Nuclear localisation of the G-actin sequestering peptide thymosin β₄

Actin is present at high concentrations in virtually every eukaryotic cell. About half of the intracellular actin is stabilised in its monomeric form (G-actin) by interaction with sequestering factors (Pollard and Cooper, 1986). This monomeric actin can be used for the fast generation of new actin filaments after an appropriate intra- or extracellular signal (Carlier and Pantaloni, 1994). The β-thymosins constitute a family of highly conserved water soluble 5-kDa polypeptides. Thymosin β₄ is the most abundant member of this family and is regarded as the main G-actin sequestering peptide in the cytoplasm of mammalian cells (Low et al., 1981; Hannappel et al., 1982; Hannappel and Leibold, 1985; Safer et al., 1991). This 43 amino acid oligopeptide forms a 1:1 complex with G-actin and thereby inhibits salt-induced polymerisation to F-actin (Hannappel and Wartenberg, 1993; Heintz et al., 1993; Yu et al., 1993; Huff et al., 1995; Huff et al., 2001). Additional members of the β-thymosin family have been identified and these peptides exhibit similar properties to thymosin β₄. Thymosin β₄ and other β-thymosins appear to be involved in a number of different processes like cancerogenesis (Hall, 1991; Bao et al., 1996; Verghese-Nikolakaki et al., 1996; Califano et al., 1998; Kobayashi et al., 2002; Otto et al., 2002) and apoptosis (Iguchi et al., 1999; Niu and Nachmias, 2000; Müller et al., 2003). In the extracellular space, thymosin β₄ participates in several physiological processes, e.g. angiogenesis (Grant et al., 1995; Malinda et al., 1997; Philp et al., 2003b), wound healing (Malinda et al., 1999; Sosne et al., 2001; Sosne et al., 2002; Philp et al., 2003a) and regulation of inflammation (Sosne et al., 2001; Sosne et al., 2002). We have recently shown that it also serves as a specific glutaminyl substrate of transglutaminases (Huff et al., 1999) which crosslink thymosin β₄ released from stimulated human platelets to fibrin and collagen (Huff et al., 2002).

Lately, there has been increasing evidence for the presence of cytoskeletal proteins in the nucleus, such as actin itself (Scheer et al., 1984; Gonsior et al., 1999), actin-related proteins (Arps) (Cairns et al., 1998; Harata et al., 1999; Harata et al., 2000) and a number of different actin binding proteins (Nowak et al., 1997; Giesemann et al., 1999; Pestic-Dragovich et al., 2000; Skare et al., 2003). Although the functions of these proteins in the nucleus are still under investigation, there is evidence that they are involved in activities ranging from nuclear assembly and shape changes to DNA replication and transcription (Rando et al., 2000; Olave et al., 2002; Pederson and Aebi, 2002; Shumaker et al., 2003). The intracellular localisation of thymosin β₄ has never been studied in detail. One study using immunofluorescence described that its intracellular localisation in macrophages was most intense in the centre of the cell but was not nuclear (Yu et al., 1994). In another study, [¹²⁵I]-labelled thymosin β₄ was injected into the cytoplasm of Xenopus laevis oocytes and the nuclear and cytoplasmic radioactivity was monitored. In these cells thymosin β₄ was distributed roughly equally between cytoplasm and nucleus (Watts et al., 1990). Recently, McCormack and colleagues (McCormack et al., 1999) studied the influence of polyamine depletion on the relationship of F-actin to G-actin and thymosin β₄ in migrating IEC-6 cells. In untreated cells, they found only cytoplasmic staining for thymosin β₄, whereas they detected nuclear staining after polyamine depletion and speculated that the translocation of thymosin β₄ into the nucleus was a result of this particular treatment (McCormack et al., 1999). In the course of our studies on the possible role of thymosin β₄ in carcinogenesis, we studied the intracellular localisation of this peptide using a newly generated monospecific antibody against thymosin β₄. Using the human mammary carcinoma MCF-7 cell line, we found variable cytoplasmic staining but also additional nuclear staining.

As the previously published discrepancies on the intracellular distribution of thymosin β₄ and our own findings may have been caused by variations in the preservation of the original localisation of the highly soluble and diffusible thymosin β₄, we decided to re-investigate its intracellular localisation by microinjecting fluorescently labelled thymosin β₄ into cells of a number of different lines. Thymosin β₄ can be labelled at two of its three glutamine-residues by the enzymatic reaction of transglutaminase without influencing its G-actin sequestering activity (Huff et al., 1999). Here this technique was used to label thymosin β₄ with Oregon Green cadaverine as a fluorescent marker. Fluorescence microscopic inspection after microinjection of the labelled peptide into cells of a number of different lines revealed that a considerable amount of thymosin β₄ was located within their nuclei. The translocation of thymosin β₄ into the nucleus is not achieved by simple diffusion, as the labelled peptide could not be detected within nuclei when the cells were previously treated with digitonin under conditions that extract the soluble components of the cytoplasm by permeabilisation of the plasma membrane while leaving the nuclear envelope intact. Nuclear localisation was observed only after subsequent treatment and permeabilisation of the nuclear membranes with Triton X100. These data are further supported by results showing that after enzymatic cleavage of bis-labelled thymosin β₄ only the larger N-terminal fragment (Tβ ^1-26₄), containing the proposed actin-binding site, was translocated to the nucleus. In contrast, the smaller C-terminal fragment (Tβ ^27-43₄) and fluorescently labelled thymosin β₄ chemically crosslinked to ADP-ribosylated actin were retained in the cytoplasm.

Reagents were obtained from the following sources: LiChroprep RP18 (40-63 μm) and trifluoroacetic acid (Uvasol) from Merck (Darmstadt, Germany); guinea pig transglutaminase and 1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide (EDC) from Sigma (Munich, Germany); AsnC-Proteinase from PanVera Corporation (Madison, WI); TRITC-phalloidin, Oregon Green-labelled deoxyribonuclease I and Oregon Green cadaverine were from Molecular Probes (Eugene, OR).

Actin was prepared from rabbit skeletal or bovine heart muscle by the method of Pardee and Spudich (Pardee and Spudich, 1982) and stored as G-actin in G-buffer (2 mM Tris, 0.2 mM ATP, 0.2 mM CaCl₂, 0.5 mM mercaptoethanol, 0.05% NaN₃, pH 8.0) at 0°C. Thymosin β₄ was isolated from pig spleen as described (Hannappel et al., 1989). The purity of the preparation was demonstrated by reverse-phase HPLC. The concentrations of thymosin β₄ and actin were determined by amino acid analysis after acid hydrolysis (6 M HCl, 155°C, 1 hour) and pre-column derivatisation with o-phthaldialdehyde/3-mercaptopropionic acid (Hannappel et al., 1988).

Fluorescently labelled thymosin β₄ was prepared by incubation of 240 μg thymosin β₄ (200 μM) with 120 μg Oregon Green cadaverine (OGC) (1 mM) and 0.2 U guinea pig transglutaminase at room temperature in 240 μl buffer consisting of 10 mM Tris-HCl, pH 7.4, 15 mM CaCl₂, 3 mM DTT. After 1 and 2 hours, 5 μl of the reaction mixture were subjected to HPLC analysis. The reaction was stopped after 4 hours by addition of 5 μl trifluoroacetic acid (TFA). Then the reaction mixture was subjected to preparative HPLC. Separated peptides were concentrated in vacuo and then characterised by amino acid analysis and MALDI-TOF mass spectrometry.

Proteolytic fragments of OGC-labelled thymosin β₄ were prepared by the following procedure: 50 μg of peptide was incubated with 20 μU AsnC-endoproteinase in 100 μl reaction buffer (50 mM sodium acetate, pH 5.0, 0.2 mM DTT, 0.2 mM EDTA) for 16 hours at room temperature. Then the reaction was stopped by adding 5 μl 10% TFA and products were separated by preparative HPLC. Prior to analysis by MALDI-TOF-MS the samples were concentrated in vacuo.

In order to avoid intracellular dissociation of microinjected actinthymosin β₄ complex and/or its polymerisation, we chemically crosslinked thymosin β₄ to ADP-ribosylated actin which is known to be polymerisation incompetent on its own (Schering et al., 1988). Therefore, fluorescently labelled thymosin β₄ was crosslinked to ADP-ribosylated actin using EDC as described previously (Ballweber et al., 2002). ADP-ribosylated actin was generated by treatment of rabbit skeletal muscle actin with iota toxin and nicotinamide-adenosine dinucleotide (NAD) as detailed (Ballweber et al., 2001).

Chromatographic conditions were controlled by a Merck-Hitachi L-6200 system supplemented with a diode-array UV detector (L-7450A, Merck-Hitachi), a reaction pump for post-column derivatisation (655A-13, Merck-Hitachi), and with a fluorometer (F-1050, Merck-Hitachi). The diode-array-detector signal was recorded on a computer using D-7000 HSM software (Merck) and the fluorescence signal on an integrator (D-2500, Merck-Hitachi). The flow rate was 0.75 ml /minute in 0.1% TFA (trifluoroacetic acid) with a 0-40% acetonitrile gradient over 60 minutes in a Beckman ODS Ultrasphere (5 μm, 4.6×250 mm) column. UV detection was at 205 nm and fluorescence was detected after post-column derivatisation with fluorescamine (Huff et al., 1997).

Mass determinations were performed with a Biflex™ III MALDI-TOF mass spectrometer (Bruker Daltonics, Bremen, Germany). The instrument is equipped with a nitrogen laser (λ=337 nm) and a reflectron. Laser-desorbed positive ions were analysed after being accelerated by 19 kV in the reflection mode. External calibration was performed by use of a standard peptide mixture. Thirty individual spectra were averaged to produce a mass spectrum. Dried peptide samples were dissolved in 0.1% TFA containing 33% acetonitrile to a final concentration of about 20 ng/μl. Each sample (1 μl) was mixed with 2 μl of a saturated solution of α-cyano-4-hydroxy-cinnamic acid (Sigma, Germany) in 0.1% TFA, 33% acetonitrile and 1 μl of this mixture was spotted onto a stainless steel target.

Actin:thymosin β₄ complex was generated by mixing equimolar concentrations of native or labelled thymosin β₄ with G-actin. These mixtures were used for a comparative determination of dissociation constants of G-actin in complex with either thymosin β₄ or its fluorescently labelled derivatives by equilibrium centrifugation as described (Huff et al., 1995).

Viscometric measurements were done with a falling-ball viscometer according to Cooper and Pollard (Cooper and Pollard, 1982) at an angle of 40° relative to the horizontal over a distance of 45 mm with a 0.794 mm diameter ball. Forty-eight microliters G-actin solution (0.18 mg/ml in G-buffer) was incubated with or without thymosin β₄ or Oregon Green cadaverine-labelled peptides for 15 minutes at room temperature and then 2 μl 50 mM MgCl₂ were added. The mixture was filled into a glass capillary (diameter 0.92 mm, 50 μl micro pipettes), sealed at one end and incubated for 4 hours before measuring.

A synthetic decapeptide representing the nine C-terminal amino acids of thymosin β₄ with an additional cysteine residue at the N-terminus was conjugated to keyhole limpet hemocyanin (KLH, Sigma, Germany). New Zealand White rabbits were immunised with the KLH conjugate, with an amount corresponding to about 63 μg of the synthetic peptide emulsified with complete Freund's adjuvant (Sigma, Munich, Germany). Following a second immunisation, serum was collected and the anti-thymosin β₄ antibody was partially purified from the serum by precipitation with 50% ammonium sulfate. The precipitate was dissolved in 5 mM phosphate buffer, pH 6.5, dialyzed against PBS and adsorbed with a 1% suspension of acetone powder from bovine heart in order to remove antibodies reacting non-specifically with cytoskeletal components. The antibody was affinity purified by passing it over a column coupled with immunogenic decapeptide. The resulting antiserum showed no cross-reactivity with other β-thymosins, actin, or any other cellular proteins of molecular weights in the range from 10-50 kDa as judged by western blot analysis and ELISA.

Kidney cells from African green monkey (Vero cells) and the human mammary cancer cell line MCF-7 were maintained in Dulbecco's MEM/F12 (Gibco, UK) supplemented with 10% (v/v) foetal calf serum (Gibco). The rat fibroblastic NRK, the human cervical cancer HeLa and epidermoid cancer A431 cell lines were maintained in DMEM supplemented with 10% FCS-gold (Invitrogen, Karlsruhe, Germany).

Microinjection was performed with an ECET cell injection system (Eppendorf, Hamburg, Germany) consisting of the micromanipulator 5170 and the microinjector 5242 adapted to an Axiovert 100 inverted microscope (Zeiss, Göttingen, Germany). Microinjections were visually controlled by a CCD camera on a TV monitor (SSM 121CE, Sony, Tokyo, Japan). Fluorescent thymosin β₄ and crosslinked ADP-ribosylated actin:thymosin β₄ complex were injected into the cytoplasm at a concentration of 32 μM and 8.27 μM respectively, in 135 mM KCl, 5 mM Na₂HPO₄, pH 7.2. The injection pressure was between 65 and 80 hPa (1 hPa=0.1 kPa) and the injection time between 0.5 and 0.7 seconds.

HeLa cells were used for in vitro nuclear translocation experiments, as most previous nuclear translocation experiments have been performed with this cell line. HeLa cells were grown on coverslips until confluence was almost reached. Subsequently the cells were treated as described (Adam et al., 1990) with slight modifications (Jäkel and Görlich, 1998), i.e. they were washed three times with ice-cold PBS, placed on ice, and treated for 12 minutes with permeabilisation buffer (0.11 M potassium acetate, 5 mM magnesium acetate, 0.25 M sucrose, 0.5 mM EGTA and 20 mM HEPES, pH 7.5) supplemented with 40 μg/ml digitonin. Then the cells were washed three times with ice-cold permeabilisation buffer for 2, 5 and 10 minutes. In order to permeabilise the nuclear envelope, the digitonin extracted cells were incubated with 0.2% Triton X100 for 10 minutes and subsequently washed three times with permeabilisation buffer. Then the cells were incubated for 2 hours at room temperature with 3 μM fluorescent thymosin β₄ in import buffer (120 mM potassium acetate, 5 mM magnesium acetate, 0.5 mM EGTA, 0.25 mM sucrose and 20 mM KHPO₄, pH 7.2).

For immunofluorescence staining of MCF-7 cells with anti-thymosin β₄ antibody cells were fixed with 1.2% paraformaldehyde, permeabilised with 0.2% Triton X100, washed in PBS and stained using anti-thymosin β₄ antibody. Cell nuclei were counterstained by treating fixed cells with 1 μg/ml Hoechst 33258. Oregon Green labelled DNase I was employed to specifically stain monomeric actin. The microfilament system was visualised by staining paraformaldehyde fixed cells with TRITC-phalloidin as described (Paddenberg et al., 2001). Fluorescence microscopy and visualisation of microinjected fluorescent thymosin β₄ was achieved either using standard or confocal microscopy with a Zeiss Axiophot equipped with epifluorescence optics, or with a Zeiss LSM510 confocal microscope (Zeiss, Göttingen, Germany).

In the course of our studies on the possible role of thymosin β₄ in carcinogenesis we studied the intracellular localisation of this peptide in MCF-7 cells using an anti-thymosin β₄ antibody in indirect immunofluorescence. In addition to the expected cytoplasmic localisation we detected a distinct staining of a number of nuclei (Fig. 1A) as verified by their simultaneous staining with the chromatin specific dye Hoechst 33258 (Fig. 1C). As the nuclear staining with anti-thymosin β₄ varied, we also tested the possibility that there might have been an unequal distribution of monomeric actin in these cells. Therefore, we counterstained the cells with Oregon Green-labelled deoxyribonuclease I (DNase I), which is known to bind with high affinity and specificity to monomeric actin (Mannherz et al., 1980). As shown in Fig. 1B almost all nuclei were stained by the labelled DNase I indicating the presence of considerable amounts of G-actin in the nuclei of these cells. To further confirm the unexpected distribution of thymosin β₄ we analysed its distribution with an alternative but more direct method. Therefore, microinjection of fluorescently labelled thymosin β₄ was attempted.

Oregon Green cadaverine (OGC)-labelled thymosin β₄ was prepared as described in the experimental section. After preparative HPLC, we isolated the two main products showing Oregon Green fluorescence. Amino acid analysis of the isolated peptides showed identical amino acid compositions for both isolated products (data not shown). The peptides were then characterised by mass spectroscopy. Table 1 summarises the observed molecular masses, assigned molecular formula, calculated masses and its deviation. These data showed, that one of the labelled products was a mono-OGC [Tβ₄(OGC)₁] and the other was a bis-OGC thymosin β₄ [Tβ₄(OGC)₂] derivative. Further analysis using enzymatic cleavage with AsnC-endoproteinase or trypsin and subsequent mass spectroscopy revealed that in the mono-labelled derivative Gln-36 and in the bis-labelled derivative Gln-23 and Gln-36 had been derivatised (data not shown). For microinjection studies only the bis-labelled peptide was used.

To ensure that this labelling does not influence the G-actin sequestering activity the dissociation constant for the complex of bis-labelled thymosin β₄ and bovine cardiac G-actin as well as its capacity to inhibit the salt-induced actin polymerisation were determined. Using equilibrium centrifugation, we found that the dissociation constant for the complex of bis-labelled thymosin β₄ with G-actin (0.47±0.1 μM) did not differ significantly from that determined for unlabelled thymosin β₄ (0.59±0.08 μM). Inhibition of salt-induced actin polymerisation was tested using falling ball viscometry and showed inhibition of actin polymerisation by the labelled thymosin β₄ at equimolar concentrations (Table 2).

To assay the intracellular distribution, bis-labelled thymosin β₄ was microinjected into the cytoplasm of MCF-7 cells. As expected, directly after microinjection the labelled peptide was evenly distributed throughout the cytoplasm (data not shown). After incubation for 1 hour a pronounced staining of the cell nucleus was detected (Fig. 2A,B). To ensure that the nuclear localisation was not a cell-specific artefact of the MCF-7 cells, microinjection experiments were also performed with Vero cells, where a comparable pattern was observed (Fig. 2C,D).

Next we analysed the distribution of fluorescently labelled intact thymosin β₄ microinjected into additional cell lines. As in Vero and MCF-7 cells, thymosin β₄ was also found within the nuclei of fibroblastic NIH-3T3 and NRK as well as the human epidermoid A431 cells after microinjection. One hour after microinjection, confocal sections of NIH-3T3 cells revealed a clear accumulation within the cell nucleus (Fig. 3A,B). Fixation of the microinjected cells after 30 minutes demonstrated nuclear accumulation together with varying fluorescence intensity within the cytoplasm. Three hours after microinjection we still observed a remaining although weaker cytoplasmic staining (data not shown). Within the nuclei we observed a homogenous distribution of the microinjected thymosin β₄ with the exception that regions of presumed nucleoli were void of thymosin β₄ (inset in Fig. 3L). A similar staining pattern was obtained with Oregon Green-labelled DNase I specific for G-actin (see Fig. 1B).

In addition, we counterstained microinjected NIH-3T3 (Fig. 3C-E) or NRK cells (Fig. 3F-H) after paraformaldehyde fixation at various time points with TRITC-phalloidin to visualise actin filaments and to analyse the distribution of thymosin β₄ in relation to the microfilament system. Besides its accumulation within the nucleus, the data obtained showed in many cases a faint punctuate staining along the cytoplasmic stress fibres of both fibroblastic NIH-3T3 and NRK cells together with a diffuse cytoplasmic localisation.

The immunohistochemical staining using the monospecific anti-thymosin β₄ antibody had indicated in a number of cases that the nuclear staining was dependent on cell density, being more intense in isolated cells than in cells within cell clusters (see also Fig. 1). Therefore, we microinjected contacting A431 cells in the middle and periphery of cell clusters. However, we did not find a dependence of the nuclear localisation of microinjected thymosin β₄ on the intensity of their cell-cell contacts (Fig. 3L-N).

The low molecular mass of thymosin β₄ (5 kDa) might suggest that it diffused through the nuclear pores by simple diffusion. To analyse its mode of nuclear translocation and accumulation, we incubated digitonin permeabilised and extracted HeLa cells with labelled thymosin β₄. Digitonin treatment of cultured HeLa cells has recently been shown to permeabilise the plasma membrane for macromolecules, but to leave the nuclear envelopes structurally intact and competent for active transport (Adam et al., 1990). Fluorescence microscopy of HeLa cells thus treated revealed a solely cytoplasmic distribution of the labelled peptide, i.e. no staining of the cell nuclei (Fig. 4A′). However, only after an additional treatment of the digitonin-extracted cells with 0.2% Triton X-100 for 10 minutes to also permeabilise the nuclear membranes was it possible to detect thymosin β₄ within the nuclei (Fig. 4B′). These data clearly indicate that the pore complexes of an intact nuclear envelope prevent the passage of thymosin β₄ through the nuclear pores in the absence of soluble cytoplasmic factors. An identical mode of thymosin β₄ exclusion was obtained with freshly isolated MCF-7 nuclei (data not shown).

The N-terminal portion of thymosin β₄ contains a sequence stretch enriched in lysine residues (¹⁴KSKLKK¹⁹) suggestive of a functional nuclear localisation signal. As an initial test as to whether translocation into the nucleus depends on this basic sequence, we digested bis-labelled thymosin β₄ using an AsnC-endoproteinase. Because thymosin β₄ possesses only one asparagine residue at position 26, this digestion produced two fragments: an N-terminal fragment (thymosin β ^1-26₄) and a C-terminal fragment (thymosin β ^27-43₄) each bearing one fluorescent label. This was confirmed by HPLC analysis with detection of Oregon Green fluorescence. After isolation and characterisation of the labelled fragments by amino acid analysis and mass spectrometry (Table 1), they were microinjected into Vero cells. The N-terminal fragment containing the ¹⁴KSKLKK¹⁹ sequence exhibited a pronounced nuclear localisation (Fig. 5A,B), whereas the C-terminal fragment was restricted to the cytoplasm (Fig. 5C,D).

The aforementioned sequence motif ¹⁴KSKLKK¹⁹ partially overlaps with the putative actin binding sequence of thymosin β₄ (¹⁷LKKTET²²). To prove the assumption that the former motif is involved in nuclear translocation and to elucidate whether thymosin β₄ is translocated into the nucleus in complex with actin, fluorescent actin:thymosin β₄ complex was generated by chemical crosslinking with EDC. ADP-ribosylated rabbit skeletal muscle actin was used, as it has been previously shown to bind thymosin β₄ but not to polymerise, in order to secure its monomeric state after microinjection (Ballweber et al., 2001). Successful crosslinking was verified by UV examination of the treated material after SDS-PAGE (data not shown). Confocal microscopy of A431 cells after microinjection of the crosslinked complex showed that actin:thymosin β₄ remained confined to the cytoplasm even after 3 hours (Fig. 6).

It is now well accepted that thymosin β₄ is the main G-actin sequestering peptide in the cytoplasm of mammalian cells. Together with other actin binding proteins that have F-actin severing and capping activities it is involved in the regulation of the ratio between monomeric (G-) and filamentous (F-) actin in the cytoplasm. Despite many publications dealing with the interaction of thymosin β₄ with G-actin, there are few reports focussing on its intracellular localisation. In an early study the intracellular localisation of thymosin β₄ was scrutinised by subcellular fractionation of rat spleen (Erickson-Viitanen et al., 1983), and described to be mainly cytosolic with negligible amounts in the nuclear, mitochondrial/lysosomal or microsomal fractions. Subsequently the intracellular localisation of thymosin β₄ and thymosin β₁₀ in mouse peritoneal macrophages has been studied using specific antibodies raised against these two peptides (Yu et al., 1994). These authors reported that thymosin β₄ immunofluorescence was most intense in the centre of the cell and lower in the periphery and the filopodia, with no staining of the nucleus. Because of its known function as a G-actin sequestering peptide, this cytoplasmic localisation seemed to be reasonable. More recently, the influence of polyamine depletion onto the actin cytoskeleton of migrating IEC-6 cells has been studied (McCormack et al., 1999). These authors used a rabbit polyclonal antibody against thymosin β₄ and described a primarily cytoplasmic staining in control cells, punctuate in appearance and close to or on the nuclear membrane, but no staining of the nucleus itself. In contrast, they found a prominent staining of the nucleus in polyamine-deprived cells or two minutes after treatment of control cells with epidermal growth factor. They concluded that the nuclear appearance resulted from a treatment-induced translocation of thymosin β₄ into the nucleus. Our results on thymosin β₄ localisation in MCF-7 cells using indirect immunofluorescent labelling of an affinity-purified antibody indicate that cells with either cytoplasmic or nuclear accumulation can be found within a growing cell population. The differences between previous studies and our results showing variable degrees of cytoplasmic and nuclear localisation of thymosin β₄ by using immunolocalisation may be caused by variations in the preservation of the original localisation of this highly soluble and diffusible peptide. Another possible explanation may be that recognition of thymosin β₄ within the nucleus may depend on the epitope detected by the anti-peptide antibody. The polyclonal anti-peptide antibody used in this study recognises an epitope comprising of the four C-terminal amino acid residues of thymosin β₄.

Recently, the nuclear localisation of actin itself (Scheer et al., 1984; Gonsior et al., 1999), actin-related proteins (Arps) (Cairns et al., 1998; Harata et al., 1999; Harata et al., 2000) and a number of actin binding proteins (Nowak et al., 1997; Giesemann et al., 1999; Pestic-Dragovich et al., 2000; Skare et al., 2003) has been reported. Although the functions of nuclear actin are far from being fully understood, it has been proposed that it might be involved in chromatin remodelling (Zhao et al., 1998; Shen et al., 2003), mRNA processing and transport (Hofmann et al., 2001; Percipalle et al., 2001; Percipalle et al., 2002). It has been repeatedly reported that in contrast to the cytoplasm, the nuclei are devoid of phalloidin-stainable actin filaments (data also obtained during this study). Indeed, it has been shown that a monoclonal antibody that presumably recognises a particular G-actin conformation yielded a punctuate nuclear staining pattern (Gonsior et al., 1999). Thymosin β₄ was evenly distributed within the nucleus except for in nucleoli, which appeared to be free of this peptide. This distribution was observed after microinjection and immunostaining (see Fig. 1 and Fig. 3L). This staining pattern coincided with the Oregon Green-DNase I staining (Fig. 1B) and was also reported by Gonsior et al. (Gonsior et al., 1999) and is suggestive of an intranuclear co-distribution of both thymosin β₄ and G-actin. DNase I binds G-actin with high affinity and practically irreversibly (Mannherz et al., 1980). We are therefore confident that the nuclear staining and retention of Oregon Green-labelled DNase I was mainly due to actin binding, otherwise chromatin staining by Hoechst 33258 should have diminished or vanished owing to endonucleolytic activity of DNase I (see Fig. 1C).

The fact that nuclear actin seems to be maintained in its monomeric form will necessitate G-actin sequestering factors inside the nucleus. However, the lack of phalloidin staining does not exclude the presence of special forms of F-actin in the nucleus. Indeed, cellular stress frequently induces the formation of nuclear actin rods (Nishida et al., 1987) that are composed of ADF/cofilin decorated actin filaments, which were shown not to bind phalloidin (McGough et al., 1997). In addition, it has also been proposed that F-actin may be present in the nucleus in the form of very short and/or highly branched filaments (Pederson and Aebi, 2003).

Our data obtained by microinjection of fluorescent thymosin β₄ into several different cell lines show a pronounced nuclear localisation of the labelled thymosin β₄. After enzymatic digestion of bis-labelled thymosin β₄ into two labelled fragments, only the larger N-terminal fragment, thymosin β ^1-26₄, was translocated to the nucleus, whereas the smaller C-terminal fragment thymosin β ^27-43₄ remained in the cytoplasm. The amino acid sequence of thymosin β₄ does not contain a canonical nuclear localisation signal, but a cluster of positively charged amino acid residues (¹⁴KSKLKK¹⁹) suggestive of a functional nuclear localisation signal, which partially overlaps with the proposed actin binding site of thymosin β₄. Indeed analysis by SubLoc v1.0 (Hua and Sun, 2001) predicted a nuclear localisation of intact thymosin β₄ (accuracy 74%) as well as of its N-terminal fragment thymosin β ^1-26₄ (accuracy 94%), whereas removal of just the above mentioned cluster from the amino acid sequence changes the prediction to cytoplasmic localisation (accuracy 95%). This assumption was further confirmed by the fact that chemically crosslinked actin:thymosin β₄ complex, in which the ¹⁴KSKLKK¹⁹ motif might be sterically blocked by actin binding, is not translocated into the nucleus. Moreover, this result argues against a possible transport of thymosin β₄ to the nucleus in complex with actin. Our data suggest that this cluster of charged amino acid residues (¹⁴KSKLKK¹⁹) may be involved in the translocation of entire thymosin β₄, as well as its N-terminal fragment into the nucleus.

As treatment of digitonin-permeabilised HeLa cells with labelled peptide resulted in a solely cytoplasmic localisation of the peptide, the translocation of thymosin β₄ to the nucleus cannot be explained by a simple diffusion mechanism of the 5 kDa peptide, nor can it be caused by the fluorescent labelling. Therefore, we propose that thymosin β₄ is translocated into the nucleus by an active transport mechanism, requiring an as yet unknown soluble cytoplasmic factor. It appears surprising that a peptide of only 5 kDa molecular mass does not freely diffuse through nuclear pores. However, structural studies have implicated that thymosin β₄ is an elongated molecule (Zarbock et al., 1990; Czisch et al., 1993; Ballweber et al., 2002) and data from gel filtration experiments have shown that it migrates like a protein of considerably higher molecular mass (Haritos et al., 1989). Future experiments will aim to identify the cytoplasmic factors involved in its nuclear translocation and binding partners.

Our data clearly indicate that distinct amounts of the G-actin sequestering peptide thymosin β₄ are translocated into the nucleus of cells by an active transport mechanism. When considered with the recent detection of G-actin and other actin binding proteins in the cell nucleus, our results suggest that thymosin β₄ is not only the main G-actin sequestering peptide in the cytoplasm of mammalian cells, but may also account for the G-actin sequestering activity within the nucleus. As well as this presumed nuclear activity of thymosin β₄ it is tempting to speculate that it may have additional functions in the nucleus as also proposed for actin itself and other actin binding proteins. Further studies into these nuclear functions may give additional insight into the proposed role of thymosin β₄ in events like carcinogenesis or apoptosis.

Alternatively, a specific function for nuclear actin has been questioned as it was shown that actin itself contains two nuclear export sequence stretches (Wada et al., 1998). Indeed, it has also been shown that nuclear accumulation of the actin binding protein profilin only serves to facilitate its nuclear export (Stüven et al., 2003). In view of these data, it might also be possible that the nuclear accumulation of thymosin β₄ stabilises actin in its monomeric state by forming a G-actin:thymosin β₄ complex that is subsequently transformed into a G-actin:profilin complex (as during actin cycling in the cytoplasm) ready for nuclear export. Thus the supposed sequestering activity of thymosin β₄ within the nucleus might also support the process of nuclear export of actin, which probably occurs after mixing cytoplasmic and nuclear contents during the open mitosis of mammalian cells.

We would like to thank K. Aktories (Freiburg, Germany) for kindly providing us with recombinant iota toxin from C. perfrigens. We express our appreciation to Doris Jaegers and Sabine Büttner for excellent technical assistance. We also want to thank Swantje Wulf for performing the microinjection experiments using NIH-3T3, MCF-7 and A431 cells. This work was supported by grants from the Deutsche Forschungsgemeinschaft, Hu 865/1-2 and Ma 807/14-1.