Proc. Natl. Acad. Sci. USA Vol. 73, No. 8, pp. 2644-2648, August 1976 Biochemistry

Electron microscopy of DNA crosslinked with trimethylpsoralen: Test of the secondary structure of eukaryotic sequences (cruciforms/electron microscopy under denaturing conditions/glyoxal) THOMAS R. CECH AND MARY Lou PARDUE Department of Biology, Massachusetts Institute of Technology, Cambridge, Mass. 02139 Communicated by Alexander Rich, May 21, 1976

ABSTRACT It has been suggested that inverted repeat strands would accomplish this if (a) the crosslinks were frequent (palindrome) sequences, which are widespread in eukaryotic enough to include the inverted repeat sequences, and (b) the , exist in two alternate configurations, a linear form and the structure. After such a cruciform. To investigate the relative frequency of these forms, crosslinking agent did not disrupt the DNA of intact mouse tissue culture cells was covalently crosslinking, the DNA base pairing would be prevented from crosslinked with 4,5',8-trimethylpsoralen (mes-psoralen) in order undergoing extensive rearrangements via branch migration to prevent rearrangement of the DNA secondary structure during DNA purification and electron microscopy. during DNA isolation. The distribution of me3rpsoralen cross- Psoralen and its derivatives intercalate in the DNA helix and links was determined by electron microscopy after denaturation undergo a photochemical reaction with pyrinidines of opposite of the DNA in the presence of glyoxal. Because of the high fre- as monoadducts quency and the relatively uniform distribution of the merpso- strands, forming covalent crosslinks as well ralen crosslinks, it could be concluded that almost all of the in- (15-17). The advantages of these Psoralen compounds for DNA verted repeat sequences had been crosslinked. In spite of this, crosslinking include their high specificity for DNA compared no significant number of cruciforms was detected by electron with proteins, the stability of the interstrand bonds, the high microscopy. To determine whether the me3-psoralen might itself efficiency of the reaction, and the lack of the DNA degradation be disrupting cruciform structures, cruciforms were first pro- that accompanies crosslinking with other reagents (17-19). duced in isolated Tetrahymena rDNA by heat treatment and can used to crosslink DNA inside of then crosslinked in vitro. The crosslinking was found to stabilize Furthermore, Psoralen be rather than disrupt these cruciforms. We conclude that the in- living cells (15-18). Psoralen crosslinking therefore seemed well verted repeat sequences of the mouse tissue culture cells we suited to a study of the secondary structure of inverted repeat tested are predominantly in linear forms rather than in cruci- sequences in intact cells. form structures inside the cell. Mouse tissue culture cells were chosen for this study because the size and frequency of the inverted repeat sequences from A small percentage of the nuclear DNA of eukaryotes is orga- this source are known in detail (3, 8). The study concerns only nized in the form of inverted repeat sequences, also called inverted repeats with sequences longer than 200 nucleotides palindromes (1-10). After the DNA is denatured, the inverted (NT), the size that can be consistently identified by electron repeat sequences undergo rapid, intramolecular renaturation microscopy (8). A homogeneous class of inverted repeat se- and are seen in the electron microscope as hairpin structures quences, the macronuclear ribosomal DNA (rDNA) of Te- containing hundreds or thousands of base pairs (4, 5, 7, 8). In trahymena (11), was used for in vitro tests of the crosslinking addition to the size of these sequences, we know their distri- of cruciforms. bution in genomes, their sequence complexity, and the fidelity with which the sequences are repeated. These facts, however, MATERIALS AND METHODS have revealed very little about their origin or function. The sequence symmetry of inverted repeats makes possible Crosslinking DNA. SVT2 (simian virus 40-transformed) a "cruciform" mouse tissue culture cells were supplied by Theodore and two base-paired configurations, a linear form and Tucker Gurney. They were grown as monolayers in 25 cm2 (30, 31; Fig. 1). These alternate secondary structures of the same flasks and labeled with palindromic sequence have recently been observed in vitro (11). plastic tissue culture (Falcon) function as [methyl-3H]thymidine as described previously (3). When the If cruciforms occurred in vivo, they-might protein cells were in the log phase of growth (1.2 X 105/cm2), the me- recognition sites (12) or sites for (13), or 3 might provide a method for conserving the complementarity dium in each flask was replaced by ml of sterile phosphate- of the two halves of inverted repeat sequences (4). Wilson and buffered saline (with MgCl2 and CaCl2) saturated with Thomas (4) saw no cruciforms in purified DNA, but suggested 4,5',8-trimethylpsoralen (mes-psoralen) (final concentration that cruciforms might be stabilized by chromosomal proteins 3 ,gg/ml). The me3-psoralen, a gift of the Paul B. Elder Drug in vivo. In purified DNA the linear form should be the ther- Co. (Brian, Ohio), was dissolved in 50% ethanol at 60 ,gg/ml. modynamically favored form (14). Control cultures received the same buffer with 50% ethanol In order to observe in purified DNA the configuration that replacing the me3-psoralen solution. The flasks were irradiated inverted repeat sequences have in vivo, be it linear or cruci- from the bottom with 365 nm light by means of a General form, the DNA configuration must be "fixed" while it is inside Electric F15T8 BLB fluorescent lamp. The lamp was 2.3 cm the cell. Covalent of the two DNA away from the cells, with a one-quarter inch (6 mm) thickness undisrupted crosslinking of Plexiglas filtering out essentially all light of wavelengths 0.72 Abbreviations: me3-psoralen, 4,5',S-trimethylpsoralen; NT, nucleotide below 340 nm (A340 > 3). The incident light intensity was or nucleotide pair; kb, kilobase, 1000 NT; rDNA, DNA sequences mW/cm2, determined by potassium ferrioxalate actinometry coding for 17S and 25S ribosomal RNA with adjacent spacer se- (20). The buffer in the flasks was at a constant temperature of quences. 28° during the irradiation. Light microscopy showed that most 2644 Downloaded by guest on September 30, 2021 Biochemistry: Cech and Pardue Proc. Natl. Acad. Sci. USA 73 (1976) 2645 of formamide, 10.ul of buffer (1.0 M Tris buffer, 0.1 M EDTA, C . pH 8.4), and 37 1l of H20 and spread by the standard 50% b' formamide method. Single-stranded, circular phage 4X174 a b c x c' b' a' +F a' DNA, provided by Claire Moore, was added to some of the : --_lcmx- Cba -F v . samples before glyoxal treatment as a molecular weight stan- dard [5.25 kilobases (kb)]. This glyoxal technique is a variation of the method developed by Hanson et al. (28) for observing me3-psoralen crosslinks. They employed formaldehyde instead of glyoxal to prevent -F rosslink base pairing. We have obtained equivalent results with the two reagents, but prefer the glyoxal technique because it gives more consistently usable preparations. FIG. 1. The cruciform hypothesis. A segment of a long DNA molecule contains an inverted repeat sequence (the sequence abc RESULTS followed by the complementary sequence a'b'c' in reverse order). A The distribution of crosslinks in DNA reacted with mex sequence which is not self-complementary (x) may separate the two psoralen inside growing cells parts of the inverted repeat sequence. Two possible alternate con- figurations ofthe molecule are shown. Factors (F) present inside the Mouse tissue culture cells were treated with me3-psoralen cell may stabilize the cruciform relative to the linear structure. Upon and/or irradiation as summarized in Table 1. A portion of the the removal ofthese factors, the cruciform DNA could rezipper into DNA extracted from each group of cells was examined by the linear form by double-stranded branch migration. Crosslinking the opposite strands of the DNA double helix should stabilize the electron microscopy under totally denaturing conditions to structure and permit cruciforms to be seen even in purified DNA, determine the number and distribution of crosslinks (28). After when the stabilizing factors have been removed. The thin lines rep- the denaturation treatment, sample SV6 showed alternating resent DNA base pairs, the thick lines are covalent crosslinks. single-stranded bubbles and apparently double-stranded regions (Fig. 2a and b). The regions that appeared double-stranded are consist of the cells remained attached to the plastic during the me3- interpreted to of DNA that is densely crosslinked, such psoralen plus irradiation treatment. Purified DNA was cross- that the two denatured strands are constrained to follow each linked in a manner similar to that described above. Drops of other closely (Fig. 2e). The sizes of bubbles and double-strand DNA solution (10 gg/ml) containing 3 ,ug/ml of me3-psoralen regions changed as expected with increasing irradiation dose were irradiated in a humidified plastic petri dish. (Fig. 2c and d and Table 1). When the control samples (SV1-3) Isolation of Crosslinked Mouse DNA. After crosslinking, were denatured and spread for microscopy by the same the SVT2 cells were removed from the flasks with trypsin and method, only single-stranded DNA was seen. harvested by centrifugation. DNA was extracted as described The electron microscopic estimates of crosslinking density previously (3), with the addition of a phenol extraction before were substantiated by denaturation-renaturation experiments. the first chloroform extraction. All steps prior to the phenol Covalent crosslinks hold the two strands of the double helix in extraction were done in low room light. register during denaturation, allowing rapid, unimolecular Isolation of Tetrahymena rDNA. Tetrahymena pyrtformis "snap-back" renaturation (25). Portions of samples SV1-6 were strain B VII, provided by Kathleen Karrer, was grown in sterile sonicated to a weight-average molecular weight of 500 NT and 1% proteose peptone. DNA was extracted by the same method then analyzed by spectrophotometric melting-cooling curves used for the mouse cells, except that two ribonuclease treat- and hydroxyapatite chromatography experiments (data not ments were performed. Each was followed by Pronase treat- shown). DNA that had not been exposed to crosslinking con- ment and chloroform extractions. The rDNA was isolated by ditions (samples SV1-3) exhibited only 3% snap-back rena- two cycles of Hg++-Cs2SO4 density gradient centrifugation turation, the amount expected from inverted repeat sequences (11). About 98% by weight of this DNA was unit length (5.9 Jim) (3, 4). The 500 NT DNA fragments from samples SV5 and SV6, molecules when spread from 50% formamide. All molecules were linear except for rare replicating forms. Table 1. Electron microscopic assay of crosslinking Electron Microscopy. DNA was spread from 50% form- me3- Irrad- Average distance amide according to the isodenaturing technique of Davis et al. psoralen iation between (21). Grids were stained with uranyl acetate and shadowed with Sample (3 gg/ml) (min) crosslinks (NT) 80% Pt-20% Pd. A Jeolco JEM1OOB electron microscope was used. Plates were exposed at a magnification of X15,000. In SV1 - 0 * experiments designed to completely denature DNA in order SV2 - 100 * to see crosslinks, the DNA was treated with glyoxal in the fol- SV3 + 0 * lowing manner. Ten microliters of DNA (dialyzed against 0.10 SV4 + 1 * M Tris.HCI buffer, 0.01 M EDTA, pH 8.4) was mixed with 73 SV5 + 12 890 (95%)t ,ul of 99% formamide (Matheson, Colman, and Bell), 1O0ul of <2004 (5%)t 0.10 M NaH2PO4-Na2HPO4 buffer, 0.01 M EDTA, pH 6.9, SV6 + 100 310 (56%)t and 7 gl of glyoxal (Eastman Technical Grade, 40% in H20). <200f (44%)t The final glyoxal concentration was 0.5 M. The solution was *After denaturation in the presence of glyoxal, these samples heated at 370 for 45-60 min. The combination of low ionic showed no evidence of crosslinks. strength and high formamide concentration results in complete t The numbers in parentheses give the percent by weight of the denaturation of DNA at 370. The glyoxal reaction (22) prevents DNA that had the stated distance between crosslinks. I These regions appeared double-stranded after denaturation. Al- base pairing (predominantly G.C base pairing) even when the though bubbles as small as 100 NT were seen with a significant DNA is returned to more stabile conditions (23, 24). After the frequency, 200 NT was the estimated size at which a bubble was denaturation, 10 Al of the above solution was diluted into 43 consistentlyctdetected. Downloaded by guest on September 30, 2021 2646 Biochemistry: Cech and Pardue Proc. Natl. Acad. Sci. USA 73 (1976) Table 2. The frequency of various structures found in mouse DNA crosslinked in intact cells (sample SV6) Molecules scored Structure Number % Simple linear* 1889 93.5 Intersecting lineart 76 3.8 Possible cruciform$ 3 0.1 Tangled § 46 2.3 Circles¶ 7 0.3 Total 2021 100.0 * Structures with two ends. t Structures with more than two ends, but no two segments having equal length, were interpreted to be two (or three) linear mole- cules crossing over each other. 0 kb - t Any four-ended structure with two of the segments having equal length. Two of the possible cruciforms had arms of length 100 FIG. 2. Crosslinked DNA spread for electron microscopy after NT, near the detection limit for this electron microscopy tech- denaturation in the presence of glyoxal. All micrographs are 26,000 nique. The other structure had arms of length 8700 NT and could X magnified. (a) Sample SV 6 (Table 1), a portion of a long molecule also be interpreted as a fortuitous intersection of two long showing typical alternation of densely crosslinked regions, which molecules. appear double-stranded, and uncrosslinked regions, which form § Molecules that contained one or more regions that were not single-stranded bubbles. (b) Another molecule from SV6, showing unambiguously traceable, usually because of multiple cross- more uniformly spaced crosslinks. This was a less common situation overs. None of these tangled regions resembled the type of than (a). (c) Sample SV5. (d) Purified SVT2 DNA crosslinked for 64 complex cruciform structure described by Kleckner et al. (26). min in 0.15 M NaCl-0.001 M EDTA-0.01 M P04 buffer, pH 6.9. The 1 Presumed to be mitochondrial DNA because of 5.0 um contour in vitro crosslinking was always much more efficient than that done length. inside cells, so this molecule is much more densely crosslinked than those in (a) and (b). (e) Interpretation of electron micrographs. Re- than resulting from the random crossover of two denatured gions u, w, and z appear double-stranded in the denatured DNA mi- strands. crographs, and are called "densely crosslinked regions." They consist The uncrosslinked regions in the SV6 sample were mostly of denatured bubbles that are too small to be resolved. Regions v, x, smaller than 400 NT with two and y contain no crosslinks and are called "bubbles." Regions x and peaks of bubble sizes at 175 and y are considered to be two separate bubbles if the two strands of each 325 NT (Fig. 3). A bimodal distribution of uncrosslinked regions bubble have equal length, which is the most common case. Indepen- of DNA would not be expected for random crosslinking and was dent evidence that adjacent bubbles (like x and y) are usually sepa- not seen when deproteinized DNA was crosslinked to a similar rated by one or more crosslinks is mentioned in the text. extent (data not shown). The uncrosslinked regions were not due to certain sequences' being resistant to crosslinking, because on the other hand, showed 47% and 87% snap-back renatura- purified SVT2 DNA could be very completely crosslinked (Fig. tion, approximately the values predicted from the electron 2d). A possible explanation for the nonrandom crosslinking of microscopy distributions. The crosslinking frequencies obtained DNA in intact cells will be discussed. by electron microscopy agreed most closely with those given Even though the crosslinks may not be located randomly with by the renaturation experiments when adjacent bubbles (Fig. respect to inverted repeat sequences, the frequency of crosslinks 2e) were interpreted as being separated by a crosslink, rather in the heavily irradiated SV6 DNA is high enough to assure that the average-sized inverted repeat sequences have been cross- linked. This DNA contained essentially no regions where ad- 175 jacent crosslinks were separated by more than 1200 NT (Table 1 and Fig. 3), while the average base-paired length of hairpins in single-stranded mouse DNA (1000 NT, ref. 8) corresponds 125 to a 2000 NT length in the native DNA. Even the smallest in- verted repeat sequences that can be consistently identified by electron microscopy (400 NT sequences which would form 200 NT hairpins in single-stranded DNA) have an estimated 94% 75- probability of being crosslinked. This is the probability that a random 400 NT DNA segment contains at least one crosslink. 52- If inverted repeat sequences in vivo were protected from Pso- ralen crosslinking to a greater or lesser extent than the bulk 25 chromatin, this estimate would be affected.

2 3 4 5 6 7 8 9 1.01.1 12 A search for cruciforms in crosslinked mouse DNA kb The SV6 sample, the DNA that was most densely crosslinked FIG. 3. Size distribution of uncrosslinked regions in sample SV6 inside of intact cells, was spread on grids using the standard 50% DNA. The lengths of 909 bubble regions on 10 randomly chosen formamide method. Each molecule in the central portion of molecules were measured using a Numonics electronic graphics cal- one was to one of four Three culator. These bubble regions comprised 56% ofthe DNA by weight. grid assigned categories (Table 2). The remainder of the DNA (the densely crosslinked regions) con- possible cruciform structures were found after scanning over tained denatured bubbles too small to be resolved by the technique 2000 molecules with an average DNA molecular weight cor- used here. responding to about 40 kb. In the mouse , the average Downloaded by guest on September 30, 2021 Biochemistry: Cech and Pardue Proc. Nati. Acad. Sd. USA 73 (1976) 2647 Table 3. Frequency of cruciforms in heat-treated Tetrahymena rDNA % of moleculest Cruciform/ Experiment Crosslinked* nt Cruciform Linear Tangled linear 1 - 750 58 39 3 1.5 2 - 550 48 37 15 1.3 + 557 64 25 11 2.5 * +, DNA treated with 3 Ag/ml of me3-psoralen and 8 min irradiation. -, DNA irradiated for 8 min without me3-psoralen in Exp. 1 and treated with neither me3-psoralen nor irradiation in Exp. 2. t The number of molecules scored. In both Exps. 1 and 2 +, the results are the pooled data from two grids. t DNA fragments shorter than rDNA unit length made up 3-7% of the total number of molecules in the various samples. They are omitted from this analysis.

distance between those inverted repeat sequences that are large the base-pairing during subsequent DNA extraction and enough to be seen in the electron microscope is 75 kb (8). preparation for electron microscopy. The high frequency of Therefore, the 2000 DNA molecules described in Table 2 crosslinks assured that most of the inverted repeat sequences contained about 1000 inverted repeat sequences, almost none in the mouse genome contained crosslinks. When the highly of which were seen as cruciforms. No cruciform structures were crosslinked DNA was examined in the electron microscope, it seen in the same sample spread after glyoxal treatment (Fig. was found to consist almost entirely of linear structures. If 1-2% 2a and b), nor were any seen in the more lightly crosslinked SV5 of the inverted repeat sequences had been cruciforms, they sample spread with or without the glyoxal treatment. could have been detected. We therefore conclude that the in- verted repeat sequences of SVT2 mouse tissue culture cells are A model system for cruciform crosslinking predominantly in linear forms rather than in cruciform struc- The extrachromosomal ribosomal DNA (rDNA) of Tetrahy- tures. It remains possible that cruciforms exist under some mena has an accurate inverted repeat sequence 20 kb in length special circumstances or in some cell type other than the rapidly (11, 27). Cruciform structures are often seen at low frequency growing, virus-transformed fibroblasts used here. We again in preparations of this rDNA, especially after it is incubated at emphasize that our study is confined to those inverted repeats 370 (11). The cruciforms can be identified as four-ended that are large enough to be identified by electron microsco- structures in which the lengths of the two short "arms" (Ls) are PY. equal, the lengths of the two long arms (LJ) are equal, and 2LS It is unlikely that cruciforms were not seen because they were + 2LL = 20 kb. obscured in tangled structures. The grid that was most exten- Tetrahymena rDNA was purified as described in Materials sively analyzed contained an unusually low frequency of un- and Methods. In order to obtain a sample with a large fraction interpretable structures (Table 2). Furthermore, cruciforms do of the molecules in the cruciform configuration, the rDNA was not seem to become preferentially tangled: in the electron partially denatured by heating and then allowed to cool to room microscopy of Tetrahymena rDNA, similar frequencies of temperature. After this treatment, the rDNA had a ratio of tangles were found in samples containing only linear molecules cruciform to linear molecules of 1.3-1.5 when spread from 50% and in samples containing a high proportion of cruciform formamide (Table 3). An example of a cruciform is shown in molecules. Fig. 4a. The LS/LL ratio varied from almost zero to about It is possible that cruciforms were present in undisturbed 0.8. chromatin but were converted to linear forms during the Ribosomal DNA that had been heated to produce cruciforms crosslinking treatment. We have no evidence that me3-psoralen was irradiated in the presence of 3 ,ug/ml of mes-psoralen for might have caused the removal of cruciform-stabilizing ele- 8 or 22 min. It was then examined by electron microscopy, using ments that could exist inside the cell. However, in the in vitro the 50% formamide and the glyoxal-denaturation techniques. studies with Tetrahymena rDNA, me3-psoralen crosslinking Cruciforms were observed after either treatment (Fig. 4b-e). did not disrupt cruciforms. The mouse inverted repeats have In order to determine whether the me3-psoralen might have a size distribution (8) and thermal stability (3, 4) that overlap disrupted some of the cruciforms, crosslinked and uncrosslinked with those of the Tetrahymena rDNA (11). This makes the portions of a single heat-treated rDNA preparation were scored rDNA a good model system for the chromosomal inverted re- for cruciform structures (Table 3, Exp. 2). The ratio of cruci- peats, even though only the former is thought to have a perfect forms to linear forms in the crosslinked portion was 2.5, com- palindromic sequence. pared to 1.3 in the uncrosslinked portion. Crosslinking a sample The techniques for directly observing Psoralen crosslinks in of linear rDNA did not result in the formation of cruciforms the electron microscope, presented in the work of Hanson et (data not shown). It therefore appears that crosslinking stabi- al. (28) and in this paper, give information about crosslink lized rather than disrupted the cruciform structures, preventing distribution that cannot be readily obtained by previous a loss of cruciforms that otherwise occurred during spreading methods. The distribution seen when the crosslinking is done of the DNA for microscopy. in intact cells (Fig. 3) appears to be nonrandom. Hanson et al. (28) have observed similar patterns in DNA crosslinked inside of Drosophila embryo nuclei. Chromosomal proteins may be DISCUSSION protecting regions of discrete sizes from me3-psoralen cross- The symmetry of inverted repeat DNA sequences makes pos- linking (28, 29). sible two base-paired configurations, a linear form and a cru- The methods described here could be applied to many other ciform. In order to determine which secondary structure ac- systems in which inverted repeat sequences have been char- tually exists inside the mouse cell, the DNA was covalently acterized. The question of whether the Tetrahymena rDNA crosslinked with mer-psoralen to prevent rearrangements of exists as a cruciform in vvo can be investigated (K. Karrer and Downloaded by guest on September 30, 2021 2648 Biochemistry: Cech and Pardue Proc. Natl. Acad. Sci. USA 73 (1976) *.fi...;..::. :..-4s...:bo: -.$. ....:.: Preliminary me3-psoralen crosslinking experiments were done by T.C. in the laboratory of John E. Hearst at the Department of Chem- *, ih ; istry, University of California, Berkeley. T.C. thanks John Hearst, James , ...... * .... . Wang, Alex Karu, Stuart Linn, Carl Hanson, James Shen, and Gary ..;. -0' ....'- l,; ; .....ah b; ...... Wiesehahn for discussions about DNA crosslinking, and Elaine Lenk for discussions about electron microscopy. Kathy Karrer and Joseph

i ;, ;.s, , < *t t- is or Gall provided Tetrahymena cells and much useful information. David % % e; . + i P * > , r z & Potter helped with the electron micrograph measurements and the ;'X'W.i.; l- A.|& i.s > * ^a1.h b{h . f,a¢ actinometry. Paul Leibowitz and Alan Spradling helped repair the manuscript. This investigation was supported by Research Fellowship 5 F32 CA05178-02, awarded to T.C. by the National Cancer Institute, Department of Health, Education, and Welfare and by National Science Foundation Grant 7519524-BMS to M.L.P. , And S- t ,, , w e $ A, ,.'e, ,,4,* ...... ,,.* .. t t P. Mol. Gen. 104, 1. Walker, M. B. & McLaren, A. (1969) Cenet. 104-106. 2. Britten, R. J. & Smith, J. (1970) Carnegie Inst. Washington Yearb. 68, 378-386. 3. Cech, T. R., Rosenfeld, A. & Hearst, J. E. (1973) J. Mol. Biol. 81, 299-325. z ; . + * ,Q w , 9 A 4. Wilson, D. A. & Thomas, C. A., Jr. (1974) J. Mol. Biol. 84, 115-138. 5. Davidson, E. H., Graham, D. E., Neufeld, B. R., Chamberlin, M. E., Amenson, C. S., Hough, B. R. & Britten, R. J. (1973) Cold Spring Harbor Symp. Quant. Biol. 38, 295-301. 6. Georgiev, G. P., Varshavsky, A. J., Ryskov, A. P. & Church, R. B. (1973) Cold Spring Harbor Symp. Quant. Biol. 38, 869- 884. 7. Schmid, C. W., Manning, J. E. & Davidson, N. (1975) Cell 5, 159-172. ..7~~~~~~~~~~~I 8. Cech, T. R. & Hearst, J. E. (1975) Cell 5,429-446. 9. Firtel, R. A. & Kindle, K. (1975) Cell 5,401-411. 10. Schmid, C. W. & Dieninger, P. L. (1975) Cell 6,345-358. 11. Karrer, K. M. & Gall, J. G. (1976) J. Mol. Biol. 104, 421-453. 12. Gierer, A. (1966) Nature 212, 1480-1481. 13. Sobell, H. M. (1972) Proc. Natl. Acad. Sci. USA 69, 2483- 0 kb 2487. 14. Tinoco, I., Jr., Borer, P. N., Dengler, B., Levine, M. D., Uhlen- FIG. 4. Cruciforms in Tetrahymena rDNA. Purified rDNA was beck, 0. C., Crothers, D. M. & Gralla, J. (1973) Nature New Biol. adjusted to 0.08 M Na+ and sealed in a capillary. In order to form 246,40-41. cruciforms, the DNA was partially denatured by heating for 15 min. 15. Musajo, L. & Rodighiero, G. (1970) Photochem. Photobiol. 11, at 80°. After cooling, the solution was adjusted to 0.18 M in Na+ and 27-35. split into two samples. One was spread immediately for electron mi- 16. Cole, R. S. (1970) Biochim. Biophys. Acta 217,30-39. croscopy, using the standard 50% formamide hyperphase. The other 17. Musajo, L. & Rodighiero, G. (1972) in Photophysiology, ed. Giese, was made 3 in me3-psoralen, irradiated, and then eth- portion Ag/ml A. C. (Academic Press, New York), Vol. VII, pp. 115-147. anol precipitated to remove unreacted me3-psoralen. It was later spread using the 50% formamide and the glyoxal-denaturation 18. Pathak, M. A. & Kramer, D. M. (1969) Biochim. Biophys. Acta techniques. (a) Cruciform structure in an rDNA sample that was not 195, 197-206. crosslinked, 50% formamide spread. The small separation at the in- 19. Cole, R. S. & Zusman, D. (1970) Biochim. 3iophys. Acta 224, tersection of the four strands was common in some spreads. (b) Cru- 660-662. ciform structure in crosslinked rDNA, 50% formamide spread. (c) 20. Hatchard, C. G. & Parker, C. A. (1956) Proc. R. Soc. London Ser. Visualization of the crosslinks in a sample of cruciforms irradiated A 235,518-536. for 8 min with me3-psoralen. The DNA was treated with glyoxal under 21. Davis. R. W., Simon, M. & Davidson, N. (1971) in Methods in denaturing conditions before being spread. Crosslinks are seen in both Enzymology, eds. Grossman, L. & Moldave, K. (Academic Press, pairs of arms; the ends of the molecule can be identified because of New York), Vol. 21, pp. 413-428. the absence of terminal crosslinks. (d) Same treatment as (c), except 22. Broude, N. E. & Budowsky, E. I. (1971) Biochim. Biophys. Acta a 22-min irradiation time gave a much higher density of crosslinks. 254,380-388. (e) The central portion of a molecule showing one of the smallest pairs 23. Hsu, M.-T., Kung, H.-J. & Davidson, N. (1973) Cold Spring of cruciform arms that could be recognized; from the same grid as Harbor Symp. Quant. Biol. 38,943-950. (d). 24. Johnson, D. (1975) Nucleic Acids Res. 2,2049-2054. 25. Geiduschek, E. P. (1961) Proc. Nati. Acad. Sci. USA 47, 950- 955. T. Cech, unpublished results). More generally, Psoralen cross- 26. Kleckner, N., Chan, R. K., Tye, B.-K. & Botstein, D. (1975) J. Mol. linking might be a useful technique for in vivo and in vitro Biol. 97, 561-575. V. & studies of a wide variety of labile DNA structures: DNA het- 27. Engberg, J., Andcersson, P., Leick, Collins, J. (1976) J. Mol. 455-470. eroduplexes and intramolecular hairpin structures with a low Biol. 104, 28. Hanson, C. V., Shen, C.-K. J. & Hearst, J. E. (1976) Science 193, degree of homology, as well as structures that might rearrange 62-64. by branch migration, such as intermediates in genetic recom- 29. Wiesehahn, G. & Hearst, J. E. (1976) Abstract, ICN-UCLA bination. After such structures were stabilized by crosslinking Conference on Molecular and Cellular Biology. under favorable conditions, they would not be disrupted when 30. Platt, J. R. (1955) Proc. Natl. Acad. Sci. USA 41, 181-183. exposed to much less stable conditions during DNA purification 31. Schwartz, D. (1955) J. Cell. Comp. Physiol. 45, suppl. 2, 171- or formamide electron microscopy. 188. Downloaded by guest on September 30, 2021