J. Mol. Biol. (1991)221, 645-667

DNase I-induced DNA

Conformation

2 A Structure of a DNase I-Octamer Armin Lahm’f and Dietrich

Complex

Suck$

European Molecular Biology Laboratory Biological Structures Division Postfach 102209, D-6900 Heidelberg, Germany (Received

20 March

1991; accepted

16 July

1991)

The structure of a complex between DNase I and d(GCGATCGC), has been solved by molecular replacement and refined to an R-factor of 0.174 for all data between 6 and 2 a resolution. The nicked octamer duplexes have lost a dinucleotide from the 3’ ends of one strand and are hydrogen-bonded across a S-fold axis to form a quasi-continuous double helix of 14 base-pairs. DNase I is bound in the minor groove of the B-type DNA duplex forming contacts in and along both sides of the minor groove extending over a total of six base-pairs. As a consequence of binding of DNase I to the DNA-substrate the minor groove opens by about 3 a and the duplex bends towards the major groove by about 20”. Apart from these more global distortions the bound duplex also shows significant deviations in local geometry. A major cause for the observed perturbations in the DNA conformation seemsto be the stacking type interaction of a tyrosine ring (Y76) with a deoxyribose. In contrast, the enzyme structure is nearly unchanged compared to free DNase I (049 A root-mean-square deviations for main-chain atoms) thus providing a rigid framework to which the DNA substrate has to adapt on binding. These results confirm the hypothesis that groove width and stiffness are major factors determining the global sequencedependence of the enzyme’s cutting rates. The nicked octamer present in the crystals did not allow us to draw detailed conclusions about the catalytic mechanism but confirmed the location of the active site near .H134 on top of the central P-sheets. A second cut of the DNA induced by diffusion of Mn2+ into the crystals may suggest the presence of a secondary active site in DNase I. Keywords:

deoxyribonuclease I; DNA flexibility; high resolution X-ray structure: protein-DNA interaction; minor groove contacts

1. Introduction The cleavage rates of bovine pancreatic deoxyribonuclease I (DNase I) vary substantially, by a factor of 100 or more, along a given DNA sequence (Lomonossoff et al., 1981; Drew & Travers, 1984). Although the enzyme is neither strictly base nor sequence-speci$c,DNase I obviously interacts with double-stranded DNA in a sequence-dependent manner, i.e. it recognizes certain sequence-dependent structural variations of the double helix. High resolution X-ray structure analyses of short pieces of double-stranded DNA of defined sequence have t Present address:Istituto di Ricerche di Biologia Molecolare,Via Pontina Km 30.6, 00040Pomezia (Roma). Italy. $ Author to whom all correspondenceshouldbe addressed, 645 0022-2836/91/230645-23

$03.00/O

indeed shown in recent years, that the variation of local helix parameters can be quite substantial and have drastically changed the common conception of the DNA as being a uniform stiff rod (Dickerson, 1983; Shakked & Kennard, 1985; Wang & Rich, 1985; Shakked & Rabinovich, 1986). On the basis of the structure of a dodecamer-duplex it was suggested that, it is the helix twist angle that determines the individual cleavage rates of DNase I (Dickerson & Drew, 1981; Lomonossoff et al., 1981). The cleavage pattern of DNase I is, however, rather complex, indicating that it is not a single varying helix parameter that is recognized. but rather a combination of several parameters. From the three-dimensional structure of the enzyme and model building studies we proposed the minor groove width to be one of the most important single parameters determining the cutting frequencies (Suck & Oefner, 1986; Oefner & Suck, 1986). 0

1991

Academic

Press

Limited

646

A.

Lahm and D. Suck

Similar conclusions were drawn by Drew & Travers from an analysis of the cleavage rates in a 160 bpt Tyr-tRNA promoter region (Drew & Travers, 1984). The groove width is, however, a slowly varying parameter and does not explain the often large difference in cleavage rates of neighbouring phosphodiester bonds. In order to fully understand the DNA-recognition pattern, we have co-crystallized DNase I with a series of self-complementary oligonucleotides of varying composition (G. Frost & D. Suck. unpublished results). Co-crystals of a selfcomplementary octanucleotide d(GCGATCGC), (“Octa I”) with DNase I allowed us to solve and refine the structure of this complex at 2 a resolution (1 A = O-1 nm). A preliminary analysis was published earlier (Suck et al., 1988). The overall structure of the complex confirms the basic features of a model suggested earlier, namely interactions in the minor groove and with both strands of a B-type DNA. The conformation of the DNA does, however, deviate significantly from ideal B-geometry and in particular shows a bend towards the major groove and widening of the minor groove. It appears therefore, that in addition to the groove width as such, the stiffness or bendability of a DNA sequencedetermines its sensitivity towards DNase I cleavage. The stacking interaction of a tyrosine side-chain with a deoxyribose ring induces a local backbone perturbation in one of the strands. To our surprise, the analysis revealed that the bound oligonucleotide is nicked and a quasi-continuous 14mer duplex formed in the crystal despite the presence of 15 mM-EDTA in the crystallization medium. The essential His134$ is located in the immediate neighbourhood of the nick; direct conclusions concerning the enzymatic mechanism are, however, not possible, since the scissile phosphodiester bond is not present in the crystal. Another, completely unexpected result, which may indicate t’he presence of a secondary active site in DNase I. is an additional cut of the oligonucleotide occurring in t,he opposite strand, four base-pairs away from the first cleavage site, when crystals are soaked with

Mn’+

2. Materials (a) Crystallization

and Methods and data

collection

The self-complementary deoxyoctanucleotide 5’ GCGATCGC 3’ was synthesized on a 200 to 300 O.D. scale using a manual solid-phase method based on phosphoramidite chemistry (Sinha et al.. 1984). Purification of the oligonucleotide included ion exchange as well as reversed t Abbreviations used: bp, base-pair(s); h.p.l.c., high pressure liquid chromatography: r.m.s.. root-mean-square. $ The amino acid numbering scheme for DBase I referred to in this paper has changed with respect to the previously published sequence (Oefner & Suck. 1986) and is shown in Fig. 1. The 14 nucleot,ides of the cleaved octamer are given residue numbers 301 to 314 (see also Fig. 5).

phase h.p.1.c. chromatography. DKasr I was purchased from Worthington and purified by chromatography on hydroxylapatite. The composition of the carefully washed and dissolved crystals was established by gel electrophoresis and h.p.1.c. analysis and confirmed by nucleasr Pl digestion and DPU’A sequencing (Rosenthal it al.. 1988). Crystals of t,he complex were grown at 4°C’ in the presence of EDTA (typically 15 mM) using S’& (w/v) polyethyleneglycol6OOO as precipitating agent’. The pH of the crystallization medium (100 mM-imidazole. 100 to 150 mM-Pu’d) was varied, but kept, below pH 60 in order to further slow down t,he endonucleast acativity of DBase T. The oligonucleotide concentration was around 1.0 mM (calculated on the basis of the single strand) corresponding to a 10 t,o 20.fold molar excess over DBase I. The crystals belonging to space group K%!2, (a = 72.92 A. b = 100.1 A, c = 92.6 A) normally grow in plates elongated in the c*-direction and diffract) to about I.9 w resolution. Assuming on 1 romplex. i.e. 1 Dh’asr I molecule and 1 nicked octamer-duplex per asymmetric, unit. the V,-parameter is 2.44 a3 dalton. well within the range normally found for proteins (Matthews, 1968). Two crystals were used to collect intensity dat,a from KJ to 4 .& and from 4 to 3 A resolution on a (‘AD4 Enraf-Nonius diffractometer operated in the w-scan mode with a scan width of 05”. Data reduction was performed using the CCP4 programs for diffractometer data processing and included an empirical absorption rorrection (North el al., 1968) and radiation damage correction (monitored by 6 periodically measured reflections). The maximum loss in intensity encountered was 3O’j/,. Scaling of 3 centric planes collected for each crystal gave an Emerge of 0.10 from x: to 4 A. This dataset consisting of 7196 reflections between 15 and 3 A (5170 = 7296 of which ha.d Z > 3a(Z)) was used for solving the structure by molrc~ular replacement. For structure refinement, data were recorded to 2 X resolution on oscillation photographs (oscillation range per film 1.5”. oscillation axis along c*) on an Elliot (:X21 rotating anode operated at 36 mA. 50 kV while t,hr crystals were cooled to about 0°C. To detect both weak and strong reflect,ions accurately each film cassette contained 3 films placed 1 after anot,her. 60 of these filmpacks were collected on 8 crystals and scanned on a 100 pm raster with an Optronics densitometer. Integrat,ed intensities (obtained by the MOSFLM program suite) were correct)ed for Lorentz and polarization et&c+ and the 3 films in eacah pack scaaled toget,her ((KS,,) = 0.04). Scaling and merging was done using programs ROTAVATA and AGR#OVATA of the C(11’4 program package resulting in a total 20,921 (=89*2q, of possible) unique reflections (Z?m,,s= 0.076) between 50 and 2 Lq resolution (Table 1). (b)

Molecular

ruplawmvnt

The orientation of the DNase 1 molecule in t,he orthorhombic cell was determined with the fast rotation fun @015 indicated by the broken line was calculated within the program SIGMAA of the CCP4 program package and estimates the r.m.s. error in atomic co-ordinates to be 021 a according to the formula given by Read (1986). (b) Real-space R-factor X[(p(obs)-p(calc))/(p(obs) +p(calc))] for all atoms as calculated by the program 0 (Jones et aE., 1991). The observed electron density p(obs) came from a a,-weighted 2F, - F, map with Fooo included. Residues 1 to 260 belong to DNase I, 301 to 314 (shown on an expanded horizontal scale) to the cleaved octamer and residues 400 to 651 to the assigned solvent molecules. related nicked DNA molecules across the a-fold axis along a at b = l/2, c = l/2 with a 2 base-pair overlap was the only consistent explanation of biochemical (h.p.1.c.) and crystallographic observations (missing phosphate peak, density for only 7 base-pairs). Calculation of partial difference maps and refinement leaving out individual base-

pairs confirmed the new model and gave no indication for an alternative model. The final 2F, - F, electron density is of good quality (Fig. 6) and there is density missing only for the flexible loop-region around Cl01 and most of the sugar moiety attached to N18. Especially at the flexible loop region the final F,-F, map still contains significant

Figure 3. Initial a,-weighted 2F,- F, map for the DNA duplex after molecular replacement and subsequent CORELS rigid-body refinement calculated from diffractometer data between 10 and 3 A. There is density present for only 7 basepairs, not, counting the symmetry-related base-pair to the left of the 2-fold axis. Around the e-fold axis parallel b at a/2,c/2 (indicated by the ‘0’ symbol) there is clear density for the bases G301 and C302. The central part of the duplex is rather well defined with weaker density at the last base-pair C308 * G309 on the right. The peak close to this base-pair is the side-chain of H121 of a symmetry-related DNaae I molecule stacking onto G309. Superimposed is the final refined model

650

A. Lahm and D. Suck

Figure 4. Part of ZF,- F, omit map after cycle 29. Part of the electron density (contoured at la) with the model after cycle 29 (top) and the final refined model (bottom) superimposed. The map was calculated from t,he refined co-ordinates after cycle 29 omitting the base-pair at the position shown. It is obvious Ohat shifting the DNA duplex 1 base-pair wit,h respect to the 2-fold axis parallel b at a/2,c/2, i.e. changing an A.T base-pair into a T. A base-pair (see Fig. 5) results in a much better fit.

unexplained features. Attempts to model this part of the molecule ‘by refining with adjacent parts of the chain omitted and calculation of difference maps with lower resolution failed and never gave an interpretable electron density. The presence of EDTA in the crystallization medium obviously further increased the flexibility in this region by removing the calcium from its binding site near D99. The result is even higher disorder compared to the native structure, where this region already showed high mobility and temperature factors >350 A’. Of the DNA-binding residues, the side-chain of R9 occupies 2 alternative conformations around the CB-CY bond. This side-chain could possibly be involved in binding of the missing phosphate at the active site and only be well defined if the latter is present. In the DNA part of the structure the base-pairs next to the 2-fold (G301. C302’ and C302. G301’) where the 2 nicked octamers overlap and the last, base-pair (C308. G309; for nomenclature see Fig. 5) show weak density in the final 2F,-F, map. At both ends of the nicked octamer partial maps calculated after refinement omitting those parts of the DNA always reproduced density at the same position and never indicated an alternative conformation. Likewise partial refinement leaving out bases G301 and C302 and the connecting phosphate as well as neighbouring protein atoms using XPLOR (Bruenger, 1988) did not result in additional or alternative density. Whereas the lower density for basepair C308. G309 is most likely a consequence of the fact

that this end of the DKA duplex is nearly free. the weak density for the base-pair-step 301/302 across the 2-fold axis could be due to a superposition of dynamic and static2 disorder. Since this part of the DNA duplex is singlestranded prior to crystallization, the base and the phosphate backbone can adopt different non-helical conformations, which are then trapped on crvstallization and are superpositioned in the electron den&y map. The initial F,- F, map (Fig. 3) unbiased by any DNA model showed. however. densit,y for this base-pair step at the expected position. Additional confirmation that the DNA conformation of the refined complex is not an artifact of the refinement comes from the structure of another DNase Ioctamer complex (d(GGTATACC),), which is presently being refined. There the asymmetric unit contains a complete octamer and no S-fold axis at the base-pair where the nick occurs in the (d(GCGATCGC),) co-crystals.

3. Resutts and Discussion (a)

Crystal packing

In contrast, to most other crystallographically examined protein-DNA complexes no continuous rods of stacked DNA duplexes are formed in the DNase I-O&a I co-crystals, which may be one reason why these co-crystals show isotropic dif%ct,ion out to 1.9 A. The dominating feature of the

Structure

of a DNase I-DNA

Complex

651

1 .

5’

"GPCPGPAPTPCPGPC"

GPCPGPA~TPCPG~C~' CPGPCPTPAPGPCPG

CPGPCPTPAPGPCPG 5’

3’

3’ t

*’

(a) 3+.

I

GPCPGPAPTPC:GPCPGPAPTPCPGPC~’ L _ __ )- _ _, CPGPCPTPAPGPCPG:CPTPAPGPCPG $.j3. 2’

5’

1

3’15’

GPCPGPAPT

lb)

5’

PC

CPGPCPTPAPGPCPG~ 3’

~GPCPGPAPTPCPGPC” .-+ __~ CPTPAPGPCPG S./S.

5’

t

309

310

311

312

313

314

301'

302'

303'

304

305'

308'

307'

308

5’ (cl 3’ 308

307

308

305

304

303

302

301

314'

313'

312

311'

310'

309

Figure 5. h.p.1.c. analysis of DNase I-Octa I co-crystals. (a) The h.p.1.c. trace of dissolved co-crystals indicates the presence of the original octamer and a hexamer at a 1 : 1 ratio (1st peak at left corresponds to EDTA). The explanation of this observation is given on the right. The dinucleoside-diphosphate pGpC is removed from the 3’ end of 1 of the strands of the octamer in solution. On crystallization, the nicked double-strands dimerize across a 2-fold axis with base-pair formation between the single-stranded CG regions to form a 14mer with phosphates missing at 2 positions. (b) The h.p.1.c. trace of co-crystals that have been soaked with Mn*+ shows essentially only a single hexamer peak. As indicated on the right, this means that the DNA has been cut a 2nd time 4 base-pairs away in the opposite strand (at ~307). This observation may suggest the presence of a 2nd active site in DNase I near 543, more than 15 ip away from the essential H134. (c) Nomenclature of the DNA sequence present in the DNase I-Octa I crystals. Primed residue numbers indicate the symmetry-related nicked octamer.

Figure 6. Active site region in the final a,-weighted 2F,--F, electron density map contoured at lo. In a complete enzyme-substrate complex the missing phosphate group would be located left to the 3’-terminal OH group of C314 with the 0-1P and 0-2P atoms at the position occupied by 2 water molecules (crosses) in the DNase I-Octa I complex. Bot,h water molecules are well defined and are in hydrogen bonding distance to H134, D168 and H252.

652

A. L&n

and D. Suck

(b)

Figure 7. Packing of the DNase I I-O&a I complex in the crystal. (a) Diagonal view along the ac-plane of the orthorhombic DNase I-Octa I crystals showing the alternating layers of protein and DNA building up the crystal in the b-direction. Protein-protein contacts are present only in the ac-plane, protein-DNA contacts stabilize the crystal along b. No continuous DNA helices are formed as often observed in crystals of repressor-operator complexes. The mostly disordered carbohydrate protrudes into continuous solvent channels along the c-direction. (b) Two nicked octamer duplexes form a quasi-continuous 14mer double-helix approximately parallel to the c-axis across the S-fold rotation axis parallel a. DNase I molecules at the upper left and lower right interact in the minor grooves of the B-type DNA. The %I121 residues of 2 other symmetry-related DNase I molecules stack onto 5’ terminal-G residues of the”1‘4mer duplex, thus stabilizing the crystal in the c as well as the b direction.

packing is instead the formation of alternating layers of protein and DNA along b (Fig, 7(a)). As a consequence, intermolecular protein-protein contacts are formed solely in the ac-plane whereas intermolecular DNA-protein contacts between the quasi-continuous 14-mer and four DNase I molecules stabilize the crystal along b and c (Fig. 7(b)).

Although the packing in the co-crystals is totally different from the one observed in the native crystals (Oefner & Suck, 1986) in both cases the sugar moiety attached to N18 and the flexible loop region Cl01 to Cl04 protrude into large solvent channels running throughout the whole crystal along C. Compared to the native crystals there are

Structure

of a DNase I-DNA

Complex

653

Table 4 Intermolecular contacts < 3.5 A ___.__.

X,

Y. 2 position

Symmetry

related

position

Residue

Atom

Residue

Atom

R27 D93 D59 N61 Y148 T188 s190 Y175 5206 Y211 H121 G301 G301 G301

NH2 OD2 OD2 ND2 OH OGl OG OH OG OH Side-chain N-2 N-l O-6

E57 K157 5178 S178 D234 A238 L241 G301’ G303’ G302’ G309 C302’ C302’ C302’

0 NZ OG OG OD2 0 0 o-5 o-1P o-2P Base o-2 N-3 N-4

Intermolecular

contacts

involving

either

polar

or charged

protein

considerably fewer intermolecular contacts in the orthorhombic DNase I-Octa I C222, crystals (Table 4), which may account for the higher overall temperature factor and reduced diffraction power of the co-crystals (1.9 A compared to 1.5 A). Attempts to co-crystallize DNase I with a selfcomplementary 14mer as present in the crystals have so far not succeeded. By seeding with Octa I crystals we were, however, able to obtain small cocrystals of a complex between DNase I and a selfcomplementary 14mer, reinforcing the finding that it is exactly a 14mer that can be accommodated in the C222, crystals. The resulting crystals were, however, not’ suitable for high-resolution X-ray analysis. (b) The conformation of DNase I doesnot change on binding of DNA Apart from a few side-chains in direct contact with the DNA and differences in side-chain conformations arising from different crystalline packing environments there is no significant conformational change taking place in DNase I on binding to the oligonucleotide (Fig. 8). The structure of. the free and DNA-bound enzyme can be superimposed with a r.m.s. difference of 0.70 A for all atomic positions and 049 A for the main-chain atoms only (omitting the flexible loop region D99 to D107). The largest movement of an individual side-chain is found at H121, which is stacking onto the 5’-terminal guanosine and is completely exposed in the native structure. The two regions of the enzyme, showing noticeable but still small movements in C”-positions are the exposed loop containing residues 72 to 75 ( ~65 A) and the turn-helix region V46 to P60. The small shift of the exposed loop is clearly a consequence of DNA binding since it closely interacts with the minor groove (see below). However, the tyrosine ring of residue 76, which is in direct

Distance l-4 33 35 2.5 32 29 2.7 33 2.8 2.9 35 3.5 2.8 F9 3.5 and DNA

residues

(except

Symmetry operator -x,

Y, l/Z-Z

1/2-x,

1/2-Y,

1/2+2

1/2+x,

112-Y.

l/Z-Z

x,

l-Y,

1/2-x. x,

l-Y,

for the stacking

I-Z

1/2+

Y, I/2+2

1-z

of H121

onto G309).

stacking contact with a deoxyribose ring, shows a surprisingly small shift, remaining well fixed through van der Waals’ contacts and a hydrogen bond to E78. The movement of helix III by about 67 A towards the N-terminal end of the helix is probably due to crystal packing effects. In the native structure the region between V45 and D55 formed tight intermolecular contacts not present in the co-crystals. Protein-DNA contacts near S43 could also contribute to the movement but this is lesslikely since residues S43 to V45 show no significant displacement. Otherwise, and in particular in the hydrophobic core of DNase I, the two structures are remarkably similar. The positions of all the buried water molecules are preserved. At the protein-DNA interface and in the active site region about half of the well-defined solvent molecules are present in bobh structures. Due to the presence of 15 mM-EDTA in the crystallization medium no calcium cations are present at the two structural Ca2+ binding sites and only at the first site (D201 to T207) was the calcium replaced by a water molecule without any structural change. In contrast, the second site (D99 to D107) near the flexible loop region containing the small disulphide bond between Cl01 and Cl04 is further destabilized and no octahedrally co-ordinated water molecule is present. As a consequence, the loop-region, which was already partially disordered in the native structure, shows even higher disorder in the complex structure and was excluded from refinement. The only other disordered part of the enzyme structure is the carbohydrate side-chain attached to N18 where only the first N-acetylglucosamine was unambiguously defined in the electron density map. This fact confirms biochemical evidence t,hat neither the carbohydrate moiety nor the region of the short S-S bridge is important for the binding to and cleavage of DNA (Price et aZ., 1969). The isotropic temperature factors of DNase I in

----------YE

_--_----

-50 A*. The ends of the duplex clearly show higher mobility compared to the central region, where DNase I binds in the minor groove (labelled m). Around the 2-fold axis in the middle of the 14mer near the missing phosphate group higher B-factors indicate higher dynamic and/or static disorder. The mean temperature factor for the phosphate groups is 36.6 A’. All other DNA atoms have a mean of 28.2 A2 compared to 302 AZ including all the DNA atoms.

coccal nuclease (Serpersu et al., 1987) the rate-acceleration by Ca2+ was estimated to be 104’6 which is in the same order of magnitude observed by us. A second completely unexpected observation was made, when we soaked DNase I-Octa I co-crystals

with divalent cations. The h.p.1.c. trace in Figure 5 showed that with Mn2+ the octamer disappeared almost completely within three days and only a single peak corresponding to the hexamer remained. The only way this h.p.1.c. pattern can be obtained is

Figure 17. DNA hydration. Solvent molecules within a distance of 4 A from any DNA atom are indicated by shaded spheres. The view shown is the same as in Fig. 16, i.e. the interaction with DNase I occurs in the minor groove at the lower left and upper right end of the quasi-continuous 14mer.

664

A. Lahm

and D. Suck

Figure 18. Protein environment of the 2 DNA cleavage sites. (a) Environment of the nick in the 14mer duplex. The essential H134 is located in the immediate neighbourhood of the expected position of the missing phosphate. The proposed acid-base catalysis of the nucleophilic attack by a water molecule through the E78-H134 pair appears likely (Suck & Oefner, 1986); it is, however, possible that H252 plays a catalytic role as well. The catalyt,ica cation could be coordinated by the scissile phosphate group and E39. which is hydrogen bonded to the O-3’ at t,he nick. confirming the results of Ca-pTp soaks of native DNase 1 crystals (Suck et al., 1984). Double circles represent water molecules. The 2 waters hydrogen bonded to H252 and H134 occupy positions where one would expect the 0-1P and 0-2P atoms of the scissile phosphodiester bond in an uncleaved substrate. Heavy lines indicate the DNA backbone with t,he bases omitted from the drawing. (b) Environment of ~307. the phosphodiester bond that is cleaved on soaking of the co-crystals with IIn’+. 543 is in direct contact with this phosphate, side-chains H44, E13. D42 and R41 are close by. This site is more than 15 a away from the essential H134.

by a second cut four phosphodiester bonds away in the opposite strand, which leaves us with just one hexamer species. No additional cleavage was obtained with Ca2+ or Mg2+, whereas a very slow second cleavage did occur also with Co2 + . How does this result fit with the reported properties and the three-dimensional structure of DNase I? The position of the cleaved octamer in the DNase I-Octa I structure as well as biochemical data from a recombinant mutant DNase I (H134Q) showing very low activity (Worrall & Connally, 1990) leaves no doubt about’ the location of the active site. It seems therefore to be inevitable to assume a secondary active site in DNase I, which apparently is activated only with Mn2+ (or other transition metal ions). Since in contrast to an active enzyme in solution the crystalline environment keeps the DNA duplex bound to DNase I, this secondary active site, which presumably has a much lower solution activity than the primary site, can actually work in a manner independent from the primary site. Possible candidates of protein residues that may contribute in a catalytic reaction at this

secondary site are E13. R41. 1142. S43 and H44 (Fig. 18). In this context it is int,eresting t,o go through the reports in the literature concerning the effect of Mn2+ on the DNase 1 reaction. Melgar & Goldthwait describe a shift of the usual double hit mechanism (one strand cut at a time) to a single hit mechanism (both strands cut simulta,neously) by changing the divalent cation from Mg” t,o ?uln’-’ (Melgar & Goldthwait, 1968a,h). Several authors describe a considerable acceleration of the cleavage reaction, but no change of the cleavage pattern. of Mn2+ (Drew & Travers, 1984; Brukner et nZ., 1990) and finally Campbell & Jackson (1980) observed cutting of both strands of circularly closed simian virus 40 DNA at or near the same site only in the presence of transition state elements. namely Mn2 + or Co’+. Up to now we were not able to collect an acceptable data set from a Mn2+ or (‘o” soaked crystal. The crystals mostly crack and show no or much reduced diffraction and the data sets collected so far neither allow identification of a metal binding site near 543 nor allowed detailed analysis of the DNA conformation after soaking. Co-crystallizing in

Structure

of a DNase

Figure 19. Packing of Y76 and R41 in the minor groove. Slice through the van der Waals’ surface illustrating the tight packing of Y76 and R41 acrossthe minor groove. A double helix exhibiting a more narrow or shallowminor groove would bind lessefficiently and cause a decrease in cleavage rate. Changing base-pair T305. A312 to a G. C or C. G could lead to unfavourable contacts betweenthe 2-aminogroup of the guanosineand the R41 side-chainand thereby locally reducethe cutting rate.

active DNase I mutants in the presence of Ca2+ or other divalent cations appears to be a more promising approach to investigate this potential secondary site.

4. Conclusions As was suspected from biochemical studies showing low cleavage rates in regions of the DNA with runs of A or T residues (Drew & Travers, 1984) and was proposed by us from model building studies (Suck & Oefner, 1956) the groove width appears to be one of the most important global helix parameters determining the binding and therefore the cutting rates. This notion is fully supported by the present X-ray study. From X-ray structure analysis of oligonucleotides, one expects a narrow minor groove in regions with runs of A or T residues (Fratini et al., 1982). The reduced binding of these regions t’o the enzyme could simply arise from steric reasons, caused by the tight packing of the exposed loop around Y76 and of R41 in the minor groove (Fig. 19), but might also reflect the inability to form the observed contacts between the two DNA phosphate backbones and the protein. At the same time, however, the structure of this DNase I-Octa I complex strongly suggests that it is not only the actual size of the groove that is important, but also the bendability or stiffness of the DNA (Hogan &, Austin, 1987). The duplex bound to DNase I is bent towards the major groove and it is this bending that widens the minor groove to the observed value of about 15 A. The bending induced by DNase I is

I-DNA

Complex

665

opposite to that expected for runs of four to five A residues according to the Trifonov-Dieckmann wedge model (Ulanovsky et aZ., 1986; Dieckmann et al., 1987) and this gives an additional explanation for the extremely low sensitivity of these regions: a narrow minor groove plus bending in the wrong direction. One could imagine that sequences showing pre-formed bends in the right direction, i.e. towards the major groove, would be particularly easily attacked by DNase I (Drew & Travers, 1985; Hochschild & Ptashne, 1986) and could possibly explain the existence of the hypersensitive sites on naked DNA. In protein-DNA complexes, these bends could be pre-formed by the interaction with a protein giving rise to a hypersensitive site near the binding site. The global, slowly varying, parameters like groove width or bendability do not explain, however, the often dramaticallv different cutting rates of neighbouring phosphodiester bonds (Drew & Travers, 1984). It has been suggestedon the basis of a DNA dodecamer that the cutting rates are proportional to the helix twist angle (Lomonossoff et al., 1981; Dickerson & Drew, 1981). The present structure does not support this hypothesis. The step where the nick occurs has in fact a very low twist angle (24*4”), although it may be affected by the missing phosphate. For GC regions, which also show somewhat lower than average cleavage rates (Drew & Travers, 1984) the groove width argument (wider, shallow minor groove) does not seem to be applicable since a recent high-resolution study of a B-type DNA with a central GGCC revealed a groove width close to canonical mixed-sequence DNA (Heinemann & Alings, 1989). For these sequences parameters affecting local DNA geometry and especially the DNA backbone conformation could play a more important role. The observed variations in cleavage rates could arise from various local distortions in DNA structure. A G. C or C. G basepair instead of an A’ T three base-pairs away from the cleavage site could lead to unfavourable contacts between the 2-NH, groups of guanosine and R41. In a recently determined structure of a homeodomain-DNA complex similar minor groove contacts of arginine residues have been observed and it is believed that this at least in part explains the preference of homeodomain proteins for A+Trich regions (Kissinger et al., 1990). The unusual base-pair stacking pattern induced by the interaction of Y76 with the phosphate backbone could also influence the cutting rates. Finally the actual orientation of the scissile phosphodiester bond might have an impact on DNase I activity towards a specific base-pair step as was pointed out already by Drew & Travers (1984). A B,, type conformation of the DNA backbone around the 0-3’-P bond would enhance the accessibility for an in-line mechanism and show higher cleavage rates than a B, type conformation. In GCGCAATTCGCG dodecamer the by far most susceptible bond has, however, a normal phosphate conformation in the crystal (Fratini et al., 1982), which again shows that in general the

666

A.

Lahm and D. Suck

observed cutting rates are influenced by several global and local parameters and cannot be explained by one simple model. Other complexes containing different, uncleaved DNA sequences as well as the corresponding free oligonucleotides have to be crystallized and solved at high resolution before we can reliably compare them and correlate local helix parameters of a specific sequence with the observed DNase T sensitivity.

We gratefully acknowledge the excellent technical assistance by Graham Frost and the expert advice of Drs Heinz Bosshard, Johan Postma and Paul Tucker (all EMBL) with the graphics display system and the film processing. respectively. The analysis of the DNA duplex was done with programs kindly provided by Drs Lavery. Sklenar and Dickerson. For drawings of the schema&c: view of the complex (Fig. 9) we used the program RIBBON written by Dr .J. Priestle (Priestle, 1988). The atomic co-ordinates have been deposited with the Brookhaven Protein Data Bank.

References Agarwal, R. C. (1978). A new least-squares refinement technique based on the fast Fourier algorithm. Acta Ckystallogr. sect. A, 34, 791-809. Aggarwal. A. K.; Rogers, D. W.. Drottar. M.. Ptashne, M. & Harrison, S. C. (1988). Recognition of a DNA operator by the repressor of phage 434: a view at high resolution. Nature (IAndon), 242, 99-107. Bruenger, A. (1988). Crystallographic refinement by simulated annealing. Application to 2.8 a resolution structure of aspartate aminotransferasr. J. Mol. Hiol. 203, 803-816. Brukner, T.. ,Jurukovski. Y. & Savic. A. (1990). Sequence-dependent structural variations of DNA revealed by DNase 1. n;~cl. Acids Res. 18, X91 -894. Campbell, V. W. & Jackson. D. A. (1980). The effect of divalent’ cations in the mode of action of DNase 1. .J. Biol.

Chem.

Chandrasekaran.

255,

R.

3726-3735.

(1989). In (Saenger, pp. 31-170. Springer-Verlag, Berlin. Heidelberg and New York. C’rowther, R. A. (1972). The fast rotation function. In The Molecular Replacement Method (Rossmann, M. (: ed.), pp. 174-I 78, Gordon BE Breach. New York. Dickerson, R. E. (1983). The DNA helix and how it, is read. Sci. ilmer. 249. No. 6. 86- 102. Dickerson, R. E. iz: Drew, H. R’. (1981). Structure of a B-DNA dodecamer. II: Influence of base sequences on helix structure. J. Mol. Biol. 149, 761.-786. Dickerson, R.. E., Bansal. M., Calladine. (1. El.. Dieckmann. S.. Hunter. W. N.. Kennard. 0.. von Kitzing, E., Lavery, R., Nelson, H. r. M., Olson, W. K., Saenger. W.. Shakked. Z., Sklenar, H.. Houmpasis, D. M., Tung. C.-S.. Wang. A. H.-.J. & Zhurkin, V. B. (1989). Definitions and nomenclature of nucleic acid structure parameters. ElMBO .J. 8, l-4. Dieckmann. S.. von Kitzing, E., McLaughlin. I,., Ott. .J. & Eckstein, F. (1987). The influence of exoeylic substituents of purine bases on DNA curvature. Proc. A’&. LanrloltBornstein,, W., ed.), vol. 16,

Acad.

Nci.,

U.S.A.

&

New

Arnott,

iSeries.

S.

Group

VII

84. 8257-8261.

Drew, H. R. $ Travers. A. A. (1984). DNA structural variations in the E’. coli tyrT promotor. Cell, 37. 491-502.

Drew. H. R. B Travers, A. A. (1985). DNA bending and its relation to nucleosome positioning. Lv:ucl. Acids Res. 13. 4445-4467. Fratini, A. V.. Kopka. M. L.. Drew. H. R. & Dickerson. R. E. (1982). Reversible bending and helix geometry in a R-DNA dodecamer: CGCGAATT%GCG. J. Riol. Chem. 257, 14686--14707. Heinemann. I!. (1991). A note on cryst,al packing and global helix structure in short il-DNA duplexes .I. Biomol. Struct. Uynam. 8, 801-811. Heinemann, U. & Alings. (‘. (1989). (Irystallogral)hic. study of one turn of G/C-rich B-DNA. J. iMo1. Biol. 210, 369-381. Hendrickson. W. A. & Konnert, ,I. H. (1980). Stereochemical restraint crystallographic leastsquares refinement of macromolecular structures. In Biological Evolution

Structure.

Fwnction,

(Srinivasan.

Conform,ation

and

R., ed.), pp. 43-57. Pergamon.

Oxford. Hochschild. A. & Pt,ashne. M. (I 986). C’ooprrativtl binding of lambda repressors t,o sites separated by integra,l t’urns of the DNA helix. (‘ell, 44, 681~-687. Hogan. At. E. h Austin, R. H. (1987). Importance ot’l)NA stiffness in prot,ein-DSA binding specificit,\. Xntwrp (London), 329. 263-266. Jones, T. A. (1978). A graphics model building and rrtirlsystem for macromolecules. .J. Appl. ment (!rystallogr. 11. 268-272. Jones. T. A.. %ou. ,I.-Y.. Cowan. S. W. & Kjeldgard. M. (1!)91). Improved methods for t)he building of protrin models in elec%ron density maps atld the loc,ation of errors in t’hese models. At-to Prqstallogr. wt. d. 47. 1 IO 119. .lordan, It. S. &, Pabo. (1. (1988). Structure of the larntia complex at 2.5 x resolution: details of the rrpresxoroperator complex. Science, 242. 893-899. Kissinger, (‘. R.. Lin. B.. Mart#in-Blanco, E.. Kornberg, T. 13. & Pabo. (‘. I). (1990). (‘rystal structure of an rngrailed homeodornain-I)NA c~omples at 2.8 .A resolution. A framework for understanding homrodomain-DXA interactions. (‘~11, 63, 579.-MM Kopka. M. L.. Fratini. A. V.. Drew. H. R. h I)ic:krrson. K. E. (1983). Ordered water strm~turr around a H-DNA dodecamer. A quantitative study. .J. Mm/. Hid. 163. 129-146. Laskowski. M.. Sr (1971). I)eoxyribonucleast~ I, Jn 7’lir Enzymes (Bayer. I’. I).. rd.). vol. -I;. pp. 28!+~ 31 I. Ac*ademic Press, Sew York. Lavery. R. & Sklenar, H. (1988). The detinitmn of generalized helicoidal parameters and of axis curvaturr for irregular nucleic acids. J. Bitmrnol. Struct. Dynnwl. 6, 63-91. Lavery. R. & Sklenar, H. (1989). Defining the structure ot irregular nucleic: acids: conventions and principles. .I. Hiow~ol.

Struck

Dynam.

6, 655-667.

Lomonossoff. (:. N.. Butler. P. ,J. G. & Klug. .I. (l!#Xl). Sequence-dependent variation in the conformation of DNA. J. Mol. Biol. 149, 745-760. Matthews. B. W. (1968). Solvent content of protein crystals. .J. Mol. Biol. 33, 491-497. Mehdi. S. & Gerlt, *J. A. (1984). Synthesis and c.onforntational analysis of thymidine 4-nitrophenyl [“O. lnO] phosphates and the stereochemical course of a rt‘act,ion catallyzed by bovine pancreat,ic drox,vrihonuc~lease 1. B~och.emistry, 23. 4844-4852. Melgar. E. 8r Goldthwait, D. A. (1968a). I)eoxyribonuc.leic arid nucleases I. The effect of metals in the mechanism of action of deoxvribonuclease 1. .J. Hiol. (‘hem. 243. 4401~4408.

Structure

qf a

DNase

Melgar, E. & Goldthwait, D. A. (19686). Deoxyribonucleic acid nucleases II. The use of a new method to observe the kinetics of DNA degradation by DNase I, DNase II and E. coli endonuclease I. J. Biol, Chem. 243, 4409-4416. Moore, S. (1981). Pancreatic DNase. In The Enzymes (Boyer, P. D., ed.). vol. 14, pp. 281-296, Academic Press, New York. Nixon. P. E. & North, A. C. T. (1976). Crystallographic relationship between human and hen-egg lysozymes. I. Methods for establishment of molecular orientation and positional parameters. Acta Crystallogr. sect. A, 32. 3206325. North, A. C. T.. Philips, D. C. & Matthews, F. S. (1968). A semi-empirical method of absorption correction. Acta Crystallogr.

sect. A, 24, 320-325.

Oefner, (1. & Suck, D. (1986). Crystallographic refinement and structure of DNase I at 2 A resolution. J. Mol. Biol. 192, 605-632. Otwinowski, Z., Schevitz, R. W., Zhang. R.-G., Lawson. (1. L., Joachimiak, A., Marmorstein, R. Q., Luisi, B. F. & Sigler, P. B. (1988). Crystal structure of the trp repressor/operator complex at atomic resolution. -Vatwe (London), 335, 321-329. Price, P. A.. Stein, W. H. & Moore, S. (1969). Effect of divalent cations on the reduction and reformation of the disulphide bonds of deoxyribonuclease. J. Biol. (‘hem.

244,

929-932.

Priestle, J. P. (1988). Ribbon: a stereo cartoon drawing program for proteins. J. Appl. Crystallogr. 21, 572576. Prive. G. Q., Heinemann. CT., Chandrasekaran, S., Kan, L. S., Kopka, M. L. & Dickerson, R. E. (1987). Helix geometry. hydration and G.A mismatch in a B-DNA decamer. fkience, 238, 498-504. Read, R. (1986). Improved Fourier coefficient for maps using phases from partial structures with errors. Acta Crystallogr. sect. A, 42, 140-149. Rosenberg, tJ. (1991). Structure and function of restriction endonucleases. C’urr. Opin. Struct. Biol. 1, 104-113. Rosenthal. A.. Schwertner, S., Hahn, V. & Hunger, H.-D. (1988). Solid-state methods for sequencing nucleic acids. I: Simultaneous sequencing of different oligodeoxyribonurleotides using a new, mechanically stable anion-exchange paper. Nucl. Acids Res. 13, 1173-1184. Saenger, W.. Hunter. W. N. & Kennard, 0. (1986). DNA conformation is determined by economics in the hydration of phosphate groups. Nature (London), 324. 385-388. Serpersu, E. H.. Shortle. D. & Mildvan, A. S. (1987).

I-DNA

Complex

667

Kinetic and magnetic resonance studies on active-site mutants of staphylococcal nuclease. Factors contributing to catalysis. Biochemistry, 26. 1289-1300. Shakked, Z. & Kennard, 0. (1985). The A-form of DNA. In Biological Macromolecules and Assemblies (McPherson, A. & Jurnak, F., eds). vol. 2. pp. l-36, Wiley and Sons, New York. Shakked, Z. & Rabinovich, D. (1986). The effect of the base sequence on the fine structure of the DNA double helix. Progr. Biophys. Mol. Biol. 47. 159-195. Sinha, N. D.: Biernat, J., McManus. J. & Koester, H. (1984). Polymer support oligonucleotide synthesis XVII: use of b-cyanoethyl-N,N-dialkylamino-/Nmorphino phosphor amidite of oligodeoxyribonucleotides for the synthesis of DNA fragments simplifies deprotection and isolation of final products. Nucl. Acids Res. 12, 4539-4557. Steitz, T. A. (1990). Structural studies of protein-nucleic interaction: the sources of sequence-specific binding. Quart. Rev. Biophys. 23, 205-280. Suck, D. & Oefner, C. (1986). Structure of DNase I at 2.0 A resolution suggests a mechanism for binding and cutting DNA. Nature (London), 312, 620-625. Suck, D.. Lahm, A. & Oefner, C. (1988). Structure refined to 2 A of a nicked DNA octanucleotide complex with DNase I. Nature (London), 332, 465-468. Suck. D., Oefner. C. & Kabsch, W. (1984). Three dimensional structure of bovine pancreatic DNase I. EMBO

J.

3. 2423-2430.

Sussmann. J. L., Holbrook. S. R.. Church. G. M. & S. H. (1977). A structure-factor LS-refinement cedure for macromolecular structures constrained and restrained parameters. Crystallogr.

Kim, prousing Acta

sect. A, 33, 800-804.

Tickle, I. J. (1985). Review of space group general translation functions that make use of known structural information and can be expanded as Fourier series. In Molecular Replacement, pp. 22-26. Proceedings of the Daresbury Study Weekend, 15-16 February 1985. SERC Daresbury Laboratory. Ulanovsky. L., Bodner, M.: Trifinov, E. ?I;. & Choder, M. (1986). Curved DNA: design, synthesis and circularization. Proc. Nat. Acad. Sci., U.S.A. 83, 862-866. Wang, A. H.-J. & Rich, A. (1985). The structure of the Z-form of DNA. In Biological Macromolecules and Assemblies (McPherson. A. & Jurnak, F.. eds), vol. 2. pp. 1288170, Wiley and Sons, New York. Worrall, A. & Connally, B. (1990). The chemical synthesis of a gene coding for bovine pancreatic DNase I and its cloning and expression in E. coli. .I. Biol. Chem. 265. 21889921895.

Edited by R. Huber