Ultrastructural rRNA localization in plant cell nucleoli RNA/RNA in situ hybridization, autoradiography and cytochemistry

The nucleolus is a widely studied nuclear structure in which ribosomal RNA synthesis, as well as assembly of preribo-somes, takes place (Hadjiolov, 1985). Under the electron microscope, three main domains are visible in this struc-ture: the dense fibrillar component (DFC), the granular component (GC) and the fibrillar centres (FCs). Different types of vacuoles and nucleolar interstices have also been found in both plant and animal cells, whereas peri- and intranucleolar chromatin are characteristic of animal cells (Goessens, 1984; Puvion and Moyne, 1980; Raska and Dundr, 1993; Raska et al., 1992; Thiry et al., 1991); con-densed chromatin is not found inside the plant nucleolus except in the heterogeneous fibrillar centres (Risueño, 1993; Risueño and Medina, 1986; Risueño et al. 1982).

Various cytochemical techniques have been used to localize ribosomal genes, their products, and the proteins involved in ribosomal transcription and RNA processing within these components (see reviews by Deltour and Mosen, 1987; Deltour and Motte, 1990; Derenzini et al., 1987, 1990; Derenzini and Ploton, 1991; Goessens, 1984;

Hadjiolov, 1985; Hernandez-Verdun, 1991; Jordan, 1991; Raska et al., 1990; Risueño and Medina, 1986; Scheer and Benavente, 1990; Schwarzacher and Wachtler, 1991; Stahl et al., 1991). High-resolution autoradiography after short pulses with tritiated uridine, together with several cyto-chemical methods for detecting DNA, histones and nucle-olar proteins, indicated that the FCs contained inactive ribo-somal genes and that the surrounding DFC was the site of rRNA synthesis (reviewed by Goessens, 1984; Medina et al., 1983; Risueño and Medina, 1986; Risueño et al., 1982). Nevertheless, in the last few years, there has been no agreement on the site of transcription in the fibrillar com-ponents. Immunocytochemical studies have demonstrated the presence of proteins that play a key role in ribosomal transcription, such as RNA polymerase I, DNA topoiso-merase I and upstream binding factors (UBF), on FCs and/or DFC (Derenzini et al., 1990; Raska et al., 1989, 1992; Rodrigo et al., 1992; Rose et al., 1988; Roussel et al., 1993; Scheer and Rose, 1984; Scheer et al., 1987).

Selective staining methods and immunogold techniques allowed the detection of DNA in FCs and in the surround-ing DFC in both animal and plant cells (Derenzini et al., 1990, 1993; Martin and Medina, 1991; Motte et al., 1991; Olmedilla et al., 1992; Raska et al., 1992; Testillano et al., 1991, 1992a; Thiry et al., 1991). From these data, differ-ent interpretations arise: transcription occurs on FCs (Scheer and Benavente, 1990; Thiry and Goessens, 1992), at the border of the FCs (Derenzini et al., 1990; Derenzini and Ploton, 1991), or in the DFC (Hartung et al., 1990; Schwarzacher and Wachtler, 1991); for other hypotheses see Jordan (1991).

Ultrastructural non-isotopic in situ hybridization is a powerful method for visualizing nucleic acids and consti-tutes a very convenient way of studying the molecular com-position of different nucleolar components. However, con-tradictory results have been obtained when this technique is applied to different materials (Dundr and Raska, 1993; Ghosh and Paweletz, 1990; Hozák et al., 1993; Motte et al., 1991; Raska and Dundr, 1993; Thiry and Thiry-Blaise, 1989, 1991; Thiry et al., 1991; Stahl et al., 1991; Watch-ler et al., 1990, 1992). The results obtained relating to the localization of rDNA using probes encompassing different regions of the transcription unit and the results obtained using different materials are not in agreement: sometimes they are located on DFC and at other times on FCs (Géraud et al., 1991; Puvion-Dutilleul et al., 1991a, 1992; Thiry and Thiry-Blaise, 1989, 1991; Watchler et al., 1990, 1992).

In studies relating to the localization of the products of transcription by non-radioactive in situ hybridization at the electron microscope level, some discrepancies are also found. The rRNAs are reported to be localized by DNA probes mainly on DFC and GC in Ehrlich, HeLa and mouse 3T3 cells (Puvion-Dutilleul, 1991a, 1992; Thiry and Thiry-Blaise, 1989) although in the Vero cell line rRNA was detected on DFC but not on GC (Escaig-Haye et al., 1989). No labelling was found inside FCs of untreated HeLa and 3T3 mouse cells whereas some infrequent hybridization sig-nals were found on FCs of cells treated with actinomycin D (Puvion-Dutilleul, 1991a,b, 1992). In Ehrlich tumour cells, the RNA is reported to be localized in the peripheral regions of the FCs (Thiry and Thiry-Blaise, 1989).

There are data on plant material showing the localization of rDNA (Motte et al., 1991; Highett et al., 1993) but the use of high-resolution in situ hybridization to detect RNA in plant material is not very extensive (Harris and Croy, 1986; Mc Fadden, 1989; Mc Fadden et al., 1988; Olmedilla et al., 1993; Sato et al., 1991). Furthermore, these investi-gations have not addressed the question of the localization of rRNAs in the nucleolus. The application of in situ hybrid-ization using 18 and 25 S rRNA probes for the first time allowed the localization at the electron microscope level of ribosomal transcripts in different components of the plant nucleoli. On the other hand, it has been proven that higher sensitivity is obtained by the use of RNA probes for RNA detection (Mc Fadden, 1989). For this reason in the present study we have developed a method of RNA/RNA in situ hybridization with our plant material in order to study the localization of the ribosomal RNAs. We have taken advan-tage of the fact that the 18 and 25 S coding regions are well conserved throughout evolution and have used clones from Arabidopsis or Raphanus to hybridize with the cellu-lar rRNA of Allium and Capsicum. The localization at the ultrastructural level of these RNAs was also achieved by high-resolution autoradiography and using different cyto-chemical methods. The combination of in situ hybridization with different cytochemical methods provides additional data on the presence of these transcripts in the various nucleolar components (dense fibrillar and granular compo-nents, and periphery of fibrillar centres).

The materials used were root-tip meristematic cells from Allium cepa L. bulbs germinated at 15°C under standard conditions, and anthers of Capsicum annuum L. flowers from plants grown in a greenhouse under controlled conditions.

For EDTA staining and autoradiography root tips were fixed in 3% glutaraldehyde (Taab Lab. Equip. Ltd, Reading, UK) in 0.025 M cacodylate buffer, pH 7, for 2 hours at room temperature. Post-fixation in 1% osmic acid for 1 hour in the same buffer was per-formed for autoradiography samples, and then, they were dehy-drated in ethanol, and embedded in Epon. Formvar-copper grids were used to collect the sections.

For the methylation and acetylation (MA) method, and in situ hybridization, meristematic onion root cells and pepper anthers were fixed in 4% formaldehyde (Merck, Germany) in PBS buffer, pH 7.3, at 4°C, for 4 hours or overnight, respectively, dehydrated in methanol and embedded in Lowicryl K4M at −20°C (Bendayan et al., 1987; Carlemalm et al., 1980). Ultrathin sections were col-lected on Formvar-coated gold grids.

Grids were floated in 5% aqueous uranyl acetate solution, 0.2 M EDTA and lead citrate according to Bernhard’s protocol (Bern-hard, 1969).

Roots attached to the bulbs were incubated for 30 minutes in water containing 100 μCi/ml [³H]uridine (sp. act. 25 Ci/mmol) (Radio-chemical Centre, Amersham, England). Grids with ultrathin sec-tions were coated with Ilford L4 emulsion by the loop technique (Bouteille et al., 1976). After exposure in the dark for 11-12 months, at 4°C, developing was performed with Phenidon. The sections were stained with uranyl acetate and lead citrate.

Sections on grids were dehydrated in a methanol series (70% for 15 minutes followed by three changes of pure methanol for 15 minutes each) and then treated with a freshly prepared methanol:acetic anhydride (5:1,v/v) mixture at room temperature overnight. After washing in water, sections were stained with 5% uranyl acetate for 60 minutes at 60°C (Tandler and Solari, 1982; Testillano et al., 1991).

The 18 S rRNA Raphanus probe was synthesized from a rDNA fragment of 1.2 kb covering almost all the 18 S rDNA region

(Delseny et al., 1983) subcloned in the pGEM-7Zf(+) plasmid. Biotinylated RNA probes for nuclear encoded 18 S and 25 S rRNA were synthesized from Arabidopsis rDNA fragments of 1.5 and 2.3 kb, respectively (Gruendler et al., 1991), subcloned in Blue-script KS+ plasmid.

RNA probes were prepared from the linearized plasmids using biotinylated UTP according to the Promega protocol. Lineariza-tion of the plasmids for synthesizing the sense or antisense probes was performed using restriction enzymes that cut at restriction sites as far as possible from the corresponding RNA Pol transcription promotor that was used in each case (see Figs 4 and 5). For each preparation linearized plasmid DNA (1 μg) was incubated for 90 minutes at 40°C with an RNA polymerase solution containing 2.5 mM ATP, CTP and GTP (Boehringer-Mannheim), and 0.5 mM biotinylated UTP (biotin-11-UTP) (Sigma), in the presence of RNasin (Promega). Non-isotopic labelled RNA was recovered using a Sephadex G 25 column, followed by ethanol precipitation. The probe was resuspended in hybridization buffer (50% formamide, 75 mM NaCl, 50 mM PIPES, pH 7.2, 5 mM EDTA, 0.02% (v/v) Ficoll, 0.02% (w/v) polyvinylpirrolidone, 0.02% (w/v) bovine serum albumin and 100 μg/ml herring sperm DNA.

Prior to hybridization, ultrathin sections were treated with 1 μg/ml proteinase K (Boehringer-Mannheim) in 100 mM Tris-HCl, pH 8.0, and 50 mM EDTA for 1 hour. RNA/RNA hybridization was carried out in a moist chamber at 55°C overnight on 50% for-mamide hybridization buffer containing 2 ng/μl RNA biotinylated probe. Non-specific RNA hybridization was removed by washing in: 4× SSC (1× SSC is 0.15 M NaCl, 0.015 M sodium citrate) (4× 2 minutes), 2 × SSC (4× 2 minutes), 1 × SSC (1× 2 hours at 55°C) and in SC buffer (50 mM PIPES, 0.5 M NaCl, 0.5% Tween-20) (1× 15 minutes).

Blocking of non-specific antigens was performed by floating the grids on 1% BSA in SC buffer for 15 minutes at room tempera-ture. The hybrids were detected with a rabbit anti-biotin antibody (Enzo Biochemicals Inc. NY, USA), 1:50 in SC for 1 hour at room temperature. The antibody was rinsed off with SC buffer (4× 2 minutes). Then, grids were floated on a drop of goat anti-rabbit IgG conjugated to colloidal gold particles of 10 nm diameter (Jansen Biotech, NV; Olen/Belgium) diluted 1:20 in SC for 45 minutes. Finally, sections were washed in SC (4× 2 minutes) and distilled water, air dried and stained with uranyl/lead.

The hybridization signal that appears on FCs, DFC and GC was evaluated in each micrograph by counting gold particles per area unit in these components and in the cytoplasmic vacuoles (where gold particles represent the background); 20-25 cell micrographs were used for each experiment. Therefore, labelling density in cytoplasmic vacuoles was used as an internal control and its main value was in estimation of the background.

FCs were counted in 20-25 nucleoli for each experiment, con-sidering FCs: (1) without gold particles; (2) with gold particles forming clusters covering also the DFC; and (3) with gold parti-cles inside. Only FCs whose labelling density value was much higher than the mean value for the estimated background (density labelling on cytoplasmic vacuoles) were counted as labelled.

After the hybridization some grids were postfixed in 3% glu-taraldehyde in PBS for 5 minutes, rinsed in the same buffer and in distilled water. Finally, the MA procedure was applied as described above, followed by EDTA/staining.

Plant nucleoli have a well established morphology (Figs 1-3). They are constituted of dense fibrillar component (DFC) and granular component (GC); the latter being almost absent in inactive nucleoli. In active nucleoli the GC is intermingled with the DFC (Fig. 3) but in moderately or less-active nucleoli the GC is present at the peripheral region of the nucleoli (Figs 1 and 2). Immersed in the dense fibrillar component clear areas are visible that constitute the fibrillar centres (FCs). Depending on the degree of activity two types of FC can be found on the nucleolus. The homo-geneous FCs are small and numerous and contain decon-densed chromatin fibres. The heterogeneous FCs are larger and less numerous, containing cores of condensed chro-matin. An intermediate type of FC can also be observed with small cores of condensed chromatin associated with decondensed chromatin fibres. Vacuoles of different sizes, sometimes connected to FCs, can also be present in the nucleolus (Figs 1 and 2). We have used different techniques to determine the composition of nucleolar components.

The application of the methylation and acetylation (MA) method followed by uranyl staining blocks the staining of proteins and results in a clear distinction between the dense fibrillar and the granular components of the nucleolus. Extranucleolar condensed chromatin masses show a high electron density. With this method we observed the con-densed chromatin cores of the heterogeneous FCs with a very high contrast as well as some fibrils and granules with a lower contrast (Fig. 1).

The nucleolus after EDTA/staining, which is preferential for ribonucleoproteins, showed the dense fibrillar and granular components well contrasted while extranucleolar chromatin patches were not contrasted. Within the fibrillar centres some bleached areas similar to those forming chromatin masses were found as well as some stained fibrils and granules (Fig. 2).

High-resolution autoradiography was performed after 30 minutes incubation of tritiated uridine and a long period of exposure (12 months). In Allium cepa 30 minutes is a short incubation time, due to the problems of reagent penetration into plant cells and shorter periods result in non-appreci-able incorporation. In the nucleolus, the dense fibrillar com-ponent was labelled very frequently in regions close to the fibrillar centres and, in some cases, silver grains also cov-ered certain areas of the fibrillar centres. The granular com-ponent was also labelled (Fig. 3). Uridine incorporation was also evident in inter- and perichromatin regions.

RNA/RNA in situ hybridization at the electron microscope level was performed using three biotinylated probes. Mature regions of 18 S rRNA from Raphanus and 18 S and 25 S rRNA from Arabidopsis were used for synthesizing the probes (Figs 4 and 5). In situ hybridization with these three probes revealed a similar pattern of labelling on both Allium cepa L. and Capsicum annuum L. pollen grains (Figs 6, 7 and 8). Most of the hybridization signal was located over the nucleolus and cytoplasm, and some labelling was also found on the nucleoplasm; in contrast, condensed chromatin masses appeared to be free of gold particles (Figs 6A,D, 7A and 8A,C). In the cytoplasm ribosome-rich areas were completely covered with particles while vacuoles and cyto-plasmic organelles such as mitochondria appeared not to be labelled (Figs 6A,D, 7A and 8A). In the nucleolus, gold particles were concentrated on the dense fibrillar and gran-ular components (Figs 6A, 7A and 8A). In order to better visualize the structures where probes are localized we have combined the MA method and EDTA/staining with the in situ hybridization technique. By this means we can see that the density of labelling is higher over the granular component, and gold particles are also present on the dense fib-rillar component. In FCs some stained fibrils were visualized covered by gold particles (Fig. 9). The areas corresponding to fibrillar centres were, in some cases, not labelled (Figs 6A,B,C, 7A,B,C and 8A,B) and in the cases where gold particles were present, the particles formed clusters that also covered the surrounding dense fibrillar com-ponent (Figs 6A,B,C, 7A,B,C and 8A,C). Nevertheless, when the particles appeared close to the central area of the FCs they were usually isolated (Figs 6B, 7C and 8B).

As mentioned above, a similar pattern of labelling was observed with each of the three probes used and also in each tissue tested: meristematic onion root and pepper pollen grains. Only slight differences were observed in the hybridization signal obtained with the different probes; gold grains are more abundant after hybridization with the 18 S rRNA probes than with the 25 S rRNA probe (compare Figs 6 and 7 with Fig. 8). The hybridization signal obtained from the 18 S probe of Raphanus was stronger than that obtained with the 18 S probe of Arabidopsis (compare Fig. 6 with 7). With respect to the differences between materi-als, the nucleoli of both cell types have some differences in ultrastructure, the nucleolus of Capsicum pollen grains being smaller and almost exclusively formed of DFC (com-pare Figs 6A, 7A and 8A with Figs 6D and 8C). The gold particles detected were more abundant in onion root meristematic cells than in pepper pollen grains, irrespective of the probe used.

Controls were performed in order to verify the specificity of the in situ hybridization reaction. Sense biotinylated rRNA probes from the three plasmids were synthesized and used for in situ hybridization in the same conditions. With each of the sense probes the labelling was not significant (Fig. 7D). Also, no signal was observed when anti-biotin antibodies were replaced by PBS buffer to detect the biotinylated hybrids.

For the quantitative study we have taken into account that there is no clear border between DFC and FCs; thus to decide to which of these components gold particles should be assigned we assumed that a gold particle is on a FC when it is not over a stained fibre connected with the adjacent DFC. The results on the distribution of labelling den-sity in the different nucleolar components (DFC, GC and FCs), as well as in cytoplasmic vacuoles, are shown in Fig. 10. The data show the highest labelling density on the DFC and GC (174.75 particles/μm²), and a much lower value on FCs (24.47 particles/μm²), while the corresponding density for cytoplasmic vacuoles is comparatively very low (2.16 particles/μm²).

We have also performed a statistical analysis of the labelling on FCs in order to differentiate the hybridization signal from the background in these structures (Fig. 11). We have taken as reference the density of label on cyto-plasmic vacuoles where gold particles correspond to back-ground. We have counted the density of gold particles per unit area in FCs and in cytoplasmic vacuoles that appear in the same micrograph. The x-axis corresponds to the den-sity of gold labelling per unit area in FCs and the y-axis to the density of gold particles per unit area in cytoplasmic vacuoles. The enormous dispersion of the cloud of points for these pairs of values indicates that there is no defined correlation between the two densities of labelling. There-fore, these data indicate that the labelling on FCs does not correspond to background but it represents hybridization signal. Moreover, by applying to this group of points the statistic computation of the regression line, a correlation coefficient of −0.43 with 14 degrees of freedom has been obtained for the best fit (after using different models of regression analysis: linear, exponential, reciprocal, multi-plicative, and trying with either the original data, or the transformed data: log, ln, square root, arcsen). This value shows no correlation between the two densities of labelling (Fig. 11).

FCs showed three types of labelling: without gold parti-cles, with gold particles forming clusters covering both DFC and FCs, and with gold particles inside (see the three classes shown schematically in Fig. 12). After quantitative analysis, we found that the most abundant pattern was that of FCs without gold particles, between 49 and 63%; FCs showing the second pattern of labelling comprised 10-21%, and FCs with gold particles inside between 23 and 36%.

In this paper the localization of transcriptional products has been studied by high-resolution in situ hybridization using different RNA biotinylated probes allowing the detection of rRNA precursors and mature 18 and 25 S rRNAs in plants. Moreover, ultrastructural cytochemistry and autora-diographical assays provide additional information on the presence of those transcripts in the different nucleolar com-ponents. On the basis of this study we find that rRNA is present in GC, DFC and at the periphery of some FCs. All these data suggest that transcription is initiated at the periphery of the FCs but a role for the DFC in this process should be considered.

The cytochemical method of MA followed by uranyl staining is a procedure that avoids the uranyl staining of proteins because it blocks their amino and carboxyl groups (Tandler and Solari, 1982). The adaptation of this method for plant material embedded in Lowicryl is a helpful cyto-chemical method to use with nucleic acid-containing struc-

tures (Risueño, 1993; Testillano et al., 1991). The MA method makes it possible to distinguish between fibrillar and granular ribonucleoproteins; this difference is not always evident on Lowicryl sections. The fibrillar material stained on FCs using this method represents DNA and RNA-containing structures. On the other hand, due to the increase in contrast of the condensed chromatin we have confirmed, using this method, that the unique condensed chromatin structures observed in the plant nucleolus are the chromatin cores of heterogeneous FCs, since there are no other areas of densely stained material except the one that appears in these FCs (Deltour and Motte, 1990; Medina et al., 1993, Risueño et al., 1982; Testillano et al., 1991).

After autoradiography with four months of exposure, RNA was mainly detected at the dense fibrillar component surrounding fibrillar centres in Allium cepa L. but no sig-nificant labelling was found on FCs (Risueño et al., 1982). However, after a longer exposure (10-12 months) the appearance of silver grains very close to, or inside the fib-rillar centres increases, indicating the presence of newly synthesized RNA. In these components ribonucleoprotein material is also detected after EDTA/staining (Bernhard, 1969). In a previous paper in which an autoradiographical and cytochemical study was made on the Allium cepa L. nucleolus, the presence of some material stained by the EDTA technique inside FCs was also described, although the presence of this material was not confirmed, by RNase digestion or uridine incorporation (Risueño et al., 1982). The differences between those data and the results pre-sented here could be explained, because only after long exposure times has autoradiographic labelling been obtained on FCs, as reported in Ehrlich tumor cells (Thiry and Goessens, 1991). The fact that the RNA detected on FCs could constitute another RNA type such as the snRNAs does not seem possible because different immunocyto-chemical studies have demonstrated the presence of these snRNAs on DFC and GC but not on FCs (Puvion-Dutilleul et al., 1992; Raska et al. 1992; Testillano et al., 1992a,b). Most of the reported studies show the presence of RNA in DFC, but there are only a few reports showing RNA in FCs: by in situ/in vitro transcription applied to Lowicryl sections of Ehrlich tumor cells (Thiry and Goessens, 1991); by polyadenylate nucleotidyl transferase/immunogold in human Sertoli cells (Thiry, 1993a,b); or by in situ hybridization in Ehrlich or actinomycin D-treated HeLa cells (Thiry and Thiry-Blaise 1989; Puvion-Dutilleul et al., 1992; Raska et al., 1992). In Ehrlich tumor cells, the presence of some RNA-containing material on FCs was also reported using RNase/gold (Thiry, 1988) or RNase treatment (Yasuzumi and Sugihara, 1965).

The presence of rRNA in the DFC and GC of the plant nucleolus demonstrated by in situ hybridization experi-ments correlates with the data on human and animal cells. In our data, the hybridization signal found in the FCs rep-resents RNA as was proven by the lack of correlation between the labelling on the FCs and the labelling on cyto-plasmic vacuoles that constitutes background. The signal in FCs is very low but the labelling density is many times higher than that of the corresponding background. The hybridization labelling found on nucleoplasm is also very low but clearly represents the very fast passage of these products from nucleolus to cytoplasm.

Based on the statistical studies, in which we have considered the labelling density of the cytoplasmic vacuoles as an internal control of the background, the regression curve obtained with a correlation coefficient of −0.43 with 14 degrees of freedom clearly indicates that the labelling density in both structures (FCs and cytoplasmic vacuoles) are not correlated. These data suggest that the hybridization signal in FCs can be considered significant and does not correspond to the background.

Thus, three different classes of labelling on FCs have been found and after counting the label almost 50% of fib-rillar centres contain rRNA in the centre and/or in the periphery. Two different interpretations can be made for the presence of in situ hybridization labelling at the periphery of the FCs, which also covers the DFC: they could be either fibres of RNA synthesized in the DFC that enter into the FCs, or vice versa. The hybridization signal on the centre of FCs is not very significant because the gold particles usually appear isolated in these cases.

With in situ hybridization using 18 and 28 S DNA probes no rRNA is reported to be detected in the interior of FCs of HeLa cells (Puvion-Dutilleul et al., 1991a,b, 1992). When DNA probes containing the external transcribed spacer were used the peripheral region of FCs appeared labelled in studies with Ehrlich and HeLa cells (Thiry and Thiry-Blaise, 1989; Puvion-Dutilleul, 1991a,b, 1992). The difference from our results using 18 S and 25 S rRNA probes may be associated not only with the difference in materials used, but also with using RNA probes, which make the signal more abundant, as is expected in RNA/RNA hybridization (Mc Fadden et al., 1988).

When we used probes encompassing mature 18 and 25 S regions of the transcriptional unit we were able to detect not only the mature rRNAs but also pre-rRNAs. In addition we were able to detect hybridization labelling on FCs. Finally, when we combined in situ hybridization with MA and EDTA we were able to find stained fibrils showing hybridization signal, which provided additional data on the presence of RNA on FCs.

On the other hand, recently, the presence of rDNA has been detected by in situ hybridization and confocal microscopy of Pisum sativum, which seems to cover areas bigger than FCs (Highett et al., 1993), this suggests that the DNA detected on DFC could be rDNA. In plant nucleoli there are no chromatin inclusions except for those that appear in the interior of heterogeneous FCs. DNA has been local-ized on DFC and on FCs (Martin and Medina, 1991; Olmedilla et al., 1991, 1992; Risueño, 1993; Risueño et al., 1991; Testillano et al. 1991, 1992a). Therefore, in this mate-rial the argument based on finding DNA only on FCs to sup-port transcription taking place only in these structures was not confirmed by those results. Moreover, although RNA polymerase I has not been directly localized by immunocy-tochemistry there are some data that seem to indicate that this enzyme can be localized on both fibrillar components in plants (Martin and Medina, 1991). The RNA polymerase I transcription factor UBF (upstream binding factor) has been localized on different phylogenetically separate species (including Allium cepa) in the DFC (Rodrigo et al., 1992). Finally, nucleolar snRNAs and fibrillarin, known to play an active role at early stages of maturation, to date have not been localized on FCs in plant material (Testillano et al., 1992b,c, 1993). These data are in favour of the hypothesis that ribosomal transcription can start at the periphery of FCs, but the possibility cannot be excluded that the DFC close to the FCs also plays a role in this process. To confirm this hypothesis, experiments using in situ hybridization with RNA probes for the transcribed external spacer in order to localize these earlier transcripts, and different rDNA probes to localize the corresponding genes, are in progress.

We thank Prof. Dr Schweizer, Institute of Botany, University of Vienna, for providing us with the E. coli strains containing the Arabidopsis rRNA genes; Mr J. Blanco for his technical assis-tence with the photographic work, Dr A. Moreno (Inst. Acústica, CSIC, Madrid), Dr R. González (Est. Exp. Zaidín, CSIC, Granada) and Dr J. Renau-Piqueras (Hospital La Fe, Valencia) for the statistical work, and Mrs B. Ligus-Walker for revision of the Eng-lish. This work was supported by projects of the DIGICYT PB87-0332-C02-01/PB92-0079-C03-01 and Spanish-Austrian interna-tional cooperation.

Results recently obtained in plant cells by different cyto-chemical and immunocytochemical methods demonstrate the presence of DNA and DNA/RNA hybrids in regions of the DFC near the FCs (Risueño and Testillano, 1993, invited review, Micron; Testillano, Gorab and Risueño, 1993, J. Histochem. Cytochem., in press). These data sup-port the hypothesis presented in this paper that the periph-ery of the FCs and/or regions of the DFC near the FCs are involved in transcription.