ClpX, an Alternative Subunit for the ATP-dependent Clp Protease of Escherichia coli SEQUENCE AND IN VIVO ACTIVITIES*

The ATP-dependent CIp protease of Escherichia coli consists of two subunits, the ClpP subunit, which has the proteolytic active site, and ClpA, which possesses ATPase activity and activates the proteolytic activity of ClpP in vitro. Recently, Zylicz and co-workers identified another E. coli protein that activated ATP-dependent degradation of X 0 protein in the presence of ClpP. The amino-terminal sequence of this protein corresponds to the translated amino-terminal sequence of a gene that we have named clpX. cipX encodes a protein with M, 46,300, similar to that observed for the protein purified by Wojtkowiak et al. clpX is in an operon with clpP; both genes are cotranscribed in a single heat- inducible 2200-base mRNA, with clpP the promoter proximal gene. The sequence of ClpX includes a single consensus ATP-binding site motif and has limited homology to regions of ClpA and other members of the ClpAIBfC family. A third group of proteins, ClpY, closely related to ClpX, has been identified by sequence homology. Mutations in either clpX or clpP abolish degradation of the highly unstable X 0 protein in vivo. clpX mutants are not defective in degradation of previously identified ClpAlClpP substrates such as a ClpA- &galactosidase fusion protein. It appears the 3' portion of the gene (23). A recent map of the E. coli chromosome (24) shows the location of clpx, based on our work; all assignments of E. coli genes to particular minutes are based on the centosome assignments used in that map.

identified in prokaryotes and eukaryotes (7, 8). E. coti has at least one such homolog of ClpA, called ClpB. ClpB has a protein-stimulated ATPase activity, but there is as yet no evidence that ClpB can combine with ClpP to carry out ATPdependent proteolysis @).' The genes encoding ClpP, ClpA, and ClpB are widely separated from each other on the E. coli chromosome. Both clpB (mapping at 58 min) and clpP (at 10 min) are at least in part regulated by a sigma-32 dependent heat shock promoter, although clpA (at 20 min) does not seem to be (4, 10-12). Evidence that ClpA and ClpP work together in vivo, as well as in vitro, is provided by the observation that a clpA-lac2 protein fusion is degraded in Clp" cells but is stabilized in cells devoid of either ClpA or ClpP (4, 13). Intracellular turnover of two other fusions is also dependent on both ClpA and ClpP? Varshavsky and co-workers (14,151 have reported that @-galactosidase constructs containing abnormal aminoterminal amino acids can be rapidly degraded in E. coli in a process which is dependent on both ClpA and ClpP. However, not all activities dependent on ClpP in uivo are also dependent on ClpA. The NHz-terminal 14 amino acids of ClpP are cleaved off during assembly of the protease; this activity is dependent on a functional ClpP, but is apparently normal in clpA mutants (8). Some growth defects have been observed in cells carrying clpP mutations in certain backgrounds; clpA mutations do not show the same defects and in fact may help relieve the clpP defects? Toussaint and coworkers (16) have recently shown that Muvir repressor i s degraded in E. coli and stimulates the rapid degradation of an endogenous wild-type repressor after infection of Muc+ lysogens; this degradation is dependent on ClpP but not ClpA.
Therefore, additional proteins capable of working with CjpP may exist in E. coli. The work presented by  in this volume provides the most direct evidence for this thus far. They identified a number of proteins able to carry out ATP-dependent degradation of X 0 protein; detection of one of these protease activities requires ClpP and a unique subunit, originally called LopC (17). As part of a collaborative effort to identify the gene encoding LopC, we determined that the NHp-terminal sequence of the LopC protein was identical to that for the predicted protein sequence of the gene just downstream of clpP, which we have called clpX. We show here that ClpX plays a central role in the in vivo degradation of X 0 protein. ClpX appears to represent a new class of Clp ATPase subunits.

B~~e r~~~
Strains, ~~t~r i o p h~e , and P~m i d s Isogenic strains containing single elp mutations or a lon mutation were constructed by PI transduction, selecting the antibiotic resist-' M. Maurizi, unpublished observations. S. Gottesman and L. Wellen, unpublished observations. V. de Crecy, unpublished observations. ance marker within the gene. clpm.:cat (131, clpA319A~an (4), c1pS::Akan (lz), and lon-146::Atet (18) have been previously described. The isolation of clpX1::kan is described in this paper. The starting strain for the canavanine protein turnover experiments was SGllOl, a wild-type E. coli strain carrying a Aarg mutation (4). derivatives of SG20250, a relative of MC4100 (19) lysogeniaed with Recipient strains for the 0 protein turnover experiments were all Xc~857Sam7 or XcI857. SG12054 is a c l p~~c a~ derivative of MZl(20). Burst size experiments were done either in SG20250 lysogenized with irnrn21 plac supF (from M. Berman) or in SA1336 (from s. Adhya) derivatives carrying the single or double clp mutations. ~i m m 4 3~I -2 Oam29 and kc1857 Sam7 are from the National Institutes of Wealth phage collection. Hosts lysogenic for SB84, a transducing phage carrying the clpA-lac2 fusion were assayed in SG20250 derivatives as previously described (21).
Cloning and Sequencing of clpX pWPC9, carrying clpP and the region downstream of clpp up to the hginning of ton, has been described (13). The EcoRV-~inD111 fragment from pWPC9 was cloned into both Ml3mp18 and Ml3mpl9 vectors for dideoxy DNA sequencing (22). Both the universal M13 primers and appropriate primers from within the sequenced region were used. The sequence for the whole gene was determined on both strands, using reactions with dITP in regions that showed compression. The sequence shown in Fig. 2 starts 1 base pair beyond the translation termination codon of clpP and extends to the region just ahead of the lon promoter. A portion of the 5' end of clpX has been published previously (13), as has the 3' portion of the gene (23). A recent map of the E. coli chromosome (24) shows the location of clpx, based on our work; all assignments of E. coli genes to particular minutes are based on the centosome assignments used in that map.
Mutagenesis of clpX pWPC9, containing both clpP and clpX, was digested with NcoI and religated. This generated a plasmid, pWPC51, that lacked 745 base pairs of the clpX gene (deleting from bp4 552 of sequence shown in Fig. 2 to bp 1290). A 12-nucleotide double-stranded oligomer (CATGGCTGCAGC) containing a PstI site and overhanging NcoI sticky ends was ligated with NcoI cut pWPC51 to create pWPC52. A BarnHI-HinDIII fragment containing the interrupted clpX gene as well as the rest of the original bacterial insert from pWPC9 was ligated with Bum~~-~~nDIII-diges~dpACYCl84, to createpWPC53. This plasmid was digested with PstI and ligated with a cassette encoding kanamycin resistance, made by digesting pUC4K with PstI (25). The AclpX;:kun plasmid, pWPC54, was linearized with PvuII and BarnHI and the digested DNA transformed into JC7623, a recBC sbcBC strain obtained from N. Trun, which does not degrade incoming linear DNA and is able to recombine it into the chromosome. Colonies which were kanamycin resistant hut ampicillin sensitive were purified and selected for further testing. One such isolate was called SG1152. P1 grown on this strain was used to introduce the clpX mutation into a variety of other backgrounds.
Purification of ClpX for Production of Antibodies ClpX was produced from a plasmid with the clpP and clpX genes under control of the p~ promoter of lambda in SG12054, a host cell carrying the heat-sensitive lambda repressor, cZ857. The plasmid pSK20 used was a derivative of the expression plasmid pREl (26) created hy S. K. Singh in the following way: an NdeI-BurnHI fragment from pWPC22, which carries clpP lacking the coding region for the first 13 amino acids, was inserted in the multiple cloning site of pRE1, Then, a double-stranded oligonucleotide coding for the first 13 amino acids but altered to include an NdeI site at the initiator methionine codon was inserted at the NdeI site (orientation was confirmed by DNA sequencing). An oligonucleotide linker carrying an SstI site was inserted at the BarnHI site. Finally, an MiuI-SstI fragment fro^ pWPC9 carrying the second half of clpP and the entire clpX gene was inserted between the MluI site in clpP and the SstI site.
For expression of ClpX, cells were grown to a density of 1.5 ODm units and induced for 2 h at 42 "C. ClpX produced under these conditions was found in inclusion bodies in cell extracts. The inclusion bodies were washed with buffer and extracted with 1% sodium dodecyl sulfate (SDS). The extract was run on a 12% acrylamide gel in SDS, and the CipX band was ehctroeluted. Antibodies were raised ' The abbreviation used is: bp, base pair(& in a rabbit after two injections of 0.5 mg each in Freund's adjuvant as described previously (4).

Western Blot Condi~ions
To detect the accumulation of X 0 protein in various mutants, cells carrying the XcI857Sam7 prophage were grown in Luria broth at 32 "C to a density of about ODm = 0.5 and transferred to 42 "6. Samples were taken after 0, 30, and 60 min and quenched with 5% trichloroacetic acid. Preparation of cellular protein for SDS gel analysis and Western blotting procedures were described previously (41.0 protein standard was the gift of R. McMacken and rabbit anti-0 protein antibody was a gift from S. Wickner. ClpX was detected in cells precipitated in trichloroacetic acid, run on SDS gels, and blotted to nitrocellulose.
In Vivo Tests of Protein ~egradation &A-lacZ-Cells lysogenic for the clpA-bc fusion phage SB84 (21) were grown in tryptone broth (27) and assayed for @-galactosidase as described by Miller (271, a reflection of the accumulation of this unstable fusion protein. C a~v a~~n e -c o n~a i n~~ Proteins-Intracellular degradation of abnormal proteins was measured by pulse labeling cells with tritiated leucine in the presence of canavanine and chasing with non-radioactive leucine after removal of canavanine as described previously (4).
X 0 Protein Turnover-The half-life of X 0 protein in vivo was determined in cells lysogenic for the heat-inducible prophage X c~~7 S a m 7 , Cells were grown in minimal medium M63 (27) with 0.2% glucose at 32 "C, and 1 ml of cells was transferrred to 42 "C when the density was about ODm = 0.5. After 8 min at 42 "C, the cells were labeled with 200 pCi of [35S~)methionine (New England Nuclear, 1190 Ciimmol) for 1 min. Medium containing 200 pg/ml non-radioactive methionine at 37 "C was added, the cells were shaken at 37 "C, and samples were removed a t intervals. Sample preparation, immunoprecipi~tion (with 5 pl of anti-0 protein antiserum and 300 pg of fixed S t a p h y~c o c c~ aureus cells (Calbiochem Pansorbin)), and analysis was performed as described previously (21).
Burst Size Experiments-Cells lysogenic for the i m m 2 l p~-s u p F phage, which carries the supF amber suppressing tRNA under the control of the plac promoter, were grown in TBMM (tryptone broth with 0.0001% vitamin B1, 0.01 M MgS04, and 0.2% maltose) in the presence of 0.02% isopropyl-1-thio-@-~-galatopyranoside overnight to induce the suppressing tRNA. Cells were collected by centri~gation and resuspended in 0.1 M MgSO,. 0.1 ml of cells was adsorbed for 10 min at room temperature with 10 pl of imm434 Oam29 or imm434 Nam7Nam53 phage to give a multiplicity of infection of less than 0.1, and then diluted 100-fold into LB (27) with 0.01 M MgSO+ The diluted infected cells were grown at 37 "C with shaking for at least 90 min, chloroform was added, and the phage titer determined by plating on the permissive host LE392 (supl: supE) (28). SA1336 ( F his relAl s~~( g u~~-X~~, )~~~~ N+ cI857 ( P -c~l A )~~~) contains a cryptic temperature-inducible X prophage encoding AN protein, 0 protein, and the immunity region, but is deleted for DNA to the right of X 0 protein. SA1336 and its clp derivatives were grown at 32 "C in TBMM with 0.003% biotin until turbid. Cells were induced for 10 min at 42 "C. Samples were removed before and after induction, adsorbed with imm434 Oam29 or imm434 Nam7Nam53 at a multiplicity of less than 0.1, diluted into 5 mi of LB with 10 mM MgS04, and grown with shaking for 150 min. CelIs were lysed with chloroform and the released phage titered on LE392 (28). Burst size in all cases was calculated by dividing the final titer of phage released by the input phage titer.

Northern Blot Methods
Total RNA was isolated from E. coti cells grown at 32 and 42 "C to a density of 0.5-0.8 ODm units using the procedure described previously (13). A 1.1% agarose gel was run under denaturing conditions with about 5 pg of each RNA in a lane and blotted to nitrocellulose as described (29). The blot was prehybridized for 4 h at 42 "C with Hybrisol I (Oncor) and then hybridized with about 50 x IO6 disintegrations/min of heat-denatured labeled primer made by random hexamer priming. The clpP-or clpX-specific DNA templates for the random priming were made by the polymerase chain reaction protocol (30) and corresponded to bases 250-950 in the clpP sequence and bases 181-1180 in the clpX sequence ( Fig. 1). After hybridization, the blots were washed twice in 2 X SSC, 0.1% SDS at 42 "C and then twice at 62 "C in 0.1 X SSC, 0.1% SDS for 15 min each. The blots were exposed to x-ray film and qu&nt~tated by scanning for 36 h with an Ambis r a d~o~a p h i c scanner.

RESULTS
Sequence of clpX-clpP maps at 10 min on the E. coli map, slightly upstream of lon, the gene for the other major ATPdependent protease of E. coli. Between the two genes is a region of about 1500 base pairs, which appears to be part of the elpP operon and encodes a protein of 424 amino acids (Figs. 1 and 2 ) . We have called this gene c b X . The DNA sequence for the entire clpX gene and the translated amino acid sequence for the ClpS protein are given in Fig. 2. The ClpX open reading frame begins 125 bases downstream from the termination codon for ClpP. There is no obvious promoter sequence in the 5' upstream region and no obvious transcription terminator between the clpP and clpX genes, There is an inverted repeat downstream of the open reading frame for ClpX, suggestive of a p-independent transcr~ption terminator (underlined in Fig. 2). CipX has a predicted M, of 46,300.
Given the possibility that a gene in the same operon with clpP might be related to the function of CfpP, we compared the N~~-t e r m i n a~ sequence of the nci, 46,000 LopC protein, found by Zylicz and co-workers to degrade X 0 protein in the presence of ClpP (accompanying paper by Wojtkowiak et al. (17)), with the sequence for the amino-term~nal portion of the ClpX protein predicted from the DNA sequence we obtained for clpX (Fig. 2). There is precise agreement between the NH~-t~rminal sequence found for LopC and the first 13 amino acids predicted for ClpX, if the NH~-terminal Met is removed after synthesis. Therefore, ClpX is the same as h p C ; both will be referred to hereafter as ClpX.
New Clp Families: ClpX and ClpY-A search of the available databases indicated that the open reading frame for a protein with extensive similarity to ClpX (53% identity and 75% similarity over 440 amino acids) has been found in A2oto~cter u i n e~~~i (31), where it is located in a cluster of genes involved in nitrogen fixation. We will refer to this protein as A. u i~~~~~ ClpX ( reading frames from P~t e u r e~a h~~~y t~c a ' and B~~~~ subtilis6 are very similar to each other, and resemble ClpX (50% identity, 68% similarity). The B. s~b~i~~ protein, called CodD, and the P. ~e m o l y t i c a protein, LapA, differ from ClpX by the presence of an insertion of 115 and 160 amino acids, respectively, in the position corresponding to about amino acid 176 of ClpX. We will refer to this subfamily of proteins as ClpY (Fig. 3A ) .
We compared the predicted amino acid sequences of E. coli and A. v~n e~n d~i ClpX with those of E. coli ClpA and ClpB. Many of the regions of sequence that are conserved between ClpA and ClpB are also highly conserved in all members of the ClpA/B/C family. A number of regions of sequence similarity between ClpX and ClpA or ClpB are apparent; the alignments for the regions with greatest similarity are shown in Fig. 3B. ClpX and ClpY contain a single consensus ATPbinding site sequence (boxed in the ClpX sequence in Fig. 2 ) which is very similar to those found in the ClpA/B/C family of proteins. While the ClpX site can be aligned with either the front or back half of ClpA, it shows a somewhat more extended alignment with the second ATP consensus site of ClpA (Fig. 323). In addition, near their COON termini, the sequences of ClpX, Y, A, and B have two extended regions of similarity which include two short sequences (boxed Tail Motifs A and B in Fig. 2) that are identical in all members of the ClpA/B/C family (7, 32).
When the ClpX protein sequence was compared to the sequences of other proteins in the database using the Blast program (331, the best matches were to two related proteins identified as chaperone-like proteins for secretion, cdc48 and a v~osin-con~ining protein (34); this alignment is primarily around the ATP consensus binding site (data not shown). Recently, Rechsteiner and co-workers (35) reported the sequence of Subunit 4 of the 26s eukaryotic ATP-dependent protease complex; this subunit shows a consensus ATP-binding site and homology to the c d c~/ v~o s i n family as well as limited homology to ClpA. Therefore, it appears that ClpA and its relatives ClpX and ClpY are part of a larger ATPase supe~amily which includes proteins involved in energy-dependent degradation in eukaryotic cells.
Close to the amino terminus of ClpX is a region unique to the two ClpX proteins with the predicted structure of a single Zn finger. This sequence motif, CxxC (x18) CxxC (boxed and marked as Zn Motifs A and B in Fig. 2), is similar to those found in the DNA-binding domain of hormone-responsive receptor proteins (see Ref. 36, for review of Zn finger domains in hormone-responsive receptors). The function of these clusters of cysteine residues in ClpX is not known.
~u t a t~~ of clpX-A deletion-substitution mutation in clpX was constructed as described under uExperimental Procedures," and crossed from the plasmid into the chromosome. Pi was used to transduce the clpX::kan mutation from SG1152, the linear transformatjon host, into hosts carrying c1pP::cat mutations or proC mutations. proC is about 20% linked to the clpP-lan region. Among kanamycin-resistant transductants into the clpP::cat host, more than 95% had lost the chloramphenical resistance marker. When Pro+ recombinants of a proC host were selected, 8% were kanamycin resistant; a parallel experiment with a clpP::cat donor gave 16% chloramphenicol-res~s~nt transductants among Pro+ recombinants (data not shown). Therefore, the kanamycin resistance marker shows the expected location very close to clpP, and Highlander, S., and Weinstock, G. (1993) Infect. Immun Fig. 3; letters in bold face type are highly conserved sequences within these motifs. The underlined nucleotide sequences form an inverted repeat (possible transcription ~rmination sequence). The boxed nucleotide sequences indicate Ncol sites used to generate the deletion/insertion clpX::kan mutation.

CGCTGTTTAATCT~AA~GTGGATCT~AATTCCGTGACGAGGCGCT~ATGCTATCG L F N L E G V D L E F R D E A L D A I A
. Tail Motif B . linkage to proC is in the range expected, suggesting that clpX to both clpP-specific and clpX-specific probes (Fig, 4). A mutations are not detrimental to growth of these strains of transcript >2 kb encoding ClpP was also observed in expo-E. coli under these conditions. nentially growing cells by others? The amount of the tran-Analysis of ~r u~c r i~t~o n and Cbx Protein ~~~~~s~script in cells grown at 42 "C was about 2-fold higher than Northern blot analysis of RNA from wild-type cells revealed ~ a single major transcript of about 2400 bases that hybridized L. Simon, personal communication.    that found in cells grown at 32 "C, which is consistent with the 2-fold increase in ClpP protein synthesis reported for cells grown at 42 "C (10). A 2700-base RNA transcript from a deletion/insertion mutant of clpP, which retained about 276 bp of clpP coding DNA, reacted weakly with a clpP probe (with homology to 214 bp of clpP coding sequence) and very strongly with a clpX probe. The size of the transcript and the lack of temperature dependence suggests that the transcript originates within the chloramphenicol insertion, and the increased clpX message correlates with the increased amount of ClpX protein detected in Western blots (data not shown). RNA from a mutant with a deletion of the clpX gene and insertion of a kanamycin cassette also lacked the 2400-base transcript but had a longer transcript that hybridized strongly to a clpP probe and weakly (by virtue of about 550 bp of clpX DNA remaining) to a clpX probe. The latter transcript must originate within the kanamycin cassette and be transcribed in the opposite direction from clpP and clpX, although this interpretation has not been confirmed. Western blots of the c1pX::kan strains confirm the loss of a 46,000 dalton band present in the wild-type strains (data not shown).

ACGAGTCGGTAATTGATGGTCMAGCAAACCGTTGCTGATTTATGGCMGCCGGMGCGC E S V I D G Q S K P L L I Y G K P E A Q A A C A~T C T~T G A A T A A T T A A C~T T C C C A T A C~T T A G T T M C C~ Q A S G E * ~T T~T T T T~C T C T A T T C T C G G C G T T G A A T G T~~C
Abnormal Protein Turnover in c1pX::kan Mutant Hosts-When canavanine is incorporated into proteins in place of arginine, the resulting proteins are subject to rapid turnover. Lon protease accounts for about 50% of the turnover observed in wild-type E. coli cells (18). We used the turnover of canavanine-containing proteins as a measure of the contribution of ClpX to abnormal protein turnover. Degradation of canavanine-containing proteins in isogenic strains bearing various clp mutations alone or in combination with a lon mutation was measured. A clpX mutation either alone or in lon mutant cells had a modest effect (10-15%) on abnormal protein turnover, similar to that seen for a clpP mutation in the same strains (Table I). We had observed earlier that clpA and clpP mutations have only modest effects on the turnover of canavanine-containing proteins when Lon is absent (4,13).
Degradation of a ClpA-Lac2 Fusion Protein in clpX::kan Mutant Hosts-We have observed that a fusion protein carrying the first 40 amino acids of ClpA fused to @-galactosidase is unstable and is stabilized in either a clpA or clpP mutant host (13, 21). The difference in instability is sufficient to result in an easily measurable difference in the accumulation  of ,@-galactosidase in cells carrying the fusion. Table I1 shows the results of ~-gala~osidase assays in isogenic strains carrying mutations in either clpX or another clp gene. The absence of any effect of the clpX mutation suggests that clpX activity is not necessary for degradation of this protein, a substrate of ClpA and ClpP. 0 Protein-de~ndent Phage Growth-In the accompanying paper, Wojtkowiak et al. (17) have reported multiple protease activities capable of degrading hO protein in uitm. They estimate that about 50% of the 0 protein degradation activity in crude extracts is due to ClpP/ClpX. We were interested in testing the contribution of ClpX to 0 protein turnover in vivo.
We had previously found that mutations in lon or clpA did not affect 0 protein stability (4, 37).
We have used two different approaches to evaluate the role of clpP and clpX in 0 protein stability and function. In the first set of experiments, conditions under which 0 protein is likely to be limiting for growth were created, and the growth of an Oam phage was determined. The results of these experiments are shown in Table 111. In the first set of conditions, a host carrying a suppressor tRNA gene under the control of a dcu: promoter was grown under inducing conditions, washed away from inducer, and used as the host for growth of the Oam phage. Growth of a wild-type phage (data not shown) or an Nam phage (Table 111) was monitored as well, to control for other effects on phage growth. In efficiency of plating tests (data not shown) or burst size experiments (Table 111), cipX and clpP mutations showed significantly better phage growth for the Oam phage than clpA mutants, wild-type cells, or h n mutant cells. The Narn phage, on the other hand, grew better in cells mutant in lon, the protease known to degrade N (37).
In the second phage growth test, cells carrying a temperature-inducible defective prophage which is O+ and N" were induced for a short period of time, and then infected with the heteroimrnune Oam or Nam phage, and allowed to continue growth at low temperature. Under these conditions, the 0 protein or N protein made from the cryptic prophage comple- In protocol A, SG20250 derivatives carrying an imm2lplac-supF prophage were grown in the presence of isopropyl-1-2 thio-6-D-galactopyranoside (IPTG) as described under "Experimental Procedures," and infected with either imm434 Oam29 or imm434 Nam7Nam53 phage. Samples were taken and lysed with chloroform at 30-min intervals, beginning at 90 min after infection. The values shown in the table are for 90 min of infection. In the absence of IPTG, the burst size at 90 min was less than 1 for all strains.
In protocol B, SA1136 derivatives, which contain a temperature inducible cryptic 0' prophage, were induced for 10 min at 42 "C, and samples were then removed, infected with phage as described under "Experimental Procedures," and diluted into medium for growth with shaking at 30 OC. Phage growth was allowed to continue for 150 min.
For cells kept at 30 "C throughout the experiment, the burst size for Oarn phage in the SA1336 cZp+ host was 0.1, and for the clpP mutant host was 0.8. The Nam phage gave a burst of 0.1 in SA1336 and 0.2 in the lon mutant host. ments the incoming phage to allow growth. We have previously used experiments of this sort to demonstrate the functional decay of hN protein and its dependence on Lon protease (37). In the experiments done here, we found that complementation of the Oam phage was extremely poor under a variety of conditions and using a number of different prophages, making it impractical to monitor the functional decay of 0 protein in these strains. However, the ability of the prophage to complement an incoming O m phage was significantly increased in either a clpP or clpX mutant host, suggesting that 0 protein either accumulates to higher levels during the brief induction period or is more available for complementation during growth of the incoming phage.
ClpP and ClpX Are Required for 0 Protein Degradation in Viuo-The half-life of 0 protein was measured by immunoprecipitation of 0 protein from cell extracts following a pulselabeling with [36S]~ethionine and chasing with non-radioactive methionine in uiuo. Isogenic hosts carrying a c1857 Sam7 prophage were grown at 30 "C and induced for 10 min at 42 "C to initiate synthesis of 0 protein. Following labeling for 1 min, cells were returned to 37 "C and chased. Immunoprecipitated 0 protein from samples removed at various times was analyzed by gel electrophoresis and autoradiography. The results of this analysis for the wild-type and clpP and ctpX mutants are shown in Fig. 5A. 0 protein had a half-life of between 1 and 2 min in wild-type cells; this result was similar to that seen previously (37)(38)(39). However, in both the clpX and clpP mutants cells the 0 protein was considerably more stable, with a half-life of more than 40 min. In clpA mutant cells, the 0 protein half-life appeared to be about twice that in wild-type cells. These results indicate that 0 protein degradation in vivo is dependent on both ClpP and ClpX and that other proteases play little if any role in its stability.
To estimate the effect of clp mutations on the amount of 0 protein accumulation in induced lysogens, the same cf857 Sam7 lysogens were grown at 30 "C and transferred to 42 "C for 30 and 60 min. Samples were analyzed for the accumuiation of 0 protein by Western blot (Fig. 5B). In

DISCUSSION
In this paper we have reported the sequence of a novel subunit of the ATP-dependent Clp protease and have shown that this protein, which we call ClpX, is required in vivo for the degradation of the highly unstable 0 protein of lambda. Wojtkowiak et al. (17) were the first to purify ClpX and to detect its ability to stimulate ClpP to degrade 0 protein in an ATP-dependent reaction in vitro. While they report a number of other activities capable of degrading 0 protein in vitro, our results show that, in vivo, clpX and clpP mutations lead to complete stabilization of 0 protein, and that clpA, clpB, and lon mutations have little or no effect on 0 protein turnover. Thus, the protease formed by ClpX and ClpP (called ClpXP protease hereafter) is responsible for essentially all 0 protein turnover in vivo.
The Clp Protease Family-Clp protease was originally identified in vitro as an ATP-dependent protease consisting of two components, ClpA and ClpP (3,4,5,40). Degradation of substrate proteins and polypeptides by the proteolytic component, ClpP, absolutely requires the activity of the ATPase, ClpA (3,4,5,40). Mutations in either clpA or clpP stabilize the ClpA-LacZ fusion protein (21), derivatives of P-galactosidase with abnormal NH2-terminal amino acids (15, 14), as well as some other fusions to P-galactosidase,' suggesting that these abnormal proteins are all substrates for the protease formed by ClpA and ClpP (which we will refer to as ClpAP).
The results in this and the accompanying paper by Wojtkowiak et al. (17) indicate that the ClpXP protease has a substrate specificity distinct from that seen for the ClpAP protease.
In two other cases, evidence suggests that ClpX may be involved in ClpP-dependent degradation of specific substrates. A virulent Mu derivative is virulent and able to induce a resident prophage because the vir repressor is unstable and interacts with wild-type repressor to cause its degradation. That degradation was reported to be dependent on ClpP but not ClpA (16), and more recent results have shown that clpX mutations are at least as effective as clpP mutations in blocking the virulence of Muvir." A post-segregational killing mechanism encoded by phage P1 also appears to depend on ClpXP activity. Cells which lose a vector carrying the pair of genes responsible for post-segregational killing normally die; hosts mutated in clpP or clpX but not clpA, clpB, or lon survive? This is presumably due to stabilization of the protective protein, Phd, in clpP or clpX mutant hosts?
Other results also indicate that ClpA is not necessary for degradation of specific substrates whose degradation is dependent on ClpP. Shapiro (41) has observed that the instability of Mucts lysogens on extended incubation of colonies is blocked by mutations in clpP, but to a much lesser extent by mutations in clpA . Damerau and St. John (42) have observed that degradation of proteins made during starvation after cells are returned to high carbon is decreased in clpP mutants but not in clpA or clpB mutants. Because most activities of ClpP appear to be dependent on an activator ATPase, it is reasonable to assume that an ATPase other than ClpA is involved in these degradation reactions; the required subunit could be ClpX. These sharp distinctions between substrates affected by clpA mutations and those affected by clpX mutations support the idea that the substrate specificity of the Clp proteases is dependent on the regulatory subunit and that only one such regulatory subunit is usually required for an active protease complex. Which of the available subunits ClpP associates with in vivo may depend on the availability of substrates as well as the availability of the regulatory subunit.
ClpX and its homologs (ClpX from A. vinelandii and the ClpY open reading frames from P . haemolytica and B. subtilis) are significantly smaller than members of the ClpA family, which contain two highly conserved regions of approximately 200 amino acids in length, each containing a consensus sequence for an ATP nucleotide-binding site. The ClpX and ClpY proteins contain a single ATP-binding site consensus, with significant homology to those found in the ClpA/B/C family (Fig. 3). Beyond the ATP consensus sequences, ClpX, the ClpA/B/C family, and the putative ClpY proteins all share scattered homologies, most strikingly two clusters of amino acids in the COOH terminus, a region which is not highly conserved for the ClpA family except for these conserved patches (Ref. 7, Fig. 3). While the function of the COOH-terminal tail has not yet been determined, it seems possible that these conserved sequences are involved in interactions of these regulatory subunits with ClpP or ClpP-like subunits.
Unique to the two ClpX proteins is a cluster of cysteine residues, possibly a zinc binding motif, in the amino-terminal region; no function has yet been assigned to this structure for ClpX. The ClpY family, for which no in vivo or in vitro activity has yet been determined, lacks the cysteine cluster and is distinguished by a >lo0 amino acid insertion between parts A and B of the ATP consensus sequence. It is an *A. Toussaint, personal communication. M. Yarmolinsky, personal communication intriguing possibility that a protein of the ClpY family might also exist in E. coli and might work with ClpP or a different protease to promote specific degradation of yet another set of substrates.
The occurrence of multiple ATPase subunits that can activate the proteolytic ClpP and direct it to different substrates has important implications for other ATP-dependent proteases. The eukaryotic ATP-dependent 26s protease contains the proteasome core associated with a number of other proteins that can modify its activity (43, 44). Among the associated proteins are one or more ATPases, which presumably carry out functions analogous to those performed by ClpA and ClpX for ClpP. Recently, Rechsteiner and co-workers (35) have reported that subunit 4 of the 265 protease, which has a single ATP-binding site consensus, is distantly related to ClpA. Subunit 4 is itself a member of a large family of ATPases which have been implicated in protein localization and gene regulation; these authors suggest that other members of this family may also be protease regulatory subunits. A second, 100-kDa subunit of the 26s ATP-dependent protease apparently has homology to the cdc48 subfamily of these ATPases (45). The Subunit 4 sequence reported by Dubiel et al. does not reveal any closer similarity between ClpX and subunit 4 than between ClpA and subunit 4. In particular, the motifs in the tail that are highly conserved between ClpX/Y and the ClpA/B/C family (Fig. 3) are not found in subunit 4 or its more closely related relatives. It will be interesting to see whether any of the other subunits of the 26s protease bear a closer resemblance to the Clp family subunits.
Unlike either ClpA or ClpB, the other member of the ClpA family found in E. coli, ClpX is synthesized as part of an operon with ClpP. Therefore, it shares with ClpP the ability to be induced as a heat shock protein. Our results indicate that a single message carrying both clpP and clpX is the primary transcript at both low and high temperatures. The 2fold increase in the amount of message at high temperature is consistent with the 2-fold increase in ClpP protein synthesis reported by Kroh and Simon (10). Since the CIpX homolog in A. uinektndii is in the middle of an operon of nif (nitrogen fixation) genes (31), the location next to ClpP is apparently not a conserved feature of this protein.
0 Protein Stabitity and X Replication"X0 protein, like a number of other important regulatory proteins in bacteriophage h, is rapidly turned over in vivo. We show here that this in vivo turnover is dependent on the activity of a novel protease composed of ClpP, the protease subunit of the Clp ATP-dependent protease (8) and ClpX, a new regulatory subunit capable of working with ClpP. The rapid turnover of hO protein has been assumed until relatively recently to reflect its role as a rate limiting step for lambda replication. Recent work by Taylor and co-workers (39,46), in which they observed the accumulation of a stable subpopulation of 0 protein, apparently in the replication complex, suggests that 0 protein instability may not limit replication after all. Here we find that stabilizing 0 protein by mutations in c1pX or clpP has no discernible effect on the burst size of an induced ~1857 Sam7 prophage. However, under conditions where the availability of 0 protein is limited by either low level suppression of phage carrying Oam29 or by complementat~on of an Oam phage by transient induction of a cryptic prophage, it is clear that clpX or clpP mutants allow s i~i f i c a n t increases in phage yield (Table 111). Therefore, it would appear that the rapid turnover of 0 protein limits replication activity only under suboptimal conditions, when 0 protein becomes limiting.
Engelberg-~ulka and co-workers (47) found that expression of h protein RexB from a plasmid stabilizes 0 protein made from either the same or a compatible plasmid. While RexB is made in small amounts in lysogens, its synthesis increases from the plit promoter when replication of X increases (48). They suggest that this increase in RexB synthesis results in 0 protein stabilization, which in turn allows further increases in replication. As we have noted here, at least under induction conditions, 0 protein stability does not limit burst size, SO if this circuit exists it must provide an additional safeguard for protection of 0 protein under special conditions. X utilizes a number of unstable proteins at important stages in its life cycle; it is striking that thus far, each of these proteins is subject to proteolysis by a different system. Thus, cI1 de~adation depends on the products of the hft locus (49, 50f, AN protein is subject to Lon degradation (37), CI degradation, part of the SOS response, is dependent on RecA, and sequences within the CI protein (51). Here we show that 0 protein is turned over by a novel Clp variant, the Clp XP protease. Axis protein is also known to be ~nctionally unstable (52) and is not stabilized in ton mutants~~' its degradation may be dependent on yet another E. coli protease. ~h e t h e r the use of these different proteases is simply accidental or reflects some ad~itional regulatory network for X is not yet clear.