic binding sites in the apical domains of each subunit of the ring 8
. It is reasonable that the two hydrophobic regions of VHL (B1 and B2) take the lead in determining the interaction with GroEL, as with TRiC 25
. Each of the seven subunits in one GroEL ring has identical hydrophobic patches, each of which can attract exposed B1 or B2 for binding. One then might guess that the B1 and B2 regions could bind to adjacent GroEL subunits, or to subunits one or two positions apart. For example, the distance between two adjacent binding sites calculated from the crystal structure is about 23 Å. Referring to the FRET efficiency distribution, it is likely that most of the B1-B2 cases bind to the subunits two monomers apart. But this is only the average case - for some molecules B1 and B2 would be brought closer together as the apical domains move, but for other single molecules, especially those whose B1-B2 positions are too closely located at the starting point after binding to GroEL, B1 and B2 would be moved further apart from their starting separation, showing expansion. On the other hand, the N- and C- termini, not being hydrophobic, would not be expected to bind strongly to GroEL. The relative conformational expansion between these two locations may be a secondary result of the changes imposed by B1 and B2 binding or may result from the changed, more hydrophilic environment of the closed GroEL/GroES cavity. It is also possible that solvent effects account for the conformational rearrangement observed upon GroES binding 40–42
Clearly, more work will be required to clarify the basis of the observed substrate remodeling and its probabilistic behavior from molecule to molecule. Taken together, these results clarify the idea that the average action of GroEL/ES on exposed hydrophobic regions is to move them closer together, an effect that would be expected to ultimately promote folding. For those molecules which did not bind in a favorable orientation, the action of GroEL/ES in promoting folding is thwarted, and therefore, additional rounds of binding and hydrolysis would be expected to stochastically sample the other possibilities available. This stochastic nature has been observed directly (for different molecules) in this single-molecule experiment. In principle, the stochastic behavior may reflect the ability of GroEL/ES to assist in exploration of a larger range of conformational space for the substrate in the attempt to fold non-evolved substrates.