PTPL1 is the largest known cytoplasmic protein tyrosine phosphatase (PTP) containing a FERM (four point-1, ezrin, radixin and moesin) domain. Enzyme localization and PTP-substrate specificity are thought to play crucial roles in the regulation of PTP activity, which determines their functions. Here we report that PTPL1 is predominantly localized at the apical face of plasma membrane enriched in dorsal microvilli when expressed in HeLa cells. By comparing localization of the full-length enzyme with its FERM domain or FERM-deleted PTPL1 construct, we first concluded that PTPL1-FERM domain is necessary and sufficient to address the wild-type enzyme at the membrane. Two potential phosphatidylinositol 4,5-biphosphate [PtdIns(4,5)P2]-binding motifs were identified within the PTPL1-FERM sequence. We further showed that mutation of both sites altered PTPL1 localization similarly to FERM domain deletion, and impaired its subcellular distribution as confirmed biochemically by cell-fractionation experiments. Using protein-lipid overlays, we demonstrated an interaction of the FERM domain of PTPL1 with PtdIns(4,5)P2, which was lost after mutation of potential PtdIns(4,5)P2-binding motifs. Moreover, neomycin, which masks PtdIns(4,5)P2 polar heads, was shown to decrease by 50% the association of PTPL1 with the cytoskeletal fraction. These results identify the crucial role of the FERM domain in PTPL1 intracellular targeting and demonstrate that localization of PTPL1 is regulated by phosphoinositide metabolism.
Many cellular processes such as cell growth, differentiation, locomotion, and cell-cell interactions are controlled by tyrosyl phosphorylation of specific cellular proteins. Reversible phosphorylation on tyrosine residues depends on a balance between protein tyrosine kinase (PTK) and protein tyrosine phosphatase (PTP) activities (for a review, see Neel and Tonks, 1997). Since any misregulation can lead to severe disease, these activities must be tightly regulated (Hunter, 1997; Zhang, 2001). Relatively little is known about the function and regulation of PTPs. Just as for PTKs, the activity of transmembrane PTPs can, in some instances, be regulated by an inter- or intra-molecular mechanism (e.g. Wallace et al., 1998). Concerning cytosolic PTPs, targeting to highly restricted subcellular micro-compartments is thought to play an important role in the control of their activity and substrate specificity. However, for some PTPs, it can be achieved in combination with an intrinsic property carried by their catalytic domain allowing specific substrate recognition (for a review, see Tonks and Neel, 2001). Cytosolic PTPs possess a remarkable variety of non-catalytic domains. Among their structural motifs, some were shown to target the enzyme towards its correct intracellular context [`zip code' model (see Mauro and Dixon, 1994; Fischer, 1999)].
PTPL1/FAP-1/hPTP1E/PTP-BAS is the largest known cytoplasmic PTP (250 kDa) and belongs to a family of PTPs characterized by the presence of a FERM (four point-1, ezrin, radixin and moesin) domain (Saras et al., 1994; Sato et al., 1995; Banville et al., 1994; Maekawa et al., 1994). This family comprises PTPH1 (Yang and Tonks, 1991), PTPMEG (Gu et al., 1991), PTP36/PTPD2/Pez (Sawada et al., 1994; Moller et al., 1994; Smith et al., 1995) and PTPD1/PTP-RL10 (Moller et al., 1994; Higashitsuji et al., 1995). PTPL1 presents no characteristic feature in its far N-terminal end except for a potential leucine zipper followed in the protein core by a FERM domain, five PDZ domains and the C-terminal catalytic domain. FERM domains are commonly found within a family of peripheral membrane proteins that mediate linkage of the cytoskeleton to the plasma membrane (for reviews, see Chishti et al., 1998; Bretscher, 1999; Mangeat et al., 1999). This family comprises the erythrocyte protein 4.1, ezrin, radixin, moesin, merlin (product of Nf2, a tumor suppressor) and talin. The FERM domain has been well characterized in ERM proteins, prototypes of membrane-cytoskeleton linkers. ERMs harbor a major actin-binding site within their C-terminus and a FERM domain at their N-terminal part, which binds a number of proteins (including EBP 50/NHE3-RF, RhoGDI) some of which are transmembrane proteins such as CD44, CD43, ICAM-2 and ICAM-3. Furthermore, ERM proteins can self-associate in an intra- or inter-molecular manner through homotypic and heterotypic interaction between their N-and C-terminal ends masking several sites involved in the binding of membrane proteins and actin. Such interactions are regulated by tyrosine or threonine phosphorylation but also by phosphatidylinositol 4,5-biphosphate [PtdIns(4,5)P2 (for a review, see Bretscher, 1999)]. Finally, ezrin localization was recently shown to be altered by mutation of PtdIns(4,5)P2-binding motifs found in its FERM domain (Barret et al., 2000).
We previously showed that the expression of PTPL1/FAP-1 was specifically induced by anti-estrogens, such as hydroxytamoxifen (OH-Tam), in estrogen receptor positive human breast cancer cells (Freiss et al., 1998). This induction was correlated with an increase of PTP activity associated with the insoluble fraction (membrane and cytoskeleton) and was involved in the early steps of anti-growth factor action of OH-Tam (Freiss and Vignon, 1994; Bompard et al., 2002).
In the present study, we investigated PTPL1 localization and evaluated the role of its FERM domain in order to understand how its enzyme activity could be regulated. We showed that the FERM domain is necessary and sufficient to target PTPL1 to the apical cell membrane and to allow it to be enriched in dorsal microvilli in HeLa cells. Analysis of PTPL1-FERM sequence suggested the presence of two potential PtdIns(4,5)P2-binding sites. We showed that these two sites, involved in phosphatidylinositol phosphate lipid (PIP) binding in general, and PtdIns(4,5)P2 in particular, play a major role in PTPL1 cellular localization and association with the cytoskeletal fraction.
Collectively, our data bring the first indication for the importance of PtdIns(4,5)P2 binding through the FERM domain for the intracellular targeting and association with cytoskeleton of the human protein tyrosine phosphatase PTPL1.
Materials and methods
Enzymes were purchased from Life Technologies (Cergy-Pontoise, France), Eurogentec (Seraing, Belgium), New England Biolabs and Promega. Monoclonal antibodies against epitope HA (12CA5) and epitope VSV-G (P5D4) were purchased from Roche. Anti-vimentin monoclonal (V9) was from Immunotech. Anti-actin (A 2066) and phalloidin-TRITC polyclonals were from Sigma. Polyclonal anti-tubulin monoclonal was a gift from N. Lamb (Montpellier, France).
Construction of expression plasmids
All proteins encoded from pHM6 vector (Roche) are HA-tagged at their N-terminus and 6His-tagged at their C-terminus. All constructs were verified by sequencing.
The expression construct pHM6-PTPL1 encoding influenza virus hemagglutinin (HA)-epitope tagged PTPL1 (HA-PTPL1) was created as follows. Oligonucleotides carrying Bsu15I and Psp5II restriction sites were introduced in SacII-NotI sites of pHM6 vector. The cDNA encoding the N-terminal part of PTPL1 was cloned by PCR using pSV7d-PTPL1 (gift from C. H. Heldin, Ludwig Institute for Cancer Research, Uppsala, Sweden) as a template in SacII-Bsu15I sites of modified pHM6 vector. The SacII site was introduced by PCR in frame with the second coding codon (removal of start codon) and the Bsu15I was an endogenous PTPL1 site. By the same approach, the cDNA encoding the C-terminal part of PTPL1 was then inserted in Psp5II-NotI sites of modified pHM6 vector. The Psp5II was an endogenous PTPL1 site and the NotI site was introduced by PCR in order to remove PTPL1 stop codon. Finally, the remaining PTPL1 cDNA from pSV7d-PTPL1 was inserted in Bsu15I-Psp5II sites generating pHM6-PTPL1 (aa 2-2459, Fig. 1).
pHM6-FERM encoding HA-tagged FERM domain (aa 511-958) of PTPL1 (HA-FERM) was cloned by PCR in EcoRI-SacII of pHM6 (Fig.1). pHM6-PTPL1 ΔFERM (deletion of aa 523-909, Fig.1) was generated as followed. pBSK-FERM (aa 511-958) was digested by BamHI and MscI, and the open reading frame was restored by introducing double-stranded oligonucleotides (hybridized oligonucleotides) generating pBSK-ΔFERM. ΔFERM was then excised by StuI/PacI digestions and the fragment was cloned in pBSK-PTPL1 (aa 2-2459) in order to generate pBSK-PTPL1 ΔFERM. Full-length PTPL1 ΔFERM was then subcloned in pHM6 through SacII/NotI sites. Mutagenesis of potential PtdIns(4,5)P2-binding sites was performed with oligonucleotides using QuickChangeTM Site-Directed Mutagenesis Kit (Stratagene, St Quentin en Yvelines, France). The changed bases are underlined and only sequences of the forward oligonucleotides are given. The oligonucleotide (5′-3′) used for introducing K645,646,647N mutations (KN1 mutant) in pHM6-FERM was GGAAAGAAGAACCAAACAACAACACCAAAGCCACTG (16 cycles, 95°C, 30 seconds; 55°C, 1 minute; 68°C, 15 minutes). K824,825,830,831,832N mutations (KN2 mutant) were introduced in two steps. First, the following oligonucleotide, GAAAATATCTTTTTCTAACAACAACATCACATTGCAAAATAC was used to mutate lysines 830, 831 and 832 to asparagines (K830, 831, 832N, same conditions described above), generating an intermediate mutant. This intermediate mutant was used as a template to introduce the K824,825N mutations with the following oligonucleotide CCATGGAGGGAAACCAACAACATATCTTTTTCTAACAAC (same conditions described above) generating KN2 mutant. KN1-2 mutant was generated using oligonucleotide carrying K645,646,647N mutations described above and using pHM6-FERM KN2 as template. pHM6-PTPL1 KN1-2 construct was generated using the same approach as described above for pHM6-PTPL1 ΔFERM.
The in vitro expression construct pRSET-FERM encoding 6His-tagged FERM domain (aa 519-958) was created by subcloning BamHI-HindIII fragment from pHM6-FERM in pRSET vector (Invitrogen). pRSET-FERM KN1, KN2 and KN1-2 were created by the same approach using respective BamHI-HindIII fragments from pHM6-FERM-KN1, pHM6-FERM-KN2 and pHM6-FERM-KN1-2.
pCB6 Nt-Ezrin (aa 1-310) expressing VSV-G epitope tagged FERM domain of ezrin was a generous gift from M. Arpin (Institut Curie, Paris, France).
HeLa and COS1 cells were cultured in Dulbecco's Modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), 2.5 units/ml penicillin and 2.5 μg/ml streptomycin (Invitrogen). Cells were transfected either with the calcium phosphate precipitation method or using LipofectAMINE PLUSTM reagents (Invitrogen) according to the manufacturer.
COS1 cells were seeded 24 hours before transfection at a density of 125,000 cells per 35 mm dish. As indicated, cells were treated with 10 mM neomycin (Sigma) for 24 hours. Forty hours after transfection, cells were washed two times with ice cold PBS and lysed in fractionation buffer [10 mM PIPES, pH 6.8, 250 mM sucrose, 3 mM MgCl2, 120 mM KCl, 1 mM EGTA, 0.15% Triton X-100, 1mM Na3VO4, 25 mM NaF, 5 mM sodium pyrophosphate, 1mM PMSF and 0.02-0.04 trypsin inhibitor unit (TIU) of aprotinin] for 5 minutes at 4°C. The total lysate was centrifuged at 25,000 g for 15 minutes at 4°C yielding a supernatant containing depolymerized tubulin and monomeric actin, and a pellet containing polymerized cytoskeletons (microfilaments and intermediate filaments) resuspended in fractionation buffer (volume comparable to supernatant) and sonicated.
Equal volumes of each fraction were separated on SDS/polyacrylamide gel and electrotransferred onto PVDF membrane. Blots were then stained with Coomassie blue (R250, Sigma) to ensure that protein amounts were comparable. Membranes were blocked with TBST (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Tween 20) containing 5% nonfat milk and immunoblotted with anti-HA, anti-actin, anti-vimentin and anti-tubulin monoclonal or polyclonal antibodies. Primary antibodies were detected using horseradish peroxydase-conjugated goat anti-mouse IgG or anti-rabbit IgG, and Western Lighting (PerkinElmer Life Sciences, Courtaboeuf, France).
Twenty hours after transfection, cells were fixed in medium with 3.7% formaldehyde for 20 minutes and permeabilized in TBS buffer (50 mM Tris, pH 7.5, 150 mM NaCl) containing 0.2% Triton X-100 for 4 minutes. Cells were incubated with anti-HA mAb for 60 minutes at room temperature, washed and incubated with FITC-conjugated anti-mouse IgG for 30 minutes at room temperature. After final wash and mounting in Gel/Mount (Biomeda, Hayward, CA), cells were examined with a laser scanning confocal microscope (Bio-Rad Laboratories) using a 63×1.4 oil immersion objective or a DMR A microscope PL APO 63× oil immersion objective (1.32 NA) with appropriate filters (Leica) and images were recorded with a cooled CCD Micromax camera (1,300×1,030 pixels, RS; Princeton Instruments, Monmouth Junction, NJ).
35S-labeled PTPL1-FERM domain and its derived mutants, generated by the TnT in vitro transcription-coupled translation system (Promega) from pRSET-FERM vector, were incubated with Ni-NTA agarose column (Qiagen) for 1 hour at room temperature. After two washes with S buffer [50 mM NaH2PO4, 20 mM Tris, 300 mM NaCl, 5 mM imidazole, 1 mM PMSF and 0.02-0.04 trypsin inhibitor unit (TIU) of aprotinin, pH 8] and W buffer (20 mM Tris, 150 mM NaCl, 30 mM imidazole), recombinants proteins were eluted in the E buffer (20 mM Tris, 100 mM NaCl, 300 mM imidazole). 40,000 cpm of 35S-labeled PTPL1-FERM domain and its derived mutants were incubated for 2 hours at room temperature in 5 ml TBST, 3% BSA with PIP-Strips (Echelon Biosciences, Salt Lake City, Utah) pre-saturated with TBST, 3% BSA. After three washes with TBST, binding of FERM domain was quantified with a Fuji BAS1000 Bioimaging Analyzer (Raytest, Paris la Défense, France).
The FERM domain of PTPL1 is necessary and sufficient for apical localization of the enzyme
FERM domains are thought to direct proteins at the interface between cell membrane and cytoskeleton. We investigated the localization of PTPL1 and the role of its FERM domain in HeLa epithelial cells by expressing various HA-epitope tagged constructs (Fig.1A), the sizes of which had been confirmed by SDS-PAGE (Fig. 1B,C). Upon transient expression in HeLa cells, PTPL1 was found almost exclusively enriched apically at the plasma membrane and little was present in the cytoplasm (Fig. 2A). In addition, PTPL1 staining presented a co-localization with F-actin in microvilli and plasma membrane (Fig. 2C). In the case of strong overexpression, PTPL1 was found to induce apical bleb-like structure where it co-localized with F-actin (data not shown). We evaluated the specific role of the FERM domain in two ways by expressing constructs encoding either only the FERM domain (HA-FERM) or the full-length enzyme deleted of the FERM domain (HA-PTPL1 ΔFERM). The FERM domain, similarly to the full-length PTPL1, localized to the apical cell membrane and was enriched in microvilli (Fig. 2D-F). Furthermore, the FERM domain also induced cell extensions (Fig. 2D) as observed with the ezrin FERM domain (Martin et al., 1997). By contrast, PTPL1 ΔFERM was found mainly cytosolic with little accumulation at the plasma membrane or in microvilli (Fig. 2G-I). These results clearly demonstrate that the FERM domain is necessary and sufficient for apical localization of PTPL1.
Mutation of two potential PtdIns(4,5)P2-binding sites altered PTPL1 localization
FERM domains of ERM family proteins are known to bind transmembrane proteins such CD44, CD43, ICAM-1, -2, -3 and also PtdIns(4,5)P2 (for reviews, see Bretscher, 1999; Mangeat et al., 1999). Furthermore, ERM-CD44 binding is enhanced, in vitro, by PtdIns(4,5)P2 (Hirao et al., 1996). Recently, it has been shown that mutation of PtdIns(4,5)P2-binding motifs in the FERM domain of ezrin altered its intracellular localization (Barret et al., 2000). Sequence analysis of the PTPL1-FERM domain revealed the presence of two potential PtdIns(4,5)P2-binding sites (Fig. 3, bold bars). PtdIns(4,5)P2-binding motifs are characterized by a cluster of basic amino acids [e.g. KKXXXXXX(K/R)K in FERM domains of ERM]. The first potential PtdIns(4,5)P2-binding motif found in PTPL1-FERM domain, KEEPKKK, was located at amino acids 641-647 and the second, KKISFSKKK, at amino acids 824-832 (Fig. 3). In order to evaluate the role of such motifs in PTPL1 localization, several mutations of the lysine residues located in the presumed sites were introduced in the FERM domain. Lysines 645, 646 and 647 in the first site and lysines 824, 825, 830, 831, 832 in the second site were mutated to asparagines generating, respectively, KN1 and KN2 mutants. KN1-2 mutant was obtained by combining both sets of mutations.
As in HeLa cells, when expressed in COS1 fibroblasts PTPL1-FERM domain localized in apical microvilli and was able to induce membrane extensions in many cells (Fig. 4A). Such effect has also been reported with the ezrin FERM domain (Martin et al., 1997) and has been shown to be lost by preventing PtdIns(4,5)P2 binding (Barret et al., 2000). On the basis of this result, we evaluated the effect of mutations of potential PtdIns(4,5)P2-binding sites within the PTPL1-FERM domain on its cellular localization (cell extensions, membrane and microvilli). KN1 and KN2 mutants partly inhibited cell extension processes but had few effects on the localization of the FERM domain at the membrane and microvilli (Fig. 4B,C). However, combination of both sets of mutations strongly altered the cellular localization of the FERM domain, and KN1-2 mutant presented in most cells a diffuse cytosolic staining (Fig. 4D). It is important to note that in some cells the KN2 mutant also presented a localization comparable to the KN1-2 mutant (data not shown). Thus, both potential PtdIns(4,5)P2-binding sites seemed to cooperate for proper intracellular targeting of the FERM domain of PTPL1.
Because the most drastic effect on FERM domain localization was obtained by mutating both potential PtdIns(4,5)P2-binding motifs, KN1-2 mutations were secondarily introduced in the full-length enzyme to evaluate their roles on PTPL1 localization in HeLa cells. As shown with COS1 cells, KN1-2 mutations abrogated membrane and microvilli localization of FERM domain as well as cell extensions (compare Fig. 2D with Fig. 5A). Similarly, HA-epitope tagged PTPL1 KN1-2 localization was strongly altered compared with the wild-type protein, presenting a diffuse cytosolic staining around the nucleus (compare Fig. 2A with Fig. 5D). Such a mislocalization was comparable to that observed for the PTPL1 ΔFERM mutant (Fig. 2G). Similar effects were obtained in COS1 cells (data not shown).
In summary, these results demonstrated that the two potential PtdIns(4,5)P2-binding sites found within the PTPL1-FERM domain are crucial for FERM and PTPL1 localization. It finally suggested that PTPL1 localization could be regulated by phosphoinositide metabolism.
PTPL1-FERM domain binds PtdIns(4,5)P2 in vitro
To directly assess the PtdIns(4,5)P2 binding property of the FERM domain of PTPL1, we carried out protein-lipid overlays. In vitro labeled transcribed and translated 6His-tagged PTPL1-FERM domains (wt, KN1, KN2 and KN1-2) were used to probe nitrocellulose membrane containing various immobilized phosphoinositides (PIPs), after their purification on and elution from nickel-agarose columns. Fig. 6A shows such eluted proteins separated on SDS-PAGE and visualized by autoradiography. Using this assay, 6His-FERM domain was found to specifically bind 4′ and/or 5′ phosphorylated PIPs, that is, PtdIns(4)P, PtdIns(5)P, PtdIns(3,4)P2, PtdIns(3,5)P2, PtdIns(4,5)P2 and PtdIns(3,4,5)P3 (Fig. 6B). A very weak interaction with PtdIns(3,4,5)P3 was sometimes observed, whereas no interaction was detected with nonphosphorylated phosphatidylinositol and other control phospholipids (PA, PS, LPA, LPC, PE, PE, S1P). The KN1 mutation slightly affects PIPs binding of the 6His-FERM domain (Fig. 6B). This effect was more pronounced with the KN2 mutation (Fig. 6B). However, KN1-2 mutation abolishes the binding of 6His-FERM domain to immobilized PIPs by 70% to 80% (Fig. 6B).
These results clearly demonstrate that the FERM domain of PTPL1 directly binds PtdIns(4,5)P2 in vitro through the two binding sites located between amino acids 641-647 and 824-832.
Subcellular distribution of PTPL1 correlates with PtdIns(4,5)P2 binding
Having established that the FERM domain of PTPL1 binds PtdIns(4,5)P2 in vitro, the subcellular distributions of wild-type, KN1, KN2, KN1-2 HA-tagged FERM domains, as well as wild-type, KN1-2, ΔFERM HA-tagged PTPL1 were compared, upon cell fractionation of COS1 cells, in order to provide biochemical evidence for the differences in localization observed by immunofluorescence. Following cell fractionation, wild-type FERM domain was found to be strongly associated with the Triton X-100 insoluble fraction containing polymerized actin (F-actin) and vimentin (Fig. 7A). The KN1 mutant displayed a comparable distribution even if the amount present in the soluble fraction, containing depolymerized tubulin and actin (G-actin), was more significant than that observed with the wild-type FERM (Fig. 7A). By contrast, KN2 and KN1-2 mutants were only found to be associated with the soluble fraction (Fig. 7A) in accordance with the poor affinity of these proteins for PtdIns(4,5)P2 (Fig. 6B). Following the same approach, effects of KN1-2 mutations and FERM deletion on full-length PTPL1 subcellular distribution were analyzed. As for its FERM domain, wild-type PTPL1 showed a marked association with the cytoskeletal fraction that is abolished in KN1-2 or ΔFERM mutants (Fig. 7B). 35% of total HA-PTPL1 was associated with Triton X-100 insoluble fraction versus 3.6% and 1.9% for HA-PTPL1 KN1-2 and HA-PTPL1 ΔFERM, respectively (Fig. 8). The subcellular distribution of full-length PTPL1 did not require its catalytic activity since the localization of an inactive mutant (C2389S) was indistinguishable from that of the native enzyme (data not shown). Our data demonstrated that association of PTPL1 with Triton X-100 insoluble fraction depended on its FERM domain and on the integrity of at least the second PtdIns(4,5)P2-binding site.
To confirm that the association of the wild-type FERM domain and PTPL1 with the particulate fraction was a result of PtdIns(4,5)P2 binding, we performed cell fractionation experiments after incubation of cells with neomycin (10 mM) known to bind, and thus to mask, PtdIns(4,5)P2 polar heads (Wang et al., 1984). It was recently shown that ERM proteins were mostly recovered in the soluble fraction after addition of neomycin in L cells (Yonemura et al., 2002). We used as a control the FERM domain of ezrin known to bind PtdIns(4,5)P2 (Niggli et al., 1995; Barret et al., 2000) and, as expected, we found that neomycin treatment strongly reduced the association of the VSV-G epitope-tagged ezrin-FERM domain with cytoskeletal fraction (Fig. 9A). First, we demonstrated that neomycin severely decreases the in vitro interaction of the PTPL1-FERM domain with PIPs on protein-lipid overlays (Fig. 9B). We next observed that such a cell treatment also strongly reduced the presence of the wild-type PTPL1-FERM domain associated with the particulate fraction without having any effects on the repartition of KN1-2 mutant (Fig. 9C). In the same manner, we showed that association of wild-type PTPL1 with the cytoskeletal fraction was inhibited by 50% with addition of neomycin (Figs 8 and 9D). Moreover, subcellular distribution of KN1-2 or ΔFERM mutants was not affected by such a treatment (Fig. 9D). Those data strongly support the necessity of PtdIns(4,5)P2 binding for cytoskeletal association of PTPL1 through motifs within its FERM domain.
The knowledge of the PTP diversity has been addressed in the past 15 years after characterization of the first family member (Tonks et al., 1988). Most attention was originally drawn on the understanding of basic enzymatic function by searching for their specific substrates. The substrate-trapping approach, developed by Tonks's laboratory, constituted a powerful and valuable approach to identify many PTP substrates and discover pathways in which PTPs are involved (Garton et al., 1996; Flint et al., 1997). However, it soon became clear that the diversity of their functions could be a consequence of the regulation of their activity in a particular subcellular environment. In fact, the substrate specificity of cytosolic PTPs is thought to depend upon their intracellular targeting and in some cases on an intrinsic property carried by their catalytic domain (for reviews, see Mauro and Dixon, 1994; Fischer, 1999; Tonks and Neel, 2001). Therefore, an important step in the understanding of a particular enzyme functions is the identification of its localization and regulatory mechanisms in different physiological contexts. In this study, we have defined human PTPL1 localization and demonstrated that phosphoinositides are playing a key role in the regulation of its membrane addressing.
PTPL1 belongs to the FERM-domain-containing PTP superfamily comprising several mammalian enzymes PTPH1, PTPMEG, PTP36/Pez and PTPD1/PTP-RL10 as well as the recently identified worm PTP-FERM. In this report, we demonstrate by immunofluorescence using different tagged-constructs that the FERM domain of human PTPL1 is necessary and sufficient to target the enzyme at the apical face of cell membrane and in microvilli in HeLa epithelial cells. In accordance with our findings, PTP36, PTPH1, and PTPMEG have been reported to associate with the cell membrane (Ogata et al., 1999; Gjorloff-Wingren et al., 2000). The FERM domains of PTP-BL, the mouse homologue of PTPL1, and of the worm PTP-FERM were shown to be sufficient for their submembranous distribution (Cuppen et al., 1999; Uchida et al., 2002). Furthermore, deletion of the FERM domain from PTPH1, PTPMEG and PTP-FERM altered their intracellular targeting (Gjorloff-Wingren et al., 2000; Uchida et al., 2002). However, it does not seem to be a general feature for all family members since PTPD1 whose FERM domain is known to interact with KIF1C, a kinesin-like protein, was shown to be mainly localized in the cytoplasm (Dorner et al., 1998; Cuppen et al., 1999), whereas FERM domain of Pez was involved in the nuclear localization of this enzyme in HEK 293 cells at low density (Wadham et al., 2000).
The membraneous PTPL1 localization raises the question of its association with the cell cytoskeleton. By COS1 cell fractionation, we show that the FERM domain of PTPL1 is mainly recovered in the particulate fraction containing polymerized actin and vimentin. Moreover 35% of total full-length PTPL1 is also found to be associated with cytoskeletal fraction in a FERM-domain-dependent manner. Having established that the FERM domain was involved in PTPL1 intracellular targeting and required for its association with cytoskeletal fraction, we next investigated how its localization could be regulated in order to identify physiological mechanisms controlling PTPL1 activity.
FERM domains of ERM proteins, composed of three subdomains (A, B and C), are known to bind transmembrane proteins such as CD44, CD43, ICAM-1, -3 and also PtdIns(4,5)P2 (Hamada et al., 2000) (for a review, see Bretscher, 1999). Ezrin binding to CD44 is increased by PtdIns(4,5)P2 in vitro (Hirao et al., 1996). Recently, it has been demonstrated that ezrin localization was altered by mutation of PtdIns(4,5)P2-binding motifs found in subdomains A and C (Barret et al., 2000). PTPL1 sequence alignment with ezrin points to two potential PtdIns(4,5)P2-binding sites in its FERM domain sequence in the same A and C subdomains (Fig. 3). The first site located at amino acids 641-647 corresponds to a KXXXKKK motif and resembles the consensus sequence in villin, (K/R)XXXX(K/R)(K/R). The second site, KKXXXXKKK, located at amino acids 824-832, is similar but not identical to the first consensus sequence in ERM, KKXXXXXX(K/R)K. By introducing mutations in these sites, we presently found that lysine to asparagine mutations of both sites (K644, 645, 646, 824, 825, 830, 831, 832N) altered the subcellular localization of the PTPL1-FERM domain as well as the full-length enzyme. Furthermore PTPL1 KN1-2 localization was similar to that observed with PTPL1 ΔFERM. In addition, we demonstrated that association of the FERM domain or full-length PTPL1 with cytoskeletal fraction was abrogated by mutation of both potential PtdIns(4,5)P2-binding motifs or FERM domain deletion. In fact, only 4% and 2% of PTPL1 KN1-2 and PTPL1 ΔFERM mutants, respectively, remained associated to the particulate fraction, whereas 35% of total wild-type PTPL1 is found in that fraction. Our results altogether indicate that the two domains, located in regions 641-647 and 824-832, were responsible for PTPL1-FERM domain, and therefore for PTPL1, localization and association with the cytoskeletal fraction.
By similarity to what has been described for ezrin (Barret et al., 2000), we hypothesized that these two domains do cooperate in PtdIns(4,5)P2 binding and we investigated the direct binding of PTPL1-FERM domain to PtdIns(4,5)P2 in vitro. Our first attempt using FERM domains expressed in bacteria failed to give us results because of the strong insolubility of this protein in absence of PtdIns(4,5)P2 (data not shown). However, labeled 6His-epitope tagged PTPL1-FERM domains obtained by coupled transcription/translation in vitro permitted the purification of low amounts of tracer proteins that remain soluble at these concentrations and thus allowed us to probe nitrocellulose membrane containing immobilized PIPs. Using such protein-lipid overlays, we showed that the FERM domain of PTPL1 binds specifically 4′ and/or 5′ phosphorylated PIPs. KN2 mutation and, to a lesser extent, KN1 mutation were found to affect PIPs binding which is almost completely abolished by combination of both mutations (KN1-2). This result clearly demonstrates that the two suspected PtdIns(4,5)P2-binding sites are functional and cooperate for PIPs binding in vitro. The in vitro ERM-FERM domain binding selectivity for PIPs is largely unknown. Among PIPs bound by the FERM domain of PTPL1, PtdIns(4,5)P2 and PtdIns(4)P are the most abundant in mammal cells (for a review, see Payrastre et al., 2001). PtdIns(5)P, which represents with PtdIns(3)P around 0.25% of cellular phosphoinositides, is absent in certain cell lines, among them COS1 cells (Shisheva, 2001). As for higher 3′ phosphorylated species known to be localized at the plasma membrane, as PtdIns(3,4)P2 and PtdIns(3,4,5)P3, their rates are generally low or absent and are strongly regulated by extracellular stimuli (Payrastre et al., 2001). Moreover, we showed that serum starvation of COS1 cells, known to reduce PI 3-kinase activity and thus PtdIns(3,4)P2 and PtdIns(3,4,5)P3 levels, does not affect the localization and subcellular distribution of PTPL1 (data not shown). The nature of PIPs could affect binding affinity and selectivity. Our preliminary results on different protein-lipid overlays show that the binding of PTPL1-FERM domain to natural PtdIns(4,5)P2 is stronger than to synthetic PtdIns(4,5)P2, which displays shorter fatty acid chains than in natural PtdIns(4,5)P2 (data not shown). Data on PIPs binding, associated to the prevalent rate of PtdIns(4,5)P2 and its proper cellular localization, strongly favor PtdIns(4,5)P2 being responsible for PTPL1 localization and subcellular distribution.
Our data suggest that the association of PTPL1 with the particulate fraction depends on PtdIns(4,5)P2 binding and that at least the second motif within its FERM domain is essential. This first evidence was strongly reinforced by the demonstration that COS1 cell treatment with neomycin (10 mM), a polycation that binds to the polar heads of PtdIns(4,5)P2 in the cell membrane and inhibits PTPL1-FERM domain interaction with PtdIns(4,5)P2 in overlay experiments, had similar effect on PTPL1 subcellular distribution as did the KN1-2 mutations. Indeed, the association of PTPL1 with the cytoskeletal fraction was reduced by 50% after neomycin treatment.
Our studies on PTPL1 bring the first argument for PtdIns(4,5)P2 binding being primarily responsible for PTP cellular targeting and subcellular distribution. It is interesting to note that among PTPs containing a FERM domain and known to localize at the plasma membrane PTPH1 and PTP-FERM also possess two potential PtdIns(4,5)P2 binding sites situated in same regions to those of PTPL1 (Fig. 3). However, PTPL1 intracellular localization is much more limited than that of PtdIns(4,5)P2, and the PH domains of PLCδ1 and oxysterol-binding protein (OSBP) known to specifically interact with PtdIns(4,5)P2 in vitro displayed different localizations when expressed in cells (for a review, see Balla et al., 2000). These data therefore suggest the existence of additional anchoring proteins that might contribute to membrane localization of PtdIns(4,5)P2-binding domains by restricting and stabilizing their recruitment into membrane specific compartments. We showed that the FERM domain as well as PTPL1 co-localized with F-actin at plasma membrane (cortical actin) and in microvilli. PTPMEG and PTP36 were reported to associate with cytoskeletal fraction (Gu and Marjerus, 1996; Ogata et al., 1999; Aoyama et al., 1999). Furthermore, many proteins interacting with PDZ domains of human PTPL1 or mouse PTP-BL have been shown to be associated with cytoskeleton such as PARG1, a Rho-GAP protein (Saras et al., 1997), ZRP-1, a zyxin-related protein (Murthy et al., 1999), or RIL, a LIM-domain-containing protein co-localizing with F-actin (Cuppen et al., 1998) suggesting a potential role of PTPL1 in the dynamic of actin cytoskeleton. It has recently been shown that PTP-BL associates with midbody during the cell cycle and is involved in cytokinesis in HeLa cells (Herrmann et al., 2003). In this study, the FERM domain of PTP-BL was found to co-localize with F-actin at the contractile ring and to bind, directly or indirectly, F-actin in vitro. Moreover, the FERM domain of ezrin has been shown to bind G-actin as well as F-actin in vitro (Roy et al., 1997). Altogether those data suggest that the association of PTPL1 with the particulate fraction could be due to a direct or indirect binding to the actin cytoskeleton through its FERM domain. However, depletion of F-actin by treatment of COS1 cells with latrunculin A does not affect the subcellular distribution of PTPL1-FERM domain or full-length enzyme (data not shown). This discrepancy in PTPL1/PTP-BL localization could be explained by the PTPL1 gene structure, which was recently described and revealed the existence of 47 exons raising many isoform possibilities (Ensembl gene ID, ENSG00000163629). Indeed, Herrmann et al. (Herrmann et al., 2003) demonstrated differential localizations of some different isoforms of PTP-BL during the cell cycle.
In this report, we have discovered two main characteristics of human PTPL1 localization. First of all, we outlined the crucial role of the FERM domain for its submembraneous addressing. Second, we demonstrated that its association with the particulate fraction is a phenomenon driven by PtdIns(4,5)P2 interaction with specific binding motifs within this domain. Our major goal is now to reconcile this regulated subcellular localization with the potential functions of this enzyme in different cellular contexts. FERM-domain deletion of PTPH1 was shown to block its inhibitory effect on T-cell antigen receptor signaling pointing that a proper localization is crucial for its physiologic effect (Han et al., 2000). In MCF7 human breast cancer cells, we previously showed that PTPL1 was involved in the inhibition of IGF-1 survival activity by blocking PI3-kinase (PI3-K) pathway and this inhibitory action occurs at early steps of IGF-I signaling (Freiss et al., 1998; Bompard et al., 2002). PTP-BL has recently been shown to localize in lipid rafts in mouse embryo and cortical neurons (Palmer et al., 2002). Futhermore, PtdIns(4,5)P2 has been demonstrated to accumulate within such lipid microdomains (Laux et al., 2000). The association of PTPL1 with the particulate fraction, not driven by F-actin binding, could thus result from its presence in lipid rafts since these domains are recovered in Triton X-100 insoluble fraction. In MCF7 cells, the IGF-I receptor (IGF-IR), as well as PI3-K, have been shown to transiently associate with lipid rafts after IGF-I stimulation (Manes et al., 1999). This transient localization could allow PTPL1, associated with lipid rafts, to inhibit the IGF-I survival pathway in human breast tumors where it could represent an interesting new therapeutic tool.
The authors wish to thank Drs Travo and Cau, from the Integrated Imaging Facility of the CRBM (Montpellier), and Nicole Lautrédou from the Imaging Service of the IGH (Montpellier) for their support. We are grateful to Dr Joubert for fruitful discussions and for critical reading of the article. This work was supported by Inserm and by the Association pour la Recherche sur le Cancer (ARC, grants 5405 and 5761). G.B. was supported by a fellowship from the ARC.
↵ ‡ Present address: School of Biosciences, University of Birmingham, Birmingham, UK
- Accepted March 3, 2003.
- © The Company of Biologists Limited 2003