Section 23.3.4. Sequence–structure relationships in B-DNA

The simplest and first-noticed sequence-dependent deformability of the B-DNA duplex was variation in minor groove width. The first B-DNA oligomer to be solved, C-G-C-G-A-A-T-T-C-G-C-G (B1–B6), had a narrow minor groove in the central A-A-T-T region, with only ca 3.5 Å of free space between opposing phosphates and sugar rings. (It has become conventional to define the free space between phosphates as the measured minimal P–P separation across the groove, less 5.8 Å to represent two phosphate-group radii. Similarly, the measured distance between sugar oxygens is decreased by 2.8 Å, representing two oxygen van der Waals radii.) The C-G-C-G ends of the helix had the 6–7 Å opening expected for ideal B-DNA, but the situation was clouded, because the outermost two base pairs at each end of the helix interlocked minor grooves with neighbours in the crystal. Hence, the wider ends could possibly be only an artifact of crystal packing.

After 1991, the situation was clarified by the structures of several decamers [Table A23.3.1.2, Part I(c)], which stack on top of one another without the interlocking of grooves. The normal minor groove opening is ca 7 Å. Regions of four or more AT base pairs can exhibit a significantly narrowed minor groove, although such narrowing is not mandatory. This behaviour is seen with the B-DNA decamer, C-A-A-A-G-A-A-A-A-G, in Fig. 23.3.4.1. The narrowing arises mainly from the larger allowable propeller twist in AT base pairs, which displaces C1′ atoms at opposite ends of the pair in different directions, and moves the backbone chains in such a way as to partially close the groove (Fig. 23.3.4.2).

Figure 23.3.4.1| top | pdf | Structure of C-A-A-A-G-A-A-A-A-G (B107). The lower half of the helix, with -A-A-A-A-G, exhibits the narrow minor groove commonly associated with the AT region of the helix and a single zigzag spine of hydration, as was first seen in C-G-C-G-A-A-T-T-C-G-C-G (B1–B6). The upper half, with C-A-A-A-G-, has the wider minor groove of general-sequence B-DNA and two separate rows of hydrating water molecules along the two walls of the wider groove.

Figure 23.3.4.2| top | pdf | Relationship between minor groove width and propeller twist. (a) View into the minor groove of B-DNA, with base pairs seen on edge and with the sugar–phosphate backbones shown schematically as inclined ladder uprights. (b) Consequences of propeller twisting the base pairs. Glycosyl bonds connected to sugar C1′ atoms are all displaced upward in the right strand and downward in the left strand. This shifts the backbone chains as indicated by the arrows. Hence, the gap between the chains is decreased, and the minor groove is narrowed.

This is an excellent example of the concept of sequence-dependent helix deformability, rather than simple deformation. The two hydrogen bonds of an AT base pair allow a larger propeller twist but do not require it. Hence, AT regions of helix permit a narrowing of the minor groove but do not demand it. Indeed, this lesson was brought home in the most dramatic way when Pelton & Wemmer (1989, 1990) showed via NMR that a 2:1 complex of distamycin with C-G-C-A-A-A-T-T-G-G-C or C-G-C-A-A-A-T-T-T-G-C-G could exist, in which two drug molecules sat side-by-side within an enlarged central minor groove. Fig. 23.3.4.3 shows a narrow minor groove with a single netropsin molecule, and Fig. 23.3.4.4 shows a wide minor groove enclosing two di-imidazole lexitropsins side-by-side. In summary, an AT-rich region of minor groove is capable of narrowing but is not inevitably narrow, in contrast to GC-rich regions where the third hydrogen bond tends to keep the base pairs flat and the minor groove wide. The AT minor groove is potentially deformable without being inevitably deformed.

Figure 23.3.4.3| top | pdf | Structure of the 1:1 complex of netropsin with C-G-C-G-A-A-T-T-C-G-C-G (B11, B12, B87). The drug binds to the central -A-A-T-T- region of the minor groove, which is barely wide enough to enclose the nearly planar polyamide molecule. The netropsin structure can be represented by where Py is a five-membered methylpyrrole ring. An even more compact representation, useful when comparing other polyamide netropsin analogues or lexitropsins, is +=Py=Py=+, where the common cationic tails are indicated only by a plus sign, and = represents a —CONH— amide.

Figure 23.3.4.4| top | pdf | Structure of the 2:1 complex of a di-imidazole lexitropsin with C-A-T-G-G-C-C-A-T-G (B108). The drug now is represented by where Im is a five-membered imidazole ring, or again more compactly by 0=Im=Im=+. The uncharged leading amide group, characteristic of distamycins, is identified by 0. Distamycin itself would be represented in this shorthand notation by 0=Py=Py=Py=+. Reprinted from B108, copyright (1977), with permission from Excerpta Medica Inc.

Sequence-dependent bendability has been reviewed recently by Dickerson (1988a,b,c) and Dickerson & Chiu (1997). The relative bendability of different regions of B-DNA sequence is an important aspect of recognition, one that is used by countless control proteins that must bind to a particular region of double helix. Catabolite activator protein or CAP (Schultz et al., 1991; Parkinson et al., 1996), lacI (Lewis et al., 1996) and purR (Schumacher et al., 1994) repressors, γδ-resolvase (Yang & Steitz, 1995), EcoRV restriction enzyme (Winkler et al., 1993; Kostrewa & Winkler, 1995), integration host factor or IHF (Rice et al., 1996), and TBP or TATA-binding protein (Kim, Gerger et al., 1993; Kim, Nikolov & Burley, 1993; Nikolov et al., 1996; Juo et al., 1996) are all sequence-specific DNA-binding proteins that bend or deform the nucleic acid duplex severely during the recognition process. IHF in Fig. 23.3.4.5 may be taken as representative of this class of DNA-binding proteins. The bend is produced by two localized rolls of ca 60° in a direction compressing the major groove and are additive, because they are spaced nine base pairs, or roughly one turn of helix, apart. In IHF, the two helix segments flanking the bend should be straight and unbent, and this is accomplished in one segment via a six-adenine A-tract: -C-A-A-A-A-A-A-G-.

Figure 23.3.4.5| top | pdf | DNA duplex (red and blue strands) looped around IHF or integration host factor. The two subunits of the IHF duplex are green and turquoise. Two antiparallel loops of protein chain, one from each subunit, insert into the minor groove of B-DNA at the sequence C-A-A-T/A-T-T-G and produce abrupt bends via local roll angles of 60°. The two localized bends are additive because they occur one helical turn apart. All other steps have roll angles of 5° or less. The two flanking helix segments pack against the IHF dimer and must be kept straight and unbent. This is accomplished in one of the two segments by an A-tract of sequence C-A-A-A-A-A-A-G. From Dickerson & Chiu (1997). Coordinates courtesy of P. Rice.

The bending locus in IHF is C-A-A-T/A-T-T-G . It is C-G in lacI and purR repressors (Fig. 23.3.4.6), C-A = T-G in CAP (Fig. 10 of Dickerson, 1998b), and T-A in EcoRV, γδ-resolvase and TBP (Fig. 23.3.4.7). Pyrimidine-purine or Y-R steps appear to be especially suitable loci for roll bending. The dashed lines in Figs. 23.3.4.6 and 23.3.4.7 plot tilt, and demonstrate its insignificance in bending, compared with roll. (This is intuitively obvious. Imagine yourself standing near a tall stack of wooden planks in a lumberyard during an earthquake. Where would you prefer to stand: alongside the stack, or at one end?)

Figure 23.3.4.6| top | pdf | Roll-angle plots for sequence-specific DNA–protein complexes with lacI (top) and purR (bottom). In each case, bending occurs via localized roll at a C-G step. Other steps of the sequence have random rolls of ca 10° or less. Note that, as with IHF, A-tracts are especially straight and unbent. Dashed lines in the lacI plot demonstrate the unimportance of tilt in production of helix bending.

Figure 23.3.4.7| top | pdf | Bending via roll at T-A steps in TBP or the TATA-binding protein (top) and in γδ-resolvase (bottom). Note that not every T-A step in TBP or γδ-resolvase is necessarily bent. Note also in γδ-resolvase that C-A = T-G steps, which in proteins such as CAP are used to generate sharp roll bends, here, frequently, are local roll maxima, even though they contribute little to the overall bending. They have a bending potential that is not used in this particular setting.

In summary, bending of the B-DNA helix nearly always involves roll, not tilt. The easier direction of bending is that which compresses the broad major groove, although examples of roll compression of the minor groove are known. Y-R steps are especially prone to roll bending. Again, the phenomenon is one of sequence-induced bendability, not mandatory bending. No one imagines that the IHF binding sequence of Fig. 23.3.4.5 is permanently kinked at its two C-A-A-T/A-T-T-G steps, wandering deformed through the nucleus, looking for an IHF molecule to bind to. Instead, this sequence has a potential bendability that other sequences, such as A-A-A-A-A-A, lack.

Table 23.3.4.1 summarizes the observed behaviour of Y-R, R-R and R-Y steps from a great many X-ray crystal structure analyses, with and without bound DNA. In the present context, these rules are termed the `Major Canon', since they are well established and generally well understood. Some understanding of the proneness of Y-R steps to bend can be obtained by looking at stereo pairs of two successive base pairs viewed down the helix axis. Fig. 23.3.4.8 gives a few representative examples; many more can be found in Figs. 4–6 of Dickerson (1988b) and in the original literature. In brief, Y-R steps, especially C-A and T-A, tend to orient so that polar exocyclic N and O atoms stack against polarizable rings of the other base pair. This is the same type of polar-on-polarizable stacking stabilization mentioned earlier in connection with O4′ and guanine in Z-DNA (Bugg et al., 1971; Thomas et al., 1982; Hunter & Sanders, 1990; B32). Base pairs in T-A steps tend not to slide over one another along their long axes, keeping pyrimidine O2 stacked over the purine five-membered ring (Fig. 23.3.4.8b). C-A steps can adopt this same stacking, or the base pairs can slide until the pyrimidine O2 sits over the purine six-membered ring instead (Fig. 23.3.4.8a).

Figure 23.3.4.8| top | pdf | Representative base-pair steps from B-DNA single-crystal X-ray analyses. (a) Pyrimidine-purine C-A step from C-C-A-A-G-A-T-T-G-G (B22, B46) (roll/slide/twist = −7.4°/2.6 Å/49.9°). Note the lack of ring-on-ring stacking, replaced by the stacking of pyrimidine O2 and purine N6 or O6, on aromatic rings of the adjacent base pair. This stacking opens up the twist angle to an unusual 50°. Note also the large +2.6 Å slide, which positions pyrimidine O2 over the six-membered rings of the neighbouring purines. (b) Pyrimidine-purine T-A step from C-G-A-T-A-T-A-T-C-G (B62) (roll/side/twist = 3.8°/−0.2 Å /39.5°). The stacking is similar to C-A, except that a near-zero slide positions pyrimidine O2 over the five-membered rings of purines. (c) Purine-purine A-A step from C-C-A-A-C-G-T-T-G-G (B46, B50) (roll/slide/twist = 8.8°/0.5 Å /28.7°). Ring-on-ring overlap now predominates, with consequently lowered twist angle and essentially zero slide. Note that purines are more extensively stacked than pyrimidines, which appear to be approaching the O2-on-ring stacking of Y-R steps. (d) Purine-pyrimidine A-T step from C-G-A-T-A-T-A-T-C-G (B62) (roll/slide/twist = 5.2°/0.0 Å/25.2°). Ring-on-ring stacking again lowers the twist angle and keeps slide around zero. Now there is no stacking of exocyclic N or O on neighbouring rings.

Purine-purine or R-R steps behave quite differently (Fig. 23.3.4.8c). They stack ring-on-ring, usually with greater overlap on the purine end than the pyrimidine. The net effect is that the pivot appears to pass through or near the purines, while pyrimidines at the other end of the pairs stack O2-on-ring as with Y-R steps. R-Y steps tend to stack ring-on-ring, with little contribution from exocyclic atoms.

El Hassan & Calladine (1997) have recently examined roll, slide and twist behaviour at 400 different steps observed in crystal structures of 24 A- and 36 B-DNA oligomers. The author has carried out a similar analysis of 1137 steps from 86 sequence-specific protein–DNA complexes (Dickerson, 1998a,c; Dickerson & Chiu, 1997). A striking feature is that trends in local parameters are just the same in DNA crystals and in protein–DNA complexes. The frequently invoked nightmare of `crystal packing deformations' appears to be of only minor significance. In both studies (El Hassan & Calladine, 1997; Dickerson, 1998b), roll versus slide, slide versus twist and twist versus roll plots are presented for all ten possible base-pair steps. Fig. 23.3.4.9 illustrates roll versus slide plots for two Y-R, two R-R and two R-Y steps.

Figure 23.3.4.9| top | pdf | Slide versus roll plots for six of the ten possible base-pair steps. Data points are from 971 steps in crystal structure analyses of 63 sequence-specific protein–DNA complexes. A complete set of 30 plots for slide/roll, twist/roll and slide/twist at all ten steps is to be found in Dickerson (1998b), and equivalent plots for DNA alone are given by El Hassan & Calladine (1997). Y-R steps exhibit a broad range of roll, slide and twist values, with roughly linear correlations between pairs of variables. Points for A-A and other R-R steps cluster tightly around the origin, showing little tendency toward roll bending. Curiously enough, points for R-Y steps tend to favour negative values of slide and twist, and, hence, to concentrate in the lower left quadrant of a slide/tilt plot.

Table 23.3.4.2 summarizes observations from these roll/slide/twist plots. These are labelled the `Minor Canon' since they are recent, approximate and not well understood. However, they provide goals for future investigations of helix behaviour.