FtsH (HflB) is an ATP-dependent protease selectively acting on SecY and some other membrane proteins.

The FtsH protein is a membrane-bound ATPase of Escherichia coli that was proposed to be involved in membrane protein assembly as well as degradation of some unstable proteins. SecY, a subunit of protein translocase, is FtsH dependently degraded in vivo when it fails to associate with its partner (the SecE protein). We constructed a series of mutants in which mutations were introduced into conserved residues in the two ATP binding consensus sequences or the zinc binding sequence of FtsH. We purified wild-type and mutant FtsH proteins by making use of a polyhistidine tag attached to their carboxyl termini. Complementation analysis and ATPase activity assays in vitro indicated that, of the two sets of ATP binding sequence motifs, the one located C-terminally (A1) is essential for ATPase activity and in vivo functioning of FtsH. Wild-type FtsH protein degraded purified SecY in an ATP hydrolysis-dependent manner in vitro. Mutant proteins without ATPase activity were inactive in proteolysis. A zinc binding motif mutant showed a decreased proteolytic activity. SecY and FtsH were cross-linkable with each other in the membrane, provided that FtsH had an ATPase-inactivating mutation. These results demonstrate that FtsH binds to and degrades SecY, its A1 motif and the zinc binding motif being important for the proteolytic activity. FtsH-dependent proteolysis was also demonstrated for SecY in crude membrane extracts, whereas a majority of other membrane proteins were not degraded, indicating that FtsH has high selectivity in protein degradation.

The components of machinery for protein secretion in Escherichia coli are encoded by the sec genes. The SecY protein is a central component of the membrane-embedded part of the secretion machinery, and functions in close interaction with other integral membrane proteins, SecE and SecG (1,2). Genetic and biochemical evidence suggests that the SecY-SecE interaction is important not only for the functioning but also for the stable existence of the SecY protein (3)(4)(5)(6). When an uncomplexed form of SecY is produced due to unbalanced expression of the secY and secE genes, it is rapidly eliminated by proteolytic degradation (6,7). A mutated form of SecY, SecY24, in which SecY-SecE interaction has been mutationally impaired, is also degraded at 42°C (3,7).
We previously reported that a membrane-bound ATPase, FtsH (HflB), is required for the degradation of the uncomplexed forms of SecY. In mutants with compromised FtsH functions, both overproduced SecY protein and the SecY24 mutant protein are stabilized (7). FtsH contains two transmembrane segments N-terminally located and a large cytoplasmic domain (8,9). The cytoplasmic domain of FtsH includes a region that is homologous to the AAA ATPase family members involved in a variety of cellular processes (10). In addition, the cytoplasmic domain contains a region homologous with a zinc protease sequence motif (11). The transmembrane region is required for the oligomeric structure of FtsH (9). The hflB mutation, known to increase lysogenization frequency of bacteriophage by stabilizing the cII protein, proved to be an allele of the ftsH gene (12). Recently, an ATP-dependent protease activity of FtsH was described (11). It is involved in the degradation of an intrinsically unstable heat shock transcription factor, 32 (11,13). It was shown that FtsH homologs (Yta10, Yta12, and Osd1/Yme1) in the yeast mitochondria also participate in degradation of some inner membrane proteins (14 -18). However, their proteolytic activities have not been characterized in vitro using purified proteins.
In this work, we characterized FtsH with respect to its activities to hydrolyze ATP and to degrade the SecY protein. The results indicate that FtsH has an ATP-dependent protease activity with high substrate specificity.
For preparation of FtsHd2-His 6 -Myc, cells of AD432 carrying pSTD179 were grown at 37°C for 6 h in 1 liter of L broth containing 1 mM isopropyl-␤-D-thiogalactopyranoside. Total membranes were prepared, and FtsHd2-His 6 -Myc was purified as described above. FtsH71-His 6 -Myc, FtsH81-His 6 -Myc, FtsH82-His 6 -Myc, FtsH83-His 6 -Myc, and FtsH84-His 6 -Myc were purified using an Ni 2ϩ -NTA spin column (Qiagen) as follows. Membranes were prepared from 100 ml of induced culture of TYE024 carrying pSTD132, pSTD143, pSTD128, pSTD147, or pSTD134, and solubilized with Nonidet P-40 as described above. The solubilized proteins (600 l) were loaded onto the Ni 2ϩ -NTA spin columns that had been equilibrated with the solubilization buffer, and the columns were centrifuged for 2 min at about 350 ϫ g. After washing the columns 2 times with 600 l of the washing buffer, proteins were eluted with 200 l ϫ 2 of 250 mM imidazole. Eluates were mixed and dialyzed extensively against dialysis buffer.
ATPase Activity Assay-ATPase activities of the wild type and mutant forms of FtsH-His 6 -Myc were assayed according to Chan et al. (30). 10 l of a sample (containing 0.2-2 g of purified protein) were mixed with 13 l of 0.5% Nonidet P-40, 4 l of H 2 O, 3 l of 20 mM ATP, and 30 l of 2 ϫ ATPase buffer (100 mM Tris/Hcl, pH7.5, 5 mM Tris acetate, 10 mM MgCl 2 , 50 M zinc acetate, 2 mM DTT) and incubated at 37°C for 1 h. Then, it was mixed with 240 l of the malachite green/polyvinyl/ ammonium molybdate reagent and 30 l of 34% Na 3 citrate/2H 2 O. After 30 min at room temperature, absorbance at 660 nm was measured.
In Vitro Degradation of Pulse-labeled Membrane Proteins in Detergent Extracts by FtsH-His 6 -Myc-Cells of AD373 carrying pKY6 were grown at 30°C to an early log phase in 5 ml of M9 medium supplemented with 18 amino acids (20 g/ml, other than Met and Cys), thiamine (2 g/ml), 0.4% glucose, and appropriate antibiotics. The cultures were shifted to 42°C for 1 h, induced with isopropyl-␤-Dthiogalactopyranoside (1 mM) and cAMP (5 mM (30 l) were removed at intervals, incubated at 37°C for 5 min in the presence of 1% SDS, and subjected to immunoprecipitation with antiserum against SecY as described previously (6). Total protein profiles were also examined without immunoprecipitation. Labeled proteins were visualized by SDS-PAGE and a BAS2000 imaging analyzer.

Phenotypes of ATP Binding Site
Mutants-We noted previously (9) that FtsH has two sets of ATP binding consensus sequences (Walker A and B motifs) in its cytoplasmic region, one (A1 and B1) within the region homologous to the AAA family members (10) and the other (A2 and B2) located more N-terminally (Fig. 1). A mutation that changes Lys 198 in A1 (resulting gene, ftsH41) or Lys 136 in A2 (ftsH51) or both (ftsH61) to Asn gave dominant Std phenotype when they were expressed from a plasmid (24). We later found that the plasmid with the ftsH51 mutation had acquired an additional spontaneous mutation causing Thr 199 to mutate to Ala in A1. The two mutations in "ftsH51" were separated, and the ftsH mutation causing the Thr 199 to Ala change was named ftsH40, whereas the one causing the Lys 136 to Asn change was named ftsH71. We also constructed, by site-directed mutagenesis, mutations that cause a change of the conserved Thr 199 in A1 to Asn (ftsH81) and Ser 137 in A2 to Asn (ftsH84). A complete listing of the mutant ftsH genes and their nomenclatures are provided in Fig. 1.
These mutant genes were recloned onto a low copy plasmid, pHSG575 (to circumvent the overproduction toxicity of FtsH), and examined for their ability to complement the temperaturesensitive ftsH1 mutant as well as for their ability to exhibit dominant negative Std phenotype. The ftsH40 (T199A) and the ftsH41 (K198N; see Ref. 24) plasmids did not complement the ftsH1(Ts) mutant, whereas the ftsH81 (T199N) plasmid did so weakly. Thus, the importance of the A1 motif for the FtsH function (24) has been confirmed. On the other hand, the conserved Lys and Ser residues in A2 can be changed to Asn without major dysfunction since both ftsH71 (K136N) and ftsH84 (S137N) were able to complement ftsH1 (the growth of ftsH84-bearing cells was slightly slower than normal). When the weak A1 mutation, ftsH81, was combined with the ftsH84 mutation, the resulting gene, named ftsH83 (T199N and S137N), proved to totally lack the complementation activity. The synthetic effects of the A1 and A2 mutations may suggest that the A2 region has some functional role in FtsH. The mutant forms of ftsH that did not complement the ftsH1 mutant all exhibited a dominant Std phenotype (data not shown), suggesting that they interfered with the functioning of the wild-type FtsH protein.
ATPase Activities of the Mutant Proteins-We introduced each of the ATP binding site mutants of ftsH into the His 6 -and Myc-tagged version of the ftsH gene (on pSTD113). Immunoblotting experiments (data not shown) showed that these FtsH derivatives accumulated in the membranes, except for FtsH51 and FtsH 61 that proved to be toxic to the cells under the induction conditions. For purification of accumulated FtsH-His 6 -Myc proteins, membranes were solubilized with a nonionic detergent, Nonidet P-40, and subjected to affinity purification using Ni 2ϩ -NTA-Sepharose or Ni 2ϩ -NTA silica, as described under "Materials and Methods." Preparations of 50 -80% purity (with respect to the full-length proteins) were obtained (Fig. 2). Some of them contained a slightly lower molecular mass protein, which should have been generated by proteolytic cleavage around the junction between FtsH and the bipartite tag during purification. We previously characterized a similar product termed FtsHЈ (9). The proteolytic product without the tag was recovered presumably through its interaction with FtsH-His 6 -Myc proteins (9). The preparations also contained several C-terminally truncated low molecular mass products, which cross-reacted with anti FtsH antibodies (data not shown). Taking this into account, the purity can be regarded as higher. In addition, they should have contained the chromosomal ftsH ϩ gene product, again because of the homooligomeric interaction between the FtsH molecules. However, the wild-type subunit should have amounted only to less than 5% of each FtsH preparation, and its contribution to the ATPase and proteolytic activities should have been at most 5% of the wild-type activity.
We measured ATPase activities of the FtsH preparations. The wild-type His 6 -Myc protein catalyzed the release of about 9.2 mol P i /mg protein/hour, a value similar to that reported for normal FtsH by Tomoyasu et al. (11). Thus, the C-terminal tag did not pronouncedly affect the ATPase activity. FtsH40-His 6 -Myc (T199A), FtsH41-His 6 -Myc (K198N), and FtsH81-His 6 -Myc (T199N) lacked significant ATPase activity (Fig. 1). The residual ATPase activity in these cases could be ascribed to the contaminating wild-type FtsH protein (see above). On the other hand, the A2 mutants, FtsH71 (K136N) and FtsH84 (S137N), had only moderately reduced or wild type level of ATPase activity. Thus, the A1 motif region is mainly responsi-ble for the total ATPase activity of FtsH. Interestingly, FtsH81, with weak but significant complementation activity, had only very week ATPase activity although the contamination by wildtype protein precluded the accurate assessment.
Demonstration of SecY Degradation Activity of FtsH-Our mutational studies strongly suggested that FtsH participates in degradation of SecY in vivo (7). We examined whether the purified FtsH-His 6 -Myc proteins can proteolyze SecY in vitro. Purified preparations of SecY and FtsH-His 6 -Myc, both in nonionic detergent solutions, were mixed and incubated at 42°C in the presence or absence of ATP. SecY was evidently degraded when incubated for 1 h in the presence of both ATP and FtsH-His 6 -Myc (Fig. 3A, lanes 4 -6). The purified SecY preparation used contained an N-terminal fragment of SecY (SecYЈ) (confirmed by antiserum against the N-terminal region of SecY, data not shown) that was generated presumably by artificial cleavage during purification. The SecYЈ fragment was degraded more rapidly than intact SecY (Figs. 3 and 4). Degradation of SecY and SecYЈ was more efficient at 42°C than at 37°C or 30°C (data not shown). Several degradation intermediates were observed upon longer exposure, but none of them substantially accumulated during shorter periods of incubation (data not shown), indicating that the proteolysis, once initiated, was very rapid.
Without ATP (Fig. 3A, lanes 1-3) or without FtsH-His 6 -Myc (lanes 7 and 8), degradation of SecY or SecYЈ was negligible. CTP and UTP substituted for ATP but ineffectively (Fig. 3B,  lanes 8 -11; see particularly the SecYЈ band). GTP and ATP␥S did not support the degradation (Fig. 3B, lanes 6, 7, 12 and 13). This nucleotide specificity was similar to that reported for the FtsH-dependent degradation of 32 , except that UTP was reported to be ineffective for 32 degradation (11). A1 site mutant proteins, FtsH40-His 6 -Myc (Fig. 4, lane 6) and FtsH41-His 6 -Myc (lane 10), did not appreciably degrade SecY. These results demonstrated that FtsH has an ATP-dependent protease activity toward the isolated SecY protein and that the ATPase region is important for the proteolytic activity.
FtsH has a zinc-metalloprotease motif (HEXXH) in its cytoplasmic domain (H 414 EAGH), and Tomoyasu et al. (11) reported a zinc ion stimulation of in vitro 32 degradation. One of our dominant negative mutants of ftsH, ftsHd2, has an amino acid alteration, Glu 415 to Lys, in the zinc binding/protease motif (24). We purified the FtsHd2-His 6 -Myc protein. FtsHd2-His 6 -Myc was found to retain an almost wild type level of ATPase activity (Fig. 1). Its proteolytic activity against SecY was markedly decreased although it did not seem to be totally abolished (Fig. 4, lane 4). FtsHd2 retained some complementation activity (Fig. 1). These results suggest that the Glu 415 to Lys mutation markedly decreases the proteolytic activity of FtsH but does not totally inactivate it.
Proteolytic Activity of FtsH Is Selective-We examined how generally membrane proteins serve as a substrate for FtsH. Cells of the temperature-sensitive ftsH1 mutant strain that carried a SecY-overproducing plasmid (pKY6) were pulse-labeled at 42°C with [ 35 S]methionine. Membranes were solubilized with Nonidet P-40, and incubated with purified wild-type FtsH-His 6 -Myc protein at 42°C in the presence or absence of ATP (Fig. 5). Labeled SecY in the detergent extract, as detected by immunoprecipitation, was degraded in a manner dependent on the exogenous FtsH and ATP (Fig. 5, lanes 13-16). Total electrophoretic profile of the labeled proteins (Fig. 5, lanes  19 -28) showed that only a few of them (marked by arrowheads) were degraded upon incubation with FtsH-His 6 -Myc and ATP. Concomitantly, low molecular weight materials were generated (see the region ahead of the dye front). Many proteins, especially those of high molecular weights, remained undigested (Fig. 5, lane 28). In other words, FtsH degrades only a selected set of E. coli membrane proteins. As most E. coli membrane proteins are stable in vivo, the selectivity of the FtsH action appears to be preserved at least partially in vitro. In a similar experiment, we found that subunit a of the F 0 part of the proton ATPase can be a substrate of FtsH. 2 In Vivo Interaction between FtsH and SecY-In vivo interaction between SecY and FtsH was examined by cross-linking experiments, using cells overproducing SecY and either FtsH-His 6 -Myc, FtsH40-His 6 -Myc, or FtsH41-His 6 -Myc protein.
Membranes were prepared from pulse-labeled cells and treated with a membrane-permeable cross-linker, dithiobis(succinimidylpropionate). After solubilization with SDS, labeled proteins that were cross-linked with FtsH or SecY were isolated by immunoprecipitation using anti-FtsH or anti-SecY, respectively. The cross-linking was cleaved with 2-mercaptoethanol before visualization by SDS-PAGE and autoradiography. When cells overproducing FtsH40-His 6 -Myc or FtsH41-His 6 -Myc were used, an FtsH-sized protein was recovered as a crosslinked partner of SecY (Fig. 6, lanes 6 and 12), and an SecYsized protein was recovered as a cross-linked partner of FtsH (lanes 2 and 10). The SecY-sized protein was also recovered when a portion of the sample for lane 2 was precipitated with anti-Myc antibodies (data not shown). Cross-linking was negligible or very weak when cells overproducing wild-type FtsH and SecY were used (lanes 4 and 8). This result may have been due to stabilization of the bound state of SecY by the ATPase mutations. In addition, SecY might have been degraded by the active FtsH. DISCUSSION FtsH is required in vivo for rapid elimination of the uncomplexed forms of SecY (7). Biochemical characterization described in this study established that FtsH directly degrades SecY. FtsH-dependent degradation of SecY was demonstrated using purified SecY protein as well as using detergent extracts of membranes. The proteolysis was ATP-dependent and not observed with the ATPase-deficient mutants of FtsH. It was demonstrated previously (11) that 32 , the heat shock factor, was degraded by FtsH in vitro. Although SecY and 32 are apparently remote in nature, they are substrates of a common FtsH protease.
The results of our mutational analyses indicate that the ATP binding motif A1 in the region homologous to the AAA family members is essential for the ATPase and proteolytic activities 2 Y. Akiyama, A. Kihara, and K. Ito, unpublished results.  9 and 10). After incubation as indicated, samples were taken and analyzed as described in the legend to Fig. 3. WT, wild type. of FtsH. The synthetic loss of FtsH functions observed for the ftsH81 and ftsH84 mutations may point to a role of the A2 motif in conjunction with the A1 motif. The inability of the A1 mutants to degrade SecY indicates that the ATPase activity is essential for the proteolytic activity of FtsH. From analogy to other ATP-dependent proteases, such as E. coli ClpP proteases, ATP hydrolysis may be coupled with unfolding of substrate proteins and their presentation to the proteolytic active site in FtsH.
Our results show that the zinc binding motif in FtsH is also important for the SecY proteolysis activity, and the mutant protein, with an alteration in the glutamine residue, that is implicated to be crucial in the catalysis of zinc metalloproteases (34) has markedly decreased proteolytic activity. A requirement for zinc was described for 32 degradation (11). Elucidation of the reaction mechanism of the FtsH protease requires further mutational and structural studies of the FtsH region around this motif as well as clarification of the primary targets of the initial hydrolysis within the SecY and 32 amino acid sequences.
In addition to 32 , the cII protein can be degraded by purified FtsH-His 6 -Myc in vitro. 3 The fact that these two cytosolic proteins are substrates of FtsH (11)(12)(13) suggests that the active site of FtsH for proteolysis is located within the cytosolic domain, which includes the ATPase region and zinc binding motif. Given this model, it seems likely that FtsH attacks cytosolic regions of SecY. Our genetic analysis of SecY-SecE interaction suggested that the fourth cytoplasmic region (C4) is important for the SecY interaction with SecE (3). An essential region of SecE has also been localized to its cytoplasmic domain (35). The interaction between the cytoplasmic domains of SecY and SecE may induce a structural change in the cytoplasmic regions of SecY, such that they become refractory to the attack of FtsH. We found that SecY-SecE complex in detergent extracts is far more resistant to even a non-specific protease (proteinase K) than singly overproduced SecY. 2 The thermolability of the SecY-SecE complex in detergent extract (36) precluded the direct examination of its resistance to FtsH, but it is conceivable that overall conformation of the cytoplasmic domains determines the FtsH-resistance of SecY. We are in the process of identifying a SecY region that is primarily attacked by FtsH in a SecE-free state.
As reported previously (7), the physiological significance of the SecY degradation by FtsH should lie in the elimination of unassembled SecY molecules that are harmful to the membrane functions. Taken together with our finding 2 that subunit a of the proton ATPase F 0 segment is also a substrate of FtsH, FtsH seems to comprise an important quality control machine in the E. coli cell.
Our recent results (21) that FtsH can associate with the hflK and hflC gene products and that this association negatively regulates the SecY-degrading activity of FtsH suggest that the proteolytic activity of FtsH is subject to regulation by other proteins. The FtsH-His 6 -Myc preparations used in this study contained a trace of HflK/HflC proteins that could be detected immunologically. 2 However, it should not affect the conclusions drawn in this paper because FtsH-His 6 -Myc that was purified from a strain where hflK and hflC were deleted showed essentially the same activity of SecY degradation (22). The product of multi-copy suppressor gene, fdrA (37), could also have some regulatory role for the FtsH activities. Under the experimental conditions used in this work, the apparent proteolytic activity of FtsH was very low since we needed FtsH in high excess over the substrate SecY protein to demonstrate its proteolytic activity. The low activity may have been due to in vitro inactivation or a lack of some activating proteins. Alternatively, full proteolytic activity may require the membrane-integrated states, such that SecY and FtsH are properly oriented to each other.
Acknowledgments-We thank T. Yoshihisa and T. Ogura for helpful discussion and K. Mochizuki for technical and secretarial assistance. 3 A. Kihara, Y. Akiyama, and K. Ito, unpublished results.

FIG. 5. Degradation of SecY and other membrane proteins by FtsH-His 6 -Myc in crude detergent extracts.
Cells of AD373/pKY6 were exposed at 42°C and pulse-labeled with [ 35 S]methionine for 5 min. Total membranes were prepared, solubilized with 0.5% Nonidet P-40, and incubated at 42°C in the presence or absence of FtsH-His 6 -Myc and ATP as indicated. Samples were subjected either directly to SDS-PAGE (lanes 19 -28) or to anti-SecY (lanes 1-16) immunoprecipitation before SDS-PAGE. Lanes 17 and 18 were for molecular mass markers (from top to bottom, 97.4 kDa, 69 kDa, 46 kDa, 30 kDa, and 18.4 kDa). Open arrowheads indicate positions of proteins affected by FtsH-His 6 -Myc.