Global Co-ordination of Protein Translocation by the SecA IRA1 Switch*

SecA, the dimeric ATPase subunit of protein translocase, contains a DEAD helicase catalytic core that binds to a regulatory C-terminal domain. We now demonstrate that IRA1, a conserved helix-loop-helix structure in the C-domain, controls C-domain conformation through direct interdomain contacts. C-domain conformational changes are transmitted to the DEAD motor and alter its conformation. These interactions establish DEAD motor/C-domain conformational cross-talk that requires a functional IRA1. IRA1-controlled binding/release cycles of the C-domain to the DEAD motor couple this cross-talk to protein translocation chemistries, i.e. DEAD motor affinities for ligands (nucleotides, preprotein signal peptides, and SecYEG, the integral membrane component of translocase) and ATP turnover. IRA1-mediated global co-ordination of SecA catalysis is essential for protein translocation.

Energy conversion to mechanical work remains a central unresolved issue in several DEAD helicases (20,21,28) as well as in protein translocation. The mechanism is expected to involve cross-talk between the ATP motor and specificity domains (13,16,17). In SecA, evidence for this is provided by the finding that, in the absence of tight C-domain association, the DEAD motor becomes a hyperactivated ATPase (15,18,19). Importantly, SecA with a short IRA1 deletion also becomes an unregulated, hyperactivated ATPase that is nevertheless incompetent for translocation (15). This observation led us to propose that IRA1 is a molecular switch essential for coupling ATP hydrolysis to translocation work (15). We now show that IRA1 contacts other SecA subdomains and through these it controls association and conformational cross-talk between the DEAD motor and the C-domain. Modulation of these physical contacts allows IRA1 to regulate DEAD motor subactivities. We propose that SecA ATP binding and hydrolysis become coupled to protein translocation through IRA1 acting as a global co-ordinator of translocase catalysis and conformation.
Fluorescence Measurements-Measurements were carried out in quartz cuvettes (1 ml; Hellma) with a Cary Eclipse fluorimeter supplemented with a four-position cuvette holder and a Peltier temperature controller (Varian). For determination of equilibrium dissociation constants, SecA was added at increasing concentrations (0.025-5 M) to MANT-ADP (0.1 M; Molecular Probes) in buffer B. Emission spectra of MANT-ADP were recorded at each step of the binding curve (380 -550 nm; excitation 356 nm; slits at 2.5 and 20 nm). K D was determined by plotting the change of MANT-ADP emission spectra upon SecA addition against SecA concentration using the equation (F 1 Ϫ F 0 )/F 0 integral values (F 0 ϭ no SecA added, F 1 ϭ SecA added). To determine apparent T m , we monitored changes in intrinsic tryptophan fluorescence emission of SecA or C34 derivatives (0.25 M in buffer B) as a function of increasing temperature (4 -82°C; ramping rate 0.8°C/min; excitation 297 nm/emission 345 nm; slits at 2.5 and 20 nm; data acquisition interval ϭ 0.5 min), in the presence or absence of ADP (2 mM). All data were collected using Cary Eclipse software (Bio Package; Varian) and analyzed by nonlinear regression using Origin 5.0 (Microcal).
Miscellaneous-Protein concentration was determined using the Bradford reagent (Bio-Rad) with BSA as a standard or by UV absorbance or by amino acid analysis. Biochemical assays, trypsinolysis experiments, preparation of SecYEG inner membrane vesicles and proteoliposomes, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), 35 S-labeled SecA binding to IMVs, and SecA reconstitution from N68 and C34 were as described (5,(15)(16)(17). CD spectroscopy assays were performed as described previously (15). Native PAGE was carried out using a Bio-Rad Mini-PROTEAN II system. ␣-NBD and ␣-IRA2 rabbit polyclonal antibodies were raised against purified N1-263 and N462-610, respectively. ␣-SSD antibodies were purified as described (17). Radioactivity was quantitated by phosphorimaging (Storm 840; Amersham Biosciences). N-terminal sequencing and amino acid analysis were done at AltaBioscience (United Kingdom). Structures were analyzed with SwissPDBViewer.
Oligohistidinyl-tagged SecA IRA1 mutants were purified and shown, like SecA⌬783-795 (15,30), to be stable, folded, and ␣-helical (far UV CD; data not shown) and dimeric (size exclusion chromatography, blue native PAGE, sedimentation equilibrium; data not shown). We concluded that IRA1 mutant proteins do not have significantly altered structures and were characterized biochemically.
To verify that the in vivo phenotypes (Fig. 1C) are the result of defective protein export, HisSecA IRA1 mutants were used in an in vitro translocation assay with SecYEG-proteoliposomes and the secretory protein proOmpA ( Fig. 1D) (6). Of all IRA1 mutants, only SecAW775A (lane 5) supports protein translocation (lanes 6 -13). SecAR792A (lane 8) and SecAF811A (lane 13) do not translocate in vitro, although they partially complement in vivo (Fig. 1C). Clearly, the more stringent and suboptimal in vitro proteoliposome assay exacerbates their defects.
IRA1 Mutations Alter SecA ATPase Activities-To test whether IRA1 mutants are defective in ATP catalysis, we determined their basal, membrane, and translocation ATPase activities (Table I). Basal ATP catalysis is enhanced, either significantly (Ͼ13-fold; SecAW775A; Ͼ5-fold SecAF811A) or slightly (up to 2-fold; all other mutants except SecAI789R). Stimulation of basal ATPase upon addition of SecYEG-proteoliposomes (membrane ATPase) or proteoliposomes plus proOmpA (translocation ATPase) is seen only with SecAW775A, SecAR792A, and SecAF811A in agreement with the in vivo complementation test (Fig. 1C). All other IRA1 mutant proteins fail to further stimulate their basal ATPase.
IRA1 Mutations Alter SecA Affinity for Nucleotide-To understand how IRA1 influences SecA ATP catalysis, we examined the effect of IRA1 single point mutations or partial deletion on nucleotide binding to SecA. To this end we developed a fluorescent MANT-ADP binding assay (Table II).
At 4°C, MANT-ADP binds to SecA with high affinity (K D ϭ 0.14 M). This is in agreement with values obtained with other methods (29,31). N68, the polypeptide carrying the complete DEAD motor bereft of the C-domain (15), exhibits similarly high affinity (K D ϭ 0.28 M). C34, the polypeptide carrying the C-domain alone (15), has no measurable nucleotide binding (data not shown). In agreement with biochemical (16) and structural analysis (18,19), these data demonstrate that the DEAD motor domain of SecA is necessary and sufficient for nucleotide binding.
Temperature does not affect nucleotide binding to the DEAD motor in SecA (37°C ; Table II). In contrast, it drastically reduces binding to the isolated DEAD motor (Ͼ350-fold). This suggests that the C-domain acts in trans to determine DEAD motor nucleotide affinity at physiological temperature. Partial deletion or single point mutations in IRA1 reduce SecA nucleotide affinity, suggesting that IRA1 may be part of this mechanism.
IRA1 Mutations Alter SecA Affinity for SecYEG-Because membranes do not stimulate the ATPase of most IRA1 mutants (Table I), their binding to SecYEG might be defective. To investigate this we determined the equilibrium dissociation constants of SecA IRA1 derivatives for SecYEG (Fig. 1E) not shown) (32). These data indicate that SecA binds to Se-cYEG through the DEAD motor, but the presence of the Cdomain optimizes binding affinity. This C-domain contribution requires an intact IRA1 because partial deletion or single point mutations in IRA1 reduce SecA DEAD motor affinity for Se-cYEG (Fig. 1E).
IRA1 Mutations Alter Signal Peptide Binding to SecA-We next examined the ability of IRA1 mutants to interact with preprotein signal peptides (Fig. 1F). SecA (lane 3), but not an unrelated control protein (lane 1), binds to a signal peptide (3K7L) biosensor (17,33). The DEAD motor of SecA is necessary and sufficient for signal peptide binding (lane 2) (17). The presence of the C-domain in SecA prevents maximal signal peptide binding to the DEAD motor (compare lanes 3 and 2) (17). IRA1 mutations seem to overcome this and render the SecA DEAD motor more competent for signal peptide binding (compare lane 3 with lanes 4 -7). IRA1 Mutants Exhibit Altered DEAD Motor Conformation-SecA DEAD motor chemistries are affected by the absence of the C-domain or by mutations in IRA1 (Tables I and II and Fig.  1, E and F). This effect could reflect alteration of DEAD motor conformation. To investigate this we developed an assay that monitors DEAD motor conformation in SecA IRA1 mutants using limited trypsinolysis (15), followed by immunostaining with domain-specific antibodies (␣-NBD, ␣-IRA2, and ␣-SSD; Fig. 2).
Despite the fact that the overall structure and organization of SecA IRA1 mutants is undisturbed, changes in the tryptic profile of the mutants were detectable. Cleavage of the DEAD motor (p67 peptide; panel A, lane 2) appears more rapid in IRA1 mutants and occurs within the NBD-IRA2 linker (aa 420; panels A-C), within IRA2 (aa 561 and 585; panel B), within NBD (aa 201; panel C), at the base of the SSD stem out (aa 220; panel C) and within the SSD bulb (aa 360; panel C). The DEAD motor of SecAI789R seems particularly sensitive to trypsinolysis (panel A, lane 5) and gives rise to slightly different IRA2 and SSD peptides (panels B and C, lane 5).
Most of these tryptic sites become inaccessible when ADP is pre-bound to SecA (compare lane 6 with lane 2; aa 220, 420, and 561) (15,17). IRA1 mutants can acquire the "ADP-bound state" characterized by enhanced p67 stability (lanes 6 -9). However, this conversion is less efficient than in SecA because some cleavage still occurs (lanes 7-9; panels B (aa 420 and 561) and C (aa 221)). None of the DEAD motor tryptic sites is in immediate contact with IRA1 (panel D) (18,19), and yet cleavage is affected by IRA1 mutations, suggesting that they may cause long range conformational effects.
Next we probed the effect of IRA1 mutations on DEAD motor conformation by a different method, thermal melting monitored by far UV CD. In this assay, SecA and N68 melt in two distinct steps (T m1(app) and T m2(app) ) that represent loss of secondary structure within the DEAD motor (Table III) (16). Both of the SecA T m(app) are stabilized significantly by ADP, whereas only T m2(app) is stabilized in N68 (Table III) (16). SecA IRA1 mutants also display two melting transitions, indicating that their DEAD motors are structurally similar to that of SecA (Table III). Nevertheless, their T m(app) are altered.
DEAD Motor Affects C-domain Conformation-IRA1 mutations affect DEAD motor conformation (Fig. 3, A-C). To test whether the DEAD motor can reciprocally affect C-domain conformation, we monitored SecA intrinsic Trp fluorescence during thermal melting. This assay specifically follows C-domain conformation because its three Trp residues (aa 701, 723, and 775) are the main contributors of SecA fluorescence (18 -34). The derived T m(app) is unrelated to those obtained by CD (see above).
Under these conditions SecA displays a characteristic T m(app) (43°C) that is significantly stabilized by ADP (48°C; Fig. 2E) (18). Clearly, the ADP-induced conformational change of the DEAD motor (Table III) (16,35) is transmitted at a distance and sensed by C-domain tryptophans (18 -35). IRA2 mutations that compromise DEAD motor conformation (G510A) or nucleotide binding (R577K) (Figs. 1B and 2D; Ref. 16) interfere with DEAD motor/C-domain conformational cross-talk because they lead to T m(app) changes in the absence or presence of ADP (Fig.  2E). SecAW775A and SecAR792A, which exhibit practically wild type T m(app) in the apoprotein state, fail to acquire specifically the ADP-induced C-domain conformation of SecA (Fig.  2E). SecAI789R exhibits a lower T m(app) than SecA, even in the absence of ADP.
Our results indicate that there is conformational cross-talk between the DEAD motor and the C-domain, and this requires a functional IRA1.

IRA1 Mutations Affect DEAD Motor/C-domain Assembly-
Interdomain conformational cross-talk in SecA ( Fig. 2; Table  III) is presumably mediated through physical contact. Because partial IRA1 deletion abolishes DEAD motor/C-domain physical association (15), IRA1 might modulate this association. To  test this we examined the ability of HisC34 derivatives to bind to N68 by native PAGE (Fig. 3A). Successful C34/N68 association generates a novel electrophoretic species that migrates slower than the individual domains (lane 4) (15) and is similar to that of native dimeric SecA (lane 1). Formation of this complex is clearly affected by some IRA1 mutations. Samples containing C34W775A (lane 6) or C34F811A or C34Y803 (data not shown) migrate as fuzzy bands suggesting that these mutant C-domains fail to form tight complexes with N68. Other mutants form reconstituted complexes of the expected size in slightly (5-20%) reduced amounts (lanes 8 and 10). Successful C34 binding to N68 suppresses its elevated ATPase (Fig. 3B, lane 1) (15). As expected, C34W775A, which fails to form stable physical complexes (Fig. 3A, lane 6), cannot suppress N68 ATPase (Fig. 3B, lane 2). Interestingly, DEAD motor ATPase suppression by C34R792A (lane 3) and C34I789R (lane 4), which form physical complexes (Fig. 3A, lanes 8 and 10), is inefficient. Taken together, our data suggest that IRA1 has a crucial role in physical and/or functional C-domain/DEAD motor assembly.
IRA1 Is Required for Scaffold Domain Binding to the DEAD Motor-SD is the only C-domain substructure that binds to the DEAD motor (18,19). Then why do IRA1 mutations affect this association (Fig. 3)? (15) To understand this we generated three truncated C34 derivatives (Fig. 3C) and examined their binding to N68 by native PAGE (Fig. 3D) Our results suggest that the concomitant presence of both IRA1 and SD is required for SD binding to the DEAD motor.
IRA1 Mutations Affect C-domain Conformation-To explain how IRA1 influences SD binding to the DEAD motor, we hypothesized that IRA1 may affect SD conformation (Figs. 2E and 3). We therefore examined the effect of IRA1 mutations on C-domain conformation by limited trypsinolysis (Fig. 4A) and intrinsic Trp fluorescence (Fig. 4, B and C).
Trypsin cleaves C34 within SD, at Lys-643, giving rise to p30 (Fig. 4A, lane 2). p30 is then cleaved within CTD, giving rise to p28 and to small peptides (lane 2). C34 IRA1 mutants display either more rapid proteolysis of p30 and p28 (e.g. W775A; lane 3) or enhanced resistance of p30 and p28 (e.g. R792A; lane 4) or delayed cleavage at Lys-643 (e.g. I789R; lane 5). We additionally observed that C34E802A undergoes cleavage within SD, at Lys-643, from an unknown cellular protease during purification, and the resulting polypeptide remains stable thereafter (data not shown). These results indicate that mutations in IRA1 affect the conformation of both SD and CTD, where the tryptic sites are located.
We further monitored the conformation of C34 IRA1 mutants by Trp fluorescence during thermal melting (Fig. 4B). This assay follows conformational changes specifically sensed by Trp-701/Trp-723 in WD (18,34) because Trp-775 only contributes to the total Trp emission of C34 (compare C34 with C34W775A; panel C) but not to the measured C34 T m(app) (panel B). C609 -834 and C669 -834 maintain the T m(app) of C34, indicating that CTD or SD either do not significantly  contribute to C-domain conformation or that their deletion is not "sensed" by the WD (panel B). In contrast, partial or complete IRA1 deletion leads to significantly increased T m(app) (compare C34 to C34⌬783-795 and C609 -757), suggesting that IRA1 mutations affect WD conformation. Finally, IRA1 mutations that do not remove any Trp residue affect C34 total Trp emission (e.g. R792A and I789R; Fig. 4C) and/or T m(app) , suggesting the occurrence of a conformational change that alters emission from the same Trp residues. An extreme case is C34E802A; emission is virtually abolished (panel C), although cleavage during purification (see above) does not remove any of the emitting Trp residues.
Our data suggest that IRA1 mutations affect C-domain conformation in toto. DISCUSSION The DEAD motor, the catalytic core of SecA, is necessary and sufficient for nucleotide binding (Table II) and hydrolysis (Table I), signal peptide binding (Fig. 1F) (17,36), and SecYEG binding (Fig. 1F) (32). The structurally independent and juxtaposed C-domain binds to the DEAD motor and suppresses its ATPase (Fig. 3B) (15). We now show that the C-domain determines DEAD motor ligand binding affinities (Fig. 1 (E and F) and Table II) and catalysis (Table I) in trans. This regulation requires specifically IRA1, a conserved C-domain substructure. IRA1 mutations or partial deletion cause measurable alterations in all DEAD motor chemistries (Tables I and II; Fig. 1, E and F) that lead to defective protein translocation (Fig. 1, C  and D).
IRA1 has a characteristic hairpin structure formed of intersecting helices (Figs. 1 (A and B) and 5 (A and B)). These assemble through highly conserved, mainly hydrophobic, residues that include the ones mutated here: Leu-785 and Ile-789 on H1, Tyr-803 on H2, and Pro-799 that defines a sharp H2-L boundary. Local deformation of the hairpin structure might prevent/alter H1/H2 relative movement. Such motion is common in other proteins with H-L-H structures such as calmodulin (37) and may be essential for IRA1 function. Mutation or deletion of these residues severely compromises SecA function (Fig. 1C), as does a three-residue insertion after Pro-799 (15). Clearly, IRA1 structural integrity is crucial for SecA function. Contact to the second protomer, as suggested by the B. subtilis crystallographic dimer (18), might additionally stabilize the IRA1 hairpin structure. Mutation of Glu-802 that mediates such an interaction leads to a severely compromised mutant ( Fig. 1 (C and D); Table I). Our data indicate that the presence of the IRA1 hairpin destabilizes the C-domain (Fig. 4B)  that mutations in IRA1 affect the conformation of all other C-domain subdomains (Fig. 4, A and C), suggesting that IRA1 is important for C-domain flexibility (38).
IRA1 is not a structural component of the DEAD motor (Fig.  1A) and makes no direct contacts with either NBD or IRA2 (18,19). How then does IRA1 affect DEAD motor chemistries? The IRA1 hairpin seems strategically positioned in the SecA structure; IRA1 H1 fits in a spacious three-sided enclave formed by SD, SSD, and WD and "capped" only by the stem of SSD, whereas the IRA1 loop and most of IRA1 H2 are completely solvent-exposed (Fig. 5, A and B). Through conserved outwardly facing residues, the IRA1 hairpin makes two main crucial contacts with surrounding subdomains (Figs. 1A and 5 (A and B). (a) The first is IRA1-SD; this is a tight interaction mediated by two bulky aromatic residues each located in one of the two IRA1 helices (Trp-775 in H1 and Phe-811 in H2). A third residue, Tyr-803, of H2 further "fixes" the IRA1 hairpin to SD by interacting with both H1 and SD. These residues fit into "sockets" formed by hydrophobic SD residues. Mutation of ei-ther one residue weakens SD binding to the DEAD motor ( Fig.  3B; data not shown). (b) The second contact is IRA1-SSD; although SSD is separated from IRA1 by a large cavity (Fig.  5B) (18), defined contacts between IRA1 and the SSD bulb and stem in occur (18,19). E806A and R792A mutations generated here are expected to interfere with IRA1-SSD bulb interaction. The inevitable consequence of such positioning is that IRA1 hairpin becomes the only physical link between the two "leverlike" appendages SSD and SD (Fig. 5, A and B) (18,19), each rooted in one of the two DEAD motor subdomains (Fig. 1A). By binding to both SD and SSD (18,19) and affecting their conformation (Figs. 2 (A-D) and 4A), IRA1 is appropriately placed to "manipulate" DEAD motor conformation and catalysis at a distance. This is achieved through a dynamic network of reciprocal conformational cross-talk ( Fig. 2; Table II) (16,17) that allows communication between the DEAD motor and its "specificity" appendages (SSD and C-domain).
The SD-IRA1-SSD interface bears the hallmarks of a coupling device that is dynamic. Although SD is the sole DEAD motor-binding determinant on the C-domain (18,19), C-domain/DEAD motor binding requires the presence of IRA1 (Fig.  3D). Presumably, by binding to it IRA1 maintains SD in a conformation that possesses the characteristic bent (Fig. 5A) and is competent for DEAD motor association. W775A and F811A, which weaken IRA1 binding to SD (Fig. 5, A and B), also weaken C-domain binding to the DEAD motor (Fig. 3A,  lane 6). Importantly, SD-mediated C-domain binding to the DEAD motor (Fig. 3A) determines DEAD motor ligand affinities (Table II; Fig. 1, E and F) and turnover (Table I). Thus, these "gain of function" mutations hyperactivate SecA ATPase in the absence of any translocation ligands. ATP catalysis by the F 1 ATPase ␤ subunit is similarly regulated by binding and release of the elongated ␣-helical ␥ subunit (39). Interestingly, Trp-775 is mobile and moves in and out of its hydrophobic environment upon ligand binding to SecA (18,26,27,34,35). Shortening of the Trp side chain in W775A presumably mimics a state of IRA1 that is detached from SD. In contrast, W775F and W775Y mutations that retain bulky hydrophobic side  Fig. 2E. On the schematic W is fluorescence-emitting Trp residues (18,35). *, main T m(app) . n ϭ 4. C, Trp fluorescence emission spectra of C34 and derivatives upon thermal melting. chains have practically wild type behavior (18,34,35). 2 We therefore anticipate that IRA1 "oscillates" laterally to and from SD throughout protein translocation. Taking into account the physical proximity of SSD to IRA1 (Fig. 5, A and B) and the effects of IRA1 mutations on SSD conformation (Fig. 2, A-D) and function (Fig. 1E), we expect that IRA1 may also bind and release from SSD during catalysis. Thus, IRA1 mutations may exert a general influence on plasticity of the SD-IRA1-SSD interface.
Taken together our data allow us to formulate the following working hypothesis: ATP-driven DEAD motor translocation work is conformationally coupled to cycles of SD association/ dissociation, which in turn are coupled to cycles of IRA1 binding and release from SD, and these may be further coupled to cycles of IRA1 binding and release from SSD. We anticipate that these events are regulated by preprotein binding to SSD (17,23) and SecYEG binding to the DEAD motor (Fig. 1E), both interactions affected by IRA1. Thus, IRA1 acts as a molecular switch (15) that "senses" translocation ligands, controls SecA conformational plasticity and subactivities. These properties render IRA1 a global co-ordinator of SecA and protein translocase catalysis.