J. Cell Sa. 72, 75-87 (1984) 75Printed in Great Britain © The Company of Biologists Limited 1984

NUCLEAR PORES AND INTERPHASE CHROMATIN:

HIGH-RESOLUTION IMAGE ANALYSIS AND FREEZE

ETCHING

C. NICOLINI

Temple University, Philadelphia, U.SA. and Eminent Chair of Biophysics, Faculty ofMedicine, University of Genova, Italy

G. VERNAZZA, A. CHIABRERAIstituto di Elettrotecnica, Sezione di Ingegneria Biofisica ed Elettronica, Universita di

Genova, Italy

I. N. MARALDI AND S. CAPITANIIstituto di Anatotnia, Universita di Bologna, Italy

SUMMARY

Computer-enhanced analysis of electron micrographs of thin-seCtioned rat liver nuclei, combinedwith three-dimensional reconstruction of the same Feulgen-stained nuclei, points to a uniqueclustering of chromatin DNA fibres near the nuclear border.

Computer-enhanced image analysis has been applied to electron micrographs of the envelopes ofthe same rat liver nuclei prepared by freeze etching and a few essential geometrical parameterscharacterizing the pores and their distribution have been determined. During interphase, clustersof nuclear pores, closely paralleling the clustering of membrane-attached chromatin fibres, havebeen identified on the envelope, the number of these being similar to the number of homologus pairsof metaphase chromosomes. Furthermore, rapid changes induced in chromatin distribution appearto be associated with rapid changes in pore number, but not in the number of pore clusters.

INTRODUCTION

The double membrane surrounding the nucleus contains as its most conspicuousfeatures the nuclear pore complexes, which have been assumed to be the sitesof molecular and ionic exchange between nucleus and cytoplasm (Maul, 1977).Curiously, these channels in the nuclear envelope are referred to as pores, even if theyare usually filled with a dense plug (Krohne, Franke & Schree, 1978; Unwin &Milligan, 1982). Biochemical characterization suggests that the nuclear pore complexis formed only of a few majdr polypeptides and some RNA (Krohne et al. 1978), butno direct identification has yet been possible within the boundaries of this structureand all speculations about its function are based on scanty, circumstantial and at timescontradictory data obtained by limited, ultrastructural and autoradiographic methods(Franke & Scheer, 1970; Maul, 1977). While the complex has long been known toconsist of a cylindrical assembly that spans the inner and outer nuclear membranesand is arranged with octagonal symmetry about a central axis, new detailed informa-tion has recently been obtained from electron micrographs of an intact oocyte nuclear

Key words: nuclear pores, chromatin, image analysis.

76 C. Nicolini and others

envelope by modern image-processing methods (Unwin & Milligan, 1982) at aresolution of 90 A. By these high-resolution methods the pore structure appears as athree-layered sandwich, consisting of a disk between two thin rings co-planar with onehalf of the double membrane. The disk, consisting of a central plug and eight broadspokes, is suspended between the two rings by extension of the spokes, at the peri-phery of which, at a diameter of about 100 nm, the two membranes fuse. Thesefindings, which also provide evidence for the attachment of eight ribosomal particlesto the cytoplasmic face of this structure (Unwin & Milligan, 1982), raise numerousfurther questions about the structure and function of the nuclear pores.

Pore complexes are no longer considered to be randomly distributed but present atspecific sites (Maul, 1977). Furthermore, rapid changes in pore pattern and numberhave been found to occur in nuclear membranes of various cell types in differentfunctional states (Markowicz, Glass & Maul, 1974). A biphasic increase in porefrequency, associated with sharp transitions from a clumped to a uniform poredistribution, has been found in synchronized HeLa cell and phytohaemagglutinin-stimulated human lymphocytes in the very early part of G\ and before DNA replica-tion (Markovicz et al. 1974). Curiously, this biphasic change is exactly correlated withabrupt changes in the higher-order chromatin structure from a clumped to a uniformdistribution, occurring at the same time in these cells (Kendall et al. 1980). Evidencesuggesting the attachment of chromatin fibres to the nuclear envelope at the pores hasrecently been provided by freeze-etching of rat liver nuclei (Nicolini et al. 1983) andby fluorescence microscopy (Agard & Sedat, 1983), but conclusive proof of thepresence of thin chromatin threads in the pore complexes is difficult by any presenttechnique of electron microscopy. In a mammalian cell, the nuclear pores are rapidlyformed in early telophase, when membrane pieces are seen attached to thechromosomes (Maul, 1977), and they are present with varying numbers anddistribution throughout interphase (G\-S-Gz). A question that arises is whether,when pore clusters appear, there is any correlation between the number of theseclusters and the number of chromosomes (even for changing chromatin and poredistributions). It may also be asked whether there is a similar number of clumps ofchromatin near the envelope during interphase.

These questions may be answered by further exploring the exact distribution ofnuclear pores, using a statistical method, in relation to the parallel changes in porenumber and chromatin distribution taking place in cells in different functional statesbut of constant chromosome number. To increase the resolution, computer-enhancedimage analysis has been utilized in conjunction with light and electron microscopy(EM), for the studies reported here.

MATERIALS AND METHODS

Isolated rat liver nuclei (Widnell & Tata, 1964) were fixed in 2-5 % (v/v) glutaraldehyde in0-1 M-phosphate buffer (pH 7-2) for 1 h and then rinsed in 0-15 M-phosphate buffer. In such isolatednuclei, before fixation, RNA synthesis is readily induced (within 1 min) by exposure to theappropriate concentration of phospholipid vesicles (namely, 1-5 min of phosphotidylserine). Thismethod has been used to obtain nuclei in a state of high metabolic activity.

The nuclear pellet was then processed in four ways.

Nuclear pores and interphase chromatin 11

(1) Directly stained with acridine-orange for differential monitoring of chromatin DNA, usingfluorescence microscopy (Nicolini, 1979).

(2) Stained with uranyl acetate following thin sectioning (500 A thick), for subsequent electronmicroscopy (Manzoli et al. 1982).

(3) Sectioned at 2//m and Feulgen-stained for differential chromatin DNA absorption, by meansof quantitative light microscopy (Kendall et al. 1980).

(4) Resuspended in 30 % glycerol in distilled water for 30 min, frozen in Freon and processed forcleaving and replication in a Balzers 360 M freeze-etch device.

Analytical image acquisition and processing

High-resolution image analysis was conducted as described in detail recently (Nicolini et al. 1983;Kendall et al. 1980; Belmont, Kendall & Nicolini, 1984), either on Feulgen-stained nuclear sectionsor on electron micrographs of freeze-etched nuclear envelopes and of sections stained with uranylacetate.

In the latter case, the EM pictures were imaged through a macroepidiascope (final opticalmagnification = 24). Individual EM pictures were acquired in an array of several thousand picturepoints. Images were acquired on a European standard TV scanner target, equipped with a Plum-bicon tube (which ensures a highly linear transfer function between light intensity and electricalsignal) and analysed by means of the ACTA system built and installed at the Biophysical andElectronic Engineering Section, Institute of Electrotechnics, University of Genova (Italy)(Beltrame et al. 1980). The final linear dimensions of each approximately square picture point,characteristic of the Plumbicon-equipped image analyser, were determined to be 0-6 or 0-9 nmunder our conditions of illumination and magnification. Individual transmittance values for eachpicture point (termed a 'pixel') were acquired in a calibrated linear scale of 256 grey levels, where0 and 256 correspond to 0 % ('black') and 100 % ('white') transmittance, respectively. The analoguevideo signal was typically fed through a fast A/D conversion group (8 bit, 30 MHz, a monoliticintegrated circuit) and each video frame could be stored in real time on a memory according to theformat 512x512pixels, 8bit resolution per pixel. Images were transferred on a mass-memorydevice, such as magnetic tape or disc, interfaced to a HP 21MX minicomputer (which controls theACTA system).

High-resolution densitometric and geometric image analysis of the Feulgen-stained sections wereperformed on a Quantimet 720-D, as previously described at length (Kendall et al. 1980; Belmontetal. 1984).

Cluster analysisMany methods can be used for cluster analysis. We have selected two different methods in order

to verify independently pore cluster assignments on the nuclear envelope. The first one is calledISODATA (Ton & Gonzales, 1974) and it is characterized by an iterative procedure, whichdetermines the number of clusters and the co-ordinates of each cluster centre by grouping samplemeans. Widely different values are given initially for these parameters to verify optimal convergenceon the 'true' final number of clusters (NC) (the term 'ISODATA' stands for Iterative Self-Organizing Data Analysis Technique A).

The second graph-theoretical method is called MST (Minimal Spanning Tree) (Dude & Hart,1973). After the construction of a tree graphically connecting all points (pores) through their closestdistance, the longest edges are progressively obtained. For the MST method the basic feature is toselect the maximum length for'cutting', i.e. the length above which the corresponding interconnectedpores belong to different clusters; normally a reference is made to the average value (d) of all treeconnections and Kd is taken as the maximum length, where 1 <K< 4.

While ISODATA seeks classical cluster configurations according to statistical properties (meanand variance), MST can also find laminar clusters, through a very lengthy computer process. Inorder to reduce border effects the entire nuclear envelope is assumed to consist of two identicalhemispheres, where the measured hemisphere is the mirror image of the unknown hemisphere; thisallows determination of the total number of pores for each nuclear sphere.

Numerous tests have been performed with positive results on the ISODATA and MSTprograms, to establish their ability to recognize automatically clusters of points grouped in differentconfigurations on the surface of a sphere (Bozzo & Risola, 1982). For these studies we have selected

78 C. Nicolini and others

nuclear envelopes only of a diameter identical to the diameter (about 7/un) of nuclei isolated fromGo rat liver tissue.

RESULTS

The pore distribution on the nuclear membrane, as apparent in freeze-etchedmicrographs, is shown in Fig. 1 for rat liver nuclei in different functional states;namely, with low and high metabolic activity. The abrupt change in metabolic activitybrought about by phospholipid treatment (Manzoli et al. 1982) is accompanied byabrupt changes in both pore number (Fig. 2 and Table 1) and chromatin distribution(Fig- 2).

However, an independent statistical analysis of each nuclear envelope, extrapolatedto the whole nucleus, yields a similar number of 19—21 total clusters of pores for bothunstimulated and phospholipid-stimulated nuclei, using either the ISODATA (seeTable 1) or the MST (see Fig. 3) algorithms.

From the ISODATA algorithm, for every cluster the distribution of the pores fromthe centre of the cluster has also been computed; a toroidal distribution consistentlyresults, with a mean value typically of about l-Ofim (the nuclear diameter is about7^tm) and with a quasi-Gaussian form. Conversely, by means of the MST algorithmthe influence of the cutting parameter (see Materials and Methods) on the resultingnumber of clusters has been investigated (see Fig. 3). As is apparent from Fig. 3, astep-like variation in the number of clusters (NC) versus the parameter K (seeMaterials and Methods) is achieved at A— 1-75, whereby for A" values greater thanthis wide variations in A." introduce relatively small changes in NC, and for K valuesless than this value wide variations in NC occur, even for small variations in K. Theaverage value of d, taking into account all the closest distances among adjacent pores(Fig. 4), i,s223nm. Therefore we obtain about 20 clusters whenKd is about 390 nm,which would represent the maximum distance between pores belonging to the same

Table 1. Average number of clusters (NC), as integer value, for each nuclear image(1M) as computed from ISODATA (see Fig. 1)

Controls

Stimulated

IM

1561

NP

580788514390

m

18-619-520-818-6

S.D.

3-55-05-62-6

iV

30392520

NC

19202119

NP, number of pores for the whole nuclear surface after the specular extrapolation;m, mean value of clusters obtained with a wide range of values for the parameters at the beginning

of the iteration process;S.D., standard deviation of the clusters obtained with different initial values;N, avarage number of pores for each cluster.

Fig. 1. Freeze-etch micrographs of nuclear membrane from rat liver cells. Nuclei arcshown before (A) and after (B) stimulation with phospholipid vesicles.

Nuclear pores and interphase chromatin 79

* •• * " / s v

i

V

P

Vv,

BFig. 1

80 C. Nicolini and others

Stimulated ;htI

J 2 •

S !"5. •

5I :

ty

II

Black Grey level White

Fig, 2. Number of pixels with given transmittance as a function of transmittance for500 A thick nuclear sections, stained with uranyl acetate, from rat liver nuclei before(A) and 1 min after (B) phospholipid vesicle stimulation. A redistribution of chromatinfrom peripheral regions towards the inner part of the nucleus is also apparent (see insert).The bars indicate the positions of states I and II of chromatin condensation, in bothA and B.

Nuclear pores and interphase chromatin 81

90 -

80

70-

60-

50-

40-

30-

20-

10-

120-

90-

60-

30-

1

1-5

1-5 4 K

Fig. 3. A. Number of clusters versus the value Kof the cutting parameter, as determinedwith the Minimal Spanning Tree algorithm (see the text for the cluster analysis), B.Derivative of NC per unit K, as a function of K.

cluster. The number of pores per cluster is shown in Table 2 for four different nuclearenvelopes.

Another reproducible finding from the frequency distribution of the closest dis-tance between adjacent pores (Fig. 4) is that most pores are separated by about160 nm, from centre to centre.

Finally, about 18 stained areas are present near the nuclear periphery in similar cellsFeulgen-stained and three-dimensionally reconstructed (Table 3 and Fig. 5). Thereis a wide variation in the fraction of total DNA present in each chromatin body(similar to the wide variation in pore number per cluster).

82 C. Nicolini and others

Table 2. Typical number of pores per cluster on the nuclear membrane for fourdifferent nuclear envelope preparations

Cluster no.

Total123456789

10

Nuclear envelopemajor diameter (^m)

1

280374755291578

141619

7-2

Number>

2

28255431824101132131611

7-0

of pores

3

251311812531539151013—

6-7

4

1982320142210198

125

—

6-2

As justified in the text and shown in Fig. 4, the combination of pores interconnected by a distanceequal to or less than 320 nm forms a cluster. Only a few pores (10%) cannot be grouped in any clusterusing this criterion and they are either isolated or interconnected to form doublets or triplets;furthermore, they are concentrated near the periphery, where some deformation of the envelope isevident and tails connected to clusters of the opposite hemisphere are possible.

Interestingly (Fig. 5 and Table 3), similar total numbers of chromatin DNA bodiesare present in both stimulated and unstimulated nuclei, which differ only in the stateof condensation of the chromatin. The quantal nature of the transition between thetwo states, upon metabolic activation, is more conclusively indicated by the uranylacetate data obtained at the electron-microscope level (Fig. 2), which show the absor-bance ratio between the two states of condensation to be 2-75 (very close to the 2-5apparent from the Feulgen-stained large chromatin bodies at the light-microscopelevel).

DISCUSSION

Rigorous statistical analysis carried out on computer-enhanced nuclear membranesfrom various types of rat liver interphase cells demonstrates that the number of poreclusters is independent of the total number of pores and the chromatin distribution,but close to the number of chromatin bodies near the nuclear border. The fraction oftotal DNA present in most chromatin clusters is about 0-9-7% (Table 3), which isin striking agreement with the fraction of total DNA present in each metaphasechromosome from the same cells (0-8-5%). Curiously, the proportion of the totalnumber of pores in each pore cluster of the envelope also ranges between 1-2 and9-9%.

C Nicolini and others

O

euo

SiE

i

iii

4 12 20 28

Average optical density (arbitrary units)

Fig. 5. Number of chromatin bodies with given average absorbance as identified in three-dimensional reconstructed nuclei from Feulgen-stained 2^m serial sections (see insert),as a function of average absorbance from Go ( ) and G\ (—) nuclei.

1979; Kendall et al. 1980; Nicolini et al. 1983). More direct evidence for a closeassociation of interphase chromatin with the nuclear membrane has been recentlyprovided in unfixed nuclei (Agard & Sedat, 1983), resulting in a predominantlychromosome-free central cavity, which confirms earlier findings in fixed nuclei(Kendall et al. 1980; Nicolini, 1979, 1980; Olins & Olins, 1979).

By optical fluorescence microscopy and three-dimensional chromosome

Nuclear pores and interphase chromatin 85

Table 3. Typical amount of DNA (integrated optical density) per identifiablechromatin body around the periphery of Feulgen-stained mammalian diploid nuclei(in Gq), reconstructed three-dimensionally for 2 \on thick sections as previously

described (Kendall et al. 1980)

Chromatin body I.O.D. (arbitrary units)

1* 106292* 149113 14724 12685 8176 6117 3118 15989 52010 207311 24112 52413 82214 92915 90116 125017 218018 3170

The total I.O.D. for the reconstructed nucleus is 44-228 arbitrary units.• Very large regions possibly resulting from the overlappiing of two or more discrete regions

('chromatin bodies' or islands of absorbance higher than the background) displaying an unusuallylarge condensation (Fig. 5).

topography of intact unfixed nuclei (Agard & Sedat, 1983) a complex mixture ofintertwined coils and parallel chromosome segments have been shown to be closelypacked against the inner surface of the nuclear membrane. The presence along thegiant 60 cm long (~10"Mr) chromatin fibres of laminar fragments - having dimen-sions centred around 120, 240 and 360 nm — and of an upper limit in the fibre lengthbetween pieces of nuclear membrane (Nicolini et al. 1983) provide an indirect justi-fication for the clustering of pores and for the upper limit to the distance between poreswithin each cluster.

Furthermore, the existence of a repeating unit of about 160 nm in the interporedistance is compatible with the observation of a repeating unit of 480 nm in thechromatin segments delimited by fragments of the nuclear lamina (as shown in figs2 and 5 of Nicolini et al. 1983), whereby a 480 nm contour length of the same fibre,when depending from the nuclear envelope, corresponds to a pore distance typicallyof 160 nm.

However, it does not escape our notice that our findings raise more questions thanthey answer, concerning the actual role of nuclear pores and the higher-orderorganization of chromatin during interphase. There is an enormous amount ofliterature on this subject (Maul, 1977; Nicolini, 1983) and a complete discussion of

86 C. Nicolini and others

the difficulties in fitting our findings with all other relevant work is beyond the scopeof this paper.

This work was partially supported by a grant from C. N. R. Finalized Project 'Oncology', ConsiglioNazionale delle Ricerche.

REFERENCES

AGARD, A. D. & SEDAT, J. (1983). Three-dimensional architecture of a polytene nucleus. Nature,Land. 302, 676-681.

Bozzo, W. & RISOLA, E. (1982). Analisi dei raggruppamenti dei pori dei pori sulla superficienucleare delle cellule. Doctoral thesis, University of Genova.

BELMONT, A., KENDALL, F. M. & NICOLINI, C. (1984). Three-dimensional intranuclear DNAorganization in situ: three states of condensation and their redistribution as a function of nuclearsize near the Gi-S border in HeLa S-3 cells. J. Cell Sci. 65, 123-138.

BELTRAME, F., CHIABRERA, A., GRATTAROLA, M., GUERRIN, P., PARODI, G., PONTA, D.,VERNAZZA, G. & VIVIANI, R. (1980). The ACTA system. 2nd Ann. Conf. of the IEEE Engineer-ing in Med. and Biol. Soc, Washington, D.C.

COMINGS, D. E. (1968). The rationale for an ordered arrangement of chromatin in interphasenucleus. Am. J . hum. Genet. 20, 440-460.

DUDE, R. O. & HART, P. E. (1973). Pattern Classification and Scene Analysis. New York: JohnWiley.

FRANKE, W. & SCHEER, U. (1970). The ultrastructure of the nuclear envelope of amphibianoocytes: a reinvestigation. J. Ultrastruct. Res. 30, 288-316.

KENDALL, F., BELTRAME, F., ZIETZ, S., BELMONT, A. & NICOLINI, C. (1980). The quinternarychromatin-DNA structure. Three-dimensional reconstruction and functional significance. CellBiophys. 2, 373-404.

KROHNE, T., FRANKE, W. & SCHREE, U. (1978). The major polypeptides of the nuclear porecomplex. Expl Cell Res. 116, 85-102.

MANZOLI, A., CAPITANI, S. & MARALDI, N. (1982). In Cell Growth (ed. C. Nicolini), pp.466-486. New York, London: Plenum.

MARKOVICZ, J., GLASS, L. & MAUL, G. (1974). Pore patterns on nuclear membranes. Expl CellRes. 85, 443-451.

MAUL, G. (1977). The nuclear and cytoplasmic pore complex: structure, dynamics, distributionand evolution. Int. Rev. Cytol. (suppl.) 6, 75-186.

MCGHEE, J. D., RAU, D. C , CHARNEY, E. & FELSENFELD, G. (1980). Orientation of the nucleo-some within the higher order structure of chromatin. Cell 22, 87—96.

NICOLINI, C. (1979). Chromatin structure from Angstrom to micron levels, and its relationship tomammalian cell proliferation. In Chromatin Structure and Function, part B (ed. C. Nicolini),pp. 613-666. New York, London: Plenum Press.

NICOLINI, C. (1980). Nuclear morphometry, quinternary chromatin structure and cell growth.J.'submicrosc. Cytol. 12, 475-505.

NICOLINI, C. (1983). Chromatin structure, from nuclei to genes. A review. Anticancer Res. 3,63-86.

NICOLINI, C , TREFILETTI, V., CAVAZZA, B., CUNIBERTI, C , PATRONE, E., CARLO, P. & BRAM-BILLA, G. (1983). Quaternary and quinternary structures of native chromatin-DNA in livernuclei: differential scanning calorimetry. Science 219, 176—178.

NICOLJNI, C , CARLO, P., MARTELLI, A., FINOLLO, R. & BRAMBILLA, G. (1982). Viscoelasticproperties of mammalian nuclear DNA. Higher order DNA packing and cell function. J. molec.Biol. 161, 155-175.

NICOLINI, C , CAVAZZA, B., TREFILETTI, V., MARALDI, N., PIOLI, F., BELTRAME, F., BRAM-BILLA, G. & PATRONE, E. (1983). Higher-order chromatin structure in rat liver nuclei: high-resolution quantitative electron microscopy. J. Cell Sci. 62, 103-115.

OLINS, A. L. & OLINS, D. E. (1979). Stereo electron microscopy of the 25-nm chromatin fibers inisolated nuclei. J. Cell Sci. 81, 260-265.

Nuclear pores and interphase chromatin 87

SEDAT, J. & MANUELIDIS, L. (1978). A direct approach to the structure of eukaryoticchromosomes. Cold Spring Harbor Symp. quant. Biol. 42, 331-350.

TON, J. T. & GONZALEZ, R. C. (1974). Pattern Recognition Principles. New York: Addison-Wesley Publishing Company.

UNWIN, P. N. & MILLIGAN, R. A. (1982). Large particles associated with the perimeter of thenuclear pore complex. .7. Cell Biol. 93, 63-75.

WIDNELL, C. & TATA, J. (1964). A procedure for the isolation of enzymatically active rat livernuclei. Biochem.J. 92, 313-317.

YOSHIDA, H. & SAGAI, T. (1972). Banding pattern analysis of polymorphic karyotype in the blackrat by a new differential staining technique. Chromosoma 37, 387-398.

(Received 12 December 1983—Accepted, in revised form, 25 April 1984)