Chemical structure of a modification of the Escherichia coli ribonucleic acid polymerase alpha polypeptides induced by bacteriophage T4 infection.

Abstract The α-polypeptides of Escherichia coli RNA polymerase are known to be chemically modified within 4 min after infection by bacteriophage T4. This paper reports a crude system derived from T4-infected E. coli which will modify the α-polypeptides of purified E. coli RNA polymerase in vitro. The product of this in vitro reaction is identical with the modified α found in vivo; it differs chemically from the "altered" α observed after T4 infection in the presence of chloramphenicol. The in vitro reaction is rapid, being 50% complete within 30 s at 37°; the enzyme activity responsible appears less than 2 min after infection at 30°. These data on kinetics of synthesis and activity are consistent with data on kinetics of α modification in vivo. Specifically labeled radioactive substrates have been used in this in vitro modification reaction to investigate the chemical substitution introduced during α modification. The chemical stability of the modification and the sequence of pronase peptides containing the modified region of α also have been examined to complement the information obtained with the in vitro system. Modification involves covalent attachment of 1 adenine nucleotide, apparently adenosine diphosphoribose, to a specific arginine in the α-polypeptide at the sequence Thr-Val-Arg. NAD+ serves as the donor of this nucleotide in the in vitro reaction. The α modification is hypothesized to involve adenosine diphosphoribose linked through its terminal ribose to a guanido nitrogen of arginine.

in vitro.
The product of this in vitro reaction is identical with the modified (Y found in viva ; it differs chemically from the "altered" a! observed after Tq infection in the presence of chloramphenicol. The in vitro reaction is rapid, being 50% complete within 30 s at 37"; the enzyme activity responsible appears less than 2 min after infection at 30". These data on kinetics of synthesis and activity are consistent with data on kinetics of cr modification in vivo. Specifically labeled radioactive substrates have been used in this in vitro modification reaction to investigate the chemical substitution introduced during a! modification. The chemical stability of the modification and the sequence of pronase peptides containing the modified region of (Y also have been examined to complement the information obtained with the in Vitro system. Modification involves covalent attachment of 1 adenine nucleotide, apparently adenosine diphosphoribose, to a specific arginine in the ol-polypeptide at the sequence Thr-Val-Arg. NAD+ serves as the donor of this nucleotide in the in vitro reaction. The a modification is hypothesized to involve adenosine diphosphoribose linked through its terminal ribose to a guanido nitrogen of arginine. RNA (12,13).
The first of these changes in RNA polymerasc to be discovered is in the o( subunits. Within 4 min after Tq infection the two 40,000-dalton a-polypeptides become more negatively charged, as shown by their altered electrophoretic mobility in both alkaline and acid polyacrylamide gels containing urea (14)(15)(16)(17).
In this process the pre-existing cu-polypeptides are modified rather than replaced (17 hours another 40 rg of pronase are added and digestion is continued for 6 to 8 hours more. After digestion, the peptide solution is lyophilized, redissolved in water, and relyophilized to remove (NHr)HCOa. The peptides are redissolved in-2 ml of water and passed slowly over a l-ml column of Dowex 50-X2 (H+ form). azP covalentlv linked to peptides is retained by the column; free phosphate andVnucleotides flow through.
Residual urea is washed through the column with water, and the peptides then are eluted from the column with icecold 1.0 N NHdOH.
Fractions containing radioactivity are pooled and quickly lyophilized.
About 80% of the 32P label originally in (Y is recovered in peptides. The lyophilized pronase peptides are dissolved in 100 ~1 of water and applied to a sheet of Whatman No. 3MM paper (46 X 57 cm) in a band (1 X 3 cm) near one corner. The paper is wetted with pH 1.8 buffer (7% formic acid) and subjected to electrophoresis at 1800 volts for 2 hours in a refrigerated tank. The phosphate-containing peptides are very slow moving at this pH. After drying at room temperature, the sample is concentrated to a narrow band near the edge of the paper sheet by wetting with electrophoresis buffer, dried again, and chromatographed in the second dimension, perpendicular to electrophoresis. Ascending chromatography requires 20 hours at 25" in a closed tank, using (by volume)-7.6yo acetic acid, 37.8yo I-butanol, 24.4% pyridine, and 30.2% distilled water. The modified neDtides are once aaain very slowmoving (see Fig. 4), presumably hue to their high charge density.
Following chromatography the paper is dried in a hood. Phosphate-containing peptides are located by autoradiography on Kodak No-Screen x-ray film; the modified peptides can also be located by the ultraviolet absorption of the adenine which they contain, if at least lo-* moles are present. Complete purification requires a third separation; a rectangle (3 X 5 cm) containing the 3aP-labeled peptides is cut from the paper, stitched into a new sheet, and subjected to electrophoresis at pH 3.5 (in 5.0% acetic acid, 0.5% pyridine).
After this electrophoresis (2500 volts, 5 hours) the labeled peptides again are located by autoradiography. Radioactive spots are cut from the sheet and soaked in 2 ml of 0.2 N acetic acid for 12 hours at 25' to elute the peptides. Amino Acid Analysis-Amino acid compositions of peptides are determined on a Beckman 116 analyzer with expanded recorder range. Aliquots of peptides (2 to 5 nmoles) are hydrolyzed at 105" for 20 to 24 hours in sealed evacuated tubes with 200 ~1 of 6 N HCl, dried down, and applied to the analyzer columns.
Assignment of acids and amides is made by dansylation following digestion of peptides with leucine aminopeptidase.
Peptides (0.2 to 0.5 nmole) are dried down in tubes (6 X 50 mm), taken up in 5 to 10 ~1 of 50 mM NaHCOa-10 mM MgC12 containing 1 mg per ml of leucine aminopeptidase (Worthington, code LAP%), and digested, 12 hours at 37". One-half volume of 0.5% dansvl chloride in acetone is added to the digest; after incubation "for 30 min at 37", the dansyl-amino acids are identified as usual on polyamide sheets (28). Leucine aminopeptidase does not interfere with the chromatography and does not generate significant background by self-digestion.
Analytical Polyacrylamide Gel Electroph.oresis-SDS gel electrophoresis (24) for estimation of protein purity is carried out on 6.5yo polyacrylamide gels. Estimates of the degree of a modification are made using pH 4.4 polyacrylamide gels containing urea, a modification of the DH 4.5 gel svstem described bv Reisfeld et al. (32). Six per cent &rylamyde gels are made up "with 6 M urea, containing 5 PM FMN instead of (NHd)&Os, and are photopolymerized with a fluorescent desk lamp for 30 min. The large pore stacking gel is omitted; samples are applied in 100 ~1 of 6 M urea-3yo p-mercaptoethanol-O.OOO~~o basic fuchsin (tracking dye); and the electronhoresis buffer. adjusted to DH 4.4. is diluted 4-fold. Electrophoresis is at room temperature and 2.5 ma per gel. All gels are stained with Coomassie brilliant blue R-250 as described by Burgess (25), then destained by gentle shaking with a small quantity of AG-501 mixed bed resin (Bio-Rad Laboratories) in 7.5% acetic acidd% methanol for 12 hours at 37'. Semiquantitative estimates of the degree of (Y modification are made by scanning stained pH 4.4 gels at 600 nm in a Gilford scanning spectrophotometer and comparing the area under modified and unmodified or-peaks.

Design of in Vitro cx Modification Assay-The in vitro modification assay described under "Experimental
Procedures" was designed on the basis of the known timing of cx modification in vivo (16) and the proven presence of an adenine nucleotide in the modified polypeptide (17). Thus 3-to 4-min infected cells were chosen as a source of the mcdifying activity, and NAD+ and ATP were both tested as potential nucleotide donors for the reaction.
In the absence of data on the properties of an a-modifying activity, essentially unfractionated cell lysates were used. Antibody precipitation and polyacrylamide gel electrophoresis were chosen as rapid, simple methods of removing unincorporated label and assuring the specificity of the assay. It was obvious that a very sensitive assay such as this would be needed since both ATP and NAD+ are present at over 1 mM concentration in Escherichia coli (33,34) and the specific activity of the added radioactive material would be greatly diluted in a crude extract.
Incubation of E. coli RNA polymerase with a concentrated extract of infected cells and adenosine-labeled NAD+ as described results in incorporation of about O.O5a/, of the input label into material undergoing electrophoresis in the position of modified cr on the SDS polyacrylamide gel (Fig. 1). There is no incorporation of ATP label into LY under these conditions (data not shown).
Characterization oj System-This assay measures incorporation of label into the cu-polypeptide and not a contaminant.
The labeled product sediments with RNA polymerase in velocity gradients (data not shown), is specifically precipitated by anti-RNA polymerase antibody, and undergoes electrophoresis with modified Q! in SDS gels. Furthermore incorporation is largely dependent on added pure RNA polymerase (see discussion of Fig. 3, below).
Slice No. was added to the gel loading buffer before' heating, and the gei was counted for both 3% and aH. Parallel stained gels predicted normal (Y to be in Slices 39 and 40, and in viva-modified LY to be in Slices 38, 39, and 40. This slight mobility difference (17)  Isotope dilution of the input label with nonradioactive NADf indicates that as expected the in vitro reaction contains approximately 0.4 mM endogenous NAD+ (data not shown). Attempts to reduce endogenous NAD+ and thereby increase the specific activity of label, either by dilution of the crude extract or by adsorption of NAD+ on Norit, have thus far caused loss of the modifying activity.
Despite the low level of incorporation and the extensive workup involved, the assay is reproducible to better than &lo% with any one extract. Different extracts from 3-to 4-min infected cells may vary up to 3-fold in activity; the most active extracts stimulate incorporation of about 0.5 mole of adenosine per mole of input a-polypeptide. The presence of initiation subunit u in the added RNA polymerase has no effect on incorporation of label into a (data not shown).
The in vitro a-modifying activity is unstable in a crude extract, with a half-life of about 8 min at 37" (data not shown). The in vitro reaction is rapid, being 50% complete within 30 s at 37", and the product is stable indefinitely in the crude extract, long after the modifying activity has disappeared (Fig. 2). The reaction is apparently irreversible under these conditions, since chasing the label with 2.5 mM NAD+ at 5 min (after the reaction is complete but before the activity has disappeared) causes no decrease in final incorporation after 15 min (373 cpm incorporated in normal reaction; 382 cpm with chase at 5 min).
The activity being assayed in this in vitro reaction is clearly phage-induced. Activity is found only in infected cells and appears rapidly after infection with kinetics expected from data on CY modification in v&o. This was shown by assaying extracts made from cells harvested at various times after infection (Fig.  3, fi1Zed circles). If these same extracts are assayed without added RNA polymerase, some incorporation is observed very early after infection (Fig. 3, open circles). I assume that this incorporation results from labeling of RNA polymerase already present in the extract (estimated to be about 20 pg per reaction). The a-polypeptides, once modified in z&o, apparently cannot accept label in vitro; if modified RNA polymerase purified from T1-infected cells is added to an in vitro modification reaction instead of the normal enzyme, essentially no labeling of its crpolypeptides occurs (data not shown). This presumably explains why the endogenous RNA polymerase can accept label at 2 min after infection, but not by 4 min; a! modification is known to be underway in vivo at 2 min and complete by 4 min (16). In vitro activity towards added RNA polymerase alone (the difference curve of Fig. 3, squares) is only slightly diminished between 2 and 4 min, but disappears by 12 min. Thus the enzyme(sj involved evidently remain active for a short time after RNA polymerase has been completely modified in vivo.
In Vitro and in Vivo Modifications Are Identical-Since in vivo modification apparently prevents in vitro labeling, as discussed above, the in vivo and in vitro reactions could involve the same site in a. If we can show that the product of the in vitro reaction is chemically identical to the modification produced in vivo, the in vitro system will provide a convenient tool for studying the structure of the LY modification.
I have compared the radioactive peptides produced by pronase digestion of in vivo-modified and in oitro-labeled a-polypeptides to prove their identity. Zero-time points were obtained by terminating the reaction as soon as possible after mixing all components at 0". The reactions were then analyzed as usual. This figure indicates the amount of label undergoing electrophoresis with 01. FIG. 3 (right). Kinetics of appearance of in vitro a-modifying activity after infection.
A culture of Escherichia coli B/r was grown at 30" and infected as described for preparation of the standard in vitro a: modification reaction, except that aliquots were harvested at various times after Td infection.
The cells were kept at 0" during centrifugation and washing to prevent continued development.
Extracts were prepared from each aliquot (Method A) and two reactions were carried out with each, one as usual containing added RNA polymerase, and the other containing only storage buffer without RNA polymerase.
After 37" incubation and termination of the reactions, RNA polymerase (25 rg) was added to the tubes without the enzyme. Antibody precipitation and subsequent analysis of all reactions were carried out as usual. This figure indicates the amount of label undergoing electrophoresis with a: in each case. l , normal reactions, with added RNA polymerase; 0, reactions without added RNA polymerase; q , difference in incorvoration with and without added RNA nolvmerase. This last curve presumably represents label incorporated into added polymerase alone, without the contribution from endogenous RNA polymerase. ered. Pronase peptides of the in vivo-and in vitro-labeled polypeptides were prepared and fingerprinted in parallel by pH 1.8 electrophoresis and chromatography as described under "Experimental Procedures." Autoradiography revealed that over 900/b of a2P moved as a single spot in the same position on both fingerprints (Fig. 4, A and B). In both cases that spot was resolved into two spots by pH 3.5 electrophoresis, and again the mobilities of in viva-and in vitro-labeled peptides were identical (Table I).
The same result was also obtained by digesting a mixture of 32P-labeled in vivo-modified a! and [adenosine-G-aH]NAD+ in vitro-labeled cr; a single radioactive spot containing both a2P and aH labels was found on the two-dimensional fingerprint. Since the radioactive peptides from in viva-and in vitro-modified a behave identically in chromatography and electrophoresis, the substitutions must be the same.
It should be noted here that the product of the in vitro and in vivo modification reactions differs from the partially "altered" a-polypeptides found in wivo after Td infection when phage protein synthesis is prevented by chloramphenicol (18). This "alteration," like modification, involves covalent addition of phosphorus and adenosine to LY, but different "P-labeled peptides were observed in tryptic digests of in vivo-labeled "altered" and modified CY (18). Reproducibly different patterns of radioactive pronase peptides are also obtained from "altered" and modified a (Fig. 4C) "alteration" and modification of (Y is still unclear, but the in vitro reaction is unambiguously modification. Therefore specifically labeled NAD+ can be used in the in vitro system to determine what port.ion of the NADf molecule is transferred to cr during modification.

Both Phosphates of NAD+ Are Covalently
Attached to Modijied a-To distinguish whether one or both phosphates of NAD+ are donated to Q! during modification, the following experiment was performed.
NAD+ doubly labeled with 3H in adenosine and 32P in both phosphates was used to modify (Y in a standard in vitro reaction.
The reaction was analyzed as usual and the SDS gel was counted for both isotopes.
An aliquot of the same doubly labeled NAD+ was incubated with diphtheria toxin and rabbit peptidyl-tRNA translocation factor EF-2. Diphtheria toxin catalyses covalent transfer of adenosine diphosphoribose from NAD+ to the EF-2 polypeptide (35) so this system provides a Because input label and counting conditions were identical, the ratio of counts of 32P to counts of 3H in cr and in EF-2 can be compared directly to determine the number of phosphorus atoms per adenosine present in modified CL The 32P:aH ratios in (Y and EF-2 are essentially identical (Table II), so both phosphates from NAD+ must be incorporated into modified 0~. This result is consistent with a previous report on the phosphate content of in vivo-modified o( (18). These reactions were analyzed as usual and counted for both isotopes.
As is clear from phosphates was used to modify (Y in a standard in vitro reaction. An aliquot of the same double-labeled NAD+ (about 5 nCi) was incubated 30 min at 37" with 10 rg of partially purified rabbit EF-2 and 1 rg of activated diphtheria toxin (both kind gifts of Dr. D. M. Gill, Harvard University) in 50 ~1 of 20 mM Tris-Cl, pH 8.0-5 mM dithiothreitol. Unincorporated label was removed by the addition of 5 mg of acid-washed Norit in 50 ~1 of 10 mM sodium phosphate, pH 7.2. After 15 min at 25" the Norit was centrifuged out. The supernatant, was made 2% in SDS, 3% in p-mercaptoethanol, and 10% in glycerol, heated 2 min at 95", and subjected to electrophoresis on a 6.5$& SDS polyacrylamide gel in parallel with the in vitro-modified a. Both gels were sliced and counted for a2P and 3H. The peak of counts in the position of EF-2 and (Y on the respective gels were summed and are presented here for two experiments (in which different batches of [3*P, 3H] NAD+ were used, with different ratios of input azP and 3H).  5. Ultraviolet spectra of modified and normal cy-polypeptides.
Purified Ta-modified and normal or-polypeptides prepared as described under "Experimental Procedures," were dissolved in 0.2% SDS-75 m&r (NHI)HCO~, pH 8, at approximately 1 mg per ml. These solutions were passed over a column of Sephadex G-25 (lo-ml bed volume per ml of solution) in 0.5% SDS, 75 mM (NH4)HC03, pH 8, to remove any ultraviolet-absorbing small molecules eluted from the preparative SDS gels and to assure that normal and modified (Y were in identical buffers. The concentrations of normal and modified 01 were adjusted to give equal absorbance at 290 nm, where the adenine nucleotide of the modification should contribute no optical density. The solutions were scanned for absorbance in a Cary 15 spectrophotometer, first against buffer, and then against each other to obtain a difference spectrum.
A, upper spectrum, modified cx; lower spectrum, normal a. B, difference spectrum, modified Q absorbance minus normal (Y absorbance.
but not nicotinamide, appears to be present in in vitro-modi$ed a Two in vitro CY modification reactions were prepared (extracts made by Method B) containing NAD+ labeled as indicated. They were analyzed as usual and counted for both 3H and W. Efficiency of incorporation of adenosine label differed in Reactions A and B because different extracts were used. adenosine diphosphoribose is linked to a! during modification. However Table III does not prove that adenosine diphosphoribose is incorporated intact in modified (Y, as the NMN ribose group could be rearranged, or an unlabeled portion of the nicotinamide ring could be retained.
Modi$ed (Y Contains One Adenine Nucleotide per Polypeptide-The in vitro system does not label input a! polypeptide quantitatively, so it cannot be used to determine whether one or more nucleotides are added to each polypeptide. Instead, 1 have purified several milligrams of the RNA polymerase Q polypeptides from normal and Th-infected E. coli B/r and compared their ultraviolet spectra to quantitate the amount of adenosine per molecule. Both normal and modified preparations were greater 6187 320 than 98"1, pure cr-polypeptide, judged by SDS gel electrophoresis, but the preparation of modified CY contained between 5 and 10% unmodified polypeptide, judged by pH 4.4 gel electrophoresis. The spectra, normalized at 290 nm (where adenine nucleotides exhibit essentially no absorbance) are shown in Fig. 5A. The difference spectrum (Fig. 5B) is unambiguously that of an adenine nucleotide (Table lV), indicating that no other purines or pyrimidines are added during modification.
The approximate concentration of adenosine present in the solution of modified (Y can be calculated from the difference spectrum and the known e2s9 of adenosine.
To calculate the concentration of cy from its spectrum in Fig. 5A, and thus determine the molar ratio of adenosine to cy, we must first obtain a molar extinction value for the unmodified polypeptide.
The amino acid composition of normal a: was determined by hydrolysis with p-toluenesulfonic acid, which allows direct quantitation of tryptophan on the amino acid analyzer (36). These data indicate 1 tryptophan and 6 tyrosines per 40,000 daltons2 and thus an ezs~ of about 12 x lo3 for unmodified (Y. Finally, we can calculate * Burgess, using spectral methods, has determined 1 tryptophan and 5 tyrosines per 40,000 daltons for the cu-polypeptide of Escherichia coli K12 RNA polymerase (37). The discrepancy may reflect either strain differences or errors in one determination. Modified and normal a-polypeptides were dissolved in the same buffer and scanned as described in the legend to Fig. 5. The data presented here for a were determined from the spectrum in Fig. 58.  Table V). As this preparation contained a small amount of unmodified 01, the ratio in completely modified (Y would be still closer to 1: 1. LY modification thus involves addition of only 1 nucleotide.

Modification
Is Linked to Arginine in cY-Polypeptide-As already mentioned, pronase digestion of Y&labeled modified (II produces two labeled peptides separable by pH 3.5 electrophoresis, Pro 1 arid Pro 2 (Table I). These modified peptides were purified from several milligrams of modified Q (see "Experimental Procedures") in order to determine the amino acid involved in the linkage.
Pro 1 was recovered from paper in 16% molar yield, and Pro 2 in 7% yield, relative to the amount of cr originally digested.
The amino acid composition of Pro 1 was determined by acid hydrolysis to be: Thr, 0.9; Gly, 0.7; Val, 1.0; and Arg, 1.0. Pro 2 was found to contain: Gly, 0.6; Val, 1.0; and Arg, 1.0. No other amino acids were present at over 0.1 residue.
Dansyl-Edman sequencing gave the following results: Pro 1, Thr-Val-Arg; Pro 2, Val-Arg. Glycine is never obtained as an NHz-terminal amino acid during Edman degradation of either Pro 1 or Pro 2. The presence of glycine in low molar ratio in acid hydrolysates of these modified peptides is an artifact produced during hydrolysis by degradation of the adenine in the modification.
Two lines of evidence support this assertion.
First, acid hydrolysis of authentic adenine under the conditions used for composition analysis produces glycine in 40 to BOY0 molar yield.
Secondly, complete digestion of Pro 1 or Pro 2 with leucine aminopeptidase reveals no glycirle.3 Peptide Pro 2 has only one reactive amino acid available for covalent addition of a rmcleotide, namely arginine, because the NH2 terminus of Pro 2 is free to react with dansyl chloride.
l'resumably Pro 1, because of its similarity to Pro 2 in chromatography and electrophoresis, contains the same linkage and is an incomplete digestion product of the same sequence in cy. Arginine is in fact the amino acid involved in modification, as proven by further digestion of the pronase pcptides from 32P-labeled modified a with leucine aminopeptidase.
The single labeled peptide observed in this digest, LAP I (Table I), was purified in low yield on paper.
It contains only arginine and glycine after acid hydrolysis and reveals NH*-terminal argiuine upon dansylation followed by acid hydrolysis.
Therefore LAP 1 must be modified arginiue.
(Normal arginine rather than the modified derivative is found iu amino acid analyses and Edman sequences because the modification is lost during acid hydrolysis.)

Chemical
Stability of Linkage to Arginine; Possible Structures- The results already presented suggest that modified (Y contains a covalent bond between some part of adenosine diphosphoribose (or a related hl>l'-sugar compound) and an arginine in the polypeptide.
The bond could reasonably involve either the guanido side chain or the carboxyl group4 of arginine, and either the phosphates or the terminal sugar residue of the nucleotide.
(Release of 5'-AMP from the modified peptides by venom phosphodiesterase (17) rules out a bond directly to the adenosiue moiety.) In order to investigate further where and how the nucleotide is linked, I have examined the chemical stability of 32P in modified o(.
For these experiments modified a doubly labeled with a2P and 3H nas Iurified as described under "Experimental Procedures." The purified protein was incubated under the desired conditions and aliquots were withdrawn at intervals, precipitated with cold 5'3& trichloroacetic acid, collected on Millipore filters, and counted. 321' counts are lost as phosphorus in the nucleotide is released from a; 3H counts, in the protein it,self, are stable and serve to measure the amount of o( precipitated.
In all cases a first order rate of loss has been observed for 92 to 95yo of input s21'. The remaining 33' is stable to all conditions tested and is probably in fragments of DNA undergoing co-electrophoresis with (Y on the preparative SDS gel. A plot of hydrolysis kinetics in acid and base at 66" is shown in Fig. 6; the data obtained from these and similar plots is summarized in Table VI. It is difficult to deduce a specific chemical structure from these data, because neighboring amino acid side chains could alter hydrolysis rates of the protein-rmcleotide bond by acid or base catalysis or by steric hinderance.
Nevertheless some general con-3 Dansylation of the leucine aminopeptidase digest reveals dansyl-Thr, dansyl-Vial, some dansyl-Arg, and a chromatographitally unknown compound from Pro 1, and dansyl-Vial, dansyl-Arg, and the same unknown from Pro 2. Acid hydrolysis of the dansylated digests appears to convert the unknown compound to dansyl-Arg. Therefore I assume that the unknown is dansyl-modified arginine.
The modification is evidently partially degraded during leucine aminopeptidase digestion, since some normal arginine is present in the digests.
4 Attachment through the carboxyl group is possible only if the arginine involved is COOH-terminal in a. This cannot be ruled out, as carboxypeptidase digestion of normal (Y has failed to release any amino acids (unpublished observations). A culture of Escherichia coli B/r w& labeled with laZPlnhosnhate and PHloroline as described under "Exnerimental @rocedur&" and in?e&d with Td phage. Doubly labkled modified LY polypeptide was prepared from the labeled cells, also as described under "Experimental Procedures." Lyophilized OLpolypeptide was dissolved in 500 pl of either 0.25 N HCl or 0.25 N NaOH and incubated at 66". At the indicated times aliquots were withdrawn to tubes containing 250 pl of 0.1 mg per ml of bovine serum albumin (to act as carrier), precipitated with 2.5 ml of cold 5oj0 trichloroacetic acid, and collected on Millipore HA filters. The filters were washed with 5% trichloroacetic acid and then 95% ethanol, then dried and counted.
Data were normalized to a constant number of aH counts per min to account for variation in the size of aliquots taken (aliquots taken late in the experiments were larger to allow more accurate SzP counting).
32P data are plotted on a logarithmic scale in this figure. l , incubation in 6.25 N NaOH. k small amount of aaP w& resistant to hydrolysis after 120.180. and 240 min of incubation.
This stable background (6.5% of'input 32P) was subtracted from all data before pl&ting. 0, incubation in 0.25 N HCl. In this experiment, a stable background of 8.0% was subtracted from all 32P data before plotting.
elusions can be drawn about the bond being broken.
First, the observed stability pf the modification to both acid and base is much too great for 'an anhydride linkage between the phosphate of a nucleotide and the carboxyl group of an amino acid (40). Secondly, a phosphoamide bond to a guanido nitrogen of arginine is unlikely.
Phosphorus in the modification is more stable to acid and to pH 4.8 hydroxylamine than to base. This is in marked contrast to the stability expected for a phosphoamide bond, such as the linkage between a lysine c-amino group and the phosphorus of 5'-AMP in the E. coli DNA ligase-adenylate covalent intermediate (40,41). Finally, the 32P is far too labile in acid to be linked i? a carbon-to-phosphorus bond such as that in 2-aminoethylphosphonic acid (42). Thus the chemical stability of 32P in the LY modification seems to rule out a bond directly between phosphorus and arginine. Because linkage through adenosine has already been ruled out, the attachment point for the protein-to-nucleotide bond is very likely within the terminal sugar of the nucleotide.
As mentioned, either the guanido or carboxyl group of arginine could be involved.
However, the modification (still judging from the azP stability data) is much more stable to 0.25 N NaOH than expected if a sugar were es- Data in acid and base were obtained as described in the legend to Fig. 6; half-times for 3zP loss were determined from semilogarithmic plots. Data in hydroxylamine and sodium acetate were obtained with 01 dissolved in 7 M urea (necessary to maintain solubility under these conditions). Modified or-polypeptide doubly labeled with 32P and 3H (prepared as described in Fig. 6) was dissolved in 7 M urea and freed of SDS as described under "Experimental Procedures." The solution of a in urea was lyophilized, then redissolved at 7 M urea and either 4 N hydroxylamine (adjusted to pH 4.8 with HCI) or 4 N sodium acetate, pH 4.8 (as a control).
These solutions were incubated and assayed as described in Fig. 6. Again, half-lives for aaP loss were determined from semilogarithmic plots. An ester linkage between the carboxyl group of arginine and a ribose l-carbon hydroxyl group would be about 3 orders of magnitude less stable under these conditions than is found for the arginine-to-nucleotide bond in cr; an ester linkage to a ribose 2-or 3-carbon hydroxyl, about 4 orders of magnitude less stable (43). Among possible alternative structures involving the sugar and the guanido group, one is especially attractive because it is consistent with the simplest hypothetical mechanism for the ar modification reaction, i.e. direct transfer of intact adenosine diphosphoribose to (Y. This structure would be generated from NAD+ and Q by a transfer of the adenosine diphosphoribose terminal-ribose l-carbon from the l-nitrogen of nicotinamide (in NADf) to a guanido nitrogen of arginine.

DISCUSSION
The Td-induced Q! modification reaction, here hypothesized to be a transfer of adenosine diphosphoribose from nicotinamide to a specific polypeptide site, seems to represent a novel utilization of NAD+ iq E. coli. The only analogous reaction yet described, ADP-ribosylation of mammalian elongation factor 2 by diphtheria toxin, is also hypothesized to involve a carbon-nitrogen bond, in this case between the adenosine diphosphoribose terminal ribose and the e-amino nitrogen of lysine (44). Interestingly, this reaction is also catalyzed by a phage-induced enzyme. Diphtheria toxin, excreted by Corynebacterium diphtheriae strains lysogenic for phage p, is known to be the product of a phage gene (45). However, diphtheria toxin and its target are extracellular and apparently not directly involved in the metabolism of the host bacterium.
The E. coli RNA polymerase a modification, in contrast, affects a very basic part of the cell's machinery for gene expression, although the physiological effects of the modification on Th and E. coli transcription are still uncertain. A poly-(adenosine diphosphoribose) polymerase present in the nuclei of eukaryotic cells can link the adenosine diphosphoribose moiety of NAD+ to histone proteins in the presence of DNA (46), but this system differs significantly from both the bacterial adenosine diphosphoribosyltransferases.
The eukaryotic enzyme exhibits very little specificity in its protein substrates, produces a relatively unstable adenosine diphosphoribose-to-protein bond (possibly an ester between ribose and a carboxyl group), and polymerizes additional adenosine diphosphoribose units onto the initial nucleotide residue (reviewed in Ref. 47).
Aside from physiological function, two important questions about (Y modification remain. First, does modification involve more than the addition of a nucleotide? One or two extra peptides, not found in digests of normal cy, have been reported in tryptic digests of modified cy (48). Moreover, modified CY appears larger than the normal ar-polypeptide, as judged by its decreased mobility in SDS gels (15-17).
However, the addition of a nucleotide alone may be sufficient to explain these results.
The efficiency of tryptic cleavage at the modified arginine may be reduced by the substitution, and the modified peptide will have altered properties because of the nucleotidc.
Addition of net negative charge to other polypeptides (by maleylation of lysines) has been shown to cause a significant decrease in SDS gel mobility, as much as au apparent molecular weight increase of 1000 per negative charge (49).
A second unanswered question, about the origin and the number of a-modifying enzymes, involves the "alteration" of the a-polypeptides occurring very rapidly after Tq infection even in chloramphenicol-treated cells. As already mentioned, "altered" (Y is chemically similar but not identical with modified (Y (Ref. 18 and Fig. 4).5 "Altered" (Y appears transiently prior to modified (Y during normal infection also (18)," suggesting that Tq may induce two sequential changes in CL It is possible that two distinct enzymes are involved, and that the "altering" enzyme (unaffected by protein synthesis inhibitors) pre-exists in either the phage particle or the host, while the modifying enzyme is synthesized de novo after infection.7 Purification of the components of the in vitro cu-modifying system should reveal how many enzyme activities are involved and whether further substitutions of cr occur duriug the modification process.