Electron-microscope observations on the organization of the nucleus in chicken erythrocytes and a superunit thread hypothesis for chromosome structure

The switching on and off of genes in higher organisms almost certainly depends on the way in which molecules of protein and possibly RNA interact with the DNA. This activity no doubt has a structural basis. Most observations have so far been confined to condensed, inactive chromatin or mitotic chromosomes. In higher organisms the DNA is itself folded up, by combination with histones and other proteins into thread-like structures, of width approximately an order of magnitude greater than that of the DNA molecule. The folding of DNA in nucleohistone and nuclei has been studied by X-ray diffraction (Wilkins & Zubay, 1963; Luzatti & Nicolaieff, 1963; Pardon, Wilkins & Richards, 1967; Richards & Pardon, 1970; Bram & Ris, 1971; Pardon & Wilkins, 1972) and by electron microscopy (sec review by Ris & Kubai, 1970; Davies, 1968; DuPraw, 1970; Moses & Wilson, 1970; Zirkin & Wolfe, 1972; Olins & Olins, 1974). There is not yet a generally accepted model for the arrangement of the DNA in the threads, although it is thought to be supercoiled, and there is little evidence about the arrangement of the protein. The X-ray diffraction method is limited by the number of reflexions, more than an order of magnitude less than that obtained from oriented specimens of the DNA molecule itself. One problem with electron microscopy is that the preparative procedure may cause alterations in the molecular details in chromosomes, which are labile structures easily affected by the ionic environment. Indeed functioning of chromosomes and nuclei probably depends on changes in structure brought about by alterations of the ionic milieu within the cell.

Our electron-microscope observations on condensed chromatin in the intact cell (Davies & Tooze, 1966; Davies, 1968; Davies & Small, 1968; Everid, Small & Davies, 1970; Haynes & Davies, 1973) showed that the threads themselves are folded up in the interphase nucleus largely in accordance with what is to be expected from the simple physical rules which govern the behaviour of structural units which are large compared with the dimensions of atoms and molecules. The overall pattern resembles very closely that found when ball-bearings are shaken in a smooth-walled container (Bernal, 1964; Bernal & Finney, 1967) or more closely that found when spaghetti is similarly shaken. When the latter is agitated in hot gelatin in a smooth bowl and the surface examined after cooling, or sections made, the resulting appearances (H. G. Davies, unpublished observations) are similar to sections made through cell nuclei. The dot-dash pattern seen in thin sections through chromatin stained with uranyllead is shown schematically and described in Fig. 1 (p. 265). Other general principles govern the appearance of cell nuclei which are spherical or asymmetric in shape (Haynes & Davies, 1973). There is a tendency, increasing with asymmetry in any one cell type, for the outer layer of units to delaminate and form membrane-limited monolayers in the cytoplasm. The approximately constant width of these monolayers, average about 35 nm (Davies & Haynes, 1975), from a wide variety of species, plant as well as animal, provides the most convincing evidence for a common structural unit.

The threads which give rise to the dots and dashes or the e-phase are separated by regions which stain less with uranyl-lead, the o-phase (see text to Fig. 1). Previously it was supposed that threads roughly 17 nm in diameter were embedded in a matrix (Davies, 1968; Tokuyasu, Madden & Zeldis, 1968; Small & Davies, 1970). Others have not commented on this matrix and conceivably these lesser-staining areas could be overlap artifacts, similar in origin to those discussed by Robertson (1966). In this paper we show by the use of 2 different electron stains, uranyl-lead and PTA, that the stain molecules are in fact distributed between 2 phases. Staining patterns are examined in reticulocytes and erythrocytes from 17-day embryo, 4-day chicks and adult hens. Certain changes in the nucleus occur both during maturation and as a result of preparation procedures which are due to changes in molecular packing. The data suggest that the e-phase contains a higher ratio of DNA to protein than the o-phase.

There are 2 possibilities regarding the o-phase (Davies, 1968; Tokuyasu et al. 1968). Either it is a matrix material, for example a component of the nuclear sap in which the chromosomes are situated and to which they are permeable, or the o-phase may form part of the molecular structuie of chromatin. The second hypothesis (see Fig IB) envisages that chromatin contains structural units which we will call superunit threads. Superunit threads consist of a central element giving rise to the e-phase, corresponding to the dots and dashes in Fig. 1 and previously referred to simply as the unit thread, with an outer shell which constitutes the o-phase. The term superunit thread is used merely to avoid confusion and because it is larger, of greater molecular complexity, than the socalled unit thread. The observations reported here on the changes which occur when erythrocytes are disrupted support the second hypothesis. The main evidence however comes from the action of solvents on chromatin, reported elsewhere (Walmsley & Davies, 1975).

We also describe certain other ultrastructural features of the nucleus in chicken erythrocytes. These include the approximately spherical cavities in the chromatin bodies and their nearly amorphous contents, structurally differentiated regions within the chromatin bodies called cores and a complex consisting of large granules linked with an amorphous body. This complex is the main morphological entity, other than a few particles of glycogen and some scattered chromatin threads, found in the nuclear sap regions separating the chromatin bodies. These observations are relevant to the general aim of this study, namely the arrangement of nucleoprotein molecules in chromosomes. This investigation also forms the basis for the biochemical study (Walmsley & Davies, 1974) concerned with the nature and location of the proteins removed from erythrocytes when the nuclei are isolated and subsequently washed with salt solutions.

The methods used in preparing whole blood for electron microscopy have been given in detail elsewhere (Small & Davies, 1970). Briefly, after centrifugation in capillary tubes, fixation was in glutaraldehyde (3 %), cacodylate buffer (01 M), with or without calcium chloride (3mM), followed by washing and fixation in 1 % OsO₄. Following dehydration, 17-day embryo and 4-day chick cells were embedded in Araldite or Epon. Blocks of cells from 17-day embryo were kindly supplied by Dr J. V. Small.

Adult hens of Light Sussex breed were used. Adult hen erythrocytes washed in SMTOG (sucrose, 0·4 M; MgCl₂ 1 mM; Tris-buffer, 0·01 M, pH 7·5; n-octanol, 41TIM; gum arabic, 3 %) were treated once in SMTOG using high-speed rotating knives in a Virtis 45 homogenizer (Arthur H. Thomas Company, Philadelphia). This is the first step in the procedure developed by Zentgraf, Deumling & Franke (1969) for isolating erythrocyte nuclei (see Walmsley & Davies, 1975). These cells were collected to a depth of 1 mm or less, in plastic centrifuge tubes spun at 2000 g for 10 min and fixed in situ. Solutions were similar to the above except that 1 mM MgClj was used. Whole blood prepared as above using capillary tubes and the treated erythrocytes were embedded in Spurr resin (1969).

Washed adult hen erythrocytes were haemolysed (see Walmsley & Davies, 1975) in NaCl (0·14 M), phosphate buffer (0·015M> pH 7·0), saponin (0·05 % w/v). After centrifugation and washing in saline-buffer the cells were resuspended for a few minutes in a 10-fold volume of water and centrifuged in plastic tubes for 10 min at 2000 g. Preparation for electron microscopy was similar to that described for treated erythrocytes except that fixatives and wash solutions were without buffer and divalent ions.

Sections were cut with diamond knives on an A. F. Huxley–Cambridge Ultramicrotome and picked up on plastic-coated grids. Sections were stained with either 2 % aqueous magnesium uranyl acetate for 20 min at 60 °C, followed by lead citrate (Reynolds, 1963) for 5 min, or with 2% PTA made up in 90% ethanol, 10% water, for 30 min, at room temperature. After depositing a coat of carbon on their surface they were examined in a Siemens Elmiskop 1 at 100 kV with a 200-μm condenser aperture and a 50-μm objective aperture. Electron intensities were measured with the photometer in approximately 0· 25-μm diameter spots corresponding to the small circular area in the fluorescent screen at x 40000. Electron densities were calculated as the logarithm of the ratio of the intensities transmitted by clear resin and specimen and expressed per unit thickness (100 nm) of section. The thickness of the section was calculated by the fold-method (Small, 1968). The electron density of resin relative to the support film was also measured and used to calculate thickness when no fold was available. The electron density per 100 nm of resin was roughly 0·18. Irradiation intensities were kept low so as to minimize variability caused by evaporation of the resin, the bulk of which seemed to occur fairly quickly before measurements were made.

A comparison of the staining properties of cell nuclei with the 2 different stains, uranyl-lead and PTA, would present no problem if there was no variability among nuclei. However, there is variability in uptake of stain and in the visibility of the dot-dash pattern. For example, the uptake of PTA stain varied in different parts of the same section taken from a block of 17-day embryo erythrocytes. Hence in order to make a valid comparison it was essential to cut pairs of, preferably serial, sections and examine the same nucleus stained with the 2 different stains. This proved simple to do.

Previous light- and electron-microscope observations on nucleated erythrocytes from amphibians, birds, fish and cyclostomes (Davies, 1961, 1968; Fawcett & Witebsky, 1964; Tooze & Davies, 1967) established that the bulk of the deoxyribo-nucleoprotein of the nucleus resides in discrete chromatin bodies, the condensed chromatin. Rough counts of the numbers of chromatin bodies in Amphiuma (Anderson & Norris, 1960) and in Rana pipiens (Davies, 1961) established that their number is approximately the same as the chromosome number. In the interchromatin region the main component of the nuclear sap is haemoglobin found at a similar concentration to that of the cytoplasm. Our findings are shown schematically in Fig. 1 A.

The ultrastructure of uranyl-lead stained sections through erythrocytes from 17-day chick embryos has been described earlier (Davies & Small, 1968). The dot-dash pattern is clearly visible throughout the bulk of the chromatin and the odd bands hi, bj are of similar electron density and width. Figs. 3, 5 and 7 are at too low a magnification to show this clearly but the point is illustrated in Fig. 22. The dense band 62, consisting of dots and dashes, often extends around most of the perimeter of the nucleus.

In sections from one block groups of cells in different areas showed a variability in PTA stain uptake. In one area, α (Figs. 2, 4), the chromatin stained much less than the cytoplasm apart from an intensely staining zone around the entire periphery of the nucleus, about the same electron density as the cytoplasm and 10-20 nm wide, that is occupying b1 and part of b2. In an adjacent area, γ (Figs. 2, 8), chromatin stained on average about the same as the cytoplasm, or sometimes more, and the density of the narrow peripheral zone was only slightly greater, if at all, than the remainder of the chromatin. In the intermediate area, β (Figs. 2, 6), between α and γ, a denser zone similar to that seen in cells in the area α, characteristically extended around the entire periphery of each nuclear or chromatin body. Figs. 3, 5, 7 are uranyl-lead stained sections through the same nuclei as Figs. 4, 6, 8, respectively, and they show a dot-dash pattern due to a relatively lower stain uptake by the o-phase compared with the e-phase. The electron densities of about 5 nuclei in each one of a pair of sections, involving 10 measurements in areas α and similar data for areas β and γ are shown in Fig. 2. The average electron density of the chromatin stained with uranyl-lead increases in going from area y to a whereas there is a decrease in the average density of the same nuclei stained with PTA. After staining with PTA the dot-dash pattern could not be seen in chromatin in area a, and appeared, if at all, with low visibility in areas x and y. This shows that the uptake of PTA is similar in both o- and e-phases. Even when the uptake of PTA is very high as in the dense peripheral zones in Figs. 4 and 6 and when the width of this zone approaches bi plus 62, the stain uptake by these 2 bands is very similar and no dot-dash pattern can be seen.

After either uranyl-lead, or PTA, and in all the cells, the electron density of the nuclear sap was very similar to that of the cytoplasm, in agreement with earlier observations on amphibian erythrocytes (Tooze & Davies, 1967) showing a similar concentration of haemoglobin in the 2 spaces.

The circulating blood from 4-day chicks contains mainly erythrocytes but also some reticulocytes characterized by relatively large numbers of cytoplasmic ribosomes in the form of polysomes and a relatively spherical nucleus (Fig. 9). After staining with uranyl-lead the dot-dash pattern in the chromatin is often visible in both types of cell but in the reticulocyte it can be seen more clearly presumably because the dots and dashes are slightly farther apart. The electron density of chromatin per 100 nm is less in reticulocytes, about 0·24 compared with 0·26 in the mature cell (Fig. 2). Also the interior outline of the chromatin bodies is less well defined in reticulocytes presumably due to incomplete condensation. In the micrograph (Fig. 9) the density of the haemoglobin in the cytoplasm of the mature cell is greater than that in the reticulocyte but the measured electron densities are similar (Fig. 2) presumably because the uranyl-lead stained ribosomes are included in the measuring area.

After staining in PTA (Fig. 10) the dot-dash pattern cannot be distinguished in the chromatin of either reticulocyte or mature cell. Now, however, the electron density of the chromatin (Fig. 2) is greater in the reticulocyte than in the mature cell. The electron density of chromatin in the mature cell is not much above that found in the unstained cell, apart from a very intense zone of stain extending around the periphery of the nucleus, width about 10–20 nm as in Fig. 4. A similar relatively denser zone of stain is found in the reticulocyte. After PTA the electron density of the cytoplasm is on average somewhat greater in the mature cell than in the reticulocyte (Fig. 2), consistent with there being a higher concentration of haemoglobin. Also the ribosomes in the reticulocyte (Fig. 10) cannot be distinguished, due presumably to their being stained to roughly the same extent as the haemoglobin.

After staining in uranyl-lead, the dot-dash pattern in the chromatin of erythrocytes, and sometimes reticulocytes, from the adult hen is often difficult to distinguish (Figs. 12, 16, 17) but it becomes visible in what we assume are cells which have become slightly swollen during fixation (Fig. 25). Characteristically the b1 layer is less dense than the 62 (Fig. 12). After staining in PTA the uptake throughout most of the chromatin is negligible apart from a very intense zone of stain at the periphery of the nucleus, about 10–20 nm wide, corresponding to the b1 and part of the b2 band (^Fig-H)-

Our observations on 17-day embryos and 4-day chicks were limited to a few animals. However blood from many adult hens was examined, including samples from blood used in biochemical experiments reported elsewhere (Walmsley & Davies, 1975). During these studies it became apparent that there are other ultrastructural features of erythrocyte nuclei, not previously described. The extent to which these structures are common to other somatic cell nuclei has yet to be investigated. After uranyl-lead staining numerous small roughly circular areas, up to about 0·15 μm in diameter, are seen (Fig. 12) scattered throughout the condensed chromatin, near the periphery as well as in the central regions of the nucleus. Similar areas occur in the chromatin of 4-day chicks (Fig. 11) and in 17-day embryo. These correspond, we suppose, to sections through roughly spherical cavities which sometimes fuse and give rise to elliptical areas in thin sections. In blood samples from 14 animals the cavities occurred with variable frequency. In 5 animals there were few cavities in the chromatin of erythrocytes. In the other 9 animals counts showed that about 10-20% of the cells in each chicken contained chromatin with appreciable numbers of cavities. The cavities contained material which stained either about equally to, or sometimes more than, the nuclear sap. Sometimes there was lesser-staining material separating the cavity contents from the surrounding chromatin (Fig. 13). Preliminary results show that the material in the cavities stains positively using the Bernhard (1969) EDTA technique for preferentially staining RNA, but, as Berhard has pointed out, this evidence is not conclusive.

Scattered throughout the nuclear sap areas in variable amounts are granules and/or, sections through threads, some of which may consist of nucleoprotein. Typically there are one or two very striking groups of large granules up to about 30 nm in diameter and heterogeneous in structure (Figs. 17, 18). Some’ granules’ may be sections through threads and some seem to be linked by fine threads. This complex of large granules appeared with the frequency that nucleoli are seen in many somatic cells. In haemolysed cells the complex regularly included an amorphous body (Walmsley & Davies, 1975), but this could be seen only rarely (Fig. 17) in the intact cell.

In 4 of the animals an appreciable number of the erythrocytes had membrane-limited regions within the nucleus containing material assumed to be haemoglobin from its homogeneous fine structure and electron density (Fig. 12). Such regions were free of particulate material but occasionally contained additional membrane-limited regions. These enclosures may originate in the nucleus or they may enter from the cytoplasm. Only once, a configuration suggesting the latter possibility was seen (Fig. 12). However, membrane-enclosed areas containing material staining like haemoglobin are found in the cytoplasm, and supposedly arise from organelles such as mitochondria which are breaking down. They may represent the end-products of lysosomal activity (Tooze & Davies, 1967). The membranous material encircling the homogeneous region in the nucleus was complex, not a single membrane. The significance of these membrane-limited regions is not yet understood.

In erythrocytes from many animals there were a few particles in the nuclear sap which stained very intensely after uranyl-lead. In many cells of one animal (Fig. 16) such particles were particularly numerous and were largely, but not entirely, confined to the nucleus. Their staining properties with lead alone (Fawcett, 1966a) and their particulate dot-like substructure seen after uranyl-lead (Fig. 15) suggested they might be glycogen (Drochmans, 1962).

Lucas & Jamroz (1961) in their light-microscope studies did not find a nucleolus in erythrocytes from adult hens. We observed, but rarely, a small spherical finely-fibrillar body, which by comparison with other studies on developing erythroblasts (Small & Davies, 1972) might correspond to a nucleolus.

After treatment with the Virtis homogenizer in SMTOG (see Methods) i-/tm sections of embedded material when examined in the light-microscope showed, in one series of counts, 45 % of the cells to be still intact and the remaining 55 % to be disrupted. Cells are defined as disrupted if they are wholly or partially haemolysed.

In about half of the disrupted cells the chromatin bodies remained condensed (Fig. 32). In the other half the chromatin bodies themselves had more or less expanded, or disrupted. Sometimes the now-separated threads comprising chromatin had spilled out into the cytoplasmic space as in Fig. 29.

It was easy to understand why a proportion of the cells remained intact. But it was puzzling why some of the chromatin bodies were disrupting, since a concentration of 1 mM magnesium ions is known (seeWalmsleySt Davies, 1975) to be sufficient to keep chromatin in a more-or-less condensed state. The explanation may be as follows. Chromatin is, we assume, kept condensed by a condensing factor. If, as a result of cell damage haemoglobin is removed from the cell the condensing factor may also diffuse out. If the rate of diffusion of magnesium ions inwards is not sufficiently rapid then the extension or disruption of chromatin will necessarily ensue and may reach a stage where the chromatin is irreversibly extended and cannot be recondensed when the magnesium ions, at a concentration of 1 mM, reach the chromatin.

When condensed nuclei in disrupted cells were examined after staining with uranyl-lead the dot-dash pattern in the chromatin bodies, with lesser-staining intermediate areas, normally difficult to sec in mature erythrocytes from adult hens, was usually very clearly seen and resembled the pattern normally found in reticulocytcs from 4-day chicks (Fig. 9), or mature erythrocytes from 17-day embryos. Fig. 19 is a good example of a nucleus in which the units are well lined up around the periphery and in which it is possible to believe that similar units occur throughout the bulk of the chromatin, apart from certain regions. In these regions referred to as cores, because they are generally more-or-less surrounded by the chromatin exhibiting the dot-dash pattern, the stain is more homogeneously distributed and each core usually contains one or more cavities within it (see also Fig. 30). Cores were very common, but we have not yet ascertained whether each chromatin body contains a core. Cores can also be distinguished in the chiomatin bodies in intact reticulocytes from 4-day chicks. Normally the cavities scattered throughout the condensed chromatin bodies also retained their contents in the disrupted cells. When the haemoglobin was absent from the cytoplasm, the haemoglobin had also left the nuclear sap areas, which is consistent with the idea that the spaces are continuous. Frequently the outer membrane of the nuclear envelope had lifted away (Fig. 19). In these Virtis-SMTOG treated cells, the bi band stained either less than or similarly to the bj band (examine Figs. 19, 20).

Pairs of serial sections from the same condensed nuclei in disrupted cells were stained in either uranyl-lead or PTA. In many of these nuclei the chromatin bodies were stained homogeneously by PTA but the same nucleus after uranyl-lead exhibited a dot-dash pattern (compare Figs. 20,21). When stained with uranyl-lead the electron density per unit thickness was similar to that in the intact cells from 17-day embryos and 4-day chicks (Fig. 2). However the uptake of PTA was considerably higher than in those cells and the resulting electron density similar to that found in the serial section when stained with uranyl-lead (see data on adult in Fig. 2). In the PTA stained nuclei there was a somewhat denser zone extending around the periphery of the nucleus, width roughly 10–20 nm (Fig. 21). Frequently a similar zone was found around the entire chromatin body giving rise to an appearance like that shown in Fig. 6.

In cells in which the chromatin bodies themselves are disrupting (e.g. Fig. 29) the organization into staining and lesser staining regions is largely destroyed. The separated dots and dashes which are now visible (Figs. 24, 26) have a width of roughly 25–30 nm. After staining in either uranyl-lead or PTA they appear more-or-less uniform in section, but occasionally have a reduced stain uptake in the outer regions. Near the nuclear envelope (Figs. 24, 26) the organization into layers is still retained in places, but the separation of the 62, 64 layers (opposite arrow in Fig. 24) is nearly twice as large as in the untreated cell (Fig. 25). It would seem logical to compare these micrographs with that of an intact cell from adult hen. However, as was pointed out above it is difficult to see the units in the well fixed intact cell. Instead, therefore, Fig. 25 was chosen and this shows an untreated erythrocyte from adult hen in which the units have become more visible as a result of changes during fixation. Its appearance is similar to that normally found in other cell types, for example the erythrocytes from 17-day embryo (Fig. 22). In the lower half the width of band 62 is about 17 nm but it has widened to 20 nm in the upper half due to further changes. These dimensions are appreciably less than the 25–30 nm found for the separated threads. In the Virtis-treated erythrocytes there were all stages of cellular and nuclear disruption. In the cell in Fig. 23, the chromatin bodies are still recognizable at low power and the units can be seen to be more closely packed when compared with Figs. 24, 26. However, due to various changes, the visibility of the units is very high and their arrangement into 2 or more layers at the periphery of the nucleus is very clearly revealed.

In certain nuclei in which the chromatin bodies are either intact or, more commonly, partly disrupting, 2 sorts of spherical bodies are found, both with staining properties similar to the contents of the cavities in condensed chromatin. First there are the intrachromatin body spheres, up to about 0-15 μm in diameter (Fig. 33) and it seems likely, on account of their size and location, that these correspond to the contents of the spherical cavities in chromatin. Walmsley & Davies (1975) show by microspectrophotometry on isolated nuclei that these cavities are unlikely to contain haemoglobin. Secondly, there are interchromatin body spheres (Fig. 31) and these may be larger, up to 0·4 μm; compare with the intrachromatin bodies in Fig. 32 printed at the same magnification. The composition of these larger spheres is unknown but they may contain haemoglobin. Studies on coelomocytes in a polychaete worm Terebella (H. G. Davies, unpublished observations) indicate that haemoglobin itself accumulates in spherical bodies in the cytoplasm and nucleus. During maturation of the coelomocytes these spheres apparently merge together to give a homogeneous appearance, like that of the cytoplasm and nucleoplasm in the intact chicken erythrocyte. Conceivably haemoglobin in the form of packed spheres may be present in erythrocytes and the spheres may only become visible during cellular disruption. Sometimes spherical bodies about the size normally found in the cavities within the chromatin can be seen lying in the spaces between the chromatin bodies and apparently escaping from the cell (Figs. 34, 35).

Another observation was made on the Virtis-treated cells which suggests that the cores are differentially stable during nuclear disruption and therefore holds out some promise that it may be possible to isolate them for chemical analysis. In those cells in which the chromatin bodies were dispersing (Fig. 29) it was common to see small regions of condensed chromatin. Occasionally, judging by their easily distinguished dot-dash pattern and variable size (see region to the right on Fig. 27) some were areas of chromatin late in expanding. However many small condensed areas resemble the cores in their size and shape and finely particulate substructure (Fig. 27). The contents of the cavity in the core shown in Fig. 28 are clearly visible.

When erythrocytes were haemolysed in a saponin (0·05 %), saline (0-14 M) solution, Brasch, Seligy & Setterfield (1971) reported that the intact chromatin bodies consisted of threads about 15–25 nm in diameter. Our observations on these nuclei are described in Walmsley & Davies (1975). When such nuclei were disrupted in water, Brasch et al. (1971) showed that these threads unfolded into finer threads about 5–10 nm in diameter. We repeated this experiment and found that after very brief treatment in water, disrupted nuclei fell into 2 classes (Fig. 37). In one the threads had a diameter of roughly 5–10 nm, similar to those reported by Brasch et al. (1971), as well as some finer threads, in size down to the limits of visibility. In the other class of disrupted nuclei (Fig. 36) the threads were larger, similar to those reported here after mechanical disruption of the cell in SMTOG, about 25–30 nm when stained with uranyl-lead or PTA. Frequently after uranyl-lead they had a lesser-staining outer region with a rather irregular shape.

The observations reported in this paper firmly establish the existence of 2 phases in the condensed chromatin of chicken erythrocytes. Since, after uranyl-lead staining, the same dot-dash pattern with lesser-staining intermediate areas is found in chromatin from tissues other than blood (Davies & Haynes, in preparation) this property is likely to be a general one. The dot-dash pattern arises from sections through threads and it was conceivable that the appearance of lesser-staining areas between the dots and dashes was an overlap artifact (see for example Robertson, 1966). If the lower staining was purely a geometric effect and there was no substance between the threads, then it would be expected that a different stain, with a greater or lesser binding, would lead to a change in the scale of the electron density profile through chromatin but not to a change in the shape. However, the even distribution of stain after treatment with PTA, compared with the dot-dash pattern after uranyl-lead (Figs. 20, 21) shows that the shape has changed and hence we conclude that the threads are actually separated by a substance, the o-phase. Near the surface of the nucleus where the units are aligned into layers or laminae, the curvature is small in thin sections and the appearance of uranyl-lead stain in the odd bands is very unlikely to be an artifact. However, it could be argued that near the envelope there are additional molecules concerned with the attachment of the chromosomes to the membranes. Our results show that the bulk of the condensed chromatin in erythrocytes consists of 2 phases.

In the intact cells (Figs. 4, 6, 10), the rather low PTA uptake makes it difficult to be absolutely sure that the shape of the electron density profile has changed, rather than merely the scale. But even there in the narrow zone with high uptake at the periphery of the nucleus, approaching bi and 62 in width in places, there is no sign of a dot-dash pattern after PTA stain (see also Fig. 14). However, due to the high stain uptake in the condensed nuclei in haemolysed cells (Fig. 21) it is possible, by visual inspection only, to be quite sure that the stain is homogeneously distributed. Since the average electron densities of sections through the same nucleus have similar high values with the 2 stains and since the eye can be relied upon to judge 2 areas as equal in intensity, a fact made use of in the construction of photometers, our conclusion that chromatin contains 2 phases is independent of possible uncertainty due to the well known problems of photographic reproduction.

In this section we offer an explanation for the changes in stain uptake during the late stages of erythrocyte maturation (Figs. 2, 9, 10) and the variations in PTA uptake found in the different areas of a section through 17-day embryo erythrocytes (Figs. 2, 4, 6, 8), as well as the relatively dense zone of stain found at the inner membrane of the nuclear envelope (Figs. 4, 6, 10, 14).

The number N, of stain atoms, or molecules, taken up per unit volume is equal to the product of 3 values, the number n of molecules per unit volume, the maximum number 5 of possible binding sites per molecule and the fraction / of binding sites which are available. That is, N = nsf. Further, the number N is proportional to the electron density per unit thickness of section. We define s as the maximum number of binding sites when the molecule is extended and separate: s depends on the nature of the molecule and the stain. The number of binding sites available is reduced by a fraction, /, when the molecule is folded up or when the concentration of molecules, n, increases leading to lower binding for steric reasons. Clearly the value of / will depend also on the molecular weight of the stain. When N changes it may not be possible to decide which of the factors n and f are involved.

During the maturation of the chicken erythrocyte, light-microscope observations by Lucas & Jamroz (1961) showed that there is a decrease in the volume of the nucleus. The amount of DNA per nucleus is unaltered and hence the concentration of DNA and presumably nucleoprotein, can be expected to increase. Consistent with this increase in n, the data in Fig. 2 show that the uptake of uranyl-lead is greater in the chromatin of the mature erythrocyte than in the reticulocyte. However, the uptake of PTA in the bulk of the chromatin is much less in the mature cell than in the reticulocyte. Hence the product nsf is less and since n is increasing this must be due to a decrease in the value off when PTA is used. Presumably, due to closer packing, fewer sites are available to this stain. This decreased availability to PTA compared with uranyl-lead is consistent with their relative molecular weights, about 3000 for PTA and about 270 for the uranyl ion.

A reduction in the value of/when PTA is used also provides an explanation for the staining patterns found in the chromatin bodies in the erythrocytes from 17-day embryos (Figs. 4, 6, 8). In the region γ (Fig. 8) the chromatin bodies are stained to about the same extent throughout. In region β (Fig. 6) a decrease in the value of/has occurred within the chromatin bodies due presumably to closer packing, apart from a narrow zone extending around the entire periphery of each chromatin body. In the region α (Fig. 4) the value of/has been further reduced throughout the entire chromatin body apart from the zone adjacent to the nuclear envelope where / remains high. This decrease in PTA binding is again accompanied by an increase in uranyllead binding (Fig. 2). In this case the alterations in volume are probably brought about by the preparation procedures used in electron microscopy. A similar explanation, involving variations in /, applies to the dense zones adjacent to the nuclear envelope seen in reticulocytes and erythrocytes from 4-day chick (Fig. 10) and adult hen (Fig. 14).

The pattern in Fig. 6, where the dense zone surrounds each chromatin body shows that there is no special substance with a high staining capacity for PTA adjacent to the nuclear envelope. It disposes of any possible suggestion that this peripheral dense zone has anything to do with the fibrous lamina, lamina densa, or zonum nucleum limitans (Fawcett, 19666; Patrizi & Poger, 1967; Kalifat, Bouteille & Delarue, 1967; Stelly, Stevens & André, 1970). This interesting differentiated region inside the nucleus near the envelope is much wider than the bi plus 62 band and occurs in certain special cells only.

In most cell types, other than the erythrocytes of the adult hen, after staining with uranyl-lead both the widths and electron densities of the bands bj, bj are approximately equal (Fig. 22). We propose as outlined in the Introduction and Fig. IB that chromatin consists of folded superunit threads. These threads have a central element about 17 nm in diameter which stains intensely with uranyl-lead and forms the e-phase and an outer shell which] stains less densely with uranyllead and which forms the o-phase. The staining patterns may be explained in terms of these units as follows. We suppose that the cylindrical units are packed hexagonally and the rare micrographs of end-on views of oriented layers of units at the surface of the nucleus support this mode. If we assume (see also later) an outer diameter of 28 nm for the superunit thread and 17 nm for the inner element then a hexagonal array (Fig. IB) leads to both widths and electron densities of bi as about 0·7 of b3. In electron micrographs it would be difficult to distinguish between the factor 0·7 and the factor 1·0 which latter corresponds to equal electron densities and widths. In the nuclei stained with PTA the dense zone has a width equal to b1 and part of b2. As already explained the appearance of a dense zone is the result of decreased staining due to closer packing within the chromatin body. Contact of the outer layer of super-units with the inner membrane of the envelope apparently inhibits a packing change in part of the outer shell and inner region.

In the mature erythrocyte of adult hen there are several differences in the staining pattern after uranyl-lead, when compared with immature erythrocytes and cells from other tissues. There is a decrease in the visibility of the dot-dash pattern, the chromatin staining nearly the same throughout except that band bi stains appreciably less (Fig. 12). We assume that there is a further increase in the closeness of packing of the molecules in chromatin, somewhat greater in the o-phase, except in the region at the nuclear envelope. This would reduce the visibility of the dot-dash pattern seen in nuclei stained in uranyl-lead and lead to the very low, or negligible uptake of PTA observed in the mature cell (Fig. 2) due to a further reduction in the value of f compared with other cells. These suggestions appear reasonable because in another nucleoprotein system, namely the nuclei in certain maturing spermatozoa (Walker & Macgregor, 1968; Walker, 1971; Henley, 1973) similar phenomena are well documented. In the sperm heads a lamellar or banded appearance after uranyl-lead progressively gives way, during maturation, to a homogeneous one. These lamellae are also thought to be due to a side-by-side packing of nucleoprotein threads (Gall & Bjork, 1958).

How are the proteins and the nucleic acid distributed between the 2 phases and hence within the hypothetical superunit thread? There are several difficulties, discussed below, in giving an unequivocal answer to this problem due to a lack of specificity in the stains, and to the fact that the uptake of the PTA is particularly sensitive to the configuration or closeness of packing of the molecules. Uranyl acetate and uranyl followed by lead are known to stain preferentially nucleic acid in thin sections through embedded biological material (Huxley & Zubay, 1961). These workers also measured the uptake of uranyl ions in aqueous solutions. DNA takes up an amount of stain almost equal to its own dry weight. Nucleohistone took up 50 % of its dry weight and purified histone only about 20 %. Probably the monovalent ion UO₂(Ac)+, binds to the phosphate groups of the DNA. Since the almost electrically neutral molecule nucleohistone contains about equal proportions of DNA and protein the above data, showing a 50 % uptake of uranyl acetate, indicate that the combination of histone with DNA has very little effect on the binding of uranyl to the DNA. Our data on uranyl staining coupled with lead (Figs. 2, 3, 5, 7) lead to similar conclusions. After uranyl-lead the electron density of chromatin is about 2·5 times greater than that of the cytoplasm. This is not due just to a high concentration of protein in the chromatin since interferometric observations have shown (Davies, 1961) that the total concentration of molecules is very similar in the 2 compartments. Evidently uranyllead does stain DNA preferentially but it is not entirely specific; it also stains protein, although to a considerably lesser extent. The stain PTA, first introduced by Schmitt, Hall & Jakus (1942) has been used in aqueous solutions at different pH levels, in ethanolic solutions and for staining blocks prior to embedding as well as on thin sections. Test-tube experiments in aqueous acidic solutions show that it combines with certain proteins but not with DNA (Silverman & Glick, 1969). The general conclusion is that PTA probably stains the positively charged groups in proteins (Hodge & Schmitt, 1960; Sheridan & Barrnett, 1969). This is consistent with the negative charge on the PTA ion. In our experiments uptake by sections in aqueous solution at room temperature was not sufficiently high to be useful. When sections are treated with ethanolic solutions one cannot be sure that the DNA is unstained but this seems likely. Comparing Figs. 7 and 8 it can be seen that PTA, unlike uranyl-lead, is a good stain for both nucleoprotein, presumably the protein moiety, and the cytoplasmic protein haemoglobin. But PTA becomes less effective in staining nucleoprotein when the molecules pack close together as in Figs. 4 and 14, for example. If we assume that the local values of/in the 2 phases, e and 0, are the same for any one stain, then the interpretation of the 2 patterns, the dot-dash after uranyl-lead and the homogeneous appearance after PTA (Figs. 20, 21) is as follows. If the protein to DNA ratio in the 2 phases is the same, then the shapes of the electron-density profiles for all stains would have to be the same. This is not so and hence the ratios of protein to DNA must be different in the 2 phases. The even-staining of the 2 phases with PTA shows that the concentration of protein in the o-phase is roughly the same as that in the e-phase assuming that the DNA remains unstained (Silverman & Glick, 1969). The combination of uranyl ions with DNA is not much affected when the DNA is combined with histone as nucleohistone (Huxley & Zubay, 1961). We must also assume that, similarly, the presence of the DNA does not affect the binding of the PTA to histone. The unequal staining with uranyl-lead shows that the e-phase contains a higher concentration of nucleic acid. Hence we conclude that the c-phase has a higher DNA to protein ratio than does the o-phase, which may or may not contain DNA. Also our data indicate that the total concentration of matter is higher in the e-phase. This conclusion regarding the molecular composition of the 2 phases cannot be regarded as well established due to uncertainties about the values of/in the e- and o-phases when PTA is used. This point needs further study. Specifically we need to learn if the same nucleoprotein thread could be folded up in the e-phase in a more compact form so as to limit PTA binding to a greater extent than occurs in the o-phase. The following argument suggests that this is improbable. If the percentage composition of the 2 phases is the same then there must be a much higher concentration in the e-phase to explain the uranyl-lead pattern. Over a range of PTA stain uptake, shown in Figs. 4–8 and Fig. 21, which corresponds to a change in the average value of f, the even-staining pattern is approximately maintained. Hence the individual values of f, which depend on packing and concentration, for each phase must vary in a similar manner. This is more likely if the packing, in so far as it affects f, is similar in the 2 phases: this can hardly be the same if the staining patterns are due to the same thread folded differently in the 2 phases.

In the intact cell the eye selects out the uranyl-lead dense regions, width roughly about 17 nm. However when the threads separate their total extent becomes visible and they are seen to be about 25–30 nm wide. These data support the concept of the superunit thread. Unfortunately, they do not provide a convincing proof because, conceivably, separation of the units comprising chromatin could be accompanied by a loss of the o-phase coupled with a shortening and thickening of the 17-nm-wide regions. Walmsley & Davies (1975) provide more convincing evidence for the super-unit thread hypothesis by showing that the o-phase largely survives both isolation of the nucleus and prolonged washing in solvents known to remove loosely bound material. It might be expected that the separated superunits would have a lesser-staining outer region after uranyl-lead staining. This is in fact sometimes seen after hypotonic disruption (Fig. 36). It is less apparent after the brief Virtis treatment in SMTOG. The exact nature of the changes occurring during disruption are not easy to document. Quite likely there are conformational changes leading to a redistribution of matter. Indeed conformational changes which lead to smaller threads, width about 5–10 nm, do take place. A further reason for postulating structural changes during separation is that the hollow-end-on views previously reported (Davies, 1968) in the unit threads in intact cells are rarely seen in the disrupted threads.

When erythrocyte nuclei are spread on an air-water interface (Gall, 1966; Ris & Kubai, 1970) the width of the threads is about 20–30 nm. Brasch et al. (1971) found that the dimensions of the threads when erythrocyte nuclei were disrupted in water are 5–10 nm. At first this was puzzling, but our results, repeating the experiments of Brasch et al. (1971) indicate that the larger thread, stable on a water surface, is an intermediate and unstable form in water. The dimensions of the spread fibres, 20–30 nm, found by others is in reasonable agreement with the dimensions 25–30 nm, found by us for the separated threads. The superunit thread hypothesis therefore helps to explain the apparent discrepancy between the dimensions of the uranyl-lead dense regions seen in intact cells, about 17 nm, and the larger threads in spread preparations (see also the discussion in Davies, 1968; Tokuyasu et al. 1968; Everid et al. 1970).

A notable feature of the ultrastructure of the nuclear envelope-limited sheets of chromatin threads is their trilaminar construction (see Fig. 114 in Davies, 1968), corresponding to bands bi, b2, bj. This geometry is in agreement with the suggestion that they are a monolayer of superunit threads (Fig. 1 c). Their width (s’ in Fig. 1 c) measured between the centres of the limiting membranes is 35·0 nm, the average measurement from a large number of different species (Davies & Haynes, 1975). If an allowance of 3·5 nm is made for each half-width of membrane, then the width of the superunit is 28 nm. This dimension agrees well with the value of 28 nm given for the centre-to-centre spacing of the units in the interior of the nuclei of erythrocytes from 17-day embryo (Davies & Small, 1968).

Further observations on the cavities in the chromatin bodies and their contents are given elsewhere (Walmsley & Davies, 1975). Cores can be distinguished in reticulocytes from 4-day chick and in mature erythrocytes from 17-day embryos by their more homogeneous structure compared with the dot-dash pattern in the surrounding chromatin. In mature erythrocytes from adult hens the dot-dash pattern is itself difficult to distinguish as noted here and elsewhere (Everid et al. 1970) and hence the structural differentiation in the chromatin body cannot be seen. In the mature erythrocytes there is an additional condensation of the already condensed chromatin, similar to that already mentioned in maturing spermatozoa. One possibility therefore is that the cores in chromatin bodies merely represent regions which contain the same structural units, the superunits, but represent the initial site of additional condensation. A more interesting possibility is that the DNA is either differently folded in these regions and, or, has different proteins associated with it. There is no doubt that these regions tend to be more stable during nuclear disruption. A differentiation along the metaphase chromosome is now well established (see reviews in Caspersson & Zech, 1973), but its physicochemical basis, whether a different folding pattern or composition, or both, is not so well understood. Theie may be a relationship between the cores in the interphase nucleus and the new patterns into which the nucleoprotein is refolded during cell division.

We thank Professor M. H. F. Wilkins, F.R.S., for his continued interest, Mrs Yvonne Buchner for preparing the material for electron microscopy, Mr D. Back for technical assistance and Mrs F. Collier and Mr Z. Gabor for help in preparing the plates. M. E. Walmsley received an M.R.C. Junior Research Fellowship and A. B. Murray is an MRC Scholar.