Importin α-Mediated Nuclear Import of Cytoplasmic Poly(A) Binding Protein Occurs as a Direct Consequence of Cytoplasmic mRNA Depletion

Abstract

Recent studies have found the cytoplasmic poly(A) binding protein (PABPC) to have opposing effects on gene expression when concentrated in the cytoplasm versus in the nucleus. PABPC is predominantly cytoplasmic at steady state, where it enhances protein synthesis through simultaneous interactions with mRNA and translation factors. However, it accumulates dramatically within the nucleus in response to various pathogenic and nonpathogenic stresses, leading to an inhibition of mRNA export. The molecular events that trigger relocalization of PABPC and the mechanisms by which it translocates into the nucleus to block gene expression are not understood. Here, we reveal an RNA-based mechanism of retaining PABPC in the cytoplasm. Expression either of viral proteins that promote mRNA turnover or of a cytoplasmic deadenylase drives nuclear relocalization of PABPC in a manner dependent on the PABPC RNA recognition motifs (RRMs). Using multiple independent binding sites within its RRMs, PABPC interacts with importin α, a component of the classical import pathway. Finally, we demonstrate that the direct association of PABPC with importin α is antagonized by the presence of poly(A) RNA, supporting a model in which RNA binding masks nuclear import signals within the PABPC RRMs, thereby ensuring efficient cytoplasmic retention of this protein in normal cells. These findings further suggest that cells must carefully calibrate the ratio of PABPC to mRNA, as events that offset this balance can dramatically influence gene expression.

INTRODUCTION

Cells regulate multiple processes through subcellular redistribution of proteins, especially upon encountering stressful conditions. Many well-characterized examples include a number of transcription factors that are transported to the nucleus to induce expression of specific genes in response to stimuli such as oxidative stress, UV radiation, heat shock, or infectious agents. For example, the activation of NF-κB, the heat shock response factor, interferon regulatory factors, and p53 leads to their nuclear relocalization and stimulation of gene expression cascades (3, 29, 30, 47, 48, 53, 55, 57, 67). Proteins can also undergo nuclear to cytoplasmic redistribution to affect gene expression posttranscriptionally. This is the case for the RNA binding protein HuR, which translocates from the nucleus to the cytoplasm upon activation of p38 or oxidative stress to stabilize mRNAs containing AU-rich elements (18, 54). Retaining these regulatory factors in the appropriate locale in the absence of activating signals is critical to preventing inappropriate alterations in gene expression that could affect cellular homeostasis. The mechanisms by which the localization of such factors is restricted in the absence of activating signals are varied, and they include interactions with inhibitor proteins, posttranslational modifications such as phosphorylation, and constitutive nuclear export (18, 44, 48, 53, 55, 57, 68, 72).

Cytoplasmic poly(A) binding protein (PABPC) has recently been shown to exert differential effects on gene expression based on its subcellular localization. At steady state, PABPC is predominantly localized within the cytoplasm, where it promotes gene expression by enhancing translation and mRNA stability (42, 51). Its interactions with the mRNA 3′ end via the poly(A) tail and the 5′ end through direct and indirect interactions with eIF4G, Paip1, and eIF4E serve to stabilize messages and enhance protein synthesis (15, 33, 74). Associations of PABPC with the release factor eRF3 promote efficient translation termination, as well as antagonize nonsense-mediated decay (NMD), a quality control pathway that degrades mRNAs bearing premature stop codons (6, 13, 20, 31, 34). PABPC also enhances message stability by protectively coating the poly(A) tail to prevent exonucleolytic degradation and influences the rate of mRNA turnover through its interactions with factors involved in deadenylation, including Pan3, TOB, and GW182 (22, 23, 78).

Though predominantly present in the cytoplasm, PABPC is a nucleocytoplasmic shuttling protein (2), and it preferentially accumulates in the nucleus upon cellular stresses such as heat shock, oxidative stress, and transcription block, as well as upon depletion of paxillin (2, 49, 64, 75). In addition, it is driven into the nucleus upon infection with a variety of viruses, including Kaposi's sarcoma-associated herpesvirus (KSHV), murine gammaherpesvirus 68 (MHV68), herpes simplex virus (HSV-1), rotavirus, and bunyamwera virus (7, 28, 46, 56, 64). Following recovery from heat shock, PABPC returns to the cytoplasm, although during viral infection it remains nuclear until cell death (46, 49). Elevated levels of nuclear PABPC promote hyperadenylation of nuclear mRNAs and a concomitant RNA export block, thereby restricting gene expression (43). Hyperadenylation is carried out by the canonical mRNA poly(A) polymerase PAPII, which presumably extends poly(A) tails on PABPC-bound nuclear mRNAs (43). Thus, PABPC appears to have generally opposing effects on gene expression when concentrated in the cytoplasm versus the nucleus, and changes in its steady-state localization are induced in response to various stresses.

Unlike many other gene expression regulators, the specific events that prompt the removal of PABPC from the cytoplasm and the mechanisms by which it translocates into the nucleus to block gene expression are not understood. Interestingly, many viruses that induce PABPC relocalization also suppress protein synthesis by promoting cytoplasmic mRNA turnover (14, 21, 25, 26, 36, 37, 63, 66, 70). Here, we reveal that PABPC localization is directly influenced by cytoplasmic mRNA abundance. Expression of viral proteins that cause widespread mRNA turnover or of a cytoplasmic deadenylase promotes nuclear accumulation of PABPC. Nuclear import requires the RNA recognition motifs (RRMs) of PABPC, which likely harbor noncanonical nuclear localization signals and which interact directly with importin α, a component of the classical nuclear import pathway. Finally, we show that binding of PABPC1 to the nuclear import machinery is antagonized by the addition of poly(A) RNA and enhanced upon RNase treatment, arguing that PABPC exhibits mutually exclusive binding to poly(A) or importin α. Collectively, our findings suggest a novel mechanism for controlling the subcellular localization of this critical protein based on the availability of poly(A) mRNA in the cytoplasm, and they have important implications for how cells may sense and respond to gross alterations in transcript accumulation.

MATERIALS AND METHODS

Plasmids.

Plasmids pCDEF3-HA-PABPC1 (43), pCDEF3-SOX, pCDEF3-muSOX, pCDNA3.1-vhs, pCDNA3.1-vhsmut (14, 35), and pTRE-d2eGFP (BD Biosciences) have been previously described. PABPC1 was cloned into the BamHI and XbaI sites of plasmid pMAL-C2X to generate pMAL-C2X-PABPC1 in order to express and purify maltose binding protein (MBP)-PABPC1 fusion protein from bacterial cells. Plasmid pCDNA3-PABPC4-HA was kindly provided by Tullia Lindsten (University of Pennsylvania) (79). Plasmid pDNR-LIB-PABPC5 was purchased from Open Biosystems (clone ID 6452933, accession number BC063113) from which PABPC5 was PCR amplified, 5′-tagged with 1× hemagglutinin (HA), and subcloned into the BamHI and XbaI sites of pCDEF3 to generate pCDEF3-HA-PABPC5. HA-PABPC1 deletion mutants were generated by overlap extension PCR and cloned into the BamHI and XbaI sites of pCDEF3 to generate pCDEF3-HA-PABPC1 ΔRRM 1 + 2 (lacking nucleotides [nt] 1 to 528), pCDEF3-HA-PABPC1 ΔRRM 3 + 4 (lacking nt 571 to 1110), pCDEF3-HA-PABPC1 ΔRRM 1–4 (lacking nt 1 to 1110), pCDEF3-HA-PABPC1 ΔLinker (lacking nt 1111 to 1629), pCDEF3-HA-PABPC1 ΔC-term (lacking nt 1630 to 1911), pCDEF3-HA-PABPC1 RRM 1 + 2 (lacking nt 571 to 1911), pCDEF3-HA-PABPC1 RRM 3 + 4 (lacking nt 1 to 525 and 1167 to 1911), and pCDEF3-HA-PABPC1 RRM 1 + 2 + 3 + 4 (lacking nt 1111 to 1911). Plasmids pGEX-importin α1, pGEX-importin α3, pGEX-importin α5, pGEX-importin α7, and pGEX-importin β were kindly provided by Riku Fagerlund (National Public Health Institute, Helsinki, Finland). Plasmids pCDNA3-FLAG-hCaf1z, pCDNA3-FLAG-hCaf1z ΔNLS, pCDNA3-FLAG-hCaf1z DEAA, pCDNA3-myc-hCaf1z ΔNLS/DEAA, and pPCβwt-Δ12 were kindly provided by Jens Lykke-Andersen (University of California, San Diego) (73).

Cells and transfections.

HEK293T, HeLa, HeLa Tet-off, and COS-7 cells were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS). DNA transfections were performed in 12-well plates using 1 μg total plasmid DNA (100 ng of indicated plasmid plus 900 ng of empty vector) using Effectene (Qiagen) or Lipofectamine (Invitrogen) as per the manufacturer's instructions. For kinetic assays, HeLa Tet-off cells were transfected with 100 ng of pPCβwt-Δ12, 100 ng of pTRE-d2eGFP, and empty vector or the indicated hCaf1z plasmid to 1 μg total DNA in the presence of 50 ng/ml tetracycline (Clontech) for 24 h. Transcription was pulsed by removal of tetracycline for 5 h and then blocked by the addition of 1 μg/ml tetracycline. Cells were harvested for total RNA at the indicated time points thereafter and analyzed by Northern blotting as described below.

IFA.

HEK293T cells were seeded on poly(l)-lysine (100 μg/ml)-coated coverslips, transfected for 24 h, and processed for immunofluorescence assays (IFA) as previously described (5). Briefly, cells were fixed in 4% formaldehyde for 20 min and incubated in permeabilization buffer (1% [vol/vol] Triton X-100 and 0.1% [wt/vol] sodium citrate in phosphate-buffered saline [PBS]) for 10 min and then in blocking buffer (1% [vol/vol] Triton X-100, 0.5% [vol/vol] Tween 20, and 3% bovine serum albumin [BSA] in PBS) for 30 min. Cells were then incubated with either mouse monoclonal anti-PABPC 10e10 (1:25 dilution; Santa Cruz Biotechnology), rabbit polyclonal anti-eIF4E (1:500 dilution; Cell Signaling Technology), goat polyclonal anti-eIF4G (1:100 dilution; Santa Cruz Biotechnology), rabbit polyclonal anti-eRF3 (1:200 dilution; Abcam), rabbit polyclonal anti-Paip1 (1:500 dilution; kindly provided by Nahum Sonenberg, McGill University), rabbit polyclonal anti-Paip2a (1:500 dilution; kindly provided by Nahum Sonenberg, McGill University), mouse monoclonal anti-HA 12CA5 (1:500 dilution; Abcam), mouse monoclonal anti-Flag M2 (1:500; Sigma), rabbit polyclonal anti-Flag (1:500; Sigma), or rabbit polyclonal anti-myc (1:200 dilution; BioVision) primary antibodies for 3 to 12 h at 37°C and washed three times with PBS. Cells were then incubated with Alexa Fluor 488-, 546-, or 594-conjugated goat anti-mouse, goat anti-rabbit, or donkey anti-goat secondary antibodies (1:1,500 dilution; Molecular Probes) for 1 h at 37°C and washed three times with PBS. Coverslips were mounted in DAPI (4′,6-diamine-2-phenylindole)-containing Vectashield mounting medium (Vector Labs) to stain cell nuclei. For all statistical analyses, cells expressing HA-PABPC1 in multiple fields of view from two to three independent experiments were counted and the percentage of nuclear HA-PABPC1 is reported graphically. Error bars indicate standard errors of the mean (SEM).

Cell extracts, Western blotting, and Northern blotting.

Cell lysates were prepared in RIPA lysis buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1% [vol/vol] Nonidet P-40, 0.5% [wt/vol] sodium deoxycholate, 0.1% [wt/vol] sodium dodecyl sulfate [SDS]) containing protease inhibitors (Roche) for Western blotting or in NETN lysis buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40) containing protease inhibitors (Roche) for immunoprecipitations and glutathione S-transferase (GST) pulldown experiments. Protein lysates were quantified by a Bradford assay (Bio-Rad).

Equivalent micrograms of each sample were resolved by SDS-polyacrylamide gel electrophoresis (PAGE), transferred to a polyvinylidene difluoride (PVDF) membrane, and subjected to Western blotting using either rabbit polyclonal anti-eIF4E (1:1,000 dilution; Cell Signaling Technology), rabbit polyclonal anti-eRF3 (1:1,000 dilution; Abcam), rabbit polyclonal anti-Paip2a (1:2,000 dilution; kindly provided by Nahum Sonenberg, McGill University), mouse monoclonal anti-PABPC 10e10 (1:2,000 dilution; Santa Cruz Biotechnology), rabbit polyclonal anti-PABPC (1:2,000 dilution; Cell Signaling Technology), mouse monoclonal anti-HA 12CA5 (1:2,000 dilution; Abcam), mouse monoclonal anti-Flag M2 (1:2,000 dilution; Sigma), rabbit polyclonal anti-myc (1:2,000 dilution; BioVision), or mouse monoclonal anti-β-tubulin primary antibodies (1:1,000 dilution; Sigma). Secondary antibodies included horseradish peroxidase-conjugated antiactin antibodies (1:5,000 dilution; Santa Cruz Biotechnology) and goat anti-mouse, goat anti-rabbit, or donkey anti-goat antibodies (1:5,000 dilution; Southern Biotechnology Associates).

For Northern blots, total cellular RNA was isolated using RNA-bee (Tel-Test), resolved on 1.2% agarose-formaldehyde gels, transferred to a nylon membrane, and probed with ³²P-labeled green fluorescent protein (GFP) or 18S DNA probes prepared with the Rediprime II random primer labeling kit (GE Healthcare) or ³²P-labeled β-globin riboprobes prepared with the MAXIscript SP6 kit (Ambion) as per the manufacturer's instructions. GFP and 18S bands were quantified using ImageJ software, and GFP transcript levels were normalized to 18S levels.

Immunoprecipitations.

Immunoprecipitations for endogenous PABPC were performed by binding 1 mg of mouse monoclonal PABPC 10e10 antibody to protein A-Sepharose 4 fast-flow beads (GE Healthcare) at 4°C with agitation for 1 h, washing the beads with NETN buffer, and mixing with lysates from transfected HEK293T cells at 4°C with agitation for 3 h. Immunoprecipitations for HA-PABPC1 were conducted by mixing monoclonal anti-HA clone HA-7 beads (Sigma) and lysates from transfected HEK293T at 4°C with agitation for 3 h. The beads were then washed 3 times with NETN buffer, and bound proteins were eluted with 40 μl 2× Laemmli sample buffer, resolved on 10% SDS-PAGE, and subjected to Western blotting.

Production of fusion proteins in Escherichia coli and importin binding assays.

E. coli strain BL21 was transformed with plasmids encoding MBP alone, MBP-PABPC1, GST alone, or GST-fused human importin α1, α3, α5, α7, or β. Overnight cultures grown from single colonies were used to inoculate 100 ml LB medium, grown at 37°C with agitation for 2 h, and induced with 1 mM IPTG (isopropyl-1-thio-β-galactopyranoside) for 4 h. Bacterial cells were pelleted at 10,000 × g and 4°C for 10 min and lysed in L buffer (50 mM Tris [pH 8.0], 120 mM NaCl, 50 mM EDTA, 1% Triton X-100, and 10 mM β-mercaptoethanol) with 3 μg/ml lysozyme (Sigma) and protease inhibitors (Roche) for 15 min with periodic vortexing. Two freeze-thaw cycles were followed by brief sonication, and the crude lysates were clarified by centrifugation. MBP and MBP-PABPC1 were purified by mixing crude lysates with amylose resin (New England BioLabs) at 4°C with agitation overnight. The resin was then packed into columns and washed three times with RIPA buffer, once with NETN buffer, and once with column buffer (20 mM Tris [pH 7.4], 200 mM NaCl, 1 mM EDTA, and 10 mM β-mercaptoethanol with protease inhibitors [Roche]), and MBP or MBP-PABPC1 was eluted with column buffer containing maltose (10 mM).

For GST pulldown experiments, GST alone or GST-fused importins were first bound to 25 μl glutathione-Sepharose 4B beads (GE Healthcare) in 300 μl NETN buffer at 4°C with agitation for 1 h. Beads were washed with NETN buffer and incubated either with protein lysates from HEK293T, HeLa, or COS-7 cells or with purified MBP or MBP-PABPC1 at 4°C with agitation overnight. Where indicated, 1 to 1,000 μg of poly(A) RNA, 1 μg of poly(C) RNA, or 150 U of RNase I_f (New England BioLabs) was added to the overnight pulldown reaction. After the beads were washed three times with NETN buffer, bound proteins were eluted with 40 μl of 2× Laemmli sample buffer, resolved by 10% SDS-PAGE, and subjected to Western blotting.

RESULTS

Viral proteins induce selective nuclear relocalization of PABPC and its dissociation from translation complexes.

One of the more dramatic examples of PABPC relocalization occurs upon viral infection. In the case of KSHV, PABPC relocalization is driven by the viral SOX protein, which promotes widespread mRNA turnover (26, 46). Because the functional consequences of PABPC nuclear relocalization have been best characterized in KSHV SOX-expressing cells, we initially used this viral protein to investigate the mechanism(s) of PABPC import. We first examined if PABPC-associated proteins were similarly relocalized in SOX-expressing cells, since during heat shock both PABPC and eIF4G are recruited to the nucleus (49). Immunofluorescence analysis (IFA) revealed that while there was a dramatic nuclear accumulation of endogenous PABPC, the PABPC-associated proteins eIF4G, eIF4E, eRF3, Paip1, and Paip2a remained strictly cytoplasmic (Fig. 1 A; quantitation in Fig. 1B). Thus, PABPC appears to be selectively relocalized to the nucleus by SOX.

Fig. 1.

PABPC is selectively relocalized to the nucleus by SOX. (A) HEK293T cells were transfected with empty vector or a plasmid expressing KSHV SOX and, 24 h later, subjected to immunofluorescence assays with anti-PABPC, anti-eIF4G, anti-eIF4E, anti-eRF3, anti-Paip1, anti-Paip2a, or anti-SOX antibodies. DAPI staining was used to visualize nuclei. Insets in panels of SOX-expressing cells stained for PABPC represent longer exposures to reveal PABPC in the cytoplasm of cells lacking SOX. (B) Percentages of nuclear PABPC harboring SOX-expressing cells were counted in multiple fields of view from three independent experiments. Error bars indicate SEM.

This observation suggested either that only the population of PABPC not associated with these proteins is targeted or that its interactions with one or more proteins are disrupted prior to its relocalization. To distinguish between these alternatives, we monitored interactions of transfected HA-PABPC1 with endogenous eIF4E, eRF3, and Paip2a in SOX-expressing cells. Associations of poly(A) RNA-bound PABPC with eIF4E (through eIF4G) and eRF3 occur during translation (13, 31, 33). However, interactions of PABPC with Paip2a, a repressor of translation, are RNA independent and separate from the association of PABPC with translation factors (38–40). A reduced amount of eIF4E and eRF3 coimmunoprecipitated with HA-PABPC1 in SOX-expressing cells; however, its level of interaction with Paip2a remained unchanged (Fig. 2 A). This was also the case in cells expressing two additional viral proteins that induce mRNA turnover and relocalize PABPC, MHV68 muSOX and HSV-1 virion host shutoff protein (vhs) (Fig. 2B and D). These observations were not an artifact of PABPC overexpression, as endogenous PABPC similarly coimmunoprecipitated dramatically reduced levels of eIF4E but not Paip2a in both KSHV SOX- and MHV68 muSOX-expressing cells (Fig. 2C). These data suggest that these viral proteins specifically target the population of PABPC in translation complexes.

Fig. 2.

Interactions of PABPC with translation factors are disrupted by viral proteins. HEK293T cells were transfected with empty vector or plasmids expressing HA-PABPC1 with or without KSHV SOX (A), MHV-68 muSOX (B), or the wild-type or catalytically inactive mutant (mut) of HSV-1 vhs (D) for 24 h. HA-PABPC1 was immunoprecipitated using anti-HA agarose beads, and bound proteins were resolved by SDS-PAGE and detected by Western blotting with anti-eIF4E, anti-eRF3, anti-Paip2a, anti-HA, and antiactin antibodies. Actin served as a loading and specificity control. (C) HEK293T cells were transfected with empty vector or plasmids expressing SOX or muSOX for 24 h. Endogenous PABPC was immunoprecipitated using anti-PABPC antibodies, and bound proteins were resolved and detected as described above.

Expression of a cytoplasmic deadenylase drives nuclear import of PABPC1.

Given that during translation, mRNA-bound PABPC associates with eIF4E and eRF3 and that each of the above viral proteins promotes cytoplasmic mRNA turnover, we hypothesized that the selective relocalization of PABPC occurs in response to mRNA depletion. However, as viral proteins are often multifunctional, PABPC relocalization might also be a consequence of additional cytoplasmic perturbations by SOX, muSOX, or vhs. Therefore, in an effort to specifically assess the extent to which binding to cytoplasmic RNA influences the subcellular distribution of PABPC1, we examined its localization upon depletion of poly(A) mRNA in a virus-independent manner. To this end, we used a cytoplasmic version of human Caf1z, a previously characterized nuclear deadenylase that can function in the cytoplasm upon mutation of its nuclear localization signal (NLS) (hCaf1z ΔNLS) (73). hCaf1z ΔNLS has been shown to catalyze rapid deadenylation of reporter mRNAs followed by slow 3′-5′ exonucleolytic decay (73). In contrast, reporter mRNA stability was unaffected in cells expressing the nuclear hCaf1z or catalytic mutant versions of hCaf1z (hCaf1z DEAA, hCaf1z ΔNLS DEAA) (73). We first confirmed that all of these proteins were expressed to similar levels and did not influence HA-PABPC1 protein expression (Fig. 3 A); then we showed that overexpression of hCaf1z ΔNLS, but not of wild-type nuclear hCaf1z or a nuclear or cytoplasmic catalytic mutant (hCaf1Z DEAA or ΔNLS DEAA, respectively), depleted a GFP reporter mRNA in cells (Fig. 3B). To determine whether the reduction in mRNA levels in hCaf1z ΔNLS-expressing cells was a consequence of enhanced deadenylation, tetracycline-responsive plasmids expressing GFP and β-globin were cotransfected into HeLa Tet-off cells together with either vector, hCaf1z ΔNLS, or hCaf1z ΔNLS DEAA. Following a pulse of transcription, the decay kinetics of each mRNA was followed over time (Fig. 3C). Characteristic of enhanced deadenylation, both the GFP and β-globin mRNAs decreased in size and abundance over the 5-h time course specifically in cells expressing hCaf1z ΔNLS (Fig. 3C).

Fig. 3.

Expression of a cytoplasmic deadenylase drives nuclear relocalization of PABPC. (A) HEK293T cells were transfected with the indicated plasmids for 24 h, and protein lysates were Western blotted with anti-Flag, anti-myc, anti-HA, or antiactin (loading control) antibodies. (B) HEK293T cells were transfected as described for panel A, and 24 h later, total RNA was isolated, resolved by agarose-formaldehyde gel electrophoresis, and Northern blotted with ³²P-labeled GFP and 18S probes. GFP and 18S levels were quantified using ImageJ software, and GFP levels were normalized to 18S levels. (C) HEK293T cells were transfected with Tet-responsive GFP and β-globin constructs along with empty vector, hCaf1z ΔNLS, or hCaf1z ΔNLS/DEAA for 24 h in the presence of 50 ng/ml tetracycline. Transcription was pulsed by the removal of tetracycline for 5 h and blocked by the addition of 1 μg/ml tetracycline. RNA was isolated from cells at the indicated time points thereafter, resolved by agarose-formaldehyde gel electrophoresis, and Northern blotted with ³²P-labeled GFP, β-globin, and 18S probes. (D) HEK293T cells were transfected with HA-PABPC1 alone or with the indicated Flag-hCaf1z plasmids for 24 h and subjected to immunofluorescence assay with anti-HA and anti-Flag antibodies. DAPI was used to visualize nuclei, and the right panels represent a merge between anti-HA and DAPI signals, anti-HA and anti-Flag signals, or anti-HA and anti-myc signals. (E) HA-PABPC1-expressing cells were counted in multiple fields of view from three independent experiments, and the percentage of transfected cells harboring nuclear HA-PABPC1 was calculated. Error bars indicate SEM.

We next tested whether enhanced cytoplasmic deadenylation was sufficient to alter the localization of PABPC. Indeed, expression of hCaf1z ΔNLS readily promoted nuclear accumulation of HA-PABPC1 (Fig. 3D; quantitation in Fig. 3E). This effect was dependent on its ability to promote cytoplasmic deadenylation, as neither the nuclear wild-type protein nor the nuclear or cytoplasmic catalytic mutants defective for deadenylation were capable of relocalizing HA-PABPC1 (Fig. 3D and E). Thus, the abundance of poly(A) RNA in the cytoplasm directly influences the steady-state localization of PABPC1.

PABPC RNA recognition motifs are required for nuclear import.

PABPC has four N-terminal RNA recognition motifs (RRMs) that mediate poly(A) binding and interactions with eIF4G, Paip1, and Paip2a (15, 33, 39, 40), followed by a central linker region involved in self-association (52), and a helical C terminus that interacts with Paip1, Paip2a, and eRF3 (13, 39, 60). In order to determine which regions of PABPC are required for nuclear import, we generated a panel of PABPC deletion mutants that lacked either the first two RRM domains (ΔRRM 1 + 2), the second two RRM domains (ΔRRM 3 + 4), all four RRM domains (ΔRRM 1–4), the linker region (ΔLinker), or the C-terminal region (ΔC-term) (Fig. 4 A and B). Interestingly, localization of some of the mutants differed from that of wild-type HA-PABPC1 (Fig. 4C). Specifically, in some cells HA-PABPC1 ΔRRM 1 + 2 was both nuclear and cytoplasmic, and HA-PABPC1 ΔRRM 1–4 was consistently present in both cellular compartments, suggesting that it lacks the determinant(s) that normally prevents nuclear accumulation (Fig. 4C). All other mutants exhibited strongly cytoplasmic staining similar to that of the full-length protein (Fig. 4C).

Fig. 4.

Expression and localization of PABPC mutants. (A) Diagram showing wild-type PABPC1, which possesses four N-terminal RRM domains followed by a central linker region and a conserved helical carboxyl terminus (5H), as well as the various HA-PABPC1 mutants. (B) HEK293T cells were transfected with the indicated wild-type and mutant HA-PABPC1 plasmids for 24 h. Equivalent amounts of protein lysate were then Western blotted using anti-HA antibodies. Actin serves as a loading control. (C) HEK293T cells were transfected with the indicated HA-PABPC1 plasmids and subjected to immunofluorescence assays with anti-HA antibodies. Nuclei are stained with DAPI. Right panels represent a merge between anti-HA IFA and DAPI signals.

To identify PABPC domains required for nuclear import by the herpesviral proteins and the cytoplasmic deadenylase, we coexpressed them with each HA-PABPC1 mutant and monitored the localization of the mutants by immunofluorescence analysis with anti-HA antibodies. Only the mutant lacking all RRMs (HA-PABPC1 ΔRRM 1–4) was not relocalized in SOX, muSOX, vhs, or hCaf1z ΔNLS-expressing cells (Fig. 5 A and B; quantitation in Fig. 5C). We have observed that additional mutants lacking RRM 2 + 3 + 4, RRM 1 + 2 + 3, RRM 1 + 3, or RRM 2 + 4 remain subject to at least partial relocalization by SOX (data not shown), indicating that at least one RRM must be present for nuclear import. Collectively, our results suggest that PABPC1 is retained in the cytoplasm via its interactions with RNA through the RRMs and that widespread depletion of cytoplasmic mRNA liberates this protein to enter the nucleus.

Fig. 5.

PABPC RNA recognition motifs are required for nuclear import. HEK293T cells were transfected with the indicated HA-PABPC1 mutants along with SOX, muSOX, and vhs (A) or Flag-hCaf1z ΔNLS (B) for 24 h and subjected to immunofluorescence assays with anti-HA antibodies. Nuclei are stained with DAPI. Right panels represent a merge between the anti-HA IFA and DAPI signals. (C) HA-PABPC1 wild-type (wt) or mutant-expressing cells were counted in multiple fields of view from two to three independent experiments, and the percentage of transfected cells harboring nuclear wt or mutant HA-PABPC1 was calculated. Error bars indicate SEM.

Nuclear import of PABPC is mediated through direct interactions with importin α.

Proteins destined for the nucleus are translocated across the nuclear pore by the classical import machinery. Proteins bind directly either to importin β or to the adaptor importin α, which then engages importin β, and the complex is transported into the nucleus in an energy-dependent manner (50, 69). We investigated whether PABPC enters the nucleus through the classical nuclear import pathway by testing its interactions with multiple isoforms of recombinant GST-tagged importin α (α1, α3, α5, and α7) and importin β proteins. GST pulldown assays with protein lysates from HEK293T, HeLa, and COS7 cells revealed that endogenous PABPC interacts predominantly with importin α3 and to some extent with other importin α proteins (particularly α7) but not with importin β (Fig. 6 A).

Fig. 6.

PABPC interacts with importin α. (A) Recombinant GST and GST-fused importin α1, α3, α5, α7, and β were purified over glutathione-Sepharose beads and incubated with 300 to 500 μg of lysate from HEK293T, HeLa, or COS-7 cells at 4°C overnight. Bound proteins were resolved by SDS-PAGE and detected by Western blotting with anti-PABPC antibodies. The input was 5% of the lysate used in the pulldown. (B to E) HEK293T cells were transfected either with the indicated HA-PABPC isoform (B) or the indicated HA-PABPC1 mutant (C to E) for 24 h, and the protein lysates were subjected to GST pulldowns with GST-importin α3 (B, C, and E) or additional GST-importin isoforms (D) and Western blotted with anti-HA antibodies. Where indicated, actin or β-tubulin served as a loading and specificity control, and the levels of GST fusion proteins used in each assay are shown by Coomassie staining.

Although PABPC1 is the most highly expressed isoform, three other cytoplasmic isoforms have been identified with variable expression patterns, namely, PABPC3, a testis-specific protein (24), PABPC4, an inducible isoform (43, 79), and PABPC5, an X-chromosome-encoded isoform (8). Similarly to PABPC1, PABPC4 and PABPC5 also accumulate in the nucleus upon expression of SOX (data not shown) (43) and therefore may be imported into the nucleus by importin α. Indeed, PABPC4-HA and HA-PABPC5 were also capable of interacting with importin α3 (Fig. 6B). Interestingly, PABPC5 is composed only of four RRM domains, suggesting that the RRMs are sufficient for interactions with importin α as well as nuclear import.

Based on our results thus far, we predicted that the interaction of importin α with PABPC1 would be mediated through its RRMs. GST pulldown assays with recombinant importin α3 confirmed that residues important for binding reside within one or more RRMs, as only the HA-PABPC1 ΔRRM 1–4 mutant failed to interact with importin α3 (Fig. 6C). This mutant was also unable to bind importin α1, α5, and α7 and importin β in GST pulldown assays (Fig. 6D). Our observation that mutants HA-PABPC1 ΔRRM 1 + 2 and HA-PABPC1 ΔRRM 3 + 4 both interact with importin α3 and accumulate in the nucleus after cytoplasmic mRNA depletion indicate that multiple importin α binding sites exist in the PABPC RRMs (Fig. 6C). Consistent with this finding, expression of PABPC RRM 1 + 2 or RRM 3 + 4 alone was sufficient for interaction with GST-importin α3 (Fig. 6E).

Protein interactions with the importins are generally mediated through a basic nuclear localization signal (NLS) (45). However, PABPC does not possess a predicted NLS and therefore may instead interact with importin α either indirectly through another NLS-containing cellular factor(s) or directly through one or more noncanonical NLS-like motifs within its RRMs. In order to distinguish between these two possibilities, we tested for an interaction between purified recombinant MBP-tagged PABPC1 and purified GST-tagged importin isoforms. GST pulldown assays demonstrated that MBP-PABPC interacts directly with multiple isoforms of importin α, including α3 and α5, but not with importin β (Fig. 7 A). The fact that we detected an interaction between PABPC1 and importin α5 in vitro but not in cells indicates that the preference for specific isoforms may be influenced by other cellular components. Collectively, these data suggest that PABPC harbors noncanonical NLS-like motifs within the RRMs that mediate direct interactions with importin α.

Fig. 7.

Direct interaction between PABPC and importin α is antagonized by poly(A) RNA. (A) Purified MBP-PABPC1 was mixed with purified GST or the indicated GST-importin bound to glutathione-Sepharose beads at 4°C with agitation for 12 h. Bound proteins were resolved by SDS-PAGE and Western blotted with anti-PABPC antibodies. The input was 50% of the purified MBP-PABPC used in the pulldown. (B and C) GST pulldowns were repeated as described for panel A, with the addition of either poly(A) RNA (1 μg [1:1 molar ratio], 10 μg, or 100 μg), 150 U of RNase I_f, or poly(C) RNA (1 μg [1:1 molar ratio]). The input was 10% (Fig. B) or 1% (Fig. C) of the purified MBP-PABPC1 used in the pulldown. GST and GST-fusion proteins were visualized by Coomassie staining.

The interaction of PABPC1 with importin α is antagonized by poly(A) RNA.

Our observations thus far indicated that the ability to bind poly(A) RNA plays a significant role in retaining PABPC in the cytoplasm. However, our prior experiments did not allow us to distinguish if this is specifically due to masking of the NLS-like motifs in PABPC by its association with RNA or as a consequence of its RNA-dependent protein interactions. To resolve these two possibilities, we assessed how addition or depletion of RNA influenced the direct interaction between purified MBP-PABPC1 and importin α3. The addition of 1 μg (representing a 1:1 molar ratio) of poly(A) RNA was, in fact, sufficient to completely block the interaction between GST-importin α3 and MBP-PABPC1 in GST pulldown assays (Fig. 7B). This disruption is not a nonspecific consequence of RNA charge, as addition of poly(C) RNA, which cannot be bound by PABPC (1, 10, 41, 42), does not block the association of PABPC with importin α3 (Fig. 7C). Although E. coli does not contain abundant polyadenylated RNA, given that PABPC can bind other RNA sequences as well (10, 42, 65), we hypothesized that any copurifying bacterial RNA might reduce its interactions with importin α. We therefore treated the purified proteins with RNase I_f, which cleaves RNA after each base, prior to performing the GST pulldowns. In agreement with our prediction, the RNase treatment significantly enhanced binding of PABPC to importin α3 (Fig. 7B). These data argue that removal of RNA unmasks sequences in PABPC that interact with the nuclear import machinery and that PABPC forms mutually exclusive interactions with poly(A) RNA and importin α3 within the cytoplasm.

DISCUSSION

PABPC has a very high affinity for poly(A) sequences (2 to 7 nM K_D [equilibrium dissociation constant]), and our observations indicate that this interaction plays a significant role in retaining PABPC in the cytoplasm (41, 42, 62). In particular, depletion of cytoplasmic mRNA, by divergent viral proteins or a cytoplasmic deadenylase, results in relocalization of PABPC to the nucleus in a manner dependent on its RRMs. Furthermore, poly(A) RNA, but not poly(C) RNA, disrupts interactions of purified PABPC with the nuclear import machinery, whereas RNase treatment enhances them. We therefore propose a model in which the localization of PABPC can be controlled by the levels of poly(A) RNA in the cytoplasm (Fig. 8). Normally, the concentration of cytoplasmic mRNA is likely sufficient to retain the majority of PABPC in this locale. However, if this balance is offset, for example by extensive mRNA degradation, an excess of non-RNA-bound PABPC accumulates, exposing sequences within the RRMs that associate with importin α, leading to nuclear translocation. It is important to note that a small fraction of PABPC is normally found in the nucleus associated with pre-mRNAs, where it is hypothesized to facilitate mRNA export (32). However, high levels of nuclear PABPC promote hyperadenylation and inhibit mRNA export (43), highlighting the importance of restricting excess import of PABPC into the nuclei of unstressed cells.

Fig. 8.

Model depicting mRNA turnover-induced nuclear accumulation of PABPC. In normal cells, PABPC is retained in the cytoplasm through its interactions with mRNA or other cellular proteins such as Paip2a and paxillin. At basal levels of mRNA turnover, a small fraction of PABPC shuttles in and out of the nucleus. However, upon widespread mRNA turnover, non-RNA bound PABPC accumulates and can interact with importin α to transit across the nuclear pore. High levels of nuclear PABPC promote hyperadenylation and nuclear retention of mRNA, thereby blocking gene expression. Abbreviations used in the model are as follows: C, PABPC; N, PABPN; α, importin α; 2a, Paip2a; Pax, paxillin; 4G, eIF4G; 4E, eIF4E; F3, eRF3.

Interestingly, PABPC levels are normally controlled in part through an autoregulatory mechanism. The 5′ UTR of the PABPC mRNA contains a 50- to 70-nt tract with stretches of 7 to 9 A residues, which are bound by PABPC, presumably once the protein concentration in the cytoplasm reaches a particular threshold (17, 61). This binding has been shown to block translation, thereby reducing PABPC accumulation (4, 17, 77). One measure of this threshold may be the amount of non-poly(A) tail-bound PABPC, an excess of which could then target the relatively short oligo(A) residues in the PABPC 5′ UTR.

PABPC has four RRMs that collectively mediate poly(A) binding, although the first two RRMs are sufficient for wild-type levels of interaction with poly(A) RNA (2, 10). In contrast, RRMs 3 and 4 have 10-fold reduced affinity for poly(A) but can bind other RNA sequences (10). Mutational analysis of PABPC supports the conclusion that the poly(A) binding capacity of PABPC is an important determinant of cytoplasmic localization. The mutant lacking RRMs 1 + 2 accumulates preferentially in the nucleus relative to other mutants, likely because it has reduced affinity for poly(A) RNA in the cytoplasm. In agreement with our model, a deletion in either the linker or the C-terminal domains of PABPC does not affect PABPC localization, as these domains do not interact with poly(A) sequences (2). Additionally, all mutants that retain at least one RRM domain are subject to nuclear relocalization upon cytoplasmic mRNA depletion, bolstering the idea that cytoplasmic poly(A) RNA is indeed a retention factor for PABPC. In the absence of all four RRMs, PABPC localization is no longer responsive to changes in mRNA levels, although it is constitutively present throughout both the nucleus and the cytoplasm. This may be a consequence of passive diffusion across the nuclear pore, as the ∼30-kDa size of the ΔRRM 1–4 mutant is below the minimum size required for active transport (71).

Interactions of PABPC with RNA and with importin α appear to be antagonistic, arguing that while bound to RNA, the sequences within PABPC that mediate nuclear import are masked. Given our mutant data, we predict that at least two noncanonical nuclear signals exist within PABPC RRMs. In particular, PABPC RRMs 1 + 2 or RRMs 3 + 4 alone are both able to bind importin α. The ability of RRM sequences to interact with the nuclear import machinery is not unprecedented, as RRMs in other proteins have been implicated in facilitating nucleocytoplasmic transport (12). For example, both the yeast Lhp1p RRM and the pab1 RRM 4 interact directly with yeast importin Sxm1/Kap108 (9, 58, 59). Furthermore, the RRM domain of the trypanosome TcUBP1 protein behaves as an NLS, mediating nuclear import of this predominantly cytoplasmic protein, particularly under conditions of arsenite-induced stress (11, 12). Interestingly, TcUBP1 has been shown to interact with TcPABP1, though whether this interaction influences its localization has yet to be determined (19).

Although we have shown that cytoplasmic mRNA abundance is an important determinant of PABPC localization, this is unlikely to be its sole regulator. RNA-dependent protein interactions could also contribute to its nuclear-cytoplasmic distribution. In this regard, it has been calculated that in HeLa cells PABPC abundance exceeds the availability of cytoplasmic mRNA (27). Presumably, a portion of the excess PABPC is complexed with other cytoplasmic proteins, such as Paip2a and paxillin, although the abundance of these proteins must not be sufficient to retain PABPC in the cytoplasm following widespread mRNA degradation (39, 76). However, depletion of paxillin, a focal adhesion adaptor protein, results in a modest nuclear accumulation of PABPC in 3T3 cells, suggesting that paxillin also influences PABPC localization (16, 75, 76). We were unable to detect an interaction between paxillin and PABPC in HEK293T cells or changes in levels of paxillin in SOX-expressing cells (data not shown), indicating that roles for paxillin in PABPC localization may be cell type or context specific. PABPC relocalization also occurs during a variety of nonviral cellular stresses, in particular upon heat shock, although it has been reported to occur during transcriptional block and oxidative stress as well (2, 49, 64). Heat shock results in a global decrease in cap-dependent translation and nuclear relocalization of both PABPC and eIF4G (3, 49, 57). The observation that eIF4G is not relocalized by the KSHV SOX protein suggests that either the mechanism of PABPC import is distinct upon viral infection versus heat shock or that heat shock also affects separate pathways that influence eIF4G.

Sensing or responding to widespread alterations in cellular transcript abundance via PABPC relocalization may be one means by which cells react to infection with a variety of pathogens. Viruses in particular often come armed with an arsenal of mechanisms to suppress activation of innate immune responses. However, select viruses that benefit from host gene expression shutdown might commandeer this response to facilitate their own replication. Thus, further delineation of the variety of pathways that control PABPC localization is certain to provide insight into how cells broadly regulate gene expression patterns during homeostasis as well as under conditions of stress.