|Home | About | Journals | Submit | Contact Us | Français|
Juxtaposed to either or both ends of the proteasome core particle (CP) can exist a 19S regulatory particle (RP) that recognizes and prepares ubiquitinated proteins for proteolysis. RP triphosphatase proteins (Rpt1-Rpt6), which are critical for substrate translocation into the CP, bind chaperone-like proteins (Hsm3, Nas2, Nas6, Rpn14) implicated in RP assembly. We used NMR and other biophysical methods to reveal that S. cerevisiae Rpt6’s C-terminal domain undergoes dynamic helix-coil transitions enabled by helix-destabilizing glycines within its two most C-terminal α-helices. Rpn14 binds selectively to Rpt6’s 4-helix bundle, with surprisingly high affinity. Loss of Rpt6’s partially unfolded state by glycine substitution (Rpt6 G360,387A) disrupts holoenzyme formation in vitro, an effect enhanced by Rpn14. S. cerevisiae lacking Rpn14 and with Rpt6 G360,387A incorporated demonstrate hallmarks of defective proteasome assembly and synthetic growth defects. Rpt4 and Rpt5 exhibit similar exchange, suggesting that conserved structural heterogeneity among Rpt proteins may facilitate RP-CP assembly.
The major process for selective protein degradation in eukaryotes is through the ubiquitin-proteasome pathway, in which substrate ubiquitination signals for proteolysis by the 26S proteasome (Hershko and Ciechanover, 1998). Dysfunctions of this pathway as well as its hijacking by pathogens are associated with numerous human diseases (Schwartz and Ciechanover, 2009), making it a major pharmaceutical target. Proteasome inhibitor Bortezomib is used to treat multiple myeloma, relapsed mantle cell lymphoma, and acute allograft rejection, while more recently developed inhibitors are in clinical trials, as reviewed in Kisselev et al., 2012.
Substrate proteolysis occurs within the 20S proteasome core particle (CP), which is capped at either or both ends with a 19S regulatory particle (RP) that recognizes ubiquitinated substrates and prepares them for translocation into the CP (Finley, 2009; Glickman et al., 1998). The RP is a multiprotein complex that contains several subunits with known function, including ubiquitin receptors Rpn10/S5a (Deveraux et al., 1994) and Rpn13 (Husnjak et al., 2008; Schreiner et al., 2008); deubiquitinating enzymes Rpn11 (Maytal-Kivity et al., 2002; Verma et al., 2002; Yao and Cohen, 2002), Ubp6/Usp14 (Leggett et al., 2002), and Uch37/UCHL5 (Lam et al., 1997); and a heterohexamic ring of six AAA+ ATPase proteins (Rpt1-Rpt6) that abuts the α-subunit ring of the CP and opens its channel for substrate entry (Rabl et al., 2008; Smith et al., 2007). The RP can be divided into a lid and base, which includes Rpt1-6, two large scaffolding proteins, Rpn1 and Rpn2, and the ubiquitin receptors, Rpn10 and Rpn13.
The best RP structural characterization has been provided by cryo-electron microscopy studies at 7.4 Å for S. cerevisiae (Beck et al., 2012) and 8.4 Å for S. pombe (Bohn et al., 2010; Lasker et al., 2012). The topological arrangement of RP subunits and of the RP-CP interface was modeled into the 3-dimensional reconstructions by incorporating intersubunit crosslinking approaches, including a study that revealed an Rpt1-Rpt2-Rpt6-Rpt3-Rpt4-Rpt5 arrangement (Tomko et al., 2010), and crystallographic structures of RP subunit orthologs, especially the archaeal AAA-ATPase regulator PAN (Djuranovic et al., 2009; Zhang et al., 2009a).
Crystal structures of PA26 (11S activator) complexed with archaeal (Förster et al., 2005) or Saccharomyces cerevisiae (Whitby et al., 2000) CP demonstrate C-terminal extensions docking into CP α-α subunit interfaces to form a salt bridge deep within the cavity (Förster et al., 2005). This interaction was also observed in an archaeal CP:PAN complex (Yu et al., 2010). Like PAN, Rpt2, Rpt3, and Rpt5 terminate in a conserved Hb-Y-X (hydrophobic-tyrosine-any amino acid) motif, and C-terminal peptides of PAN, Rpt2 and Rpt5 that preserve this motif are sufficient for CP gate opening (Smith et al., 2007). Rpt6 has a Hb-Y-X variant in which tyrosine is replaced by phenylalanine in yeast and tryptophan in humans. A recent study found that synthetic Rpt6 C-terminal peptides bind with high specificity to the α2-α3 pocket whereas the other Rpts are more promiscuous (Park et al., 2013). This finding implicates Rpt6 as having a unique role in templating RP onto CP during assembly and since Rpt6’s tail is not docked into an α pocket in mature proteasome, also suggests that the Rpt:CP interface is reconfigured upon lid addition (Park et al., 2013). Modification of the C-terminal end of Rpt4 and Rpt6 causes a biosynthetic proteasome assembly defect (Park et al., 2009) and Rpt6’s C-terminal tail is critical for base-CP complex formation (Park et al., 2013).
Three Rpt heterodimeric RP assembly intermediates (Rpt1:Rpt2, Rpt3:Rpt6, and Rpt4:Rpt5) have been identified (Funakoshi et al., 2009; Kaneko et al., 2009; Park et al., 2009; Roelofs et al., 2009; Saeki et al., 2009; Thompson et al., 2009), along with four chaperone-like proteins that interact transiently with C-terminal domains of Rpt proteins. These chaperone:Rpt complexes include Nas2 (p27):Rpt5, Nas6 (p28 or gankyrin):Rpt3, Hsm3 (S5b):Rpt1, and Rpn14 (PAAF1):Rpt6 (Funakoshi et al., 2009; Kaneko et al., 2009; Park et al., 2009; Roelofs et al., 2009; Saeki et al., 2009; Le Tallec et al., 2009). Knockdown of the chaperones causes RP assembly defects in HEK293T cells (Kaneko et al., 2009) and in yeast (Funakoshi et al., 2009; Roelofs et al., 2009; Saeki et al., 2009; Le Tallec et al., 2009).
X-ray crystallography has demonstrated significant structural diversity among Rpn14, Nas6 and Hsm3. Rpn14 forms a WD40 motif with a unique N-terminal domain (Kim et al., 2010) whereas Nas6 and Hsm3 contain a 7-fold ankyrin repeat (Nakamura et al., 2007) and eleven HEAT repeats, respectively (Barrault et al., 2012; Takagi et al., 2012). The Nas6 and Hsm3 structures were solved complexed with their Rpt binding partner (Rpt3 and Rpt1 respectively), which use the same side of a 4-helix bundle to bind their chaperone (Nakamura et al., 2007; Takagi et al., 2012). Hsm3 was recently found to contact the ATPase domain of Rpt2 as well and may thus act as a scaffold that bridges Rpt2 to Rpt1 (Barrault et al., 2012).
Passage of substrates from RP to CP is expected to require protein dynamics and multiple methods suggest that Rpt proteins govern motion at the RP-CP interface. A 3D variance map calculated from electron microscopy data for S. pombe 26S proteasome revealed the region surrounding the Rpt proteins to exhibit significantly higher conformational variation than the lid and CP (Bohn et al., 2010). Crosslinking studies have demonstrated asymmetry at the RP-CP interface in yeast, as on one side, Rpt2, Rpt6 and Rpt3 contact α4, α3 and α2, respectively, while at the other side, Rpt4, Rpt5, and Rpt1 interact dynamically with multiple α subunits (Tian et al., 2011). Furthermore, the N-terminal domain of Rpt proteins may undergo ATP-induced conformational changes (Horwitz et al., 2007). NMR experiments recently demonstrated that the CP inter-converts between multiple conformations in solution, and that this distribution is shifted to accommodate PA26 or an inhibitor (Ruschak and Kay, 2012).
In this study, we use NMR spectroscopy to investigate the structure and dynamics of the Rpt6 C-terminal domain (Rpt6-C) and its interaction with assembly chaperone Rpn14. Our data demonstrate that Rpt6-C undergoes conformational exchange that is abrogated by alanine substitution of two glycines located within its two most C-terminal α-helices. The exchange region is centered among Rpt6’s expected CP-facing, ATP-binding and Rpn14-binding surfaces, and we present evidence for a model in which Rpt6-C dynamics plays a role in its interaction with Rpn14 and proteasome assembly.
In an effort to structurally characterize Rpt6 and its interaction with Rpn14, we acquired a 1H, 15N Heteronuclear Single Quantum Coherence (HSQC) experiment on 15N-labeled Saccharomyces cerevisiae Rpt6-C (Figure 1A), which spans P318-K405. A protein with one conformational state displays in the resulting spectrum one backbone amide signal for each amino acid except proline. Rpt6-C however displayed a greater than expected number of NMR signals, indicating conformational heterogeneity. Its six glycines produced eleven signals for example (Figure 1B). Therefore, we generated Rpt6-C protein samples in which lysine, alanine, leucine, or valine were selectively 15N-labeled. These samples also displayed greater than expected numbers of NMR signals in HSQC spectra (Figure 1B and S1A). This heterogeneity was observed even at low protein concentrations (Figure S1B for a spectrum at 40 μM).
By using deuteration and state-of-the-art triple resonance and exchange experiments, 32 Rpt6-C amino acids (36% of the protein sequence) were assigned two sets of NMR signals (Figure 1C for A385-K388). A subset of these had sufficient signal-to-noise to demonstrate two Cα signals of unequal intensity in the HNCA and HNCOCA experiments with the weaker signal of each set corresponding to the stronger one of the other; this effect is illustrated for V386 and K388 (Figure 1C). To assess whether the two observable Rpt6-C states undergo dynamic exchange in solution, we acquired 1H, 15N ZZ-exchange spectroscopy (EXSY) experiments (Montelione and Wagner, 1989). We used variable mixing times to determine an optimal value of 80 ms (Figure S1C) at which we observed EXSY crosspeaks for all 30 Rpt6-C amino acids identified to have two sets of amide signals (Figure 1D and 1E). We also observed EXSY crosspeaks for M350 and F382 (Figure 1D), which appeared to have second amide signals that overlap with Y368 and K392, respectively. Altogether, these findings indicate that Rpt6-C undergoes dynamic conformational exchange between two states.
Structures of human Rpt6-C are available (PDB 2KRK and 3KW6) and demonstrate it to form the expected 4-helix bundle present in S. cerevisiae Rpt1 and Rpt3 C-terminal domains (Nakamura et al., 2007; Takagi et al., 2012). We used our chemical shift assignments (Table S1, BMRB accession number 18885) and CS-ROSETTA (Shen et al., 2008) to generate a structure of the Rpt6-C state with the most complete and dispersed set of assignments. The resulting structure contained a 4-helix bundle (Figure 1F) that matches the structure predicted by homology modeling alone via Rosetta (Rohl et al., 2004) and has a backbone root-mean-square deviation of 1.07 Å and 1.10 Å to the hRpt6-C NMR and x-ray structures, respectively (Figure S1D). We mapped the location of the amino acids found to undergo reversible chemical exchange onto our model structure to reveal that the majority forms a surface at the interface of helices α3 and α4 (Figure 1F). V346 and K349 in α2 lie adjacent to α3 and also undergo exchange, whereas no amino acid from α1 was found to undergo chemical exchange. N336, R339, G340, I341, H376, and T378 also undergo exchange and these amino acids are in the inter-helical loops bridging α1 to α2 and α3 to α4 (Figure 1F).
To determine the structural distinction between the two Rpt6-C states, we used our Cα and carbonyl chemical shift values to perform Chemical Shift Index (CSI) analyses (Wishart and Sykes, 1994), which is an established means to identify secondary structural elements. CSI analyses on the set of NMR signals used for the CS-ROSETTA 4-helix bundle structure (Figure 1F) yielded the expected secondary structure arrangement with all four predicted helices demonstrating negative Cα (Figure 2A, black) and carbonyl (Figure S2A, black) CSI values. The second set of signals, by contrast, demonstrated random coil values in regions predicted to be α-helical (Figure 2A and Figure S2A, blue). This finding suggests that the reversible exchange exhibited by Rpt6-C is principally characterized by a helix-coil transition. We used CS-ROSETTA to generate model structures for this state based on our chemical shift assignments (Table S1). The resulting structures were diverse in their tertiary fold, and among the ten lowest energy structures, we selected for demonstration that which most closely matches the 4-helix bundle (Figure 2B as compared to Figure 1F). The lack of convergence among the top ten lowest energy structures suggests structural heterogeneity for this second Rpt6-C structural state, which we henceforth refer to as partially unfolded.
We probed the internal dynamics of the two Rpt6-C structural states on the picosecond to millisecond time scale by using 1H-15N heteronuclear NOE (hetNOE), transverse relaxation (RN(NX)), and longitudinal relaxation (RN(NZ)) (Figure S2B) experiments. Rpt6-C’s C-terminal tail and its partially unfolded state exhibited significantly smaller hetNOE (Figure 2C and S2C) and RN(NX) (Figure 2D and S2D) values. These data demonstrate that as expected, Rpt6-C’s partially unfolded state and its C-terminal tail are more dynamic than its 4-helix bundle. We used the program relax (d’Auvergne and Gooley, 2008a, 2008b) to calculate a rotational correlation time of 8.5 ns for Rpt6’s 4-helix bundle, which is typical for a monomeric protein of this size.
Rpt6 sequence analysis revealed glycines within helices α3 and α4 (Figure 1A) at the heart of the exchanging region (Figure 1F). G355 and G366 are conserved among all S. cerevisiae and human Rpt proteins. G360 is present in S. cerevisiae and human Rpt6 whereas G387 is unique to S. cerevisiae Rpt6 (Figure 1A and S3A). Glycine is second only to proline as the most energetically destabilizing amino acid for helix formation; alanine, by contrast, is the most stabilizing for helices (Pace and Scholtz, 1998). We replaced G360 and G387 with alanine and acquired 1H, 15N HSQC and EXSY experiments on the resulting protein. All of the assigned amide signals for the partially unfolded state were lost (Figure S3B), as shown for the glycines in Figure 3A, as were all EXSY exchange peaks (Figure S3C). Moreover, the sedimentation coefficient of Rpt6-C increased from 1.213 ±0.003 S to 1.284 ±0.002 S upon alanine substitution of G360 and G387 (Figure 3B). The sedimentation coefficient is directly proportional to molecular weight and inversely proportional to the frictional coefficient. As the two proteins have a negligible molecular weight difference, this increase reflects a decreased frictional coefficient for Rpt6-C G360,387A, which is consistent with it having a more compact structure as averaged over time. Altogether, these results demonstrate that the observed Rpt6-C chemical exchange relies on the presence of two glycines within helices α3 and α4, and that their substitution to alanine stabilizes the Rpt6-C 4-helix bundle.
We hypothesized that Rpt6-C’s exchange to a partially unfolded and dynamic structural state would reflect a lower melting temperature, which would therefore be increased for Rpt6-C G360,387A. We used circular dichroism (CD) to evaluate secondary structure content across a temperature range spanning 5°C to 80°C. Rpt6-C and its G-to-A variants converted from a predominantly helical CD spectrum to one characteristic of a random coil across this temperature range (Figure S3D). We plotted molar ellipticity at 222 nm (Figure 3C) or 208 nm (Figure S3E) to find that the Rpt6-C melting temperature increased from 36°C to 40°C and 43°C when G387 and G360 were substituted with alanine respectively, and to 50°C when both glycines were substituted. We estimated that ~20% and ~33% of Rpt6-C is in the partially unfolded state at 25°C and 30°C, respectively, which is within the physiological range for S cerevisiae growth. Additional alanine substitution of G366 resulted in only a 2°C increase in melting temperature over Rpt6-C G360,387A (Figure S3E and S3F) and its 1H, 15N HSQC spectrum resembled that of Rpt6-C G360,387A (Figure S3G as compared to S3B). Similarly, the melting temperature of Rpt6-C G366A increased by only 1°C relative to wild-type Rpt6-C (Figure S3F), suggesting G366 does not play a role in destabilizing Rpt6-C structural integrity. Rather, G366 enables M335 and V377 side chains to pack into the protein core, thus facilitating the α1-α2 and α3-α4 loop configuration. This steric arrangement is conserved in structures of Rpt1-C (PDB 2DZN), Rpt3-C (PDB 4A3V), and PAN (PDB 3H4M).
To determine whether Rpt6-C conformational exchange influences its interaction with Rpt6’s N-terminal ATPase domain (spanning 136-317, NTD), we attempted to purify full-length S. cerevisiae Rpt6 protein from E. coli, but it precipitated during purification. Therefore, we used the structure of PAN subcomplex II from Methanocaldococcus jannaschii (Zhang et al., 2009a, 2009b) to create a homology model of full length Rpt6 using Prime (Schrödinger, Inc., Figure 3D). Rpt6-C’s α1 lies at the interface with Rpt6’s NTD and this helix does not contain any amino acids that undergo chemical exchange (Figure 1F). Interestingly, the exchange surface formed by amino acids in helices α3 and α4 is predicted to be exposed in the full length protein, but is interpolated between the Rpt6 ATP-binding pocket (Figure 3D) and the expected location of the CP interaction surface (Figure 3E). The remaining loop amino acids in conformational exchange form a peripheral ridge that may lie distal to these expected interaction surfaces.
In an effort to characterize Rpt6-C interaction with Rpn14, we acquired 1H, 15N HSQC and 1H, 15N Transverse Relaxation Optimized Spectroscopy (TROSY)-HSQC experiments on 15N labeled Rpt6-C in the presence of unlabeled S. cerevisiae Rpn14. A subset of Rpt6-C signals is significantly shifted by Rpn14 (Figure 4A), which bound to Rpt6-C in the slow exchange regime on the NMR time scale such that Rpt6-C signals from its unbound state disappear as its Rpn14-bound state signals emerge. NMR signals from Rpn14-bound Rpt6-C are attenuated due to the large molecular weight of the complex, and many emergent peaks are dispersed, suggesting that they belong to amides with structural integrity in the complex (Figure 4A, marked with asterisks). We measured the Kd for the Rpt6-C:Rpn14 interaction by sedimentation velocity. The data obtained at three Rpt6:Rpn14 ratios fit well to a 1:1 binding model but the affinity was too strong for a precise measurement by these methods. Global analysis indicates a best-fit Kd of 0.25 nM with an upper 95% confidence limit of 13 nM (Figure S4).
With Rpn14 present, we could resolve both states of 23 of the Rpt6-C signals that undergo conformational exchange. Among these, only G340, G355, A356, T378, and V386 were notably affected by the presence of Rpn14 (Figure 4A, 4B, and 4C). In all cases, the helical state of these amino acids was shifted to another location in the Rpt6-C spectrum acquired with equimolar quantities of Rpn14, as exemplified for A356 (Figure 4B) and V386 (Figure 4C). By contrast, the partially unfolded state signal is present at the same location in the spectrum (Figure 4A, 4B and 4C). This finding suggests that Rpn14 binds selectively to the structurally intact state of Rpt6.
Rpt6-C reversibly exchanges between its two conformational states (Figure 1D and 1E) and therefore, Rpn14 binding to Rpt6-C’s 4-helix bundle should lead to a reduced population of its partially unfolded state. This effect is manifested throughout the Rpn14 titration, as increasing quantity of Rpn14 leads to reduced signal intensity from Rpt6-C’s partially unfolded state, as exemplified for V386 (Figure 4C, in which all spectra are plotted at an equivalent threshold). We note that Rpn14 aggregates during this titration and the molar ratio reported here is based on protein added and that the effective Rpn14 available for Rpt6-C binding may be reduced.
The binding of Rpn14 to Rpt6-C’s 4-helix bundle suggested that this interaction should not be diminished in Rpt6-C G360,387A, unless G360 and G387 are within the Rpn14-binding surface. We tested whether Rpt6-C G360,387A affinity for Rpn14 is reduced by analytical ultracentrifugation. As in the case of wild-type Rpt6-C, this analysis revealed high affinity binding, with a Kd of 0.01 nM and an upper 95% confidence limit of 20 nM (data not shown). The confidence limits are broad relative to Kd due to the high concentrations required for detection in the experiment; however, it is clear that high affinity binding is preserved upon introducing the alanine substitutions.
Whereas the sedimentation coefficients of Rpt6-C and Rpt6-C G360,387A are significantly different (Figure 3B), that of their Rpn14-bound states (Figure 4D) differ by only 0.005 S (Figure 4D). Hydrodynamic models (De la Torre et al., 1994; de la Torre et al., 2007) suggest that if Rpt6-C were to remain partially unfolded, the sedimentation coefficient of the Rpt6-C:Rpn14 complex would be 0.06 S smaller than that of the Rpt6-C G360,387A:Rpn14 complex. This finding is consistent with the NMR evidence that Rpt6-C is folded when bound to Rpn14 (Figure 4A–C).
The slow exchange binding regime of the Rpt6-C:Rpn14 interaction prevents the tracking of Rpt6-C signals from their free to their Rpn14-bound state (Figure 4A). The Rpn14 bound-state peaks exhibited significant signal attenuation due to the increased size of this 57 kDa complex, and attempts to increase sample concentration above 0.3 mM resulted in precipitation. In an effort to identify the Rpt6-C amino acids that participate in Rpn14 binding, we examined Rpt6-C amide chemical shift changes caused by Rpn14 addition for each amino acid (Figure 5A). Fourteen amino acids were excluded from this analysis, including those suffering from severe spectral crowding and those with an unknown free-state assignment (Figure 5A, asterisks). We identified 24 Rpt6-C signals that upon Rpn14 addition shift to a location in the spectrum that is remote from their free state (Figure 5A, ^). We were unable to unambiguously assign their Rpn14-bound state signals; however we could conclude that they are significantly shifted. These amino acids most affected by Rpn14 propagate through the Rpt6-C structure and are included within the α1 - α2 and α2 - α3 loops, the N-terminal ends of α2 and α4, and the C-terminal end of α1 (Figure 5B).
Four buried hydrophobic amino acids (V346, M350, F382, and V386 (Figure 4C)) as well as G355 (Figure 4A) and A356 (Figure 4B) at the N-terminal end of α3 are shifted upon Rpn14 addition (Figure 5A); these amino acids demonstrate two conformational states in free Rpt6-C (Figure 1F). V346, M350, F382, and V386 within Rpt6-C’s hydrophobic core are not exposed for Rpn14 interaction and G355 and A356 abut the putative nucleotide-binding pocket in full length Rpt6 (Figure 5C). Since Rpn14 appears to bind to Rpt6-C’s 4-helix bundle (Figure 4), these inaccessible amino acids may be shifting upon Rpn14 addition due to stabilization of Rpt6-C’s structurally intact state rather than their direct interaction with Rpn14. We also observed significant shifting for Rpt6-C’s C-terminal amino acid, K405. A truncated Rpt6-C protein construct, Rpt6-C (P318-N393), in which its C-terminal tail spanning Q394-K405, is missing binds Rpn14 in the same manner as wild-type Rpt6-C (Figure S5A). This finding suggests that Rpt6’s C-terminus is not necessary for its binding to Rpn14. Nonetheless, the C-terminal tail shifting upon Rpn14 addition suggests that it may be reconfigured as a consequence of Rpt6 binding to Rpn14 or form additional contacts with Rpn14 that are not required for the Rpt6:Rpn14 interaction.
We combined our NMR data (Figure 5A) with previously published mutagenesis data (Figure S5B) to generate structural models of the Rpt6-C:Rpn14 protein complex by using TreeDock (Fahmy and Wagner, 2002). Complexes were excluded that yielded steric clashes with the expected location of Rpt6’s NTD and that were not affected by Rpn14 D157A and D157K amino acid substitution or alanine substitution of Rpt6 R373 and R374 (Figure S5B) (Kim et al., 2010). In the remaining model structures, Rpn14 is placed at an Rpt6-C surface formed by the N-terminal ends of α2 and α4, and the α1-α2 and α3-α4 loops (Figure S5C). It is noteworthy that the peripheral edge of Rpt6-C formed by loops α1-α2 and α3-α4, which was found to undergo chemical exchange (Figure 3E), contributes to Rpn14 binding (Figure 5D).
To test the functional significance of restricting Rpt6-C to the 4-helix bundle in vivo, we used S. cerevisiae and substituted G360 and G387 of RPT6 with alanine on its chromosomal locus; this mutant strain is referred to as rpt6AA. We examined the interaction of endogenous Rpn14 with proteasome complexes purified from wild-type or rpt6AA mutants. Strikingly, the level of Rpn14 was significantly decreased in proteasome complexes from the rpt6AA mutant (Figure 6A, lane 1 and 2). The proteasome regulatory particle can be divided into base and lid sub-complexes. The base includes the Rpt1-6, Rpn1, Rpn2, Rpn13 and Rpn10. Purified base isolated via high-salt wash (1 M NaCl) similarly showed a decrease in Rpn14 level for the rpt6AA mutant (Figure 6A, lane 3 and 4).
Since RP chaperones including Rpn14 have been proposed to temporally order RP assembly through interactions with specific Rpt subunits, we examined the status of proteasome assembly in the rpt6AA mutant. Together with the mature proteasome holoenzymes (RP2-CP, RP-CP), major assembly intermediates (RP and base) were readily detected in both wild-type and rpt6AA mutants (Figure 6B). However, Rpn14, which specifically associates with RP and base (Figure 6B; wild-type) (Park et al., 2009; Roelofs et al., 2009), was nearly absent from either assembly intermediate in the rpt6AA mutant (Figure 6B). These results were confirmed by using 2-dimensional native PAGE followed by SDS-PAGE (Figure 6C). Moreover, immunoblotting the 2D gel for Rpn1 revealed additional species in rpt6AA proteasomes, which migrate between RP2-CP and RP-CP (Figure 6C). These species may result from enhanced Ecm29 or Blm10 recruitment to RP-CP, which is charateristically observed in assembly-deficient proteasome mutants (Park et al., 2011).
We tested whether Rpn14 reduction on rpt6AA base stems from its decreased cellular abundance. Immunoblotting of whole cell extracts demonstrated slightly less Rpn14 in rpt6AA compared to wild-type cells (Figure 6D) and this might in part contribute to the near absence of Rpn14 on rpt6AA base (Figure 6B and 6C). However, RPN14 deletion abolishes the presence of base in yeast (Saeki et al., 2009), suggesting that it is essential for either its accumulation or formation in vivo and is functional in the rpt6AA strain.
To better understand the effect of Rpn14 and its interaction with Rpt6 during assembly, we compared the formation of base between rpt6AA and rpn14 null mutants via 3.5% and 5% native gel assays using whole cell extracts (Figure 6E). As expected (Funakoshi et al., 2009; Saeki et al., 2009), base was scarcely present in rpn14Δ. In rpt6AA however, base appeared at a level comparable to wild-type (Figure 6E). The rpt6AA rpn14Δ double mutant mimicked rpn14Δ, demonstrating strongly diminished levels of base (Figure 6E and S6). This finding suggests that despite its reduced levels on rpt6AA base (Figure 6B and 6C), Rpn14 is still actively contributing to base formation in this mutant.
We therefore tested the effect of Rpt6 G360,387A on base-CP assembly in the presence of excess Rpn14 in vitro by native PAGE with either Coomassie blue staining or an in-gel peptidase assay. For this experiment, CP and base purified from wt or rpt6AA strains were incubated with or without recombinant Rpn14 (Figure 6F). Whereas wt base readily integrated into base-CP complexes (Figure 6F, lane 4), rpt6AA base demonstrated poor complex formation (Figure 6F, lane 7). The addition of excess Rpn14 resulted in a reduction of base-CP species with wild-type Rpt6, consistent with the current model that RP chaperones, such as Rpn14, compete with CP for Rpt binding during assembly (Park et al., 2009; Roelofs et al., 2009). Importantly, Rpn14 was more potent in reducing the formation of base-CP complexes in the presence of rpt6AA base (Figure 6F, lanes 8–9). Our findings indicate that the intrinsic helix-coil transition of Rpt6 may provide a critical control for proper binding and release of Rpn14 during proteasome assembly.
To assess the impact of rpt6AA on substrate degradation by proteasome, we assayed for canavanine sensitivity. rpn14Δ and rpt6AA single mutants demonstrated modest canavanine sensitivity at 37°C; however, this growth defect was significantly increased in the rpn14Δ rpt6AA double mutant (Figure 7A). This finding indicates that rpt6AA does not simply mimic rpn14Δ. Nonetheless, assembled proteasome is observed in rpn14Δ rpt6AA (Figure 6E). It is possible that in contrast to in vitro CP-base assembly (Figure 6F), structural heterogeneity is not important for RP-CP assembly. Alternatively, structural heterogeneity may be enabled through other Rpt proteins (Figure 1A). To test whether other Rpt proteins exhibit conformational exchange, we acquired 1H, 15N HSQC and EXSY experiments on S. cerevisiae Rpt4 (Figure 7B) and Rpt5 (Figure 7C). Both proteins demonstrated millisecond-scale conformational exchange in the EXSY experiments (Figure 7B and 7C, red). Similar to Rpt6 (Figure 1D), correlations were observed between dispersed and non-dispersed signals, suggesting that Rpt4 and Rpt5 may also exhibit helix-coil transitions. Thus, conformational exchange may be a general phenomenon of proteasome S. cerevisiae Rpt proteins to facilitate RP-CP assembly (Figure 7D). Future experiments are needed to test this hypothesis.
We found that the two most C-terminal helices of S. cerevisiae 26S proteasome ATPase Rpt6 undergoes helix-coil transitions. Alanine substitution of G360 and G387, located within these helices, restricts Rpt6’s C-terminal domain to its structurally intact 4-helix bundle and increases its melting temperature. It is noteworthy that exchange in the α1-α2/α3-α4 loop region is abrogated in Rpt6 G360,387A, as this finding demonstrates that the α3 and α4 helix-coil transitions are propagated to this more peripheral Rpn14-binding region.
Rpt6 G360,387A affinity for Rpn14 is not decreased compared to wild-type Rpt6 and is in the low nM range, as assessed by analytical ultracentrifugation. This affinity is surprisingly high, but specific to Rpt6-C’s 4-helix bundle state. We propose that Rpt6-C conformational exchange could play a role in Rpn14 binding and release during proteasome assembly. Interestingly, base from rpt6AA exhibited reduced Rpn14, suggesting that Rpt6-C conformational heterogeneity may also help accommodate Rpn14 on base.
Rpt1-C, Rpt3-C and Rpt6-C use a similar surface to bind their molecular chaperones (Figure S5D-F), which may have evolved to position the chaperones proximal to the CP. Chaperone clashes with CP are suggested by the Rpt1:Hsm3 (Barrault et al., 2012) and Rpt3:Nas6 (Roelofs et al., 2009) structural complexes. Steric clashes are also predicted between Rpn14 and the CP when Rpt6’s C-terminal Hb-Y-X-like motif is docked into the CP α-ring pockets (Figure S5C). These structural models provide support for the hypothesis that chaperones and CP compete for Rpt binding during assembly (Park et al., 2009; Roelofs et al., 2009). Structural transitions could fine tune this competition to enable effective interconversion between Rpt:chaperone and Rpt:CP complexes. In particular, steric clashes between Rpn14 and CP may favor Rpt6 exchange to its partially unfolded state and in turn promote Rpn14 expulsion.
In vitro base-CP complex formation was inefficient with Rpt6 G360,387A, perhaps because base isolated from rpt6AA has significantly reduced levels of Rpn14. Nonetheless, addition of recombinant Rpn14 exacerbated the base-CP assembly defect, most likely because its interaction with base cannot be fine tuned by Rpt6-C dynamics. Assembly defects were also apparent for rpt6AA in vivo. Proteasome species were observed in rpt6AA extracts that migrate between RP2-CP and RP-CP by native gel; these are not present in wild-type cells and are hallmarks of assembly defects (Park et al., 2011). rpt6AA exhibited hypersensitivity to canavanine when introduced into an rpn14 null, suggesting Rpt6-C dynamics also plays an Rpn14-independent role. Altogether, these results suggest that Rpt6-C dynamics facilitate proteasome assembly in yeast. It is possible that structural heterogeneity in other Rpt proteins also contributes to proteasome assembly, as Rpt4 and Rpt5 were found to undergo conformational exchange. Future experiments are needed to evalute the effect of losing conformational exchange in these Rpt proteins.
S. cerevisiae Rpt6-C [P318-K405], Rpt5-C [L351-A434] (Lee et al., 2011), Rpt4-C [L351-L437], Rpn14, and their variants were expressed from pRSF-duet1 (Rpt) and pGEX-6p (Rpn14) vectors in E. coli fused with N-terminal polyhistidine (for Rpts) or glutathione S-transferase followed by a PreScission protease cut site (for Rpn14), as described (Roelofs et al., 2009). Affinity chromatography using Talon resin (Clontech) was followed with size-exclusion chromatography by FPLC. Buffer 1 (20 mM NaPO4, 60 mM NaCl, 14 mM β-mercaptoethanol, at pH 6.8) was used unless otherwise indicated. Selectively labeled samples were produced by growing cells in minimal media supplemented with 200 mg/L unlabeled tryptophan, 150 mg/L of 15N labeled amino acid of interest, and 150 mg/L of unlabeled amino acid for the remaining 18 types.
NMR experiments were conducted at 25°C on 800, 850, or 900 MHz spectrometers equipped with cryogenically cooled probes. 1H, 15N HSQC, 1H, 15N ZZ-Exchange, 1H-15N heteronuclear NOE, RN(Nz), and RN(NX) experiments were acquired on 15N-labeled Rpt6-C at 40–400 μM. 1H, 15N, 13C HNCA, HNCOCA, HNCO, HNCACO, and HNCACB spectra were acquired on 15N, 13C and 50% 2H-labeled Rpt6-C at 300–500 μM. 1H, 15N, TROSY-HSQC experiments were conducted on 15N, 90% 2H samples at 200 μM. RN(NX) and RN(NZ) data were collected in parallel with a pseudo-3D experiment and a repeated data point for error calculation. 1H-15N heteronuclear NOE data were acquired with a 4 s saturation transfer or control period. Relaxation data fitting and model-free analysis were performed by using the program relax (d’Auvergne and Gooley, 2008a, 2008b). The global rotational correlation time was estimated from the average of local τm values derived from the Lipari-Szabo method (Lipari and Szabo, 1982).
Hydrodynamic modeling was performed by estimating free Rpt6-C, Rpt6-C G360,387A and Rpn14 as spheres with radii corresponding to their experimentally-determined Stokes’ radii and molecular weights corresponding to their amino acid composition. The Rpt6-C:Rpn14 and Rpt6-C G360,387A:Rpn14 complexes were then modeled as two adjacent (touching) beads, and the two complexes were compared using the program HYDRO++ (De la Torre et al., 1994; de la Torre et al., 2007).
The Rpt6-C structures for each of the two states were calculated with CS-ROSETTA version 3.1 (Shen et al., 2008) and HN, N, carbonyl, Cα, and Cβ chemical shift assignments. TreeDock (Fahmy and Wagner, 2002) was used to model the Rpt6-C:Rpn14 protein complex. Rpn14’s D157 and Rpt6-C amino acids identified to be most affected by Rpn14 addition (Figure 5A) were paired in all possible arrangements and placed at optimal van der Waals distance. Rpt6-C was fixed while Rpn14 was translated and rotated at 0.7 Å increments about the anchored atoms. All non-clashing orientations that satisfied the chemical shift perturbation data were saved. The program Octopus (Fahmy and Wagner, 2011) was used to simulate the mutation experiments (Figure S5B).
Full-length Rpt6 was modeled with Prime (Schrödinger, Inc.) using our CS-Rosetta structure and the AAA-ATPase subunit of the M. janaschii Proteasome Activating Nucleotidase (PDB 3H4M). To generate a structure with Rpt6’s C-terminal Hb-Y-X-like motif docked into the CP, our Rpt6 structure was aligned 6-fold to the AAA-ATPase structure from HslU (PDB 1DO0) (Bochtler et al., 2000) with our modeled Rpt6-C:Rpn14 structure replacing one of the Rpt C-terminal domains. The resulting ATPase:Rpn14 complex was lowered onto the CP α-ring (PDB 3H4P) with Rpt6’s C-terminal twelve residues oriented based on the PA26/PAN fusion protein (PDB 3IPM) (Yu et al., 2010).
Sedimentation velocity analysis was conducted at 20°C and 50,000 RPM using interference optics with a Beckman-Coulter XL-I analytical ultracentrifuge. Double sector synthetic boundary cells equipped with sapphire windows were used to match the sample and reference menisci. The rotor was equilibrated under vacuum at 20°C and after a period of 1 hr., the rotor was accelerated to 50,000 RPM. Interference scans were acquired at 60 second intervals. Rpt6-C, Rpt6-C G360,387A and Rpn14 were each analyzed in Buffer 2 (20 mM NaPO4, 130 mM NaCl, 14 mM β-mercaptoethanol, at pH 6.8) at three concentrations spanning a ~10-fold range with Rpt6-C at 17–130 μM, Rpt6-C G360,387A at 17–160 μM, and Rpn14 at 4–33 μM. The complexes were analyzed at 20–25 μM Rpn14 and 0.5, 1 and 2 equivalents of Rpt6-C or the G360,387A variant. The c(s) sedimentation coefficient distribution plots were obtained using SEDFIT (Schuck, 2000). Sedimentation coefficients and dissociation constants were obtained by global analysis using SEDANAL (Stafford and Sherwood, 2004).
CD was conducted on a JASCO J-815 spectrophotometer between 190 and 250 nm and with Rpt6-C in Buffer 1 diluted 20-fold with H2O for a protein concentration of 10 μM. The temperature was raised in 5°C increments with 12 minutes of equilibration before acquisition.
Protein A resin was used to purify Rpt1-containing complexes by using proA-TeV-Rpt1, in which Rpt1’s N-terminal end contains a protein A affinity tag followed by a TeV protease cleavage site (proA-TeV-Rpt1), as described previously (Park et al., 2009). 4–6L of yeast culture was grown in YPD overnight to an OD600 of 15–20, harvested and ground in liquid nitrogen, and the ground powders hydrated in an equivalent volume of buffer 3 (50 mM Tris-HCl [pH 7.5], 5 mM MgCl2, 1 mM EDTA, 10% glycerol, and 1 mM ATP) supplemented with protease inhibitors. Lysates were pelleted at 4°C and the supernatant incubated with Affinity Gel Rabbit IgG resin (MP Biomedical) for 2 hr. at 4°C. Resins were washed extensively in buffer 3 containing 100 mM NaCl followed by 1 M NaCl. After a final wash with buffer 3, base was eluted by incubating the resin with 1 bed volume of buffer 3 containing TeV protease (Invitrogen) for 1 hr. at 30°C. CP was purified as described previously (Leggett et al., 2005) except that lysis was performed in liquid nitrogen.
Affinity-purified base was mixed with CP at 2:1 molar ratio in 10 μl of buffer 3 at 30°C for 15 min. The entire reactions were immediately loaded on pre-chilled 3.5% native gels and run for 2.5 hr. at 100V at 4°C (Roelofs et al., 2009). In-gel peptidase assays were conducted as described (Roelofs et al., 2009) with an AlphaImager system (ProteinSimple).
NMR data was acquired at MNMR and NMRFAM and we are grateful for technical assistance from Marco Tonelli and Youlin Xia. We thank Jeroen Roelofs for providing our Rpt5-C construct and Can Ergenekan and Xiang Chen for assistance with CS-ROSETTA and Prime, respectively. All computation was performed in the MSI BSCL. This work was funded by the National Institutes of Health (CA097004 and CA136472 to KJW and GM043601 to DF).
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.