Roles of the N-domains of the ClpA Unfoldase in Binding Substrate Proteins and in Stable Complex Formation with the ClpP Protease*

The hexameric cylindrical Hsp100 chaperone ClpA mediates ATP-dependent unfolding and translocation of recognized substrate proteins into the coaxially associated serine protease ClpP. Each subunit of ClpA is composed of an N-terminal domain of ∼150 amino acids at the top of the cylinder followed by two AAA+ domains. In earlier studies, deletion of the N-domain was shown to have no effect on the rate of unfolding of substrate proteins bearing a C-terminal ssrA tag, but it did reduce the rate of degradation of these proteins (Lo, J. H., Baker, T. A., and Sauer, R. T. (2001) Protein Sci. 10, 551-559; Singh, S. K., Rozycki, J., Ortega, J., Ishikawa, T., Lo, J., Steven, A. C., and Maurizi, M. R. (2001) J. Biol. Chem. 276, 29420-29429). Here we demonstrate, using both fluorescence resonance energy transfer to measure the arrival of substrate at ClpP and competition between wild-type and an inactive mutant form of ClpP, that this effect on degradation is caused by diminished stability of the ClpA-ClpP complex during translocation and proteolysis, effectively disrupting the targeting of unfolded substrates to the protease. We have also examined two larger ssrA-tagged substrates, CFP-GFP-ssrA and luciferase-ssrA, and observed different behaviors. CFP-GFP-ssrA is not efficiently unfolded by the truncated chaperone whereas luciferase-ssrA is, suggesting that the former requires interaction with the N-domains, likely via the body of the protein, to stabilize its binding. Thus, the N-domains play a key allosteric role in complex formation with ClpP and may also have a critical role in recognizing certain tag elements and binding some substrate proteins.

Chaperone-protease ring complexes play a major role in the cell in mediating ATP-dependent turnover of a variety of protein substrates (1,2). Substrates are specifically recognized either via intrinsic amino acid sequences, usually at terminal positions in the primary sequence (3)(4)(5)(6)(7)(8), or by added moieties. In the case of bacteria, one such addition, directed by the ssrA RNA, comprises 11 residues appended to the C terminus of translationally arrested proteins stalled at the ribosome (9). In eukaryotes, conjugation of proteins with polyubiquitin is the major step directing proteins toward proteolytic turnover (10 -14). In bacteria, tagged substrate proteins are recognized by hexameric Hsp100 ring chaperones such as ClpA, ClpX, or HslU, which are known to mediate ATP-dependent unfolding and translocation of substrate proteins into the coaxially aligned partner proteases ClpP and HslV, respectively (15,16). Analogously in eukaryotes, polyubiquitinated proteins are recognized by the 19 S proteasome cap complex containing a hexameric ring of ATPases at its base, which unfolds and translocates the substrates into the 20 S proteasome core (12,(17)(18)(19). The subunits of Hsp100 chaperones contain one or two ATPase modules of the AAAϩ family, and these modules collectively form the body of these cylindrical structures (20 -24). The ATPase activities of these modules have been shown to be critical for hexameric assembly, unfolding the substrate proteins, and translocating the substrate proteins into the associated protease (15,25). ClpX and ClpA also contain NH 2 -terminal domains (N-domains) that appear to control substrate access to the machinery by several means.
In the case of ClpX, the N-domain is a 60-residue zinc finger element that may dimerize in the context of the intact ClpX ring, considered to be a trimer of dimers (26,27). This domain appears to be critical for direct recognition of O and MuA substrates (26,28) as well as for binding a delivery agent, SspB, which directs additional substrates into ClpX for unfolding. In particular, the SspB homodimer, which can bind two ssrA moieties of tagged substrates through its dimerized NH 2 -terminal domains, can bind to a pair of ClpX N-domains via flexible COOH-terminal decapeptides (29 -32). Another delivery agent, RssB, also binds to the ClpX N-domains to deliver the stress-dependent factor S (33).
In the case of ClpA, the N-domain is composed of ϳ150 amino acids folded into an ␣-helical, 2-fold pseudosymmetric structure as revealed from both a stand alone x-ray structure and a structure of the intact subunits (22). A recent cryo-electron microscopy study comparing intact ClpA with an N-domain-deleted version suggests that the N-domains rise in the axial direction out of the body of the cylinder, emerging from the D1 AAAϩ domain (34). Simulations of this electron microscopy data suggest that the N-domains are very mobile in their position atop the cylinder, consistent with what appears from structural studies to be attachment to the D1 domain via a flexible linker segment (amino acids 143-167). As with ClpX, the ClpA N-domains can bind a partner protein, in this case a small protein called ClpS that shares an operon with ClpA (35)(36)(37). ClpS appears to modulate substrate specificity by binding 1:1 to each N-domain. This acts to inhibit the access of ssrAtagged substrates to ClpA, potentially by steric blockage. It also inhibits the "autodegradation" of ClpA itself, which occurs in the presence of ATP and ClpP. At the same time, binding of ClpS appears to activate an activity in protein disaggregation, observed in vitro. Despite the effect of ClpS on access of ssrA-tagged proteins in vitro, the fact remains that in the intact cell ClpA shares action with ClpX on ssrA-tagged protein substrates, as evidenced by an elegant experiment using a proteolytically inactive mutant ClpP as a substrate "trap" in the setting of deletion of ClpX (8).
The functional role of the N-domains has been directly evaluated by deletion analysis and cross-linking studies. N-domain-deleted ClpA has been observed to be able to support ClpP-mediated degradation of casein and two ssrA-tagged proteins, GFP-ssrA and N-cI-ssrA, although the rates of degradation of these substrates was substantially slowed (28,38). This observations suggests that the N-domains are not required for recognition of these substrates. Furthermore, cross-linking studies have shown that the ssrA tag interacts with loop structures in the ATPase domains (39). Here we have dissected the basis for the slowed turnover of ssrA-tagged proteins by N-domain-deleted ClpA-ClpP complexes, observing that whereas the rate of unfolding of GFP-ssrA is indeed unaffected by deletion of the N-domains, the rate of substrate translocation into ClpP is greatly reduced. We have also evaluated the action of the N-domain-deleted ClpA on additional substrates, both a GFP 4 derivative fused at its C terminus with the 15 N-terminal amino acids of RepA, GFP-RepA (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15), and two larger ssrA-tagged substrates, CFP-GFP-ssrA and luciferase-ssrA, all of which are efficiently acted on by wild-type ClpA-ClpP.

EXPERIMENTAL PROCEDURES
DNA Constructs-ClpA⌬143 was generated by PCR mutagenesis, removing codons for residues 2-142. A COOH-terminal ssrA-tagged version of firefly luciferase (Photinus pyralis) was amplified by PCR from the pGEM-luc vector (Promega) and cloned into the pET-16b expression vector (Novagen), which adds a 10-histidine affinity purification tag to the NH 2 terminus. CFP-GFP-ssrA was produced by adapting pECFP (Clontech) and GFP-ssrA by PCR mutagenesis to allow in-frame fusion of both coding sequences through appropriate restriction sites. GFP-RepA (1-15) was generated as reported (7). All clones were subjected to DNA sequencing to verify their identity.
Unfolding, Degradation, and Translocation Assays-Unfolding and degradation assays of GFP-ssrA, GFP-RepA (1-15), and CFP-GFP-ssrA with ClpA, ClpA⌬143, and ClpP were performed as described (5,41). Fluorescence was determined with excitation at 480 nm and emission at 510 nm for GFP or with excitation at 433 nm and emission at 475 nm for CFP. Luciferase unfolding was monitored by observing the luminescence of luciferin. His-luciferase-ssrA (1 M) and 150 M luciferin were mixed with 10 mM ATP and incubated at 25°C until the luminescence (565 nm) reached its steadily declining phase (15 min). Then, unfolding was initiated by adding 1.25 M ClpA or ClpA⌬143 and additional ATP (10 mM). Unfolding assays of luciferase-ssrA were performed in the absence of the GroEL(D87K) trap, because spontaneous refolding was not observed under these conditions (data not shown). Translocation was monitored by fluorescence resonance energy transfer (FRET) in a stopped-flow fluorometer as reported previously (40) using pyrene maleimide-modified mClpP(S111A/L139C) as the donor and the GFP chromophore as acceptor. Note that the GFP chromophore appears to act as a FRET quencher even when it is not itself fluorescent. 5 Excitation was at 338 nm and emission was recorded through a combination of long wave pass (nominal 370-nm cut-off) and short wave pass (nominal 490-nm cut-off) filters.
ATPase Assay-ATPase activity was measured by determining free phosphate with the malachite green method (42). The reaction mixture contained 0.25 M ClpA or ClpA⌬143 and 0.125 M ClpP in assay buffer (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, and 10 mM MgCl 2 ). The mixture was incubated at 25°C for 1 min, and the reaction was initiated by addition of 5 mM ATP. Aliquots were taken at various time points up to 5 min, and the phosphate released was determined. Rates were calculated from the linear plots of phosphate released versus time.

N-deleted ClpA Exhibits Normal Kinetics of Unfolding of GFP-ssrA but Supports a Decreased Rate of Degradation in the Presence of ClpP-A
ClpA construct commencing at residue 144 (ClpA⌬143) was expressed and purified. Similarly to an earlier-reported construct commencing at residue 162 (38), this molecule efficiently assembled into hexamers in either ATP␥S or ATP as observed in gel filtration (not shown), but in the presence of ATP it only supported relatively slow degradation of GFP-ssrA by the addition of ClpP (Fig. 1a, red trace). To further examine this reaction, we first measured the rate of GFP-ssrA unfolding by either wild-type ClpA or ClpA⌬143 in the presence of ATP and a trap version of GroEL, the mutant D87K (Fig. 1b). This trap mutant has been observed to bind and retain non-native forms of GFP-ssrA as they are released from ClpA, preventing their refolding and thus allowing a determination of the ability of ClpA to mediate unfolding. When GFP-ssrA was incubated with wild-type ClpA (8:1; GFP-ssrA/ClpA) and the trap, there was loss of fluorescence intensity occurring as a single exponential decay with a rate constant of 0.0021 s Ϫ1 (Fig. 1b, blue trace), similar to that reported previously (41). Remarkably, and in agreement with studies on ClpA⌬162 (38), there was a virtually identical loss of fluorescence when GFP-ssrA was incubated with ClpA⌬143 at a similar rate constant of 0.0017 s Ϫ1 (Fig. 1b, red trace). Thus, it appears that the absence of the N-domains has no effect on the ability of ClpA to recognize and unfold a substrate protein such as GFP-ssrA. This implies that recognition and unfolding of these substrate proteins is mediated by the AAA domains of ClpA, which form the barrel of the ClpA cylinder, rather than the N-domains. This hypothesis has been directly established by recent chemical cross-linking and mutational studies (39) showing that the ssrA tag is recognized by loops in the ATPase domains that extend into the central channel of the ClpA hexamer.
Defective Translocation into ClpP-Considering that the rate of recognition and unfolding by ClpA⌬143 resembles that by wild-type, we hypothesized that the slow rate of ClpA⌬143-mediated substrate deg-radation by ClpP must result from a defect of translocation into the ClpP protease. Translocation was assessed in real time by following the appearance of FRET between a donor fluorophore, pyrene, placed on the inside cavity wall of a proteolytically defective (and leader peptidedeleted, i.e. mature-sized) mutant version of ClpP, mClpP(S111A) (40), and the chromophore of GFP-ssrA, which appears to act to quench fluorescence even when GFP is unfolded and is not itself fluorescent. 5 In particular, this experiment was carried out by first assembling ClpA in ATP␥S and incubating it with GFP-ssrA and the pyrene-labeled ClpP and then adding ATP with stopped-flow mixing and measuring the acquisition of FRET by monitoring the loss of fluorescence in the donor channel (Ͼ370 nm) (Fig. 2). In the case of wild-type ClpA, there was a rapid acquisition of FRET (Fig. 2, blue trace). By contrast, the rate of acquisition of FRET was much reduced with ClpA⌬143 (Fig. 2, red trace), indicating that the transfer of GFP-ssrA into the ClpP cavity was slowed with the mutant ClpA.
Reduced Stability of Association of ClpA⌬143 with ClpP during Proteolysis-The impairment of transfer of unfolded GFP-ssrA from ClpA⌬143 to ClpP could result either from a defect of ATP-mediated translocation from the mutant into the physically associated ClpP or from decreased stability of the ClpA⌬143-ClpP complex, essentially removing access of the substrate to the translocation target. (Notably, ClpP in isolation is unable to bind and degrade unfolded polypeptides, although it can act upon small peptides; see Ref. 43). To test the relative stabilities of association of ClpP with wild-type and truncated ClpA during ClpAP-mediated degradation of GFP-ssrA, we took advantage of an experimental design used by Singh et al. (44) to measure the dissociation rate of the ClpA-ClpP complex by competition with chemically inactivated ClpP. Here, the protease-defective mClpP(S111A) (40) was used to perform a similar experiment. We first showed that the complex of this mutant with ClpA supported only a minor degree of unfolding of GFP-ssrA (Fig. 3a). Then the stability of the GFP-ssrA-ClpA-ClpP complex was tested by initiating a proteolysis reaction by the addition of ATP and, after 120 s, the addition of a large molar excess of the mutant, mClpP(S111A). Because this mutant has a normal ability to associate with ClpA, it should successfully compete with the wild-type ClpP for binding to ClpA if the ClpA-ClpP complex dissociates during the degradation reaction. This should block further degradation, with GFP fluorescence then remaining constant, because any GFP-ssrA that is unfolded by ClpA but released into solution will refold to native form (as in Fig. 3a). When this competition experiment was carried out with wild-type ClpA, the addition of mClpP(S111A) had little effect on the rate of loss of fluorescence until the reaction had proceeded for 7-8 min (Fig. 3b, compare light blue and dark blue traces), reflecting the relative stability of the wild-type ClpA-ClpP complex (44) as well as the commitment of unfolded proteins to degradation by ClpP. In striking contrast, when the mClpP(S111A) was added to a similar reaction with ClpA⌬143 after 120 s, the loss of fluorescence of GFP-ssrA was almost immediately halted (Fig. 3b, compare red and purple traces), indicating that the mClpP(S111A) had immediately begun to compete with wildtype ClpP for binding to the ClpA⌬143 mutant. This finding suggests that the association between ClpA⌬143 and ClpP is considerably less stable than that of the wild-type complex.
Further evidence of an impaired interaction between the hexameric ClpA⌬143 complex and ClpP was obtained by comparison of its ATPase activity with that of wild-type ClpA. Although a previous study with wild-type ClpA had indicated that the rate of ATP hydrolysis is slightly reduced upon association with its proteolytic partner (45), in our hands the ATPase activity of ClpA was reproducibly enhanced by a factor of 2 in a ClpA-ClpP (2:1) complex as compared with ClpA alone (Fig. 4). Deletion of the N-terminal domain did not significantly affect the ATPase activity of ClpA⌬143 alone, suggesting that neither the structure of the ATP binding pockets of the D1 and D2 domains nor the contact between neighboring subunits, which is necessary for hydrolysis  FRET between a pyrene (donor) fluorophore attached inside the mClpP(S111A) cavity and the chromophore of GFP-ssrA (acceptor) was monitored by loss of donor fluorescence after stopped-flow mixing of the reaction components (40) (blue trace, wild-type ClpA; red trace, ClpA⌬143). FRET efficiency was calculated as described previously (40) and is expressed as a percentage (i.e. 0.20 ϭ 20%). The solid lines are fits of the data to a double exponential equation, with rate constants for the initial phase of 0.086 s Ϫ1 and 0.0066 s Ϫ1 , respectively. (22), is altered by the deletion. In contrast, the addition of ClpP to ClpA⌬143 failed to stimulate the ATPase activity (Fig. 4) even in a 20-fold molar excess (not shown). The observed effect could be explained either by a structural alteration in the route of allosteric communication between ClpP and the ClpA ATPase domains or more simply by an increased dissociation constant for the complex.
We tested whether the dissociation constant for the ClpA⌬143-ClpP complex was greater than that for ClpA-ClpP by measuring individually the association and dissociation rates for the complexes using a FRETbased assay. Although the dissociation rate for the ClpA⌬143-ClpP complex was slightly greater than that measured for ClpA-ClpP (0.025 Ϯ 0.004 s Ϫ1 versus 0.019 Ϯ 0.0025 s Ϫ1 ), its association rate was also greater (0.99 Ϯ 0.16 ϫ 10 6 M Ϫ1 s Ϫ1 versus 0.63 Ϯ 0.05 ϫ 10 6 M Ϫ1 s Ϫ1 ). Thus, the dissociation constants for the two complexes as measured in the absence of substrate protein are the same (25-30 nM) within experimental error. Based on the relative instability of the association of ClpA⌬143, red and purple traces), were initiated by adding 5 mM ATP. After 120 s (arrows) a 20-fold molar excess of catalytically inactive mClpP(S111A) was added to the reactions, indicated by the light blue and purple traces. As in Fig. 1a, loss of GFP-ssrA fluorescence reflects proteolytic degradation. Note that proteolysis continues at a significant rate for some time after the mClpP(S111A) addition to the wild-type (wt) ClpA reaction (light blue trace), whereas it halts almost immediately with ClpA⌬143 (purple trace), indicating rapid exchange between active ClpP and the inactive competitor in the presence of the N-domain-deleted ClpA.   Fig. 1b. b, stable association of GFP-RepA-(1-15) with wild-type ClpA, but not ClpA⌬143, as measured by gel filtration. Wild-type (wt) or mutant ClpA (2 M) was incubated with 0.5 mM ATP␥S to assemble the respective ring complexes, and these were incubated in turn with a 4-fold excess of GFP-RepA- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). The mixtures were applied to Superose 6 gel filtration columns, and the eluates were monitored sequentially for both fluorescence emission (Ͼ490 nm; top row) and absorbance (280 nm; bottom panels). The fluorescence and absorbance peaks eluting at ϳ23-25 min correspond to the elution positions of the GFP-RepA-(1-15)-ClpA complex or ClpA alone, and those at 31-35 min correspond to GFP-RepA-(1-15) alone. As shown in the top row, GFP-RepA-(1-15) associates with wild-type ClpA (left), but not with ClpA⌬143 (right).

Roles of the ClpA N-domains in Binding and Degradation
ClpA⌬143 and ClpP observed in the ClpP exchange experiment in Fig.  3, we infer that the substrate protein itself must be contributing to the stability of the ClpA-ClpP complex via the N-domains. As observed in recent cross-linking and FRET studies (39), substrate proteins come in contact with the N-domains during binding and translocation steps, and such contact may produce allosterically directed stabilization of the association of ClpA with ClpP. The absence of the N-domains here would prevent such signaling from occurring and result in a destabilized complex. More generally, such an action is consistent with a model in which AAAϩ unfoldases allosterically regulate the proteases with which they associate (and vice versa). This has been directly demonstrated by crystallographic studies of the HslU-HslV chaperone-protease pair, where binding of HslU to HslV produces structural shifts that are transmitted to the active sites of HslV, correlating with HslU-directed activation of HslV protease activity (46).
Evidence for a complementary allosteric influence of ClpP on ClpA is apparent in the experiment in Fig. 4, where the ATPase activity of wildtype ClpA was increased by 2-fold when ClpP was added. The absence of any increase in the ATPase activity of ClpA⌬143 with added ClpP suggests that allosteric communication between ClpP and the ClpA ATPase domains was also disrupted by deletion of the N-domains, an effect that does not appear to depend on a substrate protein. Even so, it seems likely that the same pathway of signaling, from N-domains to ATPase domains to ClpP interaction sites and back, is involved in both phenomena.
N-domain-deleted ClpA Does Not Act on GFP-RepA- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) and Has Different Effects on Two Larger (ϳ60 kDa) ssrA-tagged Substrates-The plasmid P1 replication origin-binding protein RepA has previously been shown to be recruited as a dimer to ClpA for dissociation into DNA binding-competent monomers via its NH 2 -terminal region (6,47). A segment comprising the first 15 amino acids of RepA has been observed to be sufficient to promote either ClpA-mediated unfolding or ClpA-ClpP-mediated degradation of GFP fusion proteins bearing this RepA sequence at either the NH 2 or COOH terminus (5,7). Here, the COOHterminal fusion protein GFP-RepA-(1-15) was incubated with wildtype ClpA or ClpA⌬143 in the presence of ATP and the D87K GroEL trap as in the earlier study of GFP-ssrA ( Fig. 1) (41), and fluorescence was monitored in a time-dependent manner. In contrast with the rapid loss of fluorescence brought about by wild-type ClpA (Fig. 5a, blue trace), reflecting unfolding, there was little loss of fluorescence in the presence of ClpA⌬143 (Fig. 5a, red trace), reflecting that the substrate either failed to be recognized by the ClpA mutant or was recognized but then failed to be unfolded. Consistent with this observation, GFP-RepA-(1-15) was degraded very slowly by ClpA⌬143-ClpP (not shown). The step of binding could be directly evaluated because RepA and a number of GFP-RepA fusions bind stably to ClpA in ATP␥S (6), such that a fluorescent binary complex can be detected in gel filtration at the position of ClpA (e.g. Fig. 5b, upper left). When such an incubation was FIGURE 6. Unfolding of two larger substrate proteins. a, luciferase-ssrA (62 kDa) is unfolded at similar rates by wild-type (wt) and ClpA⌬143. Luciferase-luciferin reactions were initiated with ATP and incubated until the burst phase of luminescence had passed and a steady decrease in luminescence at 565 nm was achieved (black trace). ClpA (blue trace) or ClpA⌬143 (red trace) as well as additional ATP was added, and luminescence was monitored. Because continuing luciferin luminescence requires active luciferase, the rapid drop observed here indicates that luciferase-ssrA was rapidly unfolded by both chaperones. Note that no GroEL trap was required in these experiments, because luciferase does not refold spontaneously under these conditions (data not shown) (51). b, the CFP-GFP-ssrA fusion protein (58 kDa) is unfolded by wild-type (wt) ClpA (blue trace) but much more slowly by ClpA⌬143 (red trace). The substrate protein was incubated with a 6-fold molar excess of wild-type or mutant ClpA and ATP in the presence of a similar excess of GroEL D87K trap as in Fig. 1b. GFP fluorescence was monitored at 510 nm; monitoring CFP fluorescence at 475 nm gave the same result (not shown). The apparent inefficiency of unfolding of this substrate by wild-type ClpA may reflect a low efficiency of binding to the GroEL trap. C, CFP-GFP-ssrA is degraded more slowly by ClpA⌬143-ClpP than by wild-type (wt) ClpA-ClpP; the two domains of the fusion protein are degraded sequentially but at equal rates. CFP-GFP-ssrA (2 M) was mixed with 0.2 M wild-type ClpA (cyan and green traces) or ClpA⌬143 (red trace) and 0.1 M ClpP, and the reaction was initiated with 5 mM ATP. For comparison, GFP-ssrA was similarly mixed with wildtype ClpA-ClpP and ATP (blue trace). GFP fluorescence was monitored at 510 nm (red, green, and blue traces), and CFP fluorescence was monitored at 475 nm (cyan trace). Both domains of the fusion protein are degraded by wild-type ClpA-ClpP almost as rapidly as is the single domain in GFP-ssrA. The inset shows the first 70 -80 s of the CFP-GFP-ssrA reaction for each fluorescent (fluor.) domain. The CFP trace (cyan) clearly shows a delay relative to the GFP trace (green) before it begins decreasing, consistent with the sequential and directional unfolding and degradation beginning with the ssrA tag and the adjacent domain (40). The slight initial increase in CFP fluorescence may reflect a decrease in FRET-related quenching as the GFP is unfolded and its chromophore is translocated away from the neighboring domain.
carried out with ClpA⌬143, little fluorescence was observed at the migration position of ClpA⌬143 (Fig. 5b, upper right), indicating that the substrate did not associate with ClpA in the absence of its N-domains. Thus it appears that in contrast with the ssrA tag (Fig. 1b), the RepA recognition element has a strong requirement for the N-domains as sites of initial recognition. This finding has been confirmed by recent chemical cross-linking studies showing that the site of interaction of the RepA tag is within the N-domains (39).
Unfolding of Larger ssrA-tagged Substrates-Two larger ssrA tagged substrates were also examined. One was firefly luciferase, a 62-kDa protein composed of a large N-terminal ␣-␤ domain and a much smaller C-terminal one (48). When fused with a C-terminal ssrA tag, it was unfolded by the N-deleted version of ClpA as efficiently as by wild-type ClpA, as judged by the change in luciferin luminescence (Fig. 6a), and it was degraded when ClpP was added (not shown). The second larger substrate was a fusion of CFP with GFP-ssrA to produce a 58-kDa substrate with two ␤-barrel domains. In contrast to GFP-ssrA alone, which was unfolded by ClpA⌬143 at a rate similar to that of wild-type ClpA (Fig. 1b), the CFP-GFP-ssrA fusion protein was unfolded considerably more slowly by the mutant than by wild-type ClpA in the presence of ATP and D87K GroEL trap (Fig. 6b). Consistently, CFP-GFP-ssrA was also degraded much more slowly by ClpA⌬143-ClpP than by wild-type ClpA-ClpP (Fig. 6c, compare red with cyan or green traces). Interestingly, the two domains of CFP-GFP-ssrA are degraded by wild-type ClpA-ClpP at essentially the same rate (Fig. 6c, cyan and green traces) and almost as fast as is GFP-ssrA itself (blue trace). Furthermore, the data in Fig. 6c (inset) suggest that the degradation by wild-type ClpA-ClpP is sequential, with GFP fluorescence (510 nm; green) beginning to decrease immediately, whereas CFP fluorescence actually increases slightly before beginning to decay about 20 s later (475 nm; cyan). We could not evaluate whether binding of the fusion protein to ClpA⌬143 was affected, because gel filtration experiments with this and other GFP-ssrA fusions (in ATP␥S) failed to reveal the presence of binary complexes, even with wild-type ClpA (not shown). We surmise, however, that, in the absence of the N-domains, the fusion protein is unable to form a sufficiently stable initial association with ClpA to unfold efficiently, unlike a single GFP attached to an ssrA tag.
These data suggest that the N-domains likely interact with one of the CFP-GFP moieties to stabilize the overall interaction. Indeed, crosslinking studies have shown that the GFP moiety of GFP-ssrA is positioned to interact with the N-domains (39), even though this interaction is not required for unfolding. The source of the requirement for the N-domains in the case of CFP-GFP-ssrA is not clear, however. For example, the contrasting behavior of CFP-GFP-ssrA and the other similarly large protein, luciferase-ssrA, shows that larger size alone does not determine whether the ClpA N-domains are necessary. Some larger ssrA-tagged proteins, such as CFP-GFP-ssrA, may need the N-domains to stabilize binding through the "body" of the substrate protein to achieve efficient degradation. Other factors, for example the stability of individual domains in the substrate (25,49,50), have been considered to be important in determining whether the N-domains are required for individual substrate proteins, but that seems less likely to be the case here because the individual domains are of comparable stability.
Overall, it seems clear that the N-domains play several critical roles in the function of ClpA as follows. (i) They act as the binding site for the adaptor protein ClpS. (ii) They bind certain recognition tags such as the RepA N-domain; and (iii) they also are required for efficient degradation of some substrate proteins, even in the presence of a recognized terminal tag. Furthermore, the N-domains appear to allosterically modulate the interaction of ClpA with its partner protease ClpP. It remains to be seen whether other sequence elements recognized by ClpA also require the presence of the N-domains for binding and whether the N-domains themselves contribute directly to unfolding or translocation or whether they are strictly a recognition/binding element.