- Email: [email protected]
Bivalent Tethering of SspB to ClpXP Is Required for Efficient Substrate Delivery
Molecular Cell, Vol. 13, 443–449, February 13, 2004, Copyright 2004 by Cell Press
Bivalent Tethering of SspB to ClpXP Is Required for Efficient Substrate Delivery: A Protein-Design Study Daniel N. Bolon,1 David A. Wah,1 Greg L. Hersch,1 Tania A. Baker,1,2 and Robert T. Sauer1,* 1 Department of Biology and 2 Howard Hughes Medical Institute Massachusetts Institute of Technology Cambridge, Massachusetts 02139
Summary SspB homodimers deliver ssrA-tagged substrates to ClpXP for degradation. SspB consists of a substrate binding domain and an unstructured tail with a ClpX binding module (XB). Using computational design, we engineered an SspB heterodimer whose subunits did not form homodimers. Experiments with the designed molecule and variants lacking one or two tails demonstrate that both XB modules are required for strong binding and efficient substrate delivery to ClpXP. Assembly of stable SspB-substrate-ClpX delivery complexes requires the coupling of weak tethering interactions between ClpX and the SspB XB modules as well as interactions between ClpX and the substrate degradation tag. The ClpX hexamer contains three XB binding sites, one per N domain dimer, and thus binds strongly to just one SspB dimer at a time. Because different adaptor proteins use the same tethering sites in ClpX, those which employ bivalent tethering, like SspB, will compete more effectively for substrate delivery to ClpXP.
Short Article
hexamer, and experiments mixing full-length and N-terminal truncations of ClpX have been interpreted by Dougan et al. (2003) as evidence that SspB binding and efficient substrate delivery require only a single N domain of ClpX. This result suggests that as many as six SspB dimers could bind ClpX and predicts that only one XB module per SspB dimer should be needed for strong binding and substrate delivery to ClpXP. However, other studies indicate that substrate-delivery complexes contain just one SspB dimer and two ssrA-tagged substrates per ClpX hexamer (Wah et al., 2002), suggesting that more than one dimer of SspB does not interact with ClpX. Does an SspB dimer with only a single XB module mediate efficient substrate delivery to ClpXP? To address this question, we used computational design to engineer SspB heterodimers with different numbers of XB sequences and assayed these proteins for enhancement of ClpXP degradation. SspB molecules with only one XB module were found to be significantly less active in substrate delivery than those with two modules. We also found that the ClpX N domain contains the tethering sites for binding SspB. However, one dimeric N domain bound only a single XB peptide, indicating that a ClpX hexamer contains just three tethering sites for the XB module of SspB, and thus could bind strongly to only one SspB dimer at a time. These results shed light on the detailed mechanism of substrate delivery by SspB and help us understand how different intracellular adaptors might compete for ClpXP, thereby determining the priority of substrate selection.
Introduction Results In the AAA⫹ protease ClpXP, the hexameric ClpX ATPase binds protein substrates and unfolds and translocates these molecules into the degradation chamber of the ClpP peptidase (Hoskins et al., 2001). ClpX recognizes target proteins via peptide signals such as the ssrA tag, an 11 residue sequence added to the C terminus of nascent proteins on stalled ribosomes (Keiler et al., 1996; Gottesman et al., 1998; Flynn et al., 2003). SspB, a dimeric adaptor protein, also binds ssrA-tagged proteins and enhances their degradation by ClpXP (Levchenko et al., 2000; Wah et al., 2002). Each SspB subunit contains a substrate binding domain (SBD) and a flexible C-terminal tail (Levchenko et al., 2003; Song and Eck, 2003; Wah et al., 2003). An XB module at the end of each SspB tail mediates tethering interactions with ClpX (Wah et al., 2003), perhaps by binding to a site in the ClpX N-terminal domain (Wojtyra et al., 2003; Dougan et al., 2003). The mechanism by which SspB stimulates ClpXP degradation of ssrA-tagged substrates is only partially understood. It is clear that the tails of SspB form tethering interactions with ClpX, increasing the local concentration of ssrA-tagged substrates relative to the enzyme. In principle, several SspB dimers could bind one ClpX *Correspondence: [email protected]
Heterodimer Design and Properties The SBD of H. influenza SspB is a symmetric dimer (Levchenko et al., 2003), but many studies reported below required us to purify and assay heterodimeric forms of SspB. To identify sequence variants that would form heterodimers rather than homodimers, we employed computational design. Using the ORBIT software package (Dahiyat and Mayo, 1997), we repacked the dimer interface of SspB in an asymmetric fashion using a modified search algorithm to optimize the energy of the heterodimer relative to homodimers of either partner sequence. The top-ranked heterodimer design resulting from this optimization contained seven mutations relative to the wild-type SspB dimer: L12Y, A15G, Y16F, and V101M in one subunit (called YGFM) and A15S, Y16L, and V101A in the partner subunit (called SLA). Genes encoding the H. influenza YGFM and SLA SBDs with C-terminal tails from E. coli SspB were constructed. E. coli SspB and H. influenza SspB have similar SBD structures (Levchenko et al., 2003; Song and Eck, 2003), and a hybrid (HE SspB) containing the wild-type H. influenza SBD with E. coli tails delivers GFP-ssrA to E. coli ClpXP as well as both parental SspB molecules (data not shown). Following coexpression, the YGFM/SLA protein purified as a single species containing both mutant sub-
Molecular Cell 444
Figure 1. Properties of the Designed Heterodimer (A) Analytical ultracentrifugation (20⬚C; 16,000 rpm) of the YGFM/SLA protein (25 M) after 36 hr. The fitted solid line represents a value of 37.6 kDa for the heterodimer molecular weight. Dashed line, mass distribution expected for a monomer. (B) Far-UV circular dichroism spectra of 25 M YGFM/SLA or the parental HE SspB protein (25⬚C). (C) Fluorescence spectra of YGFM/SLA under native and denaturing (5 M GuHCl) conditions (25⬚C). (D) GuHCl denaturation of 3 M YGFM/SLA (⌬Gu⫽16.7 kcal/mol; m ⫽ 5.7 kcal/mol•M) and HE SspB (⌬Gu ⫽ 22.4 kcal/mol; m ⫽ 6.7 kcal/ mol•M) at 25⬚C. The fitted curves and stability parameters are for native dimer to unfolded monomer reactions. (E) YGFM/SLA heterodimer and E. coli SspB bound equally well to a fluorescent ssrA peptide (50 nM) as assayed by changes in anisotropy at 30⬚C, pH 7.6, 200 mM KCl. The solid line represents a KD of 3 M. (F) Degradation of GFP-ssrA (0.6 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) in the absence of SspB or in the presence of YGFM/ SLA or E. coli SspB (0.6 M each). The solid lines are linear fits to the kinetic data. Degradation in the presence of HE SspB was identical to that observed with YGFM/SLA (data not shown).
units, and analytical equilibrium ultracentrifugation experiments (Figure 1A) gave a molecular weight (37.6 kDa) close to the expected heterodimeric value (37.9 kDa). The circular-dichroism spectrum of YGFM/SLA (Figure 1B) was similar to that of the parent molecule, indicating similar secondary structures. Native YGFM/SLA also had a fluorescence spectrum that was blue-shifted relative to denatured protein (Figure 1C), showing that both tryptophans in the heterodimer interface are buried. In denaturation experiments (Figure 1D), YGFM/SLA displayed a single cooperative unfolding transition with stability decreased only modestly compared to the HE SspB dimer. These experiments show that the designed YGFM/SLA molecule forms a stably-folded dimer with structural properties similar to the SspB homodimer. The design of the YGFM/SLA dimer would not be considered a success unless it bound ssrA-tagged substrates and delivered them to ClpXP for degradation. A fluorescent ssrA peptide bound equally well to YGFM/ SLA and wild-type E. coli SspB (Figure 1E), showing that redesign of the dimer interface does not disturb ssrA
binding. YGFM/SLA was also essentially as active as wild-type E. coli SspB in stimulating the degradation of ssrA-tagged substrates by ClpXP (Figure 1F). We conclude that the designed YGFM/SLA protein is fully functional in substrate delivery. Heterodimers versus Homodimers When the SLA or YGFM subunits were purified to homogeneity under denaturing conditions and then allowed to refold, most protein precipitated suggesting an inability of both mutant homodimers to fold stably (data not shown). To compare the stabilities of different dimeric combinations of natural SspB with the designed heterodimer, we denatured/renatured an equal mixture of wildtype E. coli SspB and its SBD and purified the dimer containing one wild-type SBD and one wild-type fulllength subunit by ion-exchange chromatography under conditions where subunit exchange is slow. Figure 2A shows the elution profile of this purified protein following rechromatography. After 24 hr of incubation at 30⬚C, however, ion exchange revealed three peaks corre-
Designed SspB Heterodimers 445
Figure 2. Subunit Exchange Assayed by MonoQ Ion-Exchange Chromatography (A) Rechromatography of a dimer containing one full-length and one substrate binding domain of E. coli SspB immediately following the initial purification of this peak on the same ion-exchange column. (B) Rechromatography of the material from (A) following 24 hr of incubation at 30⬚C. (C) Rechromatography of a YGFM/SLA heterodimer containing one full-length and one substrate binding domain following initial purification. (D) Rechromatography of the material from (C) following 24 hr of incubation at 30⬚C. In all experiments, ion-exchange chromatography was performed at 4⬚C and was monitored by absorbance at 280 nm. The identities of the subunits in each peak were determined by SDS-PAGE.
sponding to the SBD homodimer, the SBD/SspB heterodimer, and the SspB homodimer in ratios (1:2:1) expected for unbiased equilibrium redistribution of these subunits (Figure 2B). This experiment was repeated using a heterodimer consisting of a YGFM SBD and a fulllength SLA subunit (Figure 2C). Following 24 hr, only the peak corresponding to the original YGFM-SBD/SLA heterodimer was detected (Figure 2D). SDS-PAGE confirmed that this peak contained both heterodimeric partners. To exclude the possibility of slow exchange, the YGFM-SBD/SLA heterodimer was denatured and allowed to refold. Again, only the single YGFM-SBD/SLA heterodimer peak was present (data not shown). We conclude that subunits containing the YGFM and SLA mutations form a heterodimer that must be significantly more stable than either mutant homodimer. Activity of One-Tailed SspB Variants To test the importance of the number of SspB XB modules in substrate delivery, we purified variants with two (YGFM/SLA), one (YGFM/SLA-SBD), or zero (YGFMSBD/SLA-SBD) tails and assayed ClpXP degradation
rates at different concentrations of the heterodimer and GFP-ssrA (Figure 3A). The heterodimer with two tails and XB modules was more active than the single-tailed species, which in turn was more active than the molecule with no tails. To confirm that this result was not an artifact of the heterodimer design, we also assayed wildtype E. coli SspB variants containing two, one, or zero tails. As in the YGFM/SLA background, two XB tails were required for full enhancement of ClpXP degradation of GFP-ssrA (Figure 3B). The otherwise wild-type one-tailed molecule described above might generate molecules with two tails by exchange during the degradation assay. Native gel electrophoresis showed that the exchange half-life of wild-type E. coli SspB dimers is about 30 min at 30⬚C (Figure 3C), whereas ClpXP degradation rates were determined during a 5 min reaction at 30⬚C. Thus, only a small amount of subunit exchange (ⵑ10%) probably occurred in this experiment. Nevertheless, this problem of exchange of wild-type subunits highlights the importance of being able to use the designed YFGM/SLA heterodimer, which precludes homodimer formation. The results shown in Figures 3A and 3B can be viewed in two ways. First, two tails and XB sequences are clearly required for full SspB activity. This is true with respect to both the KM and Vmax for ClpXP degradation. Second, dimers with one tail are more active than those with no tails, and thus a single tail must permit some interaction with ClpX. It is important to recall, however, that SspB is not required for ClpXP degradation of GFP-ssrA but, when present, decreases KM and increases Vmax. In fact, KM and Vmax values determined previously for ClpXP degradation of free GFP-ssrA (Levchenko et al., 2000; Wah et al., 2002) were similar to the values determined here for degradation of GFP-ssrA bound to single-tailed SspB variants. Hence, single-tailed SspB dimers should not stimulate degradation of GFP-ssrA to any appreciable extent. This prediction was confirmed. GFP-ssrA with no added SspB was degraded at about the same rate as in the presence of single-tailed YGFM/SLA, whereas degradation in the presence of the two-tailed variant was stimulated significantly (Figure 3D). GFP-ssrA bound to tail-less SspB interacts very inefficiently with ClpX (Wah et al., 2003; Figures 3A and 3B), suggesting that the SspB SBD makes unfavorable interactions with ClpX or masks recognition determinants in the bound ssrA tag. We assume that the single-tailed constructs allow weak tethering to ClpX (see below), which offsets the inhibitory effect of binding of ssrA-tagged substrates to the SBD, resulting in a substrate with properties similar to those of free GFP-ssrA. Binding of One- and Two-Tailed SspBs to ClpX To allow another probe of ClpX interaction, we designed an SspB binding-pocket mutation (A73Q) to block ssrAtag binding. A73Q variants folded and dimerized normally but did not bind ssrA peptide (KD ⬎ 500 M; data not shown). A73Q variants of YGFM/SLA heterodimers with zero, one, or two tails were used to compete for degradation of GFP-ssrA by ClpXP in the presence of wild-type E. coli SspB. As shown in Figure 3E, the twotailed A73Q heterodimer (Ki ⬇ 2.5 M) was a significantly better competitor than the single-tailed molecule (Ki ⬇
Molecular Cell 446
Figure 3. Two SspB Tails Are Required for Full Activity and Strong Binding to ClpX (A) Rate of ClpXP degradation of GFP-ssrA as a function of substrate and SspB concentration for YGFM/SLA heterodimers with two, one, or zero tails and XB modules. (B) Rate of ClpXP degradation of GFP-ssrA as a function of substrate and SspB concentration for E. coli SspB heterodimers with two, one, or zero tails and XB modules. (C) Kinetics of subunit exchange assayed by native gel electrophoresis. For E. coli SspB, a heterodimer containing one full-length subunit and one SBD subunit was incubated at 30⬚C, and samples were taken after different incubation times and placed on ice until electrophoresis on a native gel. The figure shows the approach to the equilibrium 1:2:1 distribution of subunits. When a YGFM/SLA heterodimer containing full-length subunits was mixed with His6-tagged E. coli SspB, no subunit exchange was detected by native gel following 24 hr of incubation at 30⬚C. (D) Degradation of GFP-ssrA (0.3 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) in the absence of SspB or in the presence of designed YGFM/SLA heterodimers with two, one, or zero tails/XB modules (0.3 M each). The solid lines are linear least squares fits to the kinetic data. (E) Degradation of GFP-ssrA (0.2 M) in the presence of E. coli SspB (0.6 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) was inhibited by increasing concentrations of the XB peptide or the A73Q variants of YGFM/SLA heterodimers containing one or two C-terminal tails/XB modules. The A73Q mutation prevents binding of SspB to ssrA-tagged substrates. (F) The A73Q variants of YGFM/SLA at concentrations of 55 M did not significantly inhibit ClpXP degradation of GFP-ssrA alone. Degradation conditions were identical to those in (E) except E. coli SspB was not added.
20 M). These results confirm that two tails are important for strong binding of SspB dimers to ClpX. Importantly, inhibition by the single-tailed A73Q heterodimer was essentially the same as inhibition by an isolated XB peptide (Figure 3E). This result indicates that the XB module of SspB makes all of the energetically significant interactions with ClpX. The A73Q variant of tail-less YGFM/SLA did not inhibit SspB-mediated degradation of GFP-ssrA by ClpXP (data not shown), and exchange between YGFM/SLA subunits and E. coli SspB subunits was not detected (Figure 3C). Both results indicate that the inhibition observed in Figure 3E occurs by competition for the tethering sites on ClpX and not by mixing of SspB subunits. Interestingly, the A73Q variants had little effect on the degradation of GFP-ssrA alone (Figure 3F). Interactions of SspB with the ClpX N Domain To assay for potential interactions with the XB peptide of SspB, N domain fragments of ClpX (residues 1–61 or His6-tagged 1–64) were purified and studied. Consistent with a recent report (Wojtyra et al., 2003), the untagged
N domain formed a stable dimer as assayed by analytical equilibrium centrifugation (data not shown). As monitored by fluorescence anisotropy (Figure 4A), the XB peptide bound the untagged N domain with a KD (20 ⫾ 5 M) within error of that for intact ClpX (23 ⫾ 7 M). We conclude that the N domain mediates all of the energetically significant contacts between ClpX and the XB peptide. A stoichiometry of ⵑ1 XB peptide per N domain dimer was observed in binding experiments performed at high concentrations (Figure 4B). A similar value of 1.1 XB peptide per N domain dimer was obtained in isothermal titration calorimetry (ITC) experiments using the His6tagged N domain of ClpX (Figures 4C and 4D). Because untagged and His6-tagged N domain dimers, purified by different methods, were used for the two experiments, it seems unlikely that the reduced stoichiometry results from an equal mixture of fully active and inactive protein in both cases. The N domain forms a symmetric homodimer (Donaldson et al., 2003). Therefore, any tethering site present in one subunit should also be present in the other subunit, and there should be two equivalent
Designed SspB Heterodimers 447
Figure 4. Interactions between the N Domain of ClpX and XB Modules of SspB (A) Equilibrium binding of untagged N domain or ClpX to fluorescent-XB peptide assayed by changes in anisotropy at 20⬚C (pH 7.6, 200 mM KCl, 10 mM Mg2Cl, 10 mM ATP␥S, 2 mM DTT, 10% glycerol). ClpX binding data is from Wah et al. (2003). (B) Stoichiometric binding of untagged N domain to XB peptide assayed by fluorescence anisotropy at 4⬚C (pH 7.6, 100 mM KCl). Theoretical curves for binding stoichiometries of 1 or 2 XB peptides per N domain dimer are shown. (C) Binding assayed by ITC. Aliquots (7.5 l) of an XB peptide (1.1 mM) were injected into a 1.4 ml solution containing the His-tagged N domain dimer of E. coli ClpX (80 M dimer) at 25⬚C (pH 7.6, 50 mM KCl). (D) Single-species fit of the data shown in (C) gave the KD, ⌬H, ⌬S, and n values listed. (E) Cartoon representation of SspB, one bound ssrA-tagged substrate, and ClpX, showing how interactions between the XB modules and different ClpX N domains could be shuffled. In populations, the doubly tethered molecules are favored by roughly 10:1. (F) Hierarchy of apparent ClpX affinities (estimated from KM or Ki values) of SspB molecules with different numbers of XB modules and with or without bound ssrA-tagged substrates. Two XB tethering interactions as well as a bound ssrA-tagged substrate are required for strong binding.
binding sites for the XB peptide. We assume that the observed half-of-the-sites binding reactivity occurs because the tethering sites in both subunits overlap the 2-fold axis of the N domain, and thus binding of one XB module occludes binding of a second. Discussion Computational protein design, facilitated by powerful search algorithms and improved energy functions (Desmet et al., 1992; Pierce et al., 2000), has advanced to the point that permit its use as a probe of structurefunction relationships (Shimaoka et al., 2000). Here, we have designed a pair of SspB sequences that form a
stable heterodimer but do not form homodimers and have used these variants to probe the importance of SspB dimerization in substrate delivery to the ClpXP protease. Designed heterodimers of a homing endonuclease (Chevalier et al., 2002) and the GCN4 coiledcoil (Havranek and Harbury, 2003) have also been reported recently. We repacked the SspB dimer interface, which contains two ␣ helices and two  strands, in an asymmetric manner and used negative design to disfavor homodimers. Unlike the studies of Havranek and Harbury (2003), however, the unfolded state was not modeled as part of this negative design. Our studies show that both tails of SspB with their XB modules are required for full function. Removal of
Molecular Cell 448
one SspB tail increases KM and decreases Vmax for ClpXP degradation of bound GFP-ssrA to values similar to free GFP-ssrA. Moreover, SspB variants with only one XB module are significantly poorer inhibitors of SspB-mediated degradation than the corresponding variants with two tails. These results demonstrate that both tails and XB modules of an SspB dimer engage ClpX during substrate delivery. The opposite conclusion was reached by Dougan and colleagues (2003), who argued that “complete restoration of the SspB/ClpXP system can already be achieved when only one SspB binding site is present within the ClpX hexamer… .” Their studies were based on the untested assumption that hexamers of wild-type ClpX and a mutant lacking the N domain interchange subunits in an unbiased fashion. Because the N domain forms a stable dimer (Donaldson et al., 2003; Wojtyra et al., 2003), it would contribute to both ClpX dimer and hexamer stability and interchange would not be unbiased. Our results do show that single-tailed SspB variants bind weakly to ClpX and can deliver GFPssrA, but the overall degradation activity was not significantly better than degradation of GFP-ssrA in the absence of SspB. Interestingly, single-tailed SspB variants bound ClpX with almost the same affinity as the isolated XB peptide in inhibition studies, indicating that the SspB XB module makes all of the energetically significant contacts with ClpX. The maximum rate of GFP-ssrA degradation was 25%–30% higher when this substrate was delivered to ClpXP by a two-tailed YFGM/SLA molecule than by the corresponding single-tailed variant. This finding can be explained most simply if bivalent tethering of SspB to ClpX increases the effective concentration of the bound substrate relative to its interaction site on ClpX compared with single-point tethering. This model makes intuitive sense. Imagine an object attached to one of the vertices of an equilateral triangle (approximating the positions of the three N dimers in the ClpX hexamer) by a flexible tether. This object could access a spherical volume. Now attach the object to two of the vertices of the triangle using identical flexible tethers. The object is now constrained to an ellipsoid with a volume smaller than that accessible to the singly tethered object. This decrease in accessible volume corresponds to an increase in effective concentration. Two groups have shown that deleting the N domain of ClpX prevents SspB stimulation (Wojtyra et al., 2003; Dougan et al., 2003), suggesting that this domain is required directly or indirectly for SspB interactions. Our results confirm a direct interaction and show that the N domain and intact ClpX bind the XB peptide equally well. Notably, however, only a single XB peptide binds to the dimeric N domain. This result indicates that ClpX hexamers contain just three tethering sites for the XB peptides of SspB. This conclusion combined with our finding that two SspB tails are required for full ClpXP stimulation is consistent with a model in which the two tails of the SspB dimer bind to two of the three tethering sites on ClpX, leaving one tethering site unoccupied (Figure 4E). Hence, in agreement with the observed stoichiometry of binding (Wah et al., 2002), only a single SspB dimer would form a stable complex with a single ClpX hexamer. Binding of the XB modules of SspB to ClpX is probably
a dynamic process in which single tails are released but normally rebind before the second tethering contact is broken. Indeed, the approximate 10-fold difference in the ClpX binding affinity of two-tailed and single-tailed SspB variants suggests that one tail is disengaged roughly 10% of the time. By this model, the SspB tails would constantly be detaching from and reattaching to the three N domain tethering sites of ClpX during substrate delivery (Figure 4E). This type of dynamic shuffling of XB tails and tethering sites may aid in substrate delivery. SspB-mediated delivery of ssrA-tagged substrates to ClpXP involves the use of multiple weak interactions to generate a specific interaction of significantly stronger avidity. As illustrated in Figure 4F, a single XB tethering interaction between SspB and ClpX is weak (ⵑ20 M). Moreover, ClpX recognition of the ssrA tag bound to the substrate binding domain of SspB is also very weak (⬎20 M), as shown by the feeble activity of tail-less SspB variants in substrate delivery. Nevertheless, when both XB tethering interactions are made and the SspBbound ssrA-tagged substrate also interacts with ClpX, then the overall affinity of complex formation increases to roughly 0.2 M. This strategy of coupling a number of weak interactions would result in a dynamic system because each individual contact could be broken relatively easily. Moreover, the use of several weak interactions which need to be coupled for strong binding would also permit substrate delivery to be regulated by blocking any of the weak contacts between the delivery complex and ClpX. The XB binding site in the N domain of ClpX is also used to tether other adaptors. For example, degradation of UmuD/D⬘ by ClpXP requires tethering by the UmuD subunit which can be blocked by the SspB XB peptide (Neher et al., 2003). Moreover, the RssB adaptor—which delivers S to ClpXP (Zhou and Gottesman, 1998)—has a sequence similar to the XB module of SspB (Dougan et al., 2003). RssB delivery activity requires the N domain of ClpX and can also be inhibited by the SspB XB peptide (S. Siddiqui, personal communication). The existence of one tethering site that is shared by different adaptors sets up the possibility of competition when adaptorsubstrate complexes are present in excess. Our results show that dimeric adaptors like SspB, which utilize two ClpX tethering sites, compete more effectively for ClpXP than those that use only one tethering interaction of comparable strength. Hence, the number and strength of these tethering interactions, as well as the accessibility and strength of the degradation signal in the substrate bound to the adaptor, would all be important factors in prioritizing the intracellular degradation of different substrates. Experimental Procedures Details of the design of the YGFM/SLA SspB heterodimer, using a modified version of the ORBIT software (Dahiyat and Mayo, 1997), will be presented elsewhere. HE SspB contained residues 1–106 of H. influenza SspB and residues 108–165 of E. coli SspB. The YGFM mutations, SLA substitutions, and the A73Q mutation were introduced as appropriate into a plasmid-borne gene for HE SspB or a gene encoding the substrate binding domain of H. influenza SspB (residues 1–111) by PCR mutagenesis. To aid in purification, a His6
Designed SspB Heterodimers 449
tag was added to the N terminus of some SspB constructs. Plasmid encoding E. coli ClpX N domain residues 1–61 or 1–64 with a C-terminal EH6 tag were gifts from I. Levchenko and R. Burton. Details of strain construction and protein purification are available upon request. E. coli SspB and its SBD, GFP-ssrA, E. coli ClpX, E. coli ClpP, and fluorescein-labeled ssrA and XB peptides were purified or prepared as described (Wah et al., 2002, 2003). The synthetic XB peptide used for ITC was NH2-CYRGGRPALRVVK-COOH. Peptide binding assays, degradation assays, and biophysical characterization by spectroscopy, sedimentation, and denaturation were performed essentially as described (Wah et al., 2002, 2003). For exchange experiments to monitor stability, equal concentrations of untagged SspB or variants (one subunit with and one without a tail) were mixed in GuHCl at 20⬚C and allowed to refold after desalting into 25 mM HEPES (pH 7.6), 1 mM EDTA, 100 mM KCl. SBD/full-length heterodimers were purified by monoQ chromatography at 4⬚C. Half the sample was rechromatographed immediately; the rest was incubated at 30⬚C for 24 hr and rechromatographed. Other exchange experiments were assayed by electrophoresis (150 V, 2 hr) on gels containing 11% acrylamide (37.5:1 acrylamide:bisacrylamide) in 30 mM HEPES-imidazole (pH 7.5), 1 mM EDTA. Acknowledgments We thank R. Burton, I. Levchenko, S. Mayo, and S. Siddiqui for materials, software, and communication of unpublished results. This work was supported in part by grants from the NIH (AI-16892, AI15706), NSF (0070319), and Howard Hughes Medical Institute (HHMI). D.A.W. was a Cissy Hornung American Cancer Society Postdoctoral Fellow. T.A.B. is an employee of HHMI. Received: November 10, 2003 Revised: December 5, 2003 Accepted: December 9, 2003 Published: February 12, 2004 References Chevalier, B.S., Kortemme, T., Chadsey, M.S., Baker, D., Monnat, R.J., and Stoddard, B.L. (2002). Design, activity, and structure of a highly specific artificial endonuclease. Mol. Cell 10, 895–905. Dahiyat, B.I., and Mayo, S.L. (1997). De novo protein design: fully automated sequence selection. Science 278, 82–87. Desmet, J., Maeyer, M.D., Hazes, B., and Lasters, I. (1992). The dead-end elimination theorem and its use in protein side-chain positioning. Nature 356, 539–542. Donaldson, L.W., Wojtyra, U., and Houry, W.A. (2003). Solution structure of the dimeric zinc binding domain of the chaperone ClpX. J. Biol. Chem. 278, 48991–48996. Dougan, D.A., Weber-Ban, E., and Bukau, B. (2003). Targeted delivery of an ssrA-tagged substrate by the adaptor protein SspB to its cognate AAA⫹ protein ClpX. Mol. Cell 12, 373–380. Flynn, J.M., Neher, S.B., Kim, Y.I., Sauer, R.T., and Baker, T.A. (2003). Proteomic discovery of cellular substrates of the ClpXP protease reveals five classes of ClpX-recognition signals. Mol. Cell 11, 671–683. Gottesman, S., Roche, E., Zhou, Y., and Sauer, R.T. (1998). The ClpXP and ClpAP proteases degrade proteins with carboxy-terminal peptide tails added by the SsrA-tagging system. Genes Dev. 12, 1338–1347. Havranek, J.J., and Harbury, P.B. (2003). Automated design of specificity in molecular recognition. Nat. Struct. Biol. 10, 45–52. Hoskins, J.R., Sharma, S., Sathyanarayana, B.K., and Wickner, S. (2001). Clp ATPases and their role in protein unfolding and degradation. Adv. Protein Chem. 59, 413–429. Keiler, K.C., Waller, P.R., and Sauer, R.T. (1996). Role of a peptide tagging system in degradation of proteins synthesized from damaged messenger RNA. Science 271, 990–993. Levchenko, I., Seidel, M., Sauer, R.T., and Baker, T.A. (2000). A specificity-enhancing factor for the ClpXP degradation machine. Science 289, 2354–2356.
Levchenko, I., Grant, R.A., Wah, D.A., Sauer, R.T., and Baker, T.A. (2003). Structure of a delivery protein for an AAA⫹ protease in complex with a peptide degradation tag. Mol. Cell 12, 365–372. Neher, S.B., Sauer, R.T., and Baker, T.A. (2003). Distinct peptide signals in the UmuD and UmuD⬘ subunits of UmuD/D⬘ mediate tethering and substrate-processing by the ClpXP protease. Proc. Natl. Acad. Sci. USA 100, 13219–13224. Pierce, N.A., Spriet, J.A., Desmet, J., and Mayo, S.L. (2000). Conformational splitting: a more powerful criterion for dead-end elimination. J. Comput. Chem. 21, 999–1009. Shimaoka, M., Shifman, J.M., Jing, H., Takagi, J., Mayo, S.L., and Springer, T.A. (2000). Computational design of an integrin I domain stabilized in the open high affinity conformation. Nat. Struct. Biol. 7, 614–616. Song, H.K., and Eck, M.J. (2003). Structural basis of degradation signal recognition by SspB, a specificity-enhancing factor for the ClpXP proteolytic machine. Mol. Cell 12, 75–86. Wah, D.A., Levchenko, I., Baker, T.A., and Sauer, R.T. (2002). Characterization of a specificity factor for an AAA⫹ ATPase: assembly of SspB dimers with ssrA-tagged proteins and the ClpX hexamer. Chem. Biol. 9, 1237–1245. Wah, D.A., Levchenko, I., Rieckhof, G.E., Bolon, D.N., Baker, T.A., and Sauer, R.T. (2003). Flexible linkers leash the substrate binding domain of SspB to a peptide module that stabilizes delivery complexes with the AAA⫹ ClpXP protease. Mol. Cell 12, 355–363. Wojtyra, U.A., Thibault, G., Tuite, A., and Houry, W.A. (2003). The N-terminal zinc binding domain of ClpX is a dimerization domain that modulates the chaperone function. J. Biol. Chem. 278, 48981–48990. Zhou, Y., and Gottesman, S. (1998). Regulation of proteolysis of the stationary-phase sigma factor RpoS. J. Bacteriol. 180, 1154–1158.
Bivalent Tethering of SspB to ClpXP Is Required for Efficient Substrate Delivery: A Protein-Design Study Daniel N. Bolon,1 David A. Wah,1 Greg L. Hersch,1 Tania A. Baker,1,2 and Robert T. Sauer1,* 1 Department of Biology and 2 Howard Hughes Medical Institute Massachusetts Institute of Technology Cambridge, Massachusetts 02139
Summary SspB homodimers deliver ssrA-tagged substrates to ClpXP for degradation. SspB consists of a substrate binding domain and an unstructured tail with a ClpX binding module (XB). Using computational design, we engineered an SspB heterodimer whose subunits did not form homodimers. Experiments with the designed molecule and variants lacking one or two tails demonstrate that both XB modules are required for strong binding and efficient substrate delivery to ClpXP. Assembly of stable SspB-substrate-ClpX delivery complexes requires the coupling of weak tethering interactions between ClpX and the SspB XB modules as well as interactions between ClpX and the substrate degradation tag. The ClpX hexamer contains three XB binding sites, one per N domain dimer, and thus binds strongly to just one SspB dimer at a time. Because different adaptor proteins use the same tethering sites in ClpX, those which employ bivalent tethering, like SspB, will compete more effectively for substrate delivery to ClpXP.
Short Article
hexamer, and experiments mixing full-length and N-terminal truncations of ClpX have been interpreted by Dougan et al. (2003) as evidence that SspB binding and efficient substrate delivery require only a single N domain of ClpX. This result suggests that as many as six SspB dimers could bind ClpX and predicts that only one XB module per SspB dimer should be needed for strong binding and substrate delivery to ClpXP. However, other studies indicate that substrate-delivery complexes contain just one SspB dimer and two ssrA-tagged substrates per ClpX hexamer (Wah et al., 2002), suggesting that more than one dimer of SspB does not interact with ClpX. Does an SspB dimer with only a single XB module mediate efficient substrate delivery to ClpXP? To address this question, we used computational design to engineer SspB heterodimers with different numbers of XB sequences and assayed these proteins for enhancement of ClpXP degradation. SspB molecules with only one XB module were found to be significantly less active in substrate delivery than those with two modules. We also found that the ClpX N domain contains the tethering sites for binding SspB. However, one dimeric N domain bound only a single XB peptide, indicating that a ClpX hexamer contains just three tethering sites for the XB module of SspB, and thus could bind strongly to only one SspB dimer at a time. These results shed light on the detailed mechanism of substrate delivery by SspB and help us understand how different intracellular adaptors might compete for ClpXP, thereby determining the priority of substrate selection.
Introduction Results In the AAA⫹ protease ClpXP, the hexameric ClpX ATPase binds protein substrates and unfolds and translocates these molecules into the degradation chamber of the ClpP peptidase (Hoskins et al., 2001). ClpX recognizes target proteins via peptide signals such as the ssrA tag, an 11 residue sequence added to the C terminus of nascent proteins on stalled ribosomes (Keiler et al., 1996; Gottesman et al., 1998; Flynn et al., 2003). SspB, a dimeric adaptor protein, also binds ssrA-tagged proteins and enhances their degradation by ClpXP (Levchenko et al., 2000; Wah et al., 2002). Each SspB subunit contains a substrate binding domain (SBD) and a flexible C-terminal tail (Levchenko et al., 2003; Song and Eck, 2003; Wah et al., 2003). An XB module at the end of each SspB tail mediates tethering interactions with ClpX (Wah et al., 2003), perhaps by binding to a site in the ClpX N-terminal domain (Wojtyra et al., 2003; Dougan et al., 2003). The mechanism by which SspB stimulates ClpXP degradation of ssrA-tagged substrates is only partially understood. It is clear that the tails of SspB form tethering interactions with ClpX, increasing the local concentration of ssrA-tagged substrates relative to the enzyme. In principle, several SspB dimers could bind one ClpX *Correspondence: [email protected]
Heterodimer Design and Properties The SBD of H. influenza SspB is a symmetric dimer (Levchenko et al., 2003), but many studies reported below required us to purify and assay heterodimeric forms of SspB. To identify sequence variants that would form heterodimers rather than homodimers, we employed computational design. Using the ORBIT software package (Dahiyat and Mayo, 1997), we repacked the dimer interface of SspB in an asymmetric fashion using a modified search algorithm to optimize the energy of the heterodimer relative to homodimers of either partner sequence. The top-ranked heterodimer design resulting from this optimization contained seven mutations relative to the wild-type SspB dimer: L12Y, A15G, Y16F, and V101M in one subunit (called YGFM) and A15S, Y16L, and V101A in the partner subunit (called SLA). Genes encoding the H. influenza YGFM and SLA SBDs with C-terminal tails from E. coli SspB were constructed. E. coli SspB and H. influenza SspB have similar SBD structures (Levchenko et al., 2003; Song and Eck, 2003), and a hybrid (HE SspB) containing the wild-type H. influenza SBD with E. coli tails delivers GFP-ssrA to E. coli ClpXP as well as both parental SspB molecules (data not shown). Following coexpression, the YGFM/SLA protein purified as a single species containing both mutant sub-
Molecular Cell 444
Figure 1. Properties of the Designed Heterodimer (A) Analytical ultracentrifugation (20⬚C; 16,000 rpm) of the YGFM/SLA protein (25 M) after 36 hr. The fitted solid line represents a value of 37.6 kDa for the heterodimer molecular weight. Dashed line, mass distribution expected for a monomer. (B) Far-UV circular dichroism spectra of 25 M YGFM/SLA or the parental HE SspB protein (25⬚C). (C) Fluorescence spectra of YGFM/SLA under native and denaturing (5 M GuHCl) conditions (25⬚C). (D) GuHCl denaturation of 3 M YGFM/SLA (⌬Gu⫽16.7 kcal/mol; m ⫽ 5.7 kcal/mol•M) and HE SspB (⌬Gu ⫽ 22.4 kcal/mol; m ⫽ 6.7 kcal/ mol•M) at 25⬚C. The fitted curves and stability parameters are for native dimer to unfolded monomer reactions. (E) YGFM/SLA heterodimer and E. coli SspB bound equally well to a fluorescent ssrA peptide (50 nM) as assayed by changes in anisotropy at 30⬚C, pH 7.6, 200 mM KCl. The solid line represents a KD of 3 M. (F) Degradation of GFP-ssrA (0.6 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) in the absence of SspB or in the presence of YGFM/ SLA or E. coli SspB (0.6 M each). The solid lines are linear fits to the kinetic data. Degradation in the presence of HE SspB was identical to that observed with YGFM/SLA (data not shown).
units, and analytical equilibrium ultracentrifugation experiments (Figure 1A) gave a molecular weight (37.6 kDa) close to the expected heterodimeric value (37.9 kDa). The circular-dichroism spectrum of YGFM/SLA (Figure 1B) was similar to that of the parent molecule, indicating similar secondary structures. Native YGFM/SLA also had a fluorescence spectrum that was blue-shifted relative to denatured protein (Figure 1C), showing that both tryptophans in the heterodimer interface are buried. In denaturation experiments (Figure 1D), YGFM/SLA displayed a single cooperative unfolding transition with stability decreased only modestly compared to the HE SspB dimer. These experiments show that the designed YGFM/SLA molecule forms a stably-folded dimer with structural properties similar to the SspB homodimer. The design of the YGFM/SLA dimer would not be considered a success unless it bound ssrA-tagged substrates and delivered them to ClpXP for degradation. A fluorescent ssrA peptide bound equally well to YGFM/ SLA and wild-type E. coli SspB (Figure 1E), showing that redesign of the dimer interface does not disturb ssrA
binding. YGFM/SLA was also essentially as active as wild-type E. coli SspB in stimulating the degradation of ssrA-tagged substrates by ClpXP (Figure 1F). We conclude that the designed YGFM/SLA protein is fully functional in substrate delivery. Heterodimers versus Homodimers When the SLA or YGFM subunits were purified to homogeneity under denaturing conditions and then allowed to refold, most protein precipitated suggesting an inability of both mutant homodimers to fold stably (data not shown). To compare the stabilities of different dimeric combinations of natural SspB with the designed heterodimer, we denatured/renatured an equal mixture of wildtype E. coli SspB and its SBD and purified the dimer containing one wild-type SBD and one wild-type fulllength subunit by ion-exchange chromatography under conditions where subunit exchange is slow. Figure 2A shows the elution profile of this purified protein following rechromatography. After 24 hr of incubation at 30⬚C, however, ion exchange revealed three peaks corre-
Designed SspB Heterodimers 445
Figure 2. Subunit Exchange Assayed by MonoQ Ion-Exchange Chromatography (A) Rechromatography of a dimer containing one full-length and one substrate binding domain of E. coli SspB immediately following the initial purification of this peak on the same ion-exchange column. (B) Rechromatography of the material from (A) following 24 hr of incubation at 30⬚C. (C) Rechromatography of a YGFM/SLA heterodimer containing one full-length and one substrate binding domain following initial purification. (D) Rechromatography of the material from (C) following 24 hr of incubation at 30⬚C. In all experiments, ion-exchange chromatography was performed at 4⬚C and was monitored by absorbance at 280 nm. The identities of the subunits in each peak were determined by SDS-PAGE.
sponding to the SBD homodimer, the SBD/SspB heterodimer, and the SspB homodimer in ratios (1:2:1) expected for unbiased equilibrium redistribution of these subunits (Figure 2B). This experiment was repeated using a heterodimer consisting of a YGFM SBD and a fulllength SLA subunit (Figure 2C). Following 24 hr, only the peak corresponding to the original YGFM-SBD/SLA heterodimer was detected (Figure 2D). SDS-PAGE confirmed that this peak contained both heterodimeric partners. To exclude the possibility of slow exchange, the YGFM-SBD/SLA heterodimer was denatured and allowed to refold. Again, only the single YGFM-SBD/SLA heterodimer peak was present (data not shown). We conclude that subunits containing the YGFM and SLA mutations form a heterodimer that must be significantly more stable than either mutant homodimer. Activity of One-Tailed SspB Variants To test the importance of the number of SspB XB modules in substrate delivery, we purified variants with two (YGFM/SLA), one (YGFM/SLA-SBD), or zero (YGFMSBD/SLA-SBD) tails and assayed ClpXP degradation
rates at different concentrations of the heterodimer and GFP-ssrA (Figure 3A). The heterodimer with two tails and XB modules was more active than the single-tailed species, which in turn was more active than the molecule with no tails. To confirm that this result was not an artifact of the heterodimer design, we also assayed wildtype E. coli SspB variants containing two, one, or zero tails. As in the YGFM/SLA background, two XB tails were required for full enhancement of ClpXP degradation of GFP-ssrA (Figure 3B). The otherwise wild-type one-tailed molecule described above might generate molecules with two tails by exchange during the degradation assay. Native gel electrophoresis showed that the exchange half-life of wild-type E. coli SspB dimers is about 30 min at 30⬚C (Figure 3C), whereas ClpXP degradation rates were determined during a 5 min reaction at 30⬚C. Thus, only a small amount of subunit exchange (ⵑ10%) probably occurred in this experiment. Nevertheless, this problem of exchange of wild-type subunits highlights the importance of being able to use the designed YFGM/SLA heterodimer, which precludes homodimer formation. The results shown in Figures 3A and 3B can be viewed in two ways. First, two tails and XB sequences are clearly required for full SspB activity. This is true with respect to both the KM and Vmax for ClpXP degradation. Second, dimers with one tail are more active than those with no tails, and thus a single tail must permit some interaction with ClpX. It is important to recall, however, that SspB is not required for ClpXP degradation of GFP-ssrA but, when present, decreases KM and increases Vmax. In fact, KM and Vmax values determined previously for ClpXP degradation of free GFP-ssrA (Levchenko et al., 2000; Wah et al., 2002) were similar to the values determined here for degradation of GFP-ssrA bound to single-tailed SspB variants. Hence, single-tailed SspB dimers should not stimulate degradation of GFP-ssrA to any appreciable extent. This prediction was confirmed. GFP-ssrA with no added SspB was degraded at about the same rate as in the presence of single-tailed YGFM/SLA, whereas degradation in the presence of the two-tailed variant was stimulated significantly (Figure 3D). GFP-ssrA bound to tail-less SspB interacts very inefficiently with ClpX (Wah et al., 2003; Figures 3A and 3B), suggesting that the SspB SBD makes unfavorable interactions with ClpX or masks recognition determinants in the bound ssrA tag. We assume that the single-tailed constructs allow weak tethering to ClpX (see below), which offsets the inhibitory effect of binding of ssrA-tagged substrates to the SBD, resulting in a substrate with properties similar to those of free GFP-ssrA. Binding of One- and Two-Tailed SspBs to ClpX To allow another probe of ClpX interaction, we designed an SspB binding-pocket mutation (A73Q) to block ssrAtag binding. A73Q variants folded and dimerized normally but did not bind ssrA peptide (KD ⬎ 500 M; data not shown). A73Q variants of YGFM/SLA heterodimers with zero, one, or two tails were used to compete for degradation of GFP-ssrA by ClpXP in the presence of wild-type E. coli SspB. As shown in Figure 3E, the twotailed A73Q heterodimer (Ki ⬇ 2.5 M) was a significantly better competitor than the single-tailed molecule (Ki ⬇
Molecular Cell 446
Figure 3. Two SspB Tails Are Required for Full Activity and Strong Binding to ClpX (A) Rate of ClpXP degradation of GFP-ssrA as a function of substrate and SspB concentration for YGFM/SLA heterodimers with two, one, or zero tails and XB modules. (B) Rate of ClpXP degradation of GFP-ssrA as a function of substrate and SspB concentration for E. coli SspB heterodimers with two, one, or zero tails and XB modules. (C) Kinetics of subunit exchange assayed by native gel electrophoresis. For E. coli SspB, a heterodimer containing one full-length subunit and one SBD subunit was incubated at 30⬚C, and samples were taken after different incubation times and placed on ice until electrophoresis on a native gel. The figure shows the approach to the equilibrium 1:2:1 distribution of subunits. When a YGFM/SLA heterodimer containing full-length subunits was mixed with His6-tagged E. coli SspB, no subunit exchange was detected by native gel following 24 hr of incubation at 30⬚C. (D) Degradation of GFP-ssrA (0.3 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) in the absence of SspB or in the presence of designed YGFM/SLA heterodimers with two, one, or zero tails/XB modules (0.3 M each). The solid lines are linear least squares fits to the kinetic data. (E) Degradation of GFP-ssrA (0.2 M) in the presence of E. coli SspB (0.6 M) by ClpXP (0.1 M ClpX6, 0.3 M ClpP14) was inhibited by increasing concentrations of the XB peptide or the A73Q variants of YGFM/SLA heterodimers containing one or two C-terminal tails/XB modules. The A73Q mutation prevents binding of SspB to ssrA-tagged substrates. (F) The A73Q variants of YGFM/SLA at concentrations of 55 M did not significantly inhibit ClpXP degradation of GFP-ssrA alone. Degradation conditions were identical to those in (E) except E. coli SspB was not added.
20 M). These results confirm that two tails are important for strong binding of SspB dimers to ClpX. Importantly, inhibition by the single-tailed A73Q heterodimer was essentially the same as inhibition by an isolated XB peptide (Figure 3E). This result indicates that the XB module of SspB makes all of the energetically significant interactions with ClpX. The A73Q variant of tail-less YGFM/SLA did not inhibit SspB-mediated degradation of GFP-ssrA by ClpXP (data not shown), and exchange between YGFM/SLA subunits and E. coli SspB subunits was not detected (Figure 3C). Both results indicate that the inhibition observed in Figure 3E occurs by competition for the tethering sites on ClpX and not by mixing of SspB subunits. Interestingly, the A73Q variants had little effect on the degradation of GFP-ssrA alone (Figure 3F). Interactions of SspB with the ClpX N Domain To assay for potential interactions with the XB peptide of SspB, N domain fragments of ClpX (residues 1–61 or His6-tagged 1–64) were purified and studied. Consistent with a recent report (Wojtyra et al., 2003), the untagged
N domain formed a stable dimer as assayed by analytical equilibrium centrifugation (data not shown). As monitored by fluorescence anisotropy (Figure 4A), the XB peptide bound the untagged N domain with a KD (20 ⫾ 5 M) within error of that for intact ClpX (23 ⫾ 7 M). We conclude that the N domain mediates all of the energetically significant contacts between ClpX and the XB peptide. A stoichiometry of ⵑ1 XB peptide per N domain dimer was observed in binding experiments performed at high concentrations (Figure 4B). A similar value of 1.1 XB peptide per N domain dimer was obtained in isothermal titration calorimetry (ITC) experiments using the His6tagged N domain of ClpX (Figures 4C and 4D). Because untagged and His6-tagged N domain dimers, purified by different methods, were used for the two experiments, it seems unlikely that the reduced stoichiometry results from an equal mixture of fully active and inactive protein in both cases. The N domain forms a symmetric homodimer (Donaldson et al., 2003). Therefore, any tethering site present in one subunit should also be present in the other subunit, and there should be two equivalent
Designed SspB Heterodimers 447
Figure 4. Interactions between the N Domain of ClpX and XB Modules of SspB (A) Equilibrium binding of untagged N domain or ClpX to fluorescent-XB peptide assayed by changes in anisotropy at 20⬚C (pH 7.6, 200 mM KCl, 10 mM Mg2Cl, 10 mM ATP␥S, 2 mM DTT, 10% glycerol). ClpX binding data is from Wah et al. (2003). (B) Stoichiometric binding of untagged N domain to XB peptide assayed by fluorescence anisotropy at 4⬚C (pH 7.6, 100 mM KCl). Theoretical curves for binding stoichiometries of 1 or 2 XB peptides per N domain dimer are shown. (C) Binding assayed by ITC. Aliquots (7.5 l) of an XB peptide (1.1 mM) were injected into a 1.4 ml solution containing the His-tagged N domain dimer of E. coli ClpX (80 M dimer) at 25⬚C (pH 7.6, 50 mM KCl). (D) Single-species fit of the data shown in (C) gave the KD, ⌬H, ⌬S, and n values listed. (E) Cartoon representation of SspB, one bound ssrA-tagged substrate, and ClpX, showing how interactions between the XB modules and different ClpX N domains could be shuffled. In populations, the doubly tethered molecules are favored by roughly 10:1. (F) Hierarchy of apparent ClpX affinities (estimated from KM or Ki values) of SspB molecules with different numbers of XB modules and with or without bound ssrA-tagged substrates. Two XB tethering interactions as well as a bound ssrA-tagged substrate are required for strong binding.
binding sites for the XB peptide. We assume that the observed half-of-the-sites binding reactivity occurs because the tethering sites in both subunits overlap the 2-fold axis of the N domain, and thus binding of one XB module occludes binding of a second. Discussion Computational protein design, facilitated by powerful search algorithms and improved energy functions (Desmet et al., 1992; Pierce et al., 2000), has advanced to the point that permit its use as a probe of structurefunction relationships (Shimaoka et al., 2000). Here, we have designed a pair of SspB sequences that form a
stable heterodimer but do not form homodimers and have used these variants to probe the importance of SspB dimerization in substrate delivery to the ClpXP protease. Designed heterodimers of a homing endonuclease (Chevalier et al., 2002) and the GCN4 coiledcoil (Havranek and Harbury, 2003) have also been reported recently. We repacked the SspB dimer interface, which contains two ␣ helices and two  strands, in an asymmetric manner and used negative design to disfavor homodimers. Unlike the studies of Havranek and Harbury (2003), however, the unfolded state was not modeled as part of this negative design. Our studies show that both tails of SspB with their XB modules are required for full function. Removal of
Molecular Cell 448
one SspB tail increases KM and decreases Vmax for ClpXP degradation of bound GFP-ssrA to values similar to free GFP-ssrA. Moreover, SspB variants with only one XB module are significantly poorer inhibitors of SspB-mediated degradation than the corresponding variants with two tails. These results demonstrate that both tails and XB modules of an SspB dimer engage ClpX during substrate delivery. The opposite conclusion was reached by Dougan and colleagues (2003), who argued that “complete restoration of the SspB/ClpXP system can already be achieved when only one SspB binding site is present within the ClpX hexamer… .” Their studies were based on the untested assumption that hexamers of wild-type ClpX and a mutant lacking the N domain interchange subunits in an unbiased fashion. Because the N domain forms a stable dimer (Donaldson et al., 2003; Wojtyra et al., 2003), it would contribute to both ClpX dimer and hexamer stability and interchange would not be unbiased. Our results do show that single-tailed SspB variants bind weakly to ClpX and can deliver GFPssrA, but the overall degradation activity was not significantly better than degradation of GFP-ssrA in the absence of SspB. Interestingly, single-tailed SspB variants bound ClpX with almost the same affinity as the isolated XB peptide in inhibition studies, indicating that the SspB XB module makes all of the energetically significant contacts with ClpX. The maximum rate of GFP-ssrA degradation was 25%–30% higher when this substrate was delivered to ClpXP by a two-tailed YFGM/SLA molecule than by the corresponding single-tailed variant. This finding can be explained most simply if bivalent tethering of SspB to ClpX increases the effective concentration of the bound substrate relative to its interaction site on ClpX compared with single-point tethering. This model makes intuitive sense. Imagine an object attached to one of the vertices of an equilateral triangle (approximating the positions of the three N dimers in the ClpX hexamer) by a flexible tether. This object could access a spherical volume. Now attach the object to two of the vertices of the triangle using identical flexible tethers. The object is now constrained to an ellipsoid with a volume smaller than that accessible to the singly tethered object. This decrease in accessible volume corresponds to an increase in effective concentration. Two groups have shown that deleting the N domain of ClpX prevents SspB stimulation (Wojtyra et al., 2003; Dougan et al., 2003), suggesting that this domain is required directly or indirectly for SspB interactions. Our results confirm a direct interaction and show that the N domain and intact ClpX bind the XB peptide equally well. Notably, however, only a single XB peptide binds to the dimeric N domain. This result indicates that ClpX hexamers contain just three tethering sites for the XB peptides of SspB. This conclusion combined with our finding that two SspB tails are required for full ClpXP stimulation is consistent with a model in which the two tails of the SspB dimer bind to two of the three tethering sites on ClpX, leaving one tethering site unoccupied (Figure 4E). Hence, in agreement with the observed stoichiometry of binding (Wah et al., 2002), only a single SspB dimer would form a stable complex with a single ClpX hexamer. Binding of the XB modules of SspB to ClpX is probably
a dynamic process in which single tails are released but normally rebind before the second tethering contact is broken. Indeed, the approximate 10-fold difference in the ClpX binding affinity of two-tailed and single-tailed SspB variants suggests that one tail is disengaged roughly 10% of the time. By this model, the SspB tails would constantly be detaching from and reattaching to the three N domain tethering sites of ClpX during substrate delivery (Figure 4E). This type of dynamic shuffling of XB tails and tethering sites may aid in substrate delivery. SspB-mediated delivery of ssrA-tagged substrates to ClpXP involves the use of multiple weak interactions to generate a specific interaction of significantly stronger avidity. As illustrated in Figure 4F, a single XB tethering interaction between SspB and ClpX is weak (ⵑ20 M). Moreover, ClpX recognition of the ssrA tag bound to the substrate binding domain of SspB is also very weak (⬎20 M), as shown by the feeble activity of tail-less SspB variants in substrate delivery. Nevertheless, when both XB tethering interactions are made and the SspBbound ssrA-tagged substrate also interacts with ClpX, then the overall affinity of complex formation increases to roughly 0.2 M. This strategy of coupling a number of weak interactions would result in a dynamic system because each individual contact could be broken relatively easily. Moreover, the use of several weak interactions which need to be coupled for strong binding would also permit substrate delivery to be regulated by blocking any of the weak contacts between the delivery complex and ClpX. The XB binding site in the N domain of ClpX is also used to tether other adaptors. For example, degradation of UmuD/D⬘ by ClpXP requires tethering by the UmuD subunit which can be blocked by the SspB XB peptide (Neher et al., 2003). Moreover, the RssB adaptor—which delivers S to ClpXP (Zhou and Gottesman, 1998)—has a sequence similar to the XB module of SspB (Dougan et al., 2003). RssB delivery activity requires the N domain of ClpX and can also be inhibited by the SspB XB peptide (S. Siddiqui, personal communication). The existence of one tethering site that is shared by different adaptors sets up the possibility of competition when adaptorsubstrate complexes are present in excess. Our results show that dimeric adaptors like SspB, which utilize two ClpX tethering sites, compete more effectively for ClpXP than those that use only one tethering interaction of comparable strength. Hence, the number and strength of these tethering interactions, as well as the accessibility and strength of the degradation signal in the substrate bound to the adaptor, would all be important factors in prioritizing the intracellular degradation of different substrates. Experimental Procedures Details of the design of the YGFM/SLA SspB heterodimer, using a modified version of the ORBIT software (Dahiyat and Mayo, 1997), will be presented elsewhere. HE SspB contained residues 1–106 of H. influenza SspB and residues 108–165 of E. coli SspB. The YGFM mutations, SLA substitutions, and the A73Q mutation were introduced as appropriate into a plasmid-borne gene for HE SspB or a gene encoding the substrate binding domain of H. influenza SspB (residues 1–111) by PCR mutagenesis. To aid in purification, a His6
Designed SspB Heterodimers 449
tag was added to the N terminus of some SspB constructs. Plasmid encoding E. coli ClpX N domain residues 1–61 or 1–64 with a C-terminal EH6 tag were gifts from I. Levchenko and R. Burton. Details of strain construction and protein purification are available upon request. E. coli SspB and its SBD, GFP-ssrA, E. coli ClpX, E. coli ClpP, and fluorescein-labeled ssrA and XB peptides were purified or prepared as described (Wah et al., 2002, 2003). The synthetic XB peptide used for ITC was NH2-CYRGGRPALRVVK-COOH. Peptide binding assays, degradation assays, and biophysical characterization by spectroscopy, sedimentation, and denaturation were performed essentially as described (Wah et al., 2002, 2003). For exchange experiments to monitor stability, equal concentrations of untagged SspB or variants (one subunit with and one without a tail) were mixed in GuHCl at 20⬚C and allowed to refold after desalting into 25 mM HEPES (pH 7.6), 1 mM EDTA, 100 mM KCl. SBD/full-length heterodimers were purified by monoQ chromatography at 4⬚C. Half the sample was rechromatographed immediately; the rest was incubated at 30⬚C for 24 hr and rechromatographed. Other exchange experiments were assayed by electrophoresis (150 V, 2 hr) on gels containing 11% acrylamide (37.5:1 acrylamide:bisacrylamide) in 30 mM HEPES-imidazole (pH 7.5), 1 mM EDTA. Acknowledgments We thank R. Burton, I. Levchenko, S. Mayo, and S. Siddiqui for materials, software, and communication of unpublished results. This work was supported in part by grants from the NIH (AI-16892, AI15706), NSF (0070319), and Howard Hughes Medical Institute (HHMI). D.A.W. was a Cissy Hornung American Cancer Society Postdoctoral Fellow. T.A.B. is an employee of HHMI. Received: November 10, 2003 Revised: December 5, 2003 Accepted: December 9, 2003 Published: February 12, 2004 References Chevalier, B.S., Kortemme, T., Chadsey, M.S., Baker, D., Monnat, R.J., and Stoddard, B.L. (2002). Design, activity, and structure of a highly specific artificial endonuclease. Mol. Cell 10, 895–905. Dahiyat, B.I., and Mayo, S.L. (1997). De novo protein design: fully automated sequence selection. Science 278, 82–87. Desmet, J., Maeyer, M.D., Hazes, B., and Lasters, I. (1992). The dead-end elimination theorem and its use in protein side-chain positioning. Nature 356, 539–542. Donaldson, L.W., Wojtyra, U., and Houry, W.A. (2003). Solution structure of the dimeric zinc binding domain of the chaperone ClpX. J. Biol. Chem. 278, 48991–48996. Dougan, D.A., Weber-Ban, E., and Bukau, B. (2003). Targeted delivery of an ssrA-tagged substrate by the adaptor protein SspB to its cognate AAA⫹ protein ClpX. Mol. Cell 12, 373–380. Flynn, J.M., Neher, S.B., Kim, Y.I., Sauer, R.T., and Baker, T.A. (2003). Proteomic discovery of cellular substrates of the ClpXP protease reveals five classes of ClpX-recognition signals. Mol. Cell 11, 671–683. Gottesman, S., Roche, E., Zhou, Y., and Sauer, R.T. (1998). The ClpXP and ClpAP proteases degrade proteins with carboxy-terminal peptide tails added by the SsrA-tagging system. Genes Dev. 12, 1338–1347. Havranek, J.J., and Harbury, P.B. (2003). Automated design of specificity in molecular recognition. Nat. Struct. Biol. 10, 45–52. Hoskins, J.R., Sharma, S., Sathyanarayana, B.K., and Wickner, S. (2001). Clp ATPases and their role in protein unfolding and degradation. Adv. Protein Chem. 59, 413–429. Keiler, K.C., Waller, P.R., and Sauer, R.T. (1996). Role of a peptide tagging system in degradation of proteins synthesized from damaged messenger RNA. Science 271, 990–993. Levchenko, I., Seidel, M., Sauer, R.T., and Baker, T.A. (2000). A specificity-enhancing factor for the ClpXP degradation machine. Science 289, 2354–2356.
Levchenko, I., Grant, R.A., Wah, D.A., Sauer, R.T., and Baker, T.A. (2003). Structure of a delivery protein for an AAA⫹ protease in complex with a peptide degradation tag. Mol. Cell 12, 365–372. Neher, S.B., Sauer, R.T., and Baker, T.A. (2003). Distinct peptide signals in the UmuD and UmuD⬘ subunits of UmuD/D⬘ mediate tethering and substrate-processing by the ClpXP protease. Proc. Natl. Acad. Sci. USA 100, 13219–13224. Pierce, N.A., Spriet, J.A., Desmet, J., and Mayo, S.L. (2000). Conformational splitting: a more powerful criterion for dead-end elimination. J. Comput. Chem. 21, 999–1009. Shimaoka, M., Shifman, J.M., Jing, H., Takagi, J., Mayo, S.L., and Springer, T.A. (2000). Computational design of an integrin I domain stabilized in the open high affinity conformation. Nat. Struct. Biol. 7, 614–616. Song, H.K., and Eck, M.J. (2003). Structural basis of degradation signal recognition by SspB, a specificity-enhancing factor for the ClpXP proteolytic machine. Mol. Cell 12, 75–86. Wah, D.A., Levchenko, I., Baker, T.A., and Sauer, R.T. (2002). Characterization of a specificity factor for an AAA⫹ ATPase: assembly of SspB dimers with ssrA-tagged proteins and the ClpX hexamer. Chem. Biol. 9, 1237–1245. Wah, D.A., Levchenko, I., Rieckhof, G.E., Bolon, D.N., Baker, T.A., and Sauer, R.T. (2003). Flexible linkers leash the substrate binding domain of SspB to a peptide module that stabilizes delivery complexes with the AAA⫹ ClpXP protease. Mol. Cell 12, 355–363. Wojtyra, U.A., Thibault, G., Tuite, A., and Houry, W.A. (2003). The N-terminal zinc binding domain of ClpX is a dimerization domain that modulates the chaperone function. J. Biol. Chem. 278, 48981–48990. Zhou, Y., and Gottesman, S. (1998). Regulation of proteolysis of the stationary-phase sigma factor RpoS. J. Bacteriol. 180, 1154–1158.