Skip to main content
NCBI home page
Nucleic Acids Research logoLink to Nucleic Acids Research
. 2002 Dec 1;30(23):5142–5150. doi: 10.1093/nar/gkf637

Sequence-dependent folding of DNA three-way junctions

René Assenberg 1,2,a, Anthony Weston 1, Don L N Cardy 1, Keith R Fox 2,b
PMCID: PMC137952  PMID: 12466538

Abstract

Three-way DNA junctions can adopt several different conformers, which differ in the coaxial stacking of the arms. These structural variants are often dominated by one conformer, which is determined by the DNA sequence. In this study we have compared several three-way DNA junctions in order to assess how the arrangement of bases around the branch point affects the conformer distribution. The results show that rearranging the different arms, while retaining their base sequences, can affect the conformer distribution. In some instances this generates a structure that appears to contain parallel coaxially stacked helices rather than the usual anti-parallel arrangement. Although the conformer equilibrium can be affected by the order of purines and pyrimidines around the branch point, this is not sufficient to predict the conformer distribution. We find that the folding of three-way junctions can be separated into two groups of dinucleotide steps. These two groups show distinctive stacking properties in B-DNA, suggesting there is a correlation between B-DNA stacking and coaxial stacking in DNA junctions.

INTRODUCTION

DNA rearrangements play an important role in processes such as genetic recombination and DNA damage repair and may involve the exchange of DNA segments between homologous chromosomes. These interactions often generate unusual DNA structures such as three- and four-way (Holliday) junctions. Although crystal structures have recently been determined for four-way junctions and their interaction with proteins (1–6) many aspects of their folding are still unclear. Other techniques that have been used to investigate four-way junctions include NMR, FRET and electrophoretic mobility (reviewed in 7,8). Three-way junctions fold in a similar manner, but are less complex as they only involve the coaxial stacking of two helices, whereas two pairs of arms form a four-way junction. Three-way junctions therefore provide a less complicated system for assessing the features of junction folding.

A folding scheme for three-way junctions is shown in Figure 1A. In contrast to four-way junctions, the arms of simple three-way junctions assume a Y-shaped or pyramidal conformation, even in the presence of divalent cations (9). However, addition of unpaired bases at the branch point promotes coaxial stacking of the helical arms in the presence of divalent cations (9–12). Two different conformers can be generated in which two arms are engaged in stacking interactions, leaving the third arm at an angle. Although these conformers are in dynamic equilibrium this is usually dominated by one conformer (13–16). These will be referred to as the A/C stacked and A/B stacked conformers, based on the nomenclature of Altona (17) and are illustrated in Figure 1A. The existence of these anti-parallel stacked conformers has been verified by electrophoresis (12,13,18–21) and NMR (16,22–25). In the absence of divalent cations the junctions adopt an ‘unstacked’ or ‘extended’ form (12,13).

Figure 1.

Figure 1

Open in a new tab

(A) Schematic representation of the folding of three-way junctions. The junction is formed from three oligonucleotides labelled 1, 2 and 3 in which 2 contains the unpaired bases at the junction. These generate arms A, B and C. The upper part shows the unstacked conformation, while the lower two illustrate the coaxial stacking of arms A and C (conformer A/C) or arms A and B (A/B conformer). The insets show the electrophoresis patterns expected after restriction cleavage of each of the arms in turn in which linear fragments have the greatest mobility. (B) Sequence of the three-way junctions used in this work. The locations of the restriction enzyme sites in each arm are indicated. The various junctions differ in the sequence of the six base pairs close to the junction (boxed). The loop indicates the position of the two unpaired bases.

One of the most widely used techniques for assessing the conformer adopted by a three-way junction is the long–short arm assay, first developed by Cooper and Hagerman (26). In this assay each of the arms is shortened in turn by digestion with an appropriate restriction enzyme and the structure is determined by comparing the electrophoretic mobility of the resulting fragments. A fragment in which the two intact arms are at an acute angle will migrate more slowly than one in which the arms are coaxially stacked (Fig. 1A). By comparing the mobility of three fragments, each containing a different shortened arm, it is possible to deduce the conformer preferred by the junction.

There have been several studies on the factors that control the conformational bias of a particular junction. The base composition of the unpaired bases at the branch point is not important (12,22,27), though this may affect the overall stability of a particular junction (13). Likewise, the specific cation is not important, though its valency can affect the concentration required to induce coaxial stacking (13,28). The most important factor affecting conformer selection appears to be the DNA sequence around the branch point and junctions that differ by only a single base pair can adopt different conformers (12,13,16,18,20,23,25). Van Buuren et al. (25) have postulated that the nature of the second base (purine or pyrimidine) in strand 1 of arm A has a dominant effect on conformer selection.

In this work we have investigated the sequence-dependent folding of three-way junctions by comparing a number of closely related junctions using the long–short arm assay (26). We have introduced small sequence-changes close to the branch point in order to assess the importance of purine– pyrimidine stacking interactions and the effect of strand orientation on the conformer distribution. We find that the coaxial stacking of three-way junctions is related to the stacking properties of dinucleotide steps in B-DNA.

MATERIALS AND METHODS

DNA junctions

All DNA oligonucleotides were purchased from Oswel DNA service (Southampton, UK) with HPLC purification. For the preparation of each junction 100 pmol of each oligonucleotide was added to 50 µl of NEB buffer 2 (50 mM Tris–HCl, pH 7.9 containing 50 mM NaCl, 10 mM MgCl2 and 1 mM DTT) in a 200 µl PCR tube. The mixture was placed in a PTC-200 thermal cycler (MJ Research Inc.) and incubated for 25 min at 95°C. The oligonucleotides were then annealed by lowering the temperature to 20°C at a rate of –0.1°C/s. These mixtures were either subjected to further digestion reactions or stored at –20°C. All junctions were examined on non-denaturing gels to confirm the presence of a single band before adding the restriction enzymes. The sequences of the junctions used in this work are shown in Figure 1B.

Long–short arm assay

The protocol used for determining the conformer preferred by each three-way junction was similar to that of Duckett and Lilley (29). Each arm of the junction contained a unique site for a restriction endonuclease which produces blunt ends, with the cleavage site located 9 bp away from the branch point. Restriction enzymes producing staggered cleavage were not used, as single-stranded overhangs have been shown to affect the gel mobility (21). The arms of each junction were shortened in turn by restriction enzyme cleavage. All the junctions used in this study (Fig. 1B) contained an EcoRV site in the B arm, a DraI site in the A arm and a PvuII site in the C arm. Each of these enzymes cleaved 9 bp away from the branch point. A two-base bulge was also included in strand 2 to facilitate coaxial stacking. The identity of the bases in the bulge was varied so as to minimise any alternative pairings with the bases flanking the branch point.

Restriction enzyme digestion

A 7.5 µl aliquot of each three-way junction solution (prepared as described above) was digested with 10–20 U of each restriction enzyme for 3 h at 37°C in NEB buffer 2 in a total volume of 20 µl. The reaction was stopped by adding 10 µl of TBM loading buffer [89 mM Tris pH 7.9, 89 mM boric acid, 5 mM MgCl2, 0.1% (w/v) bromophenol blue, 20% (w/v) ficoll].

Gel electrophoresis

The products of restriction enzyme digestion (3.5 µl) were resolved on a 12% (w/v) polyacrylamide gel (Accugel, National Diagnostics Inc.) and run on a Novex mini-gel cassette. Gels were prepared in either TBM (89 mM Tris pH 7.9, 89 mM boric acid, 5 mM MgCl2) or 25 mM tricine, 25 mM triethanolamine pH 7.5 containing 5 mM magnesium chloride. We found no significant differences between the results obtained in these two buffers. Gels were run for 3–4 h at room temperature at 90 V with the cassette fully immersed in running buffer on both sides. After running, the gel was stained with ethidium bromide (1.5 µg/ml) for 15 min. The gel was placed on a transilluminator and photographed using a Kodak Polaroid DS-34 camera with an ethidium bromide specific filter (Elchrom Scientific).

RESULTS

The folding of different three-way junctions was assessed by the long–short assay, measuring the relative mobility after digestion with appropriate restriction enzymes. It is worth noting that this assay yields an averaged conformation (7,13) since gel electrophoresis step takes several hours.

Swapping the central bases on the continuous and exchanging strands

It is likely that the conformer preferred by a three-way junction is determined by stacking between the base pairs that surround the junction. Swapping these pairs between the different arms should therefore lead to predictable changes in conformer distribution. This is illustrated in Figure 2. Figure 2A shows the extended form of three-way junction J1V1, which has previously been shown to prefer an A/C stacked conformer (13), illustrated in Figure 2B. In this the stacking preference is CG·TC/GACG rather than CGTG/CA·CG, which would generate an A/B conformer (· indicates the discontinuity at the strand exchange). Swapping the two base pairs close to the junction in arm B with those in arm C produces the junction shown in Figure 2C. This has the same potential base stacking combinations as the junction shown in Figure 2B, except for the location of the strand discontinuity. If conformer distribution is solely determined by the base stacking then we would expect this new junction to favour an A/B conformer as shown in Figure 2D, as this retains the stacked sequence 5′-CGTC/GA·CG. If this new junction were to adopt an A/C conformer this would generate the stacked sequence 5′-CG·TG/CACG, equivalent to the minor conformer shown for junction J1V1.

Figure 2.

Figure 2

Open in a new tab

Schematic representation of exchanging the central bases between different strands. (A) shows the extended form of PJ71A2 (J1V1), which has been shown to adopt an A/C conformer as shown in (B). The central bases are exchanged as indicated generating junction JV70A2 (C). The A/B conformer of JV70A2 is illustrated in (D).

We applied this base-swapping approach to four pairs of junctions and the results are shown in Figure 3. Figure 3A shows results for the junctions illustrated in Figure 2. PJ71A2 is equivalent to J1V1, which has previously been shown to adopt an A/C conformer. As expected, the AC fragment has the fastest mobility and AB has the slowest. This is consistent with a model in which arms A and C are coaxially stacked, with a small angle between arms A and B. However, the isomer of this junction (PJ70A2) shows a different pattern in which fragments AC and BC have similar mobilities and migrate faster than AB. The slower mobility of AB suggests that this fragment has the smallest angle between the arms, which would normally indicate an AC conformer. However, in this instance the angle between arms BC and AC appears to be similar. Although the structure of this conformer is not obvious, it is clear that this does not correspond to the predicted A/B conformer.

Figure 3.

Figure 3

Open in a new tab

Long–short arm assay for pairs of junction in which the central base pairs have been exchanged between different strands. In each case, fragment AC was generated by cutting arm B with EcoRV, BC was generated by cutting arm A with DraI and AB was generated by cutting the C arm with PvuII. The sequences of the central bases are shown above each junction.

Figure 3B shows similar experiments with another pair of junctions. PJ73C2 corresponds to J3CC (22,27) and has previously been shown to adopt an A/C stacked conformer, which is confirmed by the banding pattern. Swapping the bases at the junction generates junction PJ72C2, which shows a pattern in which fragment AB has the fastest mobility and AC the slowest, consistent with the predicted A/B conformer. However, it should be noted that fragments AB and BC are not as well separated as usual.

Figure 3C and D show the results of similar experiments with two more pairs of junctions. For PJ64C2 (Fig. 3C) the small mobility differences suggest that the conformers are in dynamic equilibrium, though the lower mobility of the A/C fragment suggests a slight preference for the A/B stacked conformer. Swapping the base pairs so as to yield junction PJ67G2 produces a pattern in which fragment BC has the slowest mobility, and therefore cannot correspond to either A/B or A/C anti-parallel conformers. This pattern shows that the smallest angle is between the B and C arms, suggesting that it adopts a parallel A/C conformer. This is the first time that this stacking mode has been proposed for an isolated DNA junction, though it has been observed for an RNA three-way junction (30) and a DNA junction as part of a pseudosquare knot (31). The results for PJ65C2 and PJ68G2 (Fig. 3D) are similar to those with PJ64C2 and PJ67G2. PJ65C2 shows a similar pattern to PJ64C2 and is most like an A/B conformer, while PJ68G2 resembles PJ67G2 for which we have suggested a parallel A/C stacked conformer.

These results show that swapping the base pairs around the junction (i.e. altering the strand on which a given sequence is located) does not always preserve the coaxial stacking preference. Figure 3B–D shows that, for these junctions, the fragment with the fastest mobility changes from AC to AB on making the switch, while the relative mobility of fragment BC is different in each case. In contrast, Figure 3A shows junctions for which fragment AB has the slowest mobility in each case. Therefore, it appears that the conformer distribution can be affected by which strand is discontinuous and which is continuous. These differences will be considered further in the Discussion.

Does the order of purines and pyrimidines around the junction determine conformer preference?

Previous studies have suggested that the conformer distribution is largely determined by the order of purines and pyrimidines around the junction, rather than the identity of the individual bases (25,32). We have tested this hypothesis using several sets of junctions that contain different R/Y sequences flanking the branch point.

Figure 4A shows results of the long–short arm assay for junctions in which the coaxial stacking is 5′-RYYY/RR·RY for conformer A/B or 5′-RY·RR/YYRY for conformer A/C (PJ61, 62 and 63 each contain two cytosines as the bulge in strand 2). The sequences around the junction are GC-rich for PJ63C2, but AT-rich for PJ62C2. The gel shows that all three junctions produce the same pattern in which fragment AB has the fastest mobility and AC has the slowest. This corresponds to an A/B coaxially stacked conformer. Hence, for these junctions it appears that changing the base sequence while retaining the order of purines and pyrimidines does not affect the conformer preference.

Figure 4.

Figure 4

Open in a new tab

Long–short arm assay for junctions containing different arrangements of purines and pyrimidines around the junction. In each case fragment AC was generated by cutting arm B with EcoRV, BC was generated by cutting arm A with DraI and AB was generated by cutting the C arm with PvuII. The sequences of the central bases are shown above each junction.

Figure 4B shows the results for a similar series of four junctions in which the base pairs close to the junction in arm A are changed from 5′-RY/RY (PJ61, 62, 63) to YR/YR (PJ40, 44, 45, 69). In this case the mobility patterns are not identical and it appears that the specific DNA sequence affects the conformational equilibrium. PJ40C2 shows a pattern suggestive of an A/B stacked conformer. However, changing all the base pairs to the alternative purine or pyrimidine, yielding PJ44C2, leads to a pattern in which the three fragments have similar mobilities, with BC migrating slightly faster. This is suggestive of rapid interconversion between the conformers. When the A-T base pair directly flanking the branch point in PJ44C2 is substituted for a G-C pair (PJ45T2) there is a subtle change in the equilibrium and fragment AB now has the slowest mobility, suggesting a slight preference for the A/C stacked conformer. Changing the second base pair from the branch point in arms A and B of PJ45T2 to the alternative purine/pyrimidine, yielding PJ69T2, appears to increase the preference for the A/C stacked conformer. This series of junctions shows a gradual change in the conformer distribution through small sequence changes while retaining the order of purines and pyrimidines. Thus, for this set of junctions, the specific base sequence is clearly important for determining the conformer bias.

Figure 4C shows the results for a series of junctions in which the stacking choice is between YY·RY/RYRR (conformer A/C) and YYYY/RR·RR (conformer A/B). PJ46T2 shows a clear preference for the A/C stacked conformer. However, changing the CG pair flanking the branch point in arm C to TA (PJ81T2) alters the pattern so that the three bands have very similar mobilities. Changing the first base pairs in arms A and C of PJ81T2, yielding PJ54C2, produces a junction with a clear preference for the A/B conformer. Comparing PJ46T2 with PJ54C2, which have identical sequences at the second base from the junction but which produce opposite stacking patterns, suggests the base pairs directly flanking the branch point contribute most to the conformational bias. We have tested the effect of changing the identity of the base pairs at this second position of PJ54C2 by examining the behaviour of junctions PJ41C2, PJ47–49C2 and PJ51–54C2. PJ53, 52 and 51 change the base pair at the second position in arms B, C and A of PJ54, respectively. PJ49 and PJ48 are single base changes from PJ51, while PJ47 and PJ41 are single base changes from PJ52 and PJ49, respectively. Each of these retains the same base pairs at the junction, so that the stacking choice is between T·G/CA (A/C conformer) or TT/A·A (conformer A/B). It can be seen that all these junctions produce the same mobility pattern in which fragment AB has the fastest and fragment AC the slowest mobility. All these junctions therefore prefer the A/B conformer. These junctions possess the same central bases as PJ61C2, which also prefers an A/B conformer (Fig. 4A). These results suggest that the first base pairs in each arm of this series have a dominant effect on the conformational preference.

Two further junctions were therefore designed to test the effect of these central bases. Junction PJ55C2 (Fig. 4D) contains the same central base pairs as those in Figure 4C, but with a different arrangement of purines and pyrimidines in the surrounding bases. The arms of this junction could stack to form GTTA/TA·AC (conformer A/B) or GT·GA/TCAC (conformer A/C). This junction again showed a fragment pattern that is consistent with the preference for an A/B conformer.

Junction PJ36C2 also has the same central core, but differs at only a single base pair in one arm from PJ61C2 and PJ53C2. PJ36C2 has the stacking choices TTTT/AA·AA (conformer A/B) or TT·GG/CCAA (conformer A/C). This junction, with a different arrangement of purines and pyrimidines at the second position compared to the other junctions shown in Figure 4D, shows a different pattern with small differences in mobility between the fragments. Fragment AC still has the slowest mobility, but the preference for the A/B conformer is reduced. Therefore, it appears that although the sequence of the central bases has a strong effect on conformer distribution, this can be affected by changes in the flanking sequences. A similar effect can be seen by comparing PJ62C2 (Fig. 4A) with PJ81T2 (Fig. 4C). These junctions have the same central core, which is flanked by different bases, and show a clear difference in their stacking preferences. A similar, though less pronounced effect, can also be seen by comparing PJ45T2 with PJ69T2 (Fig. 4B) and PJ70C2 with PJ71A2 (Fig. 3A).

DISCUSSION

Sequence arrangement around the junction

We have shown that swapping base pairs between different arms, while retaining the base sequences, can affect the preference for a particular coaxially stacked conformer. For the pair of junctions PJ72C2 and PJ73C2, the conformer preference is dictated by base stacking alone and AC/G·T and A·C/GT are preferred over A·G/CT and AG/C·T. However, for the other junctions shown in Figure 3, altering the relative positions of the two strands changes the stacking preference. In these experiments the gel patterns do not always correspond to three bands of different mobilities (see below). The pair of junctions PJ70A2 and PJ71A2 contains the same central base pairs closest to the junction and the stacking choice is between G·T/AC and GT/A·C. In each case the AB fragment has the slowest mobility suggesting that the stacking pattern in which the GT step switches between strands (AC in the same strand) is preferred over the same strand GT. The results with the pairs of junctions PJ64C2/PJ67G2 and PJ65C2/PJ68G2 are less easily interpreted as the electrophoresis assay does not show the usual pattern of three bands with different mobilities. However, it is clear that changing the order of the strands has affected the stacking preferences and the conformer distribution.

Examination of other published three-way junctions suggests that the conformer adopted by J1V4 (13) may also have been affected by the nature of the exchanging strand. For this junction, both conformers would produce two CA/TG steps across the junction. In the A/C conformer the C and A are from the exchanging strands, while in the A/B conformer these are from the continuous strand. Since J1V4 adopts the A/C conformer it appears that AC/G·T is again preferred over A·C/GT.

The observation that the relative positioning of the individual bases (i.e. on the continuous or exchanging strands) affects the stacking preference means that this needs to be considered when attempting to identify preferred stacking patterns and conformer distributions.

The role of base geometry on conformer selection (purines or pyrimidines)

Figure 4 shows that in some instances the specific base sequence, and not just the order of purines and pyrimidines, affects the conformer preference. This may not be surprising as, although some structural features of duplex DNA are determined by the order of purines and pyrimidines (33), the local conformation depends on the exact base sequence. The observation that the R/Y sequence alone is not sufficient for predicting conformer selection also argues against the pyrimidine rule (25). This predicts A/C stacking when the second base in strand 1 of arm A is a purine, and A/B stacking if this base is a pyrimidine. The most convincing evidence against the accuracy of the pyrimidine rule, however, comes from the related junctions PJ36C2, PJ41, PJ47–49 and PJ51–55 and junctions YTT (20) and PJ73C2. Each of these showed a conformational preference opposite to those predicted by the pyrimidine rule (Table 1). Moreover, our results for PJ46T2, PJ81 and PJ54 showed that changes at the first position can alter the conformer preference.

Table 1. Three-way junction sequences and conformational preferences.

Junction Conformer A/C A/B ΔProp ΔDSF ΔTSF Reference
TWJ-TC A/C GCCA TGCG –2.9 3.02 0.7 23
J1 A/C GACG CGAG –0.4 –0.71 4.7 13
J1V1 A/C GACG CGTG 0 0 13.2 13
J1V2 A/C GACG CGCG 2 5.23 7.3 13
J1V3 A/C GAGG CCAG 0.7 6.23 4.5 13
J1V4 A/C GTGG CCAG 0 0 –1.2 13
J1V5 A/C GGGG CCAG –1.4 6.01 7.3 13
PJ16C2 A/C GGAG CTGC 3.9 8.93 6.1 21
PJ42A2 A/C CAGG CCGT 0.2 3.56 3.5 Unpublished results
PJ46T2 A/C GTGA TCCC 1.4 –6.01 –3.9 This work
PJ66G2 A/C CAAA AGTG 5.7 4.15 19.5 This work
PJ69T2 A/C TTCG CGCC 2.4 5.94 9.5 This work
PJ73C2 A/C CGTC GAGC 2.8 1.99 5.4 This work
TWJ1 A/B ACCC GGCG 2.9 –3.02 –9.7 16
TWJ2 A/B ACGC GCCG –1.9 3.34 2.8 25
TWJ4 A/B ACCT AGCG 2.9 –3.02 –6 25
TWJ6 A/B GCCC GGCG 2.9 –3.02 –11 25
J3CC A/B GCTC GACG 2.8 1.99 5.4 22
YTT A/B GTGA TCTG 0.7 6.23 3.6 20
PJ40C2 A/B TCCG CGTT 4.9 2.21 7 This work
PJ41C2 A/B ACAG CTTT 9.2 12.37 13.3 This work
PJ47C2 A/B ACAA TTTT 9.2 12.37 14.5 This work
PJ48C2 A/B ACAG CTTC 9.2 12.37 13.7 This work
PJ49C2 A/B GCAG CTTT 9.2 12.37 17.3 This work
PJ51C2 A/B GCAG CTTC 9.2 12.37 17.7 This work
PJ52C2 A/B ACAA TTTC 9.2 12.37 10.0 This work
PJ53C2 A/B GCAA TTTT 9.2 12.37 15.1 This work
PJ54C2 A/B GCAA TTTC 9.2 12.37 10.6 This work
PJ55C2 A/B TCAC GTTA 9.2 12.37 10.8 This work
PJ61C2 A/B CCAC GTTT 9.2 12.37 19.3 This work
PJ62C2 A/B TTGC GCTT 0.7 6.23 5.6 This work
PJ63C2 A/B CCAT ATCC 3.9 8.93 –0.3 This work
PJ65C2 A/B AACA GTTT 5.7 4.15 18 This work
PJ67G2 A/C GTTC GAGA 8.5 6.14 –1.2 This work
PJ68G2 A/C TGTT AAAC –5.7 –4.15 –18 This work
J1V9 EQ GCTG CACG –2.8 –1.99 13 13
J1V6 EQ A/C GTTG CACG 5.7 4.15 13.4 13
PJ45T2 EQ A/C CTCA TGCC 2.4 5.94 –3.2 This work
PJ70A2 EQ A/C CACG CGTC 0 0 –13.2 This work
TWJ3 EQ A/C ACCG CGCC –2.9 3.02 6.9 25
TWJ5 EQ A/C ACCA TGCG –2.9 3.02 1 25
TWJ7 EQ A/C GCCG CGCG –2.9 3.02 4.5 25
PJ36C2 EQ A/B CCAA TTTT 9.2 12.37 10.7 This work
PJ44C2 EQ A/B CTTA TACC –5.7 –4.15 –11 This work
PJ64C2 EQ A/B GAAC TCTC –8.5 –6.14 1.2 This work
Open in a new tab

The preferred conformer is indicated as A/C, A/B, EQ (equilibrium with equal mobility of AB and AC fragments), EQ A/B and EQ A/C (equilibrium biased slightly towards the A/B and A/C conformer, respectively). The tetranucleotide sequences for the continuous strands of the two potential conformers are shown. The Δpropeller twist (Δprop), Δdinucleotide slide flexibility (ΔDSF) and Δtetranucleotide flexibility (ΔTSF) values were calculated by subtracting the corresponding values of the major conformer from those of the minor conformer, using data from Packer et al. (35,40). Propeller twist values are angles, whereas slide flexibility is an energy value (i.e. higher values signify lower slide flexibility).

Dynamic folding properties of three-way junctions

Figures 3 and 4 show that some junctions do not produce the expected banding pattern for A/B or A/C stacked conformers, and show patterns in which two or three of the fragments possess very similar mobilities. These unusual patterns could arise from unusual stacking combinations (such as the parallel conformer suggested for PJ67G2 and PJ68G2) or because the conformers are in a relatively rapid dynamic equilibrium yielding an averaged structure (7). In the latter case the relative mobility of the fragments will also be affected by the angle between the three arms, as illustrated in Figure 5. If fragment BC has a mobility that is roughly intermediate to AC and AB then dynamic equilibration will produce three fragments with very similar mobilities as shown in Figure 5A. This argument has previously been used to explain the mobility pattern of junction J1V9 (13). However, the pattern may be different if the angle between the stacked and unstacked helices is very small. This is illustrated in Figure 5B for the extreme case where the unstacked arm is collinear with the stacked arm. If the two conformers are stable, but interconvert rapidly compared to the rate of gel migration, the pattern may show fastest mobility for fragment BC. The pattern in Figure 5B is similar to that expected for the extended (unstacked) form. The long–short arm assay therefore does not discriminate between the extended form (where both A/C and B/C conformers are relatively unstable) and the form in which two stable conformers with acute angles between the stacked and the unstacked arm are in equilibrium (i.e. the two conformers have similar stabilities, both of which are more stable that the unfolded form). Junctions PJ44C2, PJ45T2, PJ81T2, PJ36C2 and J1V6 (13) could therefore be either unstacked or in this form of dynamic equilibrium. This may also help explain the pattern for PJ36C2 (Fig. 4D). This junction is very similar to PJ47C2 and PJ53C2 and differs only in the second base of arm C, which is GC, AT or CG. Since PJ47C2 and PJ53C2 adopt the same conformer (A/B) it seems unlikely that the altered pattern of PJ36C2 reflects a decrease in stability of this conformer. It is more likely that this unusual pattern is due to an increase in the stability of the other (A/C) conformer arising from changes at the second position in arm C.

Figure 5.

Figure 5

Figure 5

Open in a new tab

(A and B) Schematic representation of the effect of conformer equilibrium on gel mobility patterns. In (A) the mobility of the fragment BC is intermediate to AC and AB, so that when these are in dynamic equilibrium all three fragments appear to have the same mobility. In (B) the situation when the unstacked arm is collinear with the stacked arm is shown. In this case fragment BC will have a similar mobility to the coaxially stacked fragment (AC or AB). Further details are explained in the text. (C) Summary of the preferred stacking preferences for three-way junctions; if N6N1/N2N5 is a group 1 step and N2N3/N4N1 is group 2 then an A/C conformer will be preferred, while if N2N3/N4N1 is group 1 and N6N1/N2N5 is a group 2 then an A/B conformer will be preferred (group 1 contains AA/TT, AC/GT, AG/CT and GA/TC, while group 2 contains GG/CC, GC/GC, CG/CG and CA/TG).

Correlation of conformer preferences with DNA structural features

The preferred conformers have now been determined for a large number of three-way junctions and are listed in Table 1, and these allow some suggestions for the rules governing conformer distribution. Several of the junctions shown in Figure 4C favour stacking of AA/TT over CA/TG. Examination of the data for all junctions studied to date shows that many disfavour stacking the CA/TG step. The exceptions are J1V4 and PJ46. These two dinucleotide steps are interesting since they are known to exert different effects on duplex DNA. AA/TT steps are generally much more rigid than CA/TG steps. The latter often has unusually high positive slide values whereas the former has high values for propeller twist values (33–35). This raises the possibility that slide and/or propeller twist are important parameters in setting the conformer bias.

Close examination of the junctions shown in Table 1 suggests that it is possible to separate the dinucleotide steps at the junction into two groups. Group 1 comprises AA/TT, AC/GT, AG/CT and GA/TC, while group 2 contains GG/CC, GC/GC, CG/CG and CA/TG. Group 1 stacking is always preferred over group 2. The conformer distribution is not easily predicted when the two potential stacking combinations are members of the same group, but in these cases the outlying base pairs must contribute to the conformer distribution and the choice of stacking conformer will be more dependent on the sequence context. The simple stacking preferences are summarised in Figure 5C: if N6N1/N2N5 is a group 1 step and N2N3/N4N1 is group 2 then an A/C conformer will be preferred, while if N2N3/N4N1 is group 1 and N6N1/N2N5 is a group 2 then an A/B conformer will be preferred.

There is also a correlation between the properties of the two groups of dinucleotide steps in regular B-DNA duplexes (34–40) and their coaxial stacking preferences in three-way junctions. Group 1 steps tend to be more rigid, show higher propeller twist values, favour BI backbone conformations, have less electronegative charge on both groove sides and favour a more B-like conformation. In contrast, group 2 steps show larger slide motions, and favour more variable backbone conformations often involving the BII conformation or even A-like sugar puckering. Although many factors must affect the conformer selection, parameters that affect the backbone conformation of the exchanging strand will probably have the greatest influence. One characteristic feature of the exchanging strand in three- and four-way junctions is its unusual ε torsion angle value (2,25,41), which is similar to that observed in the BII-conformation (42,43). Previous studies have also shown that backbone modifications can influence conformer selection. For instance, when a nick was introduced at the branch point, it was found that the nick-containing strands always became an exchanging strand (44). To investigate whether backbone-coupled base stacking parameters affect the conformer distribution we have compared the differences in average propeller twist, dinucleotide slide flexibility and tetranucleotide slide flexibility (35,40) between the major and minor conformer sequences of each junction studied to date. These are summarised in Table 1. The data show that the preferred conformers generally have lower slide flexibility, evident at both the di- and tetranucleotide levels, and a tendency for higher propeller twist values than the minor conformers. This is consistent with a recent suggestion (4) that the high propeller twist of the TA step makes it more favourable for coaxial stacking in a four-way junction. Nevertheless, it is clear that these trends do not explain all stacking preferences, as for instance PJ44 would be expected to adopt an A/C conformer rather than the observed A/B conformer.

The similarities with the stacking of B-DNA may also explain the effects of sequence context on some junctions, since group 2 sequences are expected to be more dependent on sequence context than group 1 (40). This is consistent with the results in Figure 4C and D, which show that the AA/TT coaxially stacked step is insensitive to changes in the sequence context.

Finally, it is worth pointing out that the predicted higher coaxial stacking energy of group 2 steps is consistent with recent high-resolution data on four-way junctions. The NMR structures of TWJ1 (15,45) and TWJ-TC (23,24), each possessing a choice between two group 2 steps, are substantially unstacked and distorted. Group 1-containing junctions, on the other hand, show a stacking arrangement close to that expected for B-DNA (22).

Acknowledgments

ACKNOWLEDGEMENTS

We gratefully acknowledge many helpful discussions with Drs Peter Marsh, Graham Mock and Susan Warham and Professor Tom Brown. This work was funded by Cytocell Limited, UK.

REFERENCES

  • 1.Ortiz-Lombardia M., Gonzales,A., Eitja,R., Aymami,J., Azorin,F. and Coll,M. (1999) Crystal structure of a DNA Holliday junction. Nature Struct. Biol., 6, 913–917. [DOI] [PubMed] [Google Scholar]
  • 2.Eichman B.F., Vargason,J.M., Mooers,B.H.M. and Ho,P.S. (2000) The Holliday junction in an inverted repeat DNA sequence: sequence effects on the structure of four-way junctions. Proc. Natl Acad. Sci. USA, 97, 3971–3976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Ho P.S. and Eichman,B.F. (2001) The crystal structures of DNA Holliday junctions. Curr. Opin. Struct. Biol., 11, 302–308. [DOI] [PubMed] [Google Scholar]
  • 4.Vargason J.M. and Ho,P.S. (2002) The effect of cytosine methylation on the structure and geometry of the Holliday junction—the structure of d(CCGGTACm(5)CGG) at 1.5 angstrom resolution. J. Biol. Chem., 277, 21041–21049. [DOI] [PubMed] [Google Scholar]
  • 5.Thorpe J.H., Teixeira,S.C.M., Gale,B.C. and Cardin,C.J. (2002) Structural characterization of a new crystal form of the four-way Holliday junction formed by the DNA sequence d(CCGGTACCGG)2: sequence versus lattice? Acta Crystallogr. D, 58, 567–569. [DOI] [PubMed] [Google Scholar]
  • 6.Eichman B.F., Mooers,B.H.M., Alberti,M., Hearst,J.E. and Ho,P.S. (2001) The crystal structures of psoralen cross-linked DNAs: drug-dependent formation of Holliday junctions. J. Mol. Biol., 308, 15–26. [DOI] [PubMed] [Google Scholar]
  • 7.Lilley D.M.J. (2000) Structures of helical junctions in nucleic acids. Q. Rev. Biophys., 33, 109–159. [DOI] [PubMed] [Google Scholar]
  • 8.Altona C., Pikkemaat,J.A. and Overmars,F.J.J. (1996) Three-way and four-way junctions in DNA: a conformational viewpoint. Curr. Opin. Struct. Biol., 6, 305–316. [DOI] [PubMed] [Google Scholar]
  • 9.Stuhmeier F., Welch,J.B., Murchie,A.I., Lilley,D.M. and Clegg,R.M. (1997) Global structure of three-way DNA junctions with and without additional unpaired bases: a fluorescence resonance energy transfer analysis. Biochemistry, 36, 13530–13538. [DOI] [PubMed] [Google Scholar]
  • 10.Stuhmeier F., Lilley,D.M. and Clegg,R.M. (1997) Effect of additional unpaired bases on the stability of three-way DNA junctions studied by fluorescence techniques. Biochemistry, 36, 13539–13551. [DOI] [PubMed] [Google Scholar]
  • 11.Leontis N.B., Kwok,W. and Newman,J.S. (1991) Stability and structure of three-way DNA junctions containing unpaired nucleotides. Nucleic Acids Res., 19, 759–766. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Welch J.B., Duckett,D.R. and Lilley,D.M.J. (1993) Structures of bulged three-way DNA junctions. Nucleic Acids Res., 21, 4548–4555. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Welch J.B., Walter,F. and Lilley,D.M.J. (1995) Two inequivalent folding isomers of the three-way DNA junction with unpaired bases: sequence-dependence of the folded conformation. J. Mol. Biol., 251, 507–519. [DOI] [PubMed] [Google Scholar]
  • 14.Miick S., Fee,R., Millar,D. and Chazin,W. (1997) Crossover isomer bias is the primary sequence-dependent property of immobilized Holliday junctions. Proc. Natl Acad. Sci. USA, 94, 9080–9084. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Overmars F.J. (1997) Conformational aspects of branched DNA. PhD Thesis, University of Leiden, The Netherlands.
  • 16.Overmars F.J., Pikkemaat,J.A., van den Elst,H., van Boom,J.H. and Altona,C. (1996) NMR studies of DNA three-way junctions containing two unpaired thymidine bases: the influence of the sequence at the junction on the stability of the stacking conformers. J. Mol. Biol., 255, 702–713. [DOI] [PubMed] [Google Scholar]
  • 17.Altona C. (1996) Classification of nucleic acid junctions. J. Mol. Biol., 263, 568–581. [DOI] [PubMed] [Google Scholar]
  • 18.Guo Q., Lu,M., Churchill,M., Tullius,T. and Kallenbach,N. (1990) Asymmetric structure of a three-arm DNA junctions. Biochemistry, 29, 10927–10934. [DOI] [PubMed] [Google Scholar]
  • 19.Lu M., Guo,Q. and Kallenbach,N. (1991) Effect of sequence on the structure of three-arm DNA junctions. Biochemistry, 30, 5815–5820. [DOI] [PubMed] [Google Scholar]
  • 20.Zhong M., Rashes,M., Leontis,N. and Kallenbach,N. (1994) Effect of unpaired bases on the conformation and stability of three-arm DNA junctions. Biochemistry, 33, 3660–3667. [DOI] [PubMed] [Google Scholar]
  • 21.Assenberg R. and Fox,K.R. (2001) The electrophoretic mobility of DNA three-way junctions is affected by the sequence of overhanging single-stranded ends. Electrophoresis, 22, 413–417. [DOI] [PubMed] [Google Scholar]
  • 22.Rosen M.A. and Patel,D.J. (1993) Structural features of a three-stranded DNA junction containing a C-C junctional bulge. Biochemistry, 32, 6576–6587. [DOI] [PubMed] [Google Scholar]
  • 23.Leontis N.B., Hills,M.T., Piotto,M., Ouporov,I.V., Malhotra,A. and Gorenstein,D.G. (1995) Helical stacking in DNA three-way junctions containing two unpaired pyrimidines: proton NMR studies. Biophys. J., 68, 251–265. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Thiviyanathan V., Luxon,B.A., Leontis,N.B., Illangasekare,N., Donne,D.G. and Gorenstein,D.G. (1999) Hybrid–hybrid matrix structural refinement of a DNA three-way junction from 3D NOESY–NOESY. J. Biomol. NMR, 14, 209–221. [DOI] [PubMed] [Google Scholar]
  • 25.Van Buuren B.N.M., Overmars,F.J.J., Ippel,J.H., Altona,C. and Wijmenga,S.S. (2000) Solution structure of a DNA three-way junction containing two unpaired thymidine bases. Identification of sequence features that decide conformer selection. J. Mol. Biol., 304, 371–383. [DOI] [PubMed] [Google Scholar]
  • 26.Cooper J.P. and Hagerman,P.J. (1987) Gel electrophoretic analysis of the geometry of a DNA four-way junction. J. Mol. Biol., 198, 711–719. [DOI] [PubMed] [Google Scholar]
  • 27.Rosen M.A. and Patel,D.J. (1993) Conformational differences between bulged pyrimidines (C-C) and purines (A-A, I-I) at the branch point of three-stranded DNA junctions. Biochemistry, 32, 6563–6575. [DOI] [PubMed] [Google Scholar]
  • 28.Van Buuren B.N.M., Hermann,T., Wijmenga,S.S. and Westhof,E. (2002) Brownian-dynamics simulations of metal-ion binding to four-way junctions. Nucleic Acids Res., 30, 507–514. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Duckett D.R. and Lilley,D.M. (1990) The three-way DNA junction is a Y-shaped molecule in which there is no helix–helix stacking. EMBO J., 9, 1659–1664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Hammann C., Norman,D.G. and Lilley,D.M.J. (2001) Dissection of the ion-induced folding of the hammerhead ribozyme using 19F NMR. Proc. Natl Acad. Sci. USA, 98, 5503–5508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Ulyanov N.B., Ivanov,V.I., Minyat,E.E., Khomyakova,E.B., Petrova,M.V., Lesiak,K. and James,T.L. (1998) A pseudosquare knot structure of DNA in solution. Biochemistry, 37, 12715–12726. [DOI] [PubMed] [Google Scholar]
  • 32.Azaro M.A. and Landy,A. (1997) The isomeric preference of Holliday junctions influences resolution bias by λ integrase. EMBO J., 16, 3744–3755. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Dickerson R. (1999) Helix structure and molecular recognition by B-DNA. In Neidle,S. (ed.), Oxford Handbook of Nucleic Acid Structure. Oxford University Press, Oxford, pp. 145–197.
  • 34.El Hassan M.A. and Calladine,C.R. (1997) Conformational characteristics of DNA: empirical classifications and a hypothesis for the conformational behaviour of dinucleotide steps. Philos. Trans. R. Soc. Lond. A, 355, 43–100. [Google Scholar]
  • 35.Packer M.J., Dauncey,M.P. and Hunter,C.A. (2000) Sequence-dependent DNA structure: dinucleotide conformational maps. J. Mol. Biol., 295, 71–83. [DOI] [PubMed] [Google Scholar]
  • 36.Calladine C.R. (1982) Mechanics of sequence-dependent stacking of bases in B-DNA. J. Mol. Biol., 161, 343–352. [DOI] [PubMed] [Google Scholar]
  • 37.Hunter C. (1993) Sequence-dependent DNA structure: the role of base stacking interactions. J. Mol. Biol., 230, 1025–1054. [DOI] [PubMed] [Google Scholar]
  • 38.El Hassan M.A. and Calladine,C.R. (1996) Propeller-twisting of base-pairs and the conformational mobility of dinucleotide steps in DNA. J. Mol. Biol., 259, 95–103. [DOI] [PubMed] [Google Scholar]
  • 39.Hunter C. and Lu,X. (1997) DNA base-stacking interactions: a comparison of theoretical calculations with oligonucleotide X-ray crystal structures. J. Mol. Biol., 265, 603–619. [DOI] [PubMed] [Google Scholar]
  • 40.Packer M.J., Dauncey,M.P. and Hunter,C.A. (2000) Sequence-dependent DNA structure: tetranucleotide conformational maps. J. Mol. Biol., 295, 85–103. [DOI] [PubMed] [Google Scholar]
  • 41.Von Kitzing E., Lilley,D.M.J. and Diekmann,S. (1990) The stereochemistry of a four-way DNA junction: a theoretical study. Nucleic Acids Res., 18, 2671–2683. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Privé G.G., Yanagi,K. and Dickerson,R.E. (1991) Structure of the B-DNA decamer C-C-A-A-C-G-T-T-G-G and comparison with isomorphous decamers C-C-A-A-G-A-T-T-G-G and C-C-A-G-G-C-C-T-G-G. J. Mol. Biol., 217, 177–199. [DOI] [PubMed] [Google Scholar]
  • 43.Isaacs R.J. and Spielmann,H.P. (2001) NMR evidence for mechanical coupling of phosphate BI-BII transitions with deoxyribose conformational exchange in DNA. J. Mol. Biol., 311, 149–160. [DOI] [PubMed] [Google Scholar]
  • 44.Pöhler J.R.G., Duckett,D.R. and Lilley,D.M.J. (1994) Structure of 4-way DNA junctions containing a nick in one strand. J. Mol. Biol., 238, 62–74. [DOI] [PubMed] [Google Scholar]
  • 45.Overmars F.J. and Altona,C. (1997) NMR study of the exchange rate between two stacked conformers of a model Holliday junction. J. Mol. Biol., 273, 519–524. [DOI] [PubMed] [Google Scholar]

Articles from Nucleic Acids Research are provided here courtesy of Oxford University Press

RESOURCES