The intranuclear mobility of messenger RNA binding proteins is ATP dependent and temperature sensitive

How mRNAs travel from the site of transcription to the nuclear pores remains poorly understood. Immediately upon synthesis, nascent transcripts associate with proteins to form large ribonucleoprotein complexes (messenger ribonucleoprotein particles [mRNPs]^*), the protein content of which evolves as pre-mRNA is processed into mRNA and nuclear mRNA is exported to the cytoplasm (for recent reviews see Nakielny and Dreyfuss, 1999; Maquat and Carmichael, 2001). For several years, it was thought that the intranuclear environment was quite viscous and, consequently, the diffusion of large macromolecular complexes such as mRNPs was expected to be too slow (Agutter and Taylor, 1996). Based on this assumption, Blobel proposed the so-called gene gating hypothesis, which postulated a localized vectorial transport of mRNA toward the nuclear pores (Blobel, 1985). However, more recent measurements of the movement of molecules in the nucleus using photobleaching and fluorescence correlation spectroscopy techniques have indicated that macromolecule-sized solutes can diffuse in the nucleus at rates that are only approximately three- to fourfold slower than those observed in aqueous solution (Seksek et al., 1997; Politz et al., 1998). Most important, two recent reports addressing specifically the movement of nuclear mRNAs revealed properties that are characteristic of a random diffusive process (Politz et al., 1999; Singh et al., 1999). In the first study, nuclear poly(A) RNA was visualized in living cells by hybridization with an oligo(dT) probe labeled with chemically masked (caged) fluorescein (Politz et al., 1999). Upon unmasking the fluorescence by laser spot photolysis, the spread of the signal was followed by time-lapse microscopy. The second study used Chironomous tentans salivary gland cells (Singh et al., 1999). Newly synthesized RNA was labeled in vivo with BrUTP, the cells were fixed at sequential time points after the pulse, and the position of labeled Balbiani ring particles in the nucleus was analyzed by immunoelectron microscopy.

In the present work, GFP was fused to two distinct mRNP-binding proteins (poly[A]-binding protein II [PABP2] and TAP) and the mobility of the tagged complexes was analyzed by quantitative photobleaching techniques. PABP2 binds to the growing poly(A) tails formed at the 3′ ends of nearly all eukaryotic mRNAs (Wahle, 1991). PABP2 cooperates with the cleavage and polyadenylation specificity factor to stimulate the activity of poly(A) polymerase, the enzyme that catalyses polyadenylation. The cleavage and polyadenylation reactions are currently thought to be coupled to splicing of the last intron and to occur at the same time as, or just before, transcription termination (Bauren et al., 1998; Dye and Proudfoot, 1999, 2001). This implies that GFP–PABP2 will bind to nearly terminated and spliced transcripts. Because not all poly(A) RNA is exported from the nucleus, we aimed at achieving a more specific visualization of mRNAs in transit to the cytoplasm by using GFP fused to the export factor TAP. TAP binds directly to the constitutive transport element of viral RNAs and is required for the nucleo-cytoplasmic export of cellular mRNAs in vertebrates, Caenorhabditis elegans, and Saccharomyces cerevisiae (for reviews see Görlich and Kutay, 1999; Nakielny and Dreyfuss, 1999; Conti and Izaurralde, 2001).

Here we have analyzed the effect of energy depletion on the intranuclear mobility of mRNP complexes tagged with GFP–PABP2 and GFP–TAP. For comparison, we determined the mobility rates of large dextrans microinjected into the nucleus of cells depleted of ATP or incubated at 22°C. The results suggest that energy-dependent processes influence specifically the traffic of mRNPs in the living cell nucleus.

We have previously shown that the GFP–PABP2 fusion protein colocalizes with endogenous PABP2, i.e., it is distributed throughout the nucleoplasm with a higher concentration in “speckles” (Calado and Carmo-Fonseca, 2000; Calado et al., 2000). Immunoblotting analysis of total proteins from cells expressing GFP–PABP2 using an anti-GFP antibody showed a single protein band, thus excluding the presence of truncated GFP chimeras. As depicted in Fig. 1 (lane 1), the GFP–PABP2 fusion protein is detected above the expected apparent molecular weight (∼60 kD), reflecting the typical aberrant migration properties of PABP2 (Krause et al., 1994; Nemeth et al., 1995).

To determine whether GFP–PABP2 assembles into mRNP complexes, coimmunoprecipitation experiments were performed using nuclear extracts prepared from transfected cells. The binding properties of wild-type PABP2 were compared with those of a mutant variant, PABP2dm, which contains a double point mutation (a tyrosine to alanine at position 175 and a phenylalanine to alanine at position 215) in the RNA binding domain of the protein (Calado et al., 2000). Fig. 1 (lane 6) shows that a monoclonal antibody specific to hnRNP C (a protein that binds nascent transcripts; Dreyfuss et al., 1993) coprecipitates GFP–PABP2. This is no longer observed when extracts are treated with RNase A (Fig. 1, lane 7), indicating that coprecipitation results from independent binding of both hnRNP C and GFP–PABP2 to mRNA. As expected, a fusion of GFP with the dm variant of PABP2, which shows reduced binding affinity to poly(A) in vitro, fails to coprecipitate with hnRNP C in the absence of RNase treatment (Fig. 1, lane 8). Taken together, these results indicate that GFP–PABP2 binds to mRNA and assembles into mRNP complexes in vivo, whereas the mutant GFP–PABP2dm does not.

Having established that mRNP complexes can be tagged by GFP–PABP2, we next performed FRAP experiments to analyze their mobility in the nucleus of living HeLa cells. A defined area in the nucleoplasm was bleached irreversibly by a single, high-powered spot laser pulse. The recovery of fluorescence signal in the bleached area, which is a consequence of the movement of nonbleached molecules into that region, was recorded by time-lapse imaging (Fig. 2).

Contrasting with the results observed in living cells (Fig. 2 B), there is no recovery of fluorescence after bleaching chemically fixed cells (Fig. 2 A). By several criteria, the bleach pulse does not induce cytotoxic effects. First, the bleached cells showed no signs of reduced vitality when stained with rhodamine 123, a fluorescent dye that is rapidly and specifically sequestered by active mitochondria (Fig. 2 C). Second, the recovery of fluorescence signal was not affected by repeated bleaching of the same nuclear region (Fig. 2 D). Third, bleached cells continued to divide (unpublished data).

For quantitative studies, we selected cells containing the minimal detectable level of GFP fusion proteins in order to avoid potential artifacts caused by overexpression. To monitor the expression level of GFP fusion proteins, the fluorescence intensity of each transfected nucleus was determined using the same detection parameters as those applied during the imaging stage of photobleaching experiments. Because a fraction of poly(A) RNA appears to be retained in the nucleus, where it accumulates in so-called nuclear speckles or clusters of interchromatin granules (Huang et al., 1994), these domains were systematically excluded in all photobleaching measurements.

Quantitative FRAP yields information about the relative mobility of the fluorophore: the effective diffusion coefficient (D) and the fraction of fluorophore that is mobile (White and Stelzer, 1999; Reits and Neefjes, 2001). D indicates the surface area randomly sampled by the fluorophore in a given time, whereas the mobile fraction indicates how much of the fluorophore is available for recovery. To derive both D and the mobile/immobile fractions, the recovery of relative fluorescence intensity within the bleach region is plotted as a function of time (Fig. 3 a, red staining). A function is then fitted to this curve (Fig. 3 a, blue staining). GFP fusion proteins are considered to have a high mobility if they have both a high D and a high mobile fraction.

In clear contrast to the high mobility of GFP–PABP2, a chimera formed by fusing GFP–PABP2 to the Cajal body protein p80 coilin is predominantly immobile in the nucleus (Fig. 3 b). Although the mechanism underlying this immobilization is unknown, absence of recovery after photobleaching a nuclear region in cells expressing the GFP–PABP2–coilin chimera excludes the possibility that the fluorescence recovery is a consequence of GFP refolding or de novo synthesis.

For comparison purposes, we analyzed cells expressing GFP, which moves relatively freely throughout the nucleus and cytoplasm. In these cells, the fluorescence signal recovers to the prebleach value in <1 s (Fig. 4 A), and a similar behavior is observed in cells expressing GFP–PABP2dm, the mutant form of PABP2 with impaired ability to bind poly(A) RNA (Fig. 4 B). In contrast, GFP–PABP2 requires >7 s to recover completely (Fig. 4 C). The calculated D value for GFP–PABP2 is 0.6 μm²s⁻¹, whereas for the mutant it is 4.2 μm²s⁻¹. This indicates that the kinetics of wild-type GFP fusion molecules is approximately sevenfold slower than mutant proteins. Taking into account that both wild-type and mutant GFP fusion proteins have the same molecular size but differ in their ability to bind to poly(A) RNA, the retardation of GFP–PABP2 molecules is most likely a consequence of binding to mRNP complexes.

To determine what fraction of GFP–PABP2 in the nucleus is actually bound to mRNP complexes, we next performed fluorescence loss induced by photobleaching (FLIP) experiments in cells expressing either GFP–PABP2 or GFP–PABP2dm (Fig. 5). To perform FLIP, a region in the nucleus was repeatedly bleached using high laser intensity while the surrounding area was imaged between each round of bleaching. The loss of fluorescence in the area outside of the bleached region is plotted over time, providing information on the rate of displacement of the molecules from that particular space. Repetitive bleaching of a defined nucleoplasmic area for 1,000 s leads to a loss of >95% of the initial fluorescence. Kinetic analysis of FLIP data for cells expressing low levels of GFP–PABP2 revealed the existence of a single slow moving population (Fig. 5 A), indicating that the vast majority of the fusion protein is bound to RNA. In cells expressing higher levels of GFP–PABP2 (Fig. 5 B), the analysis reveals two populations, one slow (with kinetics similar to that detected in low-level expressing cells) and one faster. The fraction of molecules in the faster population increased proportionately to the expression level of GFP–PABP2 (unpublished data), suggesting that it represents a pool of excessive proteins unbound to RNA. Kinetic analysis of FLIP data for GFP–PABP2dm showed the same two populations, but with a clear predominance of molecules with faster kinetics (Fig. 5 C). The additional presence of a minor (∼28%) slow population is consistent with biochemical data indicating that the dm mutation significantly decreases, but does not entirely abolish, the ability of the protein to bind to RNA (Calado et al., 2000). Taken together, the data suggest that the FRAP recovery curves of GFP–PABP2 in cells expressing low levels of the fusion protein reflect the mobility of mRNP complexes in the living cell nucleus.

Electron microscopy of mRNPs isolated from mammalian cell nuclei reveal a repeating subunit structure of spherical particles connected by ribonuclease-sensitive strands (for review see Georgiev, 1981). Each monomer particle sediments in sucrose gradients at 30–40S, has a molecular weight of 1–2 × 10⁶ D, and contains ∼900 ribonucleotides (Pederson, 1974; Beyer et al., 1977; Georgiev, 1981). As the majority of HeLa mRNA averages 1,500 bases (Derman et al., 1976), the estimated average size of mRNP is in the range of 2–4 × 10⁶ D.

To compare the nuclear mobility of GFP–PABP2 bound to RNP complexes with that of exogenous particles with similar size, FRAP experiments were performed on cells microinjected in the nucleus with fluorescein-labeled dextrans of 0.58 and 2 × 10⁶ D. These probes, which are not degraded or metabolized by cells, have been previously used for photobleaching recovery measurements within the nucleus (Seksek et al., 1997). The estimated D value for the 580-kD dextran is 0.95 μm²s⁻¹, whereas for the 2,000-kD dextran, D is 0.53 μm²s⁻¹ (Fig. 6). In both cases, the recovery was incomplete, with immobile pools of 5 and 14%, respectively. This implies that Brownian motion is insufficient to sustain constant movement of large particles within the size range of mRNPs. Thus, the finding that poly(A) RNP complexes tagged with GFP–PABP2 are completely mobile in the nucleus (immobile fraction is 0%) argues that mRNP movement involves more than simple diffusional processes.

To investigate whether the difference observed between the intranuclear mobility of large dextrans and poly(A) RNP complexes tagged with GFP–PABP2 involves energy consumption, FRAP experiments were performed on cells depleted of ATP or incubated at a reduced temperature (22°C). To lower the cellular ATP levels, the cells were incubated with 2-deoxy-glucose and sodium azide. Within 5–10 min after treatment, most cells lost the staining with rhodamine 123, indicating that their mitochondria were no longer active (Fig. 7, A and B). As an additional control, we tested the effect of the energy depletion assay on receptor-mediated transport between the nucleus and the cytoplasm, as it is well established that this process is energy dependent (for review see Görlich and Kutay, 1999). Cells were transfected with GFP fused to a classical NLS (the SV40 large T antigen NLS). GFP is a small protein (∼27 kD) that diffuses across the pores. However, due to the presence of an NLS, it is actively imported. Consequently, under physiological conditions, GFP–NLS is predominantly detected in the nucleus. When cells expressing GFP–NLS are incubated at 4°C, the fluorescence signal is detected both in the nucleus and in the cytoplasm because GFP–NLS diffuses passively to the cytoplasm and NLS-mediated transport to the nucleus is blocked (Fig. 7 C). Within 1 h after shifting the cells to 37°C, the fluorescence is again predominantly localized in the nucleus (Fig. 7 D). However, when cells are incubated at 22°C, the fluorescence remains in the cytoplasm (Fig. 7 E). Similarly, incubation at 37°C in the presence of d-glucose and azide is unable to restore import of GFP–NLS into the nucleus against its concentration gradient (Fig. 7 F).

As illustrated in Fig. 8, when cells were either incubated at 22°C or ATP depleted, the diffusion coefficient of GFP–PABP2 was reduced from 0.6 μm²s⁻¹ to 0.2 μm²s⁻¹, which corresponds to ∼30% of the control value. Because energy depletion treatments may alter the intranuclear environment and consequently affect diffusion rates in an indirect and unspecific manner, we analyzed the recovery curves of the 2,000-kD dextran and GFP–PABP2dm in cells that were either incubated at 22°C or ATP depleted (Fig. 8). At 22°C, the diffusion coefficient of both GFP–PABP2dm and the large dextran were reduced from 4.2 and 0.5 μm²s⁻¹ to 3.2–3.7 and 0.4–0.5 μm²s⁻¹, respectively. These values correspond to 75–77% of the D values at 37°C. In contrast to the temperature effect, ATP depletion did not affect the D value of the large dextran and reduced that of GFP–PABP2dm to ∼90% of the control value. Thus, although energy-depleting treatments do reduce the mobility of both the large dextran and the PABP2 mutant with impaired ability to bind RNA, the effect is significantly more drastic on wild-type PABP2.

The next question to be addressed was whether energy depletion affects the binding of PABP2 to poly(A) RNA. This was examined by using an electrophoretic mobility shift assay with ³²P-labeled poly(A) (A₂₅₀). As depicted in Fig. 9, addition of a 1,000-fold molar excess of unlabeled substrate completely displaced PABP2 from radioactive poly(A) (A₂₅₀). The displacement occurs within seconds after addition of the unlabeled competitor, in the absence of ATP or any other source of energy. Furthermore, similar results were obtained when the assay was performed at either 37°C or 22°C (unpublished data). These results show that in vitro PABP2 is constantly exchanging between poly(A) substrates in a process that is independent of energy.

Several recent lines of evidence indicate that export of mRNA to the cytoplasm is tightly coupled to cotranscriptional processing of newly synthesized pre-mRNAs in the nucleus (Kataoka et al., 2000; Zhou et al., 2000; Lei et al., 2001). Thus, because both transcription and pre-mRNA splicing are ATP dependent, the FRAP results obtained in energy-depleted cells could be caused by inhibition of these two processes. To address this issue, FRAP experiments were performed on cells treated for 30–60 min with the transcription inhibitor actinomycin D. The results show recovery kinetics similar to that detected in nontreated cells (Fig. 10). However, longer periods of incubation with actinomycin D resulted in a progressive immobilization of both GFP–PABP2 and other GFP-tagged nuclear proteins (unpublished data), suggesting that prolonged treatment with actinomycin D causes a general obstruction to macromolecular mobility due to alterations in the intranuclear environment.

It is well known that actinomycin D treatment does not deplete the HeLa cell nucleus of poly(A) RNA (Huang et al., 1994). Thus, it is expected that GFP–PABP2 molecules remain bound to poly(A) RNP complexes synthesized before the inhibition of transcription. The similar recovery kinetics observed for GFP–PABP2 in untreated and treated cells therefore suggests that the mobility of poly(A) RNPs is not significantly affected in actinomycin D–treated cells. This indicates that on-going transcription is not essential for the movement of previously synthesized mRNP complexes in the nucleus.

Next, photobleaching experiments were performed using a GFP-tagged version of the mRNA export factor TAP. TAP associates with cellular mRNPs and is thought to promote their export by interacting with nuclear pore proteins during translocation (Bachi et al., 2000). The NH₂-terminal region of the TAP protein includes a noncanonical RNP-type RNA binding domain that exhibits general RNA binding activity (Liker et al., 2000) and four leucine-rich repeats, which play an essential role in TAP-mediated export of mRNA (Braun et al., 2001).

A TAP fragment comprising residues 371–619 (which excludes both the RNA binding domain and the leucine-rich repeats; Bachi et al., 2000) was fused to GFP and used in photobleaching experiments (Fig. 11). The recovery curve shows that this protein moves with kinetics similar to that of GFP–PABP2dm (D values of 4.3 and 4.2 μm²s⁻¹, respectively), suggesting that it does not assemble into a larger macromolecular complex. In contrast, the fluorescence recovery of GFP fused to full-length TAP was significantly slower (D = 1.2 μm²s⁻¹), as expected from its binding to mRNP complexes. Compared with the diffusion coefficient of GFP–PABP2 (0.6 μm²s⁻¹), GFP–TAP has a faster diffusion rate. FLIP analysis shows that this apparent higher mobility is caused by the presence of two distinct GFP–TAP populations in the nucleus, a slow-moving fraction of molecules bound to mRNP complexes and a fast-moving population of unbound molecules (to be described in detail elsewhere). In contrast, GFP–PABP2 forms a single population of molecules, the vast majority of which are bound to poly(A) RNPs (Fig. 5).

When cells were either ATP depleted or incubated at 22°C, the fluorescence recovery of GFP–TAP became slower and the immobile fractions increased (Fig. 11). The diffusion coefficients decreased from 1.2 μm²s⁻¹ to 0.8 and 0.6 μm²s⁻¹, respectively. Thus, as observed for GFP–PABP2, energy depletion significantly reduces the mobility of GFP–TAP molecules presumably bound to mRNP complexes.

In this work, we analyzed the mobility of GFP-tagged RNP complexes in the nucleus of living HeLa cells using photobleaching techniques. Bulk RNPs containing poly(A) RNA were visualized with GFP–PABP2. In addition, GFP–TAP was used to tag export-competent mRNP complexes. The main conclusion is that the GFP-tagged RNP complexes diffuse freely in the nucleus, whereas a large dextran, the size of which corresponds approximately to 40S mRNP particles, is partially immobilized. In addition, energy depletion reduces the intranuclear mobility of wild-type GFP–PABP2 and GFP–TAP significantly, as compared with the large dextran and a GFP–PABP2 mutant with impaired ability to bind to RNA. This suggests that although mRNPs diffuse in the nucleus, energy-dependent processes contribute to enhance their mobility.

Here we estimated a diffusion coefficient of 0.6 ± 0.03 μm²s⁻¹ for RNP complexes containing GFP–PABP2 in the nucleus. A very similar value (0.6 ± 0.1 μm²s⁻¹) was reported by Politz et al. (1999) using fluorescein-labeled oligo(dT) hybridized to nuclear poly(A) RNA. The diffusion coefficient for a particle in a solution is given by the Stokes-Einstein formula, D = kT/6πηR_h, which correlates the hydrodynamic behavior of a sphere with the absolute temperature T, the viscosity of the solution η, the Boltzmann constant k, and the hydrodynamic radius of the particle R_h (for a recent review see Reits and Neefjes, 2001). Because viscosity is highly dependent on temperature, this equation implies that changes in diffusion coefficients are expected in response to temperature variation. Because the viscosity of water increases ∼30% when temperature is lowered from 37°C to 22–23°C, the diffusion coefficient for a particle in the nucleus should decrease by a similar order of magnitude. Consistent with this prediction, we observe a reduction of ∼25% in the diffusion coefficients of the 2,000-kD dextran and the mutant form of PABP2 when the temperature is lowered from 37°C to 22°C. Likewise, an ∼30% reduction in recovery rate was reported for soluble GFP targeted to the endoplasmic reticulum when FRAP experiments were performed at 37°C and 23°C (Reits and Neefjes, 2001). In contrast to these results, other studies did not detect a significant effect of reducing the temperature to 23°C on the mobility rate of either GFP-tagged nuclear proteins (Phair and Misteli, 2000) or nuclear poly(A) RNA hybridized with fluorescein-labeled oligo(dT) (Politz et al., 1999).

Contrasting to the observed ∼25% reduction in the recovery rates of both the large dextran and the RNA binding–deficient GFP–PABP2 mutant, lowering the temperature from 37°C to 22°C caused an ∼60% reduction in the diffusion coefficient of wild-type GFP–PABP2, and a similar effect was observed after depleting the cells of ATP. Thus, energy-depleting treatments act specifically on the mobility of the fusion protein that binds to poly(A) RNA, arguing against a general effect caused by an alteration of the intranuclear environment induced by these treatments. Using a very different technical approach, Politz et al. (1999) measured the movement of nuclear poly(A) RNA and estimated the same apparent diffusion coefficient at both 37°C and 23°C. The method devised by these authors consisted of hybridizing chemically masked (caged) fluorescein to poly(A) RNA, using laser spot photolysis to uncage the fluorochrome, and measuring the radial distance that the signal traveled from the uncaging site. The radial movement of signal was calculated and plotted over time, and the diffusion coefficient was estimated from the slope of the unweighted least-squares fit line (Politz et al., 1999). The different methods used for both poly(A) RNA visualization and quantitative analysis are probably on the basis of the discrepancy between the conclusions of this earlier study and our present data.

Although the mechanism underlying the movement of poly(A) RNPs within the nucleus remains unknown, a parallel can de drawn with the process of translocation through the nuclear pore complex. Early studies showed that energy depletion blocked nuclear import and arrested the molecules in transit at the nuclear pore complexes (Newmeyer and Forbes, 1988; Richardson et al., 1988). Later, GTP hydrolysis by Ran was shown to drive transport, presumably by powering the actual translocation through the pore. However, more recent evidence revealed that the translocation step, per se, is not directly coupled to GTP hydrolysis. Rather, it appears that the translocation process occurs by facilitated diffusion, at least for relatively small cargoes (for review see Görlich and Kutay, 1999). According to this view, it has been proposed that energy depletion inhibits nuclear import by blocking the terminal release of molecules in transit from the nuclear pore into the nucleoplasm (Ribbeck and Görlich, 2001).

Although at present the mechanistic details involved in mRNP traffic inside the nucleus are completely unknown, one possibility is that in order to escape immobilization imposed by the densely packed nuclear interior, mRNPs rely on ATP-dependent enzymatic activities. Putative candidates to maintain mRNP particles in a mobile state are ATP-dependent RNA helicases, such as Dbp5p, which participates in the export of mRNAs out of the nucleus (Schmitt et al., 1999; Strahm et al., 1999). Although Dbp5p was first predominantly detected in the cytoplasmic side of nuclear pore complexes, a recent study shows that in C. tentans, this helicase binds mRNPs cotranscriptionally and accompanies the particle to the pores (Zhao et al., 2002). Additional candidates include actin-related ATPases and actomyosin-based molecular motors. A nuclear isoform of myosin I has been recently identified in a complex with RNA polymerase II (Pestic-Dragovich et al., 2000), and it is well established that mRNAs in the cytoplasm can be transported by an actomyosin-driven mechanism (Long et al., 1997; Takizawa et al., 1997). Furthermore, an increasing number of reports describe the association of actin and/or actin-related proteins (Arps) with chromatin complexes (for review see Boyer and Peterson, 2000). Actin has also been found associated with Balbiani ring mRNA (Percipalle et al., 2001) and with the nucleoplasmic filaments of nuclear pore complexes (Hofmann et al., 2001), thus suggesting an involvement in the RNA export pathway. The experimental approach described in this study should be applicable for future screening and identification of proteins that play a role in the intranuclear mobility of mRNP complexes.

HeLa cells were cultured as monolayers in DME supplemented with 10% FCS (GIBCO BRL). Cells were plated and observed in glass bottom chambers (MatTek Corporation). For imaging, the medium was changed to DME/F-12 without phenol red supplemented with 15 mM Hepes buffer (GIBCO BRL).

All GFP fusion constructs were obtained by subcloning into the appropriate pEGFP-C vector (CLONTECH Laboratories, Inc.). We have previously described the construction and properties of GFP–PABP2 and GFP–PABP2dm (Calado and Carmo-Fonseca, 2000; Calado et al., 2000). GFP–TAP and GFP–TAP 371–619 were also previously described (Bachi et al., 2000). The pEGFP-coilin–PABP2 plasmid was constructed by subcloning the cDNA of coilin into the pEGFP-PABP2 vector. The cDNA of coilin was obtained from pBluescript SK(−) coilin (Bohmann et al., 1995) through restriction with BamHI and PvuII and purification with the Qiaex II kit (QIAGEN). This double restriction deletes the DNA sequence that encodes for the 108 COOH-terminal amino acids of coilin. As a result, PABP2 is fused to a COOH-terminally truncated form of coilin. The pEGFP-PABP2 vector was first restricted with Sal I and the protruding ends generated were filled in. This vector was then restricted with Bgl II and ligated to the coilin-coding fragment. DNA for transfection was purified using the plasmid DNA midiprep kit (QIAGEN). HeLa cells were transfected with FuGene6 reagent (Roche Biochemicals) using 1 μg of DNA and analyzed 16–24 h after transfection. FITC-labeled dextrans with average molecular sizes of 580 and 2,000 kD (Sigma-Aldrich) were diluted to 200 μg/ml in water and microinjected into the nucleus of HeLa cells as previously described (Almeida et al., 1998). Actinomycin D (Sigma-Aldrich) was used at 5 μg/ml from a stock solution of 5 mg/ml in DMSO. A stock solution of 0.6 M 2-deoxy-d-glucose was prepared in water and sodium azide was used from a 1 M stock solution freshly prepared in DME/F-12. Energy depletion experiments were performed by incubating cells in 12 mM 2-deoxy-d-glucose and 20 mM sodium azide for 30–60 min at 37°C. To stain selectively active mitochondria, the cells were incubated with rhodamine 123 (Molecular Probes) for 5–10 min.

Immunopurification of hnRNP complexes from HeLa cells transfected with pEGFP-PABP2 or pEGFP-PABP2dm was performed according to Piñol-Roma et al. (1990). In brief, cells were rinsed twice with cold PBS, scraped with a rubber policeman into 1 ml of cold immunopurification buffer (IPB), and lysed by passing through a 25-gauge needle. The nuclei were pelleted, resuspended in 500 μl IPB, disrupted by sonicating, and layered in an equal volume of 30% sucrose cushion. After centrifugation at 4,000 g for 15 min at 4°C to remove nucleoli, chromatin, and other insoluble nuclear structures, the nucleoplasm-enriched fraction that overlays the cushion was collected. When required, RNase A was directly added onto this fraction at the final concentration of 100 μg/ml and digestion was allowed to proceed at 30°C for 10 min (Piñol-Roma, 1999).

Prior to immunoprecipitation, anti–hnRNP C (4F4; Choi and Dreyfuss, 1984) or anti-GFP (Roche Biochemicals) monoclonal antibodies were covalently conjugated to protein A–Sepharose beads (CL-4B; Amersham Biosciences) or protein G–Sepharose beads (Sigma-Aldrich), respectively. For the immunoprecipitation, 25 μl (packed volume) of beads were used for each 10-cm plate. The beads were first washed with IPB and then incubated with the nucleoplasmic fraction for 10 min at 4°C with rocking. After washing, the immunoprecipitation complexes were eluted by boiling the beads in SDS-PAGE sample buffer. Immunoprecipitates were analyzed by Western blotting using anti–hnRNP C and anti-GFP antibodies. SDS-PAGE and Western blotting were performed as previously described (Almeida et al., 1998).

A radiolabeled probe was synthesized by 5′ end labeling of an A₂₅₀ oligonucleotide with [γ-³²P]ATP using T4 polynucleotide kinase (Fermentas). PABP2 was purified from calf thymus as previously described (Wahle et al., 1993). 10 nM ³²P-labeled A₂₅₀ probe was mixed with an equimolar amount of PABP2 diluted in 50 mM Tris (pH 8.0), 100 mM KCl, 10% glycerin, 0.5 mM DTT, 0.01% NP-40, 0.2 mg/ml BSA, 2.6% PVA, and 2 mM MgCl₂ and incubated for 30 min at either 37°C or room temperature (22°C). A 1,000-fold excess (10 μM) of unlabeled A₂₅₀ oligonucleotide was then added to the mix. Aliquots were removed at the indicated times and directly loaded onto the running gel. Complexes were resolved by electrophoresis through nondenaturing 6% polyacrylamide gels (80:1 acrylamide/bisacrylamide ratio). Gels were run at 4°C for 2 h at 20 V/cm and analyzed using the Molecular Dynamics PhosphorImager.

Live cells were imaged at 37°C or 22°C, maintained by a heating/cooling frame (LaCon GbR) in conjunction with an objective heater (PeCon GmbH). Images were acquired on a ZEISS LSM 510 with the planapochromat 63x/1.4 objective. EGFP fluorescence was detected using the 488-nm laser line of an Ar laser (25 mW nominal output) in conjunction with an LP 505 filter. Each FRAP analysis started with three image scans, followed by a single bleach pulse of 37 ms on a spot with a diameter of 25 pixels (0.71-μm radius). A series of 97 single-section images were then collected at 78-ms intervals, with the first image acquired 2 ms after the end of bleaching. Image size was 512 × 50 pixels and the pixel width was 57 nm. For imaging, the laser power was attenuated to 0.1–0.2% of the bleach intensity.

For each time series, the background and nuclear regions were identified using an implementation of the ICM segmentation algorithm (Besag, 1986; Fwu and Djuriç, 1996) in Mathworks Matlab software. The average fluorescence in the nucleus T(t) and the average fluorescence in the bleached region I(t) were calculated for each background-subtracted image at time t after bleaching. FRAP recovery curves were normalized according to Phair and Misteli (2000),

where T_i is the fluorescence in the nucleus before bleaching and I_i is the fluorescence in the bleached region before bleaching. This normalization corrects for the loss of fluorescence caused by imaging. Typically, ∼10% of the total EGFP or FITC fluorescence was lost during the bleach pulse. During the post-bleaching scanning phase, the fluorescence lost was <5% for EGFP and 15–17% for FITC.

Nucleoplasmic diffusion coefficients were determined essentially as described previously (Axelrod et al., 1976; Phair and Misteli, 2000). We assume that fluorescent molecules can be either mobile (with a diffusion coefficient D) or immobile. The initial concentration of fluorophore inside the bleached zone is C₀, γ being the percentage of immobile molecules (ranging from 0 to 1). Total fluorescence observed inside the bleached spot is then the sum of the fluorescence of the mobile molecules and the fluorescence of the immobile ones. After bleach, the immobile fraction fluorescence is constant inside the bleached spot (Axelrod et al., 1976):

where K is the bleach constant and κ is a parameter that depends on the properties of the laser and of the detection system. The mobile fraction fluorescence is given by

(Axelrod et al., 1976), where τ_D is the characteristic time of diffusion and is related with the diffusion coefficient by D = w²/4τ_D. The total normalized fluorescence is then

i.e., the normalized fluorescence is just the weighted sum of the fluorescence recovery of the mobile portion of the fluorophore (first term) and the fluorescence after bleaching of the immobile portion of the fluorophore (second term).

The bleaching characteristics of the laser were estimated using fixed cells expressing GFP. The post-bleaching concentration of fluorophore at distance r from the center of the bleached region was fitted to a Gaussian laser profile (Axelrod et al., 1976),

where k is the bleach constant for the fixed cells and w is half the width of the laser beam at 1/e² intensity. Nonlinear regression in Mathematica yielded k = 2.14 ± 0.06 and w = 1.13 ± 0.01 μm (mean ± SEM). The FRAP recovery curves were then fitted to the recovery function, Irel*(t).

The series solution for the fluorescence recovery was truncated after 40 terms (Phair and Misteli, 2000), assuring that neglected terms made an insignificant contribution. We estimated the bleach constant, the immobile fraction, and the characteristic diffusion time for each set of FRAP replicates using a weighted least squares algorithm (Bell et al., 1996) implemented in Mathematica (WOLFRAM Research).

For FLIP experiments, cells were repeatedly bleached at intervals of 3.5 s and imaged between bleach pulses. Bleaching was performed by 278-ms bleach pulses on a spot with a diameter of 35 pixels (2.1-μm radius). Repetitive bleach pulses were achieved by taking advantage of the trigger interface for LSM 510. An electronic oscillator circuit was built to create pulses with a user-defined frequency. When connected to the LSM 510, it would then trigger the bleaching events. A series of 350 images was collected for each cell with laser power attenuated to 1% of the bleach intensity. Nuclear fluorescence of selected areas in FLIP experiments was measured using the ROI mean function of the LSM 510 physiology package. The data were then background subtracted and normalized to correct the loss of fluorescence caused by imaging, in a way similar to FRAP, but using an adjacent cell to estimate T(t) and T₀. Loss of fluorescence due to imaging could reach 20–25% over the time course of the experiment.

Nuclear fluorescence data from FLIP experiments was fitted to a single exponential with rate constant R (Fig. 5 A) and to a nonlinear curve corresponding to the sum of two exponentials,

where p is the proportion of the slower population relative to the total fluorescence, R₁ is the rate constant associated with this slower population, and R₂ the rate constant of the faster one (Fig. 5, B and C). The optimal fit for each case was chosen according to its chi-square value.

We thank Tom Misteli for stimulating discussions and critical comments. We further wish to acknowledge Rainer Pepperkok, Ernst Stelzer, and Jan Ellenberg for help with FRAP analysis, Serafin Piñol-Roma for advice on immunopurification of hnRNP complexes, and Inês Condado, Helena Pina, and Maria do Carmo Silva for technical support. We are also thankful to Elisa Izaurralde (EMBL, Heidelberg, Germnay) for providing GFP–TAP and GFP–TAP 371-619, and Gideon Dreyfuss (University of Pennsylvania, College Park, PA) for mAb 4F4.

This study was supported by Fundação para a Ciência e Tecnologia, Portugal, and by the European Commission (contract BMH4-98-3147). A. Calado was a fellow from the Gulbenkian PhD program.