Differential trafficking of Src, Lyn, Yes and Fyn is specified by the state of palmitoylation in the SH4 domain

Src-family tyrosine kinases (SFKs), which participate in a variety of signal transduction events, are known to localize to the cytoplasmic face of the plasma membrane through lipid modification. Recently, we showed that Lyn, an SFK member, is exocytosed to the plasma membrane via the Golgi region along the secretory pathway. We show here that SFK trafficking is specified by the palmitoylation state. Yes is also a monopalmitoylated SFK and is biosynthetically transported from the Golgi pool of caveolin to the plasma membrane. This pathway can be inhibited in the trans-Golgi network (TGN)-to-cell surface delivery by temperature block at 19°C or dominant-negative Rab11 GTPase. A large fraction of Fyn, a dually palmitoylated SFK, is directly targeted to the plasma membrane irrespective of temperature block of TGN exit. Fyn(C6S), which lacks the second palmitoylation site, is able to traffic in the same way as Lyn and Yes. Moreover, construction of Yes(S6C) and chimeric Lyn or Yes with the Fyn N-terminus further substantiates the importance of the dual palmitoylation site for plasma membrane targeting. Taken together with our recent finding that Src, a nonpalmitoylated SFK, is rapidly exchanged between the plasma membrane and late endosomes/lysosomes, these results suggest that SFK trafficking is specified by the palmitoylation state in the SH4 domain.

Introduction are transported to the plasma membrane via the Golgi region along the secretory pathway, whereas a large fraction of Fyn, a dually palmitoylated SFK, is directly targeted to the plasma membrane. Our findings suggest that the state of mono-or dual-palmitoylation makes a difference in trafficking among palmitoylated SFKs.

Results
Yes and Lyn localize to the Golgi region Src, Lyn, Yes and Fyn are widely expressed in various cell types (Thomas and Brugge, 1997;Kuga et al., 2007). We recently showed that Lyn is biosynthetically transported to the plasma membrane via the Golgi pool of caveolin along the secretory pathway (Kasahara et al., 2004). To examine the localization of other ubiquitously expressed SFKs, we transfected COS-1 cells with Lyn, Yes or Fyn and compared their localization. In the earlier phase of expression, localization of Lyn was predominantly seen in the perinuclear region and gradually changed to the plasma membrane in the later phase (Fig. 1A,B, top), consistent with our report (Kasahara et al., 2004). Similarly to Lyn, Yes was predominantly found at the perinuclear region during the earlier phase of expression and at the plasma membrane in the later phase (Fig. 1A,B, middle). By contrast, the majority of Fyn was found at the plasma membrane from the early phase (Fig. 1A,B, bottom), which is consistent with previous studies (van 't Hof and Resh, 1997). These results suggest that Yes, as well as Lyn, accumulates initially at the perinuclear region and is then transported to the plasma membrane.
To produce a synchronous wave of protein synthesis and trafficking, we used a HeLa cell clone (3-2) (Kasahara et al., 2007b) stably expressing the tetracycline repressor for an inducible expression system. At 12 hours after transfection with Lyn, Yes or Fyn, doxycycline (Dox) was added to medium to induce protein expression. In the course of induction, most Lyn and Yes proteins were initially observed in the perinuclear region and were subsequently detected at the plasma membrane, whereas Fyn was predominantly present on the plasma membrane from the early phase of induction (supplementary material Fig. S1A,B). These Journal of Cell Science 122 (7) results further support the possibility that Yes is transported from the perinuclear region to the plasma membrane.
Next, we characterized the perinuclear region using Golgi markers, such as β-1,4-galactosyltransferase (GalT), a trans-Golgi protein and GM130, a cis-Golgi protein. A large fraction of perinuclear Yes was also partially colocalized with GalT ( Fig. 1C), as also observed with inducible Yes and Lyn (supplementary material Fig. S1C). In addition, Yes was precisely colocalized with caveolin in the perinuclear region (supplementary material Fig. S2), indicating that perinuclear Yes is localized to the Golgi pool of caveolin similarly to the localization of Lyn in the perinuclear region. Perinuclear Yes and Lyn were dispersed by brefeldin A (BFA) (Fig.  1D), further confirming their Golgi localization.
To examine whether localization of endogenous SFKs was similar to those of overexpressed proteins, we used the human monocytic cell line THP-1 in which endogenous Lyn and Fyn were detectable by immunoblotting and immunostaining (Fig. 1E,F). Although endogenous Yes was hardly visualized owing to its low expression (data not shown), endogenous Lyn was localized to the perinuclear region and the plasma membrane. BFA treatment caused dispersal of endogenous Lyn from the perinuclear region ( Fig. 1F). Endogenous Fyn was predominantly seen at the plasma membrane and showed no apparent redistribution in the presence of BFA. These results suggest that localization of endogenous SFKs mirror those of the overexpressed proteins.

Newly synthesized Yes initially accumulates in the Golgi region
We next determined whether the presence of Yes in the Golgi region reflected newly synthesized protein in transit to the plasma membrane. COS-1 cells expressing Lyn, Yes or Fyn were treated with cycloheximide (CHX), a protein synthesis inhibitor. CHX treatment markedly reduced perinuclear Lyn within 60 minutes ( Fig.  2A,B, top), which is consistent with an earlier report (Kasahara et al., 2004). Reduction of perinuclear Yes was also found within 60 minutes of CHX treatment (middle), indicating that CHX treatment results in a time-dependent chase of newly synthesized Yes from the Golgi region to the plasma membrane. By contrast, CHX treatment did not affect Fyn localization to the plasma membrane (bottom). These results suggest that newly synthesized Yes but not Fyn is initially accumulated at the Golgi region and transported to the plasma membrane in a manner similar to Lyn trafficking.
To further examine whether the localization of endogenous Lyn and Fyn was affected by CHX treatment, THP-1 cells were treated with CHX for 3 hours and stained for Lyn and Fyn. Perinuclear accumulation of endogenous Lyn diminished upon CHX treatment (Fig. 2C, top), whereas plasma membrane localization of endogenous Fyn was not affected (Fig. 2C, bottom). These results suggest that the trafficking of endogenous Lyn is different from that of endogenous Fyn, in agreement with the results shown in Fig.  2A,B.
Yes and Lyn traffic to the plasma membrane through the exocytic pathway Since Yes is localized to the Golgi region and chased to the plasma membrane by CHX treatment, we tested whether Yes, like Lyn, was transported to the plasma membrane through the exocytic pathway. We took advantage of the property of a reduced temperature (19°C) to specifically block TGN membrane vesicle exit (Griffiths et al., 1989;Watson et al., 2003). COS-1 cells expressing Lyn or Yes were incubated at 19°C for 3 hours in the presence of CHX to block protein synthesis but allow accumulation of the synthesized proteins in the Golgi region. Incubation with CHX at 19°C caused accumulation of Yes as well as Lyn in the perinuclear region (Fig. 3A, middle), whereas perinuclear accumulation of Lyn and Yes was markedly reduced upon incubation with CHX for 3 hours at 37°C (Fig. 3A, left). When the cells incubated at 19°C were warmed to 37°C, the perinuclear accumulation of Lyn and Yes was rapidly dissipated (Fig. 3A, right). In sharp contrast, incubation at 19°C did not allow Fyn to accumulate in the perinuclear region (Fig. 3A, bottom). These results suggest that biosynthetic exocytic traffic is involved in the recruitment of Yes to the plasma membrane in a manner similar to Lyn trafficking.
Overexpression of the dominant-negative Rab11S25N mutant (Rab11DN) causes inhibition of the TGN-to-cell surface delivery (Chen et al., 1998). To ascertain whether the plasma membrane delivery of a conventional exocytic cargo was mediated by Rab11, we used green fluorescent protein (GFP)-tagged vesicular stomatitisvirus-encoded glycoprotein (VSVG-GFP), whose export from the ER occurs in a temperature-dependent manner (Presley et al., 1997). COS-1 cells transfected with VSVG-GFP or VSVG-GFP plus HA-Rab11DN were incubated at 40°C (nonpermissive temperature) overnight and then shifted to 32°C (permissive temperature) for 30 minutes or for 6 hours, including CHX for the last 3 hours. In cells transfected with VSVG-GFP alone, VSVG-GFP was predominantly found in the Golgi region at 30 minutes after temperature shift, and most VSVG-GFP was transported to the plasma membrane after incubation for 6 hours (Fig. 3Ba, left), consistent with previous reports (Hirschberg et al., 1998;Hirose et al., 2004). Upon overexpression of Rab11DN, VSVG-GFP was largely retained in the Golgi region at 6 hours after temperature shift (Fig. 3Ba, right), in agreement with previous studies (Chen et al., 1998). We then examined whether overexpression of Rab11DN inhibited the transport of Lyn and Yes to the plasma membrane. COS-1 cells transfected with Lyn alone, Lyn plus Rab11DN, Yes alone or Yes plus Rab11DN were cultured for 24 hours. At this incubation period after transfection, perinuclear Lyn and Yes were largely transported to the plasma membrane and little accumulation of Lyn and Yes was seen in the perinuclear region (Fig. 3Bb,left;see Fig. 1B). Intriguingly, overexpression of Rab11DN strongly induced perinuclear accumulation of Lyn and Yes (Fig. 3Bb, right). These results indicate that overexpression of Rab11DN blocks the transport of Lyn and Yes to the plasma membrane, suggesting the involvement of Rab11 in the exocytic transport of Lyn and Yes.
To examine whether the transport of endogenous Yes was also inhibited by temperature block, we used the human megakaryocytic cell line Dami, in which endogenous Yes as well as Lyn and Fyn was easily detected by immunoblotting and immunostaining . Despite low levels of perinuclear accumulation, endogenous Lyn and Yes were found in the plasma membrane (Fig. 3D, left), and incubation at 19°C for 3 hours increased levels of perinuclear Lyn and Yes in most cells (Fig. 3D, right; Fig. 3E). These results suggest that endogenous Yes as well as Lyn traffics to the plasma membrane through the exocytic pathway. By contrast, Fyn was primarily found at the plasma membrane in cells incubated at 37°C and no apparent redistribution was seen after incubation at 19°C (Fig. 3D, bottom), further confirming the finding that newly synthesized Fyn is directly targeted to the plasma membrane (van't Hof and . In addition, Dami cells were incubated at 15°C to examine the involvement of ER membranes in Lyn and Yes transport. Incubation Journal of Cell Science 122 (7)

Fig. 3. Temperature block of TGN membrane transport results in perinuclear accumulation of Lyn and Yes. (A) COS-1 cells transfected with Lyn, Yes or
Fyn were incubated for 15 hours, subsequently treated with 100 μg/ml CHX for 3 hours at 19°C or 37°C and stained with anti-Lyn, anti-Yes or anti-Fyn antibody. The transfected cells incubated at 19°C for 3 hours in the presence of 100 μg/ml CHX were warmed to 37°C for further 2 hours (19°Cr37°C). Arrows indicate the perinuclear region. Cells exhibiting the perinuclear localization of expressed proteins were quantified. Results (%) are means ± s.d. (n=3~5). **P<0.01, Student's t-test. NS, not significant. Scale bars: 20 μm. (B) COS-1 cells transfected with VSVG-GFP or VSVG-GFP plus HA-Rab11S25N (+Rab11DN) were incubated overnight at 40°C and then shifted to 32°C for 30 minutes or for 6 hours, including 200 μg/ml CHX during the last 3 hours. Expressed proteins were detected with GFP fluorescence (green) or anti-HA antibody (red) (a). COS-1 cells transfected with Lyn, Lyn plus Rab11DN, Yes or Yes plus Rab11DN, cultured for 24 hours. Cells were doubly stained with anti-Lyn (green) or anti-Yes (green) antibody plus anti-HA antibody (red) (b). Insets show fluorescence images of Rab11DN (red). Cells exhibiting the perinuclear localization of Lyn or Yes were quantified. Results (%) are means ± s.d. (n=3-6). ***P<0.001, Student's t-test. N, nucleus. Scale bars: 20 μm. (C) Endogenous Lyn (p56/p53), Yes (p62) and Fyn (p59) expressed in Dami cells were immunoblotted with anti-Lyn, anti-Yes and anti-Fyn antibodies. (D) Dami cells incubated for 3 hours at 37°C or 19°C and stained with anti-Lyn, anti-Yes or anti-Fyn antibody. Scale bars: 20 μm. (E) Cells incubated at 19°C for 3 hours and doubly stained for Lyn or Yes (green), and GalT (red). Arrows indicate the perinuclear region. Scale bars: 10 μm. Differential trafficking of Src kinases at 15°C, which impairs transport from the ER to the cis-Golgi (Presley et al., 1997), resulted in accumulation of Lyn and Yes in the Golgi region but not the ER (supplementary material Fig. S3), suggesting that the trafficking pathways for Lyn, Yes and Fyn do not involve ER membranes.

Newly synthesized Yes and Lyn, but not Fyn, traffic to the plasma membrane through the Golgi region
To visualize the trafficking of Lyn, Yes and Fyn just after biosynthesis in living cells, we added a GFP tag to the proteins, which preserves their N-terminal lipid attachment sites (see Materials and Methods) (Kasahara et al., 2004), and performed fluorescence recovery after photobleaching (FRAP) of Lyn-GFP, Yes-GFP and Fyn-GFP in COS-1 cells. When the whole-cell area was photobleached, rapid recovery of Lyn-GFP fluorescence at 37°C was observed in both the perinuclear region and the nonperinuclear region that excludes the perinuclear region from whole cell area. Incubation at 19°C after whole-cell photobleaching inhibited the recovery of Lyn-GFP fluorescence in the nonperinuclear region but not the perinuclear region (Fig. 4A). Intriguingly, the recovery rates of Yes-GFP fluorescence at 37°C and 19°C were comparable with those of Lyn-GFP fluorescence (compare Fig. 4B with 4A). These results suggest that, similarly to Lyn, newly synthesized Yes traffics from the Golgi region en route to the plasma membrane.
When we compared the trafficking of Fyn in cells cultured at 37°C and 19°C, we found that incubation at 19°C did not inhibit the rapid recovery of Fyn-GFP in the non-perinuclear region after whole-cell photobleaching (Fig. 4C). Rather, this rapid recovery was similar to that observed at 37°C, and the recovery rates of Fyn-GFP in the perinuclear region at 19°C and 37°C were comparable with those in the non-perinuclear region. These results suggest that most of the newly synthesized Fyn traffics directly to the plasma membrane and only a small fraction targets to the Golgi region.

Palmitoylation at Cys6 influences Fyn localization and trafficking
Palmitoylation of SFKs occurs on cysteine residues in the general motif Met-Gly-Cys, where Gly2 is myristoylated (Resh, 1999). Although Lyn and Yes are monopalmitoylated at Cys3 (Koegl et al., 1994;McCabe and Berthiaume, 1999;Kasahara et al., 2007a), Fyn is dually palmitoylated at Cys3 and Cys6 (Alland et al., 1994). Since the trafficking pathway for Lyn and Yes is different to that for Fyn (Figs 3 and 4), we asked whether the lack of the second palmitoylation site in Fyn affected its trafficking. We generated a Fyn(C6S) mutant in which Cys6 in Fyn was substituted for Ser (Fig. 5A), a substitution previously shown to decrease palmitate incorporation of Fyn (Alland et al., 1994). The levels of protein expression and kinase activity were comparable between wild-type Fyn (Fyn-wt) and Fyn(C6S) (Fig.  5B). Intriguingly, Fyn(C6S) significantly accumulated in the Golgi region 12 hours after transfection compared with Fyn-wt (Fig. 5C,D), although the level of Fyn(C6S) was lower than that for Lyn and Yes (compare with Fig. 1B). The number of cells exhibiting the Golgi localization of Fyn(C6S) was gradually decreased after transfection (Fig. 5D), similar to the results for Lyn and Yes (Fig. 1B). Accumulation of Fyn(C6S) in the Golgi region was dispersed by BFA treatment (Fig. 5E), and accumulation of inducible Fyn(C6S) was also observed in the Golgi region in the early phase of induction (supplementary material Fig. S4 and Fig. S1).
Localization of Fyn(C6S) to the Golgi region in the early phase of expression prompted us to examine whether accumulation of Fyn(C6S) in the Golgi region was diminished by CHX. Upon treatment with CHX, the number of cells exhibiting perinuclear Fyn(C6S) decreased (Fig. 5F), suggesting that following its initial biosynthesis, Fyn(C6S) can enter the secretory membranetrafficking system. We then examined the involvement of Rab11 in transport of Fyn(C6S) to the plasma membrane. COS-1 cells transfected with Fyn-wt alone, Fyn-wt plus Rab11DN, Fyn(C6S) alone, or Fyn(C6S) plus Rab11DN were incubated for 24 hours. At this incubation period after transfection, most Fyn-wt and Fyn(C6S) were localized to the plasma membrane (Fig. 5G, left, Control). Overexpression of Rab11DN induced perinuclear accumulation of Fyn(C6S) (Fig. 5G, right), similar to the results for Lyn and Yes (Fig. 3Bb). These results suggest that Rab11 is involved in transport of Fyn(C6S).
Journal of Cell Science 122 (7) We noted that perinuclear accumulation of Fyn-wt was moderately induced by long-term coexpression of Rab11DN with Fyn-wt [~38% for Fyn-wt (Fig. 5G) and 75% for Lyn or Yes (Fig. 3Bb)]. About 15% of cells expressing Fyn-wt showed perinuclear Fyn-wt, which partially colocalized with GM130, a Golgi marker ( Fig.  5G; data not shown), and weak perinuclear fluorescence of Fyn-GFP could be detected at 19°C after whole-cell photobleaching (Fig. 4C). Also, a few percentages of Dami cells showed perinuclear staining for endogenous Fyn (supplementary material Fig. S3F). These results suggest that a small fraction of Fyn-wt might traffic from the Golgi region to the plasma membrane along the Rab11-mediated exocytic pathway.
In addition, we transfected COS-1 cells with Fyn-wt, Fyn(C6S) or Lyn, and examined tyrosine phosphorylation at the early phase of expression when Golgi accumulation of Fyn(C6S) and Lyn but not Fyn-wt was visible. Similarly to results observed with Lyn, Fyn(C6S) but not Fyn-wt increased tyrosine phosphorylation of pp40 (supplementary material Fig. S5), supporting the finding that localization of Fyn(C6S) is similar to that of Lyn.

Palmitoylation at Cys6 in Fyn contributes to its plasma membrane targeting
A Fyn-Src chimera exhibits by far the greatest incorporation of the palmitate analog, whereas mutation of Ser3 and Ser6 in viral Src to Cys shows mild incorporation (Alland et al., 1994). To examine whether palmitoylation at Cys 6 affected localization of Lyn and Yes, we first created a Fyn-Lyn chimera (Fyn24-Lyn), in which the SH4 domain of Lyn was substituted for that of Fyn-wt (Fig. 6A). Expression and kinase activity were confirmed in COS-1 cells transfected with each construct (Fig. 6B). Intriguingly, localization of Fyn24-Lyn was different from that of Lyn, despite containing 95% of the Lyn protein sequence, and Fyn24-Lyn was mainly localized at the plasma membrane in the same manner as Fyn (Fig.  6C). To assess whether the palmitoylation site at Cys6 was important for the targeting of Fyn24-Lyn to the plasma membrane, we introduced a mutation of Cys6 to Ser in Fyn24-Lyn and created Fyn24(C6S)-Lyn (Fig. 6A). Although Fyn24-Lyn was mainly localized to the plasma membrane, Fyn24(C6S)-Lyn localized to the perinuclear region and the plasma membrane (Fig. 6D), similarly to Lyn localization (Figs 1-3) (Kasahara et al., 2004). Additionally, perinuclear Fyn24(C6S)-Lyn was partially  Fig. S3B,D). These results suggest that the N-terminus preserving the monopalmitoylation site is responsible for Golgi localization of SFKs.
We then created a Fyn-Yes chimera (Fyn24-Yes) and confirmed its expression and kinase activity (Fig. 6B). Similarly to the results of Fyn24-Lyn, Fyn24-Yes was also localized at the plasma membrane (Fig. 6E), suggesting that the Fyn N-terminus, which contains one myristoylation site and two palmitoylation sites, allows Lyn and Yes to traffic directly to the plasma membrane. Furthermore, we generated a Yes(S6C) mutant in which Ser6 in Yes was substituted for Cys (Fig. 6F). The levels of protein expression were comparable between wild-type Yes and Yes(S6C) (Fig. 6F). Intriguingly, Yes(S6C) was mainly localized at the plasma membrane but not at the Golgi region, similarly to Fyn localization (Fig. 6F). Taken together, these results suggest that the presence of the second palmitoylation site in the N-terminus has an important role in the direct targeting of SFKs to the plasma membrane.

Discussion
In the present study, we demonstrate that SFK trafficking is specified by its palmitoylation state. Lyn and Yes, which are both monopalmitoylated SFKs, are biosynthetically transported to the plasma membrane via the Golgi pool of caveolin along the secretory pathway, whereas Fyn, a dually palmitoylated SFK, is directly targeted to the plasma membrane. We further show that Fyn(C6S), which lacks the second palmitoylation site, behaves in a manner reminiscent of Lyn and Yes by accumulating in the Golgi region early after biosynthesis, and that Yes(S6C), in which a second palmitoylation site is created, becomes localized to the plasma membrane.

Three major pathways for SFK trafficking
Palmitoylation of SFKs has been shown to contribute to their localization (Bijlmakers et al., 1997;Carreno et al., 2000). Although Src is not palmitoylated, Lyn and Yes are monopalmitoylated at Cys3 (Koegl et al., 1994;McCabe and Berthiaume, 1999;Kasahara et al., 2007a) and Fyn is dually palmitoylated at Cys3 and Cys6 (Alland et al., 1994). We recently showed that the trafficking of Src (a nonpalmitoylated SFK) is different from that of Lyn (a monopalmitoylated SFK) because of the state of palmitoylation (Kasahara et al., 2007a).
Here, we propose a model of three major trafficking pathways for SFKs (Fig. 7). (1) The cycling pathway for myristoylated but not palmitoylated SFKs: Src is rapidly exchanged between late endosomes or lysosomes and the plasma membrane, possibly through its cytosolic release (Kasahara et al., 2007a). (2) The secretory pathway from the Golgi region to the plasma membrane for myristoylated and monopalmitoylated SFKs: newly synthesized Lyn and Yes initially enter the Golgi system, where palmitoylation probably occurs, providing entry into the membrane secretory transport pathway en route to the plasma membrane (Fig. 3A,D,  Fig. 4) (Kasahara et al., 2004). Rab11 is involved in exocytic transport of Lyn and Yes (Fig. 3B). (3) The direct plasmamembrane-targeting pathway for myristoylated and dually palmitoylated SFKs: newly synthesized Fyn is targeted directly to the plasma membrane ( Fig. 2A, Fig. 3D, Fig. 4) (van't Hof and . Importantly, the results for endogenous Lyn, Yes and Fyn in THP-1 and Dami cells ( Fig. 1F; Fig. 3D) agree with our proposed model. Although perinuclear staining for endogenous Lyn and Yes in Dami cells is not so strong as that observed in THP-1 cells, build-up of Lyn and Yes in the Golgi region after temperature block in Dami cells (Fig. 3D) allows us to assume that newly synthesized Lyn and Yes can reach the plasma membrane rapidly in Dami cells compared with THP-1 cells. Furthermore, we found that like Lyn, Fyn(C6S), which lacks the second palmitoylation site, accumulated in the Golgi region (Fig. 5), and that the dually palmitoylated N-terminus of Fyn leads Lyn and Yes directly to the plasma membrane whereas the monopalmitoylated N-terminus of Fyn redirects Lyn to the Golgi region (Fig. 6). We also found that creation of the second palmitoylation site in Yes enables its transfer directly to the plasma membrane (Fig. 6). These results suggest that the state of monopalmitoylation (Lyn and Yes) or dual palmitoylation (Fyn) makes a difference in the trafficking pathways between Lyn-Yes and Fyn. Our model is also consistent with previous findings that nonpalmitoylated p61 Hck , like Src, is localized to lysosomes whereas monopalmitoylated p59 Hck localizes to the Golgi and the plasma membrane (Carreno et al., 2000). Taken together, these results provide evidence that SFK trafficking is classified into three groups depending on the state of palmitoylation, as depicted in our model.
Notably, a very small fraction of Fyn was found to localize to the Golgi region (Fig. 4C, Fig. 5C, Fig. 5D; supplementary material Fig. S3F) (van't Hof and . Given that palmitoylation of Fyn does not occur perfectly (Liang et al., 2004), it is plausible that mono-or non-palmitoylated Fyn can follow a pathway other than the direct targeting pathway to the plasma membrane. Imperfect palmitoylation, if any, might affect the trafficking of the other SFKs including Lyn and Yes.
Cys3 is considered to have a greater role in the membrane localization of Fyn than Cys6 judging from biochemical data, although it would be difficult to detect localization changes among various cell membranes (Shenoy-Scaria et al., 1994;Wolven et al., 1997). In the present study, we reveal that mutation at Cys6 Journal of Cell Science 122 (7) influences intracellular localization of Fyn by accumulating the protein in the Golgi region (Fig. 5C). Taken together with the finding that Yes(S6C) becomes localized at the plasma membrane (Fig. 6), we thus hypothesize that the second palmitoylation site at Cys6 is important for plasma membrane targeting by keeping Fyn away from the Lyn-type trafficking. The presence of the second cysteine residue in the SFK N-terminal domain might influence recognition by proteins that are involved in protein palmitoylation or sorting to the plasma membrane. This hypothesis may be also explained by the results that Lck, which is dually palmitoylated at Cys3 and Cys5 (Koegl et al., 1994;Yurchak and Sefton, 1995), accumulates in the Golgi region by mutation of the second palmitoylation site at Cys5 in transfected NIH-3T3 cells (Bijlmakers et al., 1997).
Since the level of Golgi-accumulated Fyn(C6S) was lower than that of Lyn and Yes ( Fig. 1B; Fig. 5D), other factors besides palmitoylation might influence localization of SFK. Given that Fyn24-Lyn and Fyn24-Yes, which contain their SH3, SH2 and kinase domains, tend to stay in the Golgi, unlike Fyn-wt (Fig. 6D-E), we assume that the SH3 and SH2 domains of Lyn and Yes might contribute to retention of Lyn and Yes in the Golgi region by proteinprotein interactions with Golgi components.
Palmitoylation is mediated by protein acyltransferases (PATs), which include a large family of integral membrane proteins that contain a DHHC cysteine-rich domain (DHHC PATs) (Linder and Deschenes, 2007). More than 20 DHHC PATs were isolated in mammals, and recent progress has shown that DHHC PATs have exquisite substrate specificity (Fukata et al., 2004;Ohno et al., 2006;Linder and Deschenes, 2007). Most of the DHHC PAT members are localized to the ER and/or Golgi, but some are in fact found at the plasma membrane (Ohno et al., 2006). Although DHHC PATs are likely to account for most palmitoylation events, there might be substrate-specific PATs that are not members of the DHHC family (Linder and Deschenes, 2007). Taken together with the hypothesis that Fyn is targeted directly to the plasma membrane without a requirement for the exocytic machinery (van't Hof and Bijilmakers and Marsh, 2003), it would be reasonable to assume that PATs localizing to the Golgi region and the plasma membrane, although they have not yet been identified, are responsible for palmitoylation of Lyn-Yes and Fyn, respectively.
Lyn and Yes are precisely colocalized with the Golgi pool of caveolin (supplementary material Fig. S2) (Kasahara et al., 2004); they also partially colocalize with GalT and GM130, which are Golgi-resident proteins (Fig. 1C, Fig. 3E; supplementary material Fig. S1C and Fig. S3B,D). Inhibition of protein synthesis by CHX results in a chase of Lyn and Yes to the plasma membrane, with a concomitant chase of the Golgi pool of caveolin (this study) (Kasahara et al., 2004), but it does not allow GalT and GM130 to move from the Golgi complex to the plasma membrane (data not shown). Given that the Golgi pool of caveolin accounts for the intracellular pool of the protein in the presence of continual exit of caveolin in exocytic vesicles (Nichols, 2002), we presume that perinuclear Lyn and Yes localize to caveolin-positive exocytic vesicles in the Golgi region.

Implications for roles of the distinct trafficking of SFKs
Recent evidence provides a novel view that endomembranes, such as Golgi and ER membranes, serve as a platform of signaling molecules. For example, Ras that is restricted to endomembranes was shown to couple to the ERK-signaling cascade (Chiu et al., 2002). It is also shown that phosphatidylinositol-3,4,5-trisphosphate levels are increased to a larger extent at endomembranes than at the plasma membrane, which is triggered by endocytosed receptor tyrosine kinases (Sato et al., 2003). In the present study, we demonstrate that each SFK member localizes to various intracellular organelles because of the distinct trafficking pathway depicted in Fig. 7. Thus, our findings raise the intriguing possibility that the distinct localization brings about specific functions of SFKs by allowing SFKs to interact with differently located substrates.
Indeed, we recently showed that Src is involved in targeting of macropinosomes to lysosomes and accumulation of lysosomes (Kasahara et al., 2007b;Kasahara et al., 2008). It is also reported that Src activation on RhoB-associated late endosomes leads to Src translocation from the perinuclear region toward discrete cell peripheral structures (Sandilands et al., 2004). However, Lyn phosphorylates annexin II on endomembranes including Golgi membranes (Matsuda et al., 2006). Moreover, p61 Hck , but not p59 Hck , triggers the biogenesis of podosomes by exocytosis from lysosomes (Cougoule et al., 2005). These results strengthen our idea that specific functions among SFK members stem from their distinct intracellular localizations. In addition, we found that, like Lyn, Fyn(C6S) but not Fyn-wt increased tyrosine phosphorylation of pp40 (supplementary material Fig. S5), suggesting that an alteration in the subcellular localization of Fyn influences its access to SFK substrates. Blk and Src, both of which are nonpalmitoylated SFKs, phosphorylate coexpressed immunoglobulin-α, whereas palmitoylated SFKs do not (Saouaf et al., 1997). Taken together, the distinct trafficking of SFKs specified by the state of palmitoylation might be important for their specific roles and functional compensation.
Many signaling proteins are palmitoylated (Milligan et al., 1995;Resh, 1999;Smotrys and Linder, 2004). The notion that palmitoylation of SFKs is crucial for proper trafficking and function reinforces the possibility that the trafficking of other palmitoylated proteins, such as Ras GTPases and eNOS, is specified by their palmitoylation state and thus might lead to their specific functions. In fact, monopalmitoylation of H-Ras on Cys184 results primarily in a Golgi localization and impairs its ability to activate Raf, MEK and ERK (Roy et al., 2005). For eNOS, mutation of the palmitoylation sites blocks tight perinuclear targeting and nitric oxide release from cells (Liu et al., 1996). Palmitoylation might be essential for many proteins as well as SFKs, for proper trafficking and localization to fulfil their function. Furthermore, SFKs are abundantly expressed throughout the central nervous system (CNS) and regulate ion channel activity and synaptic transmission in the developed CNS (Wang and Salter, 1994;Thomas and Brugge, 1997). In neurons, palmitoylation has a key role in targeting proteins for transport to nerve terminals and for regulating trafficking at synapses (El-Husseini and Bredt, 2002;Huang and El-Husseini, 2005). It is therefore attractive to speculate that SFK members might be deeply involved in neuronal functions through their palmitoylation-regulated, spatiotemporal trafficking.
Considering the importance of protein trafficking in their proper localization and function, there is great interest in understanding the mechanisms that regulate palmitoylation-induced protein sorting and 'cracking' the palmitoylation codes within proteins to reveal how a palmitoylated peptide sequence relates to the final destination of a protein in the cell. Like the search for palmitoylation consensus sequences, these questions are likely to present a significant challenge.

Immunofluorescence
Immunofluorescence staining was performed as described (Yamaguchi and Fukuda, 1995;Tada et al., 1999;Yamaguchi et al., 2001;Kasahara et al., 2004;Nakayama and Yamaguchi, 2005;Kuga et al., 2008). In brief, cells were washed in warmed PBS and fixed in 3% paraformaldehyde for 20 minutes. Fixed cells were permeabilized and blocked in PBS containing 0.1% saponin and 3% bovine serum albumin for 30 minutes, and then incubated with a primary and a secondary antibody for 1 hour each. After washing with PBS containing 0.1% saponin, cells were mounted with ProLong antifade reagent (Molecular Probes). Confocal images were obtained using an LSM510 (Carl Zeiss) and a Fluoview FV500 (Olympus, Tokyo, Japan) laser-scanning microscope with a ϫ40 0.75 NA or ϫ63 1.40 NA oil-immersion objective, or a ϫ40 1.00 NA oil or a ϫ60 1.00 NA water-immersion objective. 200-400 cells were scored for each assay. For immunofluorescence of THP-1 and Dami cells, cells in suspension were directly fixed with 3% paraformaldehyde and then attached on coverslips by brief cytocentrifugation. Composite figures were prepared using Photoshop 5.0 and Illustrator 9.0 software (Adobe).

Temperature block
COS-1 cells transfected with Lyn or Yes were incubated at 37°C for 15 hours and subsequently treated with 100 μg/ml cycloheximide (CHX) and incubated at 19°C or 37°C for 3 hours in medium supplemented with 25 mM HEPES (pH 7.4). For release from temperature block, the cells incubated at 19°C for 3 hours were warmed to 37°C for further 2 hours. Dami cells were equilibrated in medium supplemented with 25 mM HEPES (pH 7.4) and then incubated at 15°C, 19°C or 37°C for 3 hours in suspension culture.

VSVG-GFP transport assay
COS-1 cells transfected with VSVG-GFP alone or VSVG-GFP plus HA-Rab11S25N were incubated at 40°C for 15 hours in medium supplemented with 25 mM HEPES (pH 7.4) and then shifted to 32°C for 30 minutes or 6 hours in the presence of 200 μg/ml CHX during the last 3 hours.

Photobleaching experiments
Fluorescence recovery after photobleaching (FRAP) experiments were performed as described (Kasahara et al., 2004;Kasahara et al., 2007a) using a Fluoview FV500 laser scanning microscope with a ϫ40 1.00 NA oil-immersion objective. COS-1 cells transfected with Lyn-GFP, Yes-GFP, or Fyn-GFP were cultured for 20 hours. Cells were maintained at 37°C or 19°C throughout photobleaching experiments. The whole cell area was photobleached at 100% laser power and recovery of fluorescence was monitored by scanning the perinuclear area or whole cell area excluding perinuclear region at low laser power (3.0%). The mean fluorescence intensities within regions of interest were quantified using the ImageJ software (National Institutes of Health, USA). Composite figures were prepared using Photoshop 5.0 and Illustrator 9.0 software (Adobe).

Western blotting
Cells were lysed in SDS-PAGE sample buffer at 4°C, and the samples were boiled for 3 minutes, subjected to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and electrotransferred onto polyvinylidene difluoride membranes (Millipore). Immunodetection was performed by enhanced chemiluminescence (Amersham Biosciences) as described (Mera et al., 1999;Matsuda et al., 2006;Kuga et al., 2007;Kasahara et al., 2007c;Ikeda et al., 2008). Sequential reprobing of membranes with a variety of antibodies was performed after the complete removal of primary and secondary antibodies from membranes in stripping buffer or inactivation of HRP by 0.1% NaN 3 , according to the manufacturer's instructions. Results were analyzed using an image analyzer LAS-1000plus equipped with Science Lab software (Fujifilm, Tokyo, Japan). For detection of tyrosine phosphorylation (supplementary material Fig. S5), cells were lysed in SDS-PAGE sample buffer containing 10 mM unbuffered HEPES and 10 mM Na 3 VO 4 at 4°C, and the samples were subjected to SDS-PAGE.