4558–4565 Nucleic Acids Research, 2000, Vol. 28, No. 22

Nuclear targeting determinants of the far upstreamelement binding protein, a c-myc transcription factorLiusheng He, Achim Weber and David Levens*

Laboratory of Pathology, DCS, NCI, Building 10, Room 2N105, Bethesda, MD 20892-1500, USA

Received June 27, 2000; Revised and Accepted September 29, 2000

ABSTRACT

FUSE binding protein (FBP) binds in vivo and in vitrowith the single-stranded far upstream element(FUSE) upstream of the c-myc gene. In addition to itstranscriptional role, FBP and its closely relatedsiblings FBP2 (KSRP) and FBP3 have been reportedto bind RNA and participate in various steps of RNAprocessing, transport or catabolism. To performthese diverse functions, FBP must traffic to differentnuclear sites. To identify determinants of nuclearlocalization, full-length FBP or fragments thereofwere fused to green fluorescent protein. Fluorescent-FBP localized in the nucleus in three patterns,diffuse, dots and spots. Each pattern was conferredby a distinct nuclear localization signal (NLS): a classicalbipartite NLS in the N-terminal and two non-canonicalsignals, an α-helix in the third KH-motif of the nucleicacid binding domain and a tyrosine-rich motif in theC-terminal transcription activation domain. Upontreatment with the transcription inhibitor actino-mycin D, FBP completely re-localized into dots, butdid not exit from the nucleus. This is in contrast tomany general RNA-binding proteins, which shuttlefrom the nucleus upon treatment with actinomycin D.Furthermore, FBP co-localized with transcriptionsites and with the general transcription factor TFIIH,but not with the splicing factor SC-35. Takentogether, these data reveal complex intranucleartrafficking of FBP and support a transcriptional rolefor this protein.

INTRODUCTION

Increasing evidence indicates that many proteins are bi- oreven multi-functional; sometimes the different roles played bya single molecule are logically connected, while other times theroles appear to be independent. Intrinsic to the reutilization of suchproteins for different intracellular or extracellular processes is thecapacity to deliver the right amount of each component to thecorrect location at the appropriate time (1,2). Even within asubcellular compartment such as the nucleus, a single proteinmay participate in several processes, and hence must besublocalized to the correct place. Proteins larger than 20–40 kDarequire nuclear localization signals (NLS) for import into the

nucleus (3–6). Two canonical types of NLS are most frequentlyemployed to direct the proteins required for replication, transcrip-tion and RNA processing and other nuclear processes to theirappropriate targets. The first ‘core’ NLS is comprised of ahexapeptide including four or more arginine and lysine residues,flanked by acidic residues or the helix-breaking residues,proline and glycine (4,7). The second common NLS is constitutedof two clusters of basic amino acids separated by ∼10 non-basic residues (4,7). Other sorts of NLSs are associated withribosomal proteins and hnRNPs. For example the M9, KNSand HNS sequences target hnRNP A1, hnRNP K and HUR,respectively, to the nucleus (8–10).

The FUSE binding protein (FBP), named for its interactionwith the far upstream element (FUSE) found upstream of the c-mycgene, has been dissected into three functional domains (11,12).Transactivation is conferred by the C-terminus of FBP (FBPC)through a tyrosine-activating sequence repeated three times. Incontrast, the N-terminus (FBPN) represses transcription in cisand in trans. These effector domains of FBP interact directlywith components of the transcription machinery in vivo andin vitro (13). The central DNA binding domain (FBPCD) iscomposed of four KH-motifs, each followed by an amphipathichelix. Replication defective adenoviruses expressing thecentral domain of FBP interfere directly with c-myc expressionvia the FUSE; similar viruses expressing antisense FBP also shutoff c-myc transcription (14). Beyond a role in transcription, somereports suggest that FBP or its close relatives FBP2 (KSRP)and FBP3 may interact with certain RNA molecules andparticipate at various steps in transcription, or in RNAprocessing, transport or catabolism, in the nucleus or in thecytoplasm (15–18). If truly multi-functional, then FBP musttraffic to assorted sites associated with different processes. Toglean more insight into the potential for FBP to play multipleroles, a series of experiments were conducted to study thelocalization and targeting of FBP within intact cells.

MATERIALS AND METHODS

DNA constructs

An FBP cDNA containing the entire 644 amino acid openreading frame cloned into the mammalian expression vectorpcDNA1 was digested by HindIII and then ligated intopEGFP-C2 (Clontech). Similarly, four deletion mutants ofFBP were excised from pcDNA1FBPN, FBPcd, FBPC andFBP∆cd with HindIII and cloned into the GFP vector. Otherprogressive deletions for mapping NLSs were generated by

*To whom correspondence should be addressed. Tel: +1 301 496 2176; Fax: +1 301 402 0043; Email: levens@helix.nih.gov

Nucleic Acids Research, 2000, Vol. 28, No. 22 4559

digesting DNA fragments with appropriate restriction endo-nucleases. The point mutations in tyrosine motif YM3 weregenerated from their pcDNA1 parent (12).

Cells culture, transfections and fluorescence microscopy

HeLa cells were grown on cover slides overnight inDulbecco’s modified minimal essential medium (DMEM)containing 10% fetal calf serum. Transient transfections wereperformed in HeLa cells using LipoFectamine plus (Life Tech-nology) with 2 µg expression plasmids as described previously(19). Sixteen hours after transfection, cells were fixed in 4%paraformaldehyde and washed twice in PBS. The slides weremounted and stored at 4°C over several months. Green fluores-cence images from green fluorescence protein (GFP) werephotographed by Zeiss fluorescence microscopy.

Cell extract and western blotting

Cells were lysed with RIPA buffer (1% v/v Nonidet P-40, 1%w/v sodium deoxycholate, 0.1% w/v SDS, 0.15 M NaCl, 2 mMEDTA, 0.01 M sodium phosphate pH 7.2, 50 mM sodiumfluoride, 0.2 mM sodium vanadate, 100 U/ml aprotinin) andincubated for 30 min on ice. Debris was removed by centrifu-gation (full speed in an Eppendorf 5415C microcentrifuge for30 min at 13 000 r.p.m.) and the supernatant was analyzed byimmunoblot with anti-GFP. The protein amounts of eachsample were quantitated by both the Bio-Rad protein assay andthe Coomassie Brilliant Blue staining of gels. Equal amountsof protein were separated by SDS–PAGE using 4–12%gradient gels (NuPAGE, Novex) and transferred onto nitro-cellulose membranes. Membranes were incubated overnight at4°C with 1:5000 primary monoclonal anti-GFP (Clontech),washed three times for 5 min with Tris-buffered salinecontaining 0.3% Tween 20 and incubated with 1:2500 dilutedhorseradish conjugated secondary antibodies for 1 h at roomtemperature. Membranes were washed as above and horseradishperoxidase was detected using enhanced chemiluminescencereagent (1.25 mM 3-aminophthalhydrazide, 0.2 mM coumaricacid, 0.3 mM hydrogen peroxide in 0.1 M Tris pH 8.5).Membranes were then exposed on Kodak X-Omat AR X-rayfilm (Eastman Kodak).

Actinomycin D treatment

HeLa cells were transfected as described (20). Sixteen hours aftertransfection, the cells were washed twice in PBS and re-incubatedwith pre-warmed DMEM with 5 µg/ml actinomycin D (Calbio-chem) for 4–6 h prior to fixation for fluorescence photography.

Labeling of transcription sites and immunofluorescenceprocedures

Nascent RNA transcription sites were labeled based onprevious procedures (21,22). Cells cultured on poly-L-lysine-coated glass-bottom chambers were used for immunostaining.All procedures were done at room temperature. The cells weretransfected with the GFPFBP construct and cultured overnight.Nascent RNA transcription sites were labeled according topublished procedures. Briefly, cells were washed with ice-coldTBS buffer (10 mM Tris–HCl pH 7.4, 150 mM NaCl, 5 mMMgCl2) and further washed with glycerol buffer (20 mM Tris–HClpH 7.4, 25% glycerol, 5 mM MgCl2, 0.5 mM EGTA, 0.5 mMPMSF) for 10 min on ice. Cells on ice were permeabilized with0.025% Triton X-100 in glycerol buffer (with 25 U/ml of

RNasin; Promega) for 3 min and immediately incubated atroom temperature for 30 min with nucleic acid synthesis buffer(50 mM Tris–HCl pH 7.4, 10 mM MgCl2, 150 mM NaCl, 25%glycerol, 0.5 mM PMSF, 25 U/ml of RNasin, 1.8 mM ATP)supplemented with 0.5 mM CTP, GTP and BrUTP (SigmaChemicals). After nucleotide incorporation, the cells werefixed with 4% freshly made paraformaldehyde in PBS on icefor 5 min, washed with ice-cold TBS-Tween buffer (10 mMTris–HCl pH 7.4, 150 mM NaCl, 0.2 mM MgCl2, 0.2% Tween20), blocked with 5% goat serum and incubated with mouseanti-BrU (IgG, Roche) followed by Rhodamine-conjugatedgoat anti-mouse IgG (1:150; Roche) to detect transcriptionsites. All incubations were performed at room temperature.

Immunostaining

HeLa cells transfected with GFP–FBP plasmids were fixedwith 4% paraformaldehyde in PBS for up to 20 min at roomtemperature and then washed twice with cold PBS. Immuno-cytochemistry was performed as described (23). The cells werethen permeabilized by treatment with 80 mM HEPES pH 6.8,5 mM MgCl2 and 0.5% Triton X-100, for 5 min at roomtemperature. The blocking solution (PBS with 0.5% Triton X-100,2% BSA) was added to the cells at 37°C for 30 min. Theprimary antibody (rabbit polyclonal anti-human p89, 1:00,Santa Cruz Biotechnology, Santa Cruz; mouse monoclonalanti-human SC-35, Sigma, St Louis) was added at a 1:1000dilution to the blocking buffer, incubated for 1 h at 37°C andthen washed twice in cold PBS. The secondary antibodyconjugated with Rhodamine (Roche) was added directly to thecover slip and covered with a plastic membrane for 1 h at 37°C.After immunostaining, cells were stained for 20 min withdiamidinophenylindole (DAPI, 1 µg/ml) and washed twicewith PBS. Finally the slides were mounted and stored at –20°Cfor microscopic examination. Fluorescent images of the cells wereobtained using a fluorescent microscope (Carl Zeiss).

RESULTS

Each of FBP’s three domains confers nuclear localization

Biochemically, FBP is enriched in the nuclear extracts. Immuno-staining with affinity purified, polyclonal anti-FBP revealedlocalization of FBP in hundreds of dots throughout the nucleo-plasm, against a background of homogeneous, diffuse nuclearstaining (Fig. 1A). FBP staining of nucleoli and the cytoplasmwas at the margin of detectability. To study the intracellulartrafficking of FBP, a means was sought to eliminate potentialinterference from the immunologically cross-reacting FBP2and FBP3. Although previous studies have demonstratednuclear staining with affinity-purified anti-FBP, the relativecontributions from each of the FBPs could not be directlyassessed. To monitor uniquely the trafficking of FBP, a plasmidexpressing a GFP–FBP fusion (GFPFBP) was transfected intoHeLa cells and the sub-cellular localization of fluorescencewas followed directly (24). GFPFBP fluorescence was tightlylocalized to the nucleus (Fig. 1A and B, panels 3 and 1–3,respectively). The pattern of GFPFBP fluorescence was essen-tially indistinguishable from the pattern of endogenous FBPimmunostaining (Fig. 1A, compare panel 1 with panel 3). Incontrast, GFP alone was distributed uniformly throughout the cell(data not shown). First, 80% of the cells showed a homogeneous

4560 Nucleic Acids Research, 2000, Vol. 28, No. 22

distribution of FBP throughout the entire nucleoplasm(excluding the nucleolus) (Fig. 1A and B, panels 1 and 3 andpanel 1, respectively; summarized in Fig. 1C). In other cellsfocal accumulations of FBP occurred. In ∼20% of the cells

FBP was seen in hundreds of tiny, brightly fluorescent dots[Fig. 1B, panels 1 (arrow) and 2; summarized in Fig. 1C].Third, in the nuclei of nearly 3% of the transfected cells, 20–40coarsely fluorescent spots were observed (Fig. 1B, panel 4;summarized in Fig. 1C). Whether the dots and spots definedistinct nuclear compartments is unknown; however, forheuristic purposes this classification is included. The intra-cellular distribution of FBP and GFPFBP were unperturbed bytreatment with leptomycin B, an inhibitor of nuclear export(data not shown). So nuclear uptake of FBP is not coupled withdelivery of nuclear cargoes to the cytoplasm. [Although thesestudies were conducted with HeLa cells, full-length FBP hasbeen observed to be nuclear in several other cell-lines trans-fected with GFPFBP and anti-FBP stained only nuclei in avariety of normal and neoplastic tissue (L.He, A.Weber andD.Levens, unpublished observations)].

In order to determine which sequences in FBP localize it tothe nucleus, each of its functional domains, FBPN, FBPCDand FBPC as well as FBP∆CD lacking the DNA-bindingcentral domain (CD), were fused with GFP by cloning intopEGFP and transfected into HeLa cells. Surprisingly each ofthese non-overlapping protein fragments was efficiently directedinto the nucleus. Both the N-terminal repression domain (Fig. 1B,panel 5; summarized in Fig. 1C) and the C-terminal activationdomain (Fig. 1B, panel 7; summarized in Fig. 1C) were almostentirely localized to the nucleus as was the protein deleted ofthe DNA-binding domain (Fig. 1B, panel 4; summarized inFig. 1C). The nucleic acid binding central domain, FBPCD,was uniformly distributed in the nucleoplasm, excepting thenucleolus, but was also found in coarse agglomerations in thecytoplasm (Fig. 1B, panel 6; summarized in Fig. 1C). Thus,each domain of FBP possessed a nuclear import signal capableof functioning autonomously. Immunoblot analysis confirmedthe sizes and levels of expression of the chimeric proteins

Figure 1. FBP possesses multiple, non-overlapping NLSs. (A) The intranuclearpattern of the GFPFBP chimera parallels endogenous FBP. Panel 1, fluorescencemicrographs of endogenous FBP stained with the FBP-specific antibody andFITC-second antibody revealing homogenous and dotted nuclear patterns inHeLa cells. A cell with a pattern of dots is marked with an arrowhead. Panel 2, thesame cells were also stained with DAPI to illustrate that FBP staining is confinedto the nucleus. Panel 3, HeLa cells were seeded on the coverslip in 6-well platesand grown overnight to 75% confluency and then transfected with theGFPFBP-plasmid. An enlarged GFPFBP-expressing cell is shown in the insetto visualize better the fine dots and to facilitate comparison with panel 1. Panel 4,FBP was immunostained with Rhodamin conjugated secondary antibody.Panel 5, merging the green and red reveals co-trafficking of transfected andendogenous FBP. (B) Wild-type FBP fused with GFP has several phenotypes,homogenous (panel 1) and focal accumulations ranging from fine dots (panel 2)to coarser spots (panel 3) while each subdomain FBPN (panel 5), FBPCD (panel 6),FBPC (panel 7) has its own NLS conferring distinct intranuclear patterns. GFPalone is uniformly distributed within the cell (data not shown). (C) The differentpatterns shown in (B) were quantitated. Dots and spots were counted separately incase this assignation distinguishes between nuclear subcompartments. Stripedboxes, solid boxes and ellipses indicate α-helix, KH-repeats and tyrosine-repeats,respectively. Hom, homogenous; speck, spots; N, nuclear; C, cytoplasmic; N + C,dispersed throughout the cell. For wild-type GFPFBP, 500 cells were countedand classified into three different nuclear phenotypes, 81.4% were homogeneouslydistributed, 16.8% dotted and 1.8% were in spots. Both FBPN and FBPC weretotally in nuclei. But only two-thirds of GFP-FBPCD distributed into nucleiwhile 28.8% formed a perinuclear ring structure surrounding the nuclear membrane.(D) GFP immunoblot was performed to verify and quantitate GFP fusion proteinexpression. Arrowheads indicate GFP chimera. An arrow marks a non-specificband (NS).

Nucleic Acids Research, 2000, Vol. 28, No. 22 4561

(Fig. 1D). Therefore, the capacity or lack thereof for nucleartrafficking was an intrinsic property of the engineered fusion-proteins and not due to spurious truncation.

FBPN contains a classical bipartite NLS

Inspection of the FBP amino acid sequence revealed a candidatebipartite NLS within its N-terminal repression domain (25). Totest whether this segment of FBP was responsible for nuclearimport, GFP–FBPN derivatives were prepared. Successivedeletions of FBPN from residue 106 toward the N-terminuswere examined. Deletions until residue 92 were fully localizedto the nucleus, whereas deletions to or beyond residue 53 werecompletely impaired for this activity (Fig. 2A, panels 1–4;summarized in Fig. 2B). The candidate NLS was indeedlocated between residues 53 and 92. To prove that this bipartiteNLS-like sequence was sufficient to target nuclear transport,

residues 63–78 comprising only this region were fused to GFPand were found to be sufficient to drive GFP into the nucleus(summarized in Fig. 2B). This classical bi-partite NLS isconserved in FBP2 (KSRP), but is lacking in FBP3 (Fig. 2D).The expression levels and sizes of the fusion proteins wereconfirmed using SDS–PAGE and immunoblot (Fig. 2C).

The FBP central domain contains an atypical NLS

The nucleic acid binding central domain of FBP supportedseveral patterns of fluorescence when fused with GFP. Almosttwo-thirds of the GFP–FBPCD expressing cells displayedhomogeneous nuclear staining with relative sparing of thenucleolus (Fig. 3A, panel 1). Almost one-third of the cellsdeveloped intensely fluorescent coarsely beaded accumulations ofGFP–FBPCD forming a peri-nuclear cuff (Fig. 3A, panel 1).The remaining cells were diffusely fluorescent eitherthroughout or just in the cytoplasm (summarized in Fig. 3B).

GFP–FBPCD was less efficiently targeted to the nucleusthan GFP fusions with full-length FBP or either effectordomain. Conceivably, the nuclear localization of FBPCDmight be due to diffusion of FBP throughout the cell withnuclear retention due to entrapment by nucleic acids. If occurring,such passive retention would require functional DNA or RNAbinding by GFP–FBPCD. A series of C-terminal truncations ofGFP–FBPCD (residue 480) were expressed in HeLa cells todetermine which FBPCD sequences allowed nuclear retention.Deletion through KH-repeat 4 (residue 416) did not impairnuclear localization (Fig. 3A, panel 2; summarized in Fig. 3B).Deletion from the C-terminus of FBPCD to residue 381 stillallowed nuclear import of FBPCD, but removal of just sevenmore residues abrogated nuclear uptake (Fig. 3A, panels 3 and4; summarized in Fig. 3B). N-terminal truncations of FBPCD(from residue 107 to 274) revealed that KH-repeats 1 and 2were dispensable for nuclear localization (Fig. 3A, panel 6;summarized in Fig. 3B). N-terminal deletions, which extendedthrough KH-repeat 3 preserved nuclear targeting untilencroaching upon α-helix 4 (Fig. 3A, panel 5; summarized inFig. 3B). Thus delimited by N- and C-terminal deletions,residues 366–386 encompassing α-helix 4 were tested andfound to confer the same pattern of nuclear targeting seen withthe entire central domain (Fig. 3A, panel 7; summarized inFig. 3B). Because this same protein fragment lacked nucleicacid binding activity, the nuclear accumulation of FBPCD wasnot due to passive ensnarement of the protein by nuclearnucleic acids. Mutation of isoleucines 378 and 379, conservedamong all the FBPs (Fig. 3D), to helix-destabilizing prolinescompletely abolished the NLS activity of α-helix 4 (Fig. 3A,panel 8; summarized in Fig. 3B). Thus, an NLS is completelyembedded within a structural element of KH-repeat 3. Similar,but not identical α-helices occur in FBP2 and FBP3. Immuno-blot analysis verified the synthesis of GFP fusion proteins(Fig. 3C). Hence, FBPCD, like FBPN, contains a shortsegment directing import into the nucleus.

The tyrosine-rich transcription activation motifs of FBPCare required for its nuclear localization

To identify the sequence elements that direct FBPC to thenucleus, a series of C-terminal deletions between residues 449and 644 of FBP were fused with GFP and each was tested fornuclear localization. Deletion to residue 576 resulted in a 95%loss of NLS activity (Fig. 4A, panel 2; summarized in Fig. 4B).

Figure 2. FBPN contains a functional bipartite NLS. (A) Segments of FBPNwere fused to GFP, transfected and assessed for nuclear localization; panel 1,residues 1–92 fused to GFP; panel 2, residues 1–53 fused to GFP; panel 3, residues1–36 fused to GFP; panel 4, residues NLS 63–78 fused to GFP. (B) Summaryof transfected GFPFBPN derivatives and quantitation of trafficking. Aminoacids 63–78 efficiently traffic into nuclei. (C) GFP immunoblot was performed toverify and quantitate GFP fusion protein expression. Arrowheads indicate GFPchimera. A non-specific band (NS) is marked by an arrow. (D) This bipartiteNLS is conserved in FBP and FBP2, but not FBP3.

4562 Nucleic Acids Research, 2000, Vol. 28, No. 22

The deletion to residue 531 retained just 0.4% NLS activity(Fig. 4A, panel 3; summarized in Fig. 4B) and beyond residue495 NLS activity was completely lost (summarized in Fig. 4B).Deletion to residue 576 removed the last two of the three

tyrosine-rich motifs, YM1, 2 and 3. These motifs individuallyor in combination have proved to be potent transcriptionaltransactivators (12). Only constructs with all three YM motifswere efficiently targeted to the nucleus. Fusion proteinsbearing wild-type YM3 retained some capacity for nuclear local-ization, but the same mutations which disabled transcriptionactivation by YM3 eliminated even this residual activity(Fig. 4A, panels 8–10; summarized in Fig. 4B). YM1 and 2were ineffective at directing nuclear fluorescence when fusedwith GFP (Fig. 4A, panels 6 and 7; summarized in Fig. 4B).Appropriate expression of all protein chimeras was confirmed(Fig. 4C).

Inhibiting transcription prompts an intranuclearredistribution of FBP

The known functional activities resident in FBP (11–14) andthe experiments reported above, support a nuclear role forFBP. But some workers have proposed a cytoplasmic role forFBP in the degradation of GAP-43 mRNA and other KH-motiffamily members are zip-code binding proteins which help todeliver particular mRNAs to appropriate intra-cytoplasmicsites (18,26). Furthermore the prototype for the KH family, whichincludes FBP, is hnRNP K, itself a bi- or multi-functional proteinshuttling between the nucleus and cytoplasm and regulatingprocesses in both compartments (9,27–30). Inhibition oftranscription often reduces the nuclear pool of hnRNA andresults in the failure of nuclear uptake or re-localization ofhnRNP proteins to the cytoplasm (31,32). hnRNP K is primarilynuclear in actively transcribing cells but contains a shuttlingdomain which can drive reporter proteins into the cytoplasm inthe absence of transcription (9). To test if FBP similarly re-localizesafter inhibiting transcription, GFPFBP-transfected cells weretreated with actinomycin D. No re-partitioning of GFPFBP tothe cytoplasm was seen. Surprisingly, a dramatic intranuclearre-distribution occurred as the diffuse nuclear fluorescenceseen in 80% of untreated cells coalesced into hundreds of fineintranuclear dots after treatment in all cells (Fig. 5A, panel 2;summarized in Fig. 5B). Prior to treatment just 20% of the cellsdemonstrated this multi-punctate pattern. FBP∆CD respondedidentically to actinomycin D treatment as did GFPFBP (Fig. 5A,panel 3; summarized in Fig. 5B).

The C-terminal activation domain, FBPC, was distributeddiffusely and in dots throughout the nucleus, prior to actino-mycin D treatment. After treatment approximately one-half toone-quarter of the fluorescent material within each compartmentescaped into the cytoplasm (Fig. 5A, panel 6; summarized inFig. 5B). The fluorescent pattern of GFPFBPN was notchanged by actinomycin D treatment (Fig. 5A, panel 4;summarized in Fig. 5B).

FBPCD, possessing the weakest NLS of FBP’s functionaldomains, partly localized to the cytoplasm prior to actino-mycin D treatment, but the bulk of the protein remainednuclear. Upon blocking RNA synthesis, however, nuclearenrichment of FBPCD was lost as the molecule became eitherentirely delocalized throughout the cell or accumulated in acytoplasmic, peri-nuclear halo (Fig. 5A, panel 5; summarizedin Fig. 5B). So the NLS activities resident within each domainof FBP were differentially coupled with ongoing transcription.The preservation of the resident NLS activities of full-lengthFBP, FBPN and FBPC despite transcription inhibition

Figure 3. The nuclear localization signal in the FBP central domain (FBPCD)resides in a single α-helix (amino acids 328–348). (A) Fluorescence micrographsof GFPFBPCD derivatives shown in (B). The II→PP mutations positioned atamino acids 338, 339 in the fourth α-helix abolished its nuclear targeting.(B) Summary of transfected GFPFBPCD derivatives and quantitation oftrafficking. (C) GFP immunoblot was performed to verify and quantitate GFPfusion protein expression. Arrowheads indicate GFP chimera. An arrow marksa non-specific band (NS). (D) Similar α-helices including a conserved isoleucinedyad are found in all FBP family members.

Nucleic Acids Research, 2000, Vol. 28, No. 22 4563

indicated that the nuclear localization of FBP was not due toadventitious association with nuclear RNA.

FBP co-localizes with nascent transcripts and thetranscription machinery

If FBP’s N-terminal transcriptional repression and C-terminalactivation domains engage the transcription machinery in vivo,then at least some of the time FBP should traffic to transcriptionsites and interact with the transcription machinery. To visualizetranscription sites in vivo, the GFPFBP-transfected cells werepermeabilized and incubated with BrUTP. Subsequent stainingwith anti-bromo-uridine revealed co-localization of FBP andnewly synthesized nascent RNA molecules (Fig. 6, top row).Because in vitro transcription experiments have indicated thatFBP’s transactivation domain operates through TFIIH (J.Liu,personal communication), experiments were conducted todetermine whether this in vitro observation has an in vivocorrelate. Therefore, GFPFBP-transfected cells were stainedwith anti-TFIIH. In fact, FBP co-localized with TFIIH using apolyclonal antibody recognizing p89/XPB, the largest subunitof this multifunctional complex (Fig. 6, middle row). TFIIHstaining was seen in hundreds of microdots pepperedthroughout the nucleus, a pattern commonly observed using

Figure 4. Tyrosine-rich transcription activating repeats (YM1–3) are required tolocalize FBPC into nuclei. (A) Deleting YM1–3 progressively impairs nuclearlocalization. Proteins with YM1–3 efficiently localize to the nucleus, panels 1and 4. Proteins with two (panel 5) or one tyrosine-rich motifs (panels 2, 6, 7 and8) localize to the nucleus less well (3). Compared with YM1 or YM2 (panels 6and 7), YM3 most strongly traffics to the nucleus (panel 8). The same mutationsthat disable YM3 transcription activation (12) abrogate nuclear localization(panel 8 versus panels 9 and 10). (B) Summary and quantitation of impairednuclear localization with progressive deletion or mutation of FBPC. (C) Expressionof GFPFBPC deletion constructs in HeLa cells was monitored by GFP immunoblot.The location of each expressed protein was pointed by arrowhead. A non-specificband is unmarked. (D) Conservation of YM1–3 through FBP family. YM3persistently exists among FBP family, which implied it is an essential role in FBPfamily function. FBP2 duplicatesYM2, whereas FBP3 loses YM1 motifs.

Figure 5. Transcription determines sub-nuclear compartmentalization of FBP.(A) Fluorescence micrographs of FBP or its subdomains before (Fig. 1B) andafter transcription inhibition with actinomycin D. Following addition ofactinomycin D, FBP (panel 2) and FBP∆CD (panel 3) completely re-localizedinto ‘dots’. FBPN remained diffusely nuclear (panel 4). More surprisingly,FBPCD was totally locked out of nuclei (panel 5). FBPC partially diffusedinto cytoplasm (panel 6). (B) The summary and quantitation of re-distributionof FBP and derivatives following treatment with actinomycin D.

4564 Nucleic Acids Research, 2000, Vol. 28, No. 22

antibodies directed against components of the transcriptionmachinery. In contrast, FBP did not co-localize in nuclearspeckles with SC-35, a protein commonly used as a markerassociated with RNA processing components (33).

DISCUSSION

The complex routes traveled by many nuclear proteins must bedelimited by functional constraints. Biochemical and geneticcharacterization of a protein’s physiological role provide aguide to understand its intracellular travels. Without such a map,even the highly regulated flow of important molecules may bedisguised or dismissed. The dynamic compartmentalizationdisplayed by many transcription factors only makes sensewithin the contexts of the upstream pathways regulating factoractivity and of the biology of the downstream target genes.NFκB, glucocorticoid receptor, β-catenin, SMADs, SREBPand many other transcription factors most of the time and inmost cells localize away from their sites of operation (34–38).Similarly, components of the splicing machinery traffic fromcytoplasmic sites of synthesis, to assembly points, and oncenuclear, a variety of RNA-binding proteins commute fromstorage depots to target RNA processing centers (8,32,39,40).Observation of these molecules as markers or indicators ofintracellular transport mechanisms without consideration offunctional context could lead to incomplete or even misleadingconclusions concerning their function and regulation. Frompromoter activity, to splicing, mRNA transport and mRNAdegradation, FBP and its close relatives FBP2 and FBP3 havebeen proposed to participate in all stages of mRNA synthesisand catabolism (11,12,14,16–18). What then is the significance

of the intranuclear distribution and multiple NLSs in FBPwithin the context of FBP function?

Although most KH-proteins such as hnRNP K, ZBP1, Vg1RBP, FMRP, SAM68, etc., participate in, or have beenproposed to participate in, nuclear processes, especially RNAprocessing, they are also believed to be involved with cyto-plasmic functions such as translation, control of RNA transport,signal trandsuction, etc. (28,41–46). In contrast, in our studiesFBP virtually always abides within the nucleus. So if FBPproves to be multi-functional, its primary function is likely tobe nuclear. FBP has three independent and non-overlappingsegments directing nuclear import, one within each of itsfunctional domain. At least two of these determinants ofnuclear localization are enmeshed with functional modules ofFBP activity; one motif is contained within the nucleic-acidbinding KH-repeat 3, while another overlaps the transcription-activating tyrosine-rich C-terminus. To a first approximation,the pattern of immunostaining exhibited by native FBP appearsto be a composite created by superimposing the activities ofeach FBP-NLS. Full-length FBP is mostly found homo-geneously dispersed and partly focused in hundreds of dots;less frequently it is concentrated in fewer spots. This patternmostly closely parallels the NLS activities found in the N- andC-termini of FBP. Therefore, all three NLSs seem to operate atleast occasionally and appear not to be latent activities. Thedifferent intranuclear patterns programmed by each FBP NLSmay indicate FBP functions or is stored at multiple nuclearsites.

Four arguments relate the nuclear sublocalization of FBP. First,the N- and C-terminal NLSs occur in domains also possessingtranscription effector function in vivo and in vitro. Second, FBPlocalizes at sites of transcription. Third, FBP co-localizes withTFIIH, an essential general transcription factor. Fourth, inhibitionof transcription did not propel endogenous FBP to cytoplasmicdestinations. Thus, FBP is dissimilar from most hnRNP moleculeswhich shuttle, with their final RNA cargoes, to the cytoplasmand remain extra-nuclear when RNA synthesis, and hencenuclear re-uptake, is inhibited (32).

FBP and FBP2/KSRP have been associated in vitro withsplicing complexes, so perhaps under particular conditions,FBP participates in RNA processing. In fact FBP binds tightlywith the FBP interacting repressor (FIR). FIR regulates theactivity of general transcription factor TFIIH to repress activatedtranscription. But alternatively spliced forms of FIR have beenreported to bind to RNA and to interact with the splicingmachinery (47). The relative roles of FBP and FIR in transcription,RNA-processing or other intranuclear processes requiresfurther illumination.

The bipartite NLS in FBP’s N-terminus is standard, theNLSs in the central and C-terminal domains are non-canonical.The non-standard NLS might be imported using standard ornon-standard nuclear receptors; or alternatively their nuclearuptake may be controlled through partner proteins importedusing conventional mechanisms. The NLS in the centraldomain neatly resides in a single α-helix; alone this NLS is notover-powering and in some cells, FBPCD is cytoplasmic.Whether nuclear uptake is less efficient in these cells or if theenhanced cytoplasmic retention is due to partner proteins is notknown. It is possible that partitioning of FBPCD between thenucleus and cytoplasm is influenced by interactions with the

Figure 6. FBP co-localizes with both transcription sites and general transcriptionfactor TFIIH, but not with splicing factor SC-35. HeLa cells, transfected withGFPFBP (middle panel) were permeabilized and transcriptionally pulse-labeledwith 5-bromo-2-uridine-5-triphosphate and then immunostained for sites of BrUTPincorporation (red, top row). The merged pictures of GFPFBP fluorescence andtranscription sites show extensive co-localization (right, top row). GFPFBPtransfected cells were also immunostained for the p89/XPB subunit of TFIIH(red, middle row), again revealing co-localization (right, middle row).However, GFPFBP localized to different sites than SC35 (red, bottom row).

Nucleic Acids Research, 2000, Vol. 28, No. 22 4565

FIR, which binds with the FBPCD (13). Such interactionsmight serve either to retard or promote nuclear transport.

The NLS in the C-terminus (FBPC) does not map neatly to adiscrete sequence; rather the efficiency of nuclear localizationbecomes progressively diminished as more and more of thisdomain is deleted. FBPC NLS activity is most impaired bydeletions or mutations, which encroach upon the third tyrosine-rich activating motif, YM3. As this domain binds with TFIIH(J.Liu, personal communication), the distribution of FBPCmay be coupled with the trafficking of TFIIH, a nine-subunitcomplex associated with transcription, DNA-repair and cell-cycleregulation (48).

The data indicate that FBP NLSs operate in parallel and not inseries. Therefore, the nuclear address of FBP is not hierarchical.FBP is not first delivered to the nucleus with a typical NLS,then sorted and sub-localized. Rather parallel nuclear importpathways seem to draw upon a common pool of FBP. Thereforethe intranuclear distribution of FBP must be determined by thetotal amount of FBP and the relative rate constants governingexchange between compartments. Interfering with the equilibriumor steady-state flux of FBP between sites might uncouple, butnot necessarily disrupt the relevant pathways. The ability ofFBP NLSs to operate in isolation from each other indicates thatFBP is unlikely to flow or carry cargo in an obligatory order tosequential nuclear stations. If acting in series, then removal ofany FBP NLSs would improperly localize FBP and disrupt theformation and regulation of downstream complexes. As functionaland genetic studies of FBP progress, the logic of the sub-cellular trafficking of FBP promises to reveal connectionsbetween specific gene transcription and other cellular processes.

ACKNOWLEDGEMENTS

We thank Drs L. Liotta, J. Yewdell and T. Misteli for discussionsand critical review of this manuscript. We thank Mary Strackefor help with figures. A.W. was supported by DFG fellowship(We 2397/1-1).

REFERENCES1. Srinivasan,M., Edman,C.F. and Schulman,H. (1994) J. Cell. Biol., 126,

839–852.2. Tuteja,R. and Tuteja,N. (1998) Crit. Rev. Biochem. Mol. Biol., 33, 407–436.3. Rosenblum,J.S., Pemberton,L.F., Bonifaci,N. and Blobel,G. (1998)

J. Cell. Biol., 143, 887–899.4. Yoneda,Y. (1997) J. Biochem., 121, 811–817.5. Seikimoto,T. and Yoneda,Y. (1997) Cytokine Growth Factor Rev., 9,

205–211.6. Boulikas,T. (1994) J. Cell. Biochem., 55, 32–58.7. Hicks,G.R. and Raikhel,N.V. (1995) Annu. Rev. Cell Dev. Biol., 11, 55–88.8. Siomi,H. and Dreyfuss,G. (1995) J. Cell. Biol., 129, 551–560.9. Michael,W.M., Eder,P.S. and Dreyfuss,G. (1997) EMBO J., 16, 3587–3598.

10. Fan,X.C. and Steitz,J.A. (1998) Proc. Natl Acad. Sci. USA, 95, 15293–15298.11. Duncan,R.D., Bazar,L., Michelotti,G., Tomonaga,T., Krutzsch,H.,

Avigan,M. and Levens,D. (1994) Genes Dev., 8, 465–480.

12. Duncan,R., Collins,I., Tomonaga,T., Zheng,T. and Levens,D. (1996)Mol. Cell. Biol., 16, 2274–2282.

13. Liu,J., He,L., Collins,I., Ge,H., Libutti,D., Li,J., Egly,J.-M. and Levens,D.(2000) Mol. Cell, 5, 331–341.

14. He,L., Liu,J., Collins,I., Sanford,S., O’Connell,B.C., Benham,C.J. andLevens,D. (2000) EMBO J., 19, 1034–1044.

15. Davis-Smyth,T., Duncan,R.C., Zheng,T., Michelotti,G.M. and Levens,D.(1996) J. Biol. Chem., 271, 31679–31687.

16. Min,H., Turck,C.W., Nikolic,J.M. and Black,D.L. (1997) Genes Dev., 11,1023–1036.

17. Grossman,J.S., Meyer,M.I., Wang,Y.C., Mulligan,G.J., Kobayashi,R. andHelfman,D.M. (1998) RNA, 4, 613–625.

18. Irwin,N., Baekelandt,V., Goritchenko,L. and Benowitz,L.I. (1997)Nucleic Acids Res., 25, 1281–1288.

19. Thompson,C.D., Frazier-Jessen,M.R., Rawat,R., Nordan,R.P. andBrown,R.T. (1999) Biotechniques, 27, 824–832.

20. D’Agostino,M., Ciminale,V., Zotti,L., Rosato,A. and Chieco-Bianchi,L.(1997) J. Virol., 71, 75–83.

21. Wei,X., Somanathan,S., Samarabandu,J. and Berezney,R. (1999) J. Cell Biol.,146, 543–558.

22. Jackson,D.A., Hassan,A.B., Errington,R.J. and Cook,P.R. (1993) EMBO J.,12, 1059–1065.

23. Pihan,G.A., Purohit,A., Wallace,J., Knecht,H., Woda,B., Quesenberry,P.and Doxsey,S.J. (1998) Cancer Res., 58, 3974–3985.

24. Chalfie,M., Tu,Y., Euskirchen,G., Ward,W.W. and Prasher,D.C. (1994)Science, 263, 802–805.

25. Dingwall,C. and Laskey,R.A. (1988) Trends Biochem. Sci. 16, 478–481.26. Wang,X.J., Avigan,M. and Norgren,R.B. (1998) Neurosci. Lett., 252,

191–194.27. Ostareck-Lederer,A., Ostareck,D. and Hentze,M.W. (1998) Trends

Biochem. Sci., 23, 409–411.28. Ostareck,D.H., Ostareck-Lederer,A., Wilm,M., Thiele,B.J., Mann,M. and

Hentze,M.W. (1997) Cell, 89, 597–606.29. Bomsztyk,K., Van Seuningen,I., Suzuki,H., Denisenko,O. and

Ostrowski,J. (1997) FEBS Lett., 403, 113–115.30. Ladomery,M. (1997) Bioessays, 19, 903–909.31. Pinol-Roma,S. and Dreyfuss,G. (1991) Science, 253, 312–314.32. Pinol-Roma,S. and Dreyfuss,G. (1992) Nature, 355, 730–732.33. Fu,X.D. and Maniatis,T. (1992) Science, 256, 535–538.34. Easwaran,V., Pishvaian,M., Salimuddin and Byers,S. (1999) Curr. Biol.,

9, 1415–1418.35. Ghosh,S. and Baltimore,D. (1990) Nature, 344, 678–682.36. Picard,D. and Yamamoto,K.R. (1987) EMBO J., 6, 3333–3340.37. Wang,X., Sato,R., Brown,M.S., Hua,X. and Goldstein,J.L. (1994) Cell,

77, 53–62.38. Tsukazaki,T., Chiang,T.A., Davison,A.F., Attisano,L. and Wrana,J.L.

(1998) Cell, 95, 779–791.39. Sandri-Goldin,R.M. (1998) Genes Dev., 12, 868–879.40. Truant,R., Fridell,R.A., Benson,E.R., Herold,A. and Cullen,B.R. (1998)

Eur. J. Cell Biol., 77, 269–275.41. Siomi,M.C., Zhang,Y., Siomi,H. and Dreyfuss,G. (1996) Mol. Cell. Biol.,

16, 3825–3832.42. Ross,A.F., Oleynikov,Y., Kislauskis,E.H., Taneja,K.L. and Singer,R.H.

(1997) Mol. Cell. Biol., 17, 2158–2165.43. Havin,L., Git,A., Elisha,Z., Oberman,F., Yaniv,K., Schwartz,S.P.,

Standart,N. and Yisraeli,J.K. (1998) Genes Dev., 12, 1593–1598.44. Zhang,Q., Yaniv,K., Oberman,F., Wolke,U., Gif,A., Fromer,M.,

Taylor,W.L., Meyer,D., Standart,N., Rez,E. and Yisreali,J.K. (1999)Mech. Dev., 88, 101–106.

45. Ceman,S., Brown,V. and Warren,S.T. (1999) Mol. Cell. Biol., 19, 7925–7932.46. Laird,A.D. and Shalloway,D. (1997) Cell. Signal., 9, 249–255.47. Page-McCaw,P.S., Amonlirdviman,K. and Sharp,P.A. (1999) RNA, 5,

1548–1560.48. Frit,P., Bergmann,E. and Egly,J.-M. (1999) Biochimie, 81, 27–38.