Loopstructures in synthetic oligonucleotides. Hairpin stability and

Proc. Int. Symp. Biomol. Struct. Interactions, Suppl. J. Biosci., Vol. 8, Nos 3 & 4, August 1985, pp. 767–780. © Printed in India.

Loopstructures in synthetic oligonucleotides. Hairpin stability and structure studied as a function of loop elongation C. A. G. HAASNOOT, S. H. DE BRUIN, C. W. HILBERS, G. A. VAN DER MAREL* and J. Η. VAN BOOM* Laboratory of Biophysical Chemistry, Faculty of Science, University of Nijmegen, Toernooiveld, 6525 ED Nijmegen, The Netherlands * Gorleaus Laboratoria, State University, P.O. Box 9502, 2300 RA Leiden, The Netherlands Abstract. The formation of hairpin structures in the homologous, (partly) selfcomplementary DNA fragments d(ATCCTATnTAGGAT), n = 0–7, was studied by means of nuclear magnetic resonance, T-jump and ultra-violet techniques. It is shown that all compounds in the series may adopt hairpin-like conformations, albeit for n < 3 this only occurs to a significant amount at relatively low concentrations (~ 10 µΜ). For the present series of oligonucleotides, hairpin formation is accompanied by an apparent loop enthalpy significantly different from zero. The stability of the DNA hairpins turns out to be at its maximum for loop lengths of four or five residues, whereas earlier experiments (Tinoco et al., 1973) indicated that loop lengths of six to seven residues are most favourable for RNA hairpins. This is explained by considering the difference in geometry of A-RNA and B-DNA helices. Keywords. DNA; nuclear magnetic resonance; T-jump; oligonucleotides; conformation.

Introduction Nucleic acids are known to occur in a rich variety of three-dimensional structures. For DNA, the basic double helical structure proposed over thirty years ago by Watson and Crick (1953) still stands, albeit in the course of time many refinements were made to this basic model. Several types of righthanded double helices are now described in the literature (baptized A, B, C, D and Ε-type), but it took almost 25 years before the monopoly of these canonical, very uniform, smooth structures was definitely broken. For instance, single crystal X-ray studies disclosed the existence of aberrant helices such as the lefthanded Z-DNA structure (Wang et al., 1979), while from the work of Dickerson and co-workers (Dickerson, 1983) it became evident that the canonical B-DNA helix is not really regular: in fact, the local structure of the B-DNA helix appeared to be dependent upon the base-sequence. But also other structural features are recognized in DNA conformations, e.g. cruciform (hairpin) structures formed at DNA sequences with short inverted repeats are hypothesized to be of importance in recognition processes, etc. Hairpin structures (in the broadest sense of the word) are certainly of importance in the RNA genus of nucleic acid molecules. Due to the single helical nature of RNA, these molecules take up very complicated secondary/tertiary structures by folding back on themselves, thus forming very complicated hairpin-like structures. Examples of such structures are abundant, we just mention 5S-RNA and tRNA as (relatively small) representatives of the type. 767

768

Haasnoot et al.

In our search to a detailed structural and thermodynamical description of hairpin formation in nucleic acids we started a systematic investigation of (partly) selfcomplementary oligonucleotides. The present paper reports on the thermostability of tailor-made synthetic oligomers, which were studied by means of UV-melting, temperature (T-) jump relaxation and NMR techniques, as a model system for hairpin formation. Hairpin formation in a series of homologous DNA fragments The model systems used in this comparative study consist of a series of synthetic homologous DNA-fragments delineated by the generic base-sequence d(ATCCTATnTAGGAT). For n = 0 this sequence is fully self-complementary; for n > 0 the molecules are only partly self-complementary. In principle, these DNA fragments are capable of adopting either a hairpin structure (monomeric form) or an interior loop structure (dimeric form), cf. figure 1.

Figure 1. (a) Generic base sequence of the series of homologous oligodeoxynucleotides studied, (b) Secondary structures accessible to the homologous molecules given under (a); the prefix d (for deoxy) is omitted for the sake of clarity. Base pair numbering as shown; note the pairwise equivalency of the base pairs in the dimeric form.

In an earlier report (Haasnoot et al., 1983a) it was assessed that for n 4 the DNAfragments adopt exclusively a hairpin-like conformation, both at high (millimolar) and 3) were less low (micromolar) oligomer concentrations. The smaller fragments (n rigorously studied: a 360 MHz NMR study (Haasnoot et al., 1979, 1980) at relatively high nucleotide (~ 3 mM) and salt (0·6 M) concentrations showed that the smallest

Hairpin stability and loop elongation

769

fragments (n = 0, 1) adopt the dimeric form; for n = 3 the molecule was in the monomeric (hairpin) form, whereas the NMR experiments indicated that for n = 2 both the monomer and the dimer appear to exist side by side under these conditions. In order to raise pertinent information with regard to the influence of the loop size on the thermodynamic stability of these DNA hairpins we investigated the thermally induced helix-to-coil transition in the smaller fragments (n 3) by means of UV melting and T-jump experiments. At UV concentrations (~ 10 µΜ) the pentadecamer (n = 3) showed a single transition (characterized by a very short relaxation time τ 5·3 µs) ascribed to the hairpin-to-coil transition. However, at high oligomer concentrations (~ 1 mM) the differential melting curve of this pentadecamer displayed a biphasic behaviour: two transitions were observed (figure 2), of which the one at the highest temperature (Tm = 54°C) corresponds with the transition recorded at low oligomer concentration (Tm = 52°C). From this concentration independency it follows that this transition is to be ascribed to the hairpin-to-coil equilibrium. The low temperature transition around 13°C points to the presence of a dimeric species at high concentrations.

Figure 2. Differential UV melting curve recorded at 260 nm for d (ATCCTATTTTAGGAT); (cf. figure 1a, n= 3). Total strand concentration 1 mM; Na+ concentration 200 mM.

This finding urged us to reinvestigate the NMR-spectrum of this pentadecamer (n = 3). The increased dispersion and higher sensitivity obtained at 500 MHz indeed allowed us to detect two species at – 1°C, (figure 3) the spectrum may be interpreted in terms oftwo subspectra, one pertaining to a major species (presumably hairpin) and one pertaining to a minor species (probably dimer). The resonances in the two subspectra were assigned (figure 3) on basis of chemical shift considerations (obtained from comparing the spectra of the complete homologous series n = 0–7, (cf. figure 1), supplemented by a number of one-dimensional nuclear Overhauser experiments. Remarkably, the imino proton resonance belonging to the A·T base pair 6 in the hairpin structure (cf. figure 1b) appears to be absent from the spectrum. The NMR experiments thus suggest that this A·T base pair 6 is not formed in the pentadecamer

770

Haasnoot et al.

1

Figure 3. Imino proton region of the 500 MHz H-NMR spectrum of the pentadecamer (n = 3); solvent H2O/D2O (95:5 v/v), pH = 6, ~ 1 mM strand concentration, [Na+] = 200 mM; Τ = 272 Κ. The numbering of the resonances corresponds to the numbering of the base pairs in figure 1.

hairpin. We take this for an indication that the nucleotides corresponding to this base pair are incorporated in the loop (thereby elongating the loop to five residues and concomitantly reducing the length of the stem to five base pairs). We shall comment upon this interpretation in a following section. The UV melting curves recorded for the tetradecamer (n = 2) and tridecamer (n = 1) 53°C, at low oligomer concentrations (~ 10 µΜ) displayed a single transition (Tm independent from oligomer concentration). T-jump experiments showed that the process involved in these transitions is characterized by a very short relaxation time 5 µs). The magnitude of these relaxation constants is in concurrence with the (τ relaxation time determined for hairpin formation in the hexadecamer (n = 4) (Haasnoot et al., 1983a). Moreover, since the helix-to-coil transition in dimeric species under these conditions is well-known to be characterized by relaxation times in the order of 10–100 ms (Tibanyenda et al., 1984) we are confident that the tetradecamer and tridecamer at low oligomer concentrations adopt exclusively a hairpin conformation. The fully self-complementary dodecamer (n = 0) was studied next by the UV technique and, surprisingly, at low oligomer concentrations (~ 10 µΜ) a biphasic melting behaviour was observed. The following findings are relevant: the temperature around which the first transition occurs increases with increasing dodecamer concen53°C) which tration, in contrast to the Tm of the high temperature transition (Tm is independent of oligomer concentration. According to T jump experiments, the low-temperature transition involves a relaxation rate in the order of 80 ms, while the 4 µs). From high temperature transition is characterized by fast relaxation (τ these observations we conclude that at low concentrations the self-complementary dodecamer d(ATCCTATAGGAT) can attain two conformations, i.e. the dimeric form and the monomeric hairpin form. The predominant conformation in solution depends on both the dodecanucleotide and the salt concentration; NMR studies (Haasnoot et al., 1979) indicated that at high oligomer concentrations (~ 3 mM) and high ionic strength ( [Na+] = 0·6 M) the dimeric form prevails. We note that our findings are in


771

keeping with a recent study of Marky et al. (1983) who found that the selfcomplementary oligomer d(CGCGAATTCGCG) displays similar characteristics. For the time being we abstain from further comments with regard to the dimeric forms as they are deemed beyond the scope of the present paper. Instead we will now deal with the monomeric form in some detail since our results offer the opportunity to evaluate the thermodynamics of hairpin formation in a comparative way. Figure 4 displays the salt-dependence of T m for the hairpins formed by the oligodeoxynu-

Figure 4. UV melting temperatures, T m , recorded at 260 nm for the hairpins d(ATCCTATnTAGGAT), n = 0–7, plotted as a function of salt concentration. Total strand concentration 10 µΜ. The hairpins are characterized by the abbreviations (number 1) Τ (number 2), where number 1 indicates the length of the oligonucleotides and number 2 the number of intervening thymidine residues. For 14T2 only one melting temperature is indicated.

772

Haasnoot et al.

cleotides (figure 1a, n = 0–7). It is seen that at the lower salt concentrations Tm increases linearly with In [Na+] for all fragments. The slopes intend to increase with increasing chain-length a phenomenon that has been observed earlier (Elson et al., 1970; Tibanyenda et al., 1984). All curves show a maximum around 1 Μ Na+; the occurrence of a maximum in the melting temperature of oligonucleotides around 1 Μ NaCl concentration seems more or less a general rule. So far we have found few exceptions. The dodecamer studied in this paper displays a maximum at slightly lower salt concentrations while tRNA’s display maxima at much higher salt concentrations. So far we have no theoretical explanation for these observations. The heats of hairpin formation (∆Η°) were calculated from the UV differential melting curves according to Gralla and Crothers (1973). The results show that for all fragments studied (figure 1a, n = 0–7) the enthalpy values for hairpin-to-coil transition are independent of salt concentration up to [Na+] 0·5 M; above this ionic strength the magnitude of ∆Η° decreases. In figure 5 the heats of hairpin formation for d(ATCCTATnTAGGAT) are summarized as a function of n ( = the number of intervening thymidine residues). For purposes of comparison the corresponding ∆Η° (= – 33 kcal/mol) determined for the duplex formed by the complementary strands d(ATCCTA) and d(TAGGAT) is also indicated (dashed line in figure 5). The base sequence of the latter hexanucleotide dimer matches that of the stem of the hairpins and may therefore serve as a reference point for the stability of the ‘unconstrained’ stem.

Figure 5. ∆Η° values obtained for the hairpin-single strand transition plotted as a function of intervening thymidine residues, n, in the homologous series d (ATCCTATnTAGGAT), [Na+] = 0·2 M. The dashed line indicates the value of ∆Η° measured for the duplex formed by d (ATCCTA) and d (TAGGAT) which in the hairpins forms the stem.

Figure 5 shows that for n > 4 the heat of hairpin formation is approximately constant (∆Η° – 43 kcal/mol) and about 10 kcal/mol more favourable than ∆Η° determined for the reference dimer d(ATCCTA)·d(TAGGAT). In other words, the


773

apparent enthalpy of loop formation in the hairpins with more than four thymidine residues in the loop can be estimated to – l0 kcal/mol. The origin of this negative enthalpy term cannot be established from the data at hand. However, it seems reasonable to hold stabilization of base pair A·T6 and stacking interactions within the loop or between the loop and the top of the hairpin stem responsible for this phenomenon (Haasnoot et al., 1983a). With respect to the large (n > 4) hairpins, the hexadecamer (n = 4) appears to be a little less stabilized (ca. – 6 kcal/mol). This reduced stabilization may point to some steric crowding of the thymidine residues in the (tighter) loop of the hexadecamer. Interestingly, the pentadecamer (n = 3) still shows a negative apparent enthalpy of loop formation (ca. – 3 kcal/mol) notwithstanding the observation (vide supra) that the stem in this hairpin has sacrificed one A·T base pair in order to elongate the loop to five residues. Obviously, the stacking interactions brought in by the loop outweigh the breaking of one A·T base pair. Note that this explanation also fits in numerically: the stabilisation caused by loops of five residues is estimated above to amount to – 10 kcal/mol; since the breaking of an A·T base pair will cost ca, 7 kcal/mol (Marky and Breslauer, 1982), loop formation in the pentadecamer will still have a favourable enthalpy of ca. – 3 kcal/mol, thus in perfect agreement with the observed value. The magnitudes of the enthalpies determined for the smallest fragments (n = 0–2) have clearly diminished with respect to that of the "unconstrained" duplex stem. Remarkably, the observed numerical values for ∆Η°hairpin can be rationalized on basis of the following equation:

where ∆Η°stem = – 33 kcal/mol, ∆Η°lοορ = –10 kcal/mol (for loops with five or more residues) or – 6 kcal/mol (for loops with four residues) and ∆Η°broken base pairs = – 7 kcal/mol for each broken A·T base pair. For instance, for the hairpin formed by the dodecamer we calculate (presuming a loop of four residues formed by the nucleotides of the central-TATA-sequence) ∆Η°hairpin= – 25 kcal/mol (experimental: ∆Η°hairpin = – 22 to – 25 kcal/mol); for the tridecamer (presuming the breaking of the A·T base pairs 5 and 6, thus forming a loop of five residues) a ∆Η°hairpin = – 29 kcal/mol is calculated (experimental: ∆Η° – 31 kcal/mol). For the tetradecamer (n = 2) the case is less clear-cut: due to the error margins in the experimental determination of ∆Η°hairpin ( – 29 to – 34 kcal/mol) it cannot be decided whether a loop of four residues (created by incorporating base pair 6 into the loop; calculated ∆Η°hairpin = – 32 kcal/mol) or a loop of six residues (created by incorporating base pairs 5 and 6 into the loop; calculated ∆Η°hairpin = – 29 kcal/mol) is formed. Be this as it may, the present study shows that loop formation in oligodeoxynucleotides may contribute favourably to the overall enthalpy content of the hairpin structure. The data at hand indicate that a loop consisting of four to five deoxynucleotides may even (over) compensate the breakage of an A·T base pair. Generally, we expect the relative preference of DNA fragments for either the hairpin or the dimer structure to depend on oligomer concentration and ionic strength. Indeed, for the smaller fragments (n = 0–3) our experiments show that this preference may be influenced by the oligomer concentration; however, for the larger fragments (n 4) no

774

Haasnoot et al.

dimer formation could be detected (oligomer concentrations up to ~ 3 mM, sodium concentrations up to 4 M). The overall stability of the hairpins formed in the homologous DNA fragments, as can be inferred from the melting temperatures summarized in figure 4, merits some discussion. For n = 0–4, i.e. with loops consisting of four to five residues and a reduced number of base pairs in the stem, the hairpins melt (at [Na+] = 0·2M) at approximately the same temperature (ca. 52°C); a further increase in loop length induces a decreasing trend in Tm. From this finding we conclude that the stability of DNA hairpins is at its maximum for loop lengths of four or five residues and declines for larger fragments (n = 5–7). Since the enthalpy contents of the latter fragments (n = 5–7) are identical (within experimental error), the observed decrease in overall stability of the larger hairpins is to be ascribed to differences in the entropy content. This can be rationalized when the hairpin formation is dissected in two steps (Cantor and Schimmel, 1980): the formation of a stable duplex nucleus (nucleation) followed by the formation of base pairs in the stem (chain growth). Since the same duplex stem is formed in all the larger (n 4) fragments, the chain growth in these fragments will involve similar thermodynamical parameters. However, because the hairpins will nucleate around the two G·C base pairs in the stem, the nucleation entropy will change upon increasing the number of nucleotides ( = distance) between the corresponding G and C bases in the base sequence and thus cause the observed decrease in overall stability upon increasing the loop size. The stability of the shorter hairpins (n = 0–3) can be explained by the same line of reasoning: the decrease in the absolute value of ∆Η° noted for these fragments (cf. figure 5) is apparently compensated by the decrease in absolute value of the nucleation entropy so that the overall stabilities in terms of observed Tm’s remain the same throughout the series (n = 0–4).

Structure and stability of DNA versus RNA hairpins Loop stabilities in RNA fragments have long since been determined (cf. e.g. Tinoco et al., 1973) and are nowadays standard textbook material (Cantor and Schimmel, 1980). It is therefore certainly of interest to compare the present results obtained for DNA hairpins with these RNA standards. Although it is not completely without risk to draw generalized conclusions from our (rather limited) dataset, the results obtained for the homologous series d(ATCCTATnTAGGAT), vide supra, clearly indicate that a loop consisting of four to five residues is the most stable one in our DNA fragments. This finding is at variance with the aforementioned data available for RNA fragments, where it was assessed that the stability in RNA hairpins reaches a maximum for loop sizes of six or seven nucleotides. When tracing the origins of differences in conformational behaviour of DNA on the one hand and RNA on the other, the first thing that comes to mind is the distinction in spatial structure for the two classes of nucleic acids: (self-) complementary RNA oligomers adopt an Α-type double helical conformation in solution, while DNA


775

fragments take tip a B-type double helical form (Pardi et al., 1981; Haasnoot et al., 1984). Presuming that the duplex stem in hairpin fragments will adhere to this precept, the difference between the DNA and RNA hairpins is then to be sought in the fact that the loop in DNA hairpins must bridge the gap between the ends of a B-type helix, whereas in RNA fragments the ends of an Α-type helix must be interconnected. At this point we turn to some distance calculations we performed on basis of the wellknown fibre diffraction structures of B-DNA and A-RNA (Arnott et al., 1976). Notably, within a top base pair, the distances betweeen the 3’-end phosphate on the one strand and the 5’-end phosphate of the other strand (which determines the gap the loop has to span) appear to be very similar in both types of helices (~ 18 Å). In other words it is not the interstrand distance in itself that is responsible for the observed differences in conformational behaviour between RNA and DNA hairpins. Therefore, these differences must originate from a more subtle disparity, such as how the loop can bridge the interstrand gap while still making the best of e.g. internucleotide stacking interactions. Following the original suggestion of Rietveld (1984) and Pleij et al. (1984) it can indeed be made plausible that in the case of an A-RNA helix the two strands can be interconnected while preserving a number of (single stranded) stacking interactions. Figure 6 illustrates the point: when for a regular A-RNA double helix the distances between a particular phosphate in one strand (P’ in figure 6) and other phosphates residing in the opposite strand are calculated, it is found that the sampling phosphate (P’) is in the proximity (10–12 Å) of a few phosphates positioned on the other side of the major groove and located five to seven base pairs away in a linear structure notation. This suggests that elongation of the 5’-end of a regular A-RNA stem of a hairpin with five to seven nucleotides (arranged in an Α-type single helical stack) will bring the 5’-end phosphate within a relatively short distance (~ 11 Å) of the 3’-end of the opposite strand. It is conceivable that this remaining 11 Å-gap is easily closed by, say, two nucleotides (which in a normal helical arrangement indeed span a distance of ca. 11 Å!). Thus a seven-membered loop is envisioned in which at least five single strand base–base interactions will contribute to the overall stability of the RNA hairpin fragment. The above rationalization accounts for the optimum stability observed for RNA hairpins with six or seven nucleotides in the loop. Smaller loop sizes will be energetically less favourable (but not impossible, see e.g. Heus et al., 1983) for several reasons: (a) less base–base stacking interactions will stabilize the loop, (b) the distance to be bridged by the two "closing" nucleotides is larger (or even more stacking interactions are to be sacrificed), (c) steric hindrance between e.g. the bulky base-moieties may occur. On the other hand, the expected increase in the probability of loop initiation with decreasing chain (loop) length (i.e. decrease in nucleation entropy) may in part counterbalance these negative effects. The larger (n > 7) loop sizes are not envisaged to suffer greatly from the effects mentioned above under (a)–(c). Obviously, it is the nucleation entropy that diminishes the overall stability of the longer (n > 7) RNA hairpins. Let us now see whether the stability of DNA hairpins can be rationalized along the lines pointed out above for RNA hairpins. The interstrand phosphate-phosphate distance calculations performed for the canonical B-DNA helix (cf. figure 7) show indeed an analogous but different situation: the shortest distance (~ 11 Å) is now encountered on the other side, spanning the minor groove and located only two to three

776

Haasnoot et al.

Figure 6. Distances between a sample phosphate (P’) in one strand and phosphates in the opposite strand in an A-RNA helix. The numbering of the phosphates in the opposite strand is indicated in the two-dimensional representation above the curve. The distances were calculated on basis of the coordinates published by Arnott et al. (1976) for a regular A-RNA helix.

base pairs away in a linear structure notation. In other words, elongating the 3’-end of the regular B-type duplex stem of the DNA hairpin with two to three nucleotides (arranged in a regular B-type single helical stack) will bring the 3’-end phosphate within approximately 11 Å of the 5’-end of the opposite strand. Again, it is presumed that this residual 11 Å-gap can be bridged by two nucleotides. Following a similar line of


777

Figure 7. Distances between a sample phosphate (P’) in one strand and phosphates in the opposite strand in a B-DNA helix. The numbering of the phosphates in the opposite strand is indicated in the two-dimensional representation above the curve. The distances were calculated on basis of the coordinates published by Arnott et al. (1976) for a regular B-RNA helix.

reasoning as given for the RNA hairpins (vide supra) it may thus be argued that a fourto five-membered loop will be the most stable one in DNA hairpin fragments (which is in perfect agreement with the experimental data, cf. the preceding section). For the sake of completeness, we note that the longer (n > 5) loops will not necessarily have to bridge the larger distances that follow from figure 7: e.g. a seven-membered loop may adapt its "apparent" length by stacking one nucleotide on the 3’-end as well as on the 5’-end of the duplex stem, the remaining 5 out of 7 residues may then form a loop as described above. At this point one may wonder whether there is experimental evidence to support the

778

Haasnoot et al.

structural interpretation of RNA/DNA hairpin stabilities presented above. The answer to this question is positive in the sense that there are indeed indications that the loops described above occur. Α classic example of an unperturbed seven-membered RNA loop is set by the anticodonloop in tRNA "Phe" from yeast. In the latter molecule the Α(31)·ψ(39) base pair (located at the bottom of the anticodon stem) is spanned by the ribonucleotide fragment r(-CmUGmAAYA-) which interconnects the 3’-side of A(31) with the 5’-side of ψ(39). The crystallographic data available for this molecule show that the sequence r(-GmAAYA-), which includes the anticodon triplet r(-GmAA-), is stacked in a regular Α-type single helical fashion upon ψ(39) and that the remaining gap between A(31) and Gm(34) is closed by the remaining r(-CmU-) dinucleotide. This conformation fits the description given above for the seven-membered loop in RNA hairpins and hence endorses the feasibility of such loops. We note in passing that this so-called 3’-stacked conformation of the anticodon loop is not confined to the solid phase. Recently, Clore et al. (1984) found evidence that this conformation may be preserved in solution. Up to now no crystal structure determination has been reported that might serve as a model for a loop structure in DNA fragments. However, we will show that some telltale information about DNA loop structures can also be extracted from a two-dimensional NMR study of the hexadecamer d(ATCCTAT4TAGGAT). In order to do so we turn to figure 8 where a contour plot of a part of the 2D NOESY spectrum of this compound is presented. In this particular region of the spectrum the NOE cross peaks that link the aromatic base protons and the l’-sugar protons are found. As was pointed out elsewhere (Scheek et al., 1983; Haasnoot et al., 1983b) the latter cross peaks can be used for a sequential assignment of the base- and 1’-protons involved when the oligonucleotide at issue is in a (double) helical conformation. Starting from the intranucleotide NOE cross peak that interconnects H8 to H1’ in the 5’-residue A1, the sequential assignment is readily performed – as expected – up to residue A6 at the end of the stem (cf. figure 8, solid line). Notably, the sequential assignment can be pursued (cf. figure 8, dotted line) for two more residues, i.e. L1 and L2 (denoting the two thymidines at the 5’-side of the loop), which may be taken for an indication that the latter two residues stack in a more or less regular fashion upon the A6 of the hairpin stem. At this point the assignment method breaks down as no cross peaks can be detected that interconnect Η1’ of L2 to H6 of L3. Alternatively, we can start at the 3’-end (T1) of the hexadecamer and repeat the procedure (cf. figure 8, dashed line): again the analysis stops just before L3 is reached. The present results are in correspondence with the above proposed loop structure for DNA hairpins: the 3’-end of the hairpin stem (A6) is elongated by two residues (L1 and L2) and the remaining gap over what can be considered to be the minor groove is closed by the thymidine residues L3 and L4. In summary, the structure of an unperturbed hairpin loop – unperturbed in the sense that no base pair formation between loop bases and other bases elsewhere in the molecule takes place (e.g. D-loop/T-loop interaction in tRNA)-is dictated by the geometrical constraints imposed by the A-RNA or B-DNA helix. The relative stability of homologous RNA and DNA fragments as a function of loop length can therefore be explained on basis of the helix geometry of the hairpin stem.


779

Figure 8. Part of a 500 MHz NOESY spectrum of d(ATCCTAT4TAGGAT) showing the cross peaks that manifest the NOE connectivities between the aromatic base protons (7·0–8·4 ppm) and the Hl’-deoxyribose protons (5·4–6·4 ppm). Solid lines indicate the sequential assignment procedure in the d(ATCCTA-) part of the molecule; dotted lines idem for the loop fragment d(-TTTT-); dashed lines idem for the d(-TAGGAT) part of the hairpin stem.

Acknowledgements This work was supported by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO). NMR spectra were recorded at the Dutch National 500/200 MHz hf-NMR Facility at Nijmegen (The Netherlands). P. A. W. van Dael, W. Guijt and J. Joordens are thanked for expert technical assistance; the experimental assistance of the students K.T.F. Janssen, G. Janssen and T. J. J. Binnendijk is gratefully acknowledged. Drs. C. W. A. Pleij, K. Rietveld and L. Bosch initiated ideas about loop folding which helped shape the developments of the structural proposals in the present paper. Finally, Dr. E. Westhof (Institut de Biologie Moleculaire et Cellulaire du

780

Haasnoot et al.

C.N.R. S., Strasbourg, France) is gratefully acknowledged for showing us around on the Α-type and B-type double helices using the Evans and Sutherland multi-picture system. References Arnott, S., Campbell Smith, P. J. and Chandrasekaran, R. (1976) in CRC Handbook of Biochemistry and Molecular Biology; Nucleic Acids, (ed. G. D. Fasnar), Vol. 2, p. 411. Cantor, C. R. and Schimmel, P. R. (1980) in Biophysical Chemistry. III. The behavior of biological macromolecules, (San Francisco: Freeman), Chapter 23. Clore, G. M., Gronenborn, A. M, Piper, Ε. Α., McLaughlin, L. W., Graeser, E. and van Boom, J. Η. (1984) Biochem. J., 221, 737. Dickerson, R. E. (1983) J. Mol. Biol., 166, 419. Elson, Ε. L., Scheffler, I. E. and Baldwin, R. L. (1970) J. Mol. Biol., 54, 401. Gralla, J. and Crothers, D. M. (1973) J. Mol. Biol., 78, 301. Haasnoot, C. A. G., den Hartog, J. Η. J., de Rooij, J. F. Μ., van Boom, J. Η. and Altona, C. (1979) Nature (London) 281, 235. Haasnoot, C. A. G., den Hartog, J. H. J., de Rooij, J. F. Μ., van Boom, J. Η. and Altona, C. (1980) Nucleic Acids Res., 8, 169. Haasnoot, C. A. G., de Bruin, S. Η., Berendsen, R. G., Janssen, Η. G. J. M., Binnendijk, Τ. J. J., Hilbers, C. W., van der Marel, G. A. and van Boom, J. Η. (1983a) J. Biomol. Struct. Dyn., 1, 115. Haasnoot, C. A. G., Westerink, Η. P., van der Marel, G. A. and van Boom, J. Η. (1983b) J. Biomol. Struct. Dyn., 1, 131. Haasnoot, C. A. G., Westerink, Η. P., van der Marel, G. A. and van Boom, J. Η. (1984) J. Biomol. Struct. Dyn., 2, 345. Heus, Η. Α., van Kimmenade, J. Μ. Α., Knippenberg, Ρ. Η. van, Haasnoot, C. A. G., Bruin, S. H. de and Hilbers, C. W. (1983) J. Mol. Biol., 170, 939. Marky, L. A. and Breslauer, Κ. J. (1982) Biopolymers, 21, 2185. Marky, L. Α., Blumenfeld, Κ. S., Kozlowski, S. and Breslauer, K. J. (1983) Biopolymers, 22, 1247. Pardi, Α., Martin, F. Η. and Tinoco, I. (1981) Biochemistry, 20, 3986. Pleij, C. W. Α., Rietveld, K. and Bosch, L. (1984) Nucleic Acids Res., (in press). Rietveld, K. (1984) Three-dimensional folding of the tRNA-like structures of some plant RNAs, Ph.D. Thesis, University of Leiden (The Netherlands). Scheek, R. M., Russo, N., Boelens, R., Kaptein, R. and van Boom, J. H. (1983) J. Am. Chem. Soc., 105, 2914. Tibanyenda, N., de Bruin, S. H., Haasnoot, C. A. G., van der Marel, G. Α., van Boom, J. H. and Hilbers, C. W. (1984) Eur. J. Biochem., 139, 19. Tinoco, I., Borer, P. N., Dengler, B., Levine, M. D., Uhlenbeck, O. C., Crothers, D. M. and Gralla, J. (1973) Nature New Biol., 246, 40. Wang, A. H.-J., Quigley, G. J., Kolpak, F. J., Crawford, J. C., van Boom, J. H., van der Marel, G. A. and Rich, A. (1979) Nature (London), 282, 680.