Differential localization patterns of myristoylated and nonmyristoylated c-Src proteins in interphase and mitotic c-Src overexpresser cells

The transforming protein of Rous sarcoma virus, pp60^v-src, and its cellular homolog, pp60^c-src, are cotranslationally myristoylated at the amino-terminal glycine 2 residue during protein synthesis (Buss et al., 1984; Buss and Sefton, 1985; Wilcox et al., 1987; Deichaite et al., 1988). Most pp60^v-src molecules are so modified although it has been estimated that as much as 16% of pp60^v-src may not be myristoylated (Buss and Sefton, 1985). For both proteins, myristoylation is required for association with the plasma membrane (Buss et al., 1986; Schuh and Brugge, 1988; Reynolds et al., 1989) but not all myristoylated proteins are associated with the plasma membrane (for review, Sefton and Buss, 1987; McIlhinney, 1990). Independent domains of Src may cooperate with myristoylation to specify association with distinctive cellular membranes (Kaplan et al.,1990). In particular, significant fractions of myristoylated pp60^c-src and pp60^v-src have been found to associate with perinuclear membranes (Resh and Erikson, 1985) and myristoylated, yet cytosolic, variants of pp60^v-src have been described (Garber et al., 1985). On the other hand, wildtype myristoylated pp60^c-src overexpressed in NIH 3T3 cells has been shown to partition between the plasma membrane and the centrosomal area in interphase cells and colocalize with endocytosed concanavalin A (ConA) at all stages of the ConA-induced endocytotic process (David-Pfeuty and Nouvian-Dooghe, 1990). More recently, pp60^c-src was also shown to be enriched in a population of late endosomes in Rat-1 c-Src overexpresser cells (Kaplan et al., 1992) and in PC12 synaptic vesicles (Lindstedt et al., 1992).

Many v-src mutations that prevent myristoylation and plasma membrane association also abrograte v-src transforming activity (for review, Jove and Hanafusa, 1987); this supports the hypothesis that important cellular targets of Src are located at the plasma membrane. However, nonmyristoylated, transformation-defective, cytosolic v-Src proteins still have mitogenic properties, promoting growth of cells to high densities and growth of some normally nondividing cells of neuronal origin (Kamps et al., 1986; Calothy et al., 1987). Myristoylation and plasma membrane association are not absolutely required for src transforming activity: two nonmyristoylated Src proteins encoded by recovered avian sarcoma viruses rASV157 and rASV1702 fractionate as soluble cytosolic enzymes yet are transformation-competent (Krueger et al., 1982, 1984). In this case, the transforming activity of these unusual mutants depends on the presence of env signal sequences at the amino termini of the retrovirus-encoded fusion proteins (Garber and Hanafusa, 1987). Moreover, a slow, time-dependent cellular transformation of cells infected with temperature-sensitive mutants of RSV has been reported to occur when the infected cells are continuously grown at the nonpermissive temperature, under conditions in which the ts-pp60^v-src is impaired in its capacity to associate with the plasma membrane and, consequently, accumulates in the cytoplasm, particularly in the centrosomal area (David-Pfeuty and Nouvian-Dooghe, 1992). The activities of cytoplasmic (either myristoylated or nonmyristoylated) mutants could reflect non-specific low-level Src-mediated phosphorylation of plasma membrane-associated targets. Alternatively, it may indicate that Src also acts at cellular sites distinct from the plasma membrane.

The observation that the activity of c-Src is transiently stimulated during mitosis (Chackalaparampil and Shalloway, 1988), partly as an indirect result of phosphorylation by p34^cdc2 or the related kinase (Morgan et al., 1989; Shenoy et al., 1989, 1992; Bagrodia et al., 1991; Kaech et al., 1991), suggests that c-Src may have special functions in this phase of the cell cycle (for review, see Taylor and Shalloway, 1993). This observation may be related to the fact that wt pp60^c-src in NIH 3T3 overexpresser cells delocalizes from the plasma membrane in late G2 to condense around the two separating centriole pairs and then to disperse in a diffuse and patchy way in the cytoplasm throughout mitosis (David-Pfeuty and Nouvian-Dooghe, 1990). These considerations prompted us to perform a detailed comparison, using immunofluorescence techniques, of the subcellular localization of wild-type (wt) and myristoylation-defective (myr⁻) c-Src proteins throughout the cell cycle both to determine which aspects of the cell cycledependent relocalization of c-Src depend on myristoylation and to get an insight into the behavior of the potential, small subfraction of nonmyristoylated Src molecules present in normal cells. This study revealed unexpected localization properties of myr⁻ c-Src that differed from those of wt csrc: (1) although primarily cytoplasmic, myr⁻ c-Src exhibits detectable nuclear localization during the G1 and S phases of the cycle; (2) it translocates to the nucleus during late G2 or early prophase; and (3) it concentrates in the spindle apparatus and interchromosomal space during mitosis. These findings raise the possibility that nonmyristoylated forms of other normally myristoylated proteins might also be targeted to the nucleus.

Plasmid-transfected NIH 3T3 cells that overexpress wild-type chicken c-Src (NIH(pMc-src/focus)B cells) have been previously described (Johnson et al., 1985). To generate cells expressing the myr⁻ mutant, NIH 3T3 cells were cotransfected with pcLN, a chimeric plasmid containing Moloney murine leukemia virus long terminal repeats and the Src coding sequence from plasmid pRSVc-SrcLN (Schuh and Brugge, 1988), and G418-resistance plasmid pSV2neo (Southern and Berg, 1982) and subjected to G418 selection and cloning as described (Kmiecik and Shalloway, 1987). A mass-culture [NIH(pcLN/ pSV2neo/MC)C] derived from about 50-100 G418-resistant colonies was used.

Cells were cultured in 35 mm dishes on 22 mm2 coverslips in Dulbecco’s modified Eagle’s medium containing 5% newborn calf serum and antibiotics in 10% CO2, at 37°C, for at least two days before immunofluorescence observation. The average cell density was 2 ×103 to 104 cells cm⁻².

Cells were synchronized at the G₁/S boundary by a thymidineaphidicolin double-block; cells were grown in normal medium for two days, then sequentially incubated with 2.5 mM thymidine (16 h), normal medium (8 h) and 5 μg/ml aphidicolin (16 h). Immunofluorescence with anti-tubulin antibody and DAPI (4′,6′diamidino-2-phenylindole) showed that G2 cells started to appear 6 h after release from the double-block and were most prevalent 7 h after release. Mitotic cells were most prevalent 7-8 h after release; a second peak of mitotic cells appeared 27-29 h after release.

Monoclonal antibody mAb327 (Lipsich et al., 1983) was from Clinisciences (Paris, France). Visualization of intracellular cytoskeletal structures was achieved using rat monoclonal antitubulin (Biosys, France) for microtubules and nitrobenzoxadiazole (NBD)-phallacidin (Molecular Probes, Inc., Beaverton, OR) for microfilaments.

Anti-p60 polyclonal rabbit antiserum (Resh and Erikson, 1985), monoclonal antibody GD11 (Parsons et al., 1984), human autoantibody to lamin B (Guilly et al., 1987) and anti-clathrin polyclonal rabbit antiserum were generous gifts from Marilyn Resh, Sarah Parsons, Françoise Danon and Paul-Henri Mangeat, respectively. Secondary antibodies were goat IgG Fab fragments from antibodies directed against either mouse or rat IgG. (The anti-rat goat IgG was preabsorbed against a mouse IgG column to eliminate cross-reactivity; a reciprocal preabsorption was used for the antimouse goat IgG.) These fragments were conjugated with either fluorescein isothiocyanate (FITC; Interchim, France) or tetramethyl rhodamine isothiocyanate (TRITC; Interchim, France). These secondary antibodies gave essentially undetectable back-ground fluorescence.

Singleor double-fluorescence cell labeling was performed as described (David-Pfeuty and Singer, 1980) following fixation with 3% formaldehyde and subsequent permeabilization by treatment with 0.1% Triton X-100 at room temperature. When specified, cells were pre-Tritonized by treatment with PHEM buffer (0.1% Triton X-100, 45 mM PIPES, 45 mM HEPES, 10 mM EGTA, 5 mM MgCl₂, 1 mM PMSF, pH 6.9) for 30-45 s before fixation (Bailly et al., 1989). Fluorescence microscopy was performed with a Leitz microscope equipped with fluorescein, rhodamine and DAPI filters using a ×40 oil objective. Photographs were taken with Kodak TMax 400 film.

NIH 3T3-derived cells, which express wild-type and nonmyristoylated chicken c-Src from transfected chimeric plasmids containing Moloney murine leukemia virus long terminal repeats, and wt and mutant src genes, were used for all studies. They expressed Src at levels 10-to 20-fold above the level of endogenous c-Src (data not shown), so almost all immunofluorescently visualized protein was plasmid-expressed. The mutant gene used for expression of myr⁻ Src contained codons for four additional residues at the amino terminus of the protein. It encodes the aminoterminal sequence Met-Ala-Ala-Ala-Met-Gly-… (where the underline denotes the first two amino acids encoded by the wt src gene). This mutant has been studied previously in chicken embryo fibroblasts by Schuh and Brugge (1988), who found, as expected, that it was not localized to the plasma membrane but had normal specific protein-tyrosine kinase activity. This was confirmed in NIH 3T3 cells (data not shown). Anti-Src monoclonal antibodies mAb327 (Lipsich et al., 1983) and, occasionally, anti-p60 polyclonal rabbit antiserum (Resh and Erikson, 1985) or mAb GD11 (Parsons et al., 1984), which react with a broad range of v-Src and chicken, mouse and human c-Src variants, were used in these studies.

In asynchronous wt c-Src overexpresser cells, mAb327 revealed the three characteristic, previously documented (David-Pfeuty and Nouvian-Dooghe, 1990) distributions: (i) uniformly dispersed at the inner face of the plasma membrane (open arrows, Fig. 1C); (ii) inside a dense spot situated at the focal point of the interphasic microtubules in the centriolar area (thin arrows, Fig. 1C and D); (iii) inside a cluster of patches surrounding the nucleus and embedded within a microtubule meshwork (arrowheads, Fig. 1C and D). In G2 cells, the mAb327 staining typically remained concentrated in part near the two separating centriole pairs (thin arrows, Fig. 1H and I).

A very different pp60^c-src localization pattern was observed in asynchronous myr⁻ c-Src overexpresser cells. Here, mAb327 gave a rather intense and uniformly distributed cytoplasmic staining (arrowheads, Fig. 1A) identical to that described by Reynolds et al. (1989). However, we also detected a nuclear staining (arrows, Fig. 1A), which was low (compared to the cytoplasmic staining) in the majority of the cells, but conspicuous in a minor cell population (∼2%). Triple labeling of the cells with anti-Src, anti-tubulin antibodies and DAPI (large arrows, Fig. 1E-G) clearly indicates that prominent nuclear staining occurs in late G2 cells (these can be unambiguously distinguished by the presence of the two centriole pairs brightly labeled with the anti-tubulin antibodies (thin arrows, Fig. 1F) and the appearance of the typically punctate chromosomal DAPI staining (large arrow, Fig. 1G) that signals the beginning of chromosome condensation). In addition, anti-Src staining concentrated in the spindle apparatus of mitotic cells (arrowheads, Fig. 1E-G).

These first observations indicated that pp60^c-src/myr− is located predominantly in the cytoplasm of interphase cells but that it translocates to the nucleus, presumably in G2 phase.

myr⁻ and wt c-Src overexpresser cells were synchronized at the G1/S boundary by thymidine-aphidicolin doubleblock (see Materials and Methods) and their distributions were observed at various times (tR) after release from the double-block. Unexpectedly, at tR = 0 (Fig. 2A and B) a very intense nuclear staining (generally partly excluded from the nucleolar region; small arrows), largely supplanted the cytoplasmic staining in the majority of the myr⁻ c-Src cells. The percentage of cells exhibiting this feature varied between 70 and 100% from one experiment to the other and decreased as cells were released from the block at later times. This abundant nuclear accumulation of pp60^c-src/myr− did not reappear at the following G1/S phase transition, which occurred about 20 h later. We infer that pp60^c-src/myr− collects intensely in the nucleus during the abnormally long arrest in G1 phase caused by thymidine-aphidicolin doubleblock but not in normal G1 phases. A similar, abnormal, myr⁻ c-Src nuclear accumulation also occurred during arrest in G0 induced by 48 h serum starvation; no nuclear accumulation of wt c-Src was observed after release from a double-block (Fig. 2E and F) or during a G0 arrest.

The strong nuclear concentration of pp60^c-src/myr− did not prevent passage into G2, which started approximately 7 h after release from the double-block. A natural control indicating that the observed nuclear accumulation of pp60^c-src/myr− is not an artifact was provided by the observation that anti-Src staining was totally excluded from the nucleus, but not from the cytoplasm, of confluent cells (which do not progress into G 2 after release from the block; Fig. 2C and D).

The first round of mitosis occurred 7 to 9 h after release from the thymidine-aphidicolin double-block. During the following G1 and S phases, that is between 9 and 27 h after release, mAb327 staining in wt and myr⁻ c-Src overexpresser cells did not differ significantly from that exhibited by unsynchronized cell populations (cf. Figs 1A to 3A, and Figs 1C to 3E).

Brief permeabilization of the cells with Triton X-100 for 30-45 s (pre-Tritonization) before fixation (see Materials and Methods) washed out almost all of the cytoplasmic mAb327 staining but preserved the nuclear one in subconfluent cultures (arrows, Fig. 3C). Such treatment did not severely alter the interphasic microtubule network (Fig. 3D), or the organization of actin into microfilament bundles (not shown). This experiment showed that the cytoplasmic pool of pp60^c-src/myr− does not associate tightly with structures that are insoluble in the presence of nonionic detergents. In contrast, it showed that the nuclear myr⁻ c-Src does interact with nuclear structures that are insoluble in Triton X-100.

For comparison, the effect of brief pre-Tritonization before fixation on wt c-Src overexpresser cells is shown in Fig. 3E-H. When the cells were fixed without pre-Tritonization, c-Src partitioned between the centriolar area (thin arrows, Fig. 3E), the perinuclear region (arrowheads, Fig. 3E) and, uniformly at the inner face of the plasma membrane (open arrow, Fig. 3E). Strikingly different c-Src features appeared when the cells were permeabilized with Triton X-100 before fixation (Fig. 3G): a residual pericentriolar concentration of c-Src occasionally persisted (thin arrows, Fig. 3G), but most often, the perinuclear pp60^c-src patches were no longer present. In addition, the previously uniform labeling of c-Src at the inner cell surface was replaced by a peculiar pattern of a multitude of rounded patches or vesicles. The relatively large size of these patches and double-labeling experiments with anti-Src and anti-clathrin (the major coat protein of coated pits) antibodies indicated that these structures were not coated pits (not shown). They could represent early endosomes. We stress that pre-Tritonizing the wt c-Src overexpresser cells did not reveal any nuclear c-Src staining.

A second peak of G2 and mitotic cells appeared 27 to 29 h after release from the thymidine-aphidicolin doubleblock. At tR = 27 h, a significant population of late G2 cells (up to 25%) were present. These were characterized by the presence of centriole pairs brightly labeled with anti-tubulin antibodies (arrows, Fig. 4J), the appearance of a typically punctate DAPI staining pattern (Fig. 4G and K) and prominent myr⁻ c-Src nuclear staining (Fig. 4E and I). As well as this population of cells that were clearly in late G2 or already in early prophase, a significant number of cells (between 40 and 50%) were observed in which the cytoplasmic myr⁻ c-Src was not uniformly distributed but was diffusely concentrated around the nucleus (arrow, Fig. 4A) or in which myr⁻ c-Src appeared to accumulate within the dense microtubule meshwork surrounding the nucleus (arrowheads, Fig. 4A and B).

In the late G2 cells, the conspicuous mAb327 nuclear staining was not seen in the early condensing chromosomes and instead filled the interchromosomal space, which was unstained by DAPI (small arrows, Fig. 4E, G, I and K). Phase-contrast microscopy (Fig. 4D, H and L) indicated that the nuclear accumulation of pp60^c-src/myr− occurred before clear evidence of nuclear membrane breakdown and before chromosomes became clearly distinguishable in phase contrast. As shown by double-immunolabeling of the cells with anti-Src and anti-lamin B antibodies (Fig. 4F), it also occurred before nuclear lamina disassembly.

Treatment of these synchronized cells, starting at tR = 25 h, with 10 μg/ml of nocadozole for 90 min did not prevent the translocation of pp60^c-src/myr− to the nucleus in late G2 or early prophase cells (arrowheads, Fig. 5C and D). This suggests that the interphasic microtubules are not involved in the late G2 phase-dependent translocation of myr⁻ c-Src from the cytoplasm to the nucleus. In addition, nocadozole treatment had no effect on the cytoplasmic distribution of myr⁻ c-Src (cf. Fig. 5A and C). In contrast, disruption of the interphasic microtubules by nocadozole treatment of wt c-Src overexpresser cells (Fig. 5E-H) induced a redistribution of pp60^c-src into patches that were scattered through the cytoplasm (Fig. 5G) as previously reported.

A chronological sequence of the variations in wt and myr⁻ c-Src localization during progression through mitosis was reconstructed from observations on selected cells (from cell populations 27 to 29 h after double-block release), whose positions in the cycle were identified by analysis of their tubulin and DAPI staining patterns. We found that the strikingly contrasting localization patterns of overexpressed pp60^c-src and pp60^c-src/myr− still persist during mitosis. At the end of the G2 phase, immediately before nuclear membrane breakdown, pp60^c-src-containing patches typically condense around the two centriole pairs, which are now symmetrically located with respect to the nucleus (arrows, Fig. 6A and B). c-Src appears to be clearly excluded from the nucleus at this time. This contrasts with myr⁻ c-Src, which is predominantly intranuclear. However, myr⁻ c-Src also starts, slightly but detectably, to accumulate at the level of the two diplosomes, which are freed from the depolymerized interphasic microtubules (arrows, Fig. 6D-F).

In prometaphase, during the process of chromosome rearrangement orchestrated by the complex dynamics of the spindle apparatus, the wt pp60^c-src-containing patches, which were previously concentrated around the two centriole pairs, partly dissolve from the centriole pairs, which now serve as spindle poles (arrows, Fig. 6J and K). In contrast, association of pp60^c-src/myr− with the spindle poles is greatly increased (large arrows, Fig. 6G, H, M and N). pp60^c-src/myr− continues to fill the interchromosomal space (thin arrows, Fig. 6G and I).

The features observed in prometaphase became even more apparent in metaphase: wt pp60^c-src was mostly excluded from the space occupied by the components of the spindle apparatus and metaphase plate (Fig. 7a-c and gi) although faint staining of the spindle fibers and spindle poles was occasionally observed (thin arrows and arrowheads, respectively, Fig. 7a, b, g and h). In contrast, all the components of the spindle apparatus, spindle poles and spindle fibers (arrowheads and thin arrows, respectively, Fig. 7d, e, j and k), the residual interchromosomal space and the contour of the metaphase plate (Fig. 7d and j), were heavily labeled by pp60^c-src/myr−.

In anaphase, wt pp60^c-src is diffusely distributed through the cytoplasm (Fig. 7m) whereas pp60^c-src/myr− remains preferentially concentrated at the spindle poles (small arrows, Fig. 7p), along the interpolar microtubules, and in the midzone separating the two sets of highly condensed chromosomes (open arrow, Fig. 7p).

In early telophase (Fig. 8A-F) wt pp60^c-src started becoming concentrated again in the centrosomal area (arrowheads, Fig. 8A and B) and at the cell surface of the two parting daughter cells (thin arrow, Fig. 8A); some pp60^c-src was still distributed diffusely through the cytoplasm (open arrow, Fig. 8A). In contrast, the pp60^c-src/myr−-associated fluorescence reappeared strongly and uniformly throughout the cytoplasm (open arrow, Fig. 8D).

In late telophase (Fig. 8G-L), the residual diffuse cytoplasmic pp60^c-src-associated fluorescence vanished and the two main areas of pp60^c-src distribution in the separated cells were at the inner cell surface and in the centrosomal area (long arrows and black arrowhead, respectively, Fig. 8G and H). pp60^c-src was also concentrated at intercellular contact areas (white arrowhead, Fig. 8G). At the same stage, the pp60^c-src/myr− distribution had not yet returned to its interphase distribution (predominently in the cytoplasm). As long as the DAPI staining retained its punctate pattern, myr⁻ c-Src remained concentrated mostly in the nucleus (Fig. 8J-L). It subsequently reverted to the distribution typical of interphase cells (lower cell, Fig. 8J-L).

The same anti-Src mAb, mAb327 (Lipsich et al., 1983), has been used to compare the subcellular localization of myristoylated (wt) and nonmyristoylated (myr⁻) c-Src in NIH 3T3-derived overexpresser cells (both expressing similar levels of wt and mutant chicken pp60^c-src, ∼15-to 20-fold higher than the level of endogeneous mouse pp60^c-src) during the various phases of the cell cycle. Similar observations have been made with a second anti-Src mAb, GD11 (Parsons et al., 1984), and with a polyclonal anti-Src antibody, αp60 (Resh and Erikson, 1985). A summary of the results obtained is presented in Fig. 9. We recall the main observations: (1) in G1 and S phases, pp60^c-src partitions primarily between the inner face of the plasma membrane and the centrosome (with somewhat higher pericentriolar concentration) while pp60^c-src/myr− is predominantly distributed diffusely in the cytoplasm with a small amount in the nucleus; (2) in G2 phase pp60^c-src leaves the plasma membrane and is redistributed diffusely throughout the cytoplasm and into two clusters of patches surrounding the two separating centriole pairs; in contrast, pp60^c-src/myr− translocates predominantly to the nucleus prior to any clear evidence of nuclear membrane breakdown or nuclear lamina disassembly; (3) during early mitosis, up to anaphase, pp60^c-src is found primarily in diffuse and patchy structures dispersed throughout the cytoplasm whereas pp60^c-src/myr− strongly associates with specific structures involved in mitosis: that is, with the main constituents of the spindle apparatus and the segregating chromosomes; (4) in telophase pp60^c-src gradually resumes its characteristic interphase, partitioning between the plasma membrane and the centrosome, but also remains concentrated in part at the intercellular contact area between the two still-attached daughter cells; apparently more time is required for pp60^c-src/myr− to recover its typical interphase distribution, since myr⁻ c-Src remains concentrated in the nucleus as long as the two daughter cells are connected.

For technical reasons the studies described here were performed using cell lines expressing abnormally high levels of the Src proteins and the possibility of artifacts must be considered. For example, saturating the highest-affinity, physiological sites of interaction could possibly result in secondary association with unsaturated, lower-affinity subcellular sites. We believe however that such a risk is minimal and, in any case, needs to be assumed when one is working with proteins (like c-Src) that are normally present at very low levels that are undetectable by ordinary available techniques. Because of the possibility of artifacts, one should be especially cautious if the overexpressed protein apparently colocalizes and/or copurifies with a major cellular constituent such as a cytoskeletal protein like tubulin or myosin or a plasma membrane-associated receptor. But the finding that overexpressed Src apparently colocalizes with minor cellular components such as the pericentriolar material or a nuclear antigen is, to us, more likely to reflect a true situation, which, without the artifice of overexpression, would have remained undiscovered. Confidence in these results obtained with overexpressed Src is supported to the extent that they overlap previous studies indicating also that wt pp60^c-src and naturally occurring nonmyristoylated forms of Src exhibit major localization sites at the plasma membrane (Loeb et al., 1987) and in the cytoplasm (Krueger et al., 1982, 1984), respectively. Furthermore, variations between the expression levels of Src in individual cells provided an internal check that indicated the lack of any significant effects of dose on subcellular distribution: even in cells in which the level of overexpression was lowest (but still detectable), the anti-Src antibodies showed that pp60^c-src preferentially occupied a centrosomal pool and was secondarily located at the plasma membrane, while pp60^c-src/myr− preferentially occupied a nuclear pool at the end of G2/early prophase and was secondarily located in the cytoplasm. We infer that this indicates the presence of high-affinity, physiological sites of interaction for pp60^c-src in the centrosomal area and for pp60^c-src/myr− in the nucleus.

Also, the significant differences observed between the cell cycle-dependent subcellular distributions of the wild-type and nonmyristoylated forms suggest that the patterns observed reflect specific associations and not merely nonspecific low-affinity sites of interaction.

Many cases are known where protein mutation leads to mislocalization, particularly when the mutation affects a region involved in the specification of localization (as for the myristoylation site in pp60 ^src) or when it induces a conformational change that exposes cryptic localization sequences (as for v-Sis, lacking its signal sequence; Lee et al., 1987). These changes in protein localization very frequently correlate with changes in the biological activity of the proteins, emphasizing the influence of the subcellular environment on the physiological function of an enzyme. These remarks are especially relevant to the Src field and further justify our present investigation, since it has previously been shown that myristoylation is required for plasma membrane localization of pp60^src (Buss et al., 1986) and that it augments the transforming potential of the viral and mutated cellular enzymes (Cross et al., 1984; Pellman et al., 1985; Schuh and Brugge, 1988) even though it does not affect their intrinsic protein-tyrosine kinase-specific activities (Schultz et al., 1985; Kamps et al., 1986; Reynolds et al., 1989). However, it has often been tacitly assumed that abrogation of myristoylation simply results in loss of a pecific (plasma membrane) localization, not that it may lead to a highly specific, alternative pattern of localization. The detailed comparative immunofluorescence study presented here clearly illustrates the complexity of the factors influencing the localization of pp60^c-src. In addition to inducing a gain of function (plasma membrane localization), myristoylation induces three losses of function. It: (1) prevents pp60^c-src translocation to the nucleus during growth arrest; (2) prevents nuclear translocation during G2 in a normal cell cycle; and (3) strongly reduces the affinity of pp60^c-src for components of the spindle apparatus and for the interchromosomal space during mitosis. (Among a dozen overexpressed c-src mutants studied, the myristoylation-defective one is the only one that induced a nuclear translocation of the protein and an association with the mitotic apparatus.) This diversity of effect is consistent with the view, arising from studies of a variety of myristoylated viral and cellular proteins (McIlhinney, 1990, for review), that myristoylation may mediate protein-protein interactions rather than protein-membrane interactions. Even the effect of myristoylation on plasma membrane association may be mediated via effects on protein-protein interactions, since pp60^src must interact with saturable high-affinity binding sites for membrane binding (Resh, 1989; Goddard et al., 1989; Resh and Ling, 1990).

It is noteworthy that wt pp60^c-src has a strong affinity for specific structures that are well developed during interphase before nuclear envelope breakdown (i.e. the plasma membrane, the centrosome and (during G2) the two pericentriolar areas). Beginning at mitosis, when the interphasic microtubules start depolymerizing, pp60^c-src is redistributed into diffuse and patchy structures throughout the cytoplasm, an effect that can be mimicked by artificial disruption of the interphasic microtubules through nocadozole treatment. These observations: first, indicate that both the maintenance of pp60^c-src association with the plasma membrane and its accumulation in the centrosomal area depend on the integrity of the interphasic microtubule network; and, second, they also suggest that the redistribution of wt pp60^c-src starting at the end of G2 could simply result from the natural change in the state of organization and the subsequent depolymerization of the interphasic microtubules. During mitosis, wt pp60^c-src does not display particularly high affinity for any specific structures participating in mitotic events. These observations are in direct contrast with those for pp60^c-src/myr−; its interphase distribution does not depend at all on the presence of a well-developed microtubule network but, starting at late G2 phase and during mitosis it exhibits a strong affinity for many components that participate in mitotic events (spindle poles and fibers, interchromosomal material).

pp60^src has been linked to control of a wide variety of events occurring throughout the cell, from the plasma membrane to the nucleus (for review, Cooper, 1990). While its varied activities might be mediated by multiple static subpopulations that participate in (tyrosine) phosphorylation cascades, it is interesting to consider the possibility that c-Src participates in intracellular signalling as a physically transported messenger between the plasma membrane and the nucleus. We have previously shown that wt pp60 ^c-src apparently associates with vesicles containing ConA-receptor complexes (endosomes) throughout the entire ConA-receptor-mediated endocytotic process (David-Pfeuty and Nouvian-Dooghe, 1990). This led us to suggest that the accumulation of pericentriolar and perinuclear pp60^c-src-containing patches during interphase might represent plasma membrane-derived vesicles or endosomes in transit between the plasma membrane and the centrosome. This hypothesis is supported, on one hand, by the finding reported here that a subpopulation of plasma membrane-associated wt pp60^c-src appears to be a component of vesicular structures that are insoluble in the presence of nonionic detergents (possibly early endosomes) and, on the other hand, by a biochemical fractionation study (Kaplan et al., 1992) showing that a subpopulation of wt pp60^c-src cofractionates with late endosomes. The report (Linstedt et al., 1992) that pp60^c-src also associates specifically with synaptic vesicles in the neuroendocrine PC12 cell line is also consistent with such an hypothesis. We find that the lack of myristoylation of pp60^c-src/myr− blocks association not only with the plasma membrane, but also with ConA-induced endocytotic vesicles and endosomes (unpublished results). This is consistent with the observation that pp60^c-src/myr− does not accumulate like wt pp60^c-src at the centrosome in interphase cells. It also implies that pp60^c-src/myr− must be transported to the nucleus by an independent mechanism.

A central question raised by this work is whether the behavior of overexpressed pp60^c-src/myr− could possibly reflect that of a naturally occurring subset of pp60^c-src in normal cells. Buss and Sefton (1985) have estimated that as much as 16% of pp60^v-src may actually not be myristoylated. If an equivalent percentage of pp60^c-src is nonmyristoylated in normal cells, it would be undetectable by conventional techniques. Overexpression of wt pp60^c-src by a factor of 10 to 20 would even be insufficient to raise such a nonmyristoylated subpopulation above the threshold level of detectability by immunofluorescence (which is a few fold higher than the endogeneous level of pp60^c-src); this may account for the fact that we did not detect such a subpopulation in wt c-Src overexpresser cells. We do not know if the molecules in this putative non-myristoylated population possess an N-terminal methionine or if, as for wild-type pp60a-40Vc-Srca+40V, it is removed by aminopeptidase activity. While there is no reason to believe that the presence or absence of this residue (or the additional alanines present in the pcLN mutant) plays any significant role or that the localization site(s) that govern nuclear translocation are at the amino end of the molecule, the possibility that small amino acid changes in this region could have effects on localization cannot be excluded. Interestingly, abundant nuclear concentration of myr⁻ c-Src in late G2/early prophase cells and within the mitotic apparatus does not interfere with mitotic progression; nor is progression through the cell cycle prevented by the massive nuclear accumulation of myr⁻ c-Src that occurs in cells that have been blocked at the G1/S transition by thymidine-aphidicolin treatment. These observations imply that a naturally occurring subset of pp60^c-src/myr− would probably not exert an inhibitory effect on cell cycle and mitotic progression. It is also possible that the behavior of pp60^c-src/myr− could mimic to some extent that of a subpopulation of wt pp60^c-src. Indeed allosteric modifications of pp60^c-src, perhaps resulting from phosphorylations or dephosphorylations within the amino-proximal region of the molecule, might interfere with the protein-protein interaction required for plasma membrane localization and could generate cytosolic enzymes capable of being transported to the nucleus. This is consistent with the report that myristate is present in soluble cytoplasmic as well as membrane-bound, pp60^src (Buss et al., 1984). Such hypothetical motions would probably be regulated in a more refined manner that would pre-vent the excessive nuclear accumulation observed here with the constitutive mutant myr⁻ c-Src. Within this model, no predictions can be made regarding the issue of whether a naturally occurring subset of c-Src proteins (whose behaviour would only be partially mimicked by mutant pp60^c-src/myr−) would exert an inhibitory or positive effect on cell cycle and mitotic progression.

It is worthwhile to mention here two recent observations that are relevant to our observations. First, Zhao et al. (1992) have reported that in vitro calcium-induced keratinocyte differentiation occurs concomitant with a marked increase in nuclear phosphotyrosine content and along with a translocation of c-Src to the nucleus. Second, C. Willman (personal communication) has shown that a naturally occurring variant of c-Fgr lacking the amino-terminal domain arises by alternative internal translation initiation in myeloid cells; this variant has a nuclear location in myeloid cells and in stably transfected fibroblasts. These data support the hypothesis that naturally occurring forms of pp60^c-src and of other members of the Src family could indeed move into the nucleus and exert a physiological function in that location. Our results also raise the intriguing suggestion that myristoylation might be required to prevent wt pp60^c-src from interacting with nuclear constituents during G2 phase and with the mitotic apparatus during mitosis. Interestingly, another well-described myristoylated enzyme is the catalytic subunit of cAMP-dependent protein kinase II (cAMP-dPKII), which translocates to the nucleus solely following cAMP-induced dissociation from the Golgi-associated cAMP-dPKII regulatory subunit (Nigg et al., 1985). Since many protein kinases (including Src) exhibit little specificity in in vitro assays, catalyzing the phosphorylation of a wide variety of substrates that apparently are never phosphorylated under physiological conditions, it is easy to conceive that the spatial and temporal distributions of such enzymes need to be subjected to very strict control in vivo in order to avoid unscheduled protein phosphorylation that could be catastrophic for cell survival.

We thank Yolande Nouvian-Dooghe for her excellent technical assistance, Mesdames Irène Gaspard and Colette Pouget for their expert artwork, Eric Bailly for his helpful advices, Sarah Parsons, Marilyn Resh, Françoise Danon and Paul-Henri Mangeat for providing antisera and Joan Brugge for plasmid pRSVc-srcLN. We also thank Madame Françoise Arnouilh for typing the manuscript. This study was supported by the Centre National pour la Recherche Scientifique, the Curie Institute and the Association pour la Recherche sur le Cancer (France) and by grants CA32317 and CA47333 and RCDA CA01139 from the National Institutes of Health (USA).