Cardiolipin Activation of dnaA Protein , the Initiation Protein of Replication in Escherichia coZi

ATP binding to dnaA protein is essential for its action in initiating the replication of plasmids that bear the unique origin of the Escherichia coli chromosome (oriC). ADP bound to that site renders dnaA protein inactive for replication. Diphosphatidylglycerol (cardiolipin), a diacidic membrane phospholipid, displaces the bound nucleotide, and in the presence of components that reconstitute replication, fully reactivates the inert ADP form of dnaA protein. The monacidic phosphatidylglycerol is one-tenth as active as cardiolipin, whereas the neutral phosphatidylethanolamine, the principal E. coli phospholipid, is inactive. Fluphenazine, a tranquilizer drug, blocks cardiolipin activation of dnaA protein, in keeping with the inhibitory action of such agents on phospholipid-dependent enzymes. With the use of this drug to terminate cardiolipin action, dependence of the activation on time, elevated temperature, and high levels of ATP was demonstrated. Cardiolipin binding of nucleotide-free dnaA protein prevents binding of ATP and initiation of oriC replication. Removal of a fatty acid from cardiolipin by phospholipase A reverses this inhibitory effect. The strong and specific interaction of cardiolipin, a cell membrane component, with an essential nucleotidebinding site of dnaA protein, the protein essential for the initiation of chromosome replication, may be an important element in regulating the cell cycle.

in initiation at the origin of the Escherichia coli chromosome (oriC) is dnaA protein (1, 2). Upon examining the properties of the purified protein, we discovered that its action depends on tightly bound ATP which it hydrolyzes to ADP in about an hour, the time interval of a cell cycle (3). Because the ADP renders the protein inert and does not exchange with ATP, this apparent suicide may be programmed to provoke the need for fresh synthesis of dnaA protein to initiate the next chromosome cycle. On the other hand, there may be mechanisms designed to rejuvenate the inactive ADP form by facilitating the replacement of ADP by ATP. To date, no protein has been found to fill this role.
In this report we describe the remarkable property of acidic * This work was supported by grants from the National Institutes of Health and the National Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $Fellow of the American Cancer Society, California Division. Present address: Faculty of Pharmaceutical Sciences, University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113, Japan. phospholipids, particularly cardiolipin, to dissociate the adenine nucleotide bound to dnaA protein and in the presence of ATP and the components of the replicative system to restore the inactive ADP-form to full activity. Despite the plausibility of linking chromosome replication to cell m e m b r~e s (4), chemical evidence for such associations has been lacking (5).
In this instance, the strong and specific interaction between a membrane phospholipid and a protein vital for initiation of replication deserves attention for the role it may perform in controlling this process.

EXPERIMENTAL PROCEDURES
Reagents-Sources were as follows: ATP, dNTPs, and HEPES,' Pharmacia P-L Biochemicals; GTP, GTP, UTP, and Tricine, Sigma; Enzymes and DNA-Purified DNA replication proteins were prepared as described previously (6). dnaA protein was purified (2) to a specific activity of 1.0 X lo6 unitsfmg and a purity greater than 90% as judged by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Plasmid pCM959 (7), a gift from M. Meijer (Unive~ity of Amsterdam, The Netherlands), is a minichromosome. It consists solely of E. coli DNA encompassing oriC (base pairs -677 to +3335).
Cardiolipin-dependent DNA replication was separated into two stages, an activation process of the ADP form of dnaA protein and a replication reaction, by using fluphenazine, a specific inhibitor of the former stage (see "Results"). dnaA protein (120 ng) was preincubated with 1 pM ADP at 0 'C for 15 min. Incubation was continued for 10 min at 38 'C with 5 mM ATP, 0.35 fig of cardiolipin, and the other replication components, then fluphenazine was added to 0.15 mM and the incubation was continued at 38 "C. DNA synthesis was for 30 min at 16 'C with dNTPs.

Cardiolipin Dissociates ADP or ATP Tightly Bound to d n d Protein-No
E. coli protein has been found that can release or exchange the ADP or ATP tightly bound to dnaA protein (& = 0.03 @). Of special interest is the fact that the ADP form generated by the slow hydrolysis of the ATP bound to dnaA protein is inert in replication (3). Two ways have been found to effect the release of the bound nucleotides. One is by the unphysiological chelation of magnesium (12). The other is through the action of certain phospholipids, particularly diphosphatidylglycerol (cardiolipin) (Fig. 1). Eighty percent of the nucleotide bound to 0.12 gg of dnaA protein (about I PI -. After exposure to a phospholipid at 38 "C for 10 min, the samples were filtered on membranes. PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PZ, phosphatidylinositol; CL, cardiolipin. 2 pmol) was dissociated after 10 min at 38 "C by 0.2 pg of cardiolipin (about 200 pmol). Without cardiolipin the nucleotide-protein complex remains intact (>go%) for at least 60 min.
Among the major phospholipids of E. coli, phosphatidylgly-cero1 was only about one-tenth as active as cardiolipin. Phosphatidylethanolamine, the principal phospholipid in E. coli membranes, was inert ( Fig. 1). Phosphatidylinositol, absent from E. coli, had an activity level near that of phosphatidylglycerol. Hydrolysis products of cardiolipin (&glyceride and glycerophosphate) were inactive (data not shown). Thus, an intact phospholipid is required. Its anionic nature is important, explaining in part why cardiolipin with two phosphate moities is the most active. Dissociation of bound ATP (or ADP) by cardiolipin required a high temperature (Fig. 2). The    Fig. 3) was incubated for 10 min a t various temperatures with 0.35 pg of cardiolipin and the c o m~n e n t s of the replication system except dNTPs. Fluphenazine (0.15 mM) was added and incubation continued at 38 "C for 10 min; then, upon the addition of [32P]dNTPs, DNA synthesis was carried out a t 16 "C for 30 min. in the presence of the components needed to reconstitute replication of the oriC plasmid resulted in complete regeneration of the replicative activity of the previously inert dnaA protein (Fig. 3A). The relative activities of the p h o s p h o~p i~ and the amounts needed were essentially the same as observed for ADP release from dnaA protein, except that at higher levels, cardiolipin and phosphatidylglycerol became inhibitory.
When the mixture of phospholipids found in E. coli was used in place of the pure phospholipids (Fig. 3B), several noteworthy facts emerged. (i) Regenerative activity on ADP.
dnaA protein was as great as with cardiolipin alone. (ii) The amount needed was consistent with that anticipated from cardiotipin being present at the usual level of 510% of E. coli phospholipids. (iii) Phospholipids in the lamellar state were active. (iv) Cardiolipin from E. coli was as active as that from a bovine source. (v) No inhibitory action was seen with high levels of the phospholipid mixture as with pure phospholipids. (vi) Despite differences between the metabolic states of phospholipids in logarithmic and stationary growth-phase cells, no distinction was observed in their action on dnaA protein.  Whether the regenerative action of cardiolipin on ADP. dnaA protein depended on an elevated temperature as observed for dissociation of ADP and whether ATP influenced the process could not be tested until a means was found (see below) to separate the stage of car~olipin action from the subsequent initiation stages known to depend on an elevated temperature and high levels of ATP (12, 13). F~u~~~z i~ ~n~~~ C~r d~o Z~p~n Release of ~u c~o t i d~s Bound to dnaA Protein-To separate the cardiolipin action on dnaA protein from subsequent events in which dnaA protein initiates replication, a specific inhibitor of the cardiolipin action was sought. Fluphenazine, a tranquilizer drug, was tried in view of the capacity of this class of agents to inhibit the actions of phospholipid-dependent enzymes (14, 15). Fluphenazine did inhibit the phospholipid activation of ADP. dnaA protein without affecting the replicative activity Activation of dnaA Protein of the ATP form of the protein (Fig. 4). The latter showed the same activity with or without cardiolipin present. The same mixture of cardiolipin and fluphenazine did not inhibit any stage of DNA synthesis (data not shown).
Using fluphenazine to terminate the action of cardiolipin, the need for a temperature near 38 "C for activation of dnaA protein was demonstrable ( Table I). ATP at a level about lo5fold higher than that which saturates the dnaA protein tightbinding site was also needed (Fig. 5).

The Nucleotide-free Form of dnaA Protein Is Inactivated by Cardiolipin and Can Be Reactivated by Phospholipase Az-
Without a nucleotide (e.g. ATP) in its tight binding site, dnaA protein was inactivated by cardiolipin at 0 "C, as measured by firm binding of ATP (Fig. 6) or by its replicative activity (Fig.  7). Thus, cardiolipin binding blocks access of ATP to the site essential for dnaA protein action. At 38 "C in the absence of the components of the replication system, cardiolipin, after displacing the ATP, blocks this site and thereby inhibits dnaA protein functions. The ATP binding (Fig. 8) and replication ( Fig. 9) activities were restored to the inert cardiolipin-dnaA protein by destroying the cardiolipin through the action of phospholipase Az. Cardiolipin denatures nucleotide-free dnaA protein at high temperatures. Probably binding of ADP (or ATP) to dnaA protein prevents the irreversible inactivation of dnaA protein by changing the conformation of the protein.

DISCUSSION
Activation of dnaA protein by cardiolipin (diphosphatidylglycerol) was discovered in our efforts to resuscitate the moribund protein after its attempt at suicide. The dnaA protein with a very tightly bound ATP, essential for its action, hydrolyzes the nucleotide slowly to ADP (3). The ADP product, held just as firmly, renders the protein inert in the initiation of replication. Replacement of the bound ADP by exchange with ATP is very feeble. No counterpart to agents that dissociate GDP from the G proteins (16) was found among the purified proteins involved in replication or those in crude cell extracts. Although the bound nucleotide can be discharged by removing M$+ with strong chelation (12), implying a requirement for the metal by the binding site, this mechanism seems unphysiological.
In view of the possible association of dnaA protein and replication events with cell membranes, the influence of phospholipids was tested. Cardiolipin proved to be strikingly effective in the rapid displacement of the bound nucleotide and rejuvenation of the ADP form of the enzyme. Among the E. coli phospholipids, cardiolipin, the least abundant (about 5% of the total), is 10 or more times as effective as the %fold more abundant phosphatidylglycerol. Phosphatidylethanolamine, comprising nearly 80% of the phospholipids, is inert. Cardiolipin is equally active whether of bovine origin presented in some micellar form or E. coli origin arranged in a lamellar (vesicular) state among the other E. coli phospholipids. Phosphatidylinositol, not found in E. coli, is about as effective as phosphatidylglycerol, indicating the importance of the acidic head group in this interaction with dnaA protein. Cardiolipin, with two ionized phosphates, is reasonably the most active. Because dnaA protein requires Mg2+ for tight binding of ATP or ADP (12), destabilization of the M$+ by cardiolipin is a possible basis for discharge of the nucleoside phosphates.
The isosteric similarity between the cardiolipin head group and the DNA backbone is striking and is likely the basis of the cross-reactivity between cardiolipin and DNA (17). In considering where on the dnaA protein cardiolipin might be located, the DNA-binding site is a possibility that deserves further study. It is noteworthy, however, that the several DNA-binding proteins in the replication system are not detectably affected by cardiolipin in the reaction.
How can we assess the physiological significance of the influence adenine nucleotides have on dnaA protein function and the relief of ADP inhibition by cardiolipin? It would be desirable to know the nucleotide form of the protein during the cell cycle. Among the thousand dnaA protein molecules in a cell (18), only a small fraction is engaged at the chromosomal origin. The rest are bound to dnaA boxes elsewhere on the chromosome or to the membrane surface. Thus, a determination of the form of the protein needs to be coupled with knowledge of its location. As for attachment of dnaA protein to cell membranes, particularly cardiolipin, and the influence of the cell cycle, information about the physiology and arrangement of phospholipids in E. coli membranes is rather limited.
Turnover of the acidic phospholipids is relatively rapid compared to that of phosphatidylethanolamine (19,20), the breakdown of phosphatidylglycerol appears to be coupled to a stage in the cell cycle (21), and a requirement for phospholipid synthesis for initiation of replication has been suggested (22). Yet a mutant deficient in cardiolipin synthesis sustains apparently normal growth (23). Still, this mutant retains a low level of cardiolipin, approximately 20,000 molecules, a number far in excess of that needed to bind the dnaA protein operating at chromosomal origins, and phosphatidylglycerol levels in this mutant are elevated. While attempts to delete the cardiolipin synthetase gene had not succeeded (24), the recent construction of a null allele of the pgsA gene renders the cell incapable of synthesizing phosphatidylglycerol or cardiolipin and is lethal unless rescued by a plasmid-borne copy of the gene (25). Inasmuch as cells manage a normal phenotype with very low levels of these acidic phospholipids, it would seem that these residual amounts are not serving a structural role in the membrane but rather are supplying some essential functions such as the orientation of proteins in key biosynthetic processes (26).
Several arguments can be cited for a physiological membrane base for dnaA protein beyond the interactions with acidic phospholipids described in this report. In a membrane fraction which selectively binds the oriC sequence (27), the dnaA protein, identified by immunoblotting, is highly enriched, and the anti-dnaA antibody blocks the binding of oriC by the membrane fraction.2 Among a number of cold-sensitive mutants (often associated with a membrane location), the dnaA gene locus was repeatedly identified with conditionally * L. I. Rothfield, personal communication.
lethal defects in replication. Finally, of the large number of proteins in a particulate fraction of a cell extract, only one was found which bound ATP with extraordinary avidity and this proved to be the dnaA protein (18).