Purification and characterization of DNA polymerase III from Bacillus subtilis.

DNA polymerase III from Bacillus subtilis has been purified about 4,500-fold. Disc gel electrophoresis of the purified fraction reveals a single major protein band which co-migrates with the polymerase activity. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis of the polymerase yields a single, 166,000 dalton band. The hydrodynamic properties of the enzyme are ionic strength-dependent. The average values from determinations in high and low salt are 7.6 S for the sedimentation coefficient and 52 A for the Stokes radius. These two parameters indicate a molecular weight for the native enzyme of 160,000. Therefore, the enzyme appears to contain a single, long, polypeptide chain. The enzyme has no endonuclease activity but does have single strand specific exonuclease activity. Hydrolysis is initiated exclusively from the 3' terminus yielding 5' mononucleotides, and a dinucleotide is the limit of digestion. The exonuclease activity has an ionic strength dependence of pH optimum similar to that of the polymerase but appears to be more fastidious in its divalent metal requirement. The mode of attack by the enzyme is strictly distributive. The activity of the exonuclease decreases markedly with increasing substrate size. Two opposing mechanisms account quantitatively for this effect--intrinsic competitive inhibition by interior substrate nucleotides and increasing accessibility of the substrate terminus to the enzyme with increasing chain length. The polymerase synthesizes DNA in the 5' leads to 3' direction and the apparent Km for each of the deoxyribonucleoside triphosphates is about 1 muM. The polymerase replicates RNA-primed, phiX174 DNA in the presence of Escherichia coli elongation Factors I and II. In contrast to polymerase III, B. subtilis DNA polymerase II has no detectable nuclease activity.

by the enzyme is strictly distributive.
The activity of the exonuclease decreases markedly with increasing substrate size. Two opposing mechanisms account quantitatively for this effect-intrinsic competitive inhibition by interior substrate nucleotides and increasing accessibility of the substrate terminus to the enzyme with increasing chain length. The polymerase synthesizes DNA in the 5' -3' direction and the apparent K, for each of the deoxyribonucleoside triphosphates is about 1 PM. The polymerase replicates RNA-primed, @Xl74 DNA in the presence of Escherichia coli elongation Factors I and II. In contrast to polymerase III, B. subtilis DNA polymerase II has no detectable nuclease activity.
Bacillus subtilis has three distinct DNA polymerases, designated as polymerases I, II, and III (l-3). Studies of DNA polymerase I mutants demonstrated that this enzyme acts in the repair of chromosomal damage (4,5). The function of DNA polymerase II is not known. The central role of DNA polymerase III in replication of the chromosome was firmly established by the isolation of an arylhydrazinopyrimidine-resistant mutant whose purified DNA polymerase III was drug-resistant (6); the arylhydrazinopyrimidines, such as OHPh(NH) JJra, * selectively inhibit DNA polymerase III (7)(8)(9). Subsequently, several temperature-sensitive polymerase III mutants have been shown to fail to synthesize DNA at nonpermissive temperatures (10,11). In addition to synthetic activity, DNA polymerase III catalyzes the hydrolysis and pyrophosphorolysis of DNA and the exchange of PP, with deoxyribonucleoside triphosphates (11,12). The intrinsic nature of these activities is shown by their OHPh(NH),Ura sensitivity, and the exonuclease and polymerase activities co-purify, have identical *This work was supported by National Institutes of Health Grants GM-21397 and CA-14599. $ Supported by Public Health Service Grant HD-COOOl. ' The abbreviation used is: OHPh(NH) ,Ura, 6-(p-hydroxyphenylhydrazinoj-uracil. thermolability in enzyme purified from wild type and temperature-sensitive mutant cells, and are equally scavenged into the ternary complex with OHPh(NH),Ura and DNA (11)(12)(13). In order to gain insight into the mechanism of chromosomal replication, the enzymological properties of B. subtilis DNA polymerase III have been examined (2,3,12,14). The enzyme has also been used to determine the mechanism of inhibition by the arylhydrazinopyrimidines with a molecular detail not yet achieved for any other DNA synthesis inhibitor (reviewed in Ref. 15). In this paper, we report the extensive purification of the enzyme to near homogeneity and some of its physical and enzymological properties. A quantitative analysis of the intrinsic exonuclease properties is also presented. Some of these results have been summarized recently (11,13). MATERIALS AND METHODS Growth of Bacteria-Bacillus subtilis DNA polymerase III was purified from the polymerase I-deficient strain BC26(F) (2). Bacteria were grown as described previously except the harvested cells were washed twice with "subtilis salts" before storage at -20" (2 (2 x 10" cpm/pmol) was prepared by the method of Glynn and Chappell (17) and purified by DEAE-Sephadex chromatography.
Unlabeled poly(dG), poly(rU),and d(T) ,I.,B were obtained from Collaborative Research. Poly(dA) was synthesized enzymatically (18); d(T), and the deoxyoligonucleotide pG-C-T-T-C-C-C-G-A (19) were gifts of Dr. K. Agarwal. E. coli DNA was isolated by the method of Marmur (20) and shows the result with the native DNA substrate. Since less than 5 x lo-' nmol of phosphodiester linkages were hydrolyzed, the polymerase to endonuclease ratio is at least 106. In a control reaction, 0.01 unit of pancreatic DNase I quantitatively converted each of the substrates to material sedimenting near the top of the gradient (data not shown).

Requirements of Intrinsic Exonuclease Activity
The intrinsic exonuclease activity has a stricter requirement for MgCl, than the polymerase activity. Essentially no exonuclease activity ( <2%) is seen in the absence of MgCl,, either in the presence or absence of 0.5 mM EDTA, whereas addition of the chelator is necessary to abolish totally polymerase activity. The MgCl, concentration optimum is quite broad; exonucleolytic activity was maximal from about 4 to 18 mM and fell to 50 and 68% maximal activity at 1. 3  Downloaded from activity appears quite low in the presence of MnCl,. In the range from 0.5 to 20 mM MnCl,, with no MgCl, present, essentially no nucleolytic activity was seen (<3%). The synthetic activity in the presence of 0.6 mM MnCl,, the sharp optimal concentration, was 60%' of that with 6.5 mM MgCl,. The polymerase and nuclease activities also respond differently to spermidine.
In the absence of MgCl, and in the presence of 0.15 mM EDTA and 3 mM spermidine, about 80% of the optimal polymerase activity with MgCl, was seen. However, in the absence of MgCl, and either in the presence or absence of 1 mM EDTA, 1 or 3 mM spermidine did not promote any exonuclease activity. In fact, in the presence of 6 mM MgCl,, 3 mM spermidine caused a 72% inhibition of nuclease activity.
The pH optimum of the exonuclease activity, pH 7.4, is the same as that of the polymerase activity (2). For the exonuclease, at pH 6.4 and pH 8.4 there was 60 and 46%, respectively of the activity seen at pH 7.4. The exonuclease is quite sensitive to ionic strength. In the presence of 30 mM Tris-HCl, pH 7.4, optimal exonuclease activity occurred at about 10 mM KCl, and at 90 mM KC1 there was 50% of maximal activity (Fig. 6). This ionic strength sensitivity is very similar to that seen with polymerase activity (2). Again like polymerase activity, the exonuclease is quite sensitive to sulfhydryl reagents. Maximal activity requires the presence of sulfhydryl compounds, such as 1 mM dithiothreitol, 2 mM GSH, or 3 mM 2-mercaptoethanol.
N-Ethylmaleimide at 7 mM and p-chloromercuriphenyl sulfonic acid at 0.8 mM caused about 90% inhibition.
While some nucleases are greatly affected by nucleotides (41), there was no effect of moderate concentrations (550 j&M) of either deoxy-or ribonucleoside triphosphates on the polymerase III exonuclease. Very high concentrations of nucleoside triphosphates are inhibitory. A mixture of each of the four common deoxyribonucleoside triphosphates at 200 PM resulted in a 40% inhibition; this concentration is over 2 orders of magnitude greater than their apparent K, values (see below). B. subtilis polymerase III does not appear to have a "nick translating" activity like the 5' --t 3' exonuclease of E. coli polymerase I (42). There is minimal degradation of nicked, minicircular DNA ( Fig. 1) and this activity, as well as that on DNase I-treated E. coli DNA, is not stimulated by deoxyribonucleoside triphosphates.
Exonuclease Activity on Various Templates The relative activity of the exonuclease on several substrates is shown in Table II. The exonuclease is considerably more active on single-stranded than double-stranded DNA. Native, duplex E. coli DNA which had been separated from singlestranded DNA fragments by benzoylated, naphthoylated DEAE-cellulose chromatography was at least 100 times less active than the same substrate which had been briefly heat-denatured for 3 min at 100". Similarly, poly(dA) essentially abolished the vigorous activity seen on poly(dT). As analyzed in detail in the subsequent section, the exonuclease prefers short single-stranded DNA, such as sonicated, heat-denatured DNA and the nonanucleotide, to long single-stranded DNA. Some physical property of the long substrate must retard hydrolysis. As also indicated in Table II, terminal 3'-phosphoryl groups on single-stranded DNA blocked hydrolysis. The low activity with an equal mixture of 3'-hydroxyl-and 3'-phosphoryl-terminated DNA indicated that 3'-phosphoryl DNA is  We have shown that the DNA polymerase III exonuclease hydrolyzes short DNA chains much faster than long DNA chains even at saturating substrate concentrations (Table II  and Ref. 12). This large effect of chain length on reaction rate, observed also with some other exonucleases (43, 44), suggests that regions of the DNA substrate distant from the site of catalysis influence activity. A plausible explanation for this general phenomenon pointed out by Huang and Lehman (43) is that the interior nucleotides provide nonproductive binding sites and thus activity falls off as the ratio of interior to 3'-hydroxyl terminal residues increases with increasing chain length. To test directly this mechanism of intrinsic competitive inhibition for the B. subtilis polymerase III exonuclease, the ratio of interior to terminal substrate residues was controlled by adding increasing amounts of circular, single-stranded, fl DNA to a fixed amount of labeled substrate (Fig. 7). According to prediction, the circular DNA is a good inhibitor of the hydrolysis of E. coli DNA; the rate was half-maximal at a concentration of circular DNA about four times that of the linear DNA. The apparently smaller effect on poly(dT) hydrolysis is in large part due to the direct proportionality of the slope of these plots to the measured K,IV,,, which, as demonstrated below, is much smaller for the homopolymer substrate. While the linearity of the plots shown in Fig. 7 indicates a simple inhibition scheme, at higher circular DNA concentrations (not shown in the figure) inhibition was less than predicted by the extrapolated line, showing secondary effects.

pG-C-T-T-C-C-C-G-
If intrinsic competitive inhibition occurring at interior nucleotides is the sole effect of chain length on hydrolytic activity, then the kinetics of the reaction should obey the following rate equation:

DNA Polymerase
IIIfrom Bacillus subtilis kc., Eo S 0: where k,,, is the catalytic rate constant; E, the total enzyme concentration; S, the substrate concentration in terms of 3'.hydroxyl termini; K,, the dissociation constant for the enzyme.inhibitor complex; and i, the concentration of internal nucleotide inhibitor.
Since here i is defined by S and n, the chain length of the substrate in nucleotides, then if only a fraction, cy, of the internal nucleotides are active competitive inhibitors, this equation can be transformed into: K kc., Eo S K, + K, a (n -1) V=

Si
Km K, K, + K, a (n -1) Thus the measured, or apparent, K, and V,,, (KM'* and Vi::",) are equal to K, K,I[K, + K, (Y (n -l)] and K, k,,, Eo/[K, + K, LY (n -l)], respectively. This simple competitive case predicts that both K&S! and Vi:", will vary inversely with n while the ratio of these two parameters will be constant. These relationships for intrinsic competitive inhibition contrast with competitive inhibition by an added compound where the inhibitor causes an increase in K%-'p and does not alter V%p,.
To test the prediction of the rate equation, K,F' and V$$% were measured for the DNA and poly(dT) substrates listed in Table III; the data for the poly(dT) and DNA substrates will be treated separately since these homopolymers are much better substrates and are free of secondary structure. The chain length of the substrates was measured by end group labeling with polynucleotide kinase and/or alkaline velocity sedimentation. Lineweaver-Burk plots for some of the substrates are shown in Fig. 8 and the calculated kinetic constants are listed in Table III. While all the double reciprocal plots were linear, the slopes and intercepts differed markedly from substrate to substrate. For the natural DNA and poly(dT) substrates, KY' varies inversely with n in the predicted fashion over at least a 3-order of magnitude range as shown by the good agreement between the measured K,P and the K,P calculated from the value of n and the rate equation (Table III). In contrast, the magnitude of V$,p! and V$~$'fK$P' does not vary as expected. While V,% does decrease with chain length for the natural DNA substrates, it does not vary according to the predicted hyperbolic function, and the effect of n on V,,\t is considerably smaller than on Kkpp. For the poly(dT) substrates, the very high Vi::: is essentially independent of n over the range of tested chain lengths. The V$$,PIK ","p instead of being constant increases with n for both DNA and poly(dT). The simplest explanation is that there is an effect of chain length in addition to internal competitive inhibition which tends to increase the reaction rate as n increases. A physical interpretation of this unexpected result in terms of accessibility of the 3' terminus. to enzymatic attack is presented in the "Discussion."

Hydrolysis Proceeds 3' + 5' Exclusively
Besides a more usual, predominant 3' + 5' exonucleolytic activity, an additional unique 5' + 3' activity has recently been found associated with E. coli DNA polymerase III (45). A 5' + 3' exonuclease may be involved in the processing of Okazaki fragments and in the repair of DNA. Three observations had       The cellulose thin layer plate was developed until the bromphenol (BP) dye marker approached the top of the plate.
extensively degraded [5'-S2P]-pG-C-T-T-C-C-C-G-A implied that the limit digest was a dinucleotide, [5'-'*P]pG-C (Fig. 11). To confirm this identification, the digest was partially degraded with snake venom phosphodiesterase and subjected to electrophoresis at pH 5.5 (Fig. 12A). Before diesterase treatment, the nonanucleotide limit digest migrated as a single peak in a position expected for a dinucleotide, while after diesterase treatment there were two peaks, the undegraded dinucleotide and 5'-dGMP.  (Fig. 12B). About 30% of the *lp migrated with an internal reference of d(T),, the limit digest, while the remainder migrated as d(T), (47), a partial digest. No [szP]dTMP was seen.

Polymerase ZZZ Can Synthesize in 5' -3' Direction
Having shown that the exonuclease degrades exclusively from the 3' terminus, a simple but novel experiment was performed to determine the direction of polymerization. The template-primer was [5'-8*P]d(T) ,,.,,:poly(dA). Polymerization of [aH]dT"I'P in the 3' -+ 5' direction will displace the **P from the 5'-terminal position to an internal phosphodiester linkage and thereby render the s*P resistant to hydrolysis by bacterial alkaline phosphatase. Synthesis in the expected 5' --+ 3' direction (41), however, should not affect the phosphatase sensitivity of this 5'-terminal a*P. As shown in Table V, only a negligible amount of the *lP was phosphatase-resistant, 0.01 pmol, while 54 pmol of [*H]dTMP were incorporated. If only a small percentage of the oligo(dT) primers were elongated, then even if synthesis were 3' + 5', an undetectable amount of $lP would be rendered phosphataseresistant. If the enzyme were completely processive, for example, only 7 x loss pmol of primers (the quantity of enzyme) would be extended. However, the poly(dT) product should then be 8 x lo6 nucleotides long and readily acid-precipitable. In fact, the incorporated tritium was not any more acidprecipitable than the s2P-labeled primers (about 4%/o), which indicates a distributive incorporation. Therefore, B. subtilis of 32P and the [aH]nucleotide incorporated were measured by adsorption to DE81 of 25 ~1 of reaction mixture; one-half of the remaining reaction mixture was used to measure acid-precipitable product; 5 mM N-ethylmaleimide was added to the other half which was followed by successive incubations with 1.3 units of bacterial alkaline phosphatase at 37", and 1.3 units of phosphatase at 65", and adsorption to Norit. We have shown for B. subtilis polymerase III that the measured K, for dGTP is 0.6 pM (7). Since the K, for a mixture of the four common deoxyribonucleoside triphosphates is at least 1 order of magnitude higher for E. coli polymerase III (49, 501, the K, was determined for dCTP, dATP, and dTTP for the B. subtilis enzyme. The results were 0.9 PM for dATP, 1.5 PM for dTTP, and 0.3 PM for dCTP, and thus the difference between the two DNA polymerases remains.

Triphosphates
Because of the utility of ribonucleoside incorporation by DNA polymerases in sequencing DNA (51), the ability of polymerase III to incorporate rGTP, rCTP, rUTP, and rATP was tested. Either in the presence of a full deoxy-or ribonucleotide complement, or in the presence of a single ribonucleotide, the polymerase was able to utilize only rGTP appreciably. The rGTP was incorporated about Ih as well as dGTP in the presence of 0.5 mM MnCl, and negligibly in the presence of 6 mM MgCl,.
In this paper we describe the 4500-fold purification of Bacillus subtilis DNA polymerase III from a mutant strain lacking DNA polymerase I. DNA polymerase II was quantitatively removed during the first five steps. The yield of polymerase III was about 3%. The alternative procedure which substituted hydroxylapatite for DNA-cellulose in the final step gave a three times higher yield of enzyme which was similarly free of contaminating nucleases but had a lower specific activity (Table I). The enzyme from the penultimate phosphocellulose step had a nuclease contamination about equal to the exonuclease activity of the polymerase. The isolation of a relatively large amount of purified enzyme has permitted a more careful characterization of the enzyme than reported previously (2,3,12,14). particularly in regard to its exonuclease activity and physical properties. We reported previously that B. subtilis DNA polymerase II had little or no nuclease activity on E. coli DNA (2). In view of the striking dependence of the B. subtilis DNA polymerase III exonuclease on the chain length, secondary structure, and base composition of the substrate, the possible presence of nuclease activity in polymerase II was reinvestigated.

Lack of Intrinsic
The enzyme elution profile from DNA-cellulose, the final purification step, shows no significant nuclease activity under the sharp polymerase peak (Fig. 13). The peak fractions of polymerase were pooled, concentrated, and tested for nuclease using long incubation times. There was still no hydrolysis of native and denatured E. coli DNA and d(T),,; an activity SW of the polymerase activity could have been readily detected.
The enzyme is nearly homogeneous, yielding a single major band on a nondenaturing gel which is coincident with enzymatic activity (Fig. 2, A and B). The sodium dodecyl sulfate-gel electrophoresis of Fraction VII enzyme showed that the intact polymerase was about 60% of the total protein. The only polypeptide seen on sodium dodecyl sulfate gels of material eluted from the nondenaturing gel has a molecular weight of 166,000. The lack of minor bands on the gel scan shown in Fig. 2C along with the good agreement between the native and denatured molecular weights indicate that the enzyme consists of a single, giant polypeptide chain. Although the mobility of the protein standards on the denaturing gel was linear with the logarithm of molecular weight (Fig. 2D), the calculated value of 166,000 for polymerase III must be viewed in light of the small number of precise standards in this high molecular weight region. Assuming that Fraction VII is 60% pure and that there are 2 x 10" bacteria/g of wet, packed cells, then one can calculate from the purification data that there are about 100 polymerase III molecules/cell during logarithmic growth.
The marked effect of phosphate concentration on the calcu-  (Fig. 3) suggests that the shape of the enzyme may be sensitive to ionic strength. The data at higher ionic strength led to a calculated native molecular weight closer to the value obtained from sodium dodecyl sulfate gels, but the difference between the two calculated values was not large since at 0.2 M phosphate the lower slO+ was partially compensated by an increased K,. The determinations were done at two ionic strengths to help control for any possible aggregation phenomena, but it was not a significant factor. Using the higher salt data, the calculated relative frictional coefficient, f/f,,, is 1.54. This high value is consistent with, but not proof of, a markedly asymmetric molecule (52). Assuming that the enzyme is a prolate ellipsoid with a "compromise" degree of solvation of 0.2 (52), then its axes would be 520 x 52 A, or large enough to extend over 150 base pairs of the DNA helix.
The intrinsic nature of the exonuclease activity in B. subtilis DNA polymerase III is well established.
The exonuclease activity co-purified with polymerase activity on DNA-cellulose ( Fig. 1) and hydroxylapatite (ll), and the polymerase to nuclease ratio remained constant upon sedimentation of Fraction VII enzyme (data not shown). Both activities were similarly inhibited by sulfhydryl reagents, high ionic strength, and the specific polymerase III antagonist, OHPh(NH) ,Ura.. The exonuclease activity was co-excluded with polymerase activity in the agarose gel monitored, OHPh(NH),Ura-promoted ternary complex (12). At elevated temperatures, the polymerase and exonuclease activities decayed at the same rate (13). Enzyme purified from a strain which is temperature sensitive because of a mutation in the structural gene for DNA polymerase III showed the same increased thermolability for both activities (13).
The distinction between processive and distributive behavior is an important property of enzymes which synthesize or degrade macromolecules.
The processiveness of exonucleases has been generally determined by comparing the relative rates of removal of the terminal and internal nucleotides; degradative intermediates were not followed directly. Hence, some reported nonprocessive exonucleases may actually remove a few hundred nucleotides processively with each enzyme-substrate encounter. Within these limits, polynucleotide phosphorylase (53), X exonuclease (54), SP3 DNase (55), and Escherichia coli RNase II (56) act processively while Escherichia coli exonuclease I (57), T4 DNA polymerase's 3' ---t 5' exonuclease (58), and snake venom phosphodiesterase (55) degrade nonprocessively. Among DNA polymerases, the synthesis catalyzed by calf thymus polymerases (Y and p and E. coli polymerases I and II appear to be strictly distributive (59)(60)(61), while the T4 enzyme is only partially so (62). Processiveness of an exonuclease has the advantage of preventing the accumulation of degradative intermediates which could interfere with other cellular processes. However, in order to avoid excessive degradation, once the desired structure has been hydrolyzed, a processive nuclease requires a stop and release signal which may be the presence of a neighboring chain, a specific base sequence, or a protein factor. In addition, a processive enzyme must be able to translocate along the DNA chain, which is a very different enzymological event than binding and catalysis. Translocation of the messenger RNA chain on the ribosome, for example, requires specific protein factors and energy in the form of GTP (63). Specific proteins involved in DNA synthesis have been shown recently to make the action of T4 DNA polymerase and E. coli DNA polymerase II much more processive (61,62).
The B. subtilis polymerase III exonucleolytic degradation of pG-C-T-T-C-C-C-G-A and d(T) 10 demonstrates a strictly nonprocessive mode of degradation in uitro.
It appears that only one or at most two nucleotides are removed per encounter of enzyme with oligonucleotide.
While the experiments were less definitive, the degradation of polynucleotides and the synthesis of DNA also seem distributive.
This nonprocessiveness implies the dissociation rate of the enzyme substrate complex is comparable to k,,, and that translocation along the DNA chain is difficult.
It is sometimes stated (e.g., Ref. 64) that all prokaryotic DNA polymerases have an intrinsic exonuclease that may function to edit mistakes in replication (41), in contrast with eukaryotic DNA polymerases where intrinsic nuclease activity has not been demonstrated (65). Certainly the nuclease activity of several prokaryotic polymerases, including B. subtilis DNA polymerase III, have the expected properties of an editor (41, 66). However we saw no nuclease activity in B. subtilis polymerase II at a detection level over 3 orders of magnitude less than the activity of B. sub&is polymerase III and there is no evidence for such an activity in the published reports on B. subtilis polymerase I (1, 2). The suggestion that the intrinsic nuclease in a B. subtilis polymerase may have been lost during purification (41) is germane. The 5' + 3' exonuclease of E. coli polymerase I can be cleaved off by proteolytic enzymes during purification or by a B. subtilis protease (41), and an improved purification of the Micrococcus luteus enzyme revealed substantially more 5' --t 3' exonuclease than found previously (67). However, the B. subtilis cells used in this report were harvested in exponential phase and washed twice with a high salt buffer to minimize extracellular protease contamination; no enzymatic, chemical, or genetic manipulation of the intensively studied E. coli polymerase I and T4 DNA polymerase has produced a polymerase devoid of 3'-5' exonuclease activity (41, 68, 69); and an intrinsic exonuclease is not needed for faithful synthesis as shown by the results with some eukaryotic polymerases (65). Perhaps in I?. subtilis an editor need only be wedded to the principal replicative polymerase, polymerase III. With the caution required in interpreting negative results, it seems that B. subtilis polymerase II is like the known eukaryotic DNA polymerases in failing to have an intrinsic nuclease activity.
The kinetic analysis indicates that the chain length of the exonuclease substrate influences the rate of the reaction by two opposmg mechanisms. One determinant is the increase in the number of interior nonproductive binding sites per 3'-hydroxyl terminus as the chain length increases. This mechanism has been invoked previously (43, 44), but this report provides quantitative proof of this possibility. First, nonproductive binding to interior nucleotides was shown clearly by the potent inhibition of the I?. subtilis polymerase III-associated exonuclease by single-stranded circular DNA (Fig. 7). Second, the strict inverse relationship over 3 orders of magnitude between K$'p and chain length (Table III) is also a reflection of this intrinsic competitive inhibition. The ratio KJa calculated both from the variation in Km 'pp with n and from the inhibition by circular DNA are in excellent agreement and are, respectively, 8 and 10 pM for the natural DNA substrates and 10 and 9 PM for the poly(dT) substrates. The data do not permit a precise evaluation of the true K, but indicate that it is of the same order of magnitude as KJa. This suggests that the binding of DNA polymerase is about as strong to interior nucleotides as to 3'-terminal nucleotides and therefore the exonucleolytic specificity of the enzyme is dictated at some step subsequent to binding. This avid nonproductive binding must also contribute both to the inhibition of E. coli and B. subtilis polymerase III synthetic activity by single-stranded DNA and to their lowered activity with substrates containing large gaps (41).
The kinetic data also show clearly that intrinsic competitive inhibition is not the sole effect of chain length. Vii! did not vary with n according to the predicted hyperbolic function and V~~$/Kp' was not independent of n. Another mechanism contributing to the chain length effect could be an excluded volume effect whereby the terminus becomes sterically buried as n increases (44). Alternatively, the probability of adventitious folding back of the 3' terminus should increase with n and such a hairpin structure would not only provide a very poor substrate for a single strand specific exonuclease but also may bind the enzyme tightly since it mimics the DNA conformation necessary for synthesis. The poly(dT) substrates were fabricated specifically to test this latter mechanism since they are devoid of secondary structure. While V,,P.F appeared independent of n for these homopolymers, Kp' did decrease with n and thus hairpin structures cannot be the sole effect of increasing n. Moreover, neither the secondary structure nor the excluded volume mechanism explains the data even in conjunction with intrinsic competitive inhibition, since Vzi$ and V,%f/Kkpp were larger than predicted by the rate equation for competitive inhibition, an unexpected result.
The simplest resolution is that the accessibility of the 3' terminus to enzymatic attack increases with n. This increased exposure is, in fact, implied by the increase of V::f'/:plK;pp with n, since the rate of an enzymatic reaction is proportional to this ratio at low substrate concentrations and thus this ratio is a measure of the intrinsic reactivity of the substrate. A possible physical interpretation is suggested by the observed increase of Vapp/Kapp with approximately the % power of n. If the submar m strate is a random coil, then its volume in solution is proportional to n8'* (52) and the local nucleotide density to n-l'*. Thus the structure of the polymer becomes more open as n increases and perhaps makes the terminus more accessible to the enzyme. This may also contribute to the higher V;if: for poly(dT) than for DNA as the lack of intramolecular hydrogen bonding may permit an even more open structure. It may not be fortuitous that the extrapolated value of V,$:: for n = 1 for natural DNA is about equal to the observed V:[f for poly(dT). While further experiments are needed to substantiate these speculations, the main conclusions are that the chain length markedly influences the rate of the reaction by the opposing mechanisms of intrinsic competitive inhibition and exposure of the terminus to attack. These conclusions about how an enzyme, in effect, measures the size of its giant substrate may extend to other enzymes with macromolecular substrates.
Besides the effect of chain length, the rate of exonuclease activity also depends upon secondary structure and possibly base composition. The exonuclease had very little activity on duplex DNA, such as native E. coli DNA, poly(dA):(dT), and nicked DNA. The minimal activity seen may be real since the analogous T4 DNA polymerase can catalyze the turnover of termini (70) or it may result from single-stranded regions in the DNA preparations.
The low activity on poly(dC) may reflect an unusual secondary structure of the homopolymer (46), since the cytidylic acid triplet in pG-C-T-T-C-C-C-G-A was hydrolyzed essentially as well as the other nucleotides. The best substrate, poly(dT), is always single-stranded.
Although the gram-negative E. coli and the gram-positive B. subtilis are evolutionarily divergent (71,72), a number of similarities appear among their DNA-related enzymes. Each bacterium has three distinct polymerases (41, 66). Previously, we pointed out that DNA polymerase III from the two sources are quite similar in function, template specificity, reaction requirements, and sensitivity to sulfhydryl reagents, singlestranded DNA, temperature, high ionic strength, and l-p-~arabinofuranosylcytosine triphosphate (2). The results in this study permit a more extensive comparison. Both enzymes purify similarly on DEAE-cellulose, phosphocellulose, DNAcellulose, and hydroxylapatite (14, 73). The native molecular weights are essentially the same and both enzymes synthesize in the 5' + 3' direction (73). The two enzymes catalyze the additional reactions of pyrophosphorolysis of DNA, the exchange of PP, with deoxyribonucleoside triphosphates, and the hydrolysis of single-stranded DNA from the 3' terminus releasing 5'-mononucleotides until the limit dinucleotide is reached (11,45,73).
Despite these similarities, B. subtilis DNA polymerase III is significantly different from the E. coli enzyme. First, only the B. subtilis enzyme is inhibited by arylhydrazinopyrimidines. Second, while the B. subtilis enzyme clearly consists of a single polypeptide chain of 166,000 daltons, the E. coli enzyme probably has two subunits of 140,000 and 40,000 daltons (73). Because the stoichiometry of these subunits is not certain (73), the physical structure of B. subtilis polymerase III is the best known for a host replicative polymerase. Third, while the purified B. subtilis polymerase III clearly has only a 3' + 5' exonuclease, the E. coli enzyme has both 3' + 5' and 5' + 3' exonucleolytic activities (45). Perhaps the E. coli 5' + 3' exonuclease resides in the 40,000 subunit. The exonucleolytic activity of the single subunit T4 DNA polymerase has virtually the same properties as the B. subtilis enzyme (41). Fourth, the apparent K, values for the deoxyribonucleoside triphosphates are about 1 order of magnitude lower for the B. subtilis polymerase III than for the E. coli enzyme.
The preferred substrate for B. subtilis and E. coli polymerase III is DNA containing small gaps (41). Replication of long single-stranded templates by the E. coli enzyme requires small molecule and protein cofactors (74, 75) whose precise nature is controversial (reviewed in Refs. 41 and 66). To determine the effect of these cofactors on B. subtilis polymerase III, we exchanged samples of purified B. subtilis polymerase III with Dr. J. Hurwitz for purified E. coli polymerase III and elongation Factors I and II, and the synthesis directed by an RNA-primed $X174 single-stranded DNA template was measured in both laboratories. The results agree, and some of the more extensive data of Vicuna and Hurwitz are shown in Table  VI. Several conclusions can be drawn. First, B. subtilis polymerase III, like the E. coli enzyme, can elongate an RNA primer, the possible in uiuo substrate (77). Second, the activity of the purified B. subtilis polymerase III is stimulated severalfold by Factors I and II to about one-half the synthetic rate with E. coli polymerase III. Both Factors I and II are required and the activity cannot be due to contaminating E. coli polymerase since the activity is sensitive to OHPh(NH)PUra. Third, this factor-dependent synthesis, remarkable considering the wide evolutionary divergence of these bacteria,