1.2.1 Early solution studies on A-tract structure

Using DNA fragments with different A-tract densities Koo and Crothers established a quantitative relationship between the extent of DNA bending and gel mobility anomaly (Koo & Crothers, Reference Koo and Crothers1988). Together with the absolute value of DNA curvature that Crothers and co-workers determined for the overall bend of molecules containing a single A₆-tract at 25°C, 18±2° (Levene et al. Reference Levene, Wu and Crothers1986; Koo et al. Reference Koo, Drak, Rice and Crothers1990), it is now possible to determine the curvature of any desired sequence by comparing its migration anomaly to that of DNA molecules containing phased A₆-tracts. Other earlier estimates for the magnitude of bending of a single A₆-tract ranged from 11° (Calladine et al. Reference Calladine, Drew and McCall1988) to about 28° (Ulanovsky et al. Reference Ulanovsky, Bodner, Trifonov and Choder1986).

Models for the structure of poly(dA)⋅poly(dT) based on fiber diffraction (Arnott & Selsing, Reference Arnott and Selsing1974; Arnott et al. Reference Arnott, Chandrasekaran, Hall and Puigjaner1983; Alexeev et al. Reference Alexeev, Lipanov and Skuratovskii1987) proposed a unique structure, distinct from that of canonical B-DNA, in which the bases are negatively inclined relative to the overall helix axis. In addition, the bases are highly propeller twisted and the minor groove is narrow. Solution studies by nuclear magnetic resonance (NMR) (Kintanar et al. Reference Kintanar, Klevit and Reid1987; Katahira et al. Reference Katahira, Sugeta, Kyogoku, Fujii, Fujisawa and Tomita1988; Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Nadeau & Crothers, Reference Nadeau and Crothers1989; Moe & Russo, Reference Moe and Russo1990) and by hydroxy-radical footprinting (Burkhoff & Tullius, Reference Burkhoff and Tullius1987) showed that the structure of A-tracts is non-uniform. There is a progressive reduction in minor groove width going from the 5′ ends towards the 3′ end of the adenine strand. The regions between the A-tracts are in B-DNA conformation and thus have a wider minor groove (however, see below on the effects of A-tracts on sequences flanking it). In addition, NMR spectroscopy (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Nadeau & Crothers, Reference Nadeau and Crothers1989) and gel electrophoresis studies (Haran & Crothers, Reference Haran and Crothers1989) provided evidence that the formation of the unique A-tract structure is cooperative. A₃ or shorter tracts show structural characteristic of B-DNA; however, as the A-tract region is lengthened to A₄ and longer tracts, there is a cooperative transition to a structure more resembling the structure of poly(dA)⋅poly(dT).

The observation that the minor groove in A-tract regions narrows down going in the 5′ to 3′ direction (Burkhoff & Tullius, Reference Burkhoff and Tullius1987; Kintanar et al. Reference Kintanar, Klevit and Reid1987; Katahira et al. Reference Katahira, Sugeta, Kyogoku, Fujii, Fujisawa and Tomita1988; Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Nadeau & Crothers, Reference Nadeau and Crothers1989; Moe & Russo, Reference Moe and Russo1990) implies structural polarity within the A-tract region. However, it was the elegant study by Hagerman (Reference Hagerman1986) that clearly demonstrated that sequence polarity is a significant factor in DNA bending. Hagerman (Reference Hagerman1986) showed that sequences of the form d(CA₄T₄G) migrate anomalously slow on polyacrylamide gels (i.e. as a bend fragment), whereas sequences of the form d(CT₄A₄G) migrate on gels as a straight fragment. These surprising observations generated many studies, discussed below, that advanced our ideas about A-tract structure and intrinsic bending. Leroy et al. (Reference Leroy, Charretier, Kochoyan and Gueron1988) showed that A₂T₂ already possess structural properties of A-tracts, similar to those of the A₄-tract. Thus, A-tracts can be symmetric (A_nT_n, n⩾2) or non-symmetric (A_n, n⩾4).

The uniqueness of A-tracts, as the main sequence element leading to intrinsic DNA bending, when repeated in phase with the DNA helical repeat, was established by selection-amplification experiments, which showed that sequences selected for reduced electrophoretic mobility contained phased A-tracts (Beutel & Gold, Reference Beutel and Gold1992).

1.2.2 Early studies using X-ray crystallography

Early X-ray studies of short DNA fragments incorporating A-tracts revealed that in contrast to predictions based on solution measurements of sequence-directed DNA bending, the base pairs that define the A-tract were orthogonal to the helical axis (Coll et al. Reference Coll, Frederick, Wang and Rich1987; Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; DiGabriele et al. Reference DiGabriele, Sanderson and Steitz1989; Edwards et al. Reference Edwards, Brown, Spink, Skelly and Neidle1992; DiGabriele & Steitz, Reference DiGabriele and Steitz1993; Shatzky-Schwartz et al. Reference Shatzky-Schwartz, Arbuckle, Eisenstein, Rabinovich, Bareket-Samish, Haran, Luisi and Shakked1997). The A⋅T base pairs in the crystalline structures were highly propeller twisted, to a degree that permits the making of interstrand bifurcated hydrogen bonds across the major groove (Coll et al. Reference Coll, Frederick, Wang and Rich1987; Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; DiGabriele et al. Reference DiGabriele, Sanderson and Steitz1989; Edwards et al. Reference Edwards, Brown, Spink, Skelly and Neidle1992; DiGabriele & Steitz, Reference DiGabriele and Steitz1993; Shatzky-Schwartz et al. Reference Shatzky-Schwartz, Arbuckle, Eisenstein, Rabinovich, Bareket-Samish, Haran, Luisi and Shakked1997). This propeller-twisted conformation has a narrow minor groove. However, this end result can be achieved by other structural means, such as negative inclination of the base pairs, or their displacement from the helix axis. These parameters were negligible in the early X-ray structures but were dominant in early models based on fiber diffraction (Alexeev et al. Reference Alexeev, Lipanov and Skuratovskii1987), NMR (Behling & Kearns, Reference Behling and Kearns1986; Lipanov & Chuprina, Reference Lipanov and Chuprina1987) or linear dichroism (Edmonson & Johnson, Reference Edmonson and Johnson1985).

1.2.3 Early models for A-tract-induced DNA bending

Early models proposed for A-tract structure were of two kinds: nearest-neighbor models that assumed independent dinucleotide additivity, both within A-tracts as well as in the B-DNA regions and cooperative or long-range models that proposed a different conformation for A–A dinucleotides vs. longer A-tracts. The prototypical nearest-neighbor model for intrinsic DNA bending is the ‘wedge model’ (Ulanovsky et al. Reference Ulanovsky, Bodner, Trifonov and Choder1986; Ulanovsky & Trifonov, Reference Ulanovsky and Trifonov1987; Bolshoy et al. Reference Bolshoy, McNamara, Harrington and Trifonov1991). According to the ‘wedge model’, all combinations of base pairs have a ‘wedge’, which is a combination of roll and tilt angles specific to this dinucleotide (Ulanovsky et al. Reference Ulanovsky, Bodner, Trifonov and Choder1986; Ulanovsky & Trifonov, Reference Ulanovsky and Trifonov1987; Bolshoy et al. Reference Bolshoy, McNamara, Harrington and Trifonov1991). Placing several A–A dinucleotides in a row is not different from any other sequence combination, except that since all the dinucleotides are of the same type, they do not cancel each other in their effect on the trajectory of the helix axis. Other nearest-neighbor models are those proposed by De Santis et al. (Reference De Santis, Palleschi, Savino and Scipioni1990) and Olson et al. (Reference Olson, Marky, Jernigan and Zhurkin1993).

A long-range model for A-tract structure, which is consistent with all solution and fiber studies, was proposed by Crothers and co-workers, called the ‘A-tract junction model’ (Koo et al. Reference Koo, Wu and Crothers1986; Koo & Crothers, Reference Koo and Crothers1988; Crothers et al. Reference Crothers, Haran and Nadeau1990). The global curvature in this model is attributed to the unique structure of A-tracts, a structure that is cooperatively formed whenever there are four or more adjacent adenine bases (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Haran & Crothers, Reference Haran and Crothers1989; Nadeau & Crothers, Reference Nadeau and Crothers1989). In this structure, the base pairs are negatively inclined relative to the global helix axis of the whole A-tract region. In contrast, the bases at the B-DNA regions, neighboring the A-tracts, are on average perpendicular to the global axis of the B-DNA region. Because of the difference in base-pair inclination in the two regions, there is a break in base stacking at the junctions. To restore proper base stacking, the two global helical axes are bent at the junctions. The net consequence is a global bending whose direction is towards the minor groove at the middle of the A-tracts (Koo et al. Reference Koo, Wu and Crothers1986; Koo & Crothers, Reference Koo and Crothers1988; Crothers et al. Reference Crothers, Haran and Nadeau1990; Haran et al. Reference Haran, Kahn and Crothers1994).

The A-tract junction model and the wedge models are not incompatible but rather use different coordinate frames. The inclination of the base pairs, with respect to a global helix axis in the junction model, is equivalent to the local wedges with their associated helical twist, when we change the coordinate frame to a local one. The only significant difference between the two models is that the A-tract junction model proposes a distinct helical structure for the A-tract region different from generic B-DNA with cooperativity in its nucleation, as observed experimentally (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Haran & Crothers, Reference Haran and Crothers1989; Nadeau & Crothers, Reference Nadeau and Crothers1989). Hence, according to the junction model, the structure of the A–A dinucleotides is context-dependent.

An equivalent junction model can be generated by assuming that the A-tracts are perpendicular to the helix axis, whereas in the flanking regions, the base pairs have an overall positive inclination relative to a global axis (or positive roll angles). Such a model was suggested (Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; Calladine et al. Reference Calladine, Drew and McCall1988; Yoon et al. Reference Yoon, Prive, Goodsell and Dickerson1988; Goodsell et al. Reference Goodsell, Kopka, Cascio and Dickerson1993, Reference Goodsell, Kaczor-Grzeskowiak and Dickerson1994; Dickerson et al. Reference Dickerson, Goodsell and Neidle1994; Goodsell & Dickerson, Reference Goodsell and Dickerson1994) based on crystal structure analyses of DNA oligomers containing A-tracts (Coll et al. Reference Coll, Frederick, Wang and Rich1987; Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; Yoon et al. Reference Yoon, Prive, Goodsell and Dickerson1988; DiGabriele et al. Reference DiGabriele, Sanderson and Steitz1989; Edwards et al. Reference Edwards, Brown, Spink, Skelly and Neidle1992; DiGabriele & Steitz, Reference DiGabriele and Steitz1993; Goodsell et al. Reference Goodsell, Kopka, Cascio and Dickerson1993, Reference Goodsell, Kaczor-Grzeskowiak and Dickerson1994; Shatzky-Schwartz et al. Reference Shatzky-Schwartz, Arbuckle, Eisenstein, Rabinovich, Bareket-Samish, Haran, Luisi and Shakked1997). This model takes the viewpoint that B-DNA sections adjoining the A-tract regions have an overall positive roll, whereas the A-tracts are rigid (Sarai et al. Reference Sarai, Mazur, Nussinov and Jernigan1989; Zhurkin et al. Reference Zhurkin, Ulyanov, Gorin and Jernigan1991; El Hassan & Calladine, Reference El Hassan and Calladine1996) and straight (i.e. have zero roll)(Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; DiGabriele & Steitz, Reference DiGabriele and Steitz1993; Dickerson et al. Reference Dickerson, Goodsell and Neidle1994). The phenomenon of global bending occurs when rigid but straight helical segments are introduced within a writhed helix. The direction of the net curvature is thus towards the major groove of B-DNA segments (Nelson et al. Reference Nelson, Finch, Luisi and Klug1987; Calladine et al. Reference Calladine, Drew and McCall1988; Goodsell et al. Reference Goodsell, Kopka, Cascio and Dickerson1993, Reference Goodsell, Kaczor-Grzeskowiak and Dickerson1994; Dickerson et al. Reference Dickerson, Goodsell and Neidle1994; Goodsell & Dickerson, Reference Goodsell and Dickerson1994). This model is called the ‘general-sequence junction model’ or the ‘non-A-tract junction model’.

It should be pointed out that according to both junction models, there are no implied specific bends at the junctions (either roll or tilt). It is only the non-perpendicular nature of the orientation of the base pairs, in either the A-tract or the general sequence B-DNA region, that forces the global axis of these two regions to be inclined relative to each other, in order to preserve the base-pair stacking across the two regions.

1.3.1 Propeller twist and bifurcated hydrogen bonds

Fritsch and Westhoff concluded, based on a molecular dynamics study of dA⋅dT homopolymers, that bifurcated hydrogen bonds are transient in nature and are formed as a consequence of the unusual propeller-twisted conformation, rather than contribute to the stability of this conformation (Fritsch et al. Reference Fritsch, Ravishanker, Beveridge and Westhof1993). A similar conclusion was reached from energy calculation and refinement of fiber-diffraction models by Lipanov et al. (Reference Lipanov, Chuprina, Alexeev and Skuratovskii1990). It has been shown that substituting some of the adenines in A-tracts by inosines does not cause a significant reduction in curvature, whereas the complete substitution of A₅ by I₅ reduces the curvature significantly (Diekmann et al. Reference Diekmann, von Kitzing, McLaughlin, Ott and Eckstein1987, Reference Diekmann, Mazzarelli, McLaughlin, von Kitzing and Travers1992; Koo & Crothers, Reference Koo and Crothers1987). Diekmann et al. (Reference Diekmann, Mazzarelli, McLaughlin, von Kitzing and Travers1992) advanced these observations as challenging the role of propeller twist and bifurcated hydrogen bonds for A-tract structure, since inosine bases were not supposed to permit the N₆–O₄ bifurcated hydrogen bonds observed in pure A-tracts. However, Shatzky-Schwartz et al. (Reference Shatzky-Schwartz, Arbuckle, Eisenstein, Rabinovich, Bareket-Samish, Haran, Luisi and Shakked1997) showed that in the crystal structure of DNA molecules with alternating adenine and inosine bases the base-stacking geometry is uniform and similar to that of pure A-tracts, including analogous bifurcated hydrogen bonds as observed in pure A-tracts. On the other hand, pure inosine tracts showed variable stacking patterns with low propeller twist that does not allow for the formation of bifurcated hydrogen bonds (Shatzky-Schwartz et al. Reference Shatzky-Schwartz, Arbuckle, Eisenstein, Rabinovich, Bareket-Samish, Haran, Luisi and Shakked1997). Propeller twist to the extent permitting the formation of bifurcated hydrogen bonds were also observed in the solution structure of an A₆-tract (MacDonald et al. Reference MacDonald, Herbert, Zhang, Polgruto and Lu2001) but not in the solution structure of an A₄-tract (Barbic et al. Reference Barbic, Zimmer and Crothers2003).

1.3.2 Spine of hydration

Lipanov and Chuprina (Reference Lipanov and Chuprina1987) used energy calculations to show that base stacking in the propeller-twisted conformation are sub-optimal and that it allows the creation of the spine of hydration, which more than compensates for the base stacking. Early NMR studies of the DDD sequence used nuclear Overhauser effects (NOEs) cross peaks between water and base protons, to show that the spine of hydration exists also in solution (Kubinec & Wemmer, Reference Kubinec and Wemmer1992; Liepinsh et al. Reference Liepinsh, Otting and Wuthrich1992). It was noted that the spine exist in A₂T₂ but not in T₂A₂ (Liepinsh et al. Reference Liepinsh, Otting and Wuthrich1992). A lower limit of one to few nanoseconds was placed on the residence time of water in the spine of hydration of the DDD sequence (Kubinec & Wemmer, Reference Kubinec and Wemmer1992; Liepinsh et al. Reference Liepinsh, Otting and Wuthrich1992), but the number of water molecules responsible for the cross peaks could not be determined. A subsequent study using both NOEs and the nuclear magnetic relaxation dispersion (NMRD) technique determined that the residence time of five water molecules in the minor groove of the DDD sequence is 0·9±0·1 ns at 4°C (Denisov et al. Reference Denisov, Carlstrom, Kandadai and Halle1997). However, subsequent NMR studies (Johannesson & Halle, Reference Johannesson and Halle1998; Phan et al. Reference Phan, Leroy and Gueron1999) showed that the residence time of water molecules in structures containing A-tracts (such as A₂T₂) and those not containing A-tracts (such as T₂A₂, ATAT, or a single TA) are similar. Thus, any A,T-rich region with a narrow minor groove will form a single file of hydration. These observations mean that the spine of hydration is important in the stability of bent and straight DNA molecules to the same extent.

1.4.1 Temperature

A temperature effect on A-tract structure is documented. Early studies showed a modest increase in bend magnitude as the temperature is decreased (Marini et al. Reference Marini, Effron, Goodman, Singleton, Wells, Wartell and Englund1984; Koo et al. Reference Koo, Wu and Crothers1986; Levene et al. Reference Levene, Wu and Crothers1986; Diekmann, Reference Diekmann1987). However, these studies did not deconvolute the effect of temperature on the DNA helical repeat (Depew & Wang, Reference Depew and Wang1975). Subsequent studies observed a temperature-dependent premelting transition in poly(dA)·poly(dT) molecules, i.e. equilibrium between at least two helical states by UV absorbance (Herrera & Chaires, Reference Herrera and Chaires1989) and by CD spectroscopy (Chan et al. Reference Chan, Breslauer, Hogan, Kessler, Austin, Ojemann, Passner and Wiles1990). These authors suggested that the temperature-dependent transition is accompanied by a disruption of bound water and by a reversion to a more standard B-DNA structure with normal electrophoretic behavior (Herrera & Chaires, Reference Herrera and Chaires1989; Chan et al. Reference Chan, Breslauer, Hogan, Kessler, Austin, Ojemann, Passner and Wiles1990). The midpoint of the observed structural change is 30°C, and hence can be of biological significance. Similar observations were made by Chan et al. (Reference Chan, Breslauer, Austin and Hogan1993) using synthetic DNA molecules containing phased A₅-tracts. The use of synthetic sequences allowed for direct thermodynamic measurement of the premelting transition by differential scanning calorimetry (DSC). This study showed (using CD, DSC and electric birefringence) that phased A₅-tracts undergo a true cooperative phase transition as a function of temperature between two conformational states – a bent one and a straight one. The transition was well modeled as a two-state structural transition (Chan et al. Reference Chan, Breslauer, Austin and Hogan1993). Using data from this study, it has been estimated that the cooperative unit in the transition is 5 bp (Crothers & Shakked, Reference Crothers, Shakked and Neidle1999), which supports earlier data (Haran & Crothers, Reference Haran and Crothers1989) on the cooperative nature of A-tracts formation and disruption.

Chan et al. (Reference Chan, Austin, Mukerji and Spiro1997) used UV resonance Raman measurements to demonstrate changes in the thymine C₄=O stretch and in the adenine N₆ amino scissor mode as a function of temperature. Chan et al. (Reference Chan, Austin, Mukerji and Spiro1997) suggested that these results support the presence of bifurcated hydrogen bonds in the major groove of A-tracts. The breaking of the bifurcated hydrogen bonds as the temperature is elevated is the cause for the reversion of the structure to a straight B-DNA conformation (Chan et al. Reference Chan, Austin, Mukerji and Spiro1997). Park and Breslauer (Reference Park and Breslauer1991) observed (using CD spectroscopy and DSC) significant premelting for DNA molecules containing the A₄T₄ tract but not for molecules containing the non-bent T₄A₄-tract. Premelting transition was also observed for A₃T₃ (Mukerji & Williams, Reference Mukerji and Williams2002), and the strength of the bifurcated hydrogen bonds in this system was estimated to be ~0·5 kcal/mol, whereas those reported for Watson–Crick hydrogen bonds are 2–3 kcal/mol (Sinden, Reference Sinden1994). A fluorescent adenosine analogue 4-amino-8-(2-deoxy-β-d-ribofuranosyl)-5′-O-dimethoxytrityl-6-methyl-7(8H)-pteridone (6MAP) was used to probe the local environment of A-tract regions during the premelting transition (Augustyn et al. Reference Augustyn, Wojtuszewski, Hawkins, Knutson and Mukerji2006). This study showed that ApT steps have a lower contribution to the premelting transition and that they undergo premelting transition at lower temperature than ApA steps (Augustyn et al. Reference Augustyn, Wojtuszewski, Hawkins, Knutson and Mukerji2006).

A temperature-dependent conformational transition was observed for poly(dA)⋅poly(dT) also by X-ray fiber diffraction. Premilat & Albiser (Reference Premilat and Albiser1997) observed a reversible transition between two discrete conformational states when the temperature was elevated above 30°C. The lower temperature structure is the one described above for poly(dA)⋅poly(dT). The high temperature structure is a structure distinct from B-DNA or the low-temperature poly(dA)⋅poly(dT) structure, with a wider minor groove and less propeller twisting of the base pairs (Premilat & Albiser, Reference Premilat and Albiser1997).

1.4.2 Organic solvent

DNA crystallization is often induced by 2-methyl-2,4-pentanediol (MPD) (Sprous et al. Reference Sprous, Zacharias, Wood and Harvey1995). The anomalous migration on polyacrylamide gels (PAG) of A-tracts, relative to a DNA fragment of identical length but having random sequence, is the hallmark of bent DNA sequences (Marini et al. Reference Marini, Levene, Crothers and Englund1982; Koo et al. Reference Koo, Wu and Crothers1986). This anomalous migration pattern is substantially reduced by the addition of MPD to the PAG matrix (Sprous et al. Reference Sprous, Zacharias, Wood and Harvey1995). The interpretation given by the authors was that MPD partially straightens the helix (Sprous et al. Reference Sprous, Zacharias, Wood and Harvey1995). In subsequent investigation of this problem, this effect has been attributed to explicit interaction of MPD with only the A-tract region of the oligomer (Dlakic & Harrington, Reference Dlakic and Harrington1996; Ganunis et al. Reference Ganunis, Guo and Tullius1996). Thus, it has been proposed that MPD is the basis of the inconsistencies between solution and crystallographic results (Sprous et al. Reference Sprous, Zacharias, Wood and Harvey1995). However, Dickerson et al. (Reference Dickerson, Goodsell and Kopka1996) showed an absence of relationship between DNA helix bending and concentration of MPD within the solution from which the crystals were grown. Using the work of Timasheff and colleagues (Reference Timasheff1993), Dickerson et al. (Reference Dickerson, Goodsell and Kopka1996) make a case that MPD is not a dehydrating agent. Instead, MPD has a preferential hydrating effect, since it is excluded from the vicinity of the macromolecule. Atomic resolution crystal structure of DNA oligomers does not show any indication of MPD (Dickerson et al. Reference Dickerson, Goodsell and Kopka1996).

Haran et al. (Reference Haran, Cohen, Spasic, Yang and Mohanty2004) analyzed the absolute mobility of both A-tracts as well as generic B-DNA fragments and showed that MPD has a larger apparent effect on the gel mobility of generic B-DNA sequences, than on A-tract-containing sequences. In addition, studies by Haran et al. (Reference Haran, Cohen, Spasic, Yang and Mohanty2003, Reference Haran, Cohen, Spasic, Yang and Mohanty2004) suggest that MPD is excluded from the surface of both A-tracts and generic B-DNA molecules. MPD affects the properties of PAG by lowering the dielectric constant and by increasing the viscosity and thus in turn the frictional coefficient of the gels.

1.4.3 Monovalent ions and A-tract bending

Based on the redetermination of the crystal structure of the original DDD at very high resolution, Shui et al. (Reference Shui, McFail-Isom, Hu and Williams1998a, Reference Shui, Sines, McFail-Isom, Vanderveer and Williamsb) proposed that monovalent cations, bound in the minor grove of A-tracts, are the primary cause for global bending at these regions. The main argument was that at the very high resolution observed in these structures (1·4 and 1·75 Å), the electron density previously attributed exclusively to water molecules in the spine of hydration in the minor groove is partially composed of Na⁺ ions (Shui et al. Reference Shui, McFail-Isom, Hu and Williams1998a) or K⁺ ions (Shui et al. Reference Shui, Sines, McFail-Isom, Vanderveer and Williams1998b) and that charge neutralization by these cations is the cause of the narrowing of the minor groove, which leads to the global bending phenomenon.

Since the initial proposal by Williams and co-workers (Shui et al. Reference Shui, McFail-Isom, Hu and Williams1998a, Reference Shui, Sines, McFail-Isom, Vanderveer and Williamsb), there has been renewed interest in high-resolution crystal structures of oligomeric A-tract DNA. Tereshko et al. (Reference Tereshko, Minasov and Egli1999a) have determined the DDD structure at 1·1 Å and have not found evidence for the presence of Na⁺ ions in the minor groove. The presence of Mg²⁺ ions bound to phosphate groups at the periphery of the minor groove near the A-tract region makes it an unlikely position for Na⁺ coordination. To further investigate the possibility of alkali-metal coordination in the minor groove and assist in its identification, Tereshko et al. (Reference Tereshko, Minasov and Egli1999b) crystallized the DDD sequence in the presence of either Rb⁺ or Cs⁺. Rb⁺, but not Cs⁺, is found in a single location within the minor groove at the center of the sequence bridging the O₂ atoms of the two central thymines. Its occupancy, however, is only fractional (Tereshko et al. Reference Tereshko, Minasov and Egli1999b). Its presence seems not to affect groove width, since the dimensions of the groove are the same in the DDD structures with and without Rb⁺. In fact, the authors suggest that Rb⁺ is bound at this location because its ionic radius (ca. 1·5 Å) is small enough to fit into the narrow minor groove and its coordination distances are optimal for replacing the water molecules that compose the spine of hydration (Tereshko et al. Reference Tereshko, Minasov and Egli1999b). They suggest that likewise K⁺ (ionic radius of ca. 1·3 Å) will bind readily there, but not Cs⁺ (too big a ion) or Na⁺ (too small a ion) (Tereshko et al. Reference Tereshko, Minasov and Egli1999b). Thus, monovalent ions bind only sporadically in the minor groove and seem to be an outcome of the width of the minor groove, not its cause. This argues against the view that ions can invade the minor groove spine of hydration along its entire length and is the cause of the global bending phenomenon by electrostatic collapse (McFail-Isom et al. Reference McFail-Isom, Sines and Williams1999).

Furthermore, when the DDD sequence was redetermined by the Dickerson group (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999), at 1·43 Å resolution, no monovalent cations were observed in the minor groove of this structure. In this dodecamer construct, a tether held the two strands bridging the adenine residues at the A–T step (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999). The tether was long enough so that no bend is induced in the helix and the structure was isostructural to the original DDD structures (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999). It has been confirmed by inductively coupled plasma mass spectrometry (ICP-MS) that the cross-linked structure contained, on the average, one sodium ion for every 160 bp and one potassium ion per 41 bp (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999). On the other hand, 11·6 Mg²⁺ ions where found by the ICP-MS method in the cross-linked DDD structure or roughly one per base pair, although only 3 ions are localized, one within the major groove and two within the interstitial space between laterally packed helices (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999). These localized Mg²⁺ ions help to stabilize the packing of adjacent helices in the crystal lattice. The effect of divalent ions on bending is further discussed below.

Hud et al. (Reference Hud, Sklenar and Feigon1999) have observed ammonium ions binding in the minor groove, using NMR and isotopically labeled ammonium ions (¹⁵NH₄⁺). The location of the ions was at the narrowest part of the minor groove. Bound NH₄⁺, as well as water, were found to be in fast exchange (sub millisecond life time) from the same binding sites (Hud et al. Reference Hud, Sklenar and Feigon1999). This correlation, however, between the width of the minor groove and monovalent ion binding, does not necessarily lead to the conclusion that ion binding is the cause of the narrowing of the minor groove. It should be noted that a narrow minor groove is not a property unique to homopolymeric A-tracts but of any A,T-rich sequence (Dickerson, Reference Dickerson and Neidle1999). In addition, as mentioned above, the single file of hydration (spine) forms in any A,T-rich region with a narrow minor groove. However, once monovalent ions are bound to a narrow minor groove, they may further narrow it and thus modulate bending further.

Denisov & Halle (Reference Denisov and Halle2000) have used ²³Na magnetic relaxation dispersion measurements on A-tract-containing oligonucleotides and showed that the occupancy of the sodium ion in the minor groove is only a few percent. Other monovalent ions, such as potassium, rubidium and cesium were also observed to be only weakly bound there, whereas ammonium ions were found to bind with slightly increased affinity (Denisov & Halle, Reference Denisov and Halle2000). However, Marincola et al. (Reference Marincola, Denisov and Halle2004) have extended these studies using DNA of higher purity and measured the full ²³Na and ⁸⁷Rb magnetic relaxation dispersion profiles in DNA solutions containing both ions. They observed 40% Rb⁺ occupancy of the ApT site of the DDD sequence and 10% Na⁺ occupancy, which increases to 50% in the absence of Rb⁺ ions.

Using fluorescence resonance energy transfer and fluorescence polarization anisotropy (FPA), Toth et al. (Reference Toth, Sauermann and Langowski1998) have estimated a bend angle of 23±3° per helical turn for an A₆ sequence at 10 mm NaCl. The curvature of the DNA helix was found to increase with salt concentration up to 41° per A₆ block at 500 mm NaCl (Toth et al. Reference Toth, Sauermann and Langowski1998). The NaCl dependence support models that postulate the modulation of A-tract bending by monovalent ions, but does not support models that suggest that monovalent ions bound at the minor groove are the primary source for A-tract bending. This is because the results at low salt agree with results on bend magnitude from other experiments carried out at low salt (Griffith et al. Reference Griffith, Bleyman, Rauch, Kitchin and Englund1986; Levene et al. Reference Levene, Wu and Crothers1986; Koo & Crothers, Reference Koo and Crothers1988).

The relationship between monovalent ions binding and A-tract bending has been investigated also by capillary electrophoresis (Stellwagen et al. Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001). It has been established that DNA oligomers and polymers migrate with different mobilities in free solution because of differences in solvent friction and effective charge (Mohanty et al. Reference Mohanty, Searls and McLaughlin1998). If there is preferential binding of ions in the minor groove of A-tracts, then the effective charge of the A-tracts should decrease relative to the effective charge in non-A-tract DNA of the same length. This should lead to a preferential decrease in the absolute free solution electrophoretic mobility of the A-tracts relative to random sequence B-DNA. This idea was explored by Stellwagen and co-workers (Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001) for four 20-bp oligonucleotides, containing the A₅, A₃T₃, T₃A₃ or CACA motifs in phase with the helical screw. To facilitate the identification of the peaks in the electropherograms, linear pUC19 was added to the solution at a small concentration. In all cases, the mobility increased linearly with concentration of NaCl (Stellwagen et al. Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001). The authors showed an apparent dependence of mobility on DNA sequence, with the mobility of the A₃T₃ motif being most affected by monovalent ions, the CACA sequence least affected and the A₅A₅ and T₃A₃ moderately affected by the ionic changes (Stellwagen et al. Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001). They interpreted the results as suggesting a preferential binding of the Na⁺ ions in the minor groove of A-tract sequences (Stellwagen et al. Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001).

We believe these results are inclusive for the following reasons: First, the measurements are relative to another DNA sequence. Indeed, Stellwagen et al. (Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001) show that both the pUC19 and the T₃A₃ sequence change their electrophoretic pattern as a function of NaCl concentration. Second, in capillary electrophoresis, there are random 1–3% changes in absolute mobility as a function of the capillary being used, but the authors base their results on less than 1% change in mobility. Third, even though the sequence A₃T₃ shows the slowest absolute mobility and CACA the largest, the slope of the changes in mobility (relative to pUC19) as a function of NaCl concentration seems to be the same (Stellwagen et al. Reference Stellwagen, Magnusdottir, Gelfi and Righetti2001). In a further study, using transient electric birefringence measurements and free solution electrophoresis of two 200-bp restriction fragments, Lu and Stellwagen (Reference Lu and Stellwagen2008) suggested that each A-tract in the fragments bind a monovalent cation, approximately one-third of the time. In this study, there appeared to be cooperative binding of monovalent ions to two or more adjacent A-tracts. However, this study did not distinguish between minor vs. major groove location of the monovalent ions (Lu & Stellwagen, Reference Lu and Stellwagen2008).

Thus, there is no conclusive evidence (from either solution or X-ray studies) for a definitive causative relationship between monovalent ions binding within the minor groove of A-tracts and induction of the global bending phenomenon by charge neutralization electrostatic collapse (McFail-Isom et al. Reference McFail-Isom, Sines and Williams1999). However, once a narrow minor groove has formed, then cations may bind there part of the time and thus further modulate bending. All in all, it seems more likely that DNA bending originates from the unique base-pair stacking arrangement of A-tracts.

1.4.4 Divalent ions and A-tract bending

In contrast to the controversy that exists over whether monovalent cations are at the root of A-tract bending or contribute significantly to it, there is a consensus that divalent metal ions can have a significant effect on the structure of DNA in general and on DNA bending (whether by A-tract or by other sequences) in particular. Here we limit ourselves to a discussion on the effect of divalent ions on A-tract-containing molecules. Discussions on the effects of divalent ions on DNA structure in general have been published (Bloomfield et al. Reference Bloomfield, Crothers and Tinoco2000).

The interactions of monovalent ions and divalent ions with DNA are known to be different. Whereas the interaction of monovalent ions with DNA is electrostatic and non-sequence-specific, due to lack of extensive hydration, that of divalent ions is sequence-specific and involves more fully hydrated ions (Chiu & Dickerson, Reference Chiu and Dickerson2000). Diekmann & Wang (Reference Diekmann and Wang1985) were the first to note the enhancing effect of Mg²⁺ ions on the anomaly in gel migration of A-tract-containing DNA fragments. Soon after, Laundon & Griffith (Reference Laundon and Griffith1987) provided evidence from electron microscopy that larger fraction of kinetoplast DNA appears bent upon addition of divalent ions. Brukner et al. (Reference Brukner, Susic, Dlakic, Savic and Pongor1994) have shown that Mg²⁺ can influence bending of A-tracts by increasing the major-groove bending of GGGCCC tracts that are adjacent to A-tracts. The first solution study to find ions in the minor groove of A-tracts was the NMR study by Hud & Feigon (Reference Hud and Feigon1997). They have observed Mn²⁺ binding at the 5′ but not the 3′ side of A-tracts. Since it is known that the width of the minor groove at A-tracts regions is narrower than the minor groove at B-DNA regions and gets progressively narrow in the 5′→3′ direction, as discussed above, this observation can mean that the groove width at the 3′ region (but not the 5′ region of the minor groove) is too narrow for the Mn²⁺ ions to penetrate.

Several groups have recrystallized the CGAATTCG motif, the central octamer of the DDD sequence, and determined the structures at atomic resolution. These structures were solved in several space groups, incorporated in sequences ranging from a 9- to a 12-bp duplex in the presence of various divalent ions (Chiu et al. Reference Chiu, Kaczor-Grzeskowiak and Dickerson1999; Liu & Subirana, Reference Liu and Subirana1999; Minasov et al. Reference Minasov, Tereshko and Egli1999; Soler-Lopez et al. Reference Soler-Lopez, Malinina, Liu, Huynh-Dinh and Subirana1999; Tereshko et al. Reference Tereshko, Minasov and Egli1999a, Reference Tereshko, Minasov and Eglib; Johansson et al. Reference Johansson, Parkinson and Neidle2000). Most structures recapitulated the main features of the DDD structure at medium resolution (Wing et al. Reference Wing, Drew, Takano, Broka, Tanaka, Itakura and Dickerson1980; Dickerson et al. Reference Dickerson, Drew, Conner, Wing, Fratini and Kopka1982). However, the increased resolution enabled a more detailed analysis of ions binding within the grooves of the DNA helix and its influence on DNA bending. In all structures, regardless of length, space group, or divalent ion identity, there is no penetration of divalent ions into the spine of hydration at the minor groove. The studies found both Mg²⁺ ions (Minasov et al. Reference Minasov, Tereshko and Egli1999; Soler-Lopez et al. Reference Soler-Lopez, Malinina, Liu, Huynh-Dinh and Subirana1999) as well as Ca²⁺ ions (Liu & Subirana, Reference Liu and Subirana1999; Minasov et al. Reference Minasov, Tereshko and Egli1999) in the minor groove, but only at the ends of the A-tract region, interacting with the ends of the spine.

Jerkovic & Bolton (Reference Jerkovic and Bolton2001) observed by NMR that Mg²⁺ enhances curvature in A-tract-containing sequences. This study showed a direct effect of Mg²⁺ on A-tract, since a sequence containing the AAAAAT motif was measured relative to a sequence containing the ATATAT motif, with the same flanking sequences. Increase in Mg²⁺ caused doubling of the shape function (Jerkovic & Bolton, Reference Jerkovic and Bolton2001). The addition of Mg²⁺ increased the curvature at all concentrations and thus Mg²⁺ increases an already present curvature in A-tracts.

Ions, especially divalent ions, can have an appreciable effect on the helical repeat of DNA (Xu & Bremer, Reference Xu and Bremer1997), as observed also for the temperature effect (discussed above). Thus, any study on the effect of divalent ions on DNA, by any solution method, should deconvolute the effect of ions on the helical repeat vs. on bending per se. The effect of magnesium on bend angle in a sequence containing five helically phased A₆-tracts has been studied using the rotational invariant method (Tchernaenko et al. Reference Tchernaenko, Halvorson and Lutter2004). In the absence of magnesium, the bend angle is reduced by 12° and the helix repeat changes from 10·38 bp/turn to 10·45 (Tchernaenko et al. Reference Tchernaenko, Halvorson and Lutter2004). Interestingly, using different buffers with varying magnesium concentrations, the fraction of charge phosphate neutralized by the counterions was estimated (Tchernaenko et al. Reference Tchernaenko, Halvorson and Lutter2004) and found to be in agreement with Manning's counterions condensation model (Manning, Reference Manning1978).

1.7.1 The A₄T₄vs. T₄A₄ puzzle

Following the pioneering study of Hagerman (Reference Hagerman1986) showing that sequences of the form d(CA₄T₄G) are bent whereas sequences of the form d(CT₄A₄G) are straight, many studies tried to understand the molecular basis of these observations. Burkhoff & Tullius (Reference Burkhoff and Tullius1988) analyzed the footprinting pattern generated by cutting DNA with hydroxy radicals. These authors observed that the cleavage pattern of sequences containing the A₄T₄ motif varied sinusoidally with position (Burkhoff & Tullius, Reference Burkhoff and Tullius1988). The frequency of cutting the DNA strands decreased going from the 5′ end to the 3′ end of each adenine tract (Burkhoff & Tullius, Reference Burkhoff and Tullius1988). In contrast, the pattern of hydroxyl radical footprinting in T₄A₄ oligomer is B-DNA like and, hence, does not resemble the cutting pattern of bent DNA (Burkhoff & Tullius, Reference Burkhoff and Tullius1988). These experiments also revealed that each adenine is in dissimilar surroundings. Park & Breslauer (Reference Park and Breslauer1991) observed a premelting transition, characteristic of A-tracts, in DNA molecules containing the A₄T₄ motif but not in DNA molecules containing the T₄A₄ motif, as discussed above.

Leroy et al. (Reference Leroy, Charretier, Kochoyan and Gueron1988) measured the base-pair lifetimes in DNA fragments containing a series of asymmetric A-tracts, from A₂ to A₆. In addition, they studied the two series of symmetric tracts, A_1–6T_1–6 and T_3–6A_3–6. Leroy et al. (Reference Leroy, Charretier, Kochoyan and Gueron1988) observed anomalously long base-pair lifetimes whenever the A-tracts had at least four consecutives adenines or included the A–T step, but not the T–A step. Therefore, A₂T₂ behaves similarly to A₄ as regards the base-pair lifetime, but not T₂A₂. These authors showed that A⋅T lifetimes in A-tracts are 10–100 times longer than in B-DNA (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988). These studies were carried out at high NH₄⁺ concentrations (>1 m). Similar results were obtained by Moe and Russo (Reference Moe and Russo1990). More recently, Leroy and co-workers showed that A⋅T lifetimes at low NH₄⁺ concentrations (<30 mm) are much longer (Snoussi & Leroy, Reference Snoussi and Leroy2002). They are 4–5 times longer in DNA fragments containing the A₂T₂ motif and 2–3 times longer in DNA fragments containing the A₄T₄ motif. Thus, at low NH₄⁺ concentrations, the structure of sequences containing the A_nT_n motif is even more different than B-DNA, with regards to base-pair lifetimes. A correlation between slow base-pair kinetics and anomalous migration on PAG has been established (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988).

Zhurkin et al. (Reference Zhurkin, Ulyanov, Gorin and Jernigan1991) carried out Monte Carlo (MC) simulations of decameric duplexes containing either the asymmetric A₅ runs or the symmetric A₄T₄ and T₄A₄-tracts. The simulation revealed that the minor groove width of the DNA fragments mirrored the results of hydroxy radical footprinting (Burkhoff & Tullius, Reference Burkhoff and Tullius1988). Lavery and co-workers (Sanghani et al. Reference Sanghani, Zakrzewska, Harvey and Lavery1996) modeled the two symmetric runs, A₄T₄ and T₄A₄, with the program JUMNA. The modeling reproduced the experimental data with A₄T₄ having a bend of 26° per decamer and a radius of curvature of 62 Å, whereas T₄A₄ had a radius of curvature of ~240 Å and 8° bend per decamer (Sanghani et al. Reference Sanghani, Zakrzewska, Harvey and Lavery1996). The minor groove width (calculated using the curve algorithm; Lavery & Sklenar, Reference Lavery and Sklenar1988; Stofer & Lavery, Reference Stofer and Lavery1994) correlated well with the hydroxy radical cleavage pattern (Burkhoff & Tullius, Reference Burkhoff and Tullius1988). Moreover, the molecular origin of the difference between A₄T₄ and T₄A₄ was investigated with Molecular dynamics (MD) simulations by Beveridge and co-workers (Sprous et al. Reference Sprous, Young and Beveridge1999). The simulations were carried out with counterions at concentrations comparable to a ligase buffer solution and started from identical straight canonical B-DNA structure. In the simulations the T₄A₄ structure exhibited less curvature than the A₄T₄ structure (Sprous et al. Reference Sprous, Young and Beveridge1999). The authors attributed the lesser curvature of the T₄A₄ structure to three steps (T–A, then C–G, and then T–A again) that were positioned half a helical turn apart from each other, such that their contribution to curvature cancelled out each other.

NMR spectroscopy, including refinement by dipolar coupling, was used to study the solution structure of d(GCAAAATTTTGC) and d(CGTTTTAAAACG) (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004). Stefl et al. (Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004) attributed the difference in bending of the two sequences to differences in the T–A vs. the A–T steps. The T–A step has a positive roll (11°), with a bend towards the major groove. The A–T step has a negative roll (−12°), with a bend towards the minor groove. The authors suggested that the two local roll bends contribute significantly to the global bend because they are out of phase (T–A) or in phase (A–T) with roll bends at the junction of the A₄ blocks with C⋅G base pairs (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004). In the ‘tail-to-tail’ arrangement of the two A₄ blocks in A₄T₄, there is a deflection of the two helical axes to produce a net curvature, whereas the ‘head-to-head’ arrangement of the two A₄ blocks in T₄A₄ produces a zigzag helical axis. To visualize the combined effect of these motifs, they created a sequence that was repeated multiple times in phase with the helical repeat. They found that A₄T₄ made a left-handed superhelix with 110 Å diameter and a helical pitch of 80 Å, whereas T₄A₄ formed a left-handed superhelix with a smaller diameter but a larger pitch (>250 Å). Thus, T₄A₄ has a gentle writhe and can go through PAG more easily. These superhelices arise because for both A₄T₄ and T₄A₄ the global bend within each repeat is not planar.

The structure of a 16-bp DNA oligomeric sequence containing T₃A₃-tract, d(CGAGGTTTAAACCTCG)₂, was investigated by NMR spectroscopy with residual dipolar coupling (McAlteer et al. Reference McAlteer, Gaona, Michalczyk, Buchko, Isern, Silks, Miller and Kennedy2004). Each half of the 16-bp oligomer was found to be significantly bent; however, the direction of bending in the two halves were in opposite directions (McAlteer et al. Reference McAlteer, Gaona, Michalczyk, Buchko, Isern, Silks, Miller and Kennedy2004). The outcome is a cancellation of overall bending, such that the molecule on the whole is nearly straight. The pattern of minor groove width was the expected one – narrow at the ends of the A₃-tracts and most wide at the T–A step (McAlteer et al. Reference McAlteer, Gaona, Michalczyk, Buchko, Isern, Silks, Miller and Kennedy2004). The inclination parameter mirrored the minor-groove width pattern, being maximal at the T–A step and minimal at the ends of the A₃-tracts. The roll and tilt angle were essentially zero throughout the A₃-tract. This is in contract to the NMR structure of T₄A₄ (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004).

1.7.2 The effect of A-tracts on B-DNA structure in flanking sequences

Initial insight into the dominant nature of A-tract structure and its influence on the sequences flanking it was provided by the study of Haran et al. (Reference Haran, Kahn and Crothers1994) that showed that changing the B-DNA sequences between phased A-tracts modulate the overall bending of the molecules by only ±10%. This implies that curvature arises from the A-tract region and validates the A-tract junction model (Koo et al. Reference Koo, Wu and Crothers1986; Koo & Crothers, Reference Koo and Crothers1988; Crothers et al. Reference Crothers, Haran and Nadeau1990). The polarity and non-uniformity of A-tract structure was clearly observed in the electrophoretic study of Merling et al. (Reference Merling, Sagaydakova and Haran2003). Merling et al. (Reference Merling, Sagaydakova and Haran2003) studied the interrelationship between G-tracts (Gt) and A-tracts. G-tracts are DNA blocks of the general form CnGm or GmCn (5⩽n+m⩽6), known to be a structural element that is cooperatively built by the whole pentamer or hexamer (Merling et al. Reference Merling, Sagaydakova and Haran2003). This study showed that, as has been previously observed, G-tracts positioned between phased A-tracts have minimal influence on the composite curvature (Merling et al. Reference Merling, Sagaydakova and Haran2003). This is true also when G-tracts are positioned at the 3′ end of A-tracts (i.e. sequences of the form A₆Gt₅N₁₀) (Merling et al. Reference Merling, Sagaydakova and Haran2003). However, when G-tracts are positioned at the 5′ end (as in Gt₅A₆N₁₀) or are one helical repeat away from A-tracts (as in Gt₅N₅A₆N₅), there is a more significant effect of the G-tracts on the composite bend.

The observation that any 5-bp sequence positioned between two A-tracts show the same mobility pattern on PAG, whereas these sequence elements modulate the composite bend angle to different degrees and directions when removed from A-tracts or when they are found only at the 5′ side of A-tracts, suggests that the structural properties of these sequences are suppressed by the dominant structure of the adjacent A-tracts. The observation that the 3′ side of A-tracts is more suppressive than the 5′ side is related to the non-uniformity of A-tract structure and in particular to the closing down of the minor groove in the 5′ to 3′ direction (Burkhoff & Tullius, Reference Burkhoff and Tullius1987; Kintanar et al. Reference Kintanar, Klevit and Reid1987; Katahira et al. Reference Katahira, Sugeta, Kyogoku, Fujii, Fujisawa and Tomita1988; Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Nadeau & Crothers, Reference Nadeau and Crothers1989; Moe & Russo, Reference Moe and Russo1990). The effect of A-tracts on the neighboring sequence extends at least 5 bp into the adjoining 3′ region (Merling et al. Reference Merling, Sagaydakova and Haran2003), such that these regions do not show their true intrinsic properties. This demonstrates that DNA structure is complex and cannot be adequately described by nearest-neighbor models.

The effect of A-tracts on their flanking sequences was also observed in the structure of TATA boxes, the DNA sequence element bound by the TATA binding protein (TBP). Haran and co-workers (Bareket-Samish et al. Reference Bareket-Samish, Cohen and Haran2000; Faiger et al. Reference Faiger, Ivanchenko, Cohen and Haran2006) studied the influence of sequences flanking TATA boxes on TBP/TATA box interactions. The half-life of TBP/TATA box complexes was dependent on the identity of sequences flanking TATA boxes from either side, when the TATA box was of the form TATATATA but not when the TATA box was of the form TATAAAAG (Bareket-Samish et al. Reference Bareket-Samish, Cohen and Haran2000; Faiger et al. Reference Faiger, Ivanchenko, Cohen and Haran2006).

What do we currently know about A-tract structure?

Below we summarize the current knowledge about A-tract structure and the molecular basis for global DNA curvature.

1. 1. The structure of the A–A step is context-dependent. Isolated A–A steps have different structure than A–A steps incorporated into A-tracts. The structure of A–A steps changes within A-tracts as one move from the 5′ end to the 3′ end. Thus, nearest-neighbor models are not useful for describing A-tract structure.

2. 2. Moreover, the formation of A-tract structure is a cooperative and a long-range phenomenon. (The structure of a base-pair step can be context-dependent without cooperativity in the formation of a DNA motif.)

3. 3. The bases within the A-tract regions are negatively inclined relative to a global helical axis, whereas the bases in the neighboring B-DNA regions are positively inclined. To preserve base stacking, there is buckling of the base pairs at the two junctions between A-tract and B-DNA regions. Relative to local helical axes, the bases have roll/tilt wedges, which have negative values within A-tracts and positive values in the B-DNA regions. The global bend is towards the minor groove at the center of the A-tract region.

4. 4. The bases are propeller twisted. The bases are usually twisted to an extent permitting the formation of bifurcated hydrogen bonds in the major groove, especially when the A_n-tract is longer than four adjacent adenines.

5. 5. The minor groove in A-tracts is narrow. Narrow minor groove has been observed in other A,T-rich sequences, but only in asymmetric A_n-tracts (i.e. no A–T steps) does the minor groove get progressively narrower going in the 5′ to 3′ direction. The minimal width of the minor groove is reached at tracts of A₇ and longer (Nadeau & Crothers, Reference Nadeau and Crothers1989). In symmetric A_nT_n-tracts, the narrowest minor groove is in the middle of the A_nT_n-tract. Spine of hydration is observed whenever the minor groove is narrow and not only in A-tract regions. Thus, it is not a stabilizing factor unique for A-tracts.

6. 6. B-DNA regions adjoining A-tracts on their 3′ end are in a constrained conformation, which is not the intrinsic structure that these sequences will adopt when they are further away from A-tracts or adjoining them on the 5′ side of A-tracts (Merling et al. Reference Merling, Sagaydakova and Haran2003). The A–T and T–A steps, found in the border between two A-tracts (i.e. in A_nT_n and T_nA_n) are also not in the same conformation as when in a different environment (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004).

7. 7. Both asymmetric A_n-tracts (n⩾4) as well as symmetric A_nT_n-tracts (n⩾2) can cause global helix bending when repeated with the helical repeat, but the details of the molecular basis of bending are not always identical. For example, in the solution structure of A₆-tract (MacDonald et al. Reference MacDonald, Herbert, Zhang, Polgruto and Lu2001), the inclination of the two strands together gets progressively more negative going in the 5′ to 3′ direction until the region beyond the 3′ end of the A-tract, where the inclination of the bases changes direction and becomes positive. In the A₂T₂ crystal structure (Hizver et al. Reference Hizver, Rozenberg, Frolow, Rabinovich and Shakked2001), the inclination is minimal at the A₂T₂ center and increases to both ends. On the other hand, the buckle in the A₆ structure is at a minimum at the second adenine of the A-tract and increases to either end. However, in the A₂T₂ structure, within the A₂T₂ region, it starts with positive values and gets progressively more negative going 5′ to 3′ (Hizver et al. Reference Hizver, Rozenberg, Frolow, Rabinovich and Shakked2001), as well as along the entire sequence in the A₄T₄ structure (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004). Buckle and inclination are related helical parameters, as inclination is the average of the individual inclination of the two strands, whereas buckle is the difference in individual inclination of the two strands (Lavery & Zakrzewska, Reference Lavery, Zakrzewska and Neidle1999). Thus, two different but related mechanisms result in similar global effects.

8. 8. Contrary to sequences containing the symmetric A_nT_n motif that cause global DNA bending when repeated in phase with the helical repeat, those containing the T_nA_n motif are straight. In comparative studies by both NMR (Stefl et al. Reference Stefl, Wu, Ravindranathan, Sklenar and Feigon2004) and molecular dynamics simulations (Sprous et al. Reference Sprous, Young and Beveridge1999), the bending in A₄T₄ was attributed to in-phase addition of roll angles at the center (i.e. at the A–T step) with roll angle half a helical repeat away on either side. In T₄A₄, there are out-of-phase cancellations of similarly positioned base-pair steps. The minor groove in A₄T₄ gradually narrows towards the central A–T step, which produces a sinusoidal pattern of closing down and widening of the minor groove, similar to that in homopolymeric A_n-tracts. In T₄A₄, the minor groove is at its maximal width both at the central T–A step and in the B-DNA regions, such that the modulation of the groove width is thus less pronounced. It should be noted, however, that the groove width is only the effect and not the cause of bending. It is the different base-stacking patterns in both oligomers, manifested by changes in helical parameters (especially inclination and propeller twist), that produce a different pattern of minor-groove width variation.

Current models for A-tract structure

In this section, we will limit ourselves to a discussion of homopolymeric, asymmetric A_n-tracts, as the mechanism of bending in symmetric A_nT_n may be different in details, as discussed above. Nearest-neighbor interactions may be inadequate to describe DNA structure in general (Faiger et al. Reference Faiger, Ivanchenko and Haran2007); however, they are certainly not adequate for a description of A-tract structure in particular, since it have been shown that A-tract structure is cooperatively formed (Leroy et al. Reference Leroy, Charretier, Kochoyan and Gueron1988; Haran & Crothers, Reference Haran and Crothers1989; Nadeau & Crothers, Reference Nadeau and Crothers1989). Currently, a synthesis of the two junction models seems to be the best description for A-tract bending. From a global viewpoint, bending is a result of a change in inclination between the two regions – negative inclination within A-tracts and positive inclination within B-DNA segments. To accommodate this change in inclination and preserve base-pair stacking interactions, there is a change in buckle at the two junctions. Thus, the current description of A-tract bending is in accord with the delocalized bending model, first described by Crothers & Shakked (Reference Crothers, Shakked and Neidle1999), which describes bending as being spread out along the whole molecule.

In the NMR structure of an A₆-tract (MacDonald et al. Reference MacDonald, Herbert, Zhang, Polgruto and Lu2001), the buckle and inclination helical parameters have extreme values one base pair into the B-DNA region at the 3′ end. The buckle is at its maximum value (13·5°) there, starting from a minimal value (−11°) at the second adenine in the A-tract. The average inclination, and more so the inclination of the adenine strand, is at its minimal value (−12°) there. This may be part of the explanation of why it is easier to influence A-tract structure at its 5′ end and why it seems that A-tract structural properties ripples into the adjacent B-DNA region at the 3′ end (Merling et al. Reference Merling, Sagaydakova and Haran2003).

High propeller twist values, with concomitant bifurcated hydrogen bonds, were observed in almost all asymmetric A_n-tract longer than an A₄-tract. Indeed, the highest propeller twist value in the A₆-tract structure (MacDonald et al. Reference MacDonald, Herbert, Zhang, Polgruto and Lu2001) appears on the sixth adenine. It could be that high propeller twist and bifurcated hydrogen bonds are more crucial for the stabilization of longer A-tracts. Spine of hydration does not seem to be a prerequisite for stabilization of A-tract structure, but a consequence of the narrow minor groove.

2.3.1 A-tracts in intergenic regions in prokaryotes

In 1993, a third bacterial promoter element was discovered and added to the known −10 and −35 elements, called the UP element (Ross et al. Reference Ross, Gosink, Salomon, Igarashi, Zou, Ishihama, Severinov and Gourse1993). It is located at position −38 to −59 upstream to the transcription start site (TSS, Ross et al. Reference Ross, Gosink, Salomon, Igarashi, Zou, Ishihama, Severinov and Gourse1993) and has the consensus sequence NNAAAWWTWTTTTNNAAANNN (Estrem et al. Reference Estrem, Gaal, Ross and Gourse1998). Thus, the sequence contains A-tracts. The consensus sequence (which can be decomposed to a proximal subsite and to a distal subsite) is recognized by the two C-terminal domains of the α-subunit (α-CTD) of bacterial RNA polymerase (RNAP, Ross et al. Reference Ross, Gosink, Salomon, Igarashi, Zou, Ishihama, Severinov and Gourse1993). Stimulation of transcription was shown to be α-CTD-dependent (Aiyar et al. Reference Aiyar, Gourse and Ross1998). When artificial phased A-tracts were substituted for the natural UP element of the rrnB P1 promoter, activation of transcription was not correlated with the extent of bending, but with similarity to the natural UP element and to the similarity to the natural distance from the −35 element (Aiyar et al. Reference Aiyar, Gourse and Ross1998). In particular, the upstream region of the rrnB P1 promoter contains many A-tracts and it had been put forward as a prime example for the effect of global intrinsic bending on transcription (reviewed by Perez-Martin et al. Reference Perez-Martin, Rojo and De Lorenzo1994). In addition to the UP element, at position −38 to −59 that interacts directly with α-CTD of RNAP, there are three A/T-rich FIS elements centered around −70, −100 and −140 in the rrnB P1 promoter (Gaal et al. Reference Gaal, Rao, Estrem, Yang, Wartell and Gourse1994). The α-CTD/UP element interaction accounts for ×30-fold in transcription activation, whereas binding to the FIS elements accounts for an additional ×10-fold activation (Gaal et al. Reference Gaal, Rao, Estrem, Yang, Wartell and Gourse1994). However, FIS binding at the −70 region is responsible for ~80% of its total effect, whereas the large bend is associated with the −100 binding site (Gaal et al. Reference Gaal, Rao, Estrem, Yang, Wartell and Gourse1994).

α-CTD interacts with the DNA double helix mainly through the minor groove, and it recognizes a degenerate consensus containing only A and T base pairs (Ross et al. Reference Ross, Gosink, Salomon, Igarashi, Zou, Ishihama, Severinov and Gourse1993; Estrem et al. Reference Estrem, Gaal, Ross and Gourse1998, Reference Estrem, Ross, Gaal, Chen, Niu, Ebright and Gourse1999; Gourse et al. Reference Gourse, Ross and Gaal2000). It has been proposed that α-CTD of RNAP recognizes the narrow minor groove characteristic of A (and T) regions (Gourse et al. Reference Gourse, Ross and Gaal2000). Contacts to the major groove do not significantly contribute to α-CTD/UP element interactions (Gourse et al. Reference Gourse, Ross and Gaal2000). We suggest that α-CTD may recognize the asymmetric narrowing of A-tracts going 5′ to 3′, characteristic of only A-tracts and not any A/T-rich region (Nadeau & Crothers, Reference Nadeau and Crothers1989). This suggestion is corroborated by the NMR study of Yasuno et al. (Reference Yasuno, Yamazaki, Tanaka, Kodama, Matsugami, Katahira, Ishihama and Kyogoku2001). Yasuno et al. (Reference Yasuno, Yamazaki, Tanaka, Kodama, Matsugami, Katahira, Ishihama and Kyogoku2001) measured DNA binding affinity of α-CTD by NMR titration curves (¹⁵N chemical shift differences for Leu295, as a function of DNA concentration) and observed an apparent correlation between high K _d values and sequences exhibiting high degree of intrinsic bending. However, a closer look at the sequences shows that the correlation is actually between the number of adjacent A's and K _d values. The largest difference in K _d values measured by Yasuno et al. (2001) is between the consensus UP element containing the A₄T₈ motif, which has the highest binding affinity to α-CTD, and a sequence with the A₆ motif, which binds α-CTD ~×10-fold weaker than the UP element. However, the A₆ motif confers a much higher bend angle on the double helix (~18°) than A₄T₈, which is essentially a straight DNA molecule. A₄T₈ is similar in sequence to the A₁₂ motif that was shown to migrate as a straight DNA molecule in polyacrylamide gels (Haran & Crothers, Reference Haran and Crothers1989). On the other hand, as A-tracts are lengthened, their minor groove becomes progressively narrower, 5′ to 3′, as discussed above. As the minor groove becomes narrower, the sugar phosphate backbone of the two strands comes closer together and then the fit of Arg265 from the α-CTD of E. coli RNAP in the minor groove becomes more intimate, as it can interact with both sides of the DNA backbone simultaneously (Katayama et al. Reference Katayama, Matsushita, Tamai, Miyata and Okabe2001). This, and not global intrinsic bending per se, is the reason for the high binding affinity of the consensus UP element to α-CTD.

Cellai et al. (Reference Cellai, Mangiarotti, Vannini, Naryshkin, Kortkhonjia, Ebright and Rivetti2007) have recently shown that the extent of compaction of DNA by RNAP depends not only on the presence of α-CTD of RNAP and the α-linker but also on the sequence. They suggest that intrinsic DNA bending contributes to compaction (Cellai et al. Reference Cellai, Mangiarotti, Vannini, Naryshkin, Kortkhonjia, Ebright and Rivetti2007). However, the highest degree of compaction (30±0·4 nm) is achieved with the λ Pr promoter (no A-tracts) and is least (2±0·5 nm) when the proximal subsite of the UP element (with an A₆ motif) is inserted in the lacUV5 promoter. Compaction is also significant with the lacUV5 promoter containing the full UP element (21±0·7 nm), which contains the straight segment A₃T₉. It should be noted, however, that the A₃T motif is found in both the λ Pr promoter and in the lacUV5(UP^full) promoter at an identical position, and thus, similar protein contacts can be made in these two promoters.

Curved DNA regions were also observed in termination regions in prokaryotes (Hosid & Bolshoy, Reference Hosid and Bolshoy2004; Kozobay-Avraham et al. Reference Kozobay-Avraham, Hosid and Bolshoy2006), but no definite role has been ascribed to these regions. It has been suggested (Hosid & Bolshoy, Reference Hosid and Bolshoy2004; Kozobay-Avraham et al. Reference Kozobay-Avraham, Hosid and Bolshoy2006) that they are either binding sites for RNAP or to other proteins, such as the histone-like protein H-NS.

2.3.2 A-tracts and DNA supercoiling

Overwinding or underwinding DNA causes it to coil in three dimensions, DNA supercoiling. The energy stored as mechanical stress, caused by DNA supercoiling, is involved in many aspects of cellular function, such as transcription, replication and recombination. A-tracts were shown to be localized at or near the apexes of supercoiled domain (Laundon & Griffith, Reference Laundon and Griffith1988; Pfannschmidt & Langowski, Reference Pfannschmidt and Langowski1998) and can thus globally orient supercoiled DNA domains, as they can fix the relative orientation of two DNA segments that are widely separated in one dimension. However, Tsen & Levene (Reference Tsen and Levene1997) have shown that it is not sequence-directed intrinsic curvature per se that directs A-tracts to apexes of superhelical domains, since both properly phased A-tracts [(A₅N₅)₅] and unphased A-tracts [(A₅N₁)₅] promoted the localization of A-tracts to the terminal loops of superhelical domains (Tsen & Levene, Reference Tsen and Levene1997). Thus, it is a structural property unique to A-tracts that forces them to prefer terminal loops of superhelical domains.

Moreover, A-tracts were shown to dampen the internal motions of supercoiled domains, such as slithering (or reptation) of the two DNA double strands relative to each other (Kremer et al. Reference Kremer, Klenin, Diekmann and Langowski1993; Diekmann & Langowski, Reference Diekmann and Langowski1995). This reduced relative motion of the two DNA strands in supercoiled domains, afforded by the introduction of A-tracts (whether phased or unphased), is the important contribution of A-tracts to transcriptional regulation via supercoiling, as it aid in fixing the relative orientation of the two strands and preventing their migration away from each other. Both A-tracts and random sequence DNA when localized to a superhelical terminal loop are in a constrained conformation relative to their conformation free in solution and both are significantly more bent than in their unconstrained conformation (Pavlicek et al. Reference Pavlicek, Oussatcheva, Sinden, Potaman, Sankey and Lyubchenko2004).

2.3.3 A-tracts and DNA looping

DNA looping is an essential mechanism in transcriptional regulation, DNA replication and in recombination (Matthews, Reference Matthews1992; Schleif, Reference Schleif1992). DNA looping can bring together DNA regions that are hundreds or thousands base pair apart. A priori, DNA looping can be stabilized by protein-induced DNA bending or by intrinsic DNA curvature or be a result of DNA supercoiling. However, it is easier to regulate protein-induced DNA looping than looping induced by A-tracts, because looping mediated by the binding of a protein is a more fully reversible process. A sequence element, such as an A-tract, which confers static bending on the DNA molecule, is continuously present in the genome. In addition, especially in eukaryotes, the presence of extended A-tract regions, especially when phased with respect to the helical repeat, can interfere with other A-tract assisted interactions, such as DNA packaging in nucleosomes (see below).

Several groups (Carmona et al. Reference Carmona, Claverie-Martin and Magasanik1997; Cheema et al. Reference Cheema, Choudhury and Das1999) reported an apparent effect of the presence of intrinsically bent regions (containing A-tracts) on transcription activation in promoters that do not contain binding sites for proteins, such as IHF, that bend DNA. Cheema et al. (Reference Cheema, Choudhury and Das1999) studied transcription activation from the nifLA promoter of Klebsiella pneumoniae. In this promoter, RNAP complexed with σ⁵⁴ factor (RNAP-σ⁵⁴) has to interact with the NtrC activator by DNA looping for transcription to occur (Cheema et al. Reference Cheema, Choudhury and Das1999). Cheema et al. (Reference Cheema, Choudhury and Das1999) have shown, by gel electrophoresis migration anomaly and electron microscopy, that the region between the NtrC binding site and the nifLA promoter of Klebsiella pneumoniae is intrinsically bent, due to the presence of two A-tracts and one T-tract. Mutations in these regions, or their deletion, resulted in a decrease both in DNA curvature and in in vivo transcription from the promoter of the same order of magnitude (Cheema et al. Reference Cheema, Choudhury and Das1999). However, later studies suggested that these effects may be due to the deletion or alteration of cryptic subsidiary NtrC binding sites (Lilja et al. Reference Lilja, Jenssen and Kahn2004). The glnAp2 promoter of E. coli and Salmonella typhinum is a prototype for looping mediate transcription activation (Rombel et al. Reference Rombel, North, Hwang, Wyman and Kustu1998) and supercoiling has been shown to be important for transcription activation in glnAp2 (Carmona et al. Reference Carmona, Claverie-Martin and Magasanik1997; Liu et al. Reference Liu, Bondarenko, Ninfa and Studitsky2001). Schulz et al. (Reference Schulz, Langowski and Rippe2000) studied transcription from the glnAp2 promoter and reaffirmed the importance of supercoiling for transcription activation in this system. However, replacing DNA sequences, between the NtrC binding sites and the promoter, with intrinsic bent sequences (A-tracts), positioning them in either a symmetric or asymmetric location relative to the NtrC site and the promoter, did not lead to any systematic results (Schulz et al. Reference Schulz, Langowski and Rippe2000). One supercoiled plasmid containing symmetrically positioned A-tracts and shown to migrate anomalously on PAG showed significant increase in transcription activation, whereas other constructs with similarly placed curved fragments did not. Moreover, one supercoiled plasmid containing end-positioned A-tracts (near the promoter) and shown not to migrate anomalously on PAG did stimulate transcription. Another construct containing additional, weaker NrtC binding sites, without any curved DNA regions activated transcription to levels similar to that of the plasmid with the end-positioned A-tracts. The results have been interpreted (Schulz et al. Reference Schulz, Langowski and Rippe2000; Lilja et al. Reference Lilja, Jenssen and Kahn2004) as evidence for the importance of the deleted weaker subsidiary sites between the strong NtrC binding sites and the promoter or to fortuitous effects of supercoiling.

Moreover, the region between the main NtrC binding sites and the core glnAp2 promoter have been replaced by A-tracts having various lengths and geometries in the study of Lilja et al. (Reference Lilja, Jenssen and Kahn2004) and the effect on transcription studied. The results showed that a variety of bending geometries could sustain some transcription activation, but none were as good as that of the wild type (50% of wild type at most). This has prompted Lilja et al. (Reference Lilja, Jenssen and Kahn2004) to probe the role of the NtrC subsidiary sites in DNA looping and transcription activation in this system. Replacing them with a straight random sequence led to transcription at 45% of that of wild type, confirming their importance. Replacing the NtrC subsidiary sites with strong NtrC binding sites led to even smaller levels of activation (Lilja et al. Reference Lilja, Jenssen and Kahn2004). This shows the importance of NtrC/DNA interactions for DNA looping in this system as well as the danger of ascribing structural roles to DNA sequences containing cryptic binding sites, which as a by-product also migrated anomalously on PAG.

Ross et al. (Reference Ross, Keating and Maher2000) studied the looping model of eukaryotic transcription activation by constructing more than 30 different transcription templates. The templates had a curved region of 72°, made from tandem A-tracts, and various phasing of activator binding sites relative to the basal promoter. The results showed that there was no significant effect on in vitro transcription from either inserting a curved DNA fragment or from helical alignment of the activator binding sites relative to the basal promoter (Ross et al. Reference Ross, Keating and Maher2000).The only variable having a significant effect on transcription activation was the length of the DNA sequence between the activator binding sites and the basal promoter, for both linearized as well as supercoiled templates (Ross et al. Reference Ross, Keating and Maher2000). This study showed that looping as a mechanism for transcription activation may be a plausible one; however, the role of global intrinsic bending in this model is probably not as significant as previously thought. In this regard, it is interesting to note that a recent study on DNA flexibility showed that DNA is significantly more flexible than previously estimated (Mathew-Fenn, Reference Mathew-Fenn, Das and Harbury2008). The measurements of Mathew-Fenn et al. (Reference Mathew-Fenn, Das and Harbury2008) are contradictory to the current elastic rod model of DNA; however, they are in agreement with a recent model that avoids weak bending approximation and takes into account the discrete nature of the phosphate charges (Guo et al. Reference Guo, Taubes, Oh, Maher and Mohanty2008). Thus, models with bent DNA (especially intrinsic bending), employed to overcome the inherent rigidity of the double helix below its persistent length, are maybe less necessary than previously envisioned.

2.3.4 A-tracts and termination of DNA synthesis by reverse transcriptase in retroviruses

Another example were functional role was first attributed to intrinsic DNA bending and then to a structural property of A-tract – the narrowing of the DNA minor groove, is the study of Lavigne et al. (Reference Lavigne, Roux, Buc and Schaeffer1997, Reference Lavigne and Buc1999) on the mechanism of plus strand synthesis by HIV-1 reverse transcriptase (RT). Synthesis of the plus strand of the HIV genome is conducted by the synthesis of two separate segments. The first (called U+) initiates from the 3′ end of the minus strand and terminates at the central termination sequence (CTS), the other (called D+) starts at the c polypurine tract (cPPT) and terminates at the 5′ end of the minus strand. There is an overlap of 100 bp between the two segments. The CTS contains two termination sites, Ter1 and Ter2. The sequences preceding these sites are A-tracts (A₅T₂ and A₄T₄ respectively). Termination occurs because of two combined effects – the presence of the sequences preceding Ter1 and Ter2 and strand displacement by RT at the 100 bp overlap between U+ and D+. Initially, Lavigne et al. (Reference Lavigne, Roux, Buc and Schaeffer1997) proposed that it is intrinsic curvature that is responsible for termination of plus strand synthesis by showing that mutations abolishing curvature abolished termination as well. However, a later study (Lavigne & Buc, Reference Lavigne and Buc1999) used hydroxy radical cleavage experiments to show a correlation between sites having compressed minor groove at the 3′ end of A-tracts and termination sites. In addition, intrinsically curved DNA without A-tracts, but only G-tract sequences, did not bring about termination of DNA synthesis by RT (Lavigne & Buc, Reference Lavigne and Buc1999). Here again, it is the unique structural properties of individual A-tracts that are responsible for in vivo function, this time for termination of DNA synthesis. Each A-tract in the sequences preceding Ter1 and Ter2 is at least 7 bp long since the width of the minor groove reaches its most constricted state only from 7 bp and longer (Nadeau & Crothers, Reference Nadeau and Crothers1989).

2.3.5 A-tracts and transcription regulation by thermosensing in pathogenic bacteria

Human pathogenic bacteria, such as Shigella and enteroinvasive E. coli (EIEC), are virulent only after shifting to the host cell temperature, 37°C (Maurelli & Sansonetti, Reference Maurelli and Sansonetti1988). The change in temperature upon entering the host cell triggers their virulent genes. The genes are activated by two transcriptional regulators, coded by the virF and virB genes (Adler et al. Reference Adler, Sasakawa, Tobe, Makino, Komatsu and Yoshikawa1989). The virF promoter is repressed by H-NS in a temperature-dependent manner (reviewed by Ussery et al. Reference Ussery, Hinton, Jordi, Granum, Seirafi, Stephen, Tupper, Berridge, Sidebotham and Higgins1994; Atlung & Ingmer, Reference Atlung and Ingmer1997). H-NS represses the expression of virulent genes only below the critical temperature, 32°C (Maurelli & Sansonetti, Reference Maurelli and Sansonetti1988). Repression by H-NS is conducted by the cooperative binding of two H-NS molecules, separated by ~250 bp (Falconi et al. Reference Falconi, Colonna, Prosseda, Micheli and Gualerzi1998). Falconi et al. (Reference Falconi, Colonna, Prosseda, Micheli and Gualerzi1998) demonstrated, both in vivo and in vitro, that H-NS binds to its two binding sites at 30°C but not at 37°C and that the two sites are simultaneously occupied and are not independently bound. The region between the two H-NS sites was shown to contain A-tracts and to migrate anomalously slow on PAG (Falconi et al. Reference Falconi, Colonna, Prosseda, Micheli and Gualerzi1998). Falconi et al. (Reference Falconi, Colonna, Prosseda, Micheli and Gualerzi1998) showed that the virF promoter undergoes a temperature-dependent conformational transition at ~32°C. It was suggested that this temperature-dependent conformational change is responsible for the temperature-dependent binding of H-NS at its two binding sites (Falconi et al. Reference Falconi, Colonna, Prosseda, Micheli and Gualerzi1998). The study of Prosseda et al. (Reference Prosseda, Falconi, Giangrossi, Gualerzi, Micheli and Colonna2004) localized the center of the intrinsically curved DNA to the region between the two H-NS binding sites and showed that the position of the bend center and the magnitude of curvature changes as a function of temperature. In addition to the two H-NS binding sites, there are four FIS binding sites, two stronger sites in the region overlapping with the downstream H-NS site (H-NS I, centered around position +1) and two in the region between the two H-NS sites (Falconi et al. Reference Falconi, Prosseda, Giangrossi, Beghetto and Colonna2001). FIS is an activator of virF gene expression and an antagonist of H-NS (Falconi et al. Reference Falconi, Prosseda, Giangrossi, Beghetto and Colonna2001). Mutations in the region between the two H-NS sites that are outside the FIS binding regions and were shown to produce more severe bending, were correlated with higher repression levels, in vitro and in vivo (Prosseda et al. Reference Prosseda, Falconi, Giangrossi, Gualerzi, Micheli and Colonna2004). Thermoregulation was not altered when FIS was absent and this lead only to a twofold reduction in overall virF transcription (Prosseda et al. Reference Prosseda, Falconi, Giangrossi, Gualerzi, Micheli and Colonna2004). Inserting an additional half a helical repeat near the upstream H-NS site (H-NS II) abolished H-NS repression at any temperature (Prosseda et al. Reference Prosseda, Falconi, Giangrossi, Gualerzi, Micheli and Colonna2004). Taken together, these studies show that the intrinsically bent region between the two H-NS sites act as a thermosensor. At low temperature, A-tracts adopt their unique structure that leads to global curvature, thus facilitating the interaction between the two H-NS molecules, which then lead to repression of the virF gene. As discussed above, a cooperative change between two states, a bent one and a straight one, in A-tract-containing sequences is documented (Herrera & Chaires, Reference Herrera and Chaires1989; Chan et al. Reference Chan, Breslauer, Hogan, Kessler, Austin, Ojemann, Passner and Wiles1990, Reference Chan, Breslauer, Austin and Hogan1993, Reference Chan, Austin, Mukerji and Spiro1997; Premilat & Albiser, Reference Premilat and Albiser1997). However, one cannot rule out at present that the change observed by Prosseda et al. (Reference Prosseda, Falconi, Giangrossi, Gualerzi, Micheli and Colonna2004) can partly be attributed to changes in DNA helical repeat as the temperature is changed (Depew & Wang, Reference Depew and Wang1975). Regardless, these studies show a clear role for intrinsic bending per se in gene regulation in bacteria.