A comprehensive study on the isolation and characterization of the hela S3 nuclear matrix

The nuclear matrix is defined as the salt-resistant proteinaceous nuclear structure that is isolated from the interphase cell (Berezney and Coffey, 1974, 1977; Shaper et al. 1979; Agutter and Richardson, 1980; Bouteille et al. 1983; Hancock, 1982; Berezney, 1984). It typically consists of residual components of the nuclear envelope, nucleoli, and a fibrogranular internal matrix. Numerous studies from a variety of different laboratories have lead to the conclusion that the nuclear matrix is involved in chromatin organization, DNA replication, and RNA transcription, processing and transport (Berezney, 1984; Jackson et al. 1984; Agutter, 1985; Nelson et al. 1986; Razin, 1987; Verheijen et al. 1988; Bodnar, 1988).

There remains controversy with regard to factors responsible for maintaining the structural integrity of the nuclear matrix. Sulfhydryl oxidation during the isolation of nuclei and subsequent nuclear matrices has been proposed to be responsible for inducing the internal matrix structure (Kaufman et al. 1981; Kaufman and Shaper, 1984). These conclusions are compromised, however, since the reported studies utilizing sulfhydryl alkylating agents (iodoacetamide and n-ethylmaleimide) and sulfhydryl reducing agents (dithiothreitol and β-mercaptoethanol) in the matrix preparations included RNase A digestion, which also may destabilize the matrix structure (Adolph, 1980; Kaufman et al. 1981; Long and Schrier, 1983; Bouvier et al. 1984; Fey et al. 1986). However, van Eekelen et al. (1982) could not detect any type of an RNase effect on the ultrastructure of matrices isolated from HeLa cells, and found no morphological differences between matrices isolated in the presence or absence of iodoacetamide, despite the inclusion of RNase A. Hodge et al. (1977) and Dijkwel and Wenink (1986) omitted RNase in their matrix preparations and found that dithiothreitol had no effect. There is need, therefore, to examine the ‘disulfide stabilization’ effect in the absence of RNase A pretreatment.

Several agents have been used to extract the histones and other soluble components of nuclei during isolation of the residual nuclear matrix. The extracting agents that have been commonly utilized are NaCl (Berezney and Coffey, 1974, 1977; Kaufman et al. 1981; Bouvier et al. 1984; Adolph, 1980; Hodge et al. 1977), (NH^SC^. (Capeo et al. 1982; van Eekelen et al. 1982), EDTA (2 HIM) (Long and Ochs, 1983), and LIS (lithium 3,5-diiodosalicyalate; Mirkovitch et al. 1984). The use of LIS as the extracting agent requires that the nuclei be stabilized by 37 °C incubation or with divalent cations such as Ca²⁺ or Cu²⁺ (Mirkovitch et.al. 1984). This poses another problem, since evidence suggested that the stabilization may cause proteins to be found associated with the matrix that were not found associated if the stabilization was omitted (Evans and Hancock, 1985; Humphrey and Pigiet, 1987; Fields et al. 1988; McConnell et al. 1987).

It is difficult to come to any major conclusions on the previous studies, since each group of investigators has different methods for isolating nuclei and nuclear matrices, in addition to the type of extracting agent used. This introduces other variables that could affect the results and the comparison of different studies. Moreover, the studies to date have been incomplete and sometimes contradictory. In an attempt to resolve some of these fundamental issues concerning the isolation of nuclear matrix and its polypeptide compositon, many previously examined parameters are studied in more detail and in a single mammalian cell system: the HeLa S3 cell. Ultrastructure, biochemical composition and polypeptide composition analysis are used to determine the role of RNA and disulfide bonds in maintaining structural integrity, the effects of the various extracting agents, and the effects of stabilization.

HeLa S3 cells were grown as suspension cultures in Joklik’s modified minimum essential medium supplemented with 10% horse serum, to a density of 1.0×10⁶ cells ml⁻¹.

HeLa cells were washed twice with ice-cold PBS. The cell pellet was warmed at room temperature for 45 s. It was then resuspended in 10 mM Tris–HCl, pH 7.4, 2ma MgCl₂, 0.5 mM phenylmethylsulfonyl fluoride (TM-2 buffer) at 10°C to a density of 10⁷ cells ml^-1 and incubated at room temperature for Imin. The cells were incubated in ice-water for 5 min. Triton X-100 was added to 0.5% (v/v), and the cells were incubated in ice-water for an additional 5 min. Cells were sheared by three passes through a 22 gauge needle and nuclei were separated from the cytosol by centrifugation at 800 revs min^-1 for 6 min at 4 °C CHS-4 rotor, Sorvall-Dupont Instruments). Nuclei were washed twice in TM-2 buffer and resuspended to a DNA concentration of approximately 1 mg ml⁻¹.

MgCl₂ was added to the nuclei in TM-2 buffer to a final concentration of 5 mM. The nuclei were digested with DNase I at 30 i.u. mg⁻¹ DNA (Cooper Biomedical) for 30 min at 4°C. An equal volume of (NHilïSOi or NaCl in 10 mM Tris-HCl, pH 7.4, 0.2 mM MgCl₂ (TM-0.2) was slowly added to a final ionic strength of at least 0.6 (0.2 M (NH₄)₂SO₄ or 0.6 M NaCl). The total volume was brought to 40 ml by adding TM-0.2 containing the appropriate high salt concentration, centrifuged at 1500 revs min⁻¹ for 15 min at 4°C (HS-4 rotor), and the matrix pellet was resuspended in TM-0.2.

Nuclei were stabilized by incubating at 37 °C for Ih followed by digestion with DNase I as previously described. An equal volume of LIS in 0.1% digitonin (Fluka), 4mM KC1, 10 mM Hepes, pH 7.4, 0.5mM spermidine, 4UM EDTA (Mirkovitch et al. 1984) was slowly added to a final concentration of at least 5mM. The total volume was brought to 40 ml by adding 0.1% digitonin, 2 mM KC1, 5 mM Hepes, pH 7.4, 0.25 mM spermidine, 2 mM EDTA containing the appropriate LIS concentration, centrifuged at 1500 revs min^- for 15 min at 4 °C, and the matrix pellet was resuspended in TM-0.2.

Protein concentration was determined as reported by Lowry et al. (1951), and RNA and DNA quantities were determined as described by Munro and Fleck (1965). 1-D 12% SDS-PAGE was performed according to the method of Laemmli (1970). Molecular weight markers were (from top to bottom): myosin, /3-galaetosi-dase, phosphorylase b, bovine serum albumin, ovalbumin, glyceraldehyde-3-phosphate dehydrogenase, carbonic anhydrase, trypsinogen, trypsin inhibitor, alpha-lactalbumin. 2-D gels were prepared using non-equilibrium isoelectric focusing gel electrophoresis in the first dimension (O’Farrell et al. 1977).

Samples were prepared for thin-section electron microscopy analysis as previously described (Smith et al. 1984). Briefly, the samples were fixed in 2.5% glutaraldehyde, 100 mM sodium cacodylate, pH7.4, 5mM MgCL (0.2 mM for matrices) for at least 2h at 4°C. The samples were rinsed with cacodylate buffer without glutaraldehyde, post-fixed with 1% OsO₄ for 30 min on ice, dehydrated (graded 30% to 100% ethanol and 100% acetone), and infiltrated with Epon-Araldite resin (Electron Microscopic Sciences). Cured blocks were thin sectioned and stained with lead citrate and uranyl acetate. Sections were examined on a Hitachi H-500 electron microscope.

HeLa cell homogenate, nuclear fraction and nuclear matrix fraction were run on 2-D gels (O’Farrell et al. 1977) and transferred to nitrocellulose filters. The filters were placed in block buffer (10 mM Tris–HCl, pH 7.5,150 mM NaCl, 0.5% Tween-20) for 1 h at 37 °C, followed by incubation for 1 h with the murine monoclonal antibody (IFA) that recognizes a common epitope to all known intermediate filament proteins (Pruss et al. 1981). Filters were washed four times (5 min each), incubated with goat anti-mouse IgM antibodies conjugated to horseradish peroxidase for 1h, washed four times, and developed with 1M Tris, pH9.5, 1M MgCh, 10 mg ml⁻¹ nitroblue tétrazolium, 100 mg ml⁻¹ bro-mochloroindolyl phosphate.

The method used to isolate nuclear matrices from intact HeLa S3 cells grown in suspension culture required only 4–5 h. An average of 70% of the nuclei were recovered on the basis of A260 measurements. More than 95% of the nuclei were free of visible cytoplasmic contamination when the cells were syringed through a 22 gauge needle (Fig. 1B), but approximately 50% were contaminated with visible cytoplasmic debris when a Dounce homogenizer was used (Fig. 1A). The isolated nucleus (Fig. 1C) retained the overall appearance, size, and shape of the nucleus in intact cells (Fig. 1D). Both the isolated nucleus and the nucleus of the intact cell consisted of a continuous peripheral nuclear lamina, nucleoli, chromatin and nonchromatin structures such as interchromatinic granules and peri chromatin fibrils.

Staufenbiel and Deppert (1982) in particular have emphasized that apparently highly purified nuclei from tissue culture are often contaminated with large amounts of intermediate filament components, which collapse upon the nuclei and resist subsequent extraction. Since intermediate filament proteins are notoriously resistant to high-salt extraction, any contaminating intermediate filament proteins should co-isolate with salt-extracted nuclear matrices, where they could potentially represent major protein components of the isolated matrices. This in turn could confuse the identification of those proteins that are truly nuclear in origin.

While our extensive electron-microscopic thin sectioning (Fig. 1) and phase-contrast observations (data not shown) suggested that our syringe technique for nuclei isolation was far superior to the Dounce-type homogenizer technique in reducing cytoplasmic contaminant, we decided to examine the question of specific contamination of nuclear and nuclear matrix preparations with intermediate filament proteins using the highly sensitive Western blot technique. A murine monoclonal antibody (IFA) reactive to all known intermediate filament proteins (Pruss et al. 1981) and the nuclear lamins (Osborn and Weber, 1987; Lebel and Raymond, 1987) was used to detect the presence of any such proteins in the isolated nuclei and nuclear matrix fractions. 2-D Western analysis of blots screened with IFA clearly label the nuclear lamins in the nucleus (Fig. 2C and D) and the nuclear matrix (Fig. 2E and F). The IFA staining of the nuclear lamins in the total cellular proteins were not visible, since they represented such a minor component, but the intermediate filament proteins, including vimentin and the cytokeratins in this fraction were stained strongly (Fig. 2A and B). These intermediate filament proteins were much less intensely labeled in the isolated nuclear fraction and represented only a very minor component in the matrix fraction. None of the major matrix proteins other than the nuclear lamins reacted with IFA.

Nuclear matrices were isolated using different concentrations of (NH4)₂S0₄ or NaCl as the extraction salt. The minimal ionic strength that effectively removed loosely associated proteins, RNA and DNA was 0.6 for either salt (Fig. 3). At this ionic strength the amount of extractable protein, RNA and DNA was 78%, 73% and 98%, respectively, using (NH₄)₂SO₄ and 86%, 76% and 99% using NaCl. Polypeptide profiles on 1-D SDS–PAGE (Fig. 4) and densiometric quantitation (data not shown) revealed that both salts effectively extracted over 99% of histone Hl and at least 95% of the core histones. The polypeptide compositions of matrices isolated with either salt were very similar except that the (NH₄)₂S0₄ matrices had noticeably more intense bands at 36, 40 and 140 × 10³M_r. However, regardless of the type and concentration of salt used, the matrix consisted of a vast array of polypeptides ranging in size from 30×10³M_r to over 205×10³M_r. Closer examination of the polypeptide composition of the HeLa 0.2 M (NH₄)₂SO₄ matrix on the 2-D gel (Fig. 2E) shows a remarkable similarity to that of rat liver and Chinese hamster Don cell nuclear matrices (Nakayasu and Berezney, unpublished data). In addition to the nuclear lamins and the heterogeneous nuclear RNP (hnRNP) core proteins, the major matrix proteins consist primarily of a basic group of polypeptides between 60 and 70×10³M_r, and several high and low molecular weight acidic proteins (Fig. 2F; and Berezney, 1984).

Nuclear matrices isolated under all conditions had a characteristic tripartite structure, consisting of a surrounding nuclear lamina, residual nucleoli and a fibro-granular internal network (Fig. 5; and Berezney and Coffey, 1974, 1977; Berezney, 1984). The nucleolar and fibrogranular internal network structure appeared to be best maintained in the 0.2 M (NH₄)SO₄ nuclear matrices (compare Fig. 5A with Fig. 1C and D), which will be referred to as AS matrices. The other preparations showed variable degrees of aggregation of the internal network, large empty regions, and diffuse nucleolar structures (Fig. 5B, C, and D).

Nuclei and AS matrices were isolated in the continuous presence of the sulfhydryl alkylating agents N-ethyl-maleimide (NEM) and iodoacetamide (IA). The swollen cells were lysed in the presence of IA/NEM, minimizing the time for sulfhydryl oxidation to occur. The IA/NEM matrices were nearly identical to untreated matrices with respect to protein, RNA and DNA content (Table 1), and polypeptide composition (data not shown). The ultrastructure of these matrices resembled that of the untreated matrices, with no indication of a disruption of the internal network (compare Fig. 6A with 5A).

AS matrices were isolated from nuclei incubated for 1 h at 4 °C in the presence of sodium tetrathionate (NaTT), a sulfhydryl oxidizing agent. NaTT matrices had a 1.6-fold increase in protein, a 1.5-fold increase in RNA, and a 4.0-fold increase in DNA compared to untreated matrices (Table 1). Similar effects were found when nuclei were incubated at 37 °C for Ih in the absence of NaTT. The fibrogranular network of stabilized nuclear matrices characteristically had thicker and more densely stained filaments than matrices not stabilized at 37 °C or with NaTT (compare Fig. 6C, D and F with A, E and Fig. 5A). When NaTT was included in the 37 °C incubation there was a slight additive effect with respect to the protein content (Table 1). The IA/NEM effectively blocked the stabilization by NaTT but not that of the 37 °C incubation, indicating that the mechanism involved in the 37 °C stabilization is distinct from that of sulfhydryl oxidation. The 2-D gel polypeptide patterns of unstabilized and stabilized matrices were virtually identical, although there were some differences in minor components (data not shown). This indicates that the increased amount of proteins and apparent internal matrix structure following stabilization is due predominantly to a corresponding increase in pre-existing major nuclear matrix proteins rather than to the appearance of new species of protein.

AS matrices were isolated from nuclei in which RNase A was included in the DNase I digestion step. Protein recoveries decreased as increasing amounts of RNase A were used up to 5 pg ml^-1 (Fig. 7). However, further increasing amounts of RNase A resulted in progressively higher amounts of protein associated with the matrices (Fig. 7) even though they had less internal structure (Fig. 9A). 1-D SDS–PAGE revealed that increasing levels of RNase A resulted in the progressive increase in a low molecular weight component which migrated at a position (M_r 17 000) identical to that of purified RNase A (Fig. 8). This was confirmed on 2-D gels (data not shown). Indeed, at the highest levels of RNase A used (300 μg ml⁻¹) this low molecular weight protein accounted for most of the protein seen on the gel (Fig. 8, lane 5). This suggested that isolated nuclear matrices were tightly binding exogenous RNase A, which might account for the progressive increase in total protein despite the apparent extraction of nuclear matrix proteins following treatment with high levels of RNase A. The RNase A resulted in matrices with an approximately 50% decrease in RNA and DNA (Table 2), and when 2.0 M NaCl was substituted for AS as the extraction agent, the resulting structures were further depleted of interna] material and RNA (Table 2) yet retained distinct residual nucleoli (Fig. 9B).

AS matrices were isolated from nuclei incubated in the presence of dithiothreitol (DTT), a sulfhydryl reducing agent, for 1 h at 4 °C. This treatment resulted in the matrix protein content dropping 46% (Table 1). The structures were perturbed, with a less extensive fibrogranular network and highly diffuse residual nucleoli (Fig. 6B). On the basis of the destabilizing effects of DTT, RNase A and 2.0 M NaCl extraction, these treatments were used in a single nuclear matrix preparation. The resultant structures were virtually empty nuclear lamina (Fig. 9D). Structures that retained some internal filaments and residual nuclei were observed when AS was substituted for 2.0 M NaCl (Fig. 9C). These findings indicate that the structural integrity of the matrix is dependent on RNA and disulfide bonds that exist in situ. The dithiothreitol but not the RNase A had a marked effect on the residual nucleoli.

Isolated nuclei were incubated at 37 °C or 4 °C for lh, digested with DNase I for 0.5 h, and extracted with different concentrations of 3,5-lithium diiodosalycylate (LIS). The minimal LIS concentration that effectively removed loosely associated proteins, RNA and DNA was 5mM (Fig. 10). At this concentration the amount of extractable protein, RNA and DNA was 88%, 90% and 99%, respectively, using unstabilized nuclei, and 70%, 74% and 95% using 37°C stabilized nuclei.

Polypeptide profiles on 1-D SDS–PAGE (Fig. 11) show that LIS matrices isolated from 37 °C stabilized nuclei consisted of many polypeptide components, whereas most of these polypeptides are solubilized when the nuclei are not stabilized prior to LIS extraction. The polypeptide profile of unstabilized 5mM LIS matrices consisted primarily of the nuclear lamins, cytoskeletal proteins and histones. A relatively high amount of histone was also associated with 37 °C stabilized 5mM LIS matrices compared to their salt-extracted counterparts (compare Figs 4 and 11). Increased concentrations of LIS were not effective in removing the histones (data not shown).

Matrix structures similar in appearance to AS matrices (Fig. 5A) were observed when stabilized nuclei were treated with 5mM LIS (Fig. 12B); 25 mM LIS resulted in full but atypical structures that had more diffuse residual nucleoli (Fig. 12D). The unstabilized 5mM LIS structures were empty and very fragile, since most of them collapsed upon preparation for electron microscopy (Fig. 12A). 25 mM LIS extraction of unstabilized nuclei resulted in empty structures (Fig. 12C) and much debris as observed with phase-contrast microscopy (data not shown).

We report that 0.2 M (NH₄)₂SO₄ is a milder extracting agent than NaCl and LIS, on the basis of the apparent preservation of the residual nucleolus, the presence of an elaborate fibrogranular network that resembles the in situ network in whole cells and nuclei, minimal aggregation and sufficient solubilization of DNA and histones. These results are consistent with the previous findings of Verheijen et al. (1986) and Fey et al. (1986).

Major differences in protein composition of matrices isolated from cultured cells are often attributed to intermediate filaments that are not completely sheared from isolated nuclei (Staufenbiel and Deppert, 1982, 1983; Capeo et al. 1982; Fey et al. 1984; Verheijen et al. 1986). The nuclei and nuclear matrices that we obtain using oui-standard HeLa cell protocol are virtually free of cytoplasmic contamination and contain relatively low amounts of cytoplasmic intermediate filament proteins. Jackson and Cook (1988) have suggested that the isolation of nuclear matrices under hypotonic conditions may introduce artefacts. The polypeptide patterns on 2-D gels of HeLa and Don nuclear matrices isolated in hypotonic conditions, however, are virtually identical to that of rat liver nuclear matrices isolated in isotonic conditions (Nakayasu and Berezney, unpublished data).

Recently, the deduced primary and predicted secondary structures of the nuclear lamins have been reported to be homologous to that of the cytoplasmic intermediate filament proteins (McKeon et al. 1986; Fisher et al. 1986; Hoger et al. 1988). In addition, the murine anti-intermediate filament monoclonal antibody (IFA) (Pruss et al. 1981) cross-reacts with the nuclear lamins (Lebel and Raymond, 1987; Osborn and Weber, 1987). Our observation that the major nuclear matrix proteins, other than the lamins, do not react with IFA suggests that these proteins are not related to the lamins in having an intermediate filamentlike property.

Mirkovitch et al. (1984) developed a low-salt extraction protocol that utilized 25 mM LIS as the extracting agent. They suggested that the high-salt extraction produced sliding of DNA at matrix attachment points. This alternate method was thought to be a milder extraction procedure, preventing DNA rearrangement. The importance of nuclear stabilization for obtaining LIS matrices is quite evident. We find that the ultrastructure of the 5 mM LIS matrices isolated from 37 °C stabilized nuclei appears similar to that of the AS matrices, but omitting the stabilization results in empty nuclear lamina-type structures, as reported by Lewis et al. (1984). The LIS matrices had a significantly higher amount of associated DNA and histones. Smith et al. (1987) also found that LIS matrices contained a high level of histones, but they noted that the ultrastructure of LIS matrices was substantially different from that of high-salt matrices despite the 37 °C stabilization and regardless of the LIS concentration. The 25 mM LIS is clearly too harsh a treatment for morphological studies, since the resultant matrices had perturbed structures.

Kaufman et al. (1981) found that when the RNase A digestion was performed prior to the high-salt extraction the resulting structures were devoid of any residual nucleoli and lacked an extensive internal network. An RNase-sensitive matrix has also been described by others (Adolph, 1980; Long and Schrier, 1983; Bouvier et al. 1984; Fey et al. 1986), although this finding has been challenged by van Eekelen et al. (1982). We also found that the internal matrix structure is sensitive to RNase A but the residual nucleolus is more resistant to it. Our finding that progressively increased levels of RNase A treatment (above 5 μg ml⁻¹) results in corresponding increases in the binding of exogenous RNase A to isolated nuclear matrices and the recovery of total nuclear matrix protein stresses that this particular treatment must be used with extreme caution and at very low levels of RNase A (Sjigml^-1). It may also account for some of the contradictory reports and confusion in the literature concerning RNase A sensitivity, since use of high levels of RNase A as reported by van Eekelen et al. (1982) could presumably lead to matrix protein recoveries similar to or higher than that of control matrices (see Fig. 9).

Consistent with the report of van Eekelen et al. (1982), the IA/NEM treatment had no effect on the morphology and chemical composition of the matrix, indicating that any disulfides formed during isolation in the absence of IA/NEM are not essential for maintaining the structural integrity of the in vitro matrix. Kaufman et al. (1981) and Kaufman and Shaper (1984) reported that disulfide bonds formed during the isolation of nuclei and subsequent nuclear matrices were responsible for inducing the internal matrix structure. Their conclusions are compromised, however, since these studies also included RNase A digestion, which in turn destabilizes the matrix structure. Indeed, when Kaufman et al. (1981) performed a nuclear matrix isolation in which IA was used but the RNase A digestion was omitted, the resultant structures were not empty. Hodge et al. (1977) and Dijkwel and Wenink (1986) reported that DTT had no effect on nuclear matrices, but we find that the DTT does disrupt the internal structure, albeit not to the extent reported by Kaufman and Shaper (1984). It thus appears that the structural integrity of the matrix may depend to some degree on disulfides that probably existed in situ.

The 37 °C and/or NaTT stabilization resulted in nuclear matrices containing approximately twice as much protein, RNA and DNA as control preparations. The NaTT stabilization effect was blocked by IA/NEM but the 37 °C stabilization was not. Isolated nuclei were exposed to IA/NEM for approximately 2 h before incubating at 37 °C, and it is well established that the IA and NEM derivatives are stable at temperatures as high as 110°C, since both alkylating agents are routinely used in amino acid analysis (Means and Feeney, 1971).

Although the 37 °C stabilization phenomenon is not understood we have shown that it is probably not due to sulfhydryl oxidation. The 37 °C and NaTT treatments may, in addition to stabilizing the same set of nuclear matrix proteins, selectively stabilize other proteins that are generally either not detected or found as only minor components in unstabilized nuclear matrix preparations. In this regard, Evans and Hancock (1985) found that the c-myc protein was associated with isolated matrices only when the nuclei were stabilized by 37 °C incubation and that sulfhydryl oxidation was probably not the stabilizing factor. A similar 37 °C stabilization phenomenon was demonstrated with polyoma large-T antigen, but NaTT also resulted in a similar effect (Humphrey and Pigiet, 1987).

We thank Dr Harold Asch for the Pruss monoclonal antibody to all known intermediate filament proteins, Dr Sei-Ichi-Matsui for his initial assistance in isolating nuclei by the syringe technique, Linda A. Buchholtz and William B. Tobin, who performed the initial experiments on HeLa nuclei and nuclear matrix isolation, and Jim Stamos for the illustrations. This work was supported by National Institutes of Health grant GM-23992.