Novel solution conformation of DNA observed in d(GAATTCGAATTC

Linda Cameron and Andrew S. Thompson. Biochemistry 2000 ... F. Leslie Boyd , Diana Stewart , William A. Remers , Mary D. Barkley , and Laurence H. Hur...
0 downloads 0 Views 883KB Size
Biochemistry 1987, 26, 1315-1 322

1315

Novel Solution Conformation of DNA Observed in d( GAATTCGAATTC) by Two-Dimensional NMR Spectroscopy K. V. R. Chary,* R. V. Hosur, and Girjesh Govil Tata Institute of Fundamental Research, Bombay 400005, India Tan Zu-kun and H. T. Miles National Institutes of Health, Bethesda, Maryland 20205 Received August 13, 1986; Revised Manuscript Received October 21, 1986

Resonance assignments of nonexchangeable base and sugar protons of the self-complementary dodecanucleotide d(GAATTCGAATTC) have been obtained by using the two-dimensional Fourier transform N M R methods correlated spectroscopy and nuclear Overhauser effect spectroscopy. Conformational details about the sugar pucker, the glycosidic dihedral angle, and the overall secondary structure of the molecule have been derived from the relative intensities of cross peaks in the two-dimensional N M R spectra in aqueous solution. It is observed that d(GAATTCGAATTC) assumes a novel double-helical structure. The solution conformations of the two complementary strands are identical, unlike those observed in a related sequence in the solid state. Most of the five-membered sugar rings adopt an unusual 01’-endo geometry. All the glycosidic dihedral angles are in the anti domain. The AATT segments A2-T5 and A8-T11 show better stacking compared to the rest of the molecule. These features fit into a right-handed D N A model for the above two segments, with the sugar geometries different from the conventional ones. There are important structural variations in the central T C G portion, which is known to show preferences for DNase I activity, and between Gl-A2 and GI-A8, which are cleavage points in the EcoRI recognition sequence. The sugar puckers for G1 and G7 are significantly different from the rest of the molecule. Further, in the three segments mentioned above, the sugar phosphate geometry is such that the distances between protons on adjacent nucleotides are much larger than those expected for a right-handed DNA. W e suggest that such crevices in the D N A structure may act as “hot points” in initiation of protein recognition.

ABSTRACT:

N u c l e i c acid-protein interactions play a vital role in the control of cellular activities. The interaction requires a specific recognition between the two molecules (Ohlendrof & Matthews, 1983). It is believed that the three-dimensional structures of individual molecules play a dominant role in the nature and specificity of interactions between proteins and nucleic acids. Nuclear magnetic resonance (NMR) spectroscopy has emerged as a very powerful technique for structural studies in solutions. It has provided information at a level of detail hitherto impossible from other methods (Govil & Hosur, 1982). With the advent of two-dimensional (2D) N M R techniques for elucidating J-coupling correlations (Jeener, 1975; Aue et al., 1976) and dipolar-coupling correlations (Ani1 Kumar et al., 1980), it has become possible to obtain unique resonance assignments in proteins and nucleic acids having molecular weights up to ca. 7000 (Wagner et al., 1981; Wagner & Wuthrich, 1982a,b; Arseniev et al., 1982; Hosur et al., 1983; Strop et al., 1983; Williamson et al., 1984; Reid et al., 1983; Clore & Gronenborn, 1984; Scheek et al., 1983, 1984; Hare et al., 1983; Feigon et al., 1983a,b; Frichet et al., 1984). The assignment strategies for nucleic acids have been described recently (Scheek et al., 1983; Hare et al., 1983; Hosur et al., 1985a; Gronenborn & Clore, 1985), and new strategies for structure determination of oligonucleotides have been proposed (Hosur et al., 1985a, 1986; Ravi Kumar et al., 1985). The new ideas make use of correlated spectroscopy for the determination of sugar geometries in the oligonucleotides. Here we report our investigations on a self-complementary dodecanucleotide, d(GAATTCGAATI’C) (Figure 1). This paired dodecamer has four specific cleavage sites 0006-2960/87/0426-1315$01.50/0

for the restriction endonuclease EcoRI, the unit sequence that is recognized being GAATTC. Also, the TCG portion of the dodecamer is of biological significance, since the enzyme DNase I is known to have a strong preference for cutting nucleic acids at TCG sites (Dickerson & Horace, 1981). The crystal structure has been solved for the self-complementary oligonucleotide d(CGCGAATTCGCG) (Dickerson & Horace, 198l), where the EcoRI recognition sequence is flanked by CGC and GCG at the two ends. In this structure, it has been observed that the TCG fragment has a large helical twist as compared to other positions and is thus postulated to be more accessible to the enzyme (Dickerson & Horace, 1981). The present studies on d(GAATTCGAATTC) indicate that a part of the molecule has a right-handed DNA conformation that is different from the earlier proposed right-handed DNA structure. There are important deviations from such a structure in the protein recognition sites, Le., GI-A2, G7-A8, and T5-C6-G7, which possibly make them accessible to the respective enzymes. EXPERIMENTAL PROCEDURES Synthesis. d(GAATTCGAATTC) was prepared by a solid-phase [ (aminomethy1)polystyrene with 1% divinylbenzene] phosphotriester method as described by Tan et al. (1982). The 3’-linked dC resin was prepared and coupled to dT monomer. The protected dimers d(AT), d(GA), d(TC), d(AT), and d(GA) were then coupled in turn to the 5’-OH of the growing chain. The oligomer was cleaved from the resin, deblocked, and purified by ion-exchange chromatography on (diet hylaminoet hy1)cellulose (DEAE-cellulose) [7 M urea, 0.15-0.4 M NaCl gradient; cf. Tan et al. (1982)]. A center 0 1987 American Chemical Society

1316

BIOCHEMISTRY EcoRI

C H A R Y ET A L .

1

DNAase1 E c o R I

I

d-GAATTCGAATTC

i 3’

C T 23

T

A A 21

I

1

I

G C T T A A G-d5’ 119

117

EcoRI DNAase1

15

113

EcoRI

FIGURE 1: Structure of d(GAATTCGAATTC), showing preferential cleavage sites for nucleases.

cut of the main peak gave a single peak on high-performance liquid chromatography (HPLC) (Hamilton PRP- 1 column; 0.10 M Et,NHOAc, 0-25% acetonitrile gradient). The sequence was confirmed by the method of Maxam and Gilbert (1980). NMR. For the ‘H NMR experiments on d(GAATTCGAATTC), 5 mg of the sample (approximately 4 mM DNA strand) was dissolved in 0.5 mL of phosphate buffer (0.02 M) having pH 7.2. Total counterion concentration was 0.08 M. Such a solution was lyophilized and redissolved in 99.97% D 2 0 twice, and finally it was made up to 0.5 mL with 99.97% D20. The spectra were recorded on a Bruker AM-500 Fourier transform (FT) NMR spectrometer whose operational frequency for proton is 500 MHz. In the case of the twodimensional correlated spectroscopy (COSY) experiment, the conventional pulse sequence was modified to (-9O0-t1-A9O0-A-t2-fid),, where A is a fixed delay of 5 ms and n is the number of transients accumulated for each t l value (Ani1 Kumar et al., 1984; Hosur et al., 1985b). Such an experiment was carried out with 2048 and 512 data points along the t 2 and tl axes, respectively, with continuous low-power irradiation of the HDO signal at 4.8 ppm to avoid dynamic-range problems of the analog to digital converter (ADC). The nuclear Overhauser effect spectroscopy (NOESY) experiments were carried out with the pulse sequence (-90°-t1-900-~,9O0-t2-fid),, where 7, is the mixing time. Experiments were carried out with 1024 and 300 data points along the t2 and t l axes, respectively. A pulse repetition time of 1.2 s was used in all the 2D experiments. The COSY and NOESY timedomain data were zero-filled to 1024 points along the t l direction and multiplied by shifted sine squared bell and sine bell window functions along t2 and t , directions, respectively, prior to the two-dimensional Fourier transformations. Thus, the frequency domain spectra consisted of 512 X 512 and 256 X 5 12 point data matrices in the cases of COSY and NOESY spectra, respectively. At this stage it may be pointed out that the appearance and intensities of the cross peaks in the NOESY spectrum eritically depend on the mixing time T, among other factors. In this endeavor, several 2D NOESY experiments were carried out by varying the value of T,, and the one with T, = 500 ms was selected for sequential assignment. It may be mentioned that spin diffusion is not a serious limitation for sequential assignment. Chemical shifts are expressed with respect to sodium 3(trimethylsilyl) [2,2,3,3-2H]propionate (TSP). RESULTS AND DISCUSSION The one-dimensional 500-MHz ‘H NMR spectrum of d(GAATTCGAATTC) in D 2 0 at 298 K is shown in Figure 2. The low-field region between 7.0 and 8.3 ppm contains base protons (H2, H6, and H8) while the region between 5.0 and 6.5 ppm contains the H5 cytosine and the sugar H1’ protons. Other sugar protons (H2’, H2”, H3’, H4’, H5’, and H5”) are contained in the highly crowded spectral region between 2.0 and 5.0 ppm. The methyl resonances of thymines are seen distinctly at 1.28, 1.51, and 1.60 ppm. The relative

//I

I 1 I

70

FIGURE

2:

IlIIY

/

5 0

DDU

3 0

1’0

1

500-MHz proton NMR spectrum of d(GA-

ATTCGAATTC) in D20solution at 298 K.

intensities indicate that two of the four methyl resonances overlap at 1.28 ppm. Resonance Assignments Figure 3 summarizes the steps involved for individual assignments by using COSY and NOESY, which have been shown to work for right-handed DNA structures. The whole assignment procedure is rather intricate and interdependent and one must go through several iterations between the COSY spectra (Figures 4 and 5) and NOESY spectra (Figures 6 and 7) before reaching the final assignments. Once the process is completed, there are adequate cross-checks to ensure the unambiguity of assignments and for the conclusions on structure based on such assignments. Spin System Identification. The spin system within the individual nucleotide units is identified from the two-dimensional COSY spectrum following the intranucleotide J-connectivity pathways: Hl’-(H2’, H2”)-H3’-H4’-(H5’, H5”), C H5-C H6, and T CH3-T H6. In each case, distinction between H2’ and H2” protons can be made from the NOESY spectrum, which, in view of the interproton distances, will show a stronger NOE between H1’ and H2” protons compared to H1’ and H2’ protons (Hosur et al., 1986). Some authors (Hare. et al., 1983) have used the multiplicity pattern in H 1’-H2’, H 1’-H”’ COSY cross peaks for such an identification. However, since multiplicities depend on coupling constants, which in turn are strongly dependent on the sugar geometry, we feel that this is not a very reliable criterion. Figure 4 shows the contour plot of the experimental COSY spectrum of d(GAATTCGAATTC) in D 2 0 at 298 K. In this spectrum, the base protons (H5 and H6) of the two cytosine residues (lower left blowup) and the methyl protons and the base H6 protons of the four thymine nucleotides (upper left blowup) are easily identified. The more complicated connectivity pattern in the COSY spectrum is the one involving sugar Hl’, H2’, H2”, H3’, and H4’ protons. Parts a, b, and c of Figure 5 show expansions of selected regions depicting H l’-(H”-H2’’), H2’-”’’, and H3’-(H2’-H4’) J-coupling correlations, respectively. These spectra enable one to group the sugar protons HI’, H2’, H2”, H3’, and H4’ of individual residues. The stereospecific assignments of H2’ and H2” and the assignment of H4’ protons are aided by NOESY spectra (Figures 6 and 7a, respectively). The spin systems are correlated with the nucleotide type (G, A, T, or C) by use of sugar-base connectivities in the NOESY spectra (Figure 7b,c). Sequential Assignments. Assignment of individual spin systems identified in the previous step to particular nucleotide

VOL. 26, NO. 5 , 1987

CONFORMATION OF D(GAATTCGAATTC)

1317

(a1

?I-

a

n2.

3: Strategies of sequential resonance assignments for B-DNA. Double-pointed arrows indicate observable COSY and N O S Y correlations that can be. used in sequential assignment. (a) Intranucleotide COSY correlations. (b) Intranucleotide NOEs (thin arrows) and internucleotide NOEs (thick arrows). Pathways d l , d2, and d3 are also indicated. FIGURE

_---

a- GAATTCGAATTC

Table I: Chemical Shifts of Nonexchangeable Protons in d(GAATTCGAATTC1 unit H1’ H2’ H2” H3’ H4’ H6/H8 H2/H5/CH3 5.51 2.44 2.64 4.84 4.17 G1 7.5w A2 5.98 2.80 2.92 5.08 4.43 8.23 8.18 6.20 2.62 2.91 5.03 4.55 A3 7.14 1.28 T4 5.88 1.97 2.51 4.83 4.16 T5 6.03 2.10 2.45 4.87 4.16 7.33 1.51 C6 5.45 1.91 2.24 4.88 4.07 7.36 5.53 G7 5.35 2.61 2.70 4.97 4.30 7.83 A8 5.92 2.67 2.87 5.05 4.41 8.11 A9 6.11 2.55 2.86 4.96 4.42 8.09 7.72a T10 5.88 1.97 2.51 4.83 4.18 7.12 1.29 .T11 6.14 2.16 2.51 4.89 4.18 7.40 1.60 C12 6.26 2.25 2.25 4.60 4.03 7.64 5.80 Observed from interstrand internucleotide NOESY cross peaks (Figure 1).

. 0

I

8 0

6 0

I

4 0

2 0

w ~ ( P P ~ )

4: Contour plot of 500-MHz proton COSY spectrum of d(GAATTCGAATTC) in D20at 298 K. The digital resolution in this spectrum is 7.50 Hz/point. The blowups on the left-hand side of this figure depict the expected J-coupling correlations between methyl and H6 base protons of the thymine residues and between H5 and H6 base protons of the cytosine residues. The sections enclosed in the boxes have been expanded in Figure 5 for clarity. FIGURE

units along the chain is obtained from the two-dimensional NOESY spectrum by employing the following intra- and internucleotide through-space connectivities. The internucleotide NOE pathways are

(H8/HS),-(H8/HS/CH,/H5),,,

(HS/H6)j-(Hl’)j-1-(H8/H6)j-I

(dl) (d2)

and (H8/H6)i-(H2’, H2”),-1-(H8/H6)1-1

(d3)

Here the subscript i refers to the position of the residue from the 5‘ end of the sequence. Such a strategy works in cases where the internucleotide configuration is such that the distances d l , d2, and d3 are within the limits needed for a reasonable NOE (e.g., in A- or B-DNA). The details of sequential connectivities are shown in Figure 7b,c: Figure 7b depicts the intra- and internucleotide (H8/H6)-Hl’ connec-

tivities, and Figure 7c shows similar (H8/H6)-(H2’,H2’’) connectivities. The cross peaks between base H6/H8 protons on adjacent nucleotides are shown in Figure 7d. A summary of all the observed sequential N O E connectivities is given in Figure 8. The sequential assignment described above provides the chemical shifts of base, Hl’, H2’, H2”, H3’, and H4’ protons of all 12 nucleotide units, and these are listed in Table I. The assignment of the H5’ and H5” protons remains difficult due to extensive overlap.

Structure of d(GAATTCGAATTC) In Figure 7 there are several unlabeled NOEs, most of which are due to intranucleotide NOEs and can be fully explained. In addition, in Figure 7d two interstrand NOESY connectivities are seen, namely, from H6 of C12 to H2 of A2 and H1’ of T4 to H 2 of A9. These confirm that the molecule is in the double-helical form under the present experimental conditions and rule out the possibilities of hairpin or circular loop structures. The fact that all the cross peaks have been assigned in the NOESY spectrum is consistent with the existence of only one type of double-helical structure in aqueous solution. Further, we observe only one set of signals for a particular nucleotide unit, and this points to the equivalence of the corresponding nucleotide units on the two strands (e.g., G1 = G13, A2 = A14, etc.). Such an equivalence is expected

1318

BIOCHEMISTRY

C H A R Y ET AL.

self-complementary oligonucleotide. Besides, the COSY and

d- GAATTCGAATTC

NOESY results have provided some finer details of the conB

T(4,IO)

C6

20@T5

@E

C6

( H 2 I, H 2” )

2 5-

3 0 1

I

formation with reference to sugar geometries, glycosidic dihedral angles, and base-base stacking of individual nucleotide units along the sequence of this molecule. These details are discussed below. Sugar Geometries. The conformation of the furanose ring plays a central role in the overall structure of double-helical DNA and RNA. At least one of the five atoms forming the furanose ring has to deviate from the plane described by the other three (or four) atoms. The conformation can be described then by the position of the puckered atom with respect to the direction of C1’-N and C4’-C5’ bonds, e.g., CZ‘-endo, C2’-exo, etc. A more generalized description of the sugar conformation is based on the concept of pseudorotation (Altona & Sundaralingam, 1972), which allows for pucker of two atoms. The conformation can then be described by the pseudorotation angle P.

CONFORMATION OF D(GAATTCGAATTC)

VOL. 26, NO. 5 , 1987

57

A? A3 A9/A8

1319

Tl! T5 Tf, ,T10

-- G7

I

8.0 5.2

5.6

6.0

I

7.6

,

bl

I

7.2

H6/ H8

a 6.4

.

I



I

H 1’ A2,A3

A?A8

7 )

G7

C12 TI! C 6 T 5 TP T l O

,,T4(CH31

‘.TIO(CH~) -.T5(CH3) --Tll(CHd

H 2’,H2 ”

-E .m _ m

--C6

a a

-

--G 1 ,-TI1

‘-TI0 ’ G 7

“A8 --A2 :‘A9

C

80

76

‘A 3

72

H6/H8

FIGURE 7: Expansion of selected regions of NOESY spectrum recorded with 7, = 500 ms. (a) This panel contains all 12 Hl’-H4’ NOESY connectivities. (b) Sequential connectivities using base H6/H8 protons and H1’ protons (d2 pathways) are shown in this panel. A vertical line to a cross peak indicates NOEs between a base proton of the nth residue to the H1’ proton of the preceding residue. A horizontal line joins NOEs between the H1’ proton of the nth residue and its own base proton and that between the former proton and the base proton of succeeding residue. Identification of H1’ protons is given on the right-hand side of the panel, and that of base protons is given at the top of the panel. (c) Sequential connectivities using base H8/H6 protons and H2’/H2’’ protons (d3 pathways) are depicted in this panel. Vertical and horizontal lines have the same meaning as in panel b. Identification of H2’/H2’’ protons is given on the right-hand side of the panel, and that of base protons is given at the top of the panel. The panel also contains intrastrand internucleotide base to base connectivities (cross peaks arising from the dl pathway), which are shown with dashed lines. (d) Interstrand internucleotide base to base connectivities, namely, cross peaks arising from the dl pathway, are shown with solid lines. The panel also contains intrastrand connectivities, which are shown with dashed lines.

resolution along o1and w2 axes which arise because of limited experimental time and storage space available, the intensities of the cross peaks in a COSY spectrum are approximately proportional to the values of the coupling constants (J)between the two nuclei. The three bond (vicinal) coupling constants between sugar protons Hl’, H2’, H2”, H3’, and H4’ depend in a predictable manner on the sugar geometry (Figure 9). Therefore, from a careful analysis of the relative intensities of different intranucleotide COSY cross peaks, it is possible

to identify the sugar pucker of individual nucleotide units. Examination of the plots in Figure 9 reveals that the values of J(Hl”H2”) and J(H2’-H3’) vary within a narrow range of 6-10 Hz and are thus relatively insensitive to sugar geometry. On the other hand, the values of J(Hl’-H23, J(H2”H3’),and J(H3’-H4’) are very sensitive to the pseudorotation angle and can be used in fixing the domains of sugar geometry. For example, in the case of C3’-endo geometry, one expects strong cross peaks in the COSY spectrum arising from all the

1320

BIOCHEMISTRY I 2 3 4 5 6 7 8 9 1 0 1 1 1 2 A A T T C G A A T T C 3’

5’d-G

FIGURE 8:

h n m a r y of all sequential connectivities observed in the

NOESY spectra due to short distances (less than ca. 5 A) between the base H8/H6 proton of the (n + 1)th nucleotide and sugar protons in the nth nucleotide: dl refers to base-base connectivity, d2 refers to base-H 1’ connectivity, and d3 refers to base“HZ’lH2’’) con-

nectivity. The blanks indicate regions where the relevant cross peak was not observed in the NOESY spectrum.

-2

t

PSEUDO ROTATION ANGLE ( P ) Plots of variation of different vicinal coupling constants among sugar ring protons as a function of pseudorotation parameter. FIGURE 9:

vicinal couplings with the exception of Hl’-H2’ [J(Hl’-H2’) = 01. In the case of C2’-endo geometry, one expects strong cross peaks for Hl’-H2’, Hl’-H2’’, and H2’-H3’ but no cross peak for H2”-H3’ and H3’-H4’, since the corresponding J values are very small (-0 Hz). In the event of an equilibrium between CY-endo and C3‘-endo geometries, a situation commonly encountered in small nucleotides, all the couplings will be weighted averages and will be large. The COSY spectrum will then show all cross peaks with almost equal intensities. In the COSY spectrum of d(GAATTCGAATTC), none of the nucleotides show any H2”-H3’ cross peaks (Figure 5c). This rules out N + S equilibrium and implies that the range of deoxyribose geometry for all the nucleotide units is limited to the so-called S region, i.e., within the pseudorotational angle P between 90’ (01’-endo) and 270’ (01’-exo). The classical C2’-endo geometry used in the B-DNA model (P= 162’) falls in the middle of this range. In addition, the COSY spectrum shows strong intranucleotide cross peaks corresponding to Hl’-H2’, Hl’-H2”, H2’-H3’, and H3’-H4’ for 9 of the 12 nucleotide units (Figure 5 ) . These are A2, A3, T4, T5, C6, AS, A9, T10, and T11. The appearance of strong H3’-H4’ cross peaks further limits the P value to the range 90-1 30’ in the case of the nine nucleotides mentioned above. Moreover, since the Hl’-H2’ and Hl’-H2” COSY cross peaks (Figure sa) have similar intensities, we conclude that the sugar geometry is in the neighborhood of 01’-endo. In the case of G1, the COSY spectrum shows strong cross peaks corresponding to Hl’-H2’ and Hl’-H2’’. Significantly,

C H A R Y ET A L .

in these two cases, the cross peaks corresponding to H3‘-H4‘ are absent. On the basis of the above arguments, the conformation for these two sugars is deduced to be such that the angle P lies between 160’ and 240° (C2’-endo to C4’-endo). The situation for C12 is complicated by the near equivalence of the chemical shifts for H2’ and H2” protons. This results in a single COSY cross peak in Figure 5a for C12. However, the intensity of this cross peak is much higher than those of all the other peaks in the figure, suggesting that the cross peak has substantial contributions from both H2’ and H2” protons. Likewise, the single cross peak between H3’ and (H2’, H2”) (Figure 5b) has the largest intensity, suggesting that again both H2’ and H2” protons make significant contributions. These observations along with the fact that the intense H3’-H4’ cross peak is also observed suggests that the sugar geometry for C12 is to the left of 01’-endo (N, P = 72’). Glycosidic Dihedral Angle. The relative magnitudes of the intranucleotide and internucleotide (H8/H6)-H1’ and (H8/H6)-(H2’, H2”) cross peaks in the NOESY spectra at different mixing times can be used to establish the domains of glycosidic dihedral angles of individual nucleotide units (Hosur et al., 1985a; Ravi Kumar et al., 1985). Below the spin-diffusion limit, the intensity patterns of the cross peaks look similar at all mixing times although the absolute intensities may vary with the mixing time. The expected intensity patterns for the above-mentioned cross peaks for different glycosidic dihedral angles (x)are given below. (i) For the syn conformation, a strong NOE between base H8/H6 and H1’ protons should be observed. At the same time, the NOES from base to H2’ and H2” protons will be relatively weak and will have different intensities. (ii) In the anti conformation, the NOE from base H8/H6 to H2’ is stronger than the NOE from base H8/H6 to H2”. Also, for right-handed structures the H2” proton shows a stronger NOE to the base proton of the next nucleotide. (iii) In the high anti conformation, the (H8/H6)-H2’ and (H8/H6)-H2” NOES will have similar intensities for CY-endo geometry and will have different intensities for C3’-endo sugar geometry. In the present case, the relative magnitudes of the abovementioned cross peaks at a mixing time of 300 ms, in conjunction with the fact that the sequential d3 connectivities (Figure 7) proceed via the H2” protons, establish that x is in the anti domain in all 12 nucleotides. This also rules out the structural possibilities of A-DNA and alternating B-DNA structures (Munt & Kearns, 1984). Base-Base Stacking. The chemical shifts given in Table I and the relative magnitudes of internucleotide base-base NOESY cross peaks in the NOESY spectrum reveal some interesting structural trends. (i) The chemical shifts of the H1’ proton reflect the symmetry of the molecule. For example, the H1’ protons of A3 and A9 appear downfield with respect to those of A2 and A8, respectively. A similar pattern is also seen in the case of thymine H1’ protons. These are consistent with the general observation that purines produce large ring current shifts on adjacent nucleotides as compared to pyrimidines. Cytosines C6 and C12 and guanines G1 and G7 show a large difference in H1’ chemical shifts, suggesting that the molecule does not undergo end-to-end aggregation. This is also consistent with the observation of sharp resonance lines for the individual protons. (ii) In the case of H2’ and H2” protons, the base type is reflected in their chemical shifts; the protons of pyrimidine nucleotide units are farther upfield than those of purine nu-

CONFORMATION OF

D(GAATTCGAATTC)

cleotide units. Such a trend has also been observed in other nucleotides, namely, d(TGAGCGG) and d(CCGCTCA) (Scheek et al., 1983), d(CGCGAATTCGCG) (Hare et al., 1983), and d(GGATCCGGATCC) (Hosur et al., 1985a,b; Ravi Kumar et al., 1985). Similarly, it is observed that the chemical shift differences between H2’ and H2” are larger in pyrimidines when compared with those in purines, the only exception being the terminal C12 nucleotide unit, which has equivalent H2’ and H2” protons. Thus, these general trends can provide useful guidelines for future assignments. (iii) In the NOESY spectrum the Ti+l H6-Ti CH3, Ti H6-Ti CH3, and A, H8-Ti+, CH3 NOESY cross peaks are of almost same intensity. This implies that these interproton distances are nearly equal, suggesting a base-stacking pattern observed in right-handed helices. Besides, the fact that we do not see any other base-base intrastrand internucleotide NOESY cross peaks indicates AATT segments have better stacking compared to the rest of the molecule. General Comments on the Solution Conformation of d(GAATTCGAATTC). One of the significant findings of our study is the dominance of the deoxyribose conformation with P values in the range 90’-130’. Such a sugar conformation, though still in the S range, deviates considerably from the C2’-endo geometry ( P = 162’) used in the proposed structure of B-DNA. It may, however, be pointed out that the proposed structures of two right-handed DNAs, Le., A and B forms from fiber-diffraction data, rely heavily on model building. In these cases, the use of idealized C3’-endo and C2’-endo geometry is an assumption that is usually made for convenience. The only known crystal structure on a molecule as large as the one studied here, namely, d(CGCGAATTCGCG), shows wide local variations of the sugar geometry ranging from C3’-endo through 01’-endo to C3’-exo. We have recently studied the structure of d(CG)6 under low-salt conditions, which can be used as an idealized model for B-DNA. In this case, the central part of the molecule shows a C1’-exo sugar pucker (Sheth et al., 1986). A significant difference between the solution structure discussed above and the crystal structure of d(CGCGAATTCGCG) (which has the underlined sequence in common with ours) is that the two strands in the double helix have different conformations in the solid state. In the fragment GAATTC, the sugar ring conformations in sequential order in the two strands are G (C2’-endo, C2’-endo), A (C2’-endo, C2’-endo), A (C1’-exo, Cl’-exo), T (01’-endo, 01’-endo), T (01’-endo, C1’-exo), and C (C2’-endo, 01’endo). Thus, the last two sugars show widely different conformations in the two strands. In solution, we observe a symmetrical structure for the two strands with a dominance of sugar pucker close to 01’-endo with the exception of G1 and G7, which acquire a conformation between C2’-endo and C3’-endo and C3’-exo. Thus, the sugar conformations exhibited by d(GAATTCGAAlTC) are more uniform, with the only wide variation in sugar pucker occurring between G1 and A2 and G7 and A8. The solution structure has most of the features of B-DNA (base in anti conformation, sugar in the S domain, etc.); however, the sugar pucker is significantly to the left of C2‘-endo conformation used in B-DNA structures. An interesting behavior is observed in the central TCG segment of the nucleotide. As shown in Figure 8, we failed to observe any of the three intrastrand internucleotide cross peaks arising from the distances d l , d2, and d3 between T5 and C6. This shows that these distances are considerably larger (>5 A), indicating a structural variation in this segment of the molecule. It may be pointed out that such a structural

VOL. 26, NO. 5 , 1987

1321

variation has also been observed in the TCG segment of d(CGCGAATTCGCG) in X-ray crystallographic analysis (Dickerson & Horace, 1981). It is possible that the strong preference of the TCG segment for the enzyme DNase I may be related to the geometrical variation of the structure. Further, the d3 connectivities in NOESY spectra recorded at low mixing times are weak in the molecular fragments G7-A8 and G1-A2 as compared to the rest of the molecule, indicating local structural variations. In addition, one finds that at these two points the G1 and G7 have sugar puckers that are significantly different from the rest of the nucleotide units. These structural variations are significant, considering that d(GAATTC) is a recognition site for EcoRI enzyme, which cleaves the molecule between G and A. A similar observation of structural variation at recognition sites was also made in d(GGATCCGGATCC) (Hosur et al., 1985; Ravi Kumar et al., 1985). The weak A2-G1 d3 connectivity may, however, be attributed to end effects. We thus conclude that the segments A2-T5 and A8-Tll in d(GAATTCGAATTC) show a uniform highly stacked right-handed DNA structure having features similar to though not identical with B-DNA. In contrast, there are very significant structural variations in the segments G 1-A2 and G7-A8, which are cleavage sites of EcoRI, and in T5-C6-G7, which is the cleavage site of DNase I. The internucleotide distances in these segments are larger than those expected for right-handed DNA. We coupled our results on d(GGATCCGGATCC) and d(GAATTCGAATTC), and it appears that local variations in solution conformations in specific protein binding sites may be important in recognition and cleavage by the corresponding nucleases. Future work on related systems may help elucidate the source of specificity of restriction endonucleases. ACKNOWLEDGMENTS The help provided by the 500-MHz FT-NMR National Facility at Tata Institute of Fundamental Research, Bombay, is gratefully acknowledged. Registry No. d(GAATTCGAATTC), 105449-07-8; nuclease, 9026-81-7.

REFERENCES Altona, C., & Sundaralingam, M. (1972) J . Am. Chem. SOC. 94, 8205-8210. Anil Kumar, Wagner, G., Ernst, R. R., & Wuthrich, K. (1980) Biochem. Biophys. Res. Commun. 96, 1156-1 163. Anil Kumar, Hosur, R. V., & Chandrasekhar, K. (1984) J . Magn. Reson. 60, 143-148. Arseniev, S. A., Wider, G., Joubert, F. J., & Wuthrich, K. (1982) J . Mol. Biol. 159, 323-351. Aue, P., Bartholdi, E., & Ernst, R. R. (1976) J . Chem. Phys. 64, 2229-2246. Clore, G. M., & Gronenborn, A. M. (1984) Eur. J . Biochem. 141, 119-129. Davies, D. B. (1978) Prog. Nucl. Magn. Reson. Spectrosc. 12, 135-225. Dickerson, R. E., & Horace, R. D. (1981) J . Mol. Biol. 149, 761-786. Feigon, J., Wright, J. M., Leupin, W., Denny, W. A., & Kearns, D. R. (1982) J . Am. Chem. SOC.104, 5540-5541. Feigon, J., Denny, W. A,, Leupin, W., & Kearns, D. R. (1983a) Biochemistry 22, 5930-5942. Feigon, J., Leupin, W., Denny, W. A., & Kearns, D. R. (1983b) Biochemistry 22, 5943-5951. Frichet, D., Cheng, D. M., Kan, L.-S., and T’so, P. 0. P. (1983) Biochemistry 22, 5194-5200.

1322

Biochemistry 1987, 26, 1322-1326

Govil, G., & Hosur, R. V. (1980) Conformation of Biological Molecules; New Results f r o m N M R , Springer-Verlag, Heidelberg. Gronenborn, A. M., & Clore, G. M. (1985) Prog. Nucl. Magn. Reson. Spectrosc. 17, 1-32. Hare, D. R., Wimmer, D. L., Cohn, S. H., Drobny, G., & Reid, B. R. (1983) J . Mol. Biol. 171, 319-336. Hosur, R. V., Wider, G., & Wuthrich, K. (1983) Eur. J . Biochem. 130, 497-508. Hosur, R. V., Ravi Kumar, M., Roy, K. B., Tan, Z., Miles, H. T., & Govil, G. (1985a) Magnetic Resonance in Biology and Medicine, pp 243-260, Tata McGraw-Hill, New Delhi. Hosur, R. V., Chary, K. V. R., Anil Kumar, & Govil, G. (198513) J . Magn. Reson. 62, 123-127. Hosur, R. V., Ravi Kumar, M., Chary, K. V. R., Sheth, A,, Govil, G., & Miles, H. T. (1986) FEBS Lett. 205, 71-76. Jeener, I. (1971) Ampere International Summer School, Basko Polje, Yugoslavia. Maxam, A. M., & Gilbert, W. (1980) Methods Enzymol. 65, 499-560. Munt, N. A., & Kearns, D. R. (1984) Biochemistry 25, 791-796. Ohlendrof, D. H., & Mathews, D. W. (1983) Annu. Rev. Biophys. Bioeng. 12, 259-284.

Ravi Kumar, M., Hosur, R. V., Roy, K. B., Miles, H. T., & Govil, G. (1985) Biochemistry 24, 7703-771 1. Reid, D. G., Salisbury, S. A., Bellard, S., Shakked, Z . , & Williams, D. H. (1983) Biochemistry 22, 2019-2025. Scheek, R. M., Russo, N., Boelens, R., Kaptein, R., & van Boom, J. H. (1983) J . A m . Chem. SOC.105, 2914-2916. Scheek, R. M., Boelens, R., Russo, N., van Boom, J. H., & Kaptein, R. (1984) Biochemistry 23, 1371-1376. Sheth, A., Ravi Kumar, M., Hosur, R. V. Tan, Z., Miles, H. T., & Govil, G. (1987) Biopolymers (in press). Strop, P., Wider, G., & Wuthrich, K. (1983) J . Mol. Biol. 166, 641-665. Tan, Z . , Ikuta, S., Huang, T., Dugaiczyk, A., & Itakura, K. (1982) Cold Spring Harbor Symp. Quant. Biol. 47, 383-391. Wagner, G., & Wuthrich, K. (1982a) J. Mol. Biol. 159, 347-366. Wagner, G., & Wuthrich, K. (1982b) J. Mol. Biol. 160, 334-340. Wagner, G., Anil Kumar, & Wuthrich, K. (1981) Eur. J . Biochem. 114, 375-384. Williamson, M. P., Marion, D., & Wuthrich, K. (1984) J . Mol. Biol. 173, 341-359.

Structure of Human Tumor Necrosis Factor

a!

Derived from Recombinant DNA

Janice M . Davis, Michael A. Narachi, N. Kirby Alton, and Tsutomu Arakawa*

Amgen, Thousand Oaks, California 91320 Received May 2, 1986; Revised Manuscript Received October 2, I986

ABSTRACT: Recombinant D N A derived tumor necrosis factor cy, when expressed at a high level in Escherichia

coli,appeared in the pellet and soluble fractions of disrupted cells. The protein was purified from the pellet fraction by solubilizing it in urea and reducing agent and was refolded into a buffer without these additives. The structure of the protein was identical with that purified from the soluble fraction without exposure to both reducing and denaturing agents, as demonstrated by circular dichroism, gel filtration, and sulfhydryl titration. As a reflection of the structural similarity, both purified proteins showed identical cytolytic activity on mouse L929 cells. The protein was characterized as an essentially nonhelical and P-sheet-rich structure and possibly as a noncovalently associating oligomer. Two cysteine residues form an intrapolypeptide disulfide bond.

x m o r necrosis factor cy (TNF-a)’ was first observed by Carswell et al. (1975) in the serum of endotoxin-treated mice and rabbits that had previously been sensitized with Bacillus Calmette-Guerin. This factor caused hemorrhagic necrosis of various tumors in mice. In addition, it has been shown that TNF-a exhibits cytolytic or cytostatic activities against animal and human transformed cell lines in vitro, but normal cell cultures seem unaffected (Carswell et al., 1975; Ruff & Gifford, 1981; Matthews, 1981, 1982; Hammerstrom, 1982; Helson et al., 1985; Green et al., 1976). The amino acid sequence of human TNF-a has been deduced from cDNA clones (Pennica et al., 1984; Marmenout et al., 1985; Wang et al., 1985; Shirai et al., 1985) and also directly determined

* Author to whom correspondence should be addressed.

from the protein purified from serum-free cell culture supernatants of the HL-60 promyelocytic leukemia cell line induced by 5P-phorbol 12-myristate 13-acetate (Aggarwal et al., 1985). The gene coding for TNF-a has been cloned, and the corresponding protein has been expressed in Escherichia coli. It has been shown that purified recombinant E . coli derived TNF-a also has cytostatic or cytolytic effects on tumor cell lines (Sugarman et al., 1985; Wang et al., 1985) and tumor 1 Abbreviations: TNF-a, tumor necrosis factor a; CD, circular dichroism; DTNB, 5,5’-dithiobis(2-nitrobenzoicacid); DTT, dithiothreitol; SDS, sodium dodecyl sulfate; SDS-PAGE, SDS-polyacrylamide gel electrophoresis; Tris-HC1, tris(hydroxymethy1)aminomethane hydrochloride.

0006-2960/87/0426-1322$01 SO10 0 1987 American Chemical Society