Determination of DNA Helical Handedness by Fluorescence Resonance Energy Transfer

J. Mol. Biol. (1996) 257, 597–617

Determination of DNA Helical Handedness by Fluorescence Resonance Energy Transfer Elizabeth A. Jares-Erijman and Thomas M. Jovin* Department of Molecular Biology, Max Planck Institute for Biophysical Chemistry P.O. Box 2841, D-37018 Go¨ttingen, Germany

Fluorescence resonance energy transfer (FRET) has been used to determine the helical handedness, twist and rise of different DNA conformations. The approach is based on the construction of a set of molecules consisting of two fused helical segments, one of which is in a known reference helical form. The duplexes are covalently labeled at one end with a donor and at the other with an acceptor. By systematically shifting the position of the junction while maintaining constant the total length in base-pairs, the variation in the efficiency of energy transfer can be shown to depend primarily and sensitively on the differences in helical twist and rise of the two constituent segments. If the latter have the same helical sense, one predicts a FRET signal that is a monotonic function of the junctional position. In contrast, a periodic function arises when two segments are of opposite handedness. The formalism includes explicit consideration of dye orientation (the dipole-dipole orientation factor k) and an implementation valid for single helix molecules, and introduces new functions of measured fluorescence signals for establishing the FRET efficiency. The method has been applied to a family of oligonucleotides forming hairpin duplexes containing an antiparallel-stranded (aps) d(m5C·G)m segment labeled at the 5' end with fluorescein (donor) and a second parallel-stranded d(A·T)N-m segment (psAT-DNA) labeled at the hairpin loop with the sulfoindocyanine dye Cy3. The segment lengths were in the range 4 to 12, but the total length N was maintained constant at 16. The d(m5C·G) sequence was chosen due to its capacity for adopting a B or a Z conformation at low and high concentrations of salt, respectively. The parallel-stranded d(A·T) sequence served as the second segment in order to determine the helical rise and twist of psAT-DNA, presumed to be right-handed from molecular modeling and a prior study of topologically constrained DNA. A Z-DNA/ps-DNA junction was created between the two segments by inducing a B-Z transition in d(m5C·G)m with MgCl2 . The range of required salt concentration was established by circular dichroism measurements. FRET efficiency values of 0.38 to 0.41 were obtained for the oligonucleotides with the d(m5C·G) segment in the B conformation. In contrast, upon induction of the B-Z transition the FRET efficiency was a decreasing function of the d(m5C·G)m content (0.38 to 0.28 for m = 6 to 12). Helical parameters were estimated from functional fits of the data, and were consistent with the known properties of B- and Z-DNAs and with the conclusion that psAT-DNA has a helical rise and twist close to that of B-DNA. The approach outlined here is not restricted to DNA but can be applied to other helical structures, e.g. RNA, proteins and protein-nucleic acid complexes. 7 1996 Academic Press Limited

*Corresponding author

Keywords: parallel-stranded DNA; left-handed Z-DNA; fluorescein; Cy3; FRET

Abbreviations used: ps, parallel-stranded; aps, antiparallel-stranded; CD, circular dichroism; FRET, fluorescence resonance energy transfer; distance variables with a caret are normalized by division with Roc . 0022–2836/96/130597–21 $18.00/0

Introduction Fluorescence methods offer unique capabilities in the elucidation of the structure and dynamics of macromolecules in solution. The range of applica7 1996 Academic Press Limited

598 bility has increased dramatically due to recent advances in solid-state chemical synthesis and the ready availability of a great variety of fluorescent reagents and selective probes. There are a number of photophysical properties characterizing the generation of fluorescence: absorption and emission spectra, quantum yield, polarization, and excited state lifetime and reactions. The proximity of two fluorophores can be assessed by exploiting the process of fluorescence resonance energy transfer (FRET), inasmuch as the efficiency exhibits a very high order (sixth power) dependence on intermolecular distances of the order of 2 to 8 nm. FRET has been applied extensively to the study of nucleic acids (for reviews, see Clegg, 1992, 1995, 1996; Selvin, 1995). By covalently attaching or specifically binding the donor and acceptor moieties to specific known loci on DNA or RNA, one can (1) probe the three-dimensional arrangement of complex structures such as four-way junctions (Gohlke et al., 1994; Lilley & Clegg, 1993) and ribozymes (Tuschl et al., 1994), (2) determine DNA strand orientation (Rippe et al., 1995), and (3) devise sensitive detection methods for DNA sequencing (Ju et al., 1995; Mergny et al., 1994). However, existing FRET techniques do not in general provide information about the sense of a helix. Here, we introduce a direct method based on FRET, designed for the determination of the handedness of nucleic acid helices in solution.

Results and Discussion Helical geometry and FRET Given a helical conformation with a fluorescent donor/acceptor pair located at the ends, extension or shortening of the sequence changes the distance between the reporter molecules in a manner reflecting the helical parameters such as rise and twist. The FRET signals display a characteristic modulation as the fluorophores ‘‘walk’’ around the helix, first demonstrated in elegant experiments by Clegg et al. (1993). However, the signals inherently lack a sensitivity to the handedness of the helix, as can be appreciated by considering that right- and left-handed helices are related by a mirror symmetry operation across the central axis. In investigations of alternative structures and their conformational transitions in solution, the determination of helical sense can be of central importance. To our knowledge, the challenge posed by this requirement has not been addressed previously with spectroscopic techniques. We have approached the handedness problem by devising a strategy based on the synthesis of a set of molecules consisting of two fused helical segments, one of which is in a known reference helical form (Figure 1, upper). The molecules are covalently labeled with a donor at (or near) one end with an acceptor at (or near) the other end, and the efficiency of fluorescence resonance energy transfer

DNA Handedness by FRET

between the dye pair is determined. By systematically shifting the position of the junction while maintaining constant the total length of the helix (in base-pairs), the change in the FRET efficiency is a function primarily of the relative helical twist (and thus handedness) and helical rise of the two segments. These parameters can be readily obtained by fitting the data to a simple model in which a detailed description of neither the helical structure (e.g. backbone) nor of the probe fluorophores is required. The demonstration of this assertion follows.

Helical geometry Consider the two helical segments, helix 1 and helix 2, m and N − m bases long, respectively, with total length N (Figure 1, lower). A fluorescent donor is attached to helix 1 and a fluorescence acceptor to helix 2. We represent the system in cylindrical coordinates and assume a unique, global, centrosymmetric and straight helical (z) axis. The y axis is defined by the vector from the helix axis to the donor, rd (the corresponding vector to the acceptor is ra ) and the helical direction as extending from donor to acceptor (helix 1 to helix 2) such that translations and right-handed angular displacements along/about the axis are positive in value. The rotation (in degrees) and translation (in nm) required to map the donor to the first reference point (e.g. a phosphate group) on helix 1 are fd and zd , respectively. The corresponding displacements from the last position on helix 2 to the acceptor are fa and za . The twist angles and helical rises of helices 1 and 2 are (f1 , z1 ) and (f2 , z2 ), respectively, and we assume a single-step junction from helix 1 to helix 2 represented by the values fj and zj . (Longer junctions can be decomposed in a manner compatible with the same representation.) From these definitions, the net twist angle u and helical translation Dz from donor to acceptor are given by: u = fd + fj + fa + (m − 1)f1 + (N − m − 1)f2 = uo + mDf1,2

(1)

Dz = zd + zj + za + (m − 1)z1 + (N − m − 1)z2 = zo + mDz1,2

(2)

where: Df1,20f1 − f2

(3)

Dz1,20z1 − z2

(4)

and several of the individual parameters are integrated into the composite variables uo and zo : uo0fd + fj + fa + (N − 1)f2 − f1

(5)

zo0zd + zj + zo + (N − 1)z2 − z1

(6)

DNA Handedness by FRET

599

Figure 1. Rationale and definitions for the FRET determination of helical handedness. D and A represent the donor and acceptor fluorophores, respectively, each attached to one of two helical segments. The total length (N) of the composite helix is maintained constant but the fractional contributions from the two constituent segments (m/N and (N − m)/N) are varied by shifting the position of the junction in the sequence. Upper left, cylindrical model for DNA helix; upper middle, side-view depicting the vector rda joining the donor D and acceptor A, the positions of which are defined by their respective emission and absorption transition moments; right, cylindrical coordinate system for describing the orientation effects. The helix is depicted in cross-section with the helix axis in the center. The donor position is fixed and the acceptor is shown as occupying a centrosymmetric position about the axis, numbered sequentially for each unit (base-pair) step. Eight arbitrary steps are depicted with their respective projections of rda on the cross-sectional plane. For step 3, the positions of the vectors from the helix axis to the donor and acceptor (rd and ra ) as well as their angular separation (u) are labeled explicitly as an example. It is assumed that the acceptor (more specifically its absorption transition moment a) bears a fixed, unique orientation to the depicted circle (cylinder) defining its helical trajectory, and can be projected onto it , ia and ir , three local base vectors oriented tangential, parallel (axial) and orthogonal (radial) to the local position. The corresponding fractional orthogonal components are at = at it , aa = aa ia and ar = ar ir . Lower A, two joined helical segments with the same handedness (and helical rise and twist). The distance between D and A remains invariant as the fractional length of the individual segments is varied. Lower B, two segments with opposite handedness. D and A adopt different orientations as the junction is translated along the helix axis.

600

DNA Handedness by FRET

From equation (2) and the law of cosines, the 2 magnitude squared, rda , of the vector extending from the donor to the acceptor, rda , is: 2 rda = Dz 2 + rd2 + ra2 − 2rd ra cos u

(7)

2 rda = (zo + mDz1,2 )2 + rd2 + ra2

− 2rd ra cos(uo + mDf1,2 )

(8)

where rd and ra are the magnitudes of the vectors rd 2 and ra . The partial derivative of rda with respect to m, the number of base-pairs in helical segment 1, is given by: 2 1rda = 2(zo + mDz1,2 )Dz1,2 1m

+ rd ra Df1,2

01

p sin(uo + mDf1,2 ) 90

(9)

FRET formalism: Ro and k 2 The FRET efficiency E varies with the sixth power of the separation between a given donor and a given acceptor normalized by Ro , the characteristic (Fo¨rster, 1948) distance for 50% transfer efficiency:

$ 0 1% rda Ro

6

−1

(10)

with: Ro6 = (8.79 × 10−28 )Jn−4k2Fd

J = 1017fqd,l ea,l l4dl

(units: nm6 mol−1 )

(11)

J is the spectral overlap integral, n the refractive index of the medium separating donor and acceptor, k the donor-acceptor orientation factor, and Fd the

(12)

where the normalized donor emission spectrum (qd,l ) is given by equation (42), the acceptor absorption spectrum (ea,l ) is in M−1 cm−1, and the wavelength l is in nm. The value of Ro is determined largely by the nature of the donor and acceptor dyes (through J and Fd ) and lies generally in the range of 2 to 6 nm (Wu & Brand, 1992). It can be regarded truly as a constant only if the various parameters entering into equation (11) are invariant. The most problematical of these is the scalar orientation factor for dipole-dipole coupling, k, which is defined by the expression: k = d·a − 3(rda ·d)(rda ·a)

In general, zomDz1,2 . Thus, the derivative (equation (9)) consists of a constant term (2zo Dz1,2 ) proportional to the helical rise difference (Dz1,2 ) and a sinusoidal component with a magnitude (rd ra Df1,2 p/90) proportional to the difference in helical twist (Df1,2 ). It is apparent from equations (7) to (9) that the variation of the donor-acceptor distance with the length m of helical segment 1 (total length N remaining constant) depends on the differences in helical rise and twist of the two segments. The twist contribution will be perceived as a sinusoidal variation only if two conditions are met: (1) both rd and ra are finite (>0), i.e. both donor and acceptor dyes lie off the helix axis; and (2) there is a finite difference in helical twist (Df1,2 $ 0). If the magnitude of Df1,2 is large (helix 1 and helix 2 are of opposite handedness), the period of the sinusoidal variation (Dm = 360°/Dz1,2 ) will be short; for the combination B- and Z-DNAs, Dm 3 6. For helical segments of equal handedness, Df1,2 will be generally small, leading to a correspondingly long period Dm that may not be evident by inspection. That is, the function given by equations (7) and (8) will be perceived as monotonic.

E= 1+

donor emission quantum yield in the absence of FRET. In order to obtain Ro6 in units of nm6:

(13)

where the dot products involve the unit vectors d and a (donor emission and acceptor transition moments, respectively) and the donor : acceptor vector rda defined earlier. A constant value of 2/3 is generally ascribed to k2 under the assumption that both donor and acceptor dye molecules randomize their local orientation during the excited state lifetime (Dale et al., 1979). This assumption has been used universally in prior treatments of FRET applied to nucleic acids with covalently bound fluorophores. However, in the event that at least one of the dyes adopts a preferential or even fixed orientation relative to the macromolecular carrier, k2 will exhibit a systematic functional dependence on the relative dispositions of the donor and acceptor defining the scalar products of equation (13). There are extensive treatments in the literature of the static and dynamic averaging of k2, and its dependence on the orientational distributions of donor and acceptor (for a recent review, see Clegg, 1996). In order to incorporate an explicit consideration of k2 in the present formalism, we found it expedient 6 to dissect Ro6 into a geometry-invariant constant Roc 2 computed from equation (11) by setting k = 2/3, and a multiplicative factor b defined in terms of the helical coordinate system introduced above (Figure 1, top). Thus: 6 b; Ro6 = Roc

b0(3/2)k2

(14)

In the experiments described here, measurements of emission anisotropy indicated that the donor but not the acceptor was freely mobile (see below). The simplest situation applicable to such DNA duplexes consists of a randomized donor and a rigid acceptor occupying a unique position relative to the helix (due to symmetry, the reciprocal situation of rigid donor and mobile acceptor is equivalent). This case exhibits a single-parameter dependence on orientation, given by Dale and Eisinger (Dale et al., 1979) as: k2 = 13 [1 + 3 cos2 c] = 13 [1 + 3(rda ·a)2 ]

(15)

601

DNA Handedness by FRET

where c is the angle between rda and a. The problem in evaluating (15) is twofold: the orientation of a is generally unknown; and the vector rda changes direction relative to a as a function of the relative helix 1/helix 2 distribution, m/(N − m). We deal with these problems by augmenting the helical coordinate system (Figure 1, top right) with a set of orthogonal local basis vectors it , ia and ir defined at every discrete position of the acceptor and with orientations tangential, parallel (axial) and orthogonal (radial), respectively, to the cylinder encompassing the helical trajectory of the acceptor. It is assumed that the absorption transition moment a of the acceptor maintains a fixed, unique orientation relative to the local coordinate system, and thus can be decomposed by projection onto the basis vectors, thereby yielding three fractional orthogonal components: at = at it ; aa = aa ia ; ar = ar ir

(16)

The coefficients are signed but subject to the condition: a2t + a2a + a2r = 1.

(17)

The corresponding projections of rda along the same basis vectors are developed using the variables and nomenclature introduced above (equations (1) to (9)): dt = dt it ; da = da ia ; dr = dr ir

(18)

with: dt =

rd sin u z + mDz1,2 r − r cos u ; da = o ; dr = a d (19) rda rda rda d2t + d2a + d2r = 1

(20)

It follows that the direction cosine in equation (15) is given by: cos c = dt at + da aa + dr ar

(21)

such that for this case (mobile donor, rigid acceptor; defined as b = bI ):

bI 0

constraining the probable position of the fluorophore. Consider, for example, a polycyclic structure with a transition moment along the long axis. If the dye is located transversely in a groove of the helix, one might expect finite values for at and aa (at > aa ) and ar00. Molecular modeling and other forms of structural analysis (NMR, crystallographic, photoselective spectroscopy) can provide additional constraints on the dye position. The orientation factor bI was computed from equation (22) for a single helix (see below) with m = 1 to 20, using standard B-DNA helical parameters and different combinations of the relative base vector contributions from the acceptor absorption transition moment (Figure 2A). bI is seen to vary greatly in magnitude (over the maximal range 0.5 to 2.0) as a function of chain length and acceptor (dipole) orientation. The values deviate significantly from unity (corresponding to dynamic positional randomization) in a manner dependent systematically on the acceptor orientation. For example, the highest values and most pronounced plateaux are exhibited for a maximally axial disposition of the acceptor transition moment. This is expected since the condition for maximal transfer (k2 = 4) is for two colinear dipoles (Clegg, 1992, 1996; Dale et al., 1979; Fo¨rster, 1951). The curves exhibit a periodicity given by the helical twist (360°/f1 , or in one case, 180°/f1 ) that damps out for longer chain lengths m as the relative change in the direction of the donor-acceptor separation vector becomes progressively smaller (the z-axis component in equation (22) dominates). The corresponding FRET efficiency calculations (Figure 2B) confirm the qualitative and quantitative importance of the orientation factor.

Integrated geometric FRET model The FRET expression (equation (10)) augmented with the explicit consideration of k2 has the form:

$ 0 1%

E= 1+

0

1 rda b Roc

6

−1

1

1 [a r sin u + ar (ra − rd cos u) + aa (zo + mDz1,2 )]2 1+3 t d 2 2 rda

From equation (22), k2 can be computed for any position in the helix (donor-acceptor separation) in terms of the same parameters introduced earlier and, at most, three additional constants, the fractional a coefficients defining the position of the acceptor relative to the helix (usually one or two suffice). These quantities often can be deduced intuitively and/or by inspection of a structural model, exploiting known physical and chemical properties of the acceptor and the nucleic acid

(23)

(22)

A transformed version of equation (23) establishes the most direct relationship between the (measured) quantity E and the stereochemical model represented in equation (10): (E−1 − 1)1/3 =

0 1

2 2 rda 1 rda 2 = 2 1/3 Ro b Roc

(24)

in which the donor-acceptor separation appears in a form normalized by Ro . The following general

602

DNA Handedness by FRET

relationship provides the basis for simulating or fitting experimental data. All distance variables bearing a caret are normalized by Roc : (E−1 − 1)1/3 =

0 1

1 [(zx o + mDzx 1,2 )2 b1/3

+ rx 2d + rx 2a − 2rx d rx a cos(uo + mDf1,2 )]

(25)

For the specific donor/acceptor geometry treated above (b = bI ):

(E−1 − 1)1/3 =

=

$0

1%

2 1 rda [a r sin u + ar (ra − rd cos u) + aa (zo + mDz1,2 )]2 1+3 t d 2 2 Roc 2 rda

(zx o + mDzx 1,2 )2 + rx 2d + rx 2a − 2rx d rx a cos(uo + mDf1,2 )

$0

1 [a rx sin u + ar (rx a − rx d cos u) + aa (zx o + mDzx 1,2 )]2 1+3 t d 2 (zx o + mDzx 1,2 )2 + rx 2d + rx 2a − 2rx d rx a cos(uo + mDf1,2 )

Note that an a priori or exact value of Roc is not required in order to extract the parameters of greatest interest, Dzx 1,2 and Df1,2 . Other models for k2, i.e. b, can be accommodated by incorporation into equation (25). For fitting simulated or experimental data by non-linear regression, equations (25) and (26) can pose problems of instability and lack of convergence as a function of input parameters. A much more stable expression with only five fit parameters (c1 to c5 ) and applicable for b = 1 (but see below), is achieved by expanding the sinusoidal term: (E−1 − 1)1/3 = (c1 + c2 m)2 + c3 cos(c5 m) + c4 sin(c5 m) (27) The desired physical quantities are extracted from the fit with the following sequence of calculations, reflecting the approximations implied by the use of equation (27): Df1,2 = c5

(28)

−1

uo = tan (−c4 /c3 )

(29)

2rx d rx a = c4 /sin uo

(30)

zx o = zc − (2rx d rx a )

(31)

2 1

Dzx 1,2 = c1 c2 /zx o =

c1 c2 zc − c4 /sin[tan−1(−c4 /c3 )] 2 1

assigned to the other fit parameters. The accuracy is typically better than 1%. Thus, foreknowledge of neither the individual quantities constituting zo and uo nor of rd , ra or Ro is required. The surprising additional finding (see below) is that this circumstance holds also for data generated with or corresponding to equation (26), i.e. with a continuously varying k2 (b = bI ). The general features of the relationships developed so far are illustrated in the simulations and analyses in Figure 3; A and B differ primarily

(32)

Equations (27) to (32) were implemented in KaleidaGraph (Abelbeck) as the macro fit FRET 21. The approximation rx 2d + rx 2a 1 2rx d rx a forming the basis for equation (31) is valid to within 50% for 0.2 < (rx d /rx a ) < 5. Inasmuch as zx o is generally considerably greater than either rx d or rx a for molecular constructs such as those used in this study (Figure 5), fits according to equation (27) lead to excellent estimates of both Df1,2 and Dzx 1,2 (depending on the data accuracy and precision, of course) with little influence from the values

1%

1/3

−1/3

(26)

in the relative handedness of the two segments constituting the helix of 20 base-pairs. In the case of the example based on a combination of left-handed and right-handed segments (Figure 3A), a strong modulation superimposed on a sloping baseline is observed for both (E−1 − 1)1/3 and E as a function of m, the number of base-pairs in helix 1. Simulations are presented for both states of the donor/acceptor: randomized orientations (k2 = 2/3; b = 1) and rigidly bound acceptor (b = bI , equation (23) to (26)). Despite the pronounced differences in the amplitude and relative phase of the sinusoidal modulation, application of equation (27) to the analysis of the data yielded excellent estimates of the helical twist and rise in both cases. The product rx d rx a was obtained with a 10% accuracy for b = 1, but as expected the deviation was much greater for b = bI . In addition, the estimates of zx o were within 10% of expectation in both cases, although the uo values were not reliable, in part due to their calculation modulo 360° (from equation (6); uo > 700° for a 20-mer). The curves generated for a helix consisting of two right-handed segments differing by 3° in helical twist (Figure 3B) offered a striking contrast to those of Figure 3A. The functions (E−1 − 1)1/3 and E showed only a slight, apparently linear dependence on m; it is interesting that the slope was reduced to 00 for b = bI . A mathematical fit with equation (27) was still feasible with the error-free simulated curves (see the legend to Figure 3), although this circumstance would not be expected to apply to real experimental data.

Single-helix systems It should be obvious that the above formalism can be applied to a single-uniform helix (denoted helix 1) in which the separation between donor and acceptor is changed by increasing/decreasing the chain length m (=N). The modulated dependence of

603

DNA Handedness by FRET

Figure 2. Computations of the orientation factor bI . A, Evaluation of bI from equation (22) for a single helix with m = 1 to 20 using parameters: rd = ra = 1 nm; z1 = 0.34 nm; Roc = 5 nm; zd + za = 1 nm; f1 = 35°; and fd + fa = 0. The acceptor orientation is defined in terms of the indicated fractional components (equation (16)) in the order given (tra: tangential, radial, axial): at =at =, ar =ar = and aa =aa =. B, Corresponding calculations of the FRET efficiency E from equations (35) to (37). The dotted line in A and B for the condition of dynamic orientation of both donor and acceptor (b = 1). In this and other Figures, the various quantities are shown as continuous functions of m, although the latter variable can only assume integral values. This feature, achieved by interpolation, facilitates the viewing process but does not confer additional physical significance upon the data.

E on m imposed by the helical geometry was first treated by Clegg et al. (1993). Equations (2) to (8) require redefinition of terms based on the substitutions Df1,2 : f1 and Dz1,2 : z1 : u = fd + fa + (m − 1)f1 = uo + mf1

(33)

Dz = zd + za + (m − 1)z1 = zo + mz1

(34)

where: uo0fd + fa − f1

(35)

zo0zd + za − z1

(36)

and the master equation (26) with b = bI takes the form:

(E−1 − 1)1/3 = =

equation (26) can be adapted slightly in order to yield better estimates of f1 and zx 1 : (E−1 − 1)1/3 = c1 + (c2 m)2 + c3 cos(c5 m) + c4 sin(c5 m) (38) Equations (28) to (30) are still valid, but: zx 1 = c2

(39)

zx + rx + rx = c1

(40)

2 o

2 d

2 a

Equation (38) was implemented in KaleidaGraph as the macro fit FRET 22. An example of its application to data reported by Clegg et al. (1993) for a series of B-DNA duplexes is shown in Figure 4. The

$0

1%

2 rda [a r sin u + ar (ra − rd cos u) + aa (zo + mz1 )]2 1 1+3 t d 2 2 Roc 2 rda

(zx o + mzx 1 )2 + rx 2d + rx 2a − 2rx d rx a cos(uo + mf1 )

$0

1 [a rx sin u + ar (rx a − rx d cos u) + aa (zx o + mzx 1 )]2 1+3 t d 2 (zx o + mzx 1 )2 + rx 2d + 2rx 2a − 2rx d rx a cos(uo + mf1 )

Inasmuch as zx o (equation (36)) will be generally
1%

1/3

−1/3

(37)

fit provides estimates for f1 (helical twist) of 33° and for z1 (helical rise) of 0.36 nm, values close to

604

DNA Handedness by FRET

those for canonical B-DNA. In the original evaluation of the data with a version of equation (10) (Clegg et al., 1993), it was necessary to fix the helical twist and rise to 36° and 0.34 nm, respectively, in order to generate estimates for some of the dye-related parameters. FRET and fluorescence signals: ratio functions Rd , Ra and Rda The emission of a fluorophore is expressed as the sum of direct excitation (reduced by any energy transfer to (an)other species, Ei: ) and of indirect excitation via energy transfer from other species (Ej:i ). Assume (1) operation in the linear regime; (2) correction of the emission spectra to signals proportional to photons emitted per constant bandwidth; and (3) normalization of the emission/ excitation spectra to constant excitation irradiance with a quantum counter:

$

%

fi A ci ei,lexc (1 − Ei: ) + s cj ej,lexc Ej:i qi,lem Fi j$i

(41)

where fi is the fluorescent emission of species i excited at wavelength lexc and measured at emission wavelength lem , ci is the concentration of the species i, ei,lexc is the extinction coefficient of species i at the wavelength lexc , Fi is the fluorescence quantum yield of species i, and qi,lem is the normalized donor emission spectrum (more formally, the spectral quantal emission distribution function), obtained by integration of the corrected emission spectrum gi,lem : qi,l =

gi,l

g

gi,l dl

;

g

qi,l dl = 1

(42)

Such an explicit distinction between spectral distribution and quantum yield is preferable to the more general use of a combined parameter, because quantum yield is defined as a global quantity. The formalism facilitates greatly the development of new functions such as Rda below (equation (55)). In the following discussion, the subscripts of l are omitted except in cases where a single excitation or

Figure 3. Simulations and fits according to the two-helix model. A, Simulation with equations (25) (b = 1; w) · and (26) (b = bI ; W) and parameters: N = 20; f1 = −28°; f2 = +35°; z1 = 0.41 nm; z2 = 0.34 nm; fj = (f1 + f2 )/2; zj = (z1 + z2 )/2; fd + fa = 0; zd + za = 1 nm; Roc = 5.4 nm; at = z2/3; aa = z1/3; ar = 0. The b = 1 data were fit to equation (27), yielding the following estimates via equations (28) to (32) (the exact input values are given in parentheses): Df1,2 = −63.0° (−63.0°); Dzx 1,2 = 0.0129 (0.0130); rx d rx a = 0.064 (0.069); ua = −24° (+46°); zx a = 1.39 (1.50). For the b = bI data: Df1,2 = −63.4°; Dzx 1,2 = 0.0132; rx d rx a = 0.157; uo = −79°; zx o = 1.34. Note that the fit does not intercept the points perfectly. B, simulation with equations (25) (b = 1; w) · and (26) (b = bI ; W) and parameters as in A except for: f1 = +35°; f2 = +38°; z1 = 0.34 nm; z2 = 0.37 nm. The fits yielded for the b = 1 data: Df1,2 = −3.1° (−3.0°); Dzx 1,2 = −0.0054 (−0.0055); rx d rx a = 0.065 (0.069); uo = 5° (153°); zx o = 1.50 (1.57). (An alternative linear regression generated a slope (=2zx a Dzx 1,2 from equation (9)) of −0.013 (expected value, −0.012).) For the b = bI curve: Df1,2 = −3.1°; Dzx 1,2 = −0.0058; rx d rx a = 0.157; uo = −57°; zx o = 1.45.

605

DNA Handedness by FRET

emission wavelength instead of a spectrum is to be designated; the above definitions also apply to species j. We now restrict the discussion to a single chemical species bearing one unique donor and one unique acceptor (the chemical concentration is omitted for clarity)

excitation spectrum (a f gaexc ) or an absorption spectrum of an equivalent single (acceptor) labeled species:

f A [ed,l (1 − E)qd,l Fd + (ed,l E + ea,l )qa,l Fa ]

There are alternatives to the use of da f aa in equation (49). For example, the existence of significant overlap between the donor and acceptor absorption spectra may dictate the use of a red-shifted gaexc that cuts off a significant part of the acceptor emission. Thus, in order to obtain Df da without spectral clipping, one can first match da f aa to another full-range acceptor reference spectrum (preferably smoothed) such as qa,l , and then use the corresponding scaling factor gref to replace da f aa with a ·q . For computing the ratio function Ra and Rda gref a,l a below, one can omit the third term in Df da entirely, producing a truncated quantity (Df a ) incorporating only the correction for the overlapping donor emission:

(43)

and adopt a nomenclature based on the symbol 1 f 23 for the fluorescence signals, where prefix 1 is d, a or da for a molecular species carrying the donor alone, the acceptor alone or both fluorophores, respectively; subscript 2 is d or a for excitation in the donor or acceptor absorption band, respectively; and superscript 3 is d or a for emission in the donor or acceptor band, respectively. Provided the existence of a band of donor emission without overlap of the acceptor and an absorption band of the acceptor without overlap of the donor, two directly excited emission spectra are available:

ga =

fa a af a

a d

or ga =

ea,ldexc ea,laexc

(50)

f d A ed,ld (1 − E)qd,l Fd

(44)

Df a0da f da − gd·d f da

f A ea,la qa,l Fa

(45)

Using equation (49), we define a normalized sensitized emission spectrum in the form of the spectral ratio Ra (Rippe et al., 1995).

da d

da

a a

The first determination of E is based on the evaluation of donor quenching from a comparison of signals in the donor emission region corresponding to solutions of molecules containing either both donor and acceptor (da f dd ) or donor alone (d f dd ), and measured at or corrected to the same concentration: Rd 0

fd

da d

fd

=1−E

(46)

d d

7 E = Ed = 1 − Rd Rd (E − 1) = 1 − Rd −1 d

Df da da

f

a a

=

(48)

(49)

The proportionality constant gd is obtained by matching da f dd to a reference spectrum d f dd obtained under the same excitation and emission conditions; the concentration need not be the same in this case (d f dd and d f da are from one and the same emission spectrum). The constant ga is derived from (preferably) a corrected reference fluorescence

0 1 ed,ld ea,la

E;

alternatively, Ra =

Df a Df a − ga = ref − ga a ga ·qa,l da f a

(47)

The evaluation of Rd according to equation (46) is subject to a number of experimental difficulties, including the requirement for exactly matching the concentrations of the different solutions (Clegg, 1992). A second determination of E is derived from the net (sensitized) acceptor emission (Df da ) corresponding to FRET generated by excitation in the donor band. From the measured acceptor emission signal, one subtracts two overlapping contributions consisting of the residual donor emission and the directly excited acceptor emission: Df da 0(da f da − gd·d f da − ga·da f aa ) A ed,ld E·qa,l Fa

Ra0

(51)

7E = Ea =

(E−1 a − 1) =

(52)

0 1

(53)

0 1

(54)

ea,la R ed,ld a

1 ed,la −1 Ra ea,ld

Ra has the virtue of being based solely on fluorescence measurements and being directly proportional to E. However, one requires care in the application of equations (52) to (54) in the event that the ratio of extinction coefficients is not constant in a series of measurements, e.g. involving a changing population of molecules with different absorption properties (see below). In the absence of such an effect, however, Ra depicts the correct functional dependence of E on a variable of interest, such as chain length, even if the absorption ratio is not known to high accuracy, as is often the case with the characteristically dilute solutions used for fluorescence spectroscopy. A third ratio function, to our knowledge introduced here for the first time, relates the fluorescence signals even more directly to the FRET expressions given in equation (24) and in the

606

DNA Handedness by FRET

ensuing treatment. Combining equations (44) and (49), one defines: Rda0

0 1>0 1 0 10 1 Df da 1−E = qa,l E

fd

da d

qd,l

Fd ; Fa

alternatively, Rda =

0 1>0 fd

da d

qd,l

Df a − ga gref a qa,l

1

(55)

where qd,l and qa,l are derived from d f dd and da f aa according to equation (42). (If da f aa does not encompass the entire emission spectrum, one requires a complete reference spectrum, a f aa , of a molecule carrying only the acceptor in order to generate qa,l .)

$ 0 1 % 0 1 Fa R Fd da

7 E = Eda = 1 +

−1

Fa R Fd da

(E−1 da − 1) =

(56)

(57)

Like Ra , Rda is a quantity based solely on fluorescence signals. Furthermore, it provides a direct measure of the quantity (E−1 − 1) required in the expressions involving the helical geometry (equation (24)). Thus, one avoids the potential of a distortion in the functional form of (E−1 − 1), proportional to (1 − E)−1, arising from any systematic, constant relative error (d ln E) in the individual E values derived from Rd , Ra , or other (Clegg, 1992) ratio functions. A very fortuitous combination of variables occurs upon introduction of equation (55) into equation (24). Rda =

6 rda

0 1

Fa R Fd 6 o

=

0 1 rda Ro,a

6

(58)

(59)

It follows that equation (24) takes the alternative form: 1/3 Rda =

0 1

2 2 1 rda rda 2 = 2 1/3 Ro,a b Roc,a

(60)

with: 6 Roc,a = (8.79 × 10−28 )Jn−4(2/3)Fa

There are additional alternative measures of FRET efficiency based on other intensity ratios, emission anisotropies, fluorescence lifetimes and photobleaching rates (Clegg, 1992, 1996). They were not used in the present study. We include for reference, however, the following identities, relating our ratio functions to those employed in numerous other studies of nucleic acids by FRET (Clegg, 1992, 1995; Gohlke et al., 1994): Df a (ratio)A0 a = Ra + ga (62) da f a (ratio)D0

in which the donor-acceptor separation is now scaled by a new operative critical transfer distance Ro,a incorporating the acceptor instead of the donor quantum yield as a factor. In fact, the donor quantum yield is no longer required at any stage for quantitative interpretations of data based on equation (58): 6 Ro,a = (8.79 × 10−28 )Jn−4k2Fa

Figure 4. Analysis of FRET efficiencies for a series of B-DNA duplexes. Data from Clegg et al. (1993) for molecules of chain lengths 8 to 19 labeled with fluorescein (donor) and tetramethylrhodamine (acceptor). The data points derived from (ratio)A determinations (equation (62)) are plotted as E and (E−1 − 1)1/3. The latter function (m = 8 to 18) was fit to equation (38), yielding a unique solution regardless of initial values varied over a wide range. The indicated fit line corresponds to f1 = 33° and zx 1 = 0.085 (c5 and c2 , equations (33), (34) and (39)); using Ro = 4.3 nm (Clegg et al., 1993), z1 = 0.36 nm. The curve through the E data points was generated with these and the other values of the fit parameters (c1 = 0.64; c3 = 0.065; c4 = −0.20).

(61)

defined by setting k2 to 2/3 in analogy to equation (14). The modifications to the expressions up to equation (40) are obvious.

da

f dd

da

f

a a

=

0 1

qd,ld RR qa,la a da

(63)

Oligonucleotide design Test molecules were designed and synthesized to apply the FRET method outlined above (Figure 5, upper). The hairpin duplexes contained two segments, the first an antiparallel-stranded d(m5C·G) sequence known to undergo a righthanded B to left-handed Z conformational transition upon increasing the salt concentration (Behe & Felsenfeld, 1981; for a review, see Jovin et al., 1987), and the second a psAT sequence presumed to be right-handed from model building (Pattabiraman, 1986) and experiments based on the topology of closed circular molecules (Klysik et al., 1990). The goal was to confirm the helical sense of psAT and obtain estimates of its rise and twist. The methylated base m5C was selected because it dramatically lowers the concentration of MgCl2 required to induce the B-Z transition (Behe &

DNA Handedness by FRET

Felsenfeld, 1981; Malfoy et al., 1982). In this manner, a reversal of helical handedness could be achieved, providing a second independent determination of helical parameters with the same molecule(s). According to the crystal structure of d(m5C·G) the Z form is favored due to various effects of the methyl group in the major groove: exclusion of water, increase of hydrophobic surface and interaction with the backbone (Fujii et al., 1982a). In the design of the two-segment molecules containing the parallel-stranded d(A·T) (psAT ) segment, we exploited the stabilization of intramolecular duplex formation and specificity inherent to DNA hairpin structures. The 3'-3' and 5'-5' linkages flanking the d(T)N−m sequence were introduced by use of 5'-phosphoramidites (Evertsz et al., 1994; van de Sande et al., 1988), thereby imposing a parallel orientation on the strands comprising the d(T·A)N−m segment. In this manner we were able to construct a system adopting either an aps-B-DNA/psAT or an aps-Z-DNA/psAT configuration depending upon the solvent conditions. At low salt concentration, aps-d(m5C·G)m should be in a right-handed B conformation, with psAT presumably in a right-handed configuration (see above). At high salt concentration (E01 M MgCl2 ) the aps-d(m5C·G)m segment should undergo a transition to the left-handed Z form but the conformation of psAT would not be expected to be affected significantly by the change in ionic strength (Klysik et al., 1991). All oligonucleotides contained a fluorescein donor (6-FAM, 6-fluorescein phosphoramidite synthesis reagent; Figure 5, lower) at the 5' end and a sulfoindocarbocyanine (Cy3) acceptor in the loop of the hairpin structure. The Cy3 was incorporated by synthesizing the oligonucleotides with an Aminomodifier C2-dT, which contained a primary amino alkyl linker to which the monofunctional succinimidyl ester of the sulfoindocarbocyanine dye Cy3 (Cy3-OSu, Biological Detection Systems) was conjugated (Figure 5, lower). The labeling reaction resulted in a covalent linkage between the Cy3 and the DNA oligonucleotide via the aminoalkyl linker. Hairpin duplex stabilities The thermodynamic stability of the hairpin duplexes was investigated by native polyacrylamide gel electrophoresis and melting curves were monitored by UV absorption. Upon annealing m5CG4 TA12 to m5CG12 TA4 , the oligonucleotides migrated as double-stranded species in native polyacrylamide gels (0.1 M NaCl, 10 mM sodium cacodylate, pH 7.5). The five oligonucleotide-hairpins exhibited the same mobility (Figure 6A), corresponding to a 16-mer double-stranded species, based on a 15-mer duplex molecule as a reference. The absence of slower or faster migration bands was indicative of a homogeneous population devoid of single-stranded or higher-order structures. The melting curves for oligonucleotides 1 to 5 are shown in Figure 6B. As expected, the thermal

607

Figure 5. A diagram of oligonucleotide sequences and structures (upper) and dye molecules (lower). The 5'-5' and 3'-3' labels indicate the positions in the sequence where a change in polarity was introduced. The broken line denotes the junction between the two segments in the DNA molecule. The structures of the dye molecules after conjugation with the oligonucleotides are shown in the lower panel. The designations d(m5C·G) and d(T·A) denote single base-pair steps in the duplexes.

stability increased with the fractional content of d(m5C·G). The first three species (m = 2, 3, 4) achieved a plateau at high temperatures and therefore could be analyzed. The curves did not conform to a simple two-state model for the helix-coil transition of a hairpin duplex. However, satisfactory fits were obtained with a formalism for two successive concerted reactions, corresponding to the two segments constituting the duplexes (fit program: fit helix 14). According to this analysis, the second (higher) tm increased from 61°C to 80°C as the number of d(m5C·G) base-pairs (m) was changed from 4 to 8; the melting point increment was 1tm /1m = 4.7 deg.C. The combined helix-coil transition DH values were 470 to 590 kJ mol−1, a range of values appropriate for the length and composition of the 16-mer hairpins, but the computed distribution between the two fractional

608

DNA Handedness by FRET

transitions did not exhibit a consistent dependence on m. The fully denatured state was not achieved up to 95°C in the case of oligonucleotides 4 and 5 (m = 10 and 12). We note that a multistate helix-coil transition has been observed in duplex molecules consisting of two distinct 15-mer blocks of ps-DNA (Evertsz et al., 1994). In the present constructs (Figure 5), the psAT component was stabilized by clamping both ends, i.e. via the closed loop at one end and the extremely stable aps-d(m5C·G) segment at the other. As a consequence, the melting of the psAT sequence would be shifted to higher temperatures with a consequent reduction in the base-pair stability constant at and near the tm , rendering the transition less cooperative and less amenable to simple quantitative analysis. However, these considerations do not affect the conclusion that the hairpins were fully paired at the low temperature (4°C) selected for the FRET measurements.

Characterization of the B-Z transition by CD spectroscopy Upon increasing the concentration of MgCl2 , alternating purine-pyrimidine DNA sequences such as d(m5C·G)m undergo a right-to-left (B-Z) conformational transition characterized by pronounced changes in the circular dichroism spectra (Pohl & Jovin, 1972; for a review, see Jovin et al., 1987). Such structural interconversions have been demonstrated in (16 to 80 bp) oligomers consisting of two sequence blocks adopting either a uniform B-DNA helical structure at low concentrations of salt or a hybrid B-Z configuration with a distinct junction at high concentrations of salt (Sheardy & Winkle, 1989). The ionic conditions under which oligonucleotides m5CG4 TA12 , m5CG6 TA10 , m5CG8 TA8 , m5CG10 TA6 and m5CG12 TA4 adopted either an aps-B-DNA/ps-AT or an aps-Z-DNA/ps-AT hybrid helical state were explored by CD spectroscopy. Spectra taken in the presence of increasing concentrations of MgCl2 are shown in Figure 7. For all oligonucleotides studied under the low-salt condition, a common spectral pattern was apparent, characterized by two positive peaks at 270 to 305 nm and 210 to 230 nm, as well as negative regions at 230 to 270 nm and 205 to 210 nm. Upon addition of Mg2+, the negative 230 to 270 peaks of DNA molecules m5CG6 TA10 , m5CG8 TA8 , m5CG10 TA6 and m5CG12 TA4 decreased in magnitude and even reverted sign in the case of m5CG12 TA4 . Furthermore, the 270 to 305 nm peak changed progressively into a negative dichroic component at 280 to 310 nm. This behavior is typical of alternating purine-pyrimidine sequences undergoing the B-Z transition (Pohl & Jovin, 1972; Reich et al., 1993; Sheardy & Winkle, 1989). The transition was evaluated further by computing the difference CD signal as a function of the Mg2+ concentration (Figure 8). The midpoint of the

Figure 6. Thermodynamic stability of the hairpin duplex structures. A, Polyacrylamide gel electrophoresis of donor-labeled DNA molecules m5CG4 TA12 , m5CG6 TA10 , m5CG8 TA8 , m5CG10 TA6 and m5CG12 TA4 (m = 4 to 12, respectively) under native conditions stabilizing the hairpin duplexes: 15% polyacrylamide gel in 10 mM Tris-borate (pH 8), 0.1 M NaCl, r, double-stranded 15-mer DNA reference molecule. B, Melting profiles for donor-labeled DNA molecules m5CG4 TA12 , m5CG6 TA10 , m5CG8 TA8 and m5CG12 TA4 in 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl indicated as 4, 6, 8 and 12, respectively. m5CG12 TA4 did not complete a transition in the range of temperatures studied. The continuous lines represent a fit to a two-stage mechanisms; see the text for details. For clarity, only every second or third measured point is plotted (w).

B-Z transition was displaced to higher ionic strength as the number (m) of d(m5C·G) base-pairs decreased. The m5CG4 TA12 hairpin did not adopt the Z conformation even at 2.5 M MgCl2 (Figure 8), providing a valuable reference molecule for the influence of salt concentration on the diverse measurements in the absence of the B-Z conformational transition. The systematic differences in amplitude between the different curves (oligonucleotides) reflected the fraction of Z-forming sequence. The inset in Figure 8 is a plot of the total transition CD signals versus m for the oligonucleotides m5CG6 TA10 to m5CG12 TA4 . The relationship was linear, with an intercept denoting the involvement of only one terminal (presumably junctional)

609

DNA Handedness by FRET

Figure 8. CD at 260 nm as a function of Mg2+ concentration. Data derived from Figure 7 and other similar experiments. The ordinate represents the difference CD signal (De − Dea ) referred to the initial state (no Mg2+ ). DNA hairpin molecules: (Q) m5CG4 TA12 , (r) m5CG6 TA10 , (R) m5CG8 TA8 , (w) m5CG10 TA6 and (W) m5CG12 TA4 . The signals reflect the B-Z transition, with a net magnitude (d(De − Deo )) proportional to the fractional contribution of the d(m5C·G) segment to the sequence. Inset: linear fit of d(De − Deo ) versus m. From the intercept we conclude that 01 base-pair at the junction does not contribute to the change in CD.

Figure 7. CD spectra of DNA hairpin duplexes m5CG4 TA12 , m5CG8 TA8 and m5CG12 TA4 with increasing concentrations of Mg2+. A, m5CG4 TA12 ; the Mg2+ concentration was increased from (w) · 0 M to (W) 2.5 M. B, m5CG8 TA8 . (w) · 0 M Mg2+, (W) 1.73 M Mg2+. C, m5CG12 TA4 . (w) · 0 M Mg2+, (W) 1.19 M Mg2+. All solutions included 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl.

base-pair of the d(m5C·G) segment not contributing to the CD signal. We tested for possible changes in the structure of the psAT segment in response to the varying ionic conditions by comparing the CD spectra of m5CG8 TA8 and poly[d(G·m5C)] at low and high concentrations of salt (Figure 9A). The difference spectra representing the contributions from the TA segment (and the hairpin loop) were compared with the corresponding CD spectra of a reference psAT hairpin (Figure 9B). The patterns were very similar both qualitatively and quantitatively. In addition, the spectra of the psAT hairpin at low and high concentrations of salt were superimposable. The small differences between the spectra at low and high concentrations of salt of m5CG8 TA8 (Figure 9B) can be attributed to the fraction of the d(m5C·G) segment not undergoing the B-Z transition. We conclude that the psAT segment in the molecules

generated in the present study (Figure 5) was structurally related to the ps-DNA constructs studied previously (Shchyolkina et al., 1989; van de Sande et al., 1988) and that its conformation was not perturbed significantly by the manipulations of the ionic conditions required for the induction of the Z conformation in the adjoining segment. The donor-acceptor pair: fluorescein-Cy3 Several dyes were investigated in order to determine an appropriate donor-acceptor pair. Among these, the combinations fluorescein-tetramethylhodamine, fluorescein-Cy3 and Cy3-Cy5 exhibited the highest values of Ro . The use of fluorescein is advantageous due to the availability of the monoisomeric phosphoramidite reagent, enabling the introduction of the fluorophore during chemical synthesis of DNA. Cy3 was chosen over tetramethylhodamine as an acceptor for fluorescein because it has a substantially higher extinction coefficient and the free dye does not exhibit a tendency for non-covalent binding to DNA. The attributes of the fluorescein-Cy3 donoracceptor pair were investigated in order to define the operative range of distances for the FRET measurements and thus the optimal lengths for the synthesized molecules. Roc (equation (14)) was determined at the extremes of salt concentration from the measured absorption and emission spectra (Figure 10) and the following values for the parameters of equation (11): n = 1.33 (value for water, inasmuch as the dyes experience a predominantly aqueous environment); k2 = 2/3 (by

610

DNA Handedness by FRET

limiting emission anisotropy for an ensemble of immobile, randomly oriented fluorophores (ro ) was determined for Cy3-DNA in an arabinose glass, yielding the value 0.375, similar to the ro of 0.35 reported for fluorescein (Weber, 1956). Fluorescence lifetimes (tf ) were also determined: Cy3-DNA in an arabinose glass, 2.5 ns; Cy3-DNA solution, 1.4 ns; fluorescein-labeled duplexes similar to those investigated here, 4.6 ns. The Perrin equation (Jablonski, 1957) relates these quantities to the effective rotational correlation times (frot ): ro t =1+ f r frot

Figure 9. CD spectra of m5CG8 TA8 (W, Q) and poly[d(m5CG)] (w, q) at low (W, w) and high (Q, q) concentrations of salt. Low-salt, 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl; high-salt, low-salt + 1.6 M Mg2+. A, CD spectra; B, difference CD spectra generated by subtracting the dichroism of poly[d(m5CG)] from that of the DNA hairpin at low (Q) and high (w) concentrations of salt. This signal (DDe) should represent the contribution from the psAT segment of the joint molecule, and is compared with the spectrum (De) of a psAT reference hairpin molecule, 3'-T15-5'-5'-CCCC-A15-3', at low-salt (- -W- -); the spectrum at 2.5 M Mg2+ was superimposable.

(64)

Experimental determinations (Eimer et al., 1990) and hydrodynamic theory (Garcia de la Torre et al., 1994) predict a frot for a 16-mer DNA duplex of 6 to 12 ns, depending on the relative contributions of motions about the two principal molecular axes (combinations of spinning and end-to-end tumbling described by two to three rotational diffusional coefficients and three to five rotational relaxation times, depending on the degree of symmetry), and the extent of internal motion. Thus, for the case of a rigid fixation of the dyes to the DNA, one would expect from equation (64) that r/ro > 0.72 for fluorescein and r/ro > 0.89 for Cy3. The measured values for the duplex molecules were: fluorescein, r/ro = 0.30; and Cy3, r/ro = 0.96. We conclude that the fluorescein donor was quite mobile (for a discussion of possible cone angles of diffusion see Clegg (1992, 1996) but that the Cy3 acceptor was essentially devoid of local motion presumably due to a (specific) interaction with the DNA. Thus, the formalism appropriate for this case would be that given by equation (26) (b = bI ). Monitoring the B-Z transition by FRET

definition; equation (14)); and Fd = 0.45 (Eis & Millar, 1993). Small salt-dependent variations in the absorption extinction coefficient of the Cy3 acceptor, reflected in the values of the overlap integral J (equation (12)), led to the following estimates of Roc from equations (11) and (14): low salt (no MgCl2 ), Roc = 5.6 nm (J = 3.58 × 1032 mol−1 nm6 ); and high salt (e1.2 M MgCl2 ): Roc = 5.7 nm (J = 4.24 × 1032 mol−1 nm6 ). These values of Roc indicate that the chain length selected for the test molecules (Figure 5) was well suited for FRET measurements in the sensitive range of distances 1Roc (equation 10). The normalized emission spectra (qd,l and qa,l , Figure 10A) are noteworthy because they should be transferable between laboratories in terms of both shape and magnitude. Emission anisotropy measurements were performed on single-labeled, double-stranded and single-stranded DNA molecules (15-mer oligonucleotide references). For the fluorescein-labeled single and double-stranded molecules, r was 0.06 and 0.12, respectively. The corresponding values for the Cy3-labeled molecules were 0.31 and 0.37. The

We evaluated the suitability of FRET as a means for sensing the B-Z transition in the set of oligonucleotides shown in Figure 5. Titration curves under the identical experimental conditions used for the CD measurements were monitored by evaluation of the various fluorescence ratio functions, Rd , Ra and Rda . A representative set of signals is depicted in Figure 11. In addition, absorbance spectra were taken in order to evaluate the absorbance ratio ea,550 /ed,490 required for the computation of Ea (equation (53)). The dependence on the Mg2+ concentration was very similar for all oligonucleotides, including m5CG4 TA12 , which did not undergo the B-Z transition. Significant changes in the absorption of the dyes (16% hyperchromicity in the case of fluorescein, and 4% hypochromicity in the case of Cy3) were observed in passing from the low-salt to the high-salt condition and stabilizing at e0.2 M MgCl2 . These effects did not differ appreciably for species with the d(m5C·G) segment in either the B or Z conformation. In addition, the ratios of fluorescence emission to absorbance of double-labeled oligonucleotides

611

DNA Handedness by FRET

Figure 10. A, Normalized donor and acceptor emission (qd,l , qa,l ) and absorption (ed,l , ea,l ) spectra and kernel of the overlap integral J (qd,l ea,l l4 ). Fluorescein (w, qd,l ; W, ed,l ); Cy3 (r, qa,l ; R, ea,l ); kernel (equation (12)) (- - -), units in M−1 cm−1 nm4. B, Absorption spectrum of a double-labeled sample (w, · m5CG12 TA4 ; continuous line, fit obtained as a vectorial sum of fluorescein (W) and Cy3 (R) absortion spectra) exemplifying the spectral decomposition required to obtain the extinction coefficients of fluorescein and Cy3 (see the text). Conditions: 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl, 4°C.

(direct excitation of the acceptor, da f aa ) and single-labeled oligonucleotides (excitation of the donor, d f dd ) were invariant, indicating that the quantum yields were unaffected by the manipulations of the ionic environment. The salt-dependent variations in the FRET efficiency values Ea (calculated from equation (52)) clearly reflected the B-Z transition of the various hairpin duplexes with the exception of m5CG4 TA12 (Figure 12A). In accordance with the CD measurements and expectation, the transition salt concentration diminished with an increase in m, the length of the d(m5C·G) segment. A composite set of CD and FRET data for m5CG12 TA4 is depicted in Figure 12B. A quantitative treatment of such titration curves will be presented elsewhere. Nonetheless, it is evident from Figure 12 that FRET provides a sensitive measure of the B-Z transition. A graphical comparison (Figure 13) of Ra and Rda according to equations (54) and (57) provided a test for internal consistence and an estimation of the relative emission quantum efficiencies (Fd /Fa ). An

Figure 11. Example of signals used in the determination of FRET (duplex m5CG4 TA12 ). (d f dd ) emission spectrum of fluorescein in the single-labeled duplex with excitation at 490 nm; (da f dda ) emission spectrum of fluorescein and Cy3 in the double-labeled duplex with excitation at 490 nm; (gd × d f dd ) calculated emission spectrum of fluorescein in the double-labeled duplex with excitation at 490 nm; (—), Df da ) difference spectrum representing the sensitized emission of Cy3 by energy transfer; (da f aa ) direct excitation a of the acceptor at 550 nm; (- - -), ga × a f lexc ) calculated emission spectrum of Cy3 in the double-labeled duplex due to direct excitation at 490 nm. The coefficient ga (0.166) was determined from an excitation spectrum of a reference duplex labeled only with Cy3 as the ratio a a a f 490 /a f 550 (equation (50)). See equations (41) to (51) for definitions.

excellent linear fit to the high-salt data, compatible with the clustered low-salt points, was observed, validating the estimations of ea,550 /ed,490 and yielding a value for Fd /Fa of 2.2. Assuming Fd = 0.45 (Eis & Millar, 1993), this result implies that the acceptor (Cy3) quantum yield Fa was 0.20, a value compatible with data reported by the manufacturer for Cy3 conjugated to proteins (F > 0.15). Determination of helical parameters from FRET The various determinations of FRET (ratio functions and transfer efficiencies) are compiled in Table 1 and displayed in Figure 14 as a function of m, the length of helix 1, and the salt conditions (affecting conformational states). Despite the very different values of the ratio functions, the derived FRET efficiencies were in satisfactory agreement, particularly in the case of Ea and Eda , as expected from Figure 13. The Ed values appeared less reliable than the other two quantities, in our judgment primarily due to the difficulty in matching the

612

DNA Handedness by FRET

Figure 13. Determination of relative donor and acceptor emission quantum yields from the correlation between Rda and Ra . The values of Rda and Ra obtained for the series of hairpin duplexes at high salt (w, · Table 1) were plotted in accordance with equations (54) and (56) so as to yield a linear relation with a slope = Fd /Fa ; the indicated errors are from Table 1. The fit was perfectly linear, passing though the origin and yielding a slope of 2.2 = Fd /Fa . The data at low salt (q, Table 1) were compatible with this value, implying that the quantum yield ratio was invariant.

Figure 12. Measured FRET efficiency Ea of the DNA hairpin duplexes (A) as a function of Mg2+ concentration and (B) correlated with the corresponding CD signals. Both curves show the transition to a plateau (Z conformation) at comparable Mg2+ concentrations. (Q) m5CG4 TA12 , (r) m5CG6 TA10 , (R) m5CG8 TA8 , (w) m5CG10 TA6 and (W) m5CG12 TA4 . The buffer was 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl.

concentrations of the single- and double-labeled DNA species. In practice, the decomposition of the combined donor-acceptor absorption spectrum into the constituent contributions of the two dyes has to be performed with great care (Figure 10B), in order to achieve the precision within a few per cent required for accurate estimations of the FRET efficiency from Ed (equation (47)). The important conclusion derived from the very satisfactory agreement between the various methods of calculation is that the system conformed to the assumptions and formalism developed in this work. We proceed to an analysis of the FRET data for the purpose of determining the relative rise and twist of the two helical segments constituting the set of molecules, one of which adopted the alternative right-handed B or left-handed Z conformation depending on the ionic conditions. It is evident from Table 1 and Figure 14 that for the low-salt condition favoring the B form of the d(m5C·G) segment, the FRET efficiency did not vary significantly with m. In contrast, in the presence of Mg2+ and consequent induction of the Z form of d(m5C·G) (Figure 14), the FRET values decreased significantly with m. The effect of salt-induced unwinding (Anderson &

Bauer, 1978) would have been negligible in our experiments. In order to apply the integrated geometric model represented by equation (26), the data were plotted in the form of (E−1 − 1)1/3 values as a function of m and for the low-salt and high-salt conditions (Figure 15). For the former case, a linear fit was applied to the data. If we assume that the contribution to the measured slope (−1.5 × 10−3 ) was mainly due to differences in helical rise (equation (9); the product rx d rx a may be quite small in this system; see below) and neglecting the effect of bI (Figure 2, equation (26)), we approximate the slope by 2zx o Dzx 1,2 (equation (9)), zx o by Nz1 /Ro 3 0.9, and thereby derive the estimate Dz1,2 3 −0.004 nm, a very small value. One can conclude that the helical rise of psAT-DNA is just slightly greater than that of B-DNA, although the quantitative interpretation depends on the possible effect of bI , i.e. a reduction of the magnitude on the measured slope (Figure 3B). We are also able to deduce that the difference in helical twist of the two segments (Df1,2 ) is small, in view of the absence of a well-defined sinusoidal component. At a high concentration of salt (MgCl2 ), the d(m5C·G) segment was left-handed for four of the five hairpin duplexes, and the FRET data plotted according to equation (26) displayed a very different behavior (Figure 15). The values of (E−1 − 1)1/3 derived from all the calculations increased substantially with m. Unfortunately, a unitary increment of m in the series of oligonucleotides was not possible due to the synthesis strategy dictated by the dinucleotide structural element of the Z-forming segment (Figure 5). The anticipated Df1,2 for a ZDNA (helix 1) and a (close to B-DNA) ps-DNA (helix 2) combination would be the difference between the mean twist angle for the m5CG

613

DNA Handedness by FRET

Table 1. Fluorescence ratio functions and computed FRET efficiencies for DNA hairpin duplexes m5CGm TAN−m labeled with fluorescein (donor) and Cy3 (acceptor) m

Rd

Ra

Rda

Ed

Ea

Eda

0.107 2 0.0010 0.109 2 0.0005 0.107 2 0.0010 0.106 2 0.0005 0.113 2 0.0010

3.43 2 0.04 3.55 2 0.05 3.45 2 0.06 3.46 2 0.05 3.30 2 0.06

0.389 2 0.017 0.404 2 0.017 0.385 2 0.018 0.387 2 0.017 0.397 2 0.017

0.390 2 0.011 0.398 2 0.009 0.390 2 0.012 0.389 2 0.012 0.410 2 0.011

0.390 2 0.003 0.383 2 0.003 0.389 2 0.004 0.389 2 0.003 0.400 2 0.004

B. High-salt condition: buffer a + end point Mg 2+ concentrations 4 0.625 2 0.018 0.126 2 0.0016 3.62 2 0.08 6 0.647 2 0.018 0.117 2 0.0009 4.07 2 0.06 8 0.668 2 0.019 0.108 2 0.0010 4.54 2 0.10 10 0.697 2 0.020 0.101 2 0.0007 5.08 2 0.07 12 0.716 2 0.020 0.097 2 0.0008 5.39 2 0.14

0.375 2 0.018 0.352 2 0.018 0.332 2 0.019 0.303 2 0.020 0.284 2 0.020

0.378 2 0.009 0.351 2 0.009 0.326 2 0.009 0.302 2 0.009 0.290 2 0.008

0.378 2 0.005 0.351 2 0.003 0.326 2 0.005 0.302 2 0.003 0.290 2 0.005

A. Low-salt condition: buffer a 4 0.611 2 0.017 6 0.596 2 0.017 8 0.615 2 0.018 10 0.612 2 0.017 12 0.603 2 0.017

a Buffer: 10 mM sodium cacodylate (pH 7.5), 0.1 M NaCl. The structures of m5CGm TAN−m are given in Figure 5. The high-salt condition corresponds to the last point of each titration (see Figure 12A). In the absence of Mg2+ the d(m5C·G)m segment was in the right-handed B form, and in the presence of Mg2+ in the left-handed Z form (except for m = 4). Rd and Ed were computed from equations (46) and (47), respectively, with errors derived from statistical analyses of the ratio da f dd /d f dd and the estimated precision in the relative concentrations due to pipetting error (according to manufacturer specifications, 22% for a P 20 Gilson Pipetman). Ra and Ea were calculated from equations (52) and (54), respectively, with errors from statistical analyses of the ratio Df da /da f aa , and in the case of Ea an additional contribution from the ratio of extinction coefficients. The errors corresponding to the quantity (E−1 − 1)1/3 used in Figure 15 were calculated by appropiate propagation methods. Rda was calculated from equation (55) with errors derived from statistical evaluation of the four associated spectral curves. Eda was calculated from Rda using Ffl /FCy3 = 2.2 (equation (56), Figure 13). All statistical analyses of spectra were performed on a fixed number of data points (40) around the emission maxima, yielding standard errors. The spectral ratio ea,550 /ed,490 was 3.63 to 3.66 and 2.99 to 3.01 under the low-salt and high-salt conditions, respectively.

dinucleotide unit ((−45 − 13)/2 = −29°, crystallographic data for d(m5CG)3 (Fujii et al., 1982b)) and +35° for B-DNA (Peck & Wang, 1981); that is, Df1,2 = −64°. The corresponding period Dm of the (E−1 − 1)1/3 function would be 5.6. For such a value, incrementing m by 2 provides only three determinations per period. Thus, a distinct sinusoidal behavior would not be expected (nor was it apparent, Figure 15), although the measured points should still correspond to the functional form of equation (26). Analysis by the fit programs was restricted to the determinations derived from Rda , which coincided perfectly with those from Ra , albeit with a more favorable estimation of the error limits 1/3 (Table 1). Repeated fits of the (E−1 versus m da − 1) curve were carried out according to equation (27) with different input values for Df1,2 . The deviations between output (fit) and input values of Df1,2 displayed a Gaussian dependence, with a peak (presumed to denote the optimal fit parameter) corresponding to Df1,2 = −58°; the standard deviation of fits with initial values in the range −55 to −70° was 2.5°. The other fit values (calculated from equations (28) to (32) and Ro = 5.7 nm) were Dz1,2 = 0.050 nm (20.002 nm over the given range of input Df1,2 values), zo = 6.0 nm, and rd ra = 0.3 nm2. The latter quantity was very small, particularly considering a possible inflationary effect from the orientation factor bI (Figure 3B) and probably reflecting the adoption by one of the dyes of a position (unfortunately) close to the helix axis. The most probable candidate is the immobilized Cy3 molecule situated at the base of the hairpin loop (Figure 5). The fit values were used to generate the continuous curve extended to the limits of m = 2 and 14. Despite the sparsity of the data defining the behavior of FRET under the high-salt condition, the

fits were internally consistent and in accordance with the known structural features of B- and ZDNA. Thus, the difference between Dz1,2 determined at high and low concentrations of salt (0.050 + 0.005 = 0.055 nm) was in excellent agreement with the value of 0.05 nm derived from crystallographic data for Z-DNA (z1  = 0.38 nm, average rise of the dinucleotide unit) and B-DNA (z1 = 0.33 nm: Dickerson et al., 1982). We cannot compute the corresponding difference in Df1,2 for the two salt conditions, but the best estimate of Df1,2 for the DNA in the presence of Mg2+ (−58°) constitutes strong evidence that psAT-DNA is right-handed with helical parameters not very different from those of B-DNA. The FRET values under the low-salt and high-salt conditions for the hairpin duplex m5CG4 TA12 (m = 4) did not coincide (arrow, Figure 15), despite the fact that Roc increased slightly upon addition of Mg2+. We cannot identify any source of systematic error in these determinations, and thus conclude that this particular DNA species underwent a partial transition to the Z form, one perceived by the FRET parameters but not detected by CD spectroscopy (Figure 7A). Concluding remarks The FRET simulations and data demonstrate that right- and left-handed conformations in short oligonucleotide sequences in solution can be discriminated based upon comparison with a second known helical segment in a joint molecule. The donor-acceptor pair selected for the experiments, fluorescein-Cy3, has a high Roc value and exhibits simple, consistent photophysical properties. The formal basis for the technique has been presented in detail, including an extensive discussion of alternative ratio functions of the various

614 fluorescent signals required for evaluating the FRET efficiency and extension to single-helix systems. In addition, we have shown that the transition between a right-handed and a left-handed conformation can be monitored sensitively by FRET, as applied in the present study (to our knowledge for the first time) of the salt-dependent B-Z transition. The method requires only low concentrations of

DNA Handedness by FRET

samples but is not restricted to solution studies, i.e. can be applied in other situations such as in situ measurements of species resolved by gel electrophoresis (unpublished data). We have confirmed in this study the right-handed character of parallelstranded DNA (psAT-DNA) based on trans A·T base-pairs and showed that its helical parameters are close to those of B-DNA.

Figure 14. Fluorescence ratio functions and corresponding FRET efficiency values of the oligonucleotides m5CG4 TA12 (m = 4), m5CG6 TA10 (m = 6), m5CG8 TA8 (m = 8), m5CG10 TA6 (m = 10) and m5CG12 TA4 (m = 12). All solutions were in 10 mM sodium sodium cacodylate (pH 7.5), 0.1 M NaCl ( ) low salt). High-salt ( ): end points from the Mg2+ titration curves in Figure 12A.

615

DNA Handedness by FRET

Figure 15. Analysis of the experimental FRET data and derivation of helical parameters for the B-DNA/psATDNA and Z-DNA/psAT-DNA states of the dual segment hairpin duplexes. The FRET efficiency values for the series of hairpin duplexes and the two ionic conditions (Table 1) are plotted according to equation (26). Data from: (q) Ed ; (w) Ea ; (W) Eda (errors for Eda are shown). The high-salt curve (Eda data alone, m = 6–12) was fit by the program fit FRET 21 using the errors of Table 1 as weighting factors. Starting parameters: c1 , 1.1; c2 , 0.01; c3 , 0.04; c4 , −0.04; c5 , −40 to −80° (see the text for discussion of results). The arrow points to the measurements for m5CG4 TA12 (m = 4), which did not undergo the B-Z transition; the points do not fall on the fit curve extended over the range m = 2–16 (continuous line). The low-salt data were fit by a straight line to the Eda data; the error bars of all points are depicted.

The approach exploited in this study is very sensitive because it is based on differences in helical rise and twist, in contrast to experiments based on single-helix constructs of varying length that are insensitive to the sign of the helical twist. Other conformational states are potential candidates for investigation, of course, including A-DNA (and the B-A transition), multistranded structures, local and global helical transformations of topologically constrained (supercoiled) DNA (Cozzarelli & Wang, 1990), polymorphic RNA structures, the myriad variants of parallel-stranded DNA (Rippe et al., 1992a,b), and protein-nucleic acid interactions. A study has been undertaken with the FRET technique of ps-d(AG)n homoduplexes (Rippe et al., 1992a,b), the results of which will be reported elsewhere. It is appropriate to discuss the difficulties associated with the FRET technique presented here. Distinct improvements can be contemplated, particularly in molecular design. We have already alluded to the problem posed by alternating dinucleotide structures such as the purine-pyrim-

idine sequences required for effecting the transition to the Z-form. A general and essential requirement for a modulation of the FRET signal reflecting the helical twist properties is that both the donor and acceptor dyes be laterally displaced from the helical axis (equation (25)). Furthermore, although randomization of the dye positions facilitates consideration of orientational effects (the k2 problem), excessive motion will blur the precision and accuracy of the distance relationships determining the FRET signals. Thus, at the level of oligonucleotide synthesis we could contemplate the use of other dye combinations (and appropriate measurement conditions, e.g. >pH 8 for fluorescein-related fluorophores); dye conjugation to the backbone moieties (sugar, phosphodiester bond) instead of the base or terminal group; and localization of the dyes to well-defined helical regions instead of at terminal positions (free ends, loops). Finally, one should emphasize that the FRET phenomena depend intimately on the molecular model. For example, deviations from a unique and straight helical axis for the composite molecules would lead to departures from the behavior predicted with the formalism presented here. On the other hand, if properly analyzed, such effects can be exploited for the purpose of detecting and characterizing specific structural perturbations (Gohlke et al., 1994).

Experimental Procedures Oligonucleotide synthesis Oligonucleotides m5CG4 TA12 , m5CG6 TA10 , m5CG8 TA8 , m5CG10 TA6 and m5CG12 TA4 (Figure 5, upper) were synthesized by conventional phosphoramidite chemistry with an Applied Biosystems model 381 A DNA synthesizer. Amino modifier C2 dT-CE phosphoramidite, 5-Me-dC-CE phosphoramidite and dT-5'-CE phosphoramidite were from Glen Research; FAM (fluorescein phosphoramidite) was from Applied Biosystems; standard dA, dT, dC phosphoramidites and CPG columns were from MWG-Biotech or from Applied Biosystems. After synthesis, the oligonucleotides were purified by electrophoresis in a denaturing (0.089 M Tris-borate (pH 8), 7 M urea) 20% polyacrylamide (19:1 (w/w) monomer/crosslinker) gel, operated typically at room temperature for five to six hours. The bands were observed by UV shadowing or by visualization of dye fluorescence. A single band corresponding to the singlelabeled (fluorescein) species was observed in every case. A fraction of these purified oligonucleotide was conjugated overnight with Cy3-Osu (Biological Detection Systems) in 0.1 M NaHCO3 (pH 8.25) at room temperature. The product was first separated from excess of unreacted dye by gel filtration (spindown Sephadex G-15 column equilibrated in water), and then purified by gel electrophoresis. A single band corresponding to the double-labeled (fluorescein, Cy3), species was isolated. The oligonucleotides were recovered by electroelution in a Schleicher & Schuell Biotrap BT 1000 system in 0.089 M Tris-borate (pH 8), and concentrated in water by using an Amicon Centricon S-3 centrifugation filter. The doublestranded hairpin state of the DNA was prepared by slowly cooling the oligonucleotides from 95°C to 5°C overnight in 0.1 M NaCl, 10 mM sodium cacodylate, pH 7.5 (low-salt condition).

616 Spectroscopic methods Absorption spectra were acquired with a Kontron Uvicon 820 spectrophotometer at 4°C and with 1 nm resolution. DNA oligonucleotide concentrations were determined using absorption extinction coefficients (in nucleotide units, mM−1 cm−1 ) calculated for singlestranded sequences at 25°C according to Puglisi & Tinoco (1989): m5CG4 TA12 , 349; m5CG6 TA10 , 343; m5CG8 TA8 , 337; m5CG10 TA6 , 331; and m5CG12 TA4 , 325. These values were corrected to 4°C and for a change in secondary structure (single-stranded to double-stranded) by first adjusting for the hyperchromism of the single-stranded state upon passing from 25°C to 70°C (multiplication with the factor 1.025 estimated from the analyses for m = 4 and 6, Figure 6B). The changes in extinction coefficients from 70°C to 4°C were then calculated from the empirical dependence of the UV absorbance of the oligonucleotides on temperature and sequence (derived from the fit curves 4° 70° for m = 4, 6, 8, Figure 6B): A260 /A260 = 0.750 + 0.027m. Measurements of A260 at 4°C were corrected for the contributions from the dyes according to a two-step procedure. (1) The individual absorption spectra of the donor and acceptor were calculated from the absorption spectra of double and single-labeled oligonucleotides by means of a decomposition into the individual spectral contributions (Figure 10B). (2) The corresponding UV absorbances of the dyes were calculated from spectral ratios determined from stock solutions of the reagents in Cy3 the absence of DNA: efl260 /efl490 = 0.30, and eCy3 260 /e550 = 0.15. The DNA concentrations, obtained from the net absorbances, yielded the extinction coefficients of the dyes in the fully structured hairpin duplexes (mean values −1 for all the oligonucleotides): eCy3 cm−1, 550 = 90 (29) mM efl490 = 25 (23) mM−1 cm−1. The internal consistency of these determinations confirmed the quantitative labeling of the single and double-labeled samples and the invariant nature of the environment sensed by the two dyes. Steady-state fluorescence measurements were performed with an SLM 8000S spectrofluorimeter at 4°C with 1 nm resolution. The emission spectra were corrected for instrument response, lamp fluctuations and solvent background contributions. Polarization artifacts were avoided by using ‘‘magic angle’’ conditions. Emission spectra were collected in the 500 to 650 nm range with excitation at 490 nm for single and double-labeled samples and 560 to 650 nm with excitation at 550 nm for double-labeled samples, using constant slit apertures and gain settings. Fluorescence anisotropies (r) were determined from measurements of fluorescence intensities using vertically polarized excitation and either vertically ( f> ) or horizontally ( f_ ) oriented emission polarizers; r = ( f> − f_ )/( f> + 2f_ ). Limiting anisotropy values were determined in a solid matrix according to Corin et al. (1987) by fusing racemic D,L-arabinose at 190°C, doping with <1% (v/v) of a Cy3-labeled DNA solution, and casting directly into a fluorescence cuvette. CD spectra were acquired at 4°C on a Jasco 720 CD spectrometer. Scanning was with a 1 nm bandwidth and 1 nm resolution. Fluorescence lifetime determinations were carried out with a phase-modulation fluorimeter (Piston et al., 1989). Salt titrations DNA duplexes were formed in 0.1 M NaCl, 10 mM sodium cacodylate (pH 7.5). Titrations were performed by adding increasing amounts of a 3.2 M stock solution of

DNA Handedness by FRET

MgCl2 in buffer containing 0.1 M NaCl, 10 mM sodium cacodylate. Absorption extinction coefficients required for the FRET calculations were corrected for changes in salt concentration as necessary. After each Mg2+ addition, samples were allowed to equilibrate for 30 minutes at 40°C, 50 minutes at room temperature, and ten minutes at 4°C, before acquisition of the spectra.

Acknowledgements The authors are grateful to D. Po¨rschke for acquisition of the thermal transition data, to C. Gohlke and E. Pap for the measurement of the Cy3 fluorescence lifetimes, to G. Heim for excellent technical assistance, and to R. M. Clegg, P. I. H. Bastiaens, V. V. Kuryavyi, E. Evertsz, A. K. Shchyolkina and L. Erijman for valuable discussions.

References Anderson, P. & Bauer, W. (1978). Supercoiling in closed circular DNA: dependence upon ion type and concentration. Biochemistry, 17, 594–601. Behe, M. & Felsenfeld, G. (1981). Effects of methylation on a synthetic polynucleotide: the B-Z transition in poly(dG-m5dC)·poly(dG-m5dC). Proc. Natl Acad. Sci. USA, 78, 1619–1623. Clegg, R. M. (1992). Fluorescence resonance energy transfer and nucleic acids. Methods Enzymol. 211, 353–388. Clegg, R. M. (1995). Fluorescence resonance energy transfer. Curr. Opin. Biotechnol. 6, 103–110. Clegg, R. M. (1996). Fluorescence resonance energy transfer (FRET). In Fluorescence Imaging Spectroscopy and Microscopy, Chemical Analysis Series (Wang, X. F. & Herman, B., eds), vol. 137, 179–251, Wiley, John & Sons, Inc., New York. Clegg, R. M., Murchie, A. I., Zechel, A. & Lilley, D. M. (1993). Observing the helical geometry of doublestranded DNA in solution by fluorescence resonance energy transfer. Proc. Natl Acad. Sci. USA, 90, 2994–2998. Corin, A. F., Blatt, E. & Jovin, T. M. (1987). Triplet-state detection of labeled proteins using fluorescence recovery spectroscopy. Biochemistry, 26, 2207–2217. Cozzarelli, N. R. & Wang, J. C. (1990). Editors of DNA Topology and its Biological Effects, Monograph Series, 20, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Dale, R. E., Eisinger, J. & Blumberg, W. E. (1979). The orientational freedom of molecular probes. The orientation factor in intramolecular energy transfer. Biophys. J. 26, 161–194. Dickerson, R. E., Drew, H. R., Wing, R. M., Fratini, A. V. & Kopka, M. L. (1982). The anatomy of A-, B- and Z-DNA. Science, 216, 475–485. Eimer, W., Williamson, J. R., Boxer, S. G. & Pecora, R. (1990). Characterization of the overall and internal dynamics of short oligonucleotides by depolarized dynamic light scattering and NMR relaxation measurements. Biochemistry, 29, 799–811. Eis, P. & Millar, D. P. (1993). Conformational distributions of a four-way junction revealed by time-resolved fluorescence resonance energy transfer. Biochemistry, 32, 13852–13860. Evertsz, E., Rippe, K. & Jovin, T. M. (1994). Parallelstranded duplex DNA containing blocks of trans

DNA Handedness by FRET

purine-purine and purine-pyrimidine base pairs. Nucl. Acids Res. 22, 3293–3303. Fo¨rster, T. (1948). Zwischenmolekulare Energiewanderung und Fluoreszenz. Ann. Phys. (Leipzig), 2, 55–75. Fo¨rster, T. (1951). Fluoreszenz Organischer Verbindungen. Vandenhoeck & Ruprecht, Go¨ttingen. Fujii, S., Wang, A. H., van der Marel, G., van Boom, J. H. & Rich, A. (1982a). Molecular structure of (m5dCdG)3 : the role of the methyl group on 5-methyl cytosine in stabilizing Z-DNA. Nucl. Acids Res. 10, 7879–7892. Fujii, S., Wang, A. H.-J., van Boom, J. H. & Rich, A. (1982b). Structural and dynamic aspects of a modified A-DNA octamer, a methylated Z-DNA and a DNA-RNA hybrid. Nucl. Acid Res. (Symp. Series), 11, 109–112. Garcia de la Torre, J., Navarro, S. & Lopez Martinez, M. C. (1994). Hydrodynamic properties of a double-helical model for DNA. Biophys. J. 66, 1573–1579. Gohlke, C., Murchie, A. I., Lilley, D. M. & Clegg, R. M. (1994). Kinking of DNA and RNA helices by bulged nucleotides observed by fluorescence resonance energy transfer. Proc. Natl Acad. Sci. USA, 91, 11660–11664. Jablonski, A. (1957). Decay of photoluminescence of solutions. Acta Physiol. Pol. 16, 471–479. Jovin, T. M., Soumpasis, D. M. and McIntosh, L. P. (1987). The transition between B-DNA and Z-DNA. Annu. Rev. Phys. Chem. 38, 521–560. Ju, J. Y., Ruan, C. C., Fuller, C. W., Glazer, A. N. & Mathies, R. A. (1995). Fluorescence energy transfer dye-labeled primers for DNA-sequencing and analysis. Proc. Natl Acad. Sci. USA, 92, 4347–4351. Klysik, J. K. R., Rippe, J. and Jovin, T. M. (1990). Reactivity of parallel-stranded DNA to chemical modification reagents. Biochemistry, 29, 9831–9839. Klysik, J. K. R. & Jovin, T. M. (1991). Parallel-stranded DNA under topological stress: rearrangement of (dA)15·(dT)15 to a d(A·A·T)n triplex. Nucl. Acids Res. 19, 7145–7154. Lilley, D. M. J. & Clegg, R. M. (1993). The structure of branched DNA species. Quart. Rev. Biophys. 26, 131–175. Malfoy, B., Rousseau, N. & Leng, M. (1982). Interaction between antibodies to Z-form deoxyribonucleic acid and double-stranded polynucleotides. Biochemistry, 21, 5463–5467. Mergny, J.-L., Boutorine, A. S., Garestier, T., Belloc, G., Rouge´ e, M., Bulychev, N. V., Koshkin, A. A., Bourson, J., Lebedev, A. V., Valeur, B., Thuong, N. T. & He´le`ne, C. (1994). Fluorescence energy transfer as a probe for nucleic acid structures and sequences. Nucl. Acids Res. 22, 920–928. Pattabiraman, N. (1986). Can the double helix be parallel? Biopolymers, 25, 1603–1606. Peck, L. J. & Wang, J. C. (1981). Sequence dependence of the helical repeat of DNA in solution. Nature, 292, 375–378.

617 Piston, D. W., Marriott, G., Radivoyevich, T., Clegg, R. M., Jovin, T. M. & Gratton, E. (1989). Wide-band acousto-optic light modulator for frequency domain fluorometry and phosphorimetry. Rev. Sci. Instrum. 60, 2596–2600. Pohl, F. M. & Jovin, T. M. (1972). Salt-induced co-operative conformational change of a synthetic DNA: equilibrium and kinetic studies with poly(dG-dC). J. Mol. Biol. 67, 375–396. Puglisi, J. D. & Tinoco, I. (1989). Absorbance melting curves of RNA. Methods Enzymol. 180, 304–325. Reich, Z., Friedman, P., Scolnik, Y., Sussman, J. L. & Minsky, A. (1993). On the metastability of left-handed DNA motifs. Biochemistry, 32, 2116–2119. Rippe, K., Fritsch, V., Westhof, E. & Jovin, T. M. (1992a). Alternating d(G-A) sequences form a parallelstranded DNA homoduplex. EMBO J. 11, 3777– 3786. Rippe, K., Kuryavyi, V. V., Westhof, E. & Jovin, T. M. (1992b). Polymorphism and possible biological functions of parallel-stranded DNA. In Structural Tools for the Analysis of Protein-Nucleic Acid Complexes. Advances in Life Sciences (ALS) (Lilley, D. M. J., Heumann, H. & Suck, D., eds), pp. 81–107, Birkha¨user Verlag, Basel, Switzerland. Rippe, K., Do¨tsch, V. & Jovin, T. M. (1995). Confirmation of strand orientation in parallel-stranded and anti-parallel-stranded DNA duplexes by fluorescence resonance energy transfer and pyrene excimer fluorescence. In DNA Interaction with Ligands and Proteins. Problems of Recognition and Self Organization. Fundamental Aspects and Technical Trends (Funck, T. et al., eds), Nova Science, Commarck, New York, U.S.A. In the press. Selvin, P. R. (1995). Fluorescence resonance energy transfer. Methods Enzymol. 246, 300–334. Shchyolkina, A. K., Lysov, Y. P., Il’ichova, I. A., Chernyi, A. A., Golova, Y. B., Chernov, B. K., Gottikh, B. P. & Florentiev, V. L. (1989). Parallelstranded DNA with AT base pairing. FEBS Letters, 244, 39–42. Sheardy, R. D. & Winkle, S. A. (1989). Temperaturedependent CD and NMR studies on a synthetic oligonucleotide containing a B-Z junction at high salt. Biochemistry, 28, 720–725. Tuschl, T., Gohlke, C., Jovin, T. M., Westhof, E. & Eckstein, F. (1994). A three-dimensional model for the hammerhead ribozyme based on fluorescence measurements. Science, 266, 785–789. (Erratum. (1995). Science, 267, 1581.) van de Sande, J. H., Ramsing, N. B., Germann, M. W., Elhorst, W., Kalisch, B. W., von Kitzing, E., Pon, R. T., Clegg, R. C. & Jovin, T. M. (1988). Parallel stranded DNA. Science, 241, 551–557. Weber, G. (1956). Photoelectronic methods for the measurement of the polarization of the fluorescence of solutions. J. Opt. Soc. Am. 46, 962–973. Wu, P. & Brand, L. (1992). Resonance energy transfer methods and applications. Anal. Biochem. 218, 1–13.

Edited by P. E. Wright (Received 22 September 1995; received in revised form 11 January 1996; accepted 17 January 1996)