Synaptotagmin 3 deficiency in T cells impairs recycling of the chemokine receptor CXCR4 and thereby inhibits CXCL12 chemokine-induced migration

The migration of leukocytes into tissues in vivo is directed by chemokines and their receptors (Rossi and Zlotnik, 2000). This occurs mainly in inflamed tissues, in which expression of certain chemokines is upregulated. However, even in the absence of inflammation, tissues contain chemokines that are probably involved in the migration of leukocytes into these tissues. In particular, CXCL12 (SDF1) is constitutively expressed in many organs. Its receptor, CXCR4, is present on most leukocytes as well as many types of tumor cells (Kakinuma and Hwang, 2006), and plays a role in the metastasis of various malignancies (Muller et al., 2001; Zeelenberg et al., 2001; Zeelenberg et al., 2003).

Cell migration has been proposed to depend on vesicle trafficking and exocytosis (Hopkins et al., 1994). Exocytosis at the leading edge was suggested to have two major roles (Bretscher, 1996). First, it provides additional membrane necessary for the generation of membrane-rich structures such as filopodia. Second, the vesicles are a source of signaling and adhesion molecules. Indeed, several proteins involved in vesicle trafficking are necessary for migration. For example, N-ethylmaleimide-sensitive factor, which is necessary for the disassembly of the SNARE (soluble NSF attachment protein receptors) complex, is also required for cell polarization and locomotion (Thompson and Bretscher, 2002). RAB11 and its downstream target rabphilin 11 regulate vesicle trafficking and play a role in migration (Mammoto et al., 1999), and RAB4-dependent recycling of αVβ3 integrin is essential for migration on substrates coated with αVβ3 ligands (Roberts et al., 2001).

Synaptotagmins (SYTs) are transmembrane proteins with a short extracellular domain that act as calcium sensors for calcium-triggered vesicle fusion (Chapman, 2002), in particular of synaptic vesicles in neurons. Fifteen synaptotagmin genes have been identified so far (Fukuda, 2003). The best studied, SYT1 and SYT2 are mainly expressed in neurons, but several other synaptotagmins are more ubiquitous and some of these, such as SYT7, have higher affinity for calcium than the neuronal SYT1, in the micromolar range (Bhalla et al., 2005). SYTs have been implicated in several processes in non-neuronal cells, such as regulated exocytosis (Gao et al., 2000; Mizuta et al., 1997), membrane repair (Martinez et al., 2000; Reddy et al., 2001) and enzyme release by sperm cells (Michaut et al., 2001). Strikingly, forced expression of synaptotagmins in fibroblasts has been reported to induce formation of filopodia (Feany and Buckley, 1993), and SYT7 was recently shown to be essential for neurite extension of neuronal cells (Arantes and Andrews, 2006).

Synaptotagmin 3 (SYT3) is localized in nerve cells on synaptic plasma membranes, but not in synaptic vesicles (Butz et al., 1999). In pancreatic beta cells, however, it has been found on both secretory granules (Mizuta et al., 1997) and the plasma membrane (Gut et al., 2001). In pancreatic cells, SYT3 plays a role in Ca²⁺-induced insulin exocytosis (Gao et al., 2000). In mast cells, SYT3 is responsible for the formation and delivery of internalized cargo from early endosomes to the perinuclear recycling compartment (Grimberg et al., 2003). So far, a role of SYT3 (or any other SYT) in T cells has not been addressed.

Given that migration depends on vesicular transport, and vesicle fusion can be regulated by synaptotagmins, we hypothesized that synaptotagmins are involved in T cell migration. Here we show that T cells express SYT3 and that chemokine-induced migration is greatly diminished in cells with impaired SYT3 function or reduced SYT3 levels. Strikingly, SYT3 was not localized in vesicles or at the plasma membrane, but mainly in multivesicular bodies (MVBs). Impaired SYT3 function led to impaired CXCR4 recycling and consequently reduced CXCR4 surface levels. We propose that SYT3 is required for CXCR4 recycling via MVBs and thereby for maintaining CXCR4 surface levels that are sufficiently high to mediate T-cell migration.

To test whether any of the synaptotagmin family members could play a role in T cell migration we first assessed their expression in TAM2D2 cells and in murine T cells by RT-PCR. The T-cell hybridoma TAM2D2 (Roos et al., 1985) expresses CXCR4 (Zeelenberg et al., 2001) and is highly migratory and invasive, like the activated T cell from which it was generated. As shown in Fig. 1A, of 11 synaptotagmins (1-11) SYT3 is the only one expressed in TAM2D2 cells. Also normal T cells express SYT3, as shown in Fig. 1B. The presence of SYT3 protein was shown by western blotting (see below).

To assess whether synaptotagmins play a role in chemokine-induced migration, we expressed the isolated C2B domain of SYT1 in the TAM2D2 cells. This C2B domain inhibits oligomerization of all synaptotagmin isoforms tested (Desai et al., 2000). We used a retroviral vector that causes correlated expression of EGFP. To obtain cells with high or medium C2B expression, transfectants with high [C2B(h)] or medium [C2B(m)] EGFP levels were FACS sorted (Fig. 2A). Of the control TAM2D2 cells, almost 35% migrated towards CXCL12. By contrast, the C2B(h) cells did not migrate at all (Fig. 2B). Migration of C2B(m) cells was partially reduced. Clearly, the C2B domain has a major impact on CXCL12-induced migration. We could not demonstrate C2B expression directly since both N- and C-terminal tags reduced the effect, and because we did not succeed in raising an antiserum against C2B. However, the clear difference between cells with medium and high GFP levels that correlate with C2B, and lack of effect of the empty vector, show that inhibition is due to the C2B domain.

The C2B domain binds calcium, so it might influence migration by acting as a calcium sink. To test this possibility, we expressed a mutant C2B (K326,327A) that does not inhibit oligomerization and fusion but still binds calcium (Desai et al., 2000). This mutant will be referred to as C2B(KA). Using FACS sorting we selected a population with similar levels of C2B(KA) as the cells expressing the non-mutated C2B, as measured by EGFP levels (Fig. 2A). CXCL12-induced migration of these cells was similar to the TAM2D2 cells (Fig. 2B), showing that inhibition by C2B was not due to calcium binding.

The T-cell hybridoma cells rapidly invade monolayers of rat embryo fibroblasts (REFs). We have shown previously that this depends on CXCL12 bound to the surface of the fibroblasts and thus on CXCR4 signaling (Zeelenberg et al., 2001). Invasive capacity in this in vitro model correlates with wide-spread dissemination to many tissues in vivo (La Rivière et al., 1988). As shown in Fig. 2C, invasion was completely blocked in C2B(h) cells. By contrast, C2B(KA), expressed at the same high level, had only a minor effect.

To confirm SYT3 involvement in CXCL12-induced migration we reduced SYT3 levels by antisense mRNA. This reduced migration (Fig. 3A). The effect was transient: about a month after transduction cells regained their original SYT3 levels and migratory activity. This was again observed in a second experiment with an independently transduced cell population (Fig. 3B). The recovery of CXCL12-induced migration, corresponding to restoration of SYT3 levels, provides strong evidence that migration depends on SYT3.

One of the effects of CXCL12 necessary for migration is actin polymerization. Therefore, we tested whether actin polymerization is induced in C2B transfectants upon stimulation with CXCL12-coated beads. TAM2D2 cells reacted to attachment of a bead coated with fibronectin and CXCL12 with local F-actin accumulation, associated with engulfment of the beads. By contrast, C2B transfectants failed to respond (Fig. 4A). Of TAM2D2 cells, 80±7% responded, but only 10±8% of C2B transfectants responded. This was associated with accumulation of CXCR4 near the beads (Fig. 4B) in 75±10% of TAM2D2 cells and only 5±5% of the C2B transfectants. TAM2D2 cells did not respond to beads coated with fibronectin but not with CXCL12 (Fig. 4C). These results show that cells with impaired SYT3 function do not respond to CXCR4 signaling with the morphological changes that are required for migration.

Synaptotagmins are known for their role in neurotransmitter release, which is an extremely rapid event. Chemokines can mediate fast responses, in less than a second, to arrest cells in flow (Grabovsky et al., 2000). Therefore, we have considered the possibility that this fast response involves rapid vesicle fusion, regulated by SYT3. To test this, we employed REF monolayers that we also used for invasion assays and that have CXCL12 bound to the upper cell surfaces. TAM2D2 cells and C2B transfectants, subjected to steady flow over a REF monolayer, adhered equally well (Fig. 5). This adhesion was strongly inhibited by the CXCR4-blocking peptide TC14012 (Tamamura et al., 2001), showing that it was dependent on CXCR4 signaling. These results indicate that SYT3 is not required for this rapid response, and that CXCR4 signals are at least partly intact in cells with impaired SYT3 function.

Since SYT3 is apparently not involved in rapid responses, we wondered whether it is localized near the membrane, as in neuronal or pancreatic cells (Butz et al., 1999; Mizuta et al., 1997). To investigate this, we expressed a SYT3-GFP fusion protein in TAM2D2 cells or stained TAM2D2 cells with SYT3 antibodies. As shown in Fig. 6A,B, in both cases SYT3 was not found at the cell membrane but dispersed in a granular pattern throughout the cell. Also immunoelectron microscopy revealed very little SYT3 at the cell membrane, or in most of the cytosol, including the Golgi area and mitochondria. By contrast, most of the SYT3 was found in multivesicular bodies (MVBs; Fig. 6C), as demonstrated by quantification of the density of gold particles (Fig. 6D). This suggested a role for SYT3 in degradation and/or recycling.

Since MVBs are involved in protein degradation, we tested whether SYT3 is involved in CXCR4 degradation by comparing CXCR4 levels in TAM2D2 cells and in C2B transfectants by western blotting (Fig. 7A). The levels were not different, arguing against a role of SYT3 in degradation. However, FACS analysis revealed that the surface level of CXCR4 was 30-50% lower in C2B cells (Fig. 7B). By contrast, surface levels of other receptors that we tested (the integrin LFA1, insulin-like growth factor 1 receptor (IGF1R) and autocrine motility factor receptor (AMFR) were not different, indicating that SYT3 is specifically required for maintenance of CXCR4 at the surface of the cell.

To test whether the reduced surface levels of CXCR4 are the cause of the migration defect, we assessed CXCL12-induced migration of C2B transfectants that also overexpressed CXCR4-GFP. These cells exhibit increased surface levels of CXCR4 (Fig. 7C). As shown in Fig. 7D, migration was in fact restored in these cells.

Since CXCR4 levels in TAM2D2 cells and C2B transfectants were not different, the reduction of CXCR4 surface levels could be due to impaired receptor recycling. To examine this, we compared CXCR4 surface levels of TAM2D2 cells and C2B transfectants after CXCL12-induced internalization and subsequent incubation in medium without CXCL12, while synthesis of CXCR4 was suppressed by cycloheximide. CXCR4 was downregulated upon CXCL12 stimulation, and similarly in both cell types, showing that SYT3 function was not required for endocytosis. In TAM2D2 cells, the amount of CXCR4 on the cell surface was restored to the original level after overnight incubation. By contrast, CXCR4 surface levels recovered only partially in the C2B transfectants (Fig. 7E). Western blotting provided no indication for extensive degradation of CXCR4 induced by CXCL12 treatment, in either control TAM2D2 cells or in the C2B transfectants (Fig. 7F).

To elucidate the role of SYT3 in recycling, we examined whether its localization in the different compartments of the recycling pathway was different in C2B transfectants, compared with TAM2D2 cells. For this we used TAM2D2 cells stained with antibodies against SYT3, and SYT3-GFP transfectants. Both cell types were stained with antibodies against markers for these compartments. SYT3 did not colocalize with either the early endosome marker EEA1 (Fig. 8A) or the late endosome marker M6PR (Fig. 8B). Also RAB11, a marker for the perinuclear endosome recycling compartment, did not colocalize (data not shown). By contrast, SYT3 (partially) co-localized with LAMP1, a marker for lysosomes and MVBs (Fig. 8C), confirming our previous results with immunoelectron microscopy (see Fig. 6C). This (co)localization was not different after treatment with CXCL12. The C2B domain had no effect on SYT3 localization.

Moreover, in both TAM2D2 cells and C2B transfectants, SYT3, CXCR4 and LAMP1 were concentrated in the same cell compartment (Fig. 8D), suggesting that impaired receptor recycling in C2B transfectants is not due to altered localization of SYT3 and/or CXCR4. Additionally, we examined the localization of CXCR4 in TAM2D2 and C2B cells expressing CXCR4-GFP using immunoelectron microscopy with GFP antibodies. Thus, we confirmed the presence of CXCR4 in MVBs (Fig. 8E). Again, there was no obvious difference between control cells and C2B transfectants.

Synaptotagmins are proteins that control rapid fusion of vesicles with membranes. We have shown here that interference with fusion mediated by SYT3 in T cells strongly inhibits migration induced by the CXCR4 chemokine receptor as well as CXCR4-dependent invasion. Surprisingly, it did not affect CXCR4-dependent arrest of cells in flow, indicating that SYT3 is not required for rapid chemokine signaling responses. Furthermore, SYT3 was not localized at or near the plasma membrane. By contrast, it was mainly associated with multi-vesicular bodies (MVBs). Inhibition of SYT3 function impaired recycling of CXCR4 and led to reduced surface levels of CXCR4 but not of several other surface proteins. We propose that in T cells, SYT3 is involved in recycling of CXCR4 through MVBs and that this is required for maintenance of CXCR4 surface levels that are sufficiently high to mediate CXCL12-induced migration. As we will discuss, it is conceivable that the newly recycled CXCR4 is the active form that is mainly responsible for migration signals.

Synaptotagmins are best known for their role in regulation of neurotransmitter release. However, several synaptotagmins are expressed in many different non-neuronal cell types, indicating a more general function. Indeed, SYT3 and SYT7 have been implicated in regulated insulin secretion in pancreatic cells (Gao et al., 2000; Mizuta et al., 1997) and SYT6 is involved in the release of enzymes from the acrosome of sperm cells that are necessary to penetrate the oocyte (Michaut et al., 2001). Furthermore, the repair of membrane defects by exocytosis of lysosomes is mediated by SYT7 (Martinez et al., 2000; Reddy et al., 2001), and recently SYT7 was shown to be required for neurite extension (Arantes and Andrews, 2006). We show here that SYT3 is expressed in T cells. Strikingly, SYT3 was the only one amongst SYTs 1-11 that we detected in the model T-cell hybridoma that we used here.

Synaptotagmins are integral membrane proteins that serve as Ca²⁺ sensors and induce vesicle fusion upon a rise in Ca²⁺ concentration (Chapman, 2002). The cytoplasmic part contains two C2 domains, C2A and C2B, both of which bind Ca²⁺. The Ca²⁺-induced conformational change enables the C2B domain to interact with C2B domains of other synaptotagmin molecules. This results in oligomerization, which is required for fusion of the docked vesicles. The isolated C2B domain binds to the C2B domains of intact synaptotagmins and thus impairs oligomerization in a dominant-negative fashion. In permeabilized PC12 cells, recombinant proteins consisting of only the C2B domain thus prevented neurotransmitter release (Desai et al., 2000). The C2B domain of SYT1 inhibits oligomerization of all synaptotagmins tested (Desai et al., 2000). We therefore expressed this C2B domain, expecting that it would inhibit oligomerization of SYT3, the only SYT present. The complete inhibition of migration by the C2B domain that we observed strongly suggested that synaptotagmin oligomerization is required for chemokine-induced migration.

Chemokines induce a rapid rise in intracellular Ca²⁺ concentration. This is mainly due to the activation of phospholipase C (Wu et al., 2000) that generates inositol trisphosphate (InsP₃) from phosphatidylinositol-4,5-bisphosphate [Ins(4,5)P₂]. This InsP₃ releases Ca²⁺ from the endoplasmic reticulum. Given the importance of increases in Ca²⁺ concentration, the inhibition by the C2B domain might be due to Ca²⁺-binding rather than impaired oligomerization. To exclude that possibility, we expressed a C2B mutant in which two lysines, which are essential for oligomerization (Chapman et al., 1998), were replaced by alanines. This mutant binds Ca²⁺ in a similar manner as intact C2B (Desai et al., 2000). The mutant, expressed at similar levels, did not inhibit CXCR4-dependent migration at all. It did inhibit CXCR4-dependent invasion somewhat, but much less than intact C2B when expressed at the same high levels, perhaps because of effects of Ca²⁺ buffering on other pathways required for invasion. This result indicates that the main effect of the C2B domain on CXCR4-dependent migration is the result of inhibition of synaptotagmin oligomerization.

To test the role of SYT3 in CXCL12-induced migration directly, we reduced SYT3 levels by expression of antisense mRNA. This approach was previously shown to result in long-lasting reduction of SYT3 levels in RBL cells (Grimberg et al., 2003). Also in the TAM2D2 T-cell hybridoma cells we used, antisense transfection resulted in substantial reduction of SYT3 levels. This reduced SYT3 was associated with a decrease in CXCR4-induced migration, in line with results obtained with the C2B transfectants. SYT3 reduction was not stable, however, and cells rapidly regained their original SYT3 level. At the same time the cells regained their original migratory capacity. This provides strong evidence for an essential role of SYT3 in CXCR4-dependent T-cell migration.

We also observed that the C2B domain affected actin polymerization induced by CXCL12. This is an essential step in the migration process and results in striking changes in T-cell morphology, from round to elongated with a uropod and a leading edge and the redistribution of integrins, chemokine receptors and signaling molecules (Sanchez-Madrid and del Pozo, 1999). CXCL12 induces accumulation of CXCR4 (van Buul et al., 2003) and F-actin at the leading edge (Takesono et al., 2004), the latter nicely visualized with CXCL12-coated beads. Using such beads, we could induce CXCL12 and CXCR4 accumulation in TAM2D2 cells, but not in the C2B transfectants. This shows that SYT3 is required for CXCL12-induced actin polymerization and might explain its effect on migration and invasion.

Synaptotagmins are involved in very rapid processes such as neurotransmitter or insulin release, whereas actin polymerization and invasion take minutes to hours. A fast response to chemokines is the arrest of cells in flow (Grabovsky et al., 2000), as required, for example, in high endothelial venules in lymph nodes. It seemed likely that SYT3 is involved in this fast response. Strikingly, however, we did not observe any reduction in CXCR4-dependent adhesion of the C2B transfectants to cell monolayers in flow. Next, we observed that CXCR4 levels were reduced by up to 50% in the C2B transfectants, as compared to the wild-type TAM2D2 cells. Apparently, these lower levels were sufficient for arrest in flow but not for induction of actin polymerization and migration.

SYT3 had been shown to be localized to the plasma membrane in neurons (Butz et al., 1999) and to secretory granules in pancreatic beta cells (Brown et al., 2000; Gao et al., 2000). In rat basophilic leukemia cells, 70% of SYT3 colocalized with early endosomal markers such as early endosome antigen 1 (EEA1), and 30% with secretory granules (Grimberg et al., 2003). In the T cells, however, SYT3 was not localized at or near the membrane. Clearly, this excluded a role in any rapid fusion event at the plasma membrane that might be involved in rapid responses to chemokines. SYT3 also did not co-localize with early (EEA1) or late (M6PR) endosomes, but it did co-localize with LAMP1, a marker for lysosomes and multi-vesicular bodies. These organelles (lysosomes/MVBs) are involved in protein degradation but may also serve as `secretory lysosomes' (Blott and Griffiths, 2002). Their presence has been described mainly in hematopoetic cells, e.g. in cytotoxic T cells that use these secretory lysosomes to store cytolytic proteins used to destroy target cells (Peters et al., 1989), as well as FAS ligand (Blott et al., 2001). In dendritic cells, major histocompatibility complex class II molecules interact with an antigen in an `MHC class II compartment', which has characteristics of a secretory lysosome (Peters et al., 1991). Localization of SYT3 in MVBs suggested a role in protein degradation or recycling. The reduced surface levels of CXCR4 in C2B transfectants could be due to increased endocytosis or degradation, or to reduced recycling or exocytosis. The total CXCR4 protein level in C2B transfectants, as measured by western blotting, was comparable with that in wild-type TAM2D2 cells, arguing against increased degradation. A general defect in recycling is also improbable, since surface levels of other receptors, namely AMFR and IGF1R as well as the integrin LFA-1, were not affected by the impaired SYT3 function. CXCL12-induced CXCR4 internalization was equally efficient in TAM2D2 cells and C2B transfectants. This argues against a role of SYT3 in endocytosis, as might be expected since it is not localized to components of the endocytic pathway.

Internalized CXCR4 recycles to the surface, but this process is generally considered to be inefficient: up to 30% after 150 minutes recovery in the human HeLa carcinoma cell line (Tarasova et al., 1998; Amara et al., 1997), and up to 50% after 60 minutes in CEM human lymphoblastoid T cells (Amara et al., 1997). In the TAM2D2 T-cell hybridoma cells we observed slow but efficient recycling, leading to complete recovery after 16 hours, and this recycling was impaired in the C2B transfectants, indicating that SYT3 is essential for this recycling. It is conceivable that recycling CXCR4 passes through MVBs/lysosomes, since intracellular CXCR4 is mainly localized in MVBs, together with SYT3. A role of SYT3 in recycling has previously been described for RBL rat basophilic leukemia cells. In those cells, however, SYT3 is localized to early endosomes as well as secretory granules, and SYT3 was proposed to be involved in delivery of internalized cargo to the perinuclear endocytic recycling compartment (Grimberg et al., 2003). Clearly, SYT3 is localized to a different compartment in T cells and may therefore be involved in a different type of recycling. The CEM lymphoma cells may have lost components of this pathway. We have also noted that SYT3 is absent from macrophages (data not shown) that do respond to CXCR4 signals. Regulation of CXCR4 function by SYT3 may therefore be restricted to T cells.

In this context, it is noteworthy that T cells were recently shown (Booth et al., 2006) to contain patches in the plasma membrane with characteristics of exosomes, vesicles released upon fusion of MVBs with the plasma membrane (Fevrier and Raposo, 2004). Retroviruses such as HIV bud in T cells from these plasma membrane patches rather than from the membrane of endosomal structures as in macrophages. This implies transport of vesicles from MVBs to the T-cell surface, and thus backfusion of internal vesicles with the limiting membrane of the MVBs, followed by pinching off of vesicles that move to the plasma membrane, in a manner similar to that proposed for recycling of antigen-loaded MHC class II molecules (Zwart et al., 2005). Given its location on the internal vesicles, SYT3 might be involved in the control of the backfusion process.

We observed a complete inhibition of CXCL12-induced migration whereas CXCR4 levels were reduced by 30-50%. This is in line with the dose-dependence of CXCL12-induced polarization (Vicente-Manzanares et al., 1998), the correlation of increases in intracellular Ca²⁺ concentrations with CXCR4 levels (Princen et al., 2003), and the positive correlation between CXCR4 levels and actin polymerization, integrin activation, adhesion and migration (Ding et al., 2003). However, our results would then suggest a rather high threshold for CXCR4 activation. Indeed, migration is restored by a shift to the original CXCR4 level in C2B transfectants that overexpress CXCR4-GFP. This may, however, also be due to incomplete inhibition of recycling by the C2B domain, allowing enough of the overexpressed receptor to recycle.

Therefore, an alternative to be considered is that not all CXCR4 molecules on the surface are equal and that migration signaling is mainly due to CXCR4 that is newly inserted into the membrane. CXCL12 binds CXCR4 in T cells only when present in lipid rafts, colocalized with the GM1 glycolipid (Nguyen and Taub, 2002), and only part of CXCR4 is colocalized. CXCR4 is not as strongly associated with rafts as, for example, CCR5 (Venkatesan et al., 2003). A possible scenario is that CXCR4 is inserted in the right environment but detaches later from the rafts and is thus inactivated. The interesting implication is that control of membrane recycling by the calcium sensor SYT3 allows for regulation of the number of active CXCR4 molecules by calcium signals.

The mouse T-cell hybridoma TAM2D2 was generated by fusion of noninvasive BW5147 lymphoma cells with normal activated T lymphocytes (Roos et al., 1985), and was cultured as described previously (La Rivière et al., 1988). Rat embryo fibroblasts (REFs) were cultured in DMEM (Gibco-BRL) and 10% newborn calf serum (Gibco/Invitrogen). They were used for invasion assays between passages 5 and 15. The virus-packaging cell line Phoenix (http://www.stanford.edu/group/nolan/retroviral_systems) was cultured in DMEM + 10% FCS (Gibco /Invitrogen) and 0.584 g/l L-glutamine (Gibco-BRL).

Brain tissue of a BALB/c nude mouse or 10⁶ TAM2D2 cells were directly lysed in 1 ml RNAzol (Teltest, Friendswood, TX). Murine T cells were isolated from a spleen using magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's protocol. The purity of the T-cell population was ∼95% as tested by FACS analysis with CD3 antibody (BD Biosciences). Total RNA was extracted and RT-PCR was performed using a one-step RT-PCR kit (Clontech, Palo Alto, CA) and the primers listed in Table S1 in supplementary material. As a control, beta-actin RNA was amplified.

The mouse SYT1 C2B domain (residues 302-421) was generated by PCR on mRNA derived from mouse brain. The C2B(K326,327A) mutant was generated using the Quick-change site-directed mutagenesis kit (Stratagene, La Jolla, CA). The full-length synaptotagmin 3 (Syt3) cDNA was generated by PCR from mouse brain mRNA. The SYT3-GFP fusion construct was made by subcloning the Syt3 cDNA as an EcoRI-ApaI fragment into the N2pEGFP vector (Clontech). The CXCR4-GFP fusion construct was a generous gift from P. Hordijk (Sanquin Research, Amsterdam, The Netherlands). The C2B and C2B (K326,327A) cDNAs were cloned into the retroviral vector pLZRS-IRES-puro-EGFP. This was made from the pLZRS-IRES-zeo vector (Michiels et al., 2000) by replacing the zeocin resistance cDNA with a cDNA encoding a puromycin resistance-enhanced green fluorescence protein (EGFP) fusion protein. It contains an internal ribosomal entry site (IRES), so EGFP levels correlate with levels of the proteins encoded by the inserted cDNAs. The Syt3 cDNA was cloned into this same vector, but in the antisense orientation, to generate antisense SYT3. The SYT3-GFP and CXCR4 cDNAs were cloned into pLZRS-IRES-zeo. The vector plasmids were transfected by calcium phosphate precipitation into the Phoenix virus-packaging cell line. After 8 hours, the medium was refreshed, and 48 hours later the virus supernatant was collected and used to infect the TAM2D2 cells. Three days later, 0.4 μg/ml puromycin (Sigma, St Louis, MO, USA) or 0.3 mg/ml zeocin (Invitrogen, Carlsbad, CA, USA) was added. After a few days the cells transduced with GFP-containing constructs were FACS-sorted to select bulk populations with high EGFP expression. To express both C2B and GFP fusion proteins in the same cell, the C2B cDNA was also cloned into pLZRS-neo, containing a neomycin-resistance gene. Transduced cells were selected in 1 mg/ml neomycin (Geneticin, G-418 sulphate, Invitrogen, UK). Cells with high expression were selected by subcloning, and C2B levels in the clones assayed by quantitative RT-PCR. The GFP fusion constructs were next transduced into these cells, selected with zeocin, and FACS-sorted for GFP expression.

SDS-PAGE-separated cell lysates were blotted to nitrocellulose, which was then blocked with 1% BSA and 3% non-fat dried milk. The membranes were incubated overnight with the primary antibodies at 4°C, followed by incubation with secondary antibodies coupled to horseradish peroxidase (Amersham Life Sciences, Little Chalfont, UK). Antibodies were directed against synaptotagmin 3 (a generous gift from B. A. Wolf and R. A. Young (Gao et al., 2000), CXCR4 (2B11, a generous gift from R. Förster, Hannover, Germany) and actin (Abcam, Cambridge, UK). Stained proteins were visualized by enhanced chemiluminescence (ECL kit, Amersham).

In addition to the antiserum that we obtained as a gift (see above), we also used SYT3 antibodies that we generated ourselves against the same C-terminal mouse SYT3 peptide (C)GGKGLSEKENSE, with the N-terminal cysteine added for conjugation to KLH. The conjugate was injected into rabbits following standard procedures. This antiserum was used specifically for confocal microscopy (see below).

Migration assays were performed as described previously (Soede et al., 1998). Briefly, Transwells with 5 μm pores were treated for 2 hours with 0.5% ovalbumin at room temperature. The lower chamber was filled with 250 μl RPMI 1640, supplemented with 0.1% ovalbumin and 100 ng/ml CXCL12 (PeproTech Inc., Rocky Hill, NJ, USA). Cells were kept in fresh medium for 30 minutes at 37°C and washed with ice-cold serum-free medium. 10⁵ cells were added to the upper chamber of the Transwells in RPMI 1640 with 0.1% ovalbumin. After incubation for 2 hours at 37°C and 5% CO₂, the migrated cells in the lower chamber were counted manually. Invasion assays were performed as described previously (La Rivière et al., 1988). Briefly, confluent rat embryo fibroblast (REF) monolayers in 24-well plates and TAM2D2 cells or transfectants were washed and the latter were added to the monolayers in medium supplemented with 10% FCS. After incubation for 1 or 4 hours at 37°C and 5% CO₂, the monolayers were extensively washed and fixed with 2% paraformaldehyde. The invaded cells were counted using phase-contrast microscopy.

We used a previously described assay (Takesono et al., 2004). Briefly, sulfate latex beads (Interfacial Dynamics, Tualatin, OR) were coated with 20 mg/ml of fibronectin (Sigma) or both fibronectin and 100 ng/ml CXCL12 at 4°C overnight and then washed three times with 3% BSA in PBS and resuspended in serum-free RPMI medium containing 20 mM HEPES at the final concentration of 10⁸ beads/ml. TAM2D2 cells and C2B transfectants were washed three times with PBS and suspended in RPMI at 10⁷ cells/ml, and 100 μl was plated per coverslip and incubated at 37°C for 30 minutes to let the cells adhere. Then 20 μl of beads were incubated with the cells for 10 minutes at 0°C. Next, the coverslips were warmed to 37°C in a water bath for 10 minutes, and then placed on ice. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 in PBS for 4 minutes at room temperature, washed with PBS, and incubated with blocking solution (3% BSA in PBS) for 20 minutes at room temperature. Cells were stained with Rhodamine-phalloidin (Molecular Probes) or with anti-CXCR4 antibody (2B11) for 30 minutes and washed twice with PBS. The CXCR4 mAb-treated cells were stained with Alexa Fluor 488-conjugated secondary antibody. The coverslips were mounted with Mowiol medium (Hoechst, Frankfurt am Main, Germany) and analyzed by confocal microscopy.

This assay was conducted as described previously (Netelenbos et al., 2003). Briefly, adhesion of TAM2D2 cells and C2B transfectants was measured under conditions of flow in a parallel plate flow perfusion chamber. REFs were cultured to confluence overnight on glass slides. Part of these monolayers were then pre-incubated 1 hour prior to perfusions with 1 μM TC14012, aCXCR4-blocking peptide (Tamamura et al., 2001). This was synthesized by the Peptide facility of the Netherlands Cancer Institute. Cells in HEPES flow buffer (20 mM HEPES, 132 mM NaCl, 6 mM KCl, 1.2 mM K₂HPO₄, 1 mM MgSO₄, 5 mM glucose, 1 mM CaCl₂ at pH 7.4) at a concentration of 2×10⁶/ml were prewarmed to 37°C. Part of the cell samples were pre-incubated with CXCR4-blocking peptide. The cells were aspirated with a syringe pump through the flow chamber, mounted on a microscope stage equipped with a video camera and recorder. After 1 minute perfusion with flow buffer, cells were added at a shear stress of 1.0 dyn/cm². Thereafter, perfusion was continued for 20 seconds with cell-free buffer to remove loosely attached cells. Video recordings were made continuously and analyzed afterwards. The total number of cells was measured in a minimum of 30 randomized high power fields, by computer analysis with Optimas 6.1 software (Media Cybernetics Systems, Silver Spring, MD). Perfusions were performed in triplicate on at least three separate occasions.

GFP levels were measured directly. Primary antibodies used were directed against: CXCR4 (2B11), LFA-1 (M17/4, ATCC), AMFR (3F3A, a generous gift from A. Raz, Detroit, MI), and IGF1R (Oncogene, San Diego, CA). Cells were stained with phycoerythrin (PE)- or Cy5-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories Inc., West Grove, PA). The analyses were performed on a Becton Dickinson FACScan using CellQuest software.

Cells (5×10⁵) were washed and incubated for 2 hours in culture medium with CXCL12 (100 ng/ml). Then, part of the cells was washed again, stained with 2B11 CXCR4 mAb and fixed. The other cells were incubated overnight in culture medium with 10 μg/ml cycloheximide (Sigma) and then stained with 2B11 and fixed, and finally all the cells were subjected to FACS analysis.

Cells were allowed to adhere to coverslips by incubating them for 30 minutes in medium without serum. Cells were permeabilized with 0.1% Triton X-100 for 10 minutes, blocked in PBS containing 3% BSA for 30 minutes, and probed with antibodies against SYT3, CXCR4 (2B11), EEA1 (Transduction Laboratories), M6PR (Transduction Laboratories), or LAMP1 (R&D), and Alexa Fluor 568- or Cy5-conjugated secondary antibodies (1:1000, Molecular Probes). Coverslips were then mounted on slides and sealed. Confocal images were acquired on a Leica TCS NT laser-scanning confocal microscope.

Cells transfected with CXCR4-GFP or SYT3-GFP were fixed for 2 hours in paraformaldehyde (4%) in 60 mM Pipes, 25 mM HEPES, 2 mM MgCl₂, 10 mM EGTA, pH 6.9 and processed for ultrathin cryosectioning (Calafat et al., 1997). For immunolabeling, sections were incubated for 10 minutes with 0.15 M glycine in PBS and for 10 minutes with 1% BSA in PBS to block free aldehyde groups and prevent aspecific antibody binding. Sections were then incubated with GFP antibody (a generous gift from J. Neefjes, Amsterdam, The Netherlands) for 1 hour, followed by a protein A-conjugated 10 nm colloidal gold probe, all in 1% BSA in PBS. After embedding in a mixture of methylcellulose and uranyl acetate, sections were analyzed with a Philips CM10 electron microscope (Eindhoven, The Netherlands). For quantification, 25 cells were examined and gold particles counted in fields of 0.2 μm². In each cell we counted at least one field containing an MVB, two to three fields with mitochondria, two to three fields containing a Golgi apparatus, five fields of cytosol, and five fields of 0.2-μm-thick perimembrane areas (including cell membrane).

We are grateful to B. A. Wolf and R. A. Young (Philadelphia), R. Förster (Hannover), A. Raz (Detroit) and J. Neefjes (Amsterdam) for providing antibodies, and P. Hordijk (Amsterdam) for the CXCR4-GFP construct. We thank Floortje Kessler for help with the adhesion in flow experiments, Laurens Oomen and Lenny Brocks for help with confocal microscopy, and Anita Pfauth and Frank van Diepen for FACS sorting.