Subcellular distribution of Ro ribonucleoprotein complexes and their constituents

Autoantibodies directed against the Ro/SS-A and the La/SS-B antigens are commonly found in sera of patients suffering from systemic lupus erythematosus (SLE) or from Sjögren’s syndrome. They have proven to be useful diagnostic markers of the disease and effective tools in the analysis of the antigens and the dissection of the complexes in which these antigens are present. In most human cell types studied so far the Ro ribonucleoprotein particles (RNPs) contain at least three proteins: the La protein, which is involved in the correct termination of RNA polymerase III transcription (Gottlieb and Steitz, 1989a,b), and the Ro RNP-specific proteins of 52 kDa (Ro52) and 60 kDa (Ro60) (Wolin and Steitz, 1984; Ben-Chetrit et al., 1988). In several hemopoietic cell types a different set of Ro proteins was detected. Red blood cells contain a 60 kDa and a 54 kDa protein, which are antigenically distinct from Ro60 and Ro52, respectively (Rader et al., 1989; Itoh et al., 1990). More recently, a novel 52 kDa Ro protein was detected in human platelets, which further increases the heterogeneity found in human Ro RNP complexes (Itoh and Reichlin, 1991). Calreticulin, formerly described as a Ro RNA-associated protein (Lieu et al., 1988; McCauliffe et al., 1990), was shown recently not to be a Ro auto antigen (Rokeach et al., 1991; Pruijn et al., 1992). Besides the proteins mentioned above, Ro RNPs may contain additional components, as the presence of single copies of the Ro proteins, together with the La protein and one RNA molecule, cannot account for the molecular masses of 230 to 350 kDa of Ro RNPs observed in gel filtration (Boire and Craft, 1990).

The RNA component of Ro RNPs in human cells consists of one out of four different small (84-112 nucleotides) polymerase III-transcribed RNAs called hY1, hY3, hY4 and hY5 RNA (hY2 is a degradation product of hY1). The distribution of these Y RNAs was also found to be heterogeneous. While nucleated human cells contain all four hY RNAs, albeit in different ratios (Pruijn et al., unpublished), red blood cells contain only hY1 and hY4 (O’Brien and Harley, 1990), while in platelets only hY3 and hY4 could be detected (Itoh and Reichlin, 1991). In other species the number of Y RNAs varies from 4 (e.g. bovine) to only 2 (e.g. mouse and duck) (reviewed in Slobbe et al., 1991a). The predicted secondary structure of the hY RNAs shows extensive base pairing between the highly conserved 5′ and 3′ ends. The Ro60 protein binds to the terminal part of this stem structure (Wolin and Steitz, 1984; Pruijn et al., 1991), while the 52 kDa Ro protein does not bind directly to the RNA but most likely associates to the particle via protein-protein interactions with Ro60 (Slobbe et al., 1992). Unlike most RNA polymerase III transcripts the Y RNAs are not subjected to post-transcriptional modification at their 3′ ter-minus (Wolin and Steitz, 1983), hence the La protein is thought to be a stable component of the Ro RNPs. In addition, the hY5 RNP appears to contain an additional unique antigenic determinant, which is absent in other Ro RNP particles (Boire and Craft, 1989).

Although the intermolecular interactions in the Ro RNPs have been studied in some detail now (Slobbe et al., 1992; Pruijn et al., 1991), the subcellular distribution of these particles is still controversial. Ro RNPs have been localized in the nucleus as well as in the cytoplasm (Clark et al., 1969; Gaither et al., 1987). Studies using anti-Ro antibodies in immunofluorescence report both nuclear and cytoplasmic staining (Ben-Chetrit et al., 1988; Slobbe et al., 1991b) while others show only nuclear staining (Harmon et al., 1984; Lopez-Robles et al., 1986), or a predominantly cyto-plasmic localization (Alspaugh and Maddison, 1979; Hen-drick et al., 1981; Bachmann et al., 1986). To unambiguously establish the subcellular distribution of the Ro RNPs and their constituents at the molecular level, we have fractionated human Jurkat and mouse 3T3 cells into a cyto-plasmic and a nuclear fraction, and analyzed these fractions separately. To avoid nuclear leakage, which is known to occur during mechanical fractionation, we have used the cell enucleation procedure described by Zieve et al. (1988). The data presented in this paper indicate that the La protein and the Ro proteins are present in both the cytoplasm and the nucleus, while the Y RNAs are exclusively cytoplasmic. Furthermore, we show that the majority of the human Y RNAs are stably associated with Ro60 and Ro52 as well as with the La protein. In the nucleus Ro60 and Ro52 are not associated with Y RNA. These results suggest a model for the assembly of Ro RNPs and strongly imply a cytoplasmic function for these complexes.

Mouse 3T3 cells and human Jurkat, HeLa and HEp-2 cells were cultured in DMEM supplemented with 10% (v/v) fetal calf serum. Jurkat cells were maintained at a density of about 0.8×10⁶ cells per ml, while the other cell types were grown in monolayer cultures to about 80% confluency before use.

Prior to immunofluorescence, cells were fixed in (i) methanol (−20°C) for 5 minutes followed by rinsing with acetone or (ii) 3% paraformaldehyde in phosphate buffered saline (PBS) for 20 min-utes at RT followed by a 5 minute incubation in 0.2% Triton X-100 in PBS. After washing in PBS, fixed cells were incubated with affinity-purified antibodies for 30 minutes, then washed with PBS and incubated with FITC-labeled rabbit anti-human IgG (Dako, F123) for 30 minutes. Finally cells were washed and pho-tographed under fluorescent illumination.

Human Jurkat and mouse 3T3 cells were fractionated into cyto-plasmic and nuclear fractions using the enucleation method essen-tially as described by Zieve et al. (1988). Briefly, about 25×10⁶ cells were treated with cytochalasin B (10 μg/ml) for 10 minutes before enucleation. Cells were then concentrated into 3 ml of 12.5% Ficoll in DMEM and layered on a Ficoll step gradient in DMEM. After centrifugation at 37°C for 90 minutes at 130,000 g (SW40 rotor) cytoplasmic and nuclear fractions were collected. The nuclear contamination of the cytoplasmic fraction was tested by Hoechst (33258) staining. Cellular fractions were resuspended in 400 μl of PBS and sonicated 4 times for 15 seconds each. Debris was removed by centrifugation at 14,000 g for 5 minutes.

Protein A-agarose beads were coated with the appropriate anti-bodies by rotation for at least 1 hour in 500 μl IPP₅₀₀ (10 mM Tris-HCl, pH 8.0, 0.5 M NaCl, 0.1% Tween-20, 0.1% NP40) fol-lowed by washing twice with IPP₅₀₀ and once with IPP₁₅₀ (10 mM Tris-Cl, pH 8.0, 0.15 M NaCl, 0.1% Tween-20, 0.1% NP40).

The coated beads were rotated with the cellular extract in 500 μl IPP₁₅₀ for 1 hour at RT. Precipitated RNA was isolated by washing 3 times with 1 ml of IPP₁₅₀ after which the beads were resus-pended in 400 μl of IPP₁₅₀ containing 0.5% SDS and phenol/chloroform extracted. After ethanol precipitation the RNA was dissolved in 10 μl of sample buffer (80% formamide/1 mM EDTA/0.025% bromophenol blue) and loaded on a 10% poly-acrylamide-8 M urea gel.

RNA was size-fractionated on 10% polyacrylamide-urea gels and transferred to Hybond N⁺ filters by electroblotting at 3 V/cm in 0.025 M phosphate (pH 6.5) for 2 hours. Filters were air dried and baked for 2 hours at 80°C. Hybridizations were performed overnight at 65°C in 6× SSC/5× Denhardt’s solution/100 μg/ml denatured herring sperm DNA. Labeled anti-sense hY RNA transcripts or random primed (Feinberg and Vogelstein, 1983) hY or U1 DNA fragments were used as probes. Following hybridization filters were washed twice at 65°C in 0.2× SSC. Filters were exposed to X-ray film at −70°C overnight.

Antibodies against either Ro60, Ro52 or the La protein were iso-lated from patient sera by affinity column chromatography using the respective recombinant proteins (Slobbe et al., 1992) coupled to CNBr-activated Sepharose 4B as antigens. Eluted antibodies were checked for reactivity and specificity by western blot analysis and precipitation of in vitro translated [³⁵S]methionine-labeled Ro60, Ro52 and La proteins. Affinity-purified antibody preparations did not show any cross-reactivities.

Cytoplasmic and nuclear enucleation extracts were run on standard 10% SDS-PAGE gels and blotted to nitrocellulose. After blocking with PT (PBS/0.5% Tween-20) containing 3% BSA for 1 hour, blots were incubated with diluted sera in PT-1% BSA for 2 hours. Filters were washed 3 times with PT and incubated with ¹²⁵I-Protein A (1 μCi/100 ml) in PT-1% BSA for 1 hour. After washing for 30 minutes in PT and for 30 minutes in water the blots were exposed to X-ray film at −70°C. The relative amount of Ro and La protein was estimated by densitometric scanning.

Reconstitution of ³²P-labeled Y RNA-protein complexes was per-formed as described (Slobbe et al., 1992), using equivalent amounts of cytoplasmic or nuclear extract.

In view of the controversial data described in previous studies, the subcellular distribution of the Ro antigen detected by immunofluorescence was likely to depend on the fixation method used. We therefore applied two different commonly used fixation methods (methanol/acetone and paraformaldehyde) on HEp-2 and HeLa cells prior to immunofluorescence microscopy. Differences both in intensity and in localization of Ro60 and La fluorescence sig-nals could be observed. After methanol/acetone fixation affinity-purified, monospecific antibodies directed against Ro60 gave a speckled nuclear and relatively strong cytoplasmic staining (Fig. 1A,B), while the fixation with paraformaldehyde gave a much brighter, diffuse nuclear fluorescence and hardly any cytoplasmic staining (Fig. 1C,D). Apart from the cytoplasmic staining essentially the same patterns were observed using monospecific antibodies against the La protein (Fig. 1E-H). Obviously, a precise determination of the localization of Ro60 and La proteins is not feasible using immunofluorescence.

As an alternative procedure to determine the subcellular localization of the Ro RNPs, human Jurkat cells and mouse 3T3 fibroblasts were separated into a nuclear and a cytoplasmic fraction using cell enucleation. This technique has previously been shown to give a reliable separation of cytoplasmic and nuclear material (Zieve et al., 1988). The efficiency of the enucleation was monitored by staining part of the resulting fractions with Hoechst (33258) reagent, which specifically stains the nuclei. While the nuclear contamination of the cytoplasmic fraction was usually between 5 and 10% for the human cells, the mouse 3T3 cytoplas-mic fraction contained hardly any nuclear material (data not shown). Extracts of the enucleation fractions were used to determine the distribution of the Y RNAs and their asssociation with proteins. The Ro RNP-specific Y RNAs appeared to be exclusively present in the cytoplasmic fraction of enucleated Jurkat and 3T3 cells (Fig. 2A). Despite the only slight contamination of the cytoplasmic fraction with nuclear material, hybridization with an U1 snRNA-(Fig. 2A) or U5 snRNA-(not shown) specific probe, which were used as controls, revealed an unexpected, approximately equal distribution of U snRNA between the cyto-plasm and the nucleus of Jurkat cells. A similar distribution of U1 snRNA was observed after enucleation of other human cells, such as Molt-4, CEM, HL-60, K-562 and U937, the fibroblast cell line HF-V32 and primary human fibroblasts (data not shown). In contrast to the human cells, the enucleation fractions of the mouse 3T3 cells showed the expected distribution of U1 snRNA; i.e. the bulk being in the nuclear fraction.

Immunoprecipitation using monospecific anti-La or anti-Ro60 antibodies revealed that the Y RNAs are not only associated with Ro60, but these RNAs could also be efficiently precipitated using antibodies against the La protein (Fig. 2B). Similar results were obtained for hY3 and hY4 RNA, although the level of association of these RNAs with the La protein showed some variation between experiments (not shown).

The cytoplasmic localization of the Ro RNPs was prob-ably not due to the treatment of the cells with cytochalasin B prior to the enucleation procedure, as the immunofluorescence of 3T3 cells using anti-Ro60 or anti-La antibodies showed no marked changes in the distribution of Ro60 or the La protein after incubation with this drug (not shown).

From the experiments described above we conclude that the Ro RNPs are located predominantly in the cytoplasm.

The Y RNAs are not processed at their 3′ termini and there-fore retain the oligouridylate stretch representing the major La protein binding site. Most if not all of the Y RNAs from Jurkat cells indeed seem to be associated not only with Ro60, but also with the La protein (Fig. 2B). To determine whether the Y RNAs were quantitatively associated with these proteins in human cells, total HeLa cell S100 extract was subjected to several consecutive immunoprecipitations using monospecific antibodies directed against either the La or Ro proteins. Northern blot analysis of the RNA isolated from the immunoprecipitates revealed that the Y RNAs could be efficiently depleted from the extract using either anti-Ro60 or anti-La antibodies, as illustrated by the absence of hY1 and hY5 RNA in the final post-immuno-precipitate supernatants (Fig. 3, top and middle). Four sub-sequent immunoprecipitations with monospecific anti-Ro52 antibodies also removed most of the Y RNA from the extract, leaving only a small portion in the supernatant (Fig. 3, bottom). Hybridization with a U1 snRNA-specific probe confirmed the specificity of these precipitations and indicated that the RNA isolated from the final post-immuno-precipitate was intact. These data strongly suggest that the majority of the Y RNAs are associated with Ro52, while almost all Y RNAs are associated with Ro60 and the La protein.

As shown above the majority of the hY RNAs appear to be present in the cytoplasm being associated with the La and the Ro proteins. Paradoxically, the studies using immunofluorescence for the detection of the Ro RNP-associated proteins suggest a primarily nuclear localization (Fig. 1A-D). Since the import of Ro RNP proteins into the nucleus during fixation of the cells seems very unlikely it was investigated whether all Ro60 was complexed with the Y RNAs or whether free (i.e. non-Y RNA-bound) Ro60 exists within the cell nucleus. This could explain the discrepancy between the nuclear localization observed using immunofluorescence and the cytoplasmic localization of the Ro RNPs obtained by the cell fractionation. To this end western blots of nuclear and cytoplasmic enucleation fractions of 3T3 and Jurkat cells were probed with affinity-purified antibody directed against the La, the Ro60 or Ro52 proteins. The results revealed that free Ro proteins are indeed present within the cell nucleus and represent about one third of the total amount in Jurkat cells, while in mouse 3T3 cells the distribution of Ro60 between cytoplasm and nucleus is about equal. For the La protein about one third of the total amount resides within the cell nucleus of both Jurkat and 3T3 cells (Fig. 4). In the enucleation fractions of mouse 3T3 cells the Ro52 protein was not detectable (not shown). The absence of Ro52 in rodent cells is in agreement with our previous results, which indicated that the Ro52 protein could only be detected by immunological techniques in cells of primate origin (Slobbe et al., 1991b). Free, non-Y RNA-bound, Ro60 and Ro52 proteins could also be detected by a different approach. When a total Jurkat cell extract was depleted of Ro RNPs by several anti-La immunoprecipitations (removing all La protein) about 30% of Ro60 and 40% of Ro52 remained in the supernatant (Fig. 5). These results, together with the presence of free nuclear Ro protein indicate that a significant amount of Ro protein is not associated with Ro RNPs.

To determine whether the non-Y RNA-associated Ro60 was capable of binding Y RNA, total cellular extract of Jurkat cells was first depleted of all Ro RNPs by several immunoprecipitations using affinity-purified anti-La anti-bodies. As illustrated by the hybridization to hY1 and hY5 RNAs (Fig. 6A) the Y RNAs are indeed completely removed from the extract, while the U1 snRNA, which is reported to be complexed with the La protein for only a small portion (Keene et al., 1987; see also Fig. 3), remains largely in the post-immunoprecipitate (P.I.) supernatant. To the depleted extract, containing no endogenous Y RNAs, in vitro-transcribed ³²P-labeled hY1 or hY5 was added. After incubation these RNAs could be efficiently precipi-tated using antibody against Ro60 (Fig. 6B), indicating that the Ro RNP-depleted total cellular extract still contained Ro60 able to bind Y RNAs.

The enucleation fractions of the mouse 3T3 cells, which showed the expected distribution of the U1 snRNA, were used to determine whether the free Ro60 was located in the cytoplasm, in the nucleus or in both compartments. To remove the Ro RNP-associated Ro60, the cytoplasmic extract was immunodepleted by anti-La precipitations as described before. In contrast to the human Jurkat and HeLa cells, the cytoplasmic fraction of the mouse 3T3 cells could not be completely depleted of Ro RNPs using anti-La immunoprecipitations. About 5% of the mY1 RNA remained after anti-La-depletion, suggesting that in these cells a small portion of the Ro RNPs is not associated with the La protein (Fig. 6C).

The addition of in vitro-transcribed ³²P-labeled hY1 RNA to a non-depleted 3T3 cytoplasmic extract, followed by immunoprecipitation with monospecific anti-Ro60 anti-bodies, resulted in a very efficient precipitation of the labeled RNA (Fig. 6D, lane 1). The signal became much weaker (lane 2) when this extract was first largely depleted of Ro RNPs by anti-La immunodepletion, as shown in Fig. 6C. Obviously, not much free Ro60 is available in the cyto-plasm for binding labeled hY RNA. This indicates that the strong signal obtained in lane 1 derives mainly from Ro RNP-associated Ro60 molecules, exchanging from an endogenous non-labeled Y RNA to the added labeled hY1 RNA. A further treatment of the La-depleted extract by an anti-Ro60 precipitation prior to the addition of ³²P-labeled hY1 RNA diminished the signal to background levels (lane 3). When labeled hY1 RNA was added to 3T3 nuclear extracts, the signal was stronger (lane 4) than the signal obtained with the La-depleted cytoplasmic extract (lane 2). If we assume that the signal obtained with the not com-pletely Ro RNP-depleted cytoplasmic extract (lane 2) results from an exchange of Ro60 from Ro RNPs that are not associated with the La protein (Fig. 6C), then the results presented above strongly suggest that the free Ro60 capable of Y RNA binding is largely, if not completely, located within the cell nucleus. Similar results were obtained using Jurkat cell enucleation fractions (not shown).

The localization of antigens using the immunofluorescence technique critically depends on the fixation method used. In this study fixation of cells following two commonly used procedures gave clear qualitative as well as quantitative differences when using antibodies against Ro60 and the La protein. The use of conventional mechanical cell fractionation in order to determine the subcellular localization of particular components is also prone to artifacts and leads to massive nuclear leakage (Zieve and Penman, 1976). Alternatively, it has been shown that cell enucleation of cytochalasin B-treated mouse L929 cells results in a reliable separation of cytoplasm and intact nuclei, showing the expected distribution of, for example, the U snRNAs and their precursors (Zieve et al., 1988). The relatively short exposure of the intact cells to cytochalasin B during the enucleation procedure has also no detectable effects on distribution of La or Ro RNPs judged from the unchanged immunofluorescence patterns when using anti-La or anti-Ro60 antibodies. The mouse 3T3 cells used in our experiments showed the expected distribution of U1 snRNA between nucleus and cytoplasm, indicating that these cells can be efficiently enucleated giving rise to a reliable cyto-plasm-nucleus separation. Remarkably, the human cells used in this study, including several cell lines and primary cells, showed an approximately equal distribution of U1 and U5 (Jurkat) snRNA between the cytoplasm and the nucleus, in spite of the minimal contamination of the cytoplasmic fraction with nuclei. Although we cannot completely exclude the possibility that nuclear leakage of (certain) RNPs does occur in human cells during fractionation we assume that the observed U snRNA distribution is an accu-rate representation of the situation in these cells.

The northern blot analysis of the human Jurkat and the mouse 3T3 enucleation fractions showed that the Y RNAs are located exclusively in the cytoplasm. The human Y RNAs were shown to be quantitatively associated with the Ro60 as well as the La protein. Furthermore, it is demon-strated that the human Y RNAs can be efficiently precipi-tated using affinity-purified antibodies directed against the Ro52 protein. Other studies describing the composition of the Ro RNPs report that the Y RNAs are only partially associated with the La protein (Mamula et al., 1989; Boire and Craft, 1990), while in human erythrocytes the Y RNAs could not be precipitated by anti-La sera at all (O’Brien and Harley, 1990). It is possible that these discrepancies are due to the use of different cell types or culturing conditions or to the fine specificity of the antibodies used.

In contrast to the predominantly nuclear localization of Ro60 revealed by immunofluorescence (see also Fig. 1), our enucleation studies indicate that about 50-70% of the Ro60 molecules are cytoplasmic and associated with Y RNAs. This discrepancy can be explained by assuming that most of the antigenic determinants of Ro60, associated with both Y RNA and Ro52, are not accessible after the fixation procedure. In contrast, the free, non-Y RNA-bound Ro60 antigen in the nucleus might still be recognized sufficiently to produce an almost exclusive nuclear immuno-fluorescent staining. An alternative, but in our view less likely possibility is to assume selective leakage of Ro RNPs from the cytoplasm during fixation.

Although the majority of Ro60 in the nucleus is not associated with Y RNA, reconstitution experiments showed that at least part of this Ro60 fraction is capable of binding Y RNA. The present data do not exclude the possibility that Ro60-Ro52 complexes exist in the nucleus. The relatively high percentage of nuclear, non-Y RNA-associated Ro protein could be indicative for additional functions of these proteins besides their role in Ro RNPs. The presence of zinc finger-like structures, similar to those found in several DNA-binding proteins, in both Ro52 and Ro60 (Deutscher et al., 1988; Ben-Chetrit et al., 1989; Chan et al., 1991; Itoh et al., 1991), together with the high degree of homology between Ro52 and the mouse rpt-1 and the human ret trans-forming protein support this idea (Chan et al., 1991).

The data presented above suggest the following model for the biosynthesis of Ro RNPs. The newly synthesized Ro proteins are, like the La protein, actively transported to the nucleus (Simons et al., unpublished) to a pool of free proteins. The newly synthesized Y RNAs, which become La-bound during transcription termination, are subsequently bound by Ro60 and Ro52. This Ro RNP complex is then rapidly and quantitatively transported from the nucleus to the cytoplasm. Although the function of Ro RNPs has not yet been determined, their cytoplasmic localization together with the high evolutionary conservation of both the Ro pro-teins and their cognate RNAs point to an important role in cytoplasmic processes.

We thank Dr G. W. Zieve for helpful discussions. This investigation was carried out with financial aid from ‘Het Nationaal Reumafonds’ of the Netherlands and was supported in part by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for Scientific Research (NWO). The research of Dr G. J. M. Pruijn has been made possible by a fellowship of the Royal Netherlands Academy of Arts and Sciences.