Localization of poly(A)-binding protein 2 (PABP2) in nuclear speckles is independent of import into the nucleus and requires binding to poly(A) RNA

Poly(A) tails have long been established as a hallmark of messenger RNA. With the single exception of histone mRNAs, all known eukaryotic protein-coding transcripts have a sequence of polyadenylic acid at the 3′ end (Adesnik et al., 1972). In addition to playing a key role in the subsequent life of mRNAs, the reaction of poly(A) tails with oligo (dT) or oligo (U) facilitates their biochemical isolation and in situ analysis.

In situ hybridization experiments using oligo (dT) probes have been performed in a variety of cell types at both light and electron microscopy level (Carter et al., 1991, 1993; Visa et al., 1993; Huang et al., 1994; Puvion and Puvion-Dutilleul, 1996; Fay et al., 1997). These studies have consistently showed that poly(A) RNA is detected throughout the cytoplasm and nucleoplasm, excluding nucleoli. Intriguingly, there are domains within the nucleoplasm where the concentration of poly(A) RNA is higher. These domains are designated nuclear speckles and are shown by electron microscopy to correspond to clusters of interchromatin granules surrounded by perichromatin fibrils (Spector et al., 1991). Perichromatin fibrils are enriched in heterogeneous high molecular mass RNA (hnRNA), they incorporate [3H]uridine after pulses for as short as 2 minutes, they are RNase-sensitive and they disappear when cells are pre-treated with transcription inhibitors (reviewed by Fakan and Puvion, 1980; Spector, 1993). Thus, it is generally accepted that perichromatin fibrils represent ribonucleoprotein complexes containing nascent pre-mRNAs (Fakan, 1994). The clusters of interchromatin granules are composed of particles with a mean diameter of 20-25 nm, they enlarge when cells are pre-treated with transcription inhibitors, and they contain splicing snRNPs (small nuclear ribonucleoprotein particles) and protein splicing factors (reviewed by Spector, 1993; Misteli and Spector, 1998). In contrast to perichromatin fibrils, clusters of interchromatin granules show little labeling following incorporation of [3H]uridine (Fakan and Bernhard, 1971; Fakan and Nobis, 1978), arguing that they are not sites of active transcription. At present, several lines of evidence suggest that most splicing occurs co-transcriptionally on perichromatin fibrils, and clusters of interchromatin granules represent reservoirs, from which splicing factors are recruited to the nascent transcripts (for recent reviews see Misteli and Spector, 1998; Sleeman and Lamond, 1999).

Pulse-chase studies following incorporation of [3H]uridine show that there is a delay of about 20 minutes between transcription and the appearance of mRNA in the cytoplasm. However, polyadenylation occurs approximately 7 minutes before the mRNA is incorporated into polysomes, suggesting that after addition of the poly(A) tail mRNAs are rapidly exported to the cytoplasm (reviewed by Lewin, 1980). It was therefore quite surprising to observe that poly(A) RNA persists localized in the nuclear speckles when cells are pre-treated with transcription inhibitors for several hours (Huang et al., 1994). This result implies that the speckles contain a population of stable poly(A) RNA that is distinct from protein-encoding mRNAs, which are rapidly exported to the cytoplasm (Huang et al., 1994). Consistent with this view, both spliced and unspliced RNAs for numerous genes have been co-localized outside or at the periphery of the nuclear speckles, but not in their interior (Zhang et al., 1994; Huang and Spector, 1991, 1995; Xing et al., 1993, 1995; Dirks et al., 1997; Lampel et al., 1997; Misteli et al., 1998; Smith et al., 1999a; Jolly et al., 1999). In contrast, transcripts encoded by herpes simplex and adenoviral genes have been shown to migrate from the sites of transcription to clusters of interchromatin granules, where they accumulate progressively during the course of infection (Puvion-Dutilleul et al., 1992, 1997; Puvion and Puvion-Dutilleul, 1996; Bridge et al., 1996; Aspegren et al., 1998). What remains to be elucidated is whether the viral mRNAs detected in the speckles are in transit to the cytoplasm or rather represent a sub-population earmarked to be retained inside the nucleus.

Messenger RNAs in the nucleus associate tightly with several proteins forming ribonucleoprotein particles (hnRNPs or mRNPs). At present it is generally assumed that RNA-binding proteins play a critical role in targeting the mRNP particles to the cytoplasm, as they contain signals that are specifically recognized by export receptors (reviewed by Lee and Silver, 1997; Nakielny et al., 1997; Stutz and Rosbach, 1998). Whether RNA-binding proteins are also involved in targeting poly(A) RNAs to the nuclear speckles is not known. Within the nucleus, a major poly(A) RNA-binding protein is poly(A) binding protein 2, PABP2 (reviewed by Kühn and Wahle, 1997; Zhao et al., 1999). PABP2 is an abundant protein that binds with high affinity to nascent poly(A) tails, stimulates processive poly(A) addition and controls the size of the tail to approximately 250 nucleotides in length (Wahle, 1991, 1995; Bienroth et al., 1993). PABP2 localizes to clusters of interchromatin granules (Krause et al., 1994), and in this work we asked whether PABP2 mediates the association of poly(A) RNA with the speckles. The results show that concentration of PABP2 in the speckles is strictly dependent upon binding to poly(A) tails, suggesting that PABP2 is localized to the speckles as a consequence of its binding to poly(A) RNA. This clearly contrasts with splicing factors, which assemble into speckles by a mechanism that apparently involves protein-protein interactions and is independent of nascent transcripts. Taking into account that at least some splicing proteins remain associated with mature mRNA (Cáceres et al., 1998), these data raise the question of whether protein splicing factors are involved in the recruitment of poly(A) RNA to the speckles.

HeLa cells were cultured as monolayers in modified Eagle’s medium (MEM) supplemented with 10% fetal calf serum (Gibco BRL, Paisley, Scotland).

Cells grown on 10×10 mm coverslips were rinsed briefly in phosphate buffered saline (PBS) and fixed in 3.7% formaldehyde (freshly prepared from paraformaldehyde) diluted in either PBS or HPEM buffer (30 mM Hepes, 65 mM Pipes, 10 mM EGTA, 2 mM MgCl₂, pH 6.9) for 10 minutes at room temperature. The cells were then permeabilized with 0.05-0.1% Triton X-100 in PBS (or HPEM buffer) for 10 minutes at room temperature.

For immunofluorescence, the cells were rinsed in PBS containing 0.05% Tween 20 (PBS-Tw), incubated for 1 hour with primary antibodies diluted in PBS, washed in PBS-Tw and incubated with appropriate secondary antibodies for 30 minutes. After washing in PBS-Tw, the samples were mounted in VectaShield (Vector Laboratories, Peterborough, UK). Secondary antibodies conjugated to FITC or TexasRed were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA).

In situ hybridization to detect poly(A) RNA was performed as described previously (Gama-Carvalho et al., 1997).

For drug experiments, fresh culture medium was added and allowed to equilibrate for 1 hour. The cells were then incubated in the presence of 5 μg/ml actinomycin D, 75 μM 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB) or 50 μg/ml cordycepin. Stock solutions were prepared (0.5 mg/ml actinomycin D in ethanol, 11 mM DRB in ethanol, 2 mg/ml cordycepin in water) and stored at –20°C. Controls were performed in the presence of 1% ethanol. All drugs were purchased from Sigma (Sigma Chemical Co., St Louis, MO).

Endogenous PABP2 was visualized using a polyclonal serum previously described (Krause et al., 1994). Affinity-purified antibodies were obtained using recombinant PABP2 expressed in E. coli from plasmid pGM10-PABP2 (Smith et al., 1999b). Additionally, the following antibodies were used: monoclonal antibody 4G3 directed against the U2 snRNP protein B′′ (Habets et al., 1989); monoclonal antibody 9H10 directed against hnRNP A1 protein (Burd and Dreyfuss, 1994); monoclonal antibody anti-SC35 splicing factor (Fu and Maniatis, 1990); monoclonal antibodies directed against CPSF-160 and CPSF-100 (Jenny et al., 1994), CstF-64 and CstF-50 (Takagaki et al., 1990); rabbit polyclonal antibodies specific for poly(A) polymerase, CPSF-73, and CPSF-30 (Barabino et al., 1997); and rabbit polyclonal antibodies anti-luciferase (Promega, Madison, WI).

To digest poly(A) RNA, the cells were fixed with 3.7% formaldehyde diluted in HPEM buffer for 10 minutes, permeabilized with 0.05% SDS in 150 mM NaCl, 12.5 mM EDTA, 100 mM Tris-HCl, pH 7.5, and incubated with 1,000 U/ml S7 (micrococcal) nuclease (Boehringer Mannheim, Mannheim, Germany) in 10 mM Tris, 25 mM CaCl₂, pH 7.5, for 1 hour at 37°C. The cells were then washed in PBS and fixed with 1% formaldehyde. Controls were performed omitting the enzyme.

The cDNA encoding bovine PABP2 was subcloned from the pGM10-PABP2 prokaryotic expression vector (Smith et al., 1999b) into the pEGFP-C1 protein fusion vector (Clontech, Palo Alto, CA). pGM10-PABP2 was cut with NdeI and the recessed 3′ ends were filled in. The linear plasmid was then cut with BamHI and the NdeI(filled in)-BamHI fragment encoding the cDNA of PABP2 was purified. The pEGFP-C1 was cut with SmaI and BamHI, and ligated to the cDNA fragment. The same strategy was used to subclone the cDNAs for PABP2 mutants.

In order to fuse PABP2 to luciferase, we have made use of plasmid pBluescript KS+LL/V. First, this plasmid was cut with HindIII and PstI, and the fragment encoding the luciferase gene was purified. This was subcloned into the HindIII and PstI sites of pSE280 (Invitrogen, Leek, The Netherlands). This plasmid was then cut with EcoNI and the recessed 3′ ends generated were filled in. Subsequently the linear plasmid was cut with BglII. The product containing the luciferase gene deleted in 31 nts at the 3´-end was purified and ligated to the cDNAs of PABP2 isolated as described above. Finally, the pSE280-luciferase-PABP2 vector was cut with HindIII and SalI, and the fragment containing the luciferase-PABP2 fusion gene was then purified and subcloned into the HindIII and SalI sites of pCMV5 (Andersson et al., 1989), thus generating the pCMV5-Luciferase-PABP2 vector. In order to produce the pCMV5-luciferase plasmid, the HindIII-PstI fragment encoding the luciferase gene was subcloned into the HindIII and PstI sites of pCMV5.

Final DNA constructs were purified using the Qiagen plasmid midi kit, and transfection of HeLa cells was performed using either the calcium phosphate method (Sambrook et al., 1989), Lipofectin (Gibco, BRL) or FuGene 6 (Boehringer Mannheim), according to the manufacturer’s instructions.

Fluorescently labeled samples were analyzed using the laser scanning microscope Zeiss LSM410 equipped with an Argon Ion laser (488 nm) to excite FITC fluorescence and a Helium-Neon laser (543 nm) to excite TexasRed fluorescence. For double labeling experiments, images from the same focal plane were sequentially recorded in different channels and superimposed.

Immunofluorescence microscopy using affinity-purified antibodies reveals PABP2 localized throughout the nucleoplasm with higher concentration in speckles (Fig. 1A), as previously reported by Krause et al. (1994). Double-labeling experiments using rabbit antibodies against PABP2 and the monoclonal antibody 4G3 directed against the U2snRNP-specific protein B′′ show that PABP2 co-localizes with snRNPs in nuclear speckles (Fig. 1A and B, arrows) but not in Cajal (coiled) bodies (Fig. 1A and B, arrowheads).

During mitosis, splicing snRNPs are excluded from the chromosomes and become predominantly diffuse in the cytoplasm with additional concentration in speckles that appear similar to clusters of interchromatin granules when observed with the electron microscope (Leser et al., 1989; Ferreira et al., 1994; Thiry, 1995). As depicted in Fig. 1C and D, analysis of PABP 2 distribution during mitosis also reveals a speckled staining pattern. Double labeling experiments further demonstrate that PABP2 co-localizes with snRNPs in the mitotic speckles (Fig. 1E and F).

Recently, it has been reported that all major 3′ end processing factors including cleavage stimulation factor (CstF), cleavage/polyadenylation specificity factor (CPSF), poly(A) polymerase (PAP) and PABP2 are detected at transcription sites (Schul et al., 1998a). However, with the single exception of PABP2 none of the other factors appears concentrated in the interior of the speckles during interphase under physiological conditions (Schul et al., 1998a,b; Chen et al., 1999). Here we confirm and extend those observations by showing that neither PAP nor any of the CPSF and CstF polypeptides accumulate in mitotic speckles (Fig. 2).

Double labeling experiments using anti-PABP2 antibodies and an oligo(U) probe show that PABP2 co-localizes with poly(A) RNA in speckles, both during interphase and mitosis (Fig. 3A and B). This finding could be explained if PABP2 piggybacks on poly(A) RNAs that are localized to the speckles. Alternatively, localization of PABP2 to the speckles could be an active process and the RNAs represent a passive cargo. An experimental approach to distinguish between these possibilities consists of determining whether localization of the protein to nuclear speckles requires binding to poly(A) RNA. To address this question, we started by analyzing the localization of PABP2 during nuclear biogenesis at the end of mitosis. As depicted in Fig. 3A, PABP2 is detected in the nucleus of daughter cells shortly after nuclear envelope assembly, even when the newly formed nuclei are largely devoid of poly(A) RNA (Fig. 3B). Since transcription does not start immediately after nuclear assembly (Ferreira et al., 1994), this implies that import of PABP2 to the nucleus occurs before the onset of RNA synthesis. Consistent with this view, PABP2 is localized in the nucleus of cells that completed mitosis in the presence of transcription inhibitors (Fig. 4A-D). As a control, double immunofluorescence was performed using antibodies specific for hnRNP A1 protein, because localization of this protein in the nucleus of post-mitotic cells requires ongoing RNA synthesis (Piñol-Roma and Dreyfuss, 1991). The results confirm that PABP2 is present in nuclei from which the hnRNP A1 protein is excluded (Fig. 4A-D). Next, we compared the nucleocytoplasmic distribution of hnRNP A1 and PABP2 in interphase cells treated with actinomycin D for 3-6 hours. Both hnRNP A1 and PABP2 are known to shuttle continuously between nucleus and cytoplasm (Piñol-Roma and Dreyfuss, 1992; Chen et al., 1999; Calado et al., 2000). Inhibition of transcription does not affect export of hnRNP A1 to the cytoplasm but prevents its import to the nucleus (Piñol-Roma and Dreyfuss, 1992). Consequently, the hnRNP A1 protein that is normally detected exclusively in the nucleus, appears in the cytoplasm (Fig. 4F). Similarly, export of PABP2 to the cytoplasm continues in the presence of transcription inhibitors (Calado et al., 2000). However, in contrast to hnRNP A1, PABP2 remains exclusively detected in the nucleus after prolonged treatment with actinomycin D (Fig. 4E), indicating that its import to the nucleus has not been blocked.

PABP2 is a small protein of approximately 33 kDa that may cross the diffusion channel of nuclear pore complexes. It is therefore possible that the observed entry of PABP2 into transcriptionally inactive nuclei occurs by passive diffusion. To address this issue, we analyzed the nucleocytoplasmic distribution of PABP2 fused to a non-nuclear protein, a form of firefly luciferase (∼60 kDa) containing a Leu?Val mutation that disrupts its peroxisomal targeting nucleoplasm of interphase cells where it shows a speckled distribution pattern similar to endogenous PABP2 (data not shown; see Calado et al., 2000). This chimera is exclusively detected in the nucleus of post-mitotic cells treated with actinomycin D (Fig. 5B and D). Since the fusion protein luciferase-PABP2 is too large (∼90 kDa) to enter the diffusion channel of nuclear pores (see Ohno et al., 1998), we conclude that at the end of mitosis import of PABP2 into the nucleus of daughter cells involves a carrier-mediated mechanism that is independent of the onset of RNA synthesis.

Although nuclear import of PABP2 occurs in the absence of transcriptional activity, the protein does not concentrate within signal (Gould et al., 1989). As expected for a relative small protein with no specific retention in any subcellular compartment, luciferase appears uniformly distributed throughout the cytoplasm and the nucleus (Fig. 5A). In contrast, the chimera luciferase-PABP2 is restricted to the speckles in the nucleus of post-mitotic cells treated with transcription inhibitors (cf Figs 3A, 4A,C). This implies that either speckles are not present in these nuclei, or PABP2 fails to localize there. To address this question we have used a monoclonal antibody directed against the splicing factor SC35, a well-known constituent of nuclear speckles (Fu and Maniatis, 1990; Spector et al., 1991). The results show that in non-treated early G₁ cells, PABP2 and SC35 co-localize in the nuclear speckles (Fig. 6A and C). After actinomycin D treatment, PABP2 no longer localizes in the nuclear speckles labeled by anti-SC35 antibody (Fig. 6B and D). This indicates that synthesis of RNA in newly assembled nuclei is necessary for recruiting PABP2 to the speckles.

Having established that PABP2 fails to localize to speckles in the absence of RNA synthesis, we next asked whether poly(A) tails are specifically involved in the subnuclear distribution of the protein. To address this question, cells were treated with cordycepin (deoxyadenosine), which prevents poly(A) synthesis but does not significantly affect the synthesis of nuclear hnRNA (reviewed by Lewin, 1980). Mitotic cells were incubated with cordycepin, double labeled for PABP2 and hnRNP A1, and analyzed at the early G₁ stage. Since A1 is excluded from the nucleus of post-mitotic cells that are transcriptionally inactive, we selected for early G₁ cells containing A1 in the nucleus. In the two daughter cells depicted in Fig. 7A and B, the A1 protein is present in the nucleus indicating that RNA synthesis was not blocked; in these cells, PABP2 appears diffuse throughout the nucleoplasm with no concentration in speckles. This suggests that targeting of PABP2 to the speckles does not depend on transcription, but specifically requires synthesis of poly(A) tails.

Possibly, PABP2 localizes in speckles as a consequence of its binding to the poly(A) tails of RNAs present in these structures. A prediction from this model is that removal of poly(A) tails should de-localize PABP2 from the speckles. Poly(A) tails can be enzymatically removed by treatment with an exoribonuclease specific for 3′-OH ends (see Lewin, 1980). Consistent to the model, digestion of cells with S7 (micrococcal) nuclease abolishes the speckled distribution of PABP2, without interfering with the structure of the nuclear speckles, as demonstrated by staining with anti-SC35 antibody (Fig. 7C-J).

According to the view that recruitment of PABP2 to the speckles requires binding to poly(A) RNAs, PABP2 mutants unable to bind to poly(A) are not expected to localize in speckles. To address this prediction, wild-type and mutant forms of PABP2 were fused to the green fluorescent protein (GFP) and transiently expressed in HeLa cells. In agreement with the data obtained using anti-PABP2 antibodies, the fusion GFP-PABP2 co-localizes with SC35 in speckles but remains diffuse in the nucleoplasm of post-mitotic cells exposed to actinomycin D (Fig. 8A,B,C). PABP2 contains a single RNA binding domain in its middle region (Nemeth et al., 1995) and it was previously shown that a double-point mutant (a tyrosine to alanine at position 175 and a phenylalanine to alanine at position 215) in this domain reduces significantly binding of the protein to poly(A) in vitro (Nemeth, 1998). This mutant (dm-PABP2) was fused to GFP and expressed in HeLa cells. The chimera GFP-dmPABP2 distributes uniformly throughout the nucleoplasm of cells that contain nuclear speckles labeled by anti-SC35 antibody (Fig. 8D and E) or poly(U) probe (Fig. 8F and G).

In addition, we have previously reported that a fusion protein GFP-ΔNPABP2 localizes to speckles, whereas the fusion proteins GFP-ΔCPABP2 and GFP-CPABP2 are not detected in speckles (Calado et al., 2000). Biochemical analysis revealed that ΔNPABP2 (which contains a deletion of amino acids 1-160) binds to poly(A) as efficiently as the wild-type protein, while ΔCPABP2 contains a deletion of amino acids 257-306 and binds to poly(A) about 1000-fold less than the wild-type protein and CPABP2 (which contains only amino acids 256-306) is devoid of the entire RNA-binding domain (Smith et al., 1999b; A. Nemeth and E. Wahle, personal communication). Thus, there is a tight correlation between ability to bind to poly(A) and localization in speckles.

In conclusion, the data show that PABP2 fails to localize to speckles either in the absence of poly(A) tails or when the protein has reduced binding affinity to poly(A). This strongly suggests that recruitment of PABP2 to the speckles is a consequence of its binding to poly(A) RNAs localized in these structures.

An intriguing feature of mammalian cell nuclei consists of domains termed clusters of interchromatin granules or nuclear speckles, which are enriched in stable poly(A) RNAs of unknown function (Carter et al., 1991; Visa et al., 1993; Huang et al., 1994). In addition to poly(A) RNA, these domains are enriched in PABP2, an abundant nuclear protein that binds with high affinity to nascent poly(A) tails, stimulating their extension and controlling their length (Wahle, 1991, 1995; Bienroth et al., 1993). In this work we show that binding of PABP2 to poly(A) tails is essential for localization of the protein to the speckles.

An additional conclusion from this work is that import of PABP2 to the nucleus is independent of localization of the protein to nuclear speckles. In fact, the data demonstrate that the nuclear import of PABP2 involves a carrier-mediated mechanism that is independent of ongoing RNA synthesis, whereas targeting to the speckles requires interaction with poly(A) RNA. Previous studies have shown that PABP2 contains a nuclear localization signal in its C-terminal domain and that it binds directly to transportin in a RanGTP-sensitive manner (Calado et al., 2000). As transportin is the nuclear transport receptor that mediates nuclear import of some shuttling hnRNP proteins (Pollard et al., 1996; Siomi et al., 1997), this finding suggests that PABP2 and hnRNP proteins share the same nuclear import pathway. It was therefore surprising to observe that the facilitated import of PABP2 to the nucleus is independent of ongoing RNA synthesis, whereas that of hnRNP A1 is blocked in the presence of transcription inhibitors (Piñol-Roma and Dreyfuss, 1991). This suggests that transcription-dependent nuclear import is determined by the cargo and not by the transport receptor. Consistent with this view, the nucleocytoplasmic distribution of transportin is not affected by transcription-inhibitors (Siomi et al., 1997).

When viewed with the electron microscope the nuclear speckles correspond to clusters of interchromatin granules surrounded by perichromatin fibrils, which represent nascent transcripts. During mitosis clusters of interchromatin granules are detected in the cytoplasm, while perichromatin fibrils disappear as a consequence of transcriptional inactivity (Leser et al., 1989; Ferreira et al., 1994; Thiry, 1995). Presumably, the mitotic clusters of interchromatin granules represent nuclear speckles that were released to the cytoplasm upon nuclear envelope disassembly at prophase. The mitotic speckles persist in the cytoplasm of daughter cells for some time after nuclear assembly, and we show here that they contain PABP2 and poly(A) RNA (Fig. 3). Thus, the poly(A) RNAs present in nuclear speckles from the mother nucleus are transmitted to the cytoplasm of daughter cells. Whether these RNAs will be eventually degraded in the cytoplasm or re-imported to the nucleus is not known. However, the finding that poly(A) RNA is not detected in the nucleus of post-mitotic cells before the onset of transcription (Fig. 3 and data not shown) suggests that most, if not all, nuclear poly(A) RNA is newly synthesized by the daughter cells.

Although the nature of poly(A) RNAs localized in nuclear speckles is as yet unclear, it is unlikely that they correspond to protein-encoding mRNAs in transit to the cytoplasm. In fact, [3H]uridine pulse-labeling experiments have demonstrated that nascent RNA can be chased to the cytoplasm within 20 minutes after transcription inhibition by actinomycin D (reviewed by Lewin, 1980); however, poly(A) RNA persists in the nuclear speckles for several hours after treatment of cells with transcription inhibitors (Huang et al., 1994). Moreover, in situ hybridization experiments using probes for specific cellular mRNAs have consistently failed to localize spliced or unspliced transcripts in the interior of the speckles (reviewed by Misteli and Spector, 1998). Finally, hnRNP A1, a protein that binds nascent pre-mRNAs and accompanies them to the cytoplasm (Visa et al., 1996), is not detected concentrated in nuclear speckles (Piñol-Roma and Dreyfuss, 1993; Mintz et al., 1999).

With the exception of PABP2, other components of the cleavage/polyadenylation machinery are not normally detected in speckles (Krause et al., 1994; Schul et al., 1998a,b; and Fig. 2). Taking into account that a direct interaction has been demonstrated in vitro between CPSF-30 and PABP2 (Chen et al., 1999), these results suggest that PABP2 dissociates from the 3′ end processing machinery upon completion of cleavage/polyadenylation and accompanies the poly(A) RNA after its release from the site of transcription. Interestingly, immuno-localization studies using antibodies directed against CstF and CPSF show staining of nuclear speckles when cells are treated with α-amanitin (Krause et al., 1994; Schul et al., 1998a,b). In this regard it is noteworthy that both CstF and CPSF interact in vitro with the carboxy-terminal domain (CTD) of the large subunit of RNA pol II (McCracken et al., 1997; recently reviewed by Bentley, 1999), and anti-CTD antibodies label the nuclear speckles in cells treated with transcription inhibitors (Bregman et al., 1995). Therefore, it can be speculated that under normal conditions, CstF and CPSF associate with the CTD at sites of transcription, whereas in the presence of transcription inhibitors the CTD redistributes into the speckles and, consequently, the cleavage/ polyadenylation factors pre-assembled on the CTD will become localized to nuclear speckles.

Contrasting with PABP2, targeting of splicing snRNPs, members of the family of SR-protein splicing factors, and splicing factor U2AF65 to the speckles is independent of the presence of poly(A) RNA (Ferreira et al., 1994; Gama-

Carvalho et al., 1997; see also Fig. 6). In particular, SR-protein splicing factors are RNA binding proteins with one or two amino-terminal ribonucleoprotein-type RNA-binding domain and a carboxy-terminal RS domain, which consists largely of repeating arginine-serine dipetides that are thought to function in protein-protein interactions (reviewed by Tacke and Manley, 1999). For two members of this family, the Drosophila melanogaster Tra protein and the mammalian SRp20 protein, the RS domain has been shown to be both necessary and sufficient for a speckled localization (Li and Bingham, 1991; Hedley et al., 1995; Cáceres et al., 1997). Although SR proteins bind to nascent transcripts and can even be exported to the cytoplasm presumably in association with mature mRNAs (Cáceres et al., 1998), their targeting to the speckles appears to rely on protein-protein interactions and to be independent of RNA binding. Thus, a key issue to be addressed in future studies is whether splicing factors are recruiting poly(A) RNA to the speckles.

We thank Prof. Elmar Wahle (University of Halle, Germany) and Dr Iain Mattaj (EMBL, Heidelberg, Germany) for stimulating discussions. We acknowledge Prof. David Ferreira (University of Lisbon, Portugal) for encouragement, and Mrs Inês Condado (University of Lisbon, Portugal) for technical support. We are also grateful to Prof. E. Wahle for generously providing PABP2 cDNAs and anti-PABP2 serum; Prof W. Keller (University of Basel, Switzerland) for antibodies directed against CPSF subunits and poly(A) polymerase; Prof. J. Manley (Columbia University, New York, USA) for anti-CstF64 and anti-CstF50 monoclonal antibodies; Prof. G. Dreyfuss (University of Pennsylvania, Philadelphia, USA) for mAb 9H10 (anti-hnRNP A1); Dr O. Bensaude (École Normale Supérieure, Paris, France) for luciferase plasmid; Prof. A. Lamond (University of Dundee, UK) for poly(U) riboprobe; Prof. T. Maniatis (Harvard University, USA) for mAb SC35; Prof. W. van Venrooij (University of Nijmengen, The Netherlands) for 4G3 monoclonal antibody and anti-Sm autoimmune serum C45. This study was supported by grants from Fundação para a Ciência e Tecnologia/ PRAXIS XXI (SAU/1310/95, Portugal), and from the European Union (BMH4-98-3147). A.C. is a fellow of the Gulbenkian PhD Program.