Eur. J. Biochem. 97. 593- 602 (1979)

Higher-Order Structures of Chromatin in Solution Pedro SUAU, E. Morton BRADBURY, and John P. BALDWIN Biophysics Laboratories, Portsmouth Polytechnic (Received December 22, 1978)

Neutron scatter studies have been made on gently prepared chicken erythrocyte chromatin over a range of ionic strength. At low ionic strength the mass per unit length of the ‘10-nm’ nucleofilament corresponds to one nucleosome per 8 - 12 nm and a DNA packing ratio of between 6 and 9. From the contrast dependence of the cross-section radius of gyration of the nucleofilament the following parameters have been obtained; RgDNA,the cross-section radius of gyration (R,) when DNA dominates the scatter; RgP, the cross-section R, when protein dominates the scatter; R,, the cross-section R, at infinite contrast and a, the constant which describes the dependence of the cross-section R, on contrast variation. From our understanding of the structure of the core particle, various arrangement of core particles in the nucleofilament have been tested. In models consistent with the above parameters the core particles are arranged edge-to-edge or with the faces of the core particles inclined to within 20* to the axis of the nucleofilament. With increase of ionic strength the transition to the second-order chromatin structure has been followed. This gave the interesting result that above 20 mM NaCl or 0.4 mM MgCI2 the cross-section R, increases abruptly to about 9 nm with a packing ratio of 0.2 nucleosome/nm and with further increase of ionic strength the R, increases to 9.5 nm while the packing ratio increases threefold to 0.6 nucleosome/nm. This suggests a family of supercoils of nucleosomes which contract with increasing ionic strength. In its most contracted form the diameter of the hydrated supercoil has been found from the radial distribution function to be 34 nm. Models for the arrangements of core particles in the 34-nm supercoil are discussed. There is now growing evidence that the DNA in metaphase chromosomes and in the interphase nucleus is organised into discrete loops which contain some 30000-90000 basepairsofDNA[1,2]or inchromatin domains of average size 34000 bases of DNA [3,4]. The DNA of these domains or loops when complexed with chromosomal proteins in inactive regions of chromatin and in metaphase chromosomes is highly compacted and several orders of chromatin structure must exist above the linear array of nucleosomes. In the electron microscope chromatin has been visualised as fibrils of different diameters. At low ionic strength the fibrils have a diameter of about 10 nm [5-81. On the addition of monovalent or divalent cations a transition to a higher-order structure is observed and the diameter of the fibril increases to 25 - 30 nm IS, 9 - 131. From electron microscopy studies a solenoid model has been proposed [ 121which consists of the 10-nm fibril or nucleofilament coiled with a pitch of 11 nm and a diameter of 30 nm. It has also been shown that the 10- 1I-nm semimeridional arc in the neutron diffraction of H 1-depleted chromatin [14] and of total chromatin fibres [15] contain maxima which are located off the meridian and these observa-

tions have been explained by a coil model with similar parameters. This type of model is one of several different types of models proposed to explain the characteristic series of diffraction maxima observed in the X-ray diffraction pattern of chromatin fibres [16]. We have applied neutron scattering techniques to large multimers of nucleosomes under a variety of pH and ionic strength conditions to obtain information on the transition from the 10-nm nucleofilament to the ‘30-nm’ coil of nucleosomes and on the arrangement of nucleosomes in both of these chromatin structures. The results are discussed in the context of the recent progress in the knowledge of the structure of the chromatin core particle [17 - 201. EXPERIMENTAL PROCEDURE AND RESULTS Multimer Preparation

Chicken erythrocyte nuclei were isolated by the procedure of Murray et al. [21]. Chicken blood was collected in 10% citrate buffer and transported on ice.

594

Higher-Order Structures of Chromatin in Solution

x

5.35

y

HI

H5

A

B

Fig. 1 . Polyacrylamide gel electrophoresls patterns of ( A ) protiJiii, from the isolated chromatin fraction, and ( B ) the proteins from chicken erythrocyte nuclei. NHP= non-histone protein

The preparation began within an hour of killing the chickens. Purified nuclei in the concentration range (2-3) x lo8 nuclei/ml were digested with 400 U/ml of micrococcal nuclease (Microbiology Research Establishment, Salisbury, England) in 0.25 M sucrose, 150 mM NaC1, 10 mM Tris-HCl, 0.3 mM CaC1, , pH 7.6 at 37 "C. Digestion was stopped by adding EDTA to 10 mM and cooling on ice. Digestion times ranged from 5 s to 20 s. The enzyme used in the digestion was eliminated by several washes of the digested nuclei in 0.25 M sucrose, 150 mM NaCl, 10 mM Tris-HC1, 2 mM EDTA, pH 7.6. The final nuclear pellet was homogenized with a wide-bore pasteur pipette in 1 mM Tris-EDTA buffer, p H 7.0 after addition of NaCl to give a final concentration of 40- 50 mM NaCl [13]. The insoluble material was separated by centrifugation at 25 000 x g for 20 min. The supernatant contained between 10% and 50% of the nuclear DNA, depending on the digestion time. Chemical Characterisation

Proteins were analysed by dodecylsulphate/polyacrylamide gel electrophoresis according to Laemmli [22]. Coomasie blue was used as stain. Purified nuclei showed the presence of non-histone proteins (Fig. 1B), the proportion of which was considerably reduced in the gently isolated chromatin fraction (Fig. 1A). The histone/DNA ratio of the multimers was found to be 1.07+0.05 by the method of Zamenof and Chargaff [23], using total chicken erythrocyte histones as standard. DNA concentrations were measured by

O

0

L

5

10 -Migration

15

I

1

'

I

2

'

0

(crn)

'

'

3 4 5 6 7 Length of DNA (pm)

I

8

'

9

Fig. 2. ( A ) Densitonieter scan of 1.6% agarose gel of ( I ) the DNA lengths from chromatin pieces used for studies of the 30-nm coil compared to ( I I ) 1 DNA and EcoRI restriction fragments; ( B ) histogram of DNA lengths measured directly from electron micrographs of DNA isolated from chromatin pieces used in studies of the '10-nm' nucleofilament. In (A) the numbers on the peaks correspond to the number Of bases X

ultraviolet absorption spectrophotometry using E260 =20 cm2 mg-'. The DNA length distribution was established by electron microscopy and by 1.6% agarose gel electrophoresis. The lengths of DNA in the multimers of nucleosomes varied depending on the time of digestion. Multimers used in the neutron scatter studies of the '30-nm' fibril contained a distribution of DNA molecular weights with the highest being 2.5 x lo', i.e. about 180 nucleosomes (Fig. 2A). There was material with much higher-molecular-weight DNA but very little material containing DNA with molecular weight less than 5 x lo6, i.e. about 40 nucleosomes. To reduce entanglement of 10-nm fibrils, smaller multimers con-

P. Suau, E. M. Bradbury, and J. P. Baldwin

taining about 50 nucleosomes were used for the neutron studies at low ionic strength. The histogram (Fig. 2 B) of DNA lengths for this material showed that the most frequent length was 3.0- 3.5 pm, corresponding to 50 nucleosomes. The spread of DNA lengths extended from a minimum length of 1.5-2.0 pm to an upper length of around 8.0-9.0 pm, i.e. 25-150 nucleosomes. Neutron Measurements

Neutron experiments were performed on the smallangle scattering apparatus D 11 [24,25] at the Institute Laue-Langevin (Grenoble, France). Neutrons from the high-flux reactor were moderated by a liquid deuterium cold source and monochromatized by a helical-slot velocity selector to give a mean wavelength of 0.8 nm with full half-width of 8%. The scattered neutrons were detected in a BF, ionization chamber with gas amplification by proportional wires. It consisted of 4096 individual counters. The samples in the quartz cells of I-nm or 2-mm thickness were measured at three detection positions: 0.84 M, 2.5 M and 10 M. The data were scaled using overlapping regions and merged to obtain the complete scattering curve. The data were not corrected for distortions due to wavelength distribution because the slow variation of the distribution of scattered neutrons remains unaffected by the width of the wavelength distribution. Variation of detector response were corrected by division of the observed spectrum by the water spectrum after a small correction for the empty cell scatter. Neutron Scattering from Rod-Like Macromolecules in Solution

The raw data is a measurement of the intensity of neutrons Z(k) scattered from the solution as a function of the scattering angle 20 and plotted against the parameter k =471 sin 012 where 2 is the wavelength. In the case of rod-like particles in solution the Guinier analysis [26] is modified and k l ( k ) is plotted against k 2 . From the slope of the linear part of this plot the cross-section radius of gyration of the rods is obtained while its intercept on the In [kZ(k)] axis, i. e. [kZ(k)]k+O, gives the mass per unit length of rod. The theory of X-ray scattering from rigid rods in solution [27,28] can be modified for neutron scattering to give the following expression for the mass per unit length of rod in terms of [kZ(k)],,, and the incoherent scattering of neutrons from H 2 0 , IHzo :

(see also Stuhrman et al. [29] and Suau et al. [20]) where M , is the molar mass in g/mol of the oligomers of nucleosomes, I is the length of the filament of

595

nucleosomes, k = 471 sin 011 the wave vector where 1=wavelength, N A=Avagadro's constant of 6.0231 x mol-', D=cell thickness (1 mm), ZH2,, is the observed water scattering from a I-mm cell, Twis the transmittance of the water sample, I ( k ) is the scatter from the chromatin solutions, [kZ(k)],,, is the intercept obtained by extrapolating the linear portion of the plot In [kI(k)] versus k 2 to k 2 = 0 , c is the concentration of the chromatin solutions (g ml-'), eMis the scattering-length density of the chromatin oligomers measured at zero contrast, esolis the scattering length density of the solvent, v2 is the partial specific volume of the chromatin oligomers, r;Z, is a Factor depending on the mean level of exchangeable proton sites in the chromatin oligomers and T,= transmittance of the chromatin solutions. If eMis the mean neutron scattering-length density of the rods then eM- e,,, = ij represents the contrast of the rods relative to the solvent scattering-length density which can be varied over a wide range by using different mixtures of 2 H 2 0 and H,O for the solvent; hence a plot of I/[kl(k)], versus eSolfrom Equ. (1) above for Moil can be used to estimate the mass per unit length. The cross-section radius of gyration of rods in solution will change with contrast ij; for example, when the solvent scattering-length density Qso, is equal to the mean scattering-length density of DNA, then the measured cross-section radius of gyration of the filament will correspond largely to that of the protein component of chromatin; when Qsol is equal to the mean scattering-length density of the protein then its radius of gyration will be largely that of the DNA component. The dependence of cross-section radius of gyration on the contrast ij is given by essentially the same equation as that used for three-dimensional radii of gyration of particles in solution [30,311 :

A plot of Rf versus lie allows R,, a and p to be determined. R , is the cross-section radius of gyration of the particle at infinite contrast, a is the first moment of the internal structure of scattering-length density projected on the axis of the filament or coil. A nonzero B implies an asymmetric distribution about the rod axis of the scattering-length density of the protein compared to the DNA. The experimental scatter curves of kZ(k) versus k can be integrated by performing an inverse transform of the Hankel integral. The result of the integration is a function D(R)/R [=2nP(R)]: D(R), the chord distribution function has the property that it reaches zero at a value of R equal to the longest chord that can be drawn in the cross-section of the rod. P(R) is the circular average of the self-convolution function (Patterson function) of the cross-section of the rod.

596

Higher-Order Structures of Chromatin in Solution

Fig. 4. Dependence of the square root of the zero-angle values of k l ( k ) taken from Fig. 3 on the solvent scattering length density, e,,, . The zero-angle intensities were calibrated with the scatter from chicken erythrocyte core particle obtained during the same neutron run

L

OO

0.2

0.4

0.6

k2(nm-2)

Fig. 3. Modified Guinier plots for rodsfrom neutron scatter curves of chromatin at low ionic strength in the range of ' H 2 0JH,O mixtures indicated. The solutions were buffered with 10 mM Tris-HC1, 1 mM EDTA at pH 7.8

It is possible to propose models for the crosssection of the rod and to predict the scatter curve k I ( k ) , or the D(R) function derived from it, for comparison with the observed data. This procedure is equivalent to that applied to scatter curves of threedimensional particles in solution [32].

Multimers of Nucleosoirzes at Low Ionic Strength Reproducible cross-section information has been obtained on the structure of the nucleofilament at low ionic strength in 10 mM Tris-HC1, 1 mM EDTA and pH 7.8 for six 'H2O/HZ0mixtures. N o concentration dependence of the scattering curves was found over the range 2 - 10 mg DNA/ml and no effect was observed on reducing the ionic strength to 0.5 mM TrisHCl, 0.5 mM EDTA and pH 7.8. The Guinier plots for rods [26], In [ k l ( k ) ]versus k 2 , obtained from the scatter curves of the nucleofilaments in the different 2H,0/H,0 ratios are given in Fig. 3. From these modified Guinier plots two parameters are obtained: firstly, the zero-angle values of the product k l ( k ) by extrapolation of the linear portion of the plots to zero angles and, secondly, the crosssection radius of gyration from the slopes of the linear

portions of the plots. The peak at k2= 0.5 in the 30% 'H,O curve comes from the internal structure of chromatin which shows up at low contrast. The down turn of the Guinier plots at the lowest values of kZ we attribute to the persistence of chromatin segments of 2 - 3 nucleosomes at low ionic strength. Fig. 4 shows a linear dependence of the square-root of the zero-angle values of k l ( k ) , i.e. V[kZ(k)lk+, on the solvent scattering density. The intercept on the abscissa gives a value for the mean scattering length density of the oligomers of nucleosomes and associated proteins, emean.of (2.80 & 0 . 1 0 ) ~ 10" cm-2, equivalent to the scattering length density of the water mixtures 48 2H20/52'%, H20. This corresponds to a protein/DNA ratio of about 1 . 1 which is in good agreement with the chemical analysis. The products [kZ(k)], were calibrated with the zero-angle intensities of chicken erythrocyte chromatin core particles obtained during the same neutron run and also by comparison with the incoherent scattering from H20. From this the mass per unit length of the nucleofilament could be estimated as 26 000 _+ 6000 g mol-' nm-'. A similar value has been reported for the mass per unit length of gently isolated rat liver chromatin at low ionic strengths from low-angle X-ray scatter [33]. For chicken erythrocyte chromatin the mass per unit length corresponds to one nucleosome per 8- 12 nm and because the DNA repeat is 212212 base pairs of DNA [341 this gives a DNA packing ratio of between 9.0 and 6.0. The cross-section radius of gyration, R,, of the nucleofilament is obtained from the Guinier plots of Fig. 3 through the relationship, slope= R,2/2. The variation of the cross-section R, with contrast has been analysed by Equ. (2) given above. The plot of R,"against lie given in Fig. 5, is linear with a positive slope. Three parameters are obtained from this plot. First, the intercept at l / @ = O , i.e. at

597

P. Suau, E. M. Bradbury, and J. P. Baldwin

1

-1

1 0 ' O / p (an2)

Fig. 5. Dependence of the square of the cross-section R, obtained from Fig. 3, on the reciprocal of contrast I I@.

infinite contrast, gives Rf = R: where R, in effect is the cross-sectional radius of gyration of the pure shape of the nucleofilament. Internal fluctuations of scattering density, i. e. from the details of internal structure, are negligibly small at infinite contrast and R, is the radius of gyration of the cross-sectional shape of the nucleofilament as though it were filled with uniformly dense neutron-scattering material. From the intercept, R,=2.60& 0.04 nm. This is substantially lower than the radius of gyration of gently isolated rat liver chromatin of 3.8 nm [33] from X-ray scatter studies. One major difference between chicken erythrocyte chromatin and rat liver chromatin is the latter contains a much higher amount of non-histone proteins. If the location of the non-histone protein was on the outside of the 10-nm filament it would increase the apparent cross-section radius of gyration. Secondly, the positive value of the tangent a=(2.7 indicates that the more strongly scatter20.3) x ing component (i. e. DNA) is weighted to the outside of the nucleofilament in relation to the more weakly scattering component, the histones. Lastly, the plot of R,' versus l/G is close to linear which shows that the parameter fl is close to zero. This implies that the centres of scattering mass of the histone and DNA in the cross-section are close together. Mode.for the Nucleojlament

Various models have been explored to explain the low-angle neutron scattering parameters of the nucleofilament which incorporate the known features of the shape and internal structure of the core particles [18-201. In these models nucleosomes are spaced by 11 nm. Other parameters which have to be taken into account are: (a) the cross-section radius of gyration when DNA dominates the neutron scatter, i. e. RgoNA=3.4 nm, (b) the cross-section radius of gyration

of the nucleofilament when protein dominates the scatter, R,, = 2.1 nm, (c) R, = 2.6 nm, and (d) the parameter a=2.7. It is instructive to compare these parameters with the radii of gyration of the protein and DNA components about different axes in the core particle using the model obtained from neutron scatter studies of core particles [20]. Consider first the DNA component : from the model for the disc-shaped core particle, the radius of gyration is a maximum about the minor axis and has a calculated value of 4.5 nm. This is very much larger than the cross-section R,,,, of the nucleofilament of 3.4 nm. For any axis inclined at an angle 6 to the minor axis of the core particle the radius of gyration is approximately2-v times the radius of gyration about the minor axis. From this the lowest value for R , is about a major axis and is calculated to be 3.2 nm which is close to the observed value for the cross-section RgDNA.In the nucleofilament we have to take the linker DNA into account. The extreme situation is with all the linker DNA lying on theaxis ofthenucleofilament. With this arrangement the linker DNA would not contribute significantly to 1e(r)?d3r in the expression R g = j @(r)?d3r/je(r)d3r because of the small distance r of the scattering densities of the linker DNA from the axis; but it would increase j @(v)d3rby about 312, thus reducing R, by m r 0 . 8 2 . This gives minimum values of 3.7 nm with the core particles face-to-face and 2.6 nm with the core particles edge-to-edge. Because of the mass per unit length data, it would be impossible for all the linker DNA to lie along the axis of the nucleofilament and it follows that the cross-section R,,,, would be larger than these minimum values. From these arguments therefore an arrangement of core particles is with which is consistent with the measured RgDNA the core particles edge-to-edge, or close to edge-toedge, i.e. with the faces inclined within an angle of 20" to the axis of the nucleofilament. Consider now the radius of gyration of the nucleofilament when the histone dominates the scatter. The model for the core particle which gave the best fit to the fundamental neutron scatter functions, I , and I s , contained 75% of the histones, the apolar regions of the histones in the protein core of the core particle and 25% of the histone, the basic N-terminal regions located in the outer DNA-rich region of the core particle [20]. With this model the calculated R, for the protein about the minor axis of the core particle would be 3.0 nm which is substantially larger than the measured value for the cross-section radius of gyration RgDNaof 2.1 nm. Following the arguments outlined above in the case of the DNA component, the minimum R, value, about the major axis of the core particle, is 2.3 nm which is closer to the observed cross-section R,,. Now histone H 1 and H 5 are present in the nucleofilament but not the core particles. With core particles arranged face-to-face in the nucleofilament

598

Higher-Order Structures of Chromatin in Solution

A

B

C

Fig. 6 . Models f o r the arrangement of nucleosomes in the nucleofilament. The nucleosomes are separated by 11 nm. (A) Faces of the core particles are inclined within 20" to the axis of the nucleofilament; (B) core particles are edge-to-edge and parallel; (C) core particles are face-to-face and separated by a linker region

the transverse radius of gyration, which is too high compared to the measured value, could be reduced by locating these very-lysine-rich histones on the axis of the nucleofilament. With this arrangement the reduction in the values of radius of gyration would be by a factor of 0.9 which would reduce the radius of gyration about the minor axis to 2.7 nm, still much larger than the experimental value. However when the core particles are arranged edge-to-edge the crosssection radius of gyration is reduced to 2.1 nm which is in agreement with the experimental value of 2.1 nm. Thus the arrangement of core particles which gives a cross-section radius of gyration for the protein component consistent with the experimental values is with the face of the disc nearly parallel to the axis of the nucleofilament, i. e. with the core particles edge-toedge. A similar argument applies to the R,determination. The cross-section R, for the nucleofilament is 2.60 nm which is very much lower than the value of R , for the core particle of 3.94 nm. If in the nucleofilament the core particles were face-to-face then the cross-section R , would be very close to the value obtained for the isolated core particle. Again the only way to reduce the cross-section R, is to incline the faces of the core particles to the axis of the nucleofilament or to arrange them edge-to-edge. In this extreme arrangement the calculated R, would be about 2.8-3.0 nm which would be reduced further if the linker DNA and histones H 1 and H 5 were located close to axis of the nucleofilament. If a model gives values for R,, RgDNAand R,, which are close to the experimental values, then it follows that the calculated value of a must also be close to its experimental value. A model which gives agreement for all these parameters has the disc-shaped

core particles parallel and edge-to-edge (Fig. 6B) or with the faces of the core particles inclined to within 20" to the nucleofilament axis (Fig. 6A). In the arrangement of Fig. 6 B the discs could be twisted with respect to each other about the axis of the nucleofilament. Models which d o not give values for the above parameters in agreement with the experimental values include the arrangement of Fig. 6 C where the core particles are face-to-face with the linker DNA and having H 1 and H 5 located at a smaller radius than the major axis of the core particles. Various coilings of DNA are possible in the models of Fig. 6 A and 6B. In the model of Fig. 6A with the DNA coil around the core particles always in the same direction, the linker DNA can span from the 'top' of one core particle to the 'bottom' of the next core particle with a relatively small change in the geometry of the DNA coil. This change could be induced by the N-terminal regions of the core histones or by the interactions with histone H 1. In the model of Fig. 6 B, an additional distortion of the DNA linker would be required to incorporate a similar DNA coiling. Other types of DNA coiling and linker arrangements can be envisaged but, as yet, there is no hard data to distinguish between the different possibilities. The '30-nm' Chromatin Filament

X-ray microanalysis of frozen nuclei [35] indicate that a large variety of monovalent and divalent cations are probably involved in stabilizing the three-dimensional structure of chromatin. It is also known that the nucleofilament undergoes a transition to a higherorder coil on increasing the ionic strength of aqueous solutions (5,121. This transition has been studied by increasing the ionic strength of the low-ionic-strength solutions of the nucleofilament either with sodium chloride or with magnesium chloride. Initially, as the ionic strength with NaCl was increased from the low-ionic-strength solution used for the 10-nm nucleofilament, there was no change in the neutron scattering curves. However quite suddenly, above 20 mM NaCl the Guinier region shifted to lower k values and an apparent cross-section radius of gyration of about 9 nm was observed. We have determined the cross-section radius of gyration and the nucleosome packing ratio for a family of structures which are induced in solution in a reproducible manner by increasing the MgCl, or NaCl concentration up to the point where chromatin precipitates. A convenient way to study the folding of the nucleofilament with increase of ionic strength is to assume that the neutron rigid rod analysis can be applied in an approximate manner, although at some intermediate ionic strengths the partially folded structures may not be well-defined rods of tightly packed

599

P. Suau, E. M. Bradbury, and J. P. Baldwin

r

4.0

3.0

2

-

2.0

Y b-

Y

c

1.0

0

0.05

0 k2( nm-?

Fig. 7. ModifiedGuinier plots for chromatin at different ionic strengths. (A) 0.4 mM MgClz, pH 7.8; (B) 0.8 mM MgCl,, pH 7.8; (C) 70 mM NaC1, 1 mM EDTA, pH 7.8; (D) 2.0 m M MgCI,, pH 7.0. All solutions were buffered with 10 m M Tris-HC1, 1 m M sodium cacodylate

Table 1. Packing ratios and cross-section radii of gyration in various ionic conditions Data were obtained as described in Fig. 7 Curve Ionic conditions of Fig. 7 salt concn

pH ~

~

mM D B A C

MgC1, MgClz MgC1, NaCl

2.0 0.8 0.4 70.0

7.0 7.8 7.8 7.8

Nucleosomes packing ratio

Radius of gyration

nm-'

nm

0.6 0.24 0.2 0.38

9.50 9.30 9.20 9.40

~

nucleosomes. Fig. 7 shows the Guinier plots for gently prepared multiniers of nucleosomes in a variety of ionic strengths. It can be seen that as the ionic strength is increased the linear Guinier region becomes more extended, probably indicating that the higher-order structure becomes more regular. Table 1 summarises the values of the cross-section R, and the packing ratio of the nucleosomes per unit length for each of the ionic conditions. The most striking feature of these results is the nearly constant value for the cross-section R, while the packing ratio varies from 0.2 to 0.6 nucieosome/nm. This suggests that in solution there is a family of supercoils of nucleosomes of similar radius of gyration but with different degrees of compaction. Additional data consistent with the view o f a family of chromatin supercoils which respond to the ionic environment in an aqueous solution is given by the neutron scattering curves at higher angles as shown in Fig. 8. All curves show an inflection or shoulder which moves to higher angles with increasing ionic strength

0

0.1

0.2 0.3 5 ( = 2 sin Oii.) (nnr-')

0.4

0.5

Fig. 8. Neutron scatter curvesfor chromatin in 100% 'H,O at carious ionic conditions. 0.8 mM MgCI,, pH 7.8; (-----)70 mM NaC1, 1 mM EDTA; (-) 2 mM MgCI,. All solutions were buffered against 10 m M Tris-HC1 1 mM sodium cacodylate

in parallel with the increase in packing ratio : at intermediate ionic strengths, 0.4 and 0.8 mM Mg2', when the packing ratio is 0.2-0.24 nucleosome/nm, a shoulder is observed at 20-25 nm; at 70 mM NaCl, when the packing ratio is 0.38 nucleosome/nm, an inflection is found a t 12.5 nm; while for 2 mM MgZ+, when the packing ratio is 0.6 nucleosome/nm, the shoulder has moved to 10-11 nm. The observed packing ratios for the most compact supercoil requires 7 + 1 nucleosomes per turn of the supercoil. The secondary maximum in the three scatter curves at about 3.3 nm is known to come from the internal structure of the core particle [20]. Thus this feature is expected to be present in the neutron, or X-ray, scattering curves of all higher-order chromatin structures which preserve the intact core particle structure. Transverse Radial Distribution Function

We have carried out a n inverse transform of the Hankel integral for the k l ( k ) versus k neutron scatter curves of the most contracted form of the nucleosome supercoil in H20/2 mM Mg2+. The plot of D ( R ) [ = 2 n R P ( R ) ]versus R is shown in Fig. 9. The transverse radial distribution function D(R) goes to zero at the maximum transverse dimension of this secondorder chromatin structure. From Fig. 9 it can be seen that D(R) goes to zero at 34 nm. This observation has been repeated and we conclude that the maximum transverse dimension of the hydrated second-order chromatin structure is 34 nm.

600

Higher-Order Structures of Chromatin in Solution

folded structure. Neutron scatter studies of this transition gave the interesting result that above 20 mM NaCl or 0.4 mM MgCl, the cross-section R, increases abruptly to about 9 nm with a packing ratio of 0.2 nucleosome/nm and then with increasing ionic strength I I 1 \ the R, increases slowly to about 9.5 nm while the 0 0 10 20 30 40 50 packing ratio increases threefold to 0.6 nucleosome/ R(nm) nm. This suggests a transition to a supercoil whose Fig. 9. The cross-section radial distribution function, D( R), ,for pitch depends on ionic strength of the solution. In cliromcrtin in H 2 0 , 2 mM MgC12, p H 6.9, I 0 m M Tris-HCI, I m M parallel with the increase in packing ratio with ionic cacodylate strength, an inflection is observed in the experimental scatter curves which moves from 20 - 25 nm at 0.8 mM MgC1, to 10-11 nm by 2 mM MgCI,. The most contracted state of the supercoil corresponds with the DISCUSSION coil model proposed from neutron diffraction studies of fibres [ 14,151 or the solenoid model proposed from It is difficult to determine the precise arrangement electron microscope studies 1121. In many electron of nucleosomes in the nucleofilament or in the secondmicroscope studies the range of diameters of this order supercoiled chromatin structure. Ideally higherhigher-order coil has been given as 25 - 30 nm which order chromatin structures should be studied in the is smaller than the diameter of the coil from the radial same ionic environments as the cell nucleus but at distribution function of 34 nm. These differences can present these conditions are not known with any probably be attributed to the different states of the certainty. We have used neutron scatter techniques samples which are dehydrated for electron microscope to characterize further the structure of the linear array studies and hydrated for neutron studies. The diffracof nucleosomes at low ionic strength and the secondtion at about 10- 11 nm observed in X-ray and neutron order chromatin structure induced by increase of ionic diffraction of fibres, gels and solutions has been attristrength. buted to the pitch of a coil [14,15] or solenoid 1121. If At low ionic strength the mass per unit length of this is correct then changes in the spacing would 26 000 +_ 6000 g mol - nm - ' corresponds to one nuindicate that the pitch of this coil is variable. Studies cleosome/8 - 12 nm and a DNA compaction ratio in of the effect of hydration on the structure of chromatin the chicken erythrocyte nucleofilament of 6 - 9. This have shown [40] that on dehydration of chromatin is close to the initial proposals of Kornberg [37], fibres in 2 H 2 0a neutron diffraction peak observed at although this proposal was based on a probably in10.2 nm at 40% (w/w) concentration moves progrescorrect assumption concerning the origin of the semisively to higher angles to an equivalent spacing of 8.4 meridional arc at 10- 11 nm in the fibre X-ray difnm in the 'dry' state at 87 (w/w) concentration. Also fraction patterns of chromatin. on hydration of fibres of H 1-depleted chromatin, a The dependence of the cross-section radius of gyration of the nucleofilament on contrast gives RgDNA, neutron diffraction peak at 10.5 nm at 44% (w/w) moves progressively to lower angles and is found at when DNA dominates the scatter, R,, when protein 13.5 nm at 26% (w/w) concentration 1411. These obdominates the scatter, R, and a. From our understandservations are consistent with a supercoil of variable ing of the structure of the core particle [18 -201, and pitch which responds to the ionic conditions of aqueous in particular using the model proposed from neutron solutions or the s t t e of hydration of fibres of chroscatter studies [20], various arrangements of the core matin. From the mass per unit length measurements particles in the nucleofilament could be tested. Models there are about seven nucleosomes per turn of the consistent with these parameters are shown in Fig. 6A supercoil of pitch 11 nm. and 6 B in which the core particles were arranged With a supercoil in which every nucleosome exparallel and edge-to-edge or with their faces inclined periences the same environment there must be a hole to within 20" to the axis of the nucleofilament. Various along the axis of the coil and the diameter of this hole models have been proposed for the arrangement of will depend on the arrangement of the nucleosomes in nucleosomes in the nucleofilament 138,391. A general the nucleofilament. If the models suggested for the feature of these models is that they consist of DNA nucleofilament shown in Fig. 6 A and B persist at higher coiled around octameric cores of histones which are in contact along the axis of the nucleofilament. Such ionic strengths when supercoiled then there are two and a than are models give higher values for R,,,, possible arrangements of nucleosomes : either the axes observed. of the disc-shaped core particles are arranged parallel As found by others [5,12], on increase of ionic to the axis of the supercoil, i. e. a thick-walled cylinder, strength with MgCl, or NaCl there is a transition or the axes are arranged radially and perpendicular to from the linear array of nucleosomes to a higher-order the axis of the supercoil, i.e. a thin-walled cylinder.

.. ..

.... ..........-

'

601

P. Suau, E. M . Bradbury, and J. P. Baldwin

Taking the maximum transverse dimension of the supercoil as 34 nm, then with the former arrangement of nucleosomes a supercoil would have a hole of diameter 12 nm while in the latter arrangement the hole would be 23 nm. The approximate transverse radius of gyration for these arrangements would be 11.9 and 14.2 nm respectively, both higher than the observed value of 9.5 nm, particularly the latter value. Further, a supercoil with a hole of diameter 22 nm would be expected to give a pronounced inflection in the radial distribution function which is not observed. It would appear therefore that a model with the core particles coiled so that they face out from the supercoil, i.e. a thin-walled cylinder, is not in accord with the data. With the core particles having their axes parallel to the axis of the supercoil, the smaller hole of 12-nm diameter may be dificult to detect for the following reasons. a) There are non-histone proteins or, as suggested [12], lysine-rich histones in the hole and their presence ‘softens’ the discontinuity in the neutron scattering density at the inner radius of the coil. b) The most contracted form of the coil is not uniform over extended regions and this non-uniformity causes a ‘blurring’ of the radial distribution function. c) A possibility which cannot be excluded is that the ‘hole’ does not exist but contains nucleosomes. In this situation about one in seven of the nucleosomes would lie on the axis of the coil and would thus be different from the nucleosomes coiled on the outside. This difference could reside in the histones and chromosomal proteins associated with the linker DNA region. It is relatively easy to coil the edge-to-edgearrangement of nucleosomes depicted in the nucleofilament model of Fig. 6B into a flat supercoil with a pitch of 10-11 nm and diameter of 33-34 nm. If every seventh nucleosome can be ‘flipped’ out of the linear array then it can be incorporated into the centre of the supercoil without difficulty. Each supercoil turn would then contain a close packed array of six nucleosomes coiled around a central nucleosome and the flat nucleosomes would be stacked, in effect, along the axis of the supercoil. Such a model would explain many of the features of the second-order chromatin structure: (a) taking the diameter of the core particle as 11 nm, a close-packed supercoil would have a diameter of at least 33 nm in agreement with the value obtained from the radial distribution function; (b) the packing of seven nucleosomes/turn of pitch 11 nm is close to the observed value of six nucleosomes per 10 nm for the supercoil in its most compact form in 2 mM MgC1, ; (c) no inflection would be expected in the radial distribution function, ‘solid cylinder’models have been found to give a close approximation to this function 1421; (d) the partial disruption of such a

close-packed coil would give the appearance of ‘superbeads’ [ 131 or ‘clustered arrays’ [43] with each of these entities consisting of six nucleosomes coiled around a central nucleosome. It follows that it may be possible to isolate a stable complex of about seven nucleosomes by brief nuclease digestion under conditions of high salt concentration. Clusters of nucleosomes containing about eight nucleosomes have been found by Smulson and coworkers (T. R. Butt, D. B. Jump & M. E. Smulson, personal communication). With increase of ionic strength the transition of the nucleofilament to the supercoiled form may be accompanied by a rearrangement of the nucleosomes in the nucleofilament. This possibility and further tests of the model described above are in progress through neutron scatter studies of gently isolated chromatin which has not been exposed to low-salt conditions. It should be emphasized that in these studies ofchromatin higher-order structures, we are investigating the properties and conformational states of pieces of chromatin dissected out from chicken erythrocyte nuclei by brief nuclease digestion. Whether the additional constraints of the incorporation of chromatin into loops [1,2] or domains [3,4] modifies these conformational states remains to be shown. We thank Mrs E. Boulter and Dr B. G. Carpenter for their help in sizing the DNA lengths of the chromatin pieces used in these studies, Mr A. W. Thorne for unpublished work and Dr K. Ibel, Dr G . G. Kneale and Mr G. W. Braddock for discussion of the work and criticisms of this manuscript. The neutron experiments were performed at the Institute Laue-Langevin (Grenoble, France). These experiments were supported by the Science Research Council of the U.K.

REFERENCES 1. Poulson, J. R. & Laemmli, U. K. (1977) Cell, 12, 817. 2. Laemmii, U. K., Cheng, S. M., Adolph, K. W., Poulson, J. R.. Brown, J. A. & Baumbach, W. R. (1978) Cold Spring Harbor Symp. Quant. Biol. 42, 351. 3. Igo-Kemenes, T., Greil, W. & Zachau, H. G . A. (1977) Nucleic Acids Res. 4, 3387. 4. Igo-Kemenes, T. & Zachau, H. G. (1978) Cold Spring Harbor Symp. Quant. Biol. 42, 109. 5. Ris, H. & Kubai, D. F. (1970) Annu. Rev. Genet. 4, 263-294. 6. Woodcock, C. L. F. (1973) J . Cell. Biol. 59, 737a. 7. Olins, A. L. & Olins, D. E. (1974) Science (Wash. D. C . ) lX3, 330. 8. Oudet, P., Gross-Bellard, M. & Chambon, P. (1975) Cell, 4 , 281. 9. Gall, J. G. (1966) Chromosoma, 20, 221 -233. 10. DuPraw, E. J. (1970) in D N A and Chromosomes, pp. 132 - 185, Holt, Rinehart and Winston, New York. 11. Davies, H. G., Murray, A. B. & Walmsley, M. R. (1974) J . Cell Sci. 16, 261 -299. 12. Finch, J. T. & Klug, A. (1976) Proc. Nut1 Acad. Sci. U . S . A . 73, 1897 - 1901. 13. Renz, M., Nehls, P. & Hozier, J. (1977) Proc. Natl Acad. Sci. U.S.A. 74, 1879-1883.

602

P. Suau, E. M. Bradbury, and J. P. Baldwin : Higher-Order Structures of Chromatin in Solution

14. Carpenter, B. G., Baldwin, J. P., Bradbury, E. M. & Ibel, K. (1976) Nucleic Acids Res. 3, 1739- 1746. 15. Baldwin, J. P., Carpenter, B. G., Crespi, H., Hancock, R., Sthephens, R. M., Simpson, J. K. & Bradbury, E. M. (1978) J . Appl. Crystallogr. 11, 484 - 486. 16. Carlson, R. D. & O h , D. E. (1976) Nucleic Acids Res. 3, 89. 17. Hjelm, R. P., Kneale, G. G., Sum, P., Baldwin, J. P. & Bradbury, E. M. (1977) Celt, 10, 139-151. 18. Finch, J. T., Lutter, L. C., Rhodes, D., Brown, R. S., Rushton, B., Levitt, M. & Klug, A. (1977) Nature (Lond.) 269,29-36. 19. Pardon, J. F., Worcester, D. L., Wooley, J. C., Cotter, R. I., Lilley, D. M. J. & Richards, B. M. (1977) Nucleic Acids Res. 4, 3199-3214. 20. Suau, P., Kneale, G. G., Braddock, G. W., Bladwin, J. P. & Bradbury, E. M. (1977) Nucleic Acids Res. 4, 3769-3786. 21. Murray, K., Vidali, G. & Neelin, J. M. (1968) Biochem. J . 107, 207. 22. Laemmli, U. K. (1970) Nature (Lond.) 227, 680-685. 23. Zamenhof, S. & Chargaff, E. (1957) Methods Enzymol. 3, 696 - 704. 24. Schmatz, W., Springer, T., Shelten, J. & Ibel, K. (1974) J . Appl. Crystallogr. 7, 96. 25. Ibel, K. (1976) J. Appl. Crysrallogr. 9, 630. 26. Guinier, A. & Fournet, G . (1955) Small-Angle Scattering of X-Rays, Wiley and Sons, New York. 27. Luzzati, V. (1960) Acta Crystallogr. 13, 939-945. 28. Federov, B. A. & Aleshin, V. G. (1966) Vysokornol. Soyed. 8, 1657- 1666.

29. Stuhrmann. H. B., Haas, J., Ibel, K., DeWolf, B., Koch, M. H. J., Parfait, R. & Crichton, R. R. (1976) Proc. Natl Acad. Sci. U.S.A. 73, 2379-2383. 30. Ibel, K. & Stuhrmann, H. B. (1975) J . Mol. Biol. 93, 255-266. 31. Luzzati, V., Tardieu, A., Mateu, L. & Stuhrmann, H. B. (1976) J . Mol. Biol. 101, 115. 32. Kneale, G. G., Baldwin, J. P. & Bradbury, E. M. (1977) Qu. Reo. Biophys. 10, 485 - 527. 33. Sperling, L. &Tardieu, A . (1976) FEBS Lett. 64, 89-91. 34. Morris, N. R. (1976) Cell, 8, 357 -363. 35. Cameron, I. L., Sparks, K. L., Horn, K. L. & Smith, N. R. (1977) J . Cell Biol. 73, 193-199. 36. Lindigkeit, R., Bottger, M., Mickwitz, C. U., Fenske, H., Karawajew, L. & Karawajew, K . (1977) Acta Biol. Med. Germ. 36, 275 -279. 37. Kornberg, R. D. (1974) Science (Wash. D.C.) 184, 868-871. 38. Kornberg, R. D. (1977) Annu. Reo. Biochem. 46, 931 -954. 39. Worcell, A. (1978) Cold Spring Harbor Symp. Quant. Biol.42, 31 3 - 323. 40. Baldwin, J. P., Boseley, P. G., Bradbury, E. M. & Ibel, K. (1975) Nature (Lond.) 253, 245. 41. Bram, S., Butler-Browne, G., Bradbury, E. M., Baldwin, J. P., Reiss, C. & Ibel, K. (1974) Biochimie (Paris) 56, 987. 42. Baldwin, J. P., Boulter, E., Braddock, G. W., Bradbury, E. M., Cary, P. D., Carpenter, B. G., Kneale, G. G. & Suau, P. (1978) FERS S,LW~. 51, 171 -191. 43. Olins, A. L. (1978) Cold Spring Harbor Symp. Quanf. Biol. 42, 325.

P. Suau, Laboratoire de Biophysique, Museum National d’Histoire Naturelle, 61 Rue Buffon, F-75005 Paris, France E. M. Bradbury* and J. P. Baldwin, Biophysics Laboratories, Portsmouth Polytechnic, St Michael’s Building, White Swan Road, Portsmouth, Great Britain, PO1 2 D T

* To whom correspondence should be addressed