Proc. Nati. Acad. Sci. USA Vol. 75, No. 11, pp. 5334-5338, November 1978 Biochemistry RNA sequencing with radioactive chain-terminating (phage Qft replicase/MDV-1 RNA/3'- 5'4a-3Pltriphosphates/persistence of secondary structures during electrophoresis) FRED RUSSELL KRAMER AND DONALD R. MILLS Institute of Cancer Research and Department of Human Genetics and Development, College of Physicians and Surgeons, Columbia University, New York, New York 10032 Communicated by S. Spiegelman, August 14, 1978

ABSTRACT A rapid method for determining Maxam and Gilbert (6): The DNA molecules are usually en- sequences in RNA is described. It employs the 3'-deoxy ana- zymatically labeled with 32P at their 5' ends, and fragment sets logues of the triphosphates as specific chain terminators during RNA synthesis. For example, the inclusion are generated by preferential chemical cleavage. of 3'-deoxyuridine 5'-triphosphate in an RNA synthesis reaction Until now, only one approach has been explored for the rapid in addition to the four usual ribonucleoside triphosphate pre- sequencing of RNA: Homogeneous RNA molecules are enzy- cursors results in the synthesis of a set of ifferent-length matically labeled with 32P at their 5' ends, and fragment sets product strands that terminate in a 3-deoxyuridine that has been are generated by endonucleolytic cleavage (7, 8). This approach incorporated in place of . To sequence an RNA, four separate reactions are run, each employing a different 3'-deoxy to RNA sequencing is limited by the specificity of the available terminator. Parallel electrophoretic analysis of the resulting four ribonucleases. Although endonucleases exist that are specific sets of specifically terminated product chains leads to a direct for guanosines (ribonuclease T1) and (ribonuclease reading of the nucleotide sequence. We tested this method by U2), there are no known ribonucleases that are specific for ur- sequencing MDV-1 (-) RNA, a molecule that is synthesized in idines or . Furthermore, the ribonucleases employed vitro by phage QjB replicase. The sequence read from the re- are single-strand specific, and thus have difficulty in generating sulting gels agreed completely with the known sequence of MDV-1 (-) RNA. The bands in some regions of the sequencing fragments in regions that form secondary structures. gels were unusually close to one another, as has also been ob- This paper describes another technique for sequencing RNA served in other rapid sequencing procedures, making order as- molecules. Fragment sets are generated in vitro by the presence signment in these regions very difficult. Because the secondary of specific chain-terminating analogues during structure of MDV-1 (-) RNA was known, it was shown that the product synthesis. In addition, the chain-terminating nucleo- compression of the bands is due to the persistence of secondary tides are 32P labeled, resulting in the labeling of the structures during electrophoresis. Thus, structured regions of fragments nucleic acids may intr uce difficulties for sequencing tech- at their 3' ends as they are terminated. This method was used niques that employ the currently available methods of gel to sequence MDV-1 (-) RNA (9), a molecule that is synthesized electrophoresis. in nitro by QB replicase (10). Because the complete nucleotide sequence of MDV-1 (-) RNA had already been determined Rapid sequencing procedures depend on the (11) by classical procedures (12), we were able to test the ac- generation of a set of radioactive fragments of the molecule to curacy of the new method. Moreover, we were able to examine be sequenced. These fragments have a common 5' end and the effect of the presence of secondary structures in the RNA specifically terminate in one of the four bases. Four sets of on the sequencing procedure. Many secondary structures have specifically terminated fragments are prepared and electro- been identified and located in MDV-1 (-) RNA with the aid phoretically separated, in parallel, on polyacrylamide slab gels, of reagents (e.g., sodium bisulfite) that react specifically with under conditions designed to resolve chains differing in length that are in a single-stranded conformation (un- by only one nucleotide (1). Because mobility in the gel is a published data). Although secondary structures had no effect function of chain length, the nucleotide sequence can be read on the generation of fragments, the results indicate that the directly from an autoradiograph of the gel. formation of structures in the terminated product chains does This approach was used by Sanger and his colleagues (2-5) influence their electrophoretic mobility. in the sequencing of DNA. Sets of DNA fragments were gen- erated in vitro by using DNA-dependent DNA polymerase. A MATERIALS AND METHODS common 5' end was ensured by the use of specific oligonu- cleotide primers (2). Specific termination at 3' ends was ac- Nucleotides. All four 3'-deoxyribonucleoside 5'-[a-32P]tri- complished in four separate ways: extension of partially syn- phosphates were custom synthesized at specific activities of thesized reaction products in the absence of one of the four 25-170 Ci/mmol (1 Ci = 3.7 X 1010 becquerels) by the Inter- nucleotide precursors (3), exonucleolytic cleavage of the par- national Chemical and Nuclear Corp., Irvine, CA. [a-32P]GTP tially synthesized reaction products (3), the presence of was purchased from the same company. Cordycepin triphos- chain-terminating analogues of the phate (3'-dATP) was obtained from Miles Laboratories. Unla- during the reaction (4), and the presence of ribonucleotides beled ribonucleoside triphosphates were purchased from P-L during the reaction, followed by alkaline hydrolysis of the re- Biochemicals. action products (5). In all of these methods, the product strands QB Replicase. This RNA-dependent RNA polymerase was were labeled by [a-32P]deoxynucleotides that were present at isolated from Qfl bacteriophage-infected Escherichia coli Q13 some time during product synthesis. DNA that is not synthe- by the procedure of Eoyang and August (13), with the hy- sized in vitro can be sequenced by a procedure developed by droxylapatite step omitted. MDV-1 RNA. "Midivariant" RNA was originally isolated The publication costs of this article were defrayed in part by page from a Qf replicase reaction that was not provided with ex- charge payment. This article must therefore be hereby marked "ad- ogenous template RNA (9). It consists of two antiparallel, sin- vertisement" in accordance with 18 U. S. C. §1734 solely to indicate gle-stranded complements. Its biological origin and cellular this fact. function are not yet known. However, once it was isolated, it 5334 Downloaded by guest on September 30, 2021 Biochemistry: Kramer and Mills Proc. Natl. Acad. Sci. USA 75 (1978) 5335 was shown to be an excellent template for Q# replicase. The synthesis of MDV-1 RNA and the separation of the comple- mentary (+) and (-) strands by electrophoresis in the presence o 12 MLoI rj of Mg2+ ions has been described previously (14). In order to E 12_ 20Ji employ a homogeneous template in these experiments, we used E a mutant RNA whose sequence (15) differs from the published ._ 0 sequence of MDV-1 RNA (11) at positions 87 and 88 in the (-) 108 strand. For simplicity, this mutant RNA is referred to as CL "MDV-1 RNA" throughout this report. reaction mix- Sequencing Reactions. Four separate 100-IAI ) 6 tures were prepared, each containing: 84 mM Tris-HCl (pH U) 200 M 7.4), 12 mM MgC12, 100 ,M ATP, 100,uM UTP, 100,uM GTP, ji 100,uM UTP, 500MM of one of the four 3'-deoxyribonucleoside z 4 - [a-32P]triphosphates, MDV-1 (+) RNA template at 43 Asg/ml, and Qfl replicase at 61 tig/ml. Each reaction was incubated at 370C for 8 min and then terminated by bringing the mixture o 2 2,000 jjM to 1 mg of sodium dodecyl sulfate per ml, 10 mM EDTA, and 400 mM NaCl. Unlabeled yeast RNA from Calbiochem (10 Mg) was then added to each as carrier, and each mixture was ex- 0 2 4 6 8 10 tracted with an equal volume of phenol/cresol solution (16). Time (minutes) Deproteinization results in the annealing of partially synthe- FIG. 1. Inhibitory effect of 3'- 5'-triphosphate sized product strands to their template strands (17). Product on the synthesis of MDV-1 RNA by Qfl replicase. The concentration strands shorter than 10 nucleotides do not remain hybridized of triphosphate in each reaction was 200 ,gM. [a-32P]GTP to their templates and were lost during isolation. The RNA was was present to label the product strands. Samples of each reaction previously were taken every 30 sec, and the RNA was freed of unincorporated isolated by gel filtration and alcohol precipitation as nucleotides by absorption onto filter paper, followed by washing with described (15). The RNA from each reaction was dissolved in 300 mM trichloroacetic acid. The reaction conditions and the details 50 Ail of a solution containing: 7 M urea, 5 mM Tris-borate (pH of sample preparation have been described previously (15). 8.3), 200 MM Na2EDTA, and bromophenol blue and xylene cyanol tracking dyes (Eastman Organic Chemicals), each at 500 minate in a 3'-deoxyadenosine that had been incorporated in ug/iml. These solutions were then heated to 1000C for 60 sec, place of adenosine. By keeping the effective concentration of to melt the partially synthesized product strands free of their 3'-dATP low, compared with the concentration of ATP, ter- templates, and they were then chilled to 0°C. One- to 10-,Ml mination would be an infrequent event, and would occur with samples of each product (the volume dependent on the specific approximately equal probability at all adenosines in the se- activity) were then analyzed, in parallel, by electrophoresis on quence. To sequence an RNA, four separate reactions would 12% polyacrylamide slab gels containing 7 M urea. be carried out, each containing one of the four chain-termi- Electrophoresis. Polyacrylamide (12%) slab gels, 405 mm nating analogues. Parallel electrophoretic separation of the high X 306 mm wide X 1.5 mm deep, with slots having a 6.3 products of each reaction would lead to a direct reading of the X 1.5 mm cross section, were cast between glass plates, in an nucleotide sequence. apparatus obtained from Dan-Kar Plastic Products, Reading, The Qft Replicase/MDV-1 RNA System. This system was MA. The solution poured into the mold contained: 116 mg of chosen to test the chain-termination sequencing procedure for acrylamide and 4 mg of N,N'-methylene-bisacrylamide per the following reasons: MDV-1 RNA is a short template (221 ml (both purchased from Bio-Rad Laboratories), 600 Mg of nucleotides) whose sequence and secondary structure are ammonium persulfate per ml, 50 mM Tris-borate (pH 8.3), 2 known and whose complementary strands can be readily sep- mM Na2EDTA, 7 M urea, and 0.06 Ml of NN,N',N'-tetra- arated. Although synthesis of product RNA does not require methylethylenediamine (Eastman Organic Chemicals) per ml. a primer oligonucleotide, it always begins at the same 5'-gua- Gels were run in a buffer containing: 50 mM Tris-borate (pH nosine residue (9), ensuring that all the product strands possess 8.3) and 2 mM Na2EDTA. Each chamber of the electrophoresis the same 5' end. And finally, our recent work on the mechanism apparatus contained 400 ml of the buffer. The buffer was not of replication (19) permitted the design of a reaction in which, recirculated and the gel was not cooled. Gels were first prerun essentially, only one of the two complementary strands is syn- at 800 V for 2 hr. The samples were then loaded and run at 1000 thesized, through many cycles of replication. We found that, V. After electrophoresis, each gel was covered with polyeth- after the completion of product chain elongation, QB replicase ylene film and autoradiographed at -800C with Kodak X- molecules do not remain bound to either the template or the Omat R film in the presence of a Cronex Lighting-Plus inten- product strand, and are free to select a new template, at ran- sifying screen (Du Pont). dom, from the pool of RNAs present in the reaction tube. If reactions are prepared that contain only one type of strand, and EXPERIMENTAL DESIGN if the number of template molecules far exceeds the number of active replicase molecules, then the product will consist of Principle of the Method. The chain-terminating properties essentially only strands that are complementary to the original of the 3'-deoxyribonucleoside triphosphates were demonstrated template. by Shigeura and Boxer (18). They showed that 3'-deoxyaden- Effect of Terminators on MDV-1 RNA Synthesis. Fig. 1 osine 5'-triphosphate (-dATP) can be incorporated into a shows the effect of three different concentrations of 3'-dATP growing RNA chain in place of . How- on the time course of synthesis of MDV-1 RNA. An excess of ever, because this nucleotide lacks a 3'-hydroxyl group, its in- template RNA over active replicase molecules was present, so corporation precludes any further extension of the chain, and that the rate of synthesis would be proportional to the concen- synthesis is terminated. We therefore reasoned that the addition tration of active replicase. All three reactions contained the of 3'-dATP to an RNA synthesis reaction would result in the same concentration of the four normal ribonucleoside tri- synthesis of a set of different-length product chains that ter- phosphates. Each of the reactions was inhibited by 3'-dATP. Downloaded by guest on September 30, 2021 5336 Biochemistry: Kramer and Mills Proc. Natl. Acad. Sci. USA 75 (1978)

I 4 hours 7 hours

_i __ _ - __ A_ _b-w A_ A C G U A C G U A_ 41 -tu .0 _w--*_A|||_ _ ~ -A'---C- G 95 _. -=As A Z------wG 90 e __ a_ _ -A:-G 55 ~c. G-~U 85 _ qmm 50 U 80

i_ _ .... _L _ _ _-_ __ GCGCGAA U3 Cm-: ------U _W 44f U' _ 1 .- - -G-UG 75 A_ -c - G;CGUGGCG __ -A-___-_____ r ~-C ------__,... C *a,. -- 2 -A- L GCG _ 4 m -U __. SCA 55 - 30 qmb -* I-- GGG - 71 -J-a r -C 5 -- _r-- -G,, 50 and- - G: - G -G" - -c .-" 15 CGCGAA G- 40

"ftft i "U, dw*"" - C ------G^ b. - a--- go - FIG. 2. Sequence analysis of MDV-1 (-) RNA by electrophoretic separation of terminated reaction products: 4- and 7-hr gels. The sequence of the RNA can be read in the 5' to 3' direction, beginning with bands at the bottom of the 4-hr gel, and proceeding upward. The sequence can be read further towards the 3' end by continuing upwards on the 7-hr gel. Broken lines are used in the keys to indicate the order ofthe nucleotides. Boxes indicate the regions of band compression. The sequence that should be represented by the compressed bands is written in each box. The more terminator that was present, the greater was the in- bonucleoside [a-32P]triphosphates results in the labeling of hibition. In the absence of 3'-dATP, the rate of synthesis, as fragments at their 3' ends, as they are terminated. Strands that measured by the slope of the kinetic plot, is constant (15). Sig- are not specifically terminated will not be labeled. nificantly, the rate of synthesis in each reaction decreased with Test of the Sequencing Method. Four reaction mixtures time, indicating that the concentration of active replicase were prepared, each containing an excess of pure MDV-1 (+) molecules was decreasing. We interpret this result to indicate RNA as template, and each containing the four normal ribo- that, when a termination does occur, the replicase remains triphosphates at a concentration of 100,uM each. bound to the replication complex and is unavailable for further One of the four 3'-deoxyribonucleoside [a-32P]triphosphate synthesis. Thus, these reactions are self-limiting, and the amount terminators was present in each reaction mixture at a concen- of terminated product that can be obtained is determined by tration of 500 IM. The reaction mixtures were incubated for the initial concentration of replicase. 8 min. The 32P-labeled products were isolated from the reaction Selection of the Terminator Concentration. The frequency mixture as template-product hybrids. These hybrids were of termination is governed by the ratio of terminator to normal melted apart in 7 M urea, and the resulting mixtures were an- triphosphate. We found that Q#t replicase discriminates against alyzed, in parallel, by electrophoresis on 12% polyacrylamide terminator triphosphates. A concentration of roughly 100 times slab gels containing 7 M urea. Four separate gels were run, each as much terminator as normal triphosphate is required to give for a different length of time, in order to obtain information an equal probability of incorporation at a given site. In order about different regions of the sequence of MDV-1 (-) RNA. to ensure that termination would occur with approximately the same frequency at all positions in the sequence, it was desirable that the ratio of terminator to normal triphosphate be kept low. RESULTS AND DISCUSSION We decided to aim for a ratio that would result in a 50% prob- Interpretation of the Sequencing Gels. Figs. 2 and 3 show ability of termination occurring in each cycle of replication. autoradiographs of the gels obtained after 4, 7, 11, and 17 hr We found that a ratio of 5 times as much terminator as normal of electrophoresis. With the exception of regions of "band triphosphate gave the desired result, and that all four termi- compression," which will be discussed in the next section, the nators were about equally effective at this ratio. nucleotide sequence of MDV-1 (-) RNA can easily be read Labeling the Product Strands. The presence of terminators from the pattern of the bands. The sequence that can be read in a reaction affords a unique means of obtaining end-labeled from the gels is in complete agreement with the known se- fragments, because the terminators can be labeled with 32P. quence shown in Fig. 4 . The resolution of the gels permits Labeling the fragments during their synthesis avoids an extra unambiguous sequence determination through the first 143 step in the preparation of the RNA. The use of 3'-deoxyri- nucleotides. Downloaded by guest on September 30, 2021 Biochemistry: Kramer and Mills Proc. Natl. Acad. Sci. USA 75(1978) 5337

11 hours 17 hours

A C G U A C G U

U 105 _A, * -A -A' -----C;;=8 140 A------pmp4 GCACGA -A-- 135 ,WW4_-m - *At=- -::- _f -Az -A- JOG 90 CC--2--U 130 M AA.r = CC ------115 UL~U. z. .- ~~~~~~~~~~~~G.- 4- -. u , _ A ------M U .. -C,------110 U" ----C E L--C: 85 CGCACGUAA~~ ------w 105

U i-G.' -A- -G 80 jjj. I CGCACGAA -A, _ M~~~~~~~~U _A -- -G 90

-. G MU' 75 GCGU5GCG _4d. -A

FIG. 3. Sequence analysis of MDV-1 (-) RNA: 11- and 17-hr gels. No region of the sequence is represented to a greater extent Persistence of Secondary A:*'-Structures. fGThe sequence in a than any other, indicating that the~~~~~~~~~~------frequency of termination number of regions cannot be determined because the bands was sufficiently low. All nucleotides are represented, including representing these nucleotides are unusually close to one an- runs of the same type of nucleotide, such as the four guanosines othbr, and in some cases, they are also out of their expected at positions 47-50. There is some variation in the relative in- order. These regions of band compression are identified by to in a se- in the to 2 and 3. BecauseU85band tensities of the bands, but it is not sufficient result boxes keys Figs. compression quencing error. has been observed with other rapid sequencing techniques that

60 AA CG GC 160 UU UA UC Uc c c CG C G UC CG GC A U CG210 10CG CG GC CG CG A Up UA CG CG GC CG C GE CG 200ACG G U qG GGAA iUUc GAGAGA UuAGCCGUCGAAC UCCCGUACGUCCCCAo,A C U CG 150 170 1 220 8OGC CG GC UAA GC AU GC CG GC CG 140G A GC GC AA UA U A GC U A CG A A C1C CcG9o 190 FIG. 4. Nucleotide sequence of MDV-1 (-) RNA, showing the secondary structures that it forms. Nucleotides that are not resolved well in the sequencing gels are identified by heavy lines. These regions of band compression are associated with nucleotides on the 3' side of hairpin stems. Although the region of nucleotides 58 through 60 at first appears to be an exception to this rule, it is not, because partially synthesized strands between 57 and 65 nucleotides in length form a hairpin (not shown above) by the pairing of nucleotides 29-39 with nucleotides 49-60. Nucleotides 58-60 are on the 3' side of this hairpin structure. Downloaded by guest on September 30, 2021 5338 Biochemistry: Kramer and Mills Proc. Natl. Acad. Sci. USA 75 (1978) employ a similar method of electrophoresis (3, 4), it is probably well as all other, sequencing procedures. Because the nucleotide caused by the conditions in the gel during electrophoresis, and sequence in structured regions is often of particular biological not by the procedure employed to generate the fragments. Fig. interest, more effective denaturing conditions should be 4 shows where in the sequence of MDV-1 (-) RNA the regions sought. of band compression occur. All of these regions are located on the 3' side of hairpin stems. We therefore believe that band We thank Professor Sol Spiegelman for support and encouragement. compression is caused by the persistence of secondary structures We are grateful to Fred Andrea, Carol Williams, and Dan Dimitri- during electrophoresis, despite the presence of 7 M urea and jevich, of the International Chemical and Nuclear Corporation, for high running temperatures in the gels. In regions of a sequence the custom synthesis of the 3'-deoxyribonucleoside 5'-[a-32P]tri- where structures cannot form, each additional nucleotide in- phosphates. We also thank Laurene Moise for expert technical assis- creases the apparent size of the molecule, resulting in a de- tance, Carl Dobkin, Allan Maxam, and Wolf Prensky for helpful iid- creased mobility in the gel. If, however, the additional vice, and John Mack and Edward Hajjar for the preparation of the il- nucleotides can form an intrastrand hybrid, the apparent size lustrations. This work was supported by American Cancer Society of the molecule is not increased, and may even be slightly de- Grant NP-229, National Science Foundation Grant PCM-76-22220, creased. In this sense, the formation of a 3'-terminal hairpin and National Institutes of Health Grant CA-02332-22. "conceals" the 3'-terminal nucleotides. 1. Maniatis, T., Jeffrey, A. & van deSande, H. (1975) Biochemistry Band compression is a striking feature of our gels because of 14,3787-3794. the many strong secondary structures that occur in MDV-1 (-) 2. Sanger, F., Donelson, J. E., Coulson, A. R., Kossell, H. & Fisher, RNA. The structures that are responsible for band compression D. (1973) Proc. Nati. Acad. Sci. USA 70, 1209-1213. in our gels are the same structures that are formed during 3. Sanger, F. & Coulson, A. R. (1975) J. Mol. Biol. 94, 4441- MDV-1 (-) RNA chain elongation (14). Thus, band compres- 4448. sion obscures sequences that occur in structures that are actually 4. Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Natl. Acad. present during replication. Unfortunately, it is often the se- Sci. USA 74,5463-5467. quences in structured regions that are of greatest biological 5. Barnes, W. M. (1978) J. Mol. Biol. 119,83-99. interest. It is possible that the use of thinner gels that can be run 6. Maxam, A. M. & Gilbert, W. (1977) Proc. Nati. Acad. Sci. USA at higher temperatures, as recommended by Sanger and 74,560-564. Coulson (20), will result in more denaturation. However, the 7. Donis-Keller, H., Maxam, A. M. & Gilbert, W. (1977) Nucleic severity by band compression, as indicated Acids Res. 4,2527-2538. of the problem posed 8. Simoncsits, A., Brownlee, G. G., Brown, R. S., Rubin, J. R. & by our results with a molecule of known structure, suggests that Guilley, H. (1977) Nature (London) 269,833-836. stronger measures may be required. 9. Kacian, D. L., Mills, D. R., Kramer, F. R. & Spiegelman, S. (1972) Applications to DNA Sequencing. It may be possible to use Proc. Natl. Acad. Sci. USA 69,3038-3042. the 3'-deoxyribonucleoside triphosphates to sequence DNA by 10. Haruna, I. & Spiegelman, S. (1965) Proc. Natl. Acad. Sci. USA the generation of RNA fragment sets during in vitro synthesis 54,579-587. with DNA-dependent RNA polymerase. Van Kreijl and his- 11. Mills, D. R., Kramer, F. R. & Spiegelman, S. (1973) Science 180, colleagues (21) have recently reported conditions under which 916-927. the DNA-directed synthesis of RNA by E. coli RNA polymerase 12. Sanger, F., Brownlee, G. G. & Barrell, B. G. (1965) J. Mol. Biol. is primer-initiated. Specific primers could be isolated from 13,373-398. restriction digests of the DNA, followed by their ribosylation 13. Eoyang, L. & August, J. T. (1971) in Procedures in Nucleic Acid Research, eds. Cantoni, G. L. & Davies, D. R. (Harper and Row, with the aid of terminal transferase. Or unique oligonucleotides New York), Vol. 2, pp. 829-839. could be isolated from ribonuclease digests of the RNA tran- 14. Mills, D. R., Dobkin, C. & Kramer, F. R. (1978) Cell 15,541- scripts of the DNA, followed by their dephosphorylation with 550. the aid of alkaline phosphatase. In either case, the primers 15. Kramer, F. R., Mills, D. R., Cole, P. E., Nishihara, T. & Spiegel- should be labeled to ensure that any transcripts that are not man, S. (1974) J. Mol. Biol. 89,719-736. primer-initiated will not interfere with the pattern of bands 16. Kirby, K. S. (1968) Methods Enzymol. 12B, 87-100. generated by the terminators. 17. Feix, G., Slor, H. & Weissmann, C. (1967) Proc. Natl. Acad. Sci. Conclusions. The chain-termination procedure is a rapid and USA 57, 1401-1408. very simple means of generating fragment sets for the se- 18. Shigeura, H. T. & Boxer, G. E. (1964) Biochem. Biophys. Res. quencing of RNAs that can be synthesized in vitro. Every hu- Commun. 17,758-763. 19. Dobkin, C., Mills, D. R., Kramer, F. R. & Spiegelman, S. (1978) cleotide is well represented in the fragment set, indicating that Biochemistry, in press. secondary structures do not significantly affect the generation 20. Sanger, F. & Coulson, A. R. (1978) FEBS Lett. 87, 107-110. of fragments. However, the persistence of secondary structures 21. Van Kreijl, C. F., Beelen, R. H. J. & Borst, P. (1977) Nucleic Acids during electrophoresis interferes with the resolution of this, as Res. 4, 445-455. Downloaded by guest on September 30, 2021