Distinct roles of the C₂A and the C₂B domain of the vesicular Ca²⁺ sensor synaptotagmin 9 in endocrine β-cells

Antibodies

The following antibodies were employed: MAb (monoclonal antibody) anti-TfR (transferrin receptor; H684, Sigma), MAb anti-syt2, MAb anti-syt9 (BD Biosciences), anti-syntaxin (clone HPC1, Sigma), anti-β-actin (Abcam), polyclonal anti-eGFP (enhanced green fluorescent protein; SantaCruz Biotechnologies) and anti-SNAP-25 (25 kDa synaptosome-associated protein; Sternberger Monoclonals, Covance Research Products). Antibodies against chromogranin A, insulin, VAMP (vesicle-associated membrane protein) 2 and synaptophysin were as previously described [25–27]. HRP (horseradish peroxidase)-linked anti-mouse or anti-rabbit secondary antibodies were purchased from Amersham Biosciences. A Cy3-labelled anti-mouse secondary antibody was obtained from Jackson ImmunoResearch.

Plasmids

The plasmid peGFP-Rab7 was provided by B. van Deurs (Department of Medical Anatomy, Panum Institute, University of Copenhagen, Copenhagen, Denmark), peGFP-CD63 by G. Griffiths (Sir William Dunn School of Pathology, University of Oxford, Oxford, U.K.), mRFP-Phogrin by W. Almers (Vollum Institute, Oregon Health Science University, Portland, OR, U.S.A.), BONT/C and BOTN/E by H. Niemann, H. Binz and T. Binz, (Institut für Biochemie, Medizinische Hochschule Hannover, Hannover, Germany) and peGFP-PKC by C. Larsson (Department of Laboratory Medicine, University of Lund, Sweden). The vector pRSET-BmStrawberry was donated by R. Tsien [28]. The plasmid pKS-Syt2 was as described previously [15].

Plasmid construction and mutagenesis

cDNA encoding mouse syt9 was generated by RT (reverse transcriptase)-PCR from total brain RNA and inserted in a pGEMT-T cloning vector. Truncated syt2 or syt9 was generated by PCR from pGEMT-syt9 or pKS-syt2 plasmids respectively, using the corresponding sense and antisense primers with an EcoRI site (underlined) inserted in the forward primer and a BamHI in the reverse primer. Primers were as follows: syt9C₂AB (amino acids 77–386), sense primer 5′-TTGGAATTCATGCTGGGCCGGAGTTACATAG-3′, antisense primer 5′-TCTGGATCCTCCGGGTGCAGGTATTGGCC-3′; syt9C₂A (77–233): sense primer 5′-TTGGAATTCATGCTGGGCCGGAGTTACATAG-3′, antisense primer 5′-TCTGGATCCAGCCACCTGCAGCTCTCTC-3′; syt9C₂B (215–386): sense primer 5′-TTGGAATTCATGAGTTCAGTGAACCTGG-3′, antisense primer 5′-TCTGGATCCTCCGGGTGCAGGTATTGGCC-3′; syt2C₂AB (101–422): sense primer 5′-TTGGAATTCATGAAGGGCAAAGGCATGAAG-3′, antisense primer 5′-AGTGGATCCTTGTTCTTGCCCAGAAGAG-3′. The amplified coding sequences were inserted in the EcoRI/BamHI cloning sites either of peGFPN3 (Clontech Laboratories) for syt9 or in the EcoRI/BamHI cloning sites of peGFPN1 for syt2. Syt9C₂AB-mStrawberry was constructed by PCR amplification of syt9C₂AB using the primers 5′-CCCAAGCTTATGCTGGGCCGGAGTTACATAG-3′ and 5′-TCTGGATCCGGTCCGGGTGCAGGTATTGGCC-3′, HindIII and BamHI sites are underlined. After digestion with HindIII/BamHI, the amplicon was inserted in-frame into pCDNA₃-mStrawberry. The pcDNA3-mStrawberry vector was constructed by insertion of the mStrawberry coding sequence into the restriction sites for BamHI and EcoRI of pcDNA3. All constructs were verified by sequencing of both strands. Point mutations were introduced into the syt9 coding sequence leading to the substitution of an aspartic acid residue to an asparagine residue using the Quikchange® site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. The forward primers corresponding to each mutation were as follows (exchanged nucleotides given in bold) D145N: 5′-CCTAGGAGGTTCCTCAAATCCCTATGTTAGTGTCT-3′; D197,199N: 5′-GGTCATGGCGGTGTATAACTTTAATCGGTTCTCCCGCAACG-3′; D330,332N: 5′-CTGACTGTTCTGAATTATAACAAACTGGGGAAGAATGAG-3′.

Cell culture, transient transfection and biochemical membrane translocation assay

MIN6 cells (passage 21–30) and HIT-T15 cells (passage 75–85) were cultured as described previously [28,29]; after plating (72 h), MIN6 and HIT-T15 cells were transfected with plasmids using Lipofectamine™2000 (Invitrogen) or JetPEI™ (PolyPlus-Transfection) respectively according to the manufacturer's instructions. HIT-T15 cells seeded at 5×10⁴ cells/ml 72 h prior transfection were grown in 24-well dishes and transfected with plasmid encoding for different cytosolic syt9 fusion proteins. After transfection (72 h), cells were washed with PBS (pH 7.4), detached by PBS containing 10 mM EDTA and centrifuged at 3000 g for 5 min at 4 °C. Pelleted cells were resuspended and incubated in intracellular buffer [140 mM L-glutamic acid/monopotassium salt, 5 mM NaCl, 7 mM MgSO₄, 20 mM Hepes (pH 7), 1 μg/ml aprotinin, 1 μg/ml pepstatin, 1 μg/ml leupeptin and 0.5 mM PMSF] containing either 2 mM EGTA (non-stimulating conditions/in the absence of Ca²⁺) or 2 mM CaCl₂ with 10 μM ionomycin (stimulating conditions/in the presence of Ca²⁺) for 5 min at 37 °C. Cells were then disrupted by sonication using pulses of 10 s for 1 min at 4 °C [27,29]. The cell debris and the nuclei were eliminated by a centrifugation at 500 g for 10 min at 4 °C. Supernatants were centrifuged at 55000 rev./min for 1 h at 4 °C (using a TLA120.1 rotor) to separate the membrane and the cytosolic fractions. The corresponding samples were suspended in Laemmli buffer, heated at 95 °C for 5 min and resolved on SDS/PAGE (10% gels). Proteins were electroblotted on to PVDF, stained, blocked and incubated with the appropriate antibodies [27,29]. The signal obtained with the ECL (enhanced chemiluminescence) kit Lumi-Light^Plus Western blotting substrate (Roche Diagnostics) was detected using a CCD (charged-coupled-device)-camera (Roperts Scientific) and quantified with FluorChem v2.00 (Alpha Innotech) [26].

In vivo membrane translocation assay

MIN6 cells (1×10⁴ cell/cm²) grown on 25 mm round glass coverslips were washed twice with 1 ml of KRB (Krebs–Ringer buffer), pH 7.4, supplemented with 0.05% BSA and 3 mM glucose at 37 °C [26,27]. In experiments using digitonin, cells were incubated in intracellular buffer supplemented with EGTA (0.4 mM) and then stimulated in the same buffer with a defined Ca²⁺ concentration in the presence of 30 μM digitonin. Buffered Ca²⁺ solutions were obtained as described previously [26,27] or, in the case of concentrations of free Ca²⁺ above 10 μM, calculated using Winmaxc (http://www.stanford.edu/∼cpatton/maxc.html) employing the chelators NTA (nitrilotriacetic acid) and HEDTA (N-hydroxyethylethylenediaminetriacetic acid) [29]. When ionomycin was used, cells were first incubated in KRB (pH 7.4) supplemented with 0.05% BSA and 3 mM glucose and stimulation was performed with 1 mM Ca²⁺ and 10 μM ionomycin in KRB. Cells were kept on a heated microscope stage during acquisition as described previously [26,27]. The stimulating buffer was pressure ejected (≈20.7 kPa for 10 s) from a micropipette held at approximately 20 μm from the cell. Imaging of living cells was performed at 1 frame/0.5 s using an inverted microscope (Nikon TD300 equipped with a Z-drive) coupled to a monochromator (Till Photonics) and appropriate emission filters. Images were recorded by a CCD-camera (Micromax 1300Y HS, Roperts Scientific) using Metamorph software (Universal Imaging) [30]. The percentage of cells in which translocation was observed was determined. Due to leakage of unbound syt9-constructs through the pores generated by the detergent, translocation is an all-or-nothing event in digitonin-permeabilized cells.

Fluorescence microscopy of fixed cells

MIN6 cells seeded at 2.6×10⁴ cells/cm² were grown on 12-mm round glass coverslips and transfected with the corresponding plasmid. After transfection (72 h), cells were washed twice with KRB and exposed to either KRB (non-stimulating conditions) or KRB containing 5 mM Ca²⁺ and 10 μM ionomycin (stimulating conditions) for 10 min at room temperature (22 °C). Cells were then fixed for 20 min at room temperature in 4% paraformaldehyde/PBS in the absence (non-stimulating conditions) or presence of 5 mM CaCl₂ (stimulating conditions). Coverslips were either directly mounted or first stained with antibodies. In the latter case, coverslips were washed once with KRB and three times with PBS containing 2% BSA. Permeabilization was performed with 0.1% saponin in PBS containing 2% BSA for 30 min at room temperature and cells were stained with corresponding antibodies. Coverslips were mounted using an antifade kit (Vectashield® mounting medium, Vector Laboratories). Imaging was performed on a LMS 510 Meta confocal laser microscope (Zeiss). Co-localization of antibody staining was quantified using Metamorph software (Universal Imaging) by analysing localization and intensities of pixels (n=3) after thresholding. Values are given as means±S.E.M. [28,30]. Co-localization masks were generated with ImageJ using the plugin ‘RG2B co-localization’ with a minimum ratio between channels of 0.2 and thresholds of 26 (over 255) for each channel [30].

Molecular modelling

Simulations were carried out for following systems: (i) domain C₂B of syt1 (NM001033680, amino acids 272–421 for modelling) [32] with 12323 water molecules plus 5 chloride ions, giving a total of 38.544 atoms in an initial box size of 7.37 nm³; (ii) C₂B domain of syt9 (NM016908, amino acids 239–386) with 10967 water molecules plus 12 chloride ions giving a total of 34.388 atoms in an initial box size of 7.06 nm³. These systems were solvated with SPC (simple point charge) water and energy minimized. Molecular dynamic simulations were run using Gromacs [33]. A twin-range cut-off was used for longer range interactions: 1 nm for Van der Waals interactions and 1.8 nm for electrostatic interactions. The time step was 2 fs using Shake and NPT conditions were used to constrain bond lengths (i.e. constant number of particles, pressure and temperature in the simulation). A constant pressure of 100 kPa was employed in all three directions with a coupling constant of θ_r=1 ps [34]. Water and protein were coupled separately to a temperature bath at 300 K, with a coupling constant θ_r=0.1 ps. The system was equilibrated for 2 ns and the simulation run for another ns. Results were depicted as ribbon diagrams using VMD 1.8.4 (http://www.ks.uiuc.edu/Research/vmd/).

Statistical analyses

For the comparison of two groups, the Student's t test was used.

Biochemical characterization of syt9C₂ domains in situ

In our attempt to investigate the behaviour and translocation of the cytosolic domain of syt9 we have generated constructs and mutants encompassing the C₂A and the C₂B domain (see Figure 1) and compared it with syt2C₂AB. All constructs harboured a fluorescent protein at their C-terminus, either eGFP or Strawberry [28]. Moreover, point mutations were introduced into the Ca²⁺-binding sites. Indeed, replacement of cognate aspartic acid residues by asparagine residues in the C₂A domain of syt1 has been shown to partially (D¹⁴⁵, numbering for syt9) or completely (D^197/199 in syt9) abolish Ca²⁺-mediated effects of this domain [15,35]. Similarly, a double mutation in the Ca²⁺-binding site of the C₂B domain of syt1 (corresponding to D^330/332 of syt9) completely abolished its function in vivo [11]. The encoded proteins were expressed in insulin-secreting cells (Figure 1B) as demonstrated by the use of antibodies against syt9, which also recognized the endogenous forms (right-hand panel and middle panel), except for syt9C₂B. The protein expressed by this construct no longer contained the antigenic epitope for the anti-syt9 antibody, but still reacted with anti-eGFP (Figure 1B, middle panel). Note that degradation products of fluorescent proteins were not detected.

We subsequently used these constructs to examine the translocation of syt9 using a biochemical approach to characterize the behaviour of syt9C₂AB. To this end cells were incubated in the absence or presence of Ca²⁺ (2 mM) and ionomycin (10 μM) prior to their fractionation into supernatant (cytosol) and membrane pellets. As shown in Figure 2(A), a considerable amount of syt9C₂AB was already present at membranes in the absence of Ca²⁺ and the cation induced a complete shift of syt9C₂AB to the membrane fraction. Membrane binding of syt9C₂AB was sensitive to a high concentration of salt (Figure 2B), indicating the electrostatic nature of the interaction. Ca²⁺-sensitive synaptotagmins bind to membrane SNARE proteins such as syntaxin or SNAP-25 in a Ca²⁺-sensitive manner and this interaction is sensitive to the action of clostridial neurotoxins [36–38]. To test whether these SNARE proteins were involved in the translocation observed in the present study, we co-expressed syt9C₂AB and botulinum neurotoxin E or C prior to analysis. Neither of the two toxins altered the Ca²⁺-sensitive distribution of syt9C₂AB despite cleavage of syntaxin 1 and SNAP-25 (Figure 2C). Note that cleavage was not complete, as transient transfection led to expression only in a fraction of cells. Taken together these data demonstrate that syt9C₂AB translocates to membranes in response to Ca²⁺ in these insulin-secreting cells. This event occured independently of SNARE proteins and most likely implies ionic interactions with membrane phospholipids.

Next we addressed the role of the different domains of syt9C₂AB in translocation by biochemical means and compared first the C₂ domain of PKC (protein kinase C) with the C₂AB domain of syt2 and syt9 (Figure 3A). Interestingly both PKC-C₂α and syt2C₂AB did not bind to membranes in the absence of Ca²⁺ in contrast with syt9C₂AB. As expected, a rise in Ca²⁺ resulted in a complete translocation for all three constructs. The first C₂ domain of syt9 alone, syt9C₂A, was largely cytosolic in the absence of Ca²⁺ and fully translocated. It thus behaved as syt2C₂AB and PKC-C₂α. Note that a double band was observed (indicated with an arrowhead) as syt9C₂A migrated just below endogenous syt9 and both were stained by the antibody used. In contrast with the C₂A domain, the C₂B domain remained mainly cytosolic despite the increase in Ca²⁺. Mutation of Asp¹⁴⁵ to an asparagine residue in syt9C₂AB did not alter translocation, whereas mutation of Asp¹⁹⁷ and Asp¹⁹⁹ rendered syt9C₂AB insensitive to Ca²⁺ (Figure 3C). We also tested the role of the aspartic acid residues in the second C₂ domain known to be required for Ca²⁺ co-ordination in syt1 and its function in neuroendocytosis [12]. Contrary to the mutations in syt9C₂A, the mutations in the second C₂ domain did not influence the Ca²⁺-induced distribution of syt9C₂AB. However, syt9C₂AB_{D330N, D332N} exhibited a considerable increase in membrane binding in the absence of Ca²⁺ similar to the reported observations on the cognate mutation in syt1C₂AB [35].

Modelling of the syt9C₂B domain

To get more insight into the differences between syt1 and syt9 we modelled the structure of syt9C₂B using the published crystal structure of syt1C₂B as a template, and performed molecular dynamic simulations following equilibration of the system (Figure 4). The syt9 sequence showed a striking overall resemblance to syt1, though some differences were apparent. The main distinction concerned the β-sheets 3 and 4 (Figure 4) which were partially or completely absent in syt9C₂B. Sheet 4 contains seven lysine residues, which are generally unfavourable for β-sheet formation and modelling has probably failed to detect the sheets documented previously by NMR and crystallography for this reason (http://rsb.info.nih.gov/ij/ and [37]. As a control, we modelled syt1C₂B alongside syt9C₂B and again, both β-sheets were partially unfolded or were absent (results not shown). The sole additional, albeit minor, difference concerned the C-terminal acidic α-helix H2, which was shorter in syt9C₂B compared with syt1C₂B. In contrast, the main Ca²⁺ co-ordination sites were well-conserved and no difference was seen in the spatial arrangement around β-sheets 6 and 7, from which they emerge. Actually exchange of this portion in syt9C₂B by the corresponding sequence of syt1C₂B conferred ‘syt1-like’ biochemical properties to the mutated syt9C₂B. Thus differences in spatial arrangement do not seem to be responsible for the particular behaviour of syt9C₂B.

Translocation of syt9C₂ domains in living cells

These biochemical assays suggest that the C₂B domain does not positively influence translocation of syt9C₂AB but rather impedes it. However, this approach may be influenced by the lengthy fractionation protocol and does not allow determination of the nature of the membrane(s) syt9 is binding to. We therefore resorted to the observation of membrane translocation in living cells by the use of fluorescent videomicroscopy. Moreover, we combined this approach with the use of a detergent, digitonin, to permeabilize the membrane and the use of defined Ca²⁺ buffers to equilibrate intracellular concentrations of the cation [26,37]. A typical experiment is shown in Figure 5(A) depicting the translocation observed for syt9C₂A and syt9C₂AB. Subsequent to the stimulus and membrane attachment, we observed the disappearance of fluorescent proteins over time. This is most likely due to leakage of fluorescent proteins through the pores formed by digitonin as it was also observed for eGFP itself and only minor bleaching occurred (results not shown). Indeed, chelation of the small volume of Ca²⁺ ejected from the pipette by EGTA present in the bath solution will rapidly decrease the concentration of free Ca²⁺. Consequently, even translocated syt9C₂ constructs will detach and dilute in the bath. In the case of syt9C₂B we only observed disappearance of the fluorescence without any attachment to membranes. We also observed some nuclear localization of the fluorescent proteins. This probably represents transiently expressed full-length syt9C₂ constructs, as no degradation products containing eGFP were apparent on immunoblots (see Figure 1) and the proteins dissolved through the pores arguing against aggregation.

Syt2C₂AB and syt9C₂AB exhibited distinct Ca²⁺ sensitivities EC₅₀s at around 1 μM and 10 μM free Ca²⁺ respectively (Figure 5B). In contrast, syt9C₂A behaved as syt2C₂AB in terms of Ca²⁺ sensitivity in line with our observations in the biochemical assays (see above). Again syt9C₂B remained purely cytosolic and no evidence, even on close direct inspection during the experiment or by confocal microscopy (results not shown), could be found for translocation (Figure 5B). This approach also revealed that syt9C₂AB carrying the mutation D145N exhibited some degree of translocation but only at millimolar concentrations of free Ca²⁺, whereas translocation was completely abolished by the double mutation D197N/D199N. Again, the mutation D330N/D332N in the C₂B domain did not alter the Ca²⁺ sensitivity of syt9C₂AB. Thus in both the biochemical and the in vivo approaches the C₂B domain of syt9 actually reduced the Ca²⁺-sensitivity and the extent of translocation.

Syt9C₂B is required for late translocation to endosomes

On close inspection of films it became apparent that syt9C₂AB rapidly detached from the membrane once the perfusion with buffers containing elevated Ca²⁺ was stopped, and translocated to granular intracellular structures. However, as a large amount of syt9C₂AB leaked out of the cell when using digitonin, imaging of those structures was difficult. We therefore used ionomycin and Ca²⁺ as a stimulus to improve visualization of these structures (Figures 6A–6D). At about 5 s after the beginning of membrane translocation, staining of these structures became apparent. Most interestingly, the D330N/D332N mutant of syt9C₂AB translocated to the membrane, but never subsequently moved to intracellular structures, thus indicating a role for the C₂B domain in the second step. Syt9C₂B alone never translocated and both mutants of the C₂A Ca²⁺-binding site (syt9C₂AB D145N and D197N, D199N) failed to translocate to the plasma membrane under the conditions used here (intact cells, ionomycin) and were not found on intracellular structures (results not shown). It seems therefore that prior passage to the plasma membrane is required for localization on intracellular structures.

In view of the requirement of the C₂B domain for this event we attempted to characterize the compartment(s) involved by co-expression of fluorescent fusion proteins and the use of antibodies. As the fusion proteins used as markers were tagged with eGFP, we used syt9C₂AB linked to the fluorescent protein Strawberry (syt9C₂AB–S) known to be monomeric [28]. Syt9C₂AB–S behaved as syt9C₂AB–eGFP and co-localized with eGFP-tagged syt9 upon co-transfection (results not shown). To exclude a potential interaction between fluorescent proteins, we tested the combination of syt9C2AB–S and LPH–eGFP, a plasma-membrane marker [29,30] and SVP38–eGFP [39]. Neither of these constructs co-localized with syt9C₂AB–S suggesting that interactions between fluorescent proteins did not play any significant role (results not shown). A considerable number of compartments tested using antibodies also did not reveal any substantial degree of co-localization such as SNAP-25, syntaxin 1, VAMP2/synaptobrevin 2, insulin and chromogranin A. This excluded compartments harbouring these SNARE proteins or secretory granules as localization for syt9C₂AB–eGFP. In contrast, we observed a high degree of co-localization for the endosomal markers CD63–eGFP [40] and rab7–eGFP [41] in fixed cells (see Figures 7A and 7C). In addition, co-localization was evident for phogrin–eGFP, known to be recycled through endosomal pathways [42], and to a lesser degree for the TfR. We also determined whether co-localization can be observed in living cells by videomicroscopy (see Figure 7B). Indeed, a considerable degree of co-localization was apparent for Rab7–eGFP and syt9C₂AB–S after stimulation of cells with ionomycin. These observations are in line with an attachment of syt9C₂AB to endosomal membranes.