Subcellular Localization of Multiple PREP2 Isoforms Is Regulatedby Actin, Tubulin, and Nuclear Export*

Received for publication, June 1, 2004, and in revised form, August 18, 2004Published, JBC Papers in Press, August 30, 2004, DOI 10.1074/jbc.M406046200

Klaus Haller‡§¶, Isabel Rambaldi§, Eugene Daniels�, and Mark Featherstone‡§**‡‡§§

From the ‡McGill Cancer Centre and §Department of Biochemistry, Departments of **Medicine, ‡‡Oncology, and�Anatomy and Cell Biology, McGill University, Montreal, Quebec H3G 1Y6, Canada

The PREP, MEIS, and PBX families are mammalianmembers of the TALE (three amino acid loop extension)class of homeodomain-containing transcription factors.These factors have been implicated in cooperative DNAbinding with the HOX class of homeoproteins, but PREPand MEIS interact with PBX in apparently non-HOX-de-pendent cooperative DNA binding as well. PREP, MEIS,and PBX have all been reported to reside in the cyto-plasm in one or more tissues of the developing verte-brate embryo. In the case of PBX, cytoplasmic localiza-tion is due to the modulation of nuclear localizationsignals, nuclear export sequences, and interaction witha cytoplasmic anchoring factor, non-muscle myosinheavy chain II B. Here we report that murine PREP2exists in multiple isoforms distinguished by interactionwith affinity-purified antibodies raised to N- and C-ter-minal epitopes and by nuclear versus cytoplasmic local-ization. Alternative splicing gives rise to some of thesePREP2 isoforms, including a 25-kDa variant lacking theC-terminal half of the protein and homeodomain andhaving the potential to act as dominant-negative. Wefurther show that cytoplasmic localization is due to theconcerted action of nuclear export, as evidenced by sen-sitivity to leptomycin B, and cytoplasmic retention bythe actin and microtubule cytoskeletons. CytoplasmicPREP2 colocalizes with both the actin and microtubulecytoskeletons and coimmunoprecipitates with actin andtubulin. Importantly, disruption of either cytoskeletalsystem redirects cytoplasmic PREP2 to the nucleus. Wesuggest that transcriptional regulation by PREP2 ismodulated through the subcellular distribution of mul-tiple isoforms and by interaction with two distinct cy-toskeletal systems.

Patterning of the embryonic antero-posterior axis comes un-der the control of HOX and TALE1 class homeoproteins (1).

Members of the TALE class are characterized by sequencerelatedness and by a three amino acid loop extension betweenhelices one and two of the homeodomain (2). Families withinthe TALE group include PBC, MEIS, and PREP. Among thePBC family members, the fly extradenticle and vertebrate PBXproteins have been most intensively studied. PBC proteins bindDNA cooperatively with many but not all HOX proteins. Addi-tionally, they form stable DNA-binding heterodimers withMEIS and PREP family members (2). The ability of PBC pro-teins to interact with both HOX and MEIS/PREP partnersresults in DNA-bound heterotrimeric complexes that have beenimplicated in gene control in vitro and in vivo (3–8). However,the instances of HOX-independent gene control by TALE het-erodimers have also been documented (9, 10), consistent withthe presence of these proteins in tissues devoid of Hox geneexpression.

In addition to the homeodomain, two conserved N-terminaldomains, PBC-A and PBC-B, have been defined in PBC familyproteins (11). These regions are implicated in numerous pro-cesses including nuclear export, transcriptional repression,MEIS/PREP interaction, and binding to non-muscle myosinheavy chain II B (2). Vertebrate MEIS proteins and fly homo-thorax are members of the MEIS family, whereas PREP isrepresented in vertebrates but not in flies. The related MEISand PREP families share N-terminal regions termed homologyregion 1 (HR1) and HR2 (9) with HR2 implicated in binding toPBC family proteins (12).

Many TALE class homeoproteins have been shown to residein the cytoplasm at certain times or sites of embryonic devel-opment. For example, extradenticle and PBX are cytoplasmicin the distal limb primordia of insects and vertebrates, respec-tively (13, 14). This distribution is critical for normal develop-ment, because forced nuclear accumulation of PBC proteinscorrelates with limb proximalization (13, 15, 16). Subcellularlocalization is controlled by protein kinase A (17) and by inter-action with MEIS family proteins (14). MEIS binding masksNES while exposing NLS within the PBC N terminus (18–20).Reciprocally, MEIS is dependent on PBX for nuclear localiza-tion (21–24).

Recently, it has been shown in fly embryos and mammalian cellsthat PBX is retained in the cytoplasm by interaction with non-muscle myosin heavy chain II B (22), raising the possibility thatsubcellular localization of TALE class proteins could be influencedby signaling pathways impinging on the cytoskeleton.

The PREP (also known as PKNOX) (25) family has twomembers in mice and humans (9, 25–29). In complexes withPBX, PREP1 regulates the transcription of HOX-dependent(Hoxb2) and HOX-independent (urokinase plasminogen activa-tor, glucagon) target genes (5, 9, 10). Both PREP1 and PREP2

* This work was supported by Operating Grant 36409 from the Ca-nadian Institutes of Health Research (CIHR). The costs of publication ofthis article were defrayed in part by the payment of page charges. Thisarticle must therefore be hereby marked “advertisement” in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

¶ Supported by a McGill Graduate Studies Fellowship.§§ A Chercheur-National of the Fonds de la Recherche en Sante du

Quebec. To whom correspondence should be addressed: McGill CancerCenter, McGill University, McIntyre Medical Science Bldg., Rm. 714,3655 Promenade Sir W. Osler, Montreal, Quebec H3G 1Y6, Canada.Tel.: 514-398-8937; Fax: 514-398-6769; E-mail: mark.featherstone@mcgill.ca.

1 The abbreviations used are: TALE, three amino acid loop extension; HR,homology region; NLS, nuclear localization signal; NES, nuclear exportsequence; PREP2NAB, N-terminal PREP2 antibody; PREP2CAB, C-termi-nal PREP2 antibody; DBD, DNA-binding domain; HA, hemagglutinin;TRITC, tetramethylrhodamine isothiocyanate; LMB, leptomycin B; GFP,green fluorescent protein; siRNA, small interfering RNA; GAPDH, glycer-

aldehyde-3-phosphate dehydrogenase; RT, reverse transcriptase; DAPI,4�,6-diamidino-2-phenylindole; dpc, days post coitum.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 279, No. 47, Issue of November 19, pp. 49384–49394, 2004© 2004 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org49384

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

have been shown to accumulate in the cytoplasm of culturedand embryonic cells (19, 28, 29). When we used an affinity-purified antibody directed against an N-terminal epitope toassess PREP2 distribution in the mouse embryo, we notedwidespread but essentially cytoplasmic expression (29). Wereport here that a second affinity-purified antibody directedagainst the PREP2 C terminus reveals nuclear staining andthat between them the two antibodies recognize at least fivedistinct PREP2 isoforms. One of these isoforms has a molecularmass of 25 kDa and is the result of alternative splicing. Wefurther show that cytoplasmic PREP2 is associated with theactin and microtubule cytoskeletons and that this cytoplasmiclocalization is dependent on CRM-1-mediated nuclear exportand the integrity of cytoskeletal networks.

EXPERIMENTAL PROCEDURES

Primary Antibodies—The PREP2 antibodies were raised as previ-ously described (29). The N-terminal PREP2 antibody, PREP2NAB, istargeted against the region encoded by exon 2 (29), whereas the C-terminal PREP2 antibody (PREP2CAB) is targeted against residues341–392. Cross-reactivity to PREP1 was removed from PREP2CAB bypassing over a column bearing a fusion of glutathione S-transferase tomurine PREP1 residues 322–380. The following antibodies were alsoused: anti-GAL4 DNA-binding domain (DBD) mouse monoclonal anti-body clone RK5C1 raised against residues 94–147 (sc-510, Santa CruzBiotechnology); mouse monoclonal anti-�-tubulin antibody clone tub 2.1(T-4026, Sigma); mouse monoclonal anti-HA-11 antibody (MMS-101R,BabCO); mouse monoclonal anti-actin antibody, clone AC 40 (A470,Sigma); anti-PBX1 (P-20) antibody (sc-889, Santa Cruz Biotechnology);anti-TATA-binding protein antibody (SI 1) (sc 273, Santa Cruz Biotech-nology); anti-FLAG-agarose beads (A2220, Sigma); and rhodamine-con-jugated phalloidin (R-415, Molecular Probes).

Secondary Antibodies—The secondary antibodies used were as fol-lows: marina blue-conjugated anti-mouse IgG (M-10991, MolecularProbes); rhodamine (TRITC)-conjugated goat anti-mouse IgG (115-025-003, Jackson Laboratories); peroxidase-conjugated goat anti-mouse IgG(115-035-003, Jackson Laboratories); peroxidase-conjugated anti-rabbitIgG (A0545, Sigma); and fluorescein isothiocyanate-conjugated anti-rabbit IgG (F-7512, Sigma).

Drugs—The drugs used were as follows: cytochalasin D (C8273,Sigma); swinholide A (350-088-C010, Alexis Biochemicals); phalloidin(P2141, Sigma); paclitaxel (T1912, Sigma); nocodazole (M1404, Sigma);leptomycin B (LMB) (L2913, Sigma) kindly provided by Dr. M. Yoshida;and taxol (T1912, Sigma).

Constructs—The full-length Prep2 coding region was PCR-amplifiedwith NheI site-containing primers, subcloned into the vector TOPO II(Invitrogen), and sequenced (pPrep2-Topo). The Prep2-containing NheIfragment was then subcloned into the vector pFRED 143, 5� to theGFP-coding region (construct A). pFRED 143 contains a humanizedversion of a strong mutant of GFP (Kyoji Horie and George N. Pavlakis,National Cancer Institute-Advanced Bioscience Laboratories.) A DNAfragment encoding the N terminus of PREP2 (residues 1–267) wasfused to the GFP-coding region in the same manner (construct 2).Construct A was partially digested with NheI/XbaI, and the resultingfull-length Prep2-gfp fusion was introduced into XbaI-digested pRC/CMV plasmid resulting in construct 1. Gal-O45 is a mammalian expres-sion vector for the GAL4 DBD (30). For construct 3, pPrep2-Topo wasdigested with EcoRI/BamHI subcloned into BamHI-digested and-blunted Gal-045 3� to the GAL-DBD-coding region. The cloning ofPBX1-HA has been reported previously (18). The expression vector forthe G13R actin mutant was a generous gift from the Treisman labora-tory. For constructing the plasmid HA-PREP2-Myc, pPrep2-Topo wasdigested with NheI and subcloned into pHAv (18). This construct wasfurther cleaved with BamHI/ApaI and inserted into pcDNA4/Myc-His A(Invitrogen).

Cell Culture and Transfection—COS-7, NIH3T3, and primary mouseembryonic fibroblasts were cultured in Dulbecco’s modified Eagle’s me-dium supplemented with 10% fetal bovine serum (Invitrogen). Drosoph-ila S2 cells were cultured in Schneider’s Drosophila medium (Invitro-gen) with 10% fetal bovine serum. For immunohistological analysis,cells were grown on 12-mm-diameter round coverslips. Transfectionwas performed using LipofectAMINE Plus (NIH3T3) or LipofectAMINE2000 (COS-7) reagents (Invitrogen).

Cytoplasmic and Nuclear Fractionation of Cell Extracts—NIH3T3cells were scraped from confluent tissue culture dishes and washed

twice in phosphate-buffered saline, and cytoplasmic extracts were ob-tained by resuspending in hypotonic buffer (10 mM Tris, pH 7.4, 10 mM

NaCl, 3 mM MgCl, 0.5 mM dithiothreitol, 0.05% Nonidet P-40). Nucleiwere collected by centrifugation, and nuclear proteins were extracted inradioimmune precipitation assay buffer (50 mM HEPES, pH 7.5, 150mM NaCl, 0.1% SDS, 1% sodium deoxycholate, 1% Triton X-100). Com-plete EDTA-free protease inhibitor tablets (1873580, Roche Diagnos-tics) were added to each buffer. 40 �g of total protein were separated ona 10% SDS gel and analyzed by Western blotting.

RNA Interference—A mixture of two double-stranded siRNA oligonu-cleotides (Qiagen) was transfected into NIH3T3 cells using Lipo-fectAMINE Plus, and transfection efficiency was monitored by non-silencing fluorescein-labeled control siRNA. RNA was extracted usingTRIzol reagent followed by real-time PCR using SYBR Green TaqReadyMixTM for quantitative PCR (S1816, Sigma) in a Roche LightCy-cler as described previously (31). Quantification of the LightCycler datawas performed using standard curves of serially diluted plasmid DNAof known concentration. Sequences for primers are as follows: GAPDHsense primer 5�-AACGACCCCTTCATTGAC-3�; GAPDH antisenseprimer 3�-TCCACGACATACTCAGCAC-5�; Prep2 sense primer 5�-CCAGGTCCTGAGGATC-3� (16 bp); and Prep2 antisense primer (5�-CATGGAATTGGGGGAATTCTGCAGGAGGTC-3� (30 bp). PCR prod-ucts were run on a 1% agarose gel and stained with ethidium bromide.The sequences of siRNA oligonucleotides (one strand only) were asfollows: (a) 5�-CCCAGAUCCUGCUCCCAAA-3�; (b) 5�-UGUCUG-GAGUCUCCAAUAA-3�; and (c) 5�-CCAAGAUGCACAGUGAUAA-3�.Mixtures of siRNAs a and c or b and c gave comparable results.

Northern Analysis—Total RNA was isolated from P19 embryonalcarcinoma cells using TRIzol reagent. Poly(A)� RNA was then purifiedusing the GenElute mRNA kit (Sigma). 3 �g of poly(A)� RNA per wellwere run on a 1.2% formaldehyde-agarose gel and transferred to aNytran plus membrane (Schleicher & Schuell). Blots were sliced to giveone well of fractionated mRNA per membrane strip and prehybridizedand hybridized in Express hybridization solution and washed accordingto the manufacturer’s instructions. 2 � 106 cpm/ml of 32P-labeled anddenatured Prep2 probe were used in the hybridizations. Blots wereexposed to x-ray film with screens either overnight (375-nucleotide 5�probe) or for 7 days (156-nucleotide 3� probe and 153-nucleotide intron4 (int) probe). An adult mouse tissue poly(A)� blot purchased fromClontech was described previously (29). The region used for probemarked int is shown in Fig. 5A, whereas the 3� probe spanned thesequence coding for PREP2 amino acids 340–391.

Coimmunoprecipitation—NIH3T3 cells grown to confluency in 10-cmtissue culture dishes were harvested and washed in phosphate-bufferedsaline. Cells were lysed in buffer, pH 7.9, containing 50 mM NaCl, 1 mM

EDTA, 0.5% Triton X-100, and 20 mM HEPES. Cell debris was removedby a - min spin in a microcentrifuge, and the lysate was precleared with100 �l of protein A-agarose (16-156, Upstate Biotechnology). The im-munoprecipitation was performed for 1 h at 4 °C using 1.5–2 mg ofwhole cell extract and 1 �g of antibody/mg protein. Protein A-agarosebeads were applied for 2 h in a volume of 100 �l at 4 °C. The complexeswere washed twice with cold phosphate-buffered saline and eluted witha 1:1 ratio of water and 4� SDS sample buffer. The proteins wereseparated on an 8% SDS gel and analyzed by Western blotting with theappropriate antibodies. Taxol was used at a concentration of 2�M overnight.

Actin Fractionation—The protocol for actin fractionation wasadapted from Posern et al. (32). Phalloidin was added for 10 min to thelysis buffer by using 10 �M drug. Cyochalasin D was applied at aconcentration of 10 �M for 1 h to cells in culture.

Immunofluorescence Staining—Cells were grown on coverslips andfixed in methanol for 15 min at �20 °C, rinsed in phosphate-bufferedsaline, and permeabilized with 0.1% Triton X-100 for 2 min. Primaryantibodies were applied for 2 h, whereas secondary antibodies wereapplied for 1 h. Cells were mounted with GelTol water-based mountingmedium from Fisher (catalog number 28-607-87). Pictures were takenwith a Zeiss Axiovert 100 M confocal microscope and a Nikon digitalcamera DXM 1200 attached to a Nikon ECLIPSE E800 microscope. Theprotocol for immunohistological analysis of embryonic sections as wellas the procedures for fixation and sectioning has been described previ-ously (29).

RESULTS

A C-terminal Antibody Detects PREP2 in the Nucleus—Wehave previously reported that an affinity-purified antibody(PREP2NAB) directed against the murine PREP2 N terminus(amino acids 18–64) gives largely cytoplasmic reactivity in

Subcellular Localization of PREP2 49385

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

numerous tissues of 8.5–12.5-dpc mouse embryos (29). Al-though this result was consistent with other reports (19, 28), itwas surprising for a protein with a transcriptional function.Therefore, we raised and tested a second affinity-purified an-tibody (PREP2CAB) directed against the murine PREP2 Cterminus (amino acids 341–392). In Western analysis, bothPREP2NAB and PREP2CAB were highly specific for PREP2,failing to recognize the closely related PREP1, two additionalmembers of the TALE class of homeoproteins, and numerouspeptides within rabbit reticulocyte lysates (Fig. 1). A signal wasnot detected with these antibodies in untransfected DrosophilaS2 cells or when preimmune sera were used instead of primaryantibodies (data not shown). Moreover, preabsorption of theanti-sera with PREP2 peptides abolished reactivity in Westernand immunofluorescence studies (Fig. 2, K and L) (data notshown). Thus, PREP2NAB and PREP2CAB are specific forPREP2.

To assess the embryonic pattern of PREP2CAB reactivity,we performed immunohistological analysis on embryonic sec-tions of 10.5-dpc mouse embryos (Fig. 2, A–F). The tissue dis-tribution of PREP2 protein was consistent with our previousobservations (29). However, in striking contrast to results ob-tained with PREP2NAB, we now observed predominantly nu-clear staining. Tissues with marked nuclear signals includedthe epithelium of the choroid plexus (Fig. 2A), precartilagenousmesenchyme (Fig. 2C), cartilage of the intervertebral discs(Fig. 2D), dermal mesenchyme in the hindbrain area (Fig. 2E),and mesenchyme between the spinal ganglia (Fig. 2F). Asnoted previously (29), PREP2NAB gave cytoplasmic signals inthese same tissues (Fig. 2B). These data suggested that PREP2isoforms differing at their N and C termini localize to differentsubcellular compartments.

To confirm and extend these results, we used PREP2NABand PREP2CAB to examine PREP2 distribution in cells inculture (Fig. 2, G–L). In NIH3T3 cells, the PREP2NAB signalwas consistently cytoplasmic but with faint and reproduciblereactivity in the nucleus (Fig. 2G), whereas that of PREP2CABwas strongly nuclear (Fig. 2H). These observations were con-firmed in a number of cell lines including COS-7, transfectedS2, HEK293, and primary mouse embryonic fibroblasts (datanot shown). Preincubation with fusions of glutathione S-trans-ferase to the PREP2 N or C termini abolished reactivity (Fig. 2,K and L). Thus, PREP2 isoforms distinguished by N- andC-terminal-specific antibodies localize to distinct subcellularcompartments in both embryonic and cultured cells.

Distinct Isoforms of PREP2 Localize to Different SubcellularCompartments—To understand the basis of this differentiallocalization, we used Western blot analysis to examine PREP2

in cytoplasmic and nuclear fractions of cultured cells. Confirm-ing the results observed in immunofluorescence studies,PREP2NAB- and PREP2CAB-reactive species were distributedin a mutually exclusive fashion between the nuclear and cyto-plasmic compartments of untransfected NIH3T3 (Fig. 3A) andCOS-7 (data not shown). Strikingly, five PREP2 isoforms weredetected, none of which was recognized by both antibodies andnone of which was found in both subcellular compartments.Consistent with data from immunofluorescence, the majority ofPREP2NAB-reactive material was found as high molecularmass species in the cytoplasmic fraction but also as a lessabundant 25-kDa nuclear isoform.

To confirm the PREP2 identity of these isoforms, we per-formed siRNA analysis. A robust knockdown at the mRNAlevel could be detected (Fig. 3, B and C) and was accompaniedby reduced protein levels for all PREP2 species (Fig. 3D). Theseresults strongly support the PREP2 identity of all of the iso-forms detected by PREP2NAB and PREP2CAB.

Alternative Splicing Gives Rise to a 25-kDa PREP2 IsoformLacking the Homeodomain—We and others have identifiedcDNAs encoding the same PREP2 isoform in mouse and hu-man. To determine whether the product of this open reading

FIG. 1. Affinity-purified anti-PREP2 antibodies specificallyrecognize PREP2 but not other PREP2-related proteins of theTALE class. Top panel, 35S-labeled products produced by in vitrotranslation. Bottom panels, in vitro translated products from parallelreactions were probed with affinity-purified anti-PREP2 antibodies.Mock, unprogrammed reticulocyte lysate.

FIG. 2. PREP2 isoforms localize to different subcellular com-partments. A and C–F, immunohistological analysis of 10.5-dpc mouseembryonic sections with PREP2CAB. The tissue distribution ofPREP2CAB reactivity was as previously reported for PREP2NAB (29)but was in contrast predominantly nuclear. This is shown in the nu-cleus (arrow) of choroid plexus epithelium (A), in the precartilage mes-enchyme (C), by the cartilage in the intervertebral discs (D), in thedermal mesenchyme in the hind brain area (E), and in the mesenchymebetween the spinal ganglia (SG) (F). Note the absence of nuclear stain-ing with PREP2NAB in choroid plexus epithelium (B). G–L, immuno-fluorescent detection of PREP2 in NIH3T3 cells. G, PREP2NAB reac-tivity is predominantly cytoplasmic (93% cells). H, nuclear reactivity ofPREP2CAB (95% cells). I and J, DAPI-stained nuclei corresponding toG and H. K and L, preabsorption with glutathione S-transferase-PREP2 fusion protein abolishes immunoreactivity of both PREP2NABand PREP2CAB.

Subcellular Localization of PREP249386

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

frame is proteolytically processed to place the N- and C-termi-nal epitopes on different peptides, we expressed PREP2 simul-taneously tagged at its N terminus with HA and at its Cterminus with GFP (Fig. 4, A–C). Importantly, both tags werefound predominantly in the nucleus, arguing against proteo-lytic cleavage. Furthermore, expression and immunoblotting ofa similar construct bearing a C-terminal Myc tag revealed noevidence of cleavage products (Fig. 4C). When HA tags were

individually placed at either PREP2 terminus, again no cleav-age products were observed (data not shown). These data sug-gest that PREP2 isoforms distinguished by their N and Ctermini are not derived by proteolytic cleavage of the product ofthe known open reading frame.

PBX family members are the only known binding partners ofPREP proteins (5, 9, 10, 29). Thus, if the immunoreactivitydetected by PREP2NAB is indeed PREP2, it should be redis-tributed to the nucleus by overexpression of PBX. As previouslyshown with PREP1 (19), PBX directed strong nuclear localiza-tion of the PREP2 N terminus (Fig. 4D, compare with Fig. 6I).Importantly, PBX had a similar effect on endogenous PREP2detected with PREP2NAB (Fig. 4E). These results further con-firm the specificity of PREP2NAB.

To assess whether alternative splicing might give rise todifferent PREP2 isoforms differentially detected by our anti-bodies, we searched the publicly available databases. This scanyielded an alternatively spliced murine cDNA (GenBankTM

accession number AK084178). As a result of the retention ofintron 4, this transcript encodes a truncated peptide bearing anovel C terminus (Fig. 5A). This presumptive PREP2 isoformwould be recognized by PREP2NAB but not by PREP2CAB andhas a predicted molecular mass of 25 kDa corresponding to thesmallest isoform detected by PREP2NAB (Fig. 3A).

To determine whether the alternatively spliced transcriptencoding a 25-kDa PREP2 isoform is normally present in thecell, we designed primers specific to this form for use in RT-PCR. A specific band of the predicted size was detected fromRNA of murine P19 embryonal carcinoma cells. To assesswhether this message was present in mature processed mRNA,we performed Northern analysis on the poly(A)� fraction ofRNA from murine P19 embryonal carcinoma cells. We usedthree independent probes: a 5� probe previously used to detectPrep2 transcripts in adult and embryonic tissues (29); a 153-bpprobe corresponding precisely to the intronic sequencesuniquely present in the alternatively spliced cDNA; and a 3�probe downstream of the homeobox. As shown in Fig. 5B, the 5�probe detects two bands of 4.5 and 1.8 kb, comparable in size tothe two transcripts noted previously (29). The probe, which lieswithin intron 4 and which is specific to the alternatively splicedcDNA, detects a 1.4-kb band in poly(A)� mRNA of P19 cells. Asimilar band is likewise detected in the poly(A)� fraction ofadult liver (Fig. 5B) but not in several other tissues known toexpress Prep2 (data not shown). These results confirm thebiological relevance of this alternatively spliced cDNA andstrongly suggest that it is responsible for the 25-kDa PREP2peptide detected with PREP2NAB. Interestingly, the 3� probe,located downstream of the homeobox, detected further RNAs of4.7, 3.0, 2.6, and 1.4 kb, demonstrating multiple Prep2 tran-scripts that may well account for the additional peptide iso-forms we detect in Western analysis.

To determine the subcellular localization of the 25-kDaPREP2 isoform encoded by the alternatively spliced transcript,we cloned the cDNA by RT-PCR and fused it to sequencesencoding an HA tag. When expressed in 3T3 cells and detectedby an anti-HA antibody, the 25-kDa isoform was found tolocalize predominantly to the cytoplasm (Fig. 5C), consistentwith the largely cytoplasmic reactivity of PREP2NAB.

The PREP2 N Terminus Directs Cytoplasmic Localizationvia Multiple Mechanisms—Cytoplasmic localization directedby the PREP2 N terminus could be mediated by non-exclusiveactivities including nuclear export and cytoplasmic anchoring.Prep2 products are cytoplasmic in insect S2 cells (19). To testfor a role for active nuclear export, we expressed PREP2protein in S2 cells followed by treatment with the CRM-1inhibitor LMB. S2 cells showed clear nuclear accumulation of

FIG. 3. Multiple PREP2 isoforms are partitioned between thenucleus and cytoplasm and are sensitive to RNA interference. A,Western analysis of PREP2 isoforms in nuclear (N) and cytoplasmic (C)fractions of NIH3T3 cells. Lysates were run in parallel and probed withPREP2CAB or PREP2NAB (left panel). Cytoplasmic and nuclear frac-tions were also probed with antibodies against cytoplasmic (tubulin)and nuclear (TATA-binding protein, TBP) markers, whereas actin wasused as a loading control (right panel). B, reduction in the level of Prep2transcript-specific RT-PCR products from small interference-oligonu-cleotide-transfected NIH3T3 cells versus mock (upper panel) in compar-ison to gapdh control (lower panel). C, quantification of relative RT-PCRproduct abundance before (black bar) and after (gray bar) interference.D, Western analysis of NIH3T3 whole cell extracts detected withPREP2NAB and PREP2CAB before (lanes 1 and 3) and after (lanes 2and 4) RNA interference. Actin was used as loading control.

Subcellular Localization of PREP2 49387

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

PREP2NAB reactivity upon a 25 nM treatment for 4 h (data notshown). We then scored LMB-treated NIH3T3 cells for nuclearlocalization of endogenous PREP2NAB reactivity. Punctate nu-

clear staining appeared at concentrations of 5–25 nM LMB after4 h of treatment (Fig. 6, B and C), whereas strong nuclearstaining was observed at a concentration of 100 nM over the

FIG. 4. PREP2 subcellular localization is regulated on multiple levels. A, schematic representation of PREP2 and PBX1A constructs fusedto HA (red oval), GFP (green oval), or GAL-DBD (red rectangle). B, D, and E, fluorescence microscopy of COS-7 cells transfected with the indicatedconstructs. B, Double labeling of PREP2 protein at both termini reveals predominantly nuclear localization of both labels. C, substitution of theC-terminal GFP tag with a Myc tag followed by transfection into NIH3T3 cells and immunoblotting for both labels revealed no apparent cleavageproducts (lanes 3 and 4). The band indicated by the arrow (lane 4) was nonspecific, because it could also be detected in untransfected cells.Endogenous PREP2 detected by NAB and CAB is shown in lanes 1 and 2. D, coexpression of HA-tagged PBX1A (red; middle panel) directsPREP2-(1–267)-GFP (construct #2, green) to the nucleus (left panel; compare with Fig. 5I). Colocalization appears as yellow (right panel). E,endogenous NAB-reactive PREP2 is redirected to the nucleus (green) upon overexpression of HA-tagged PBX1A, which localizes to the nucleus(red), and the overlap is shown in yellow (right panel). N, nuclear; C, cytoplasmic.

Subcellular Localization of PREP249388

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

same time (Fig. 6D). To test whether the PREP2 N terminuscan direct cytoplasmic localization, we fused residues 1–267 toGFP (Fig. 4A) and assessed subcellular distribution. Whereasunfused GFP is found throughout the cell, the addition of thePREP2 N terminus results in predominantly cytoplasmic local-ization (Fig. 6I). Although PREP2-(1–267)-GFP did not relocateto the nucleus upon 25 nM LMB treatment for 4 h, nuclearaccumulation was detected after increasing the concentration

to 100 nM (Fig. 6J). To extend these observations, we fusedPREP2 residues 1–131, encompassing the PREP2NAB epitope,conserved regions HR1, and part of HR2, to the GAL4 DBD.The GAL4 DBD harbors two NLSs (33) and is normally foundin the nucleus. However, fusion to the PREP2 N terminusresults in almost total restriction of GAL4 immunoreactivity tothe cytoplasm (Fig. 6K), supporting nuclear export and/or cy-toplasmic retention roles for this region.

FIG. 5. An alternative splice product encodes a PREP2 isoform of 25 kDa. A, partial sequence of a murine Prep2 cDNA (GenBankTM

accession number AK084178) plus conceptually translated C terminus. Inverted triangles denote exon-exon and exon-intron junctions as indicated.Intron 4, which is spliced out of the cDNA encoding the originally reported PREP2 isoform, is retained in this variant. The probe used for Northernanalysis (below) is underlined. B, Northern analysis of poly(A)� RNA from undifferentiated murine P19 cells or adult mouse liver. For P19 cellpoly(A)� RNA, blots were hybridized with one of three probes: 5�, the same probe from the 5� end of the full-length Prep2 cDNA as used previously(29); int, a 153-bp probe corresponding to the unique portion of the alternative splice variant (underlined in A) within intron 4; 3�, a 156-bp probefrom the 3� end of the originally reported Prep2 cDNA. For liver, a commercial blot of poly(A)� RNA from adult tissues was hybridized with the“int” probe. Black arrowheads denote the position of hybridizing bands. C, the presumptive 25-kDa PREP2 isoform was tagged with HA, expressedin transiently transfected NIH3T3 cells, and detected by immunofluorescence with an anti-HA antibody. Localization was predominantlycytoplasmic.

Subcellular Localization of PREP2 49389

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

PREP2 Interacts with the Actin and Microtubule Cytoskel-etons—Three-dimensional reconstruction of GAL4-PREP2-(1–131) confocal immunofluorescent images revealed a cytoplas-

mic staining pattern highly reminiscent of cytoskeletalstructures (Fig. 6K). This suggested that interaction with oneor more components of the cytoskeleton may retain PREP2 inthe cytoplasm, as has been noted for the action of non-musclemyosin heavy chain II B on PBX (22). To test this proposal, weperformed colocalization studies in synchronized NIH3T3 cellsusing PREP2NAB and an anti-tubulin antibody. NIH3T3 cellsshowed a clear overlap of PREP2 with microtubules at all of thestages of mitosis (Fig. 7, A and B). To further verify theseobservations, we transfected GAL4-PREP2-(1–131) (Fig. 4, con-struct 3) into COS-7 cells and assayed for colocalization withtubulin (Fig. 7C). Again, we observed strong colocalization ofthe portion representing the PREP2 N terminus and tubulin.As expected, PREP2CAB-reactive signal did not colocalize withtubulin (Fig. 7F).

FIG. 6. Cytoplasmic forms of PREP2 are sensitive to the nu-clear export inhibitor LMB. NIH3T3 cells were treated with LMBfor 4 h at 0 (A), 5 (B), 25 (C), and 100 nM (D) and scored for nuclearaccumulation of PREP2 using anti-PREP2NAB. E–H, DAPI staining(blue) reveals nuclei. I, the C-terminal deletion mutant PREP2-(1–267)-GFP (construct #2) localizes almost exclusively to the cytoplasm butrelocates to the nucleus upon 150 nM LMB treatment (J). K, the PREP2N terminus (amino acids 1–131) directs GAL-DBD to the cytoplasm asrevealed by rhodamine (red)-conjugated immunoreactivity. EndogenousPREP2CAB reactivity (green) is nuclear, as expected.

FIG. 7. PREP2 colocalizes with the microtubule and actin cy-toskeletons. A, NIH3T3 cells were synchronized at G2/M, released for40 min, and examined in M phase. PREP2 (green) and DAPI (blue)labeling was scored at different mitotic stages as indicated. B, doublelabeling using PREP2NAB (green) and anti-tubulin-specific antibody(red) reveals colocalization (yellow) at the mitotic spindle aster. Themonochromatic image is also shown (lower right). C, colocalization ofGAL-tagged PREP2-(1–131) (green) with microtubules (red) in trans-fected COS-7 cells is shown in the overlay (yellow). D, PREP2NAB(green) and rhodamine-conjugated phalloidin (red) were used to detectPREP2 and F-actin in untransfected COS-7 cells. Overlay demonstratesthat the endogenous PREP2 N terminus colocalizes with actin (yellowand arrows). E, GAL-tagged PREP2-(1–131) (green) colocalizes withF-actin (red) in the periphery of COS-7 cells as shown in the overlay(lower panel, yellow). F, endogenous nuclear PREP2 recognized byPREP2CAB does not colocalize with tubulin. Images C and F arethree-dimensional reconstructions from multiple confocal “z” planes.

Subcellular Localization of PREP249390

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

To verify the association of PREP2 with tubulin, we per-formed coimmunoprecipitation assays on untransfectedNIH3T3 cell extracts. PREP2NAB coprecipitated �-tubulin (55kDa) from untreated cells but not from cells treated with themicrotubule-stabilizing agent, taxol (Fig. 8A, upper panel). In-teraction with PREP2CAB was similar but much weaker (Fig.8A, lower panel). Thus, PREP2 isoforms bearing the N-termi-nal epitope associate with microtubules in a specific mannersensitive to tubulin-interacting drugs.

We then asked whether PREP2 might also associate with theactin cytoskeleton. Fluorescent detection of F-actin with rho-damine-conjugated phalloidin revealed colocalization with the

PREP2NAB epitope (Fig. 7D). Furthermore, overlapping sig-nals could be detected in the cell periphery of GAL4-PREP2-(1–131)-transfected COS-7 cells (Fig. 7E). In addition,PREP2NAB revealed a strong interaction between PREP2 andactin (42 kDa) in coimmunoprecipitation experiments (Fig. 8A).Again, the signal was much weaker with PREP2CAB.

Based on this result, we expected PREP2 to copurify withF-actin in a F/G-actin fractionation assay. Under physiologicalconditions, the F/G-actin ratio in NIH 3T3 cells is roughlyequal. To shift the F/G-actin ratio, we used cytochalasin D, anF-actin-disrupting agent. PREP2 was detected in the F-actin-containing pellet by PREP2NAB (Fig. 8B) and was reduced by

FIG. 8. PREP2 is associated with F- and G-actin. A, coimmunoprecipitation assay of NIH3T3 cell extracts using PREP2NAB (top panel)versus PREP2CAB (bottom panel). AB, antibody. Proteins were separated by 10% SDS-PAGE and immunoblotted simultaneously for actin (42 kDa,black arrowhead) and �-tubulin (55 kDa, gray arrowhead) with specific antibodies. Lane 1, input; lane 2, beads only; lane 3, anti-PREP2NAB, lane4, anti-PREP2NAB on taxol-treated cells. Note the absence of the tubulin signal (lane 4) when cells were treated with the microtubule-stabilizingdrug, taxol. IP, immunoprecipitation; IB, immunoblotting. B, fractionation of cellular extracts from NIH3T3 cells into 100,000-g supernatant (S)and pellet (P) fractions reveals association of PREP2 with F-actin. Cells were treated with the following drugs: lanes 1 and 3, 10 �M F-actin-depolymerizing drug cytochalasin D added to lysis buffer; lanes 2 and 4, untreated. Equal amounts were separated by 10% SDS-PAGE, andproteins were detected by Western blotting. Upper panel, PREP2 full-length and 25-kDa isoforms sediment with the F-actin-containing pellet(lanes 4). Lower panel, note the shift in G- to F-actin ratios following drug treatment (lanes 1 versus 3). WB, Western blotting. C, NIH3T3 cells wereeither transfected with the FLAG-tagged actin mutant (G13R) or assayed for endogenous actin. Cells were lysed in actin stabilization buffer withand without the F-actin stabilization drug phalloidin and fractionated as in B. WT, wild type. Equal volumes of the fractions were separated by10% SDS-PAGE and actin detected by Western blotting. Under physiological conditions, the distribution of F- versus G-actin was approximatelyequal (top left panel), whereas phalloidin shifted the balance toward F-actin (top right panel). G13R does not contribute to the F-actin pool (bottomleft panel). Even under F-actin-stabilizing conditions, the signal in the supernatant (G-actin) remains stronger (bottom right panel, compare withupper right panel). D, immunoprecipitation assays show the association of PREP2 with G-actin. FLAG-tagged actin G13R-expressing cell extractswere precipitated with PREP2NAB (upper panel) or PREP2CAB (lower panel), blotted, and probed with anti-FLAG antibody. Lane 1, input; lane2, immunoprecipitation; lane 3, control. E, precipitation of FLAG-tagged actin G13R-expressing cell extracts with M2-anti-FLAG-agarose beadsand detection of PREP2 confirming the association of PREP2 and G-actin. Cyto D, cytochalasin D.

Subcellular Localization of PREP2 49391

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

treatment with cytochalasin D (compare lanes 3 and 4). To-gether, these results support an interaction between PREP2and F-actin but do not exclude association with G-actin.

To test whether PREP2 also interacts with G-actin, we madeuse of a non-polymerizing �-actin mutant (G13R) that does notcontribute to the cellular F-actin pool (Fig. 8C) (32). Extracts ofNIH3T3 cells expressing FLAG-tagged actin G13R were pre-cipitated with PREP2NAB and PREP2CAB and Western blot-ted against anti-FLAG antibody (Fig. 8D). Although coprecipi-tation was observed for both PREP2 immunoreactive species, amuch stronger interaction with the actin mutant was revealedwith PREP2NAB (Fig. 8D, lane 2). These results were con-firmed by the reverse experiment in which immunoprecipita-tion was performed with anti-FLAG beads followed by detec-tion with the anti-PREP2 antibodies (Fig. 8E and datanot shown).

The embryonic limb bud displays tight regulation of TALEclass homeoproteins at a number of levels. We have previouslyshown that cytoplasmically restricted PBX is colocalized withassemblies bearing non-muscle myosin heavy chain II B in thedistal mouse limb bud. Moreover, we have shown that PREP2species detected with PREP2NAB are also found in the cyto-plasm of limb bud cells (29). Therefore, we examined the dis-tribution of actin, tubulin, and PREP2NAB-reactive species inthis structure. PREP2 isoforms recognized by PREP2NABstrongly colocalize with the microtubule (Fig. 9, A–D) and actin(Fig. 9, E–H) cytoskeletons in this embryonic tissue. Takentogether, these findings indicate that the N-terminally distinctisoforms of PREP2 are associated with microtubules, tubulin,and polymerized actin (F-actin) as well as G-actin.

Disruption of the Actin or Microtubule Cytoskeletons Pro-vokes PREP2 Relocalization—The actin and microtubule cy-toskeletons control gene expression by influencing the nucleo-cytoplasmic transport of transcriptional regulators (22, 34–37).Therefore, we asked whether perturbation of these structures

might trigger a change in PREP2 subcellular localization. In-deed, a 2-h treatment with 10 �g/ml nocodazole, a drug thatdisrupts microtubules by binding to �-tubulin and preventingthe formation of interchain-disulfide bridges, relocated the N-terminal epitope of PREP2 to the nucleus of COS-7 cells (Fig.10, A–D). Under these conditions, the PREP2 N-terminalepitope was also observed in nucleoli (data not shown), con-trary to the nuclear staining pattern observed withPREP2CAB. Similarly, cytochalasin D, a drug that disruptsactin microfilaments, provoked a dose-dependent relocalizationof PREP2NAB reactivity. A 2-h exposure to 10 �M cytochalasinD redistributed PREP2 within the cytoplasm (Fig. 10, E–G),whereas 1 h at twice this concentration relocated the protein tothe nucleus (Fig. 10, H–J). It has been reported that actinomy-cin D can block nuclear export (38). To confirm that relocaliza-tion of PREP2 to the nucleus was due to the loss of actincytoskeleton integrity, we also treated cells with swinholide A,an F-actin-severing drug that does not affect nuclear export(38). A 2-h exposure to 50 nM swinholide A also induced effi-cient nuclear accumulation of endogenous PREP2NAB-reactivespecies as well as an ectopically expressed fusion of PREP2residues 1–267 linked to GFP (data not shown). Together, thesedata demonstrate that both the actin and microtubule cytoskel-etons coregulate the nuclear availability of PREP2 throughinteraction with distinct PREP2 isoforms.

DISCUSSION

A fundamental means to regulate transcription is throughthe control of nuclear access. In addition to the expected role ofPBX in controlling PREP2 subcellular localization, here weshow that CRM-1-dependent nuclear export and associationwith the actin and microtubule cytoskeletons also impinge on

FIG. 9. PREP2 colocalizes with actin and tubulin in the mouseembryo. Double staining with PREP2NAB (A) and anti-tubulin-spe-cific antibodies (B) of frontal limb-bud sections of 10.5-dpc embryosdemonstrates colocalization in the distal anterior region (C). Highermagnification reveals a clear overlap (examples given by arrows, D).Double labeling with anti-PREP2NAB (E) and anti-actin (F) revealscolocalization in the apical ectodermal ridge (AER) of transverse limb-bud sections (G). H, higher magnification of the area boxed in G.

FIG. 10. Disrupting the actin or microtubule cytoskeleton re-locates PREP2. COS-7 cells were treated with the microtubule-desta-bilizing drug, nocodazole (33 �M or 10 �g/ml for 2 h) and scored forlocalization of PREP2 using PREP2NAB (green) and anti-�-tubulin(red). Nuclear relocalization of PREP2NAB reactivity was clearly visi-ble following treatment with nocodazole (A). The microtubule cytoskel-eton was depolymerized and diffusely distributed (B), whereas tubulinremained in newly formed plasma membrane processes (arrows) thatoverlapped with residual cytoplasmic PREP2NAB reactivity (C). Anintact microtubule cytoskeleton is shown in untreated cells (D). Treat-ment with the F-actin-disrupting drug cytochalasin D (10 �M for 2 h) ledto a redistribution of PREP2 (E). Doubling the drug concentration (1 h)relocated the majority of PREP2 (H) and actin (I) immunoreactivity tothe nucleus. Actin was visualized with an anti-actin antibody (F and I).G, overlap of E and F. J, nuclei in H and I stained with DAPI.

Subcellular Localization of PREP249392

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

this process. Throughout this study, we used two antibodiesraised against N-and C-terminal epitopes of PREP2. We havevalidated the specificity of our antibodies in several ways in-cluding the use of Western blotting of in vitro and in vivotranslated proteins, immunoprecipitation, preadsorption, andpreimmune serum. Perhaps most tellingly for the specificity ofPREP2NAB, the material reactive with this antibody is copre-cipitated with PBX in immunoprecipitation assays (data notshown), is strongly translocated to the nucleus upon overex-pression of PBX1A, a known PREP2 partner (Fig. 4E), and isdecreased following siRNA knockdown of Prep2 message(Fig. 3D).

A key finding from immunofluorescence studies is thatPREP2NAB reactivity is largely restricted to the cytoplasm,whereas that of PREP2CAB is nuclear. Additionally, up to fivedistinct PREP2 isoforms are distributed between cytoplasmicand nuclear fractions in Western blots. These findings raisetwo major questions. First, how are multiple PREP2 isoformsproduced? Second, what are the mechanisms governing thesubcellular distribution of PREP2 isoforms?

Processes that could generate PREP2 isoforms of differingmobility in SDS-PAGE include alternative start codon usage,alternative splicing, and post-translational modification,whereas we have shown that proteolytic cleavage is less likely.Although the start codon at position 2 of the published Prep2sequence appears optimal, there are other AUGs both up-stream and downstream. The use of these starts could add orremove 2.5 or 1.8 kDa, respectively, and may well account forsome of the isoforms we detect. Importantly, Northern analysesusing additional probes located in intronic and 3� sequenceshave detected multiple Prep2 transcripts in mature poly(A)�

RNA in addition to those noted previously. Clearly, these ad-ditional products have the potential to encode the differentPREP2 isoforms detected with our antibodies. Following theobservation of alternatively spliced intron-4-containing Prep2cDNAs in the public data base, we used a specific probe toconfirm that intron 4 is indeed facultative and present in a1.4-kb transcript in poly(A)� RNA of adult liver and P19 cells.One such intron-4-containing transcript would in fact encode a25-kDa PREP2 isoform recognized by PREP2NAB but notPREP2CAB, just as we observe for the 25-kDa isoform in ourWestern analyses. Together, these observations strongly sug-gest that at least some PREP2 isoforms and specifically the25-kDa variant are the result of alternative splicing.

Additionally, post-translational modification or processingcould change the apparent molecular mass while masking orremoving the N- or C-terminal epitopes. For example, phospho-rylation can alter protein conformation and mobility in SDS-PAGE. Moreover access to a given epitope could be directly orindirectly blocked. Lastly, proteolytic cleavage could generateisoforms of differing molecular masses that contain only one orthe other (or neither) of the epitopes studied here. For example,the 25-kDa N-terminal and 47-kDa C-terminal peptides de-tected in our study could arise by cleavage of the 62-kDa“full-length” protein. However, our experiments with N- andC-terminally tagged PREP2 suggest that this is not the case.

PREP2NAB detects a 25-kDa PREP2 isoform in nuclearextracts despite the fact that immunofluorescent signals de-tected with this antibody are predominantly cytoplasmic. Be-cause the extraction protocol (buffers and temperatures) usedto obtain cytoplasmic and nuclear fractions is known to disruptthe cytoskeleton and given our demonstration that cytoskeletaldisruption triggers nuclear translocation, the procedure mayprovoke rapid relocalization of normally cytoplasmic 25-kDaPREP2 to the nucleus. Moreover, protein-protein interactionsmasking N- and C-terminal epitopes may be relatively stable in

immunofluorescence but not in denaturing SDS-PAGE,thereby minimizing nuclear immunofluorescent detection.Lastly, the staining detected with PREP2NAB is not exclu-sively cytoplasmic (Figs. 2G and 9B). Given that immunofluo-rescent signal strength does not necessarily correlate with thatobtained by Western analysis, it may be that the weak nuclearreactivity of PREP2NAB in immunofluorescence is indeed dueto the 25-kDa isoform. Nonetheless, the expression of a tagged25-kDa product shows marked cytoplasmic accumulation, sug-gesting that the endogenous 25-kDa isoform does contribute toPREP2NAB immunoreactivity outside the nucleus.

Our observation of a 25-kDa PREP2 isoform residing largelyin the cytoplasm but capable of accumulating in the nucleussuggests a role in transcriptional regulation. Whereas the 25-kDa PREP2 isoform lacks the homeodomain, it retains all ofHM1 and HM2. This region of the MEIS/PREP family is knownto be sufficient for interaction with PBX. Thus, the N terminusof MEIS family members binds to PBX family members in vitroand alters PBX function in vivo (39, 40). Moreover, the deletionof the PREP1 homeodomain actually augments associationwith PBX (8). Thus, the 25-kDa PREP2 isoform could reason-ably function as a dominant-negative capable of preventingfull-length forms of PREP and MEIS family members frominteracting with PBX family peptides. Given the importance ofDNA binding by MEIS/PREP family members for the functionof HOX-PBX-MEIS/PREP trimers (3) or PBX-MEIS/PREPdimers (5), significant expression of the 25-kDa PREP2 isoformcould significantly modulate target gene regulation. This doesnot exclude possible novel functions conferred on the 25-kDapeptide by the unique 50 residues at its C terminus. However,such a function may not be evolutionarily conserved because asimilar splicing event in human PREP2 (in which intron 4 isretained in the mature transcript) would encode a C terminusonly poorly conserved with its murine counterpart.

The masking and unmasking of NESs and NLSs throughintermolecular and intramolecular interactions have beenshown to play a crucial role in the localization of PBX andextradenticle (18–20). In this study, we provide evidence thatthe N-terminal portion of PREP2 harbors either a NES or aregion that associates with NES-containing protein(s). In thepresence of the nuclear export inhibitor LMB, PREP2 accumu-lates in the nucleus of diverse cell types including NIH3T3,COS-7, and S2. This is also true for the C-terminally truncatedversion PREP2-(1–267)-GFP, although higher LMB concentra-tions were necessary to achieve nuclear staining. This mayhave been due to the absence of two putative NLSs in theC-terminal portion of PREP2. One of these NLSs was proposedto lie between residues 348 and 354 (PKAKKIK) (27), and wetentatively assigned the other NLS to residues 274–280(KKSKNKR). Substitution of the two presumptive NLSs in thePREP2 C terminus with the GAL4 DBD resulted in cytoplas-mic staining only. Therefore, PREP2 N-terminal signals direct-ing cytoplasmic localization dominate over NLSs in GAL-DBDand in larger PREP2 isoforms.

We showed intimate association between PREP2NAB-reac-tive material and the actin and microtubule cytoskeleton byboth immunofluorescent colocalization and coimmunoprecipi-tation. Additionally, we demonstrated cytochalasin D-sensitivecosedimentation of PREP2 with F-actin. Overall, our data showthe association between PREP2 with F- and G-actin and withmicrotubules/tubulin. A functional role for these associations isimplied by our observation that drug-induced perturbation ofthe actin or microtubule cytoskeletons results in nuclear accu-mulation of PREP2NAB-reactive species.

How might association with cytoskeletal elements controlPREP2 distribution in the cell? First, direct association with

Subcellular Localization of PREP2 49393

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

G-actin or tubulin in the nucleus could facilitate exposure ofone or more NESs in the PREP2 N terminus. This would beconsistent with our demonstration that PREP2 interacts withG-actin and with the fact that G-actin can be found in thenucleus (41). Alternatively, PREP2 may not encode an NES butcould be exported from the nucleus because of its associationwith G-actin, which does have NES (41). Second and non-exclusively, direct binding to F-actin or microtubules couldanchor PREP2 in the cytoplasm. We note that the action ofnocodazole and cytochalasin D does not allow easy discrimina-tion between these possibilities, because their effect may notbe attributed to depolymerization of cytoskeletal elements butto the masking of interaction sites on G-actin or free tubulin (37).Nonetheless, interaction with the filamentous forms of actin andtubulin strongly supports a cytoplasmic anchoring function.

Association of PREP2 with cytoskeletal structures could alsobe indirect via one of the many peptides that contact thesepolymers. Some proteins, such as bullous pemphigoid antigen,have binding sites for both F-actin and microtubules (42). Link-age to such a protein could explain our observation that dis-ruption of either cytoskeletal system is sufficient to direct themajority of PREP2 to the nucleus. This implies a mechanisminvolving cross-talk between the two cytoskeletal systems,which could be provided by a cross-linking protein, and isconsistent with extensive cooperation between actin and micro-tubules in numerous cellular processes (43).

A number of studies have established the importance ofcytoskeletal interactions in regulating nuclear access to tran-scription factors. Thus, Myc-interacting zinc finger 1, a regu-lator of the low density lipoprotein receptor gene, is seques-tered in the cytoplasm by association to microtubules but isdirected to the nucleus upon treatment with microtubule-depolymerizing agents (36). In a variation on this theme, bind-ing to tubulin inhibits TGF-�-mediated phosphorylation andperhaps nuclear translocation of Smad2, Smad3, and Smad4(34). On the other hand, actin filaments retain the Nrf2 (NF-E2-related factor 2) in the cytoplasm until filament disruptionprovoked by phosphatidylinositol 3-kinase signaling sends aNrf2-actin complex to the nucleus (44). Conversely, associationof the MAL coactivator with G-actin prevents nuclear translo-cation in a manner reversible by Rho-induced F-actin forma-tion (37). In the case of the glucocorticoid receptor, associationwith microtubules, including the mitotic spindle, may allowdirect control of cell growth by glucocorticoids (35). These ex-amples highlight the importance of the cytoskeleton in relayingcell-signaling cues to a variety of transcription factors and raisethe possibility that multiple signaling pathways may impingeon interactions with the cytoskeleton to control PREP2 func-tion. With regard to the ubiquitous embryonic expression pro-file of PREP2 protein (29), we speculate that PREP2 may bepreferentially regulated at the protein level. This could havethe advantage of making certain isoforms, perhaps with dis-tinct regulatory potential, rapidly functional, both because theprotein is already present and because it could be directedquickly to the nucleus by the cytoskeletal network.

In summary, our results, along with reports of cytoplasmi-cally localized PBX, MEIS, and HOX family members (24,45–47), suggest that the control of subcellular distribution isbroadly employed to regulate homeoprotein function.

Acknowledgments—We thank R. Treisman and G. Posern for theG13R actin mutant, K. Yoshida for LMB, U. Stochaj, W. Mushynski,Feng Gu, and A. Nepveu for helpful discussions.

REFERENCES

1. Krumlauf, R. (1994) Cell 78, 191–2012. Featherstone, M. (2003) in Murine Homeobox Gene Control of Embyronic

Patterning and Organogenesis. (Lufkin, T., ed) pp. 1–42, Elsevier SciencePublishers B.V., Amsterdam

3. Jacobs, Y., Schnabel, C. A., and Cleary, M. L. (1999) Mol. Cell. Biol. 19,5134–5142

4. Gebelein, B., Culi, J., Ryoo, H. D., Zhang, W., and Mann, R. S. (2002) Dev. Cell3, 487–498

5. Ferretti, E., Marshall, H., Popperl, H., Maconochie, M., Krumlauf, R., andBlasi, F. (2000) Development 127, 155–166

6. Shanmugam, K., Green, N. C., Rambaldi, I., Saragovi, H. U., and Feather-stone, M. S. (1999) Mol. Cell. Biol. 19, 7577–7588

7. Swift, G. H., Liu, Y., Rose, S. D., Bischof, L. J., Steelman, S., Buchberg, A. M.,Wright, C. V., and MacDonald, R. J. (1998) Mol. Cell. Biol. 18, 5109–5120

8. Berthelsen, J., Zappavigna, V., Ferretti, E., Mavilio, F., and Blasi, F. (1998)EMBO J. 17, 1434–1445

9. Berthelsen, J., Zappavigna, V., Mavilio, F., and Blasi, F. (1998) EMBO J. 17,1423–1433

10. Herzig, S., Fuzesi, L., and Knepel, W. (2000) J. Biol. Chem. 275, 27989–2799911. Burglin, T. R. (1998) Dev. Genes Evol. 208, 113–11612. Knoepfler, P. S., Calvo, K. R., Chen, H., Antonarakis, S. E., and Kamps, M. P.

(1997) Proc. Natl. Acad. Sci. U. S. A. 94, 14553–1455813. Gonzalez-Crespo, S., Abu-Shaar, M., Torres, M., Martinez, A. C., Mann, R. S.,

and Morata, G. (1998) Nature 394, 196–20014. Rieckhof, G. E., Casares, F., Ryoo, H. D., Abu-Shaar, M., and Mann, R. S.

(1997) Cell 91, 171–18315. Mercader, N., Leonardo, E., Azpiazu, N., Serrano, A., Morata, G., Martinez, C.,

and Torres, M. (1999) Nature 402, 425–42916. Capdevila, J., Tsukui, T., Rodriquez Esteban, C., Zappavigna, V., and Izpisua-

Belmonte, J. C. (1999) Mol. Cell 4, 839–84917. Kilstrup-Nielsen, C., Alessio, M., and Zappavigna, V. (2003) EMBO J. 22,

89–9918. Saleh, M., Huang, H., Green, N. C., and Featherstone, M. S. (2000) Exp. Cell

Res. 260, 105–11519. Berthelsen, J., Kilstrup-Nielsen, C., Blasi, F., Mavilio, F., and Zappavigna, V.

(1999) Genes Dev. 13, 946–95320. Abu-Shaar, M., Ryoo, H. D., and Mann, R. S. (1999) Genes Dev. 13, 935–94521. Choe, S. K., Vlachakis, N., and Sagerstrom, C. G. (2002) Development 129,

585–59522. Huang, H., Paliouras, M., Rambaldi, I., Lasko, P., and Featherstone, M. (2003)

Mol. Cell. Biol. 23, 3636–364523. Vlachakis, N., Choe, S. K., and Sagerstrom, C. G. (2001) Development 128,

1299–131224. Buske, C., Feuring-Buske, M., Abramovich, C., Spiekermann, K., Eaves, C. J.,

Coulombel, L., Sauvageau, G., Hogge, D. E., and Humphries, R. K. (2002)Blood 100, 862–868

25. Chen, H., Rossier, C., Nakamura, Y., Lynn, A., Chakravarti, A., and Antonara-kis, S. E. (1997) Genomics 41, 193–200

26. Ferretti, E., Schulz, H., Talarico, D., Blasi, F., and Berthelsen, J. (1999) Mech.Dev. 83, 53–64

27. Imoto, I., Sonoda, I., Yuki, Y., and Inazawa, J. (2001) Biochem. Biophys. Res.Commun. 287, 270–276

28. Fognani, C., Kilstrup-Nielsen, C., Berthelsen, J., Ferretti, E., Zappavigna, V.,and Blasi, F. (2002) Nucleic Acids Res. 30, 2043–2051

29. Haller, K., Rambaldi, I., Kovacs, E. N., Daniels, E., and Featherstone, M.(2002) Dev. Dyn. 225, 358–364

30. Rambaldi, I., Nagy Kovacs, E., and Featherstone, M. S. (1994) Nucleic AcidsRes. 22, 376–382

31. Rastegar, M., Kobrossy, L., Nagy Kovacs, E., Rambaldi, I., and Featherstone,M. (2004) Mol. Cell. Biol. 24, 8090–8103

32. Posern, G., Sotiropoulos, A., and Treisman, R. (2002) Mol. Biol. Cell 13,4167–4178

33. Chan, C. K., Hubner, S., Hu, W., and Jans, D. A. (1998) Gene Ther. 5,1204–1212

34. Dong, C., Li, Z., Alvarez, R., Jr., Feng, X. H., and Goldschmidt-Clermont, P. J.(2000) Mol. Cell 5, 27–34

35. Akner, G., Wikstrom, A. C., and Gustafsson, J. A. (1995) J. Steroid Biochem.Mol. Biol. 52, 1–16

36. Ziegelbauer, J., Shan, B., Yager, D., Larabell, C., Hoffmann, B., and Tjian, R.(2001) Mol. Cell 8, 339–349

37. Miralles, F., Posern, G., Zaromytidou, A. I., and Treisman, R. (2003) Cell 113,329–342

38. Hofmann, W., Reichart, B., Ewald, A., Muller, E., Schmitt, I., Stauber, R. H.,Lottspeich, F., Jockusch, B. M., Scheer, U., Hauber, J., and Dabauvalle,M. C. (2001) J. Cell Biol. 152, 895–910

39. Ryoo, H. D., Marty, T., Casares, F., Affolter, M., and Mann, R. S. (1999)Development 126, 5137–5148

40. Choe, S. K., and Sagerstrom, C. G. (2004) Dev. Biol. 271, 350–36141. Olave, I. A., Reck-Peterson, S. L., and Crabtree, G. R. (2002) Annu. Rev.

Biochem. 71, 755–78142. Dalpe, G., Leclerc, N., Vallee, A., Messer, A., Mathieu, M., De Repentigny, Y.,

and Kothary, R. (1998) Mol. Cell Neurosci. 10, 243–25743. Rodriguez, O. C., Schaefer, A. W., Mandato, C. A., Forscher, P., Bement, W. M.,

and Waterman-Storer, C. M. (2003) Nat. Cell Biol. 5, 599–60944. Kang, K. W., Lee, S. J., Park, J. W., and Kim, S. G. (2002) Mol. Pharmacol. 62,

1001–101045. Komuves, L. G., Shen, W. F., Kwong, A., Stelnicki, E., Rozenfeld, S., Oda, Y.,

Blink, A., Krishnan, K., Lau, B., Mauro, T., and Largman, C. (2000) Dev.Dyn. 218, 636–647

46. Komuves, L. G., Ma, X. K., Stelnicki, E., Rozenfeld, S., Oda, Y., and Largman,C. (2003) Dev. Dyn. 227, 192–202

47. Immergluck, K., Lawrence, P. A., and Bienz, M. (1990) Cell 62, 261–268

Subcellular Localization of PREP249394

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from

Klaus Haller, Isabel Rambaldi, Eugene Daniels and Mark FeatherstoneTubulin, and Nuclear Export

Subcellular Localization of Multiple PREP2 Isoforms Is Regulated by Actin,

doi: 10.1074/jbc.M406046200 originally published online August 30, 20042004, 279:49384-49394.J. Biol. Chem. 

  10.1074/jbc.M406046200Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/279/47/49384.full.html#ref-list-1

This article cites 46 references, 18 of which can be accessed free at

by guest on March 30, 2020

http://ww

w.jbc.org/

Dow

nloaded from