The C2 domains of granuphilin are high-affinity sensors for plasma membrane lipids

Abstract

Membrane-targeting proteins are crucial components of many cell signaling pathways, including the secretion of insulin. Granuphilin, also known as synaptotagmin-like protein 4, functions in tethering secretory vesicles to the plasma membrane prior to exocytosis. Granuphilin docks to insulin secretory vesicles through interaction of its N-terminal domain with vesicular Rab proteins; however, the mechanisms of granuphilin plasma membrane targeting and release are less clear. Granuphilin contains two C2 domains, C2A and C2B, that interact with the plasma membrane lipid phosphatidylinositol-(4,5)-bisphosphate [PI(4,5)P₂]. The goal of this study was to determine membrane-binding mechanisms, affinities, and kinetics of both granuphilin C2 domains using fluorescence spectroscopic techniques. Results indicate that both C2A and C2B bind anionic lipids in a Ca²⁺-independent manner. The C2A domain binds liposomes containing a physiological mixture of lipids including 2% PI(4,5)P₂ or PI(3,4,5)P₃ with high affinity (apparent K_d,PIPx of 2-5 nM), and binds nonspecifically with moderate affinity to anionic liposomes lacking phosphatidylinositol phosphate (PIP_x) lipids. The C2B domain binds with sub-micromolar affinity to liposomes containing PI(4,5)P₂ but does not have a measurable affinity for background anionic lipids. Both domains can be competed away from their target lipids by the soluble PIP_x analogue inositol-(1,2,3,4,5,6)-hexakisphosphate (IP₆), which is a positive regulator of insulin secretion. Potential roles of these interactions in the docking and release of granuphilin from the plasma membrane are discussed.

1. Introduction

Protein domains that interact reversibly with lipid membranes serve as signaling modules in a variety of cellular processes (Cho and Stahelin, 2005; Lemmon, 2008). C2 domains represent one of the most abundant families of membrane-targeting domains and typically dock to anionic lipid membranes in response to Ca²⁺, although some members sense phosphatidylinositol phosphate (PIP_x) lipids in addition to, or instead of, cytosolic Ca²⁺ (Corbalan-Garcia et al., 2003; Murray and Honig, 2002; Nalefski and Falke, 1996). C2 domains are highly represented among proteins in regulated secretory pathways, including synaptotagmin and related proteins that serve as Ca²⁺ and/or lipid sensors in secretory vesicle docking and exocytosis (Fukuda and Mikoshiba, 2001; Gustavsson and Han, 2009; Martens, 2010; Sudhof, 2002).

The synaptotagmin-like protein granuphilin (also known as Slp-4) contains an N-terminal domain that attaches to Rab27a/Rab3 on secretory vesicles, a linker region of unknown structure that mediates binding to syntaxin and/or Munc18 on the plasma membrane, and two C-terminal C2 domains, termed C2A and C2B (Coppola et al., 2002; Torii et al., 2004; Tsuboi and Fukuda, 2006; Wang et al., 1999; Yi et al., 2002). Both C2 domains have been shown qualitatively to bind lipid vesicles containing phosphatidylinositol-(4,5)-bisphosphate [PI(4,5)P₂ or PIP₂] independently of Ca²⁺ (Yu et al., 2007), but little else is known about their membrane binding properties. The role of C2 domain–plasma membrane interaction for granuphilin function is somewhat controversial; some models indicate direct C2 domain interaction with plasma membrane lipids, while others suggest that granuphilin plasma membrane docking is mediated solely by its linker region binding Munc18/syntaxin (Izumi, 2011; Tsuboi, 2009). Recent evidence indicates that deletion of both C2 domains from granuphilin leads to a cytosolic distribution, whereas each C2 domain localizes to the plasma membrane when expressed individually (Galvez-Santisteban et al., 2012). This observation supports the idea that membrane binding by one or both C2 domains is required for granuphilin plasma membrane localization. Interestingly, a splice variant lacking the C2B domain, termed granuphilin-b, functions and localizes similarly to the granuphilin-a variant containing both C2 domains (Torii et al., 2004; Tsuboi and Fukuda, 2006; Wang et al., 1999).

Granuphilin functions in secretory vesicle – plasma membrane docking prior to Ca²⁺-triggered membrane fusion (Fukuda, 2006; Izumi, 2011). It is found in insulin-secreting pancreatic β cells and other cell types that undergo large dense-core granule exocytosis (Bierings et al., 2012; Tomas et al., 2008; Yi et al., 2002). Altered granuphilin expression produces two seemingly contradictory effects: knockdown or knockout leads to decreased vesicle – plasma membrane docking but increased rates of secretion, while overexpression leads to increased vesicle docking with diminished secretion (Coppola et al., 2002; Gomi et al., 2005; Torii et al., 2004). This pattern suggests that granuphilin is part of an apparatus that tethers secretory vesicles to the plasma membrane and acts as a “brake” to exocytosis, which must be overcome during secretory stimulation. The molecular mechanisms of this braking behavior are unclear but likely involve syntaxin/Munc18 interaction (Gomi et al., 2005; Tsuboi and Fukuda, 2006). Importantly, the mechanism by which braking is released during active secretion is currently unknown (Izumi, 2011; Tomas et al., 2008).

In order to better understand granuphilin-mediated secretory vesicle tethering and inhibition of membrane fusion, more information is needed on the contributions of its C2 domains to plasma membrane docking and release. In this study, we report docking mechanisms, target lipids, and affinities of both individual granuphilin C2 domains using a combination of equilibrium and kinetic fluorescence spectroscopy. In addition, we identify a product of insulin secretory signaling, inositol (1,2,3,4,5,6)-hexakisphosphate (IP₆), that displaces both C2 domains from lipid vesicles. The results suggest potential mechanisms by which C2 domains contribute significantly to granuphilin plasma membrane docking and release.

2. Materials and Methods

2.1 Reagents

Lipids were purchased from Avanti Polar Lipids (all are synthetic unless otherwise indicated): 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (phosphatidylcholine, PC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoserine (phosphatidylserine, PS), phosphatidylinositol (PI) from liver, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (phosphatidylethanolamine, PE), phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂, PIP₂] from brain, 1,2-dioleoyl-sn-glycero-3-phospho-(1'-myo-inositol-3',4',5'-trisphosphate) [PI(3,4,5)P₃, PIP₃], cholesterol, and sphingomyelin (SM) from brain. N-[5-(dimethylamino)-naphthalene-1-sulfonyl]-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (dansyl-PE, dPE) was from Invitrogen. d-myo-inositol 1,2,3,4,5,6,-hexakisphosphate dodecasodium salt (IP₆) was from Sigma. D-myo-inositol-1,4,5-triphosphate tripotassium salt (IP₃) was from Cayman Chemical.

2.2 Cloning, Expression and Purification of Granuphilin C2A and C2B domains

cDNA encoding the human granuphilin C2A domain (Gly352 – Ser488) flanked by a C-terminal His₆ tag was purchased from the DNASU plasmid repository (DNASU: HsCD00321956). This domain was expressed in E. coli BL-21 cells and purified using Ni-NTA column chromatography. For C2B, the region of human granuphilin cDNA (ATCC: 10700678) encoding this domain (Glu487 – Leu671) was PCR-amplified and cloned into a previously described N-terminal glutathione S-transferase expression vector (Corbin et al., 2004). Protein was expressed in E. coli BL-21 cells and purified using glutathione sepharose as described (Brandt et al., 2012). High salt washes were used to remove contaminating nucleic acid, and the free C2B domain was eluted following thrombin cleavage. Purified proteins were concentrated, treated with benzonase (Sigma) for 12 h at 4 °C to remove remaining nucleic acid, and dialyzed into assay buffer (140 mM KCl, 0.5 mM MgCl₂, 150 mM NaCl, 25 mM HEPES, pH 7.4) including 1 mM 2-mercaptoethanol and 0.02% NaN₃. Purity of the isolated proteins was >95% by SDS-PAGE, and the absence of significant contaminating nucleic acid was verified via absorbance measurement at 260 nm. Concentration was determined from absorbance of denatured protein at 280 nm ( = 19940 M⁻¹ cm⁻¹ and 34950 M⁻¹ cm⁻¹ for C2A and C2B, respectively).

2.3 Preparation of Lipid Vesicles

Small unilamellar vesicles (SUVs) with the lipid compositions listed in Table 1 were prepared by sonication as described previously (Brandt et al., 2012).

Table 1.

Vesicle lipid compositions used in this study (mol %)

	PC	PS	PE	PI	PG	CH	SM	PIP2	PIP3	dansyl-PE
PM(+)PIP2	10.5	21	28	4	0	25	4.5	2	0	5
PM(+)PIP3	10.5	21	28	4	0	25	4.5	0	2	5
PM(−)PIPx	10.5	21	28	6	0	25	4.5	0	0	5
PM(−)PIPx/PS/PI	37.5	0	28	0	0	25	4.5	0	0	5
PM(−)PIPx/PS/PI(+)PG	10.5	0	28	0	27	25	4.5	0	0	5
PC/dPE	95	0	0	0	0	0	0	0	0	5
PC/PS/dPE	71	24	0	0	0	0	0	0	0	5
PC/PI/dPE	71	0	0	24	0	0	0	0	0	5
PC/PG/dPE	71	0	0	0	24	0	0	0	0	5
PC/PS/PI/dPE	68	21	0	6	0	0	0	0	0	5

2.4 Equilibrium fluorescence measurements

Measurements were performed in a Photon Technology International QuantaMaster fluorescence spectrometer at 25 °C, with excitation at 284 nm (1 nm slit width) and emission slit width 8 nm.

Qualitative protein-to-membrane FRET measurements were performed using 125 μM total accessible lipid in assay buffer containing either 100 μM EDTA or 1 mM CaCl₂. First, a blank spectrum of each sample was measured in order to quantify fluorescence emission due to direct dansyl excitation at 284 nm. Subsequent emission spectra were measured after addition of (a) 1 μM C2A or C2B domain and (b) 8 mM IP₆. This non-physiological concentration of IP₆ has been sufficient to competitively remove PIP_x-bound proteins in previous studies, including those with very high PIP_x affinity (Kavran et al., 1998; Landgraf et al., 2008). Samples were equilibrated for 40 s with stirring after each addition. All spectra shown are corrected for dilution.

For measurement of IP₆ or IP₃ binding to the free granuphilin C2A domain (0.2 μM), the change in intrinsic Trp emission at 330 nm was measured upon titration with ligand. (Corbin et al., 2004; Landgraf et al., 2008) Samples were equilibrated for 40 s with stirring following each addition, and emission intensities were averaged over 10 s and corrected for dilution. The resulting plot of intensity F vs. ligand concentration [I] was a subject to a non-linear fit best-fit analysis in order to calculate the equilibrium dissociation constants (K_I) for these inhibitors of membrane binding:

$$F=\Delta F_{\max }\left(\frac{\left[\mathrm{I}\right]}{K_{\mathrm{I}}+\left[\mathrm{I}\right]}\right)+C$$ (1)

where ΔF_max represents the calculated maximum fluorescence change from the unbound protein and C is a constant. To simplify graphical representations, data are normalized such that C = 0 and ΔF_max = 1.

For FRET competition binding assays, C2A or C2B domain (1 μM) was pre-mixed with liposomes (125 μM total accessible lipid) in assay buffer containing 100 μM EDTA. Decreased protein-to-membrane FRET upon IP₆ titration was assessed based on acceptor emission at 520 nm, using a 10-s time average for each measurement. In a separate cuvette measured in parallel, IP₆ was titrated into a sample containing lipid but no protein. After correcting for dilution, intensities from this blank sample were subtracted from the experimental samples. Resulting plots were fit to a hyperbolic model for single-site competitive inhibition:

$$F=\Delta F_{\max }\left(1-\frac{\left[\mathrm{I}\right]}{{\mathrm{IC}}_{50}+\left[\mathrm{I}\right]}\right)+C$$ (2)

where ΔF_max is the total FRET signal before ligand titration, [I] is the concentration of added IP₆, and IC₅₀ is the IP₆ concentration at which FRET is 50% of the initial value. To simplify graphical representations, data are normalized such that C = 0 and ΔF_max = 1.

In these competition binding measurements, equilibrium constants for protein-lipid interaction can be calculated from the best-fit IC₅₀ value if the K_I of the inhibitor is known. Affinities measured in this manner can be expressed either as mole-fraction partition coefficients (K_x) (White and Wimley, 1999) or as PIP_x dissociation constants (K_d,PIPx). Here, both values are reported and are calculated using eq. 3 and 4, respectively:

$$K_{\mathrm{x}}=\left(\frac{{\mathrm{IC}}_{50}}{K_{\mathrm{I}}}-1\right)\times \frac{\left[{\mathrm{H}}_2\mathrm{O}\right]}{\left[\mathrm{L}\right]}$$ (3)

$$K_{\mathrm{d},\mathrm{PIPx}}=\frac{\left[{\mathrm{PIP}}_{\mathrm{x}}\right]K_{\mathrm{I}}}{{\mathrm{IC}}_{50}-K_{\mathrm{I}}}$$ (4)

where [L] is the total concentration of accessible lipids in the outer leaflet (i.e., half the total lipid concentration) and [PIP_x] is the concentration of free accessible PI(4,5)P₂ or PI(3,4,5)P₃ when [I] = IC₅₀ (i.e., the total outer leaflet [PIP_x] less half the total protein concentration). For error propagation, the uncertainty in [L] or [PIP_x] was estimated to be 10%. G° values are calculated as described (White and Wimley, 1999).

2.5 Stopped-flow FRET measurement of association and dissociation kinetics

Kinetic measurements were performed using an Applied Photophysics SX.17 stopped-flow fluorescence spectrophotometer in assay buffer with 100 μM EDTA at 25 °C. Excitation was at 284 nm with a slit width of 3 nm, and dansyl emission was measured using a 475-nm longpass filter. All data prior to the instrument dead time of 1 ms were discarded prior to plotting and fitting.

To determine dissociation rate constants (k_off), a solution containing protein (0.3 μM) and liposomes (75 μM total accessible lipid, concentrations after mixing) was rapidly mixed with an equal volume of a solution containing a 4-fold excess of unlabeled vesicles, which lacked dansyl-PE but were otherwise identical to the vesicles to which the protein was initially bound. Kinetic profiles of decreasing protein-to-membrane FRET were fit to single-exponential decay functions, in which the resulting approach to equilibrium is described by the intrinsic dissociation rate constant k_off:

$$F=\Delta F_{\max }\left(e^{-k_{\mathrm{off}}t}\right)+C.$$ (5)

For simplified presentation, the best-fit offset C was subtracted from all data points and ΔF_max was normalized to unity.

To determine the apparent association rate constants (k_obs) of the C2 domains, C2A or C2B domain (0.3 μM) was rapidly mixed with liposomes (75 μM total accessible lipid, concentrations after mixing). Each resulting time course was fit to a single-exponential binding model (Corbin et al., 2004):

$$F=\Delta F_{\max }\left(1-e^{-k_{\mathrm{obs}}t}\right)+C.$$ (6)

The best-fit offset C was subtracted from all data points and ΔF_max was normalized to unity.

The protein-membrane association rate constant k_on was calculated based on k_obs and k_off according to either the membrane partitioning (k_on,x) or PI(4,5)P₂ binding (k_on,PIP2) model using eq. 7 or 8, respectively:

$$k_{\mathrm{on},\mathrm{x}}=(k_{\mathrm{obs}}-k_{\mathrm{off}})\times \frac{\left[{\mathrm{H}}_2\mathrm{O}\right]}{\left[\mathrm{L}\right]}$$ (7)

$$k_{\mathrm{on},\mathrm{PIP}2}=\frac{k_{\mathrm{obs}}-k_{\mathrm{off}}}{\left[{\mathrm{PIP}}_2\right]}$$ (8)

where [L] is the total outer leaflet lipid concentration and [PIP₂] is the concentration of unbound PI(4,5)P₂ in the outer leaflet, approximated as the total outer leaflet PI(4,5)P₂ concentration less half the total protein concentration.

2.6 Electrostatic surface calculations for C2A

Electrostatic surface calculations were based on the available structure of granuphilin C2A (PDB: 3FDW). The positions of missing atoms in the structure (some sidechain atoms from K365, Q369, T370, K390, K391, R407, S435, Q439, D466, K469, and K483) were modeled using default settings of Swiss PDB Viewer. The electrostatic surface was generated using the Advanced Poisson-Boltzmann Solver plugin (APBS 0.5.1) for PyMol, with calculations performed using 0.15 M monovalent cation and anion concentrations.

3. Results

3.1 Identification of target lipids using protein-to-membrane FRET

In order to measure the lipid composition dependence of membrane targeting for granuphilin C2 domains, an established protein-to-membrane fluorescence resonance energy transfer (FRET) assay was used with SUVs of defined lipid composition (Nalefski and Falke, 2002). In this measurement, one or more Trp residues intrinsic to the protein domain (2 total in C2A, 5 in C2B) serve as FRET donors, and dansyl-phosphatidylethanolamine (dPE) lipids included in the liposomes serve as FRET acceptors. Protein-membrane binding leads to an increased dPE emission at 512 nm upon Trp excitation at 284 nm and a decrease in Trp emission at 350 nm. In order to determine target lipid preferences for granuphilin C2A and C2B domains, liposomes were prepared with the compositions listed in Table 1.

The granuphilin C2A domain shows significant protein-to-membrane FRET when added to liposomes closely resembling the lipid composition of the plasma membrane cytosolic leaflet (Corbin et al., 2007; Voelker, 2008). Since some C2 domains are known to dock preferentially to PIP_x lipids, such physiologically relevant lipid mixtures are used throughout this study either including 2% PI(4,5)P₂ [PM(+)2%PIP₂] or lacking PIP_x [PM(−)PIP_x]. In qualitative measurements, addition of the C2A domain led to an increase in dPE emission in vesicles with either of these compositions (Figure 1A,B), but not vesicles composed only of phosphatidylcholine (PC) and dPE (Figure 1C).

Figure 1. Qualitative FRET measurement of granuphilin C2A binding to lipid vesicles.

Fluorescence emission spectra are shown of C2A with (A) PM(−)PIP_x, (B) PM(+)PIP₂, and (C) PC/dPE vesicles (lipid compositions given in Table 1). Thin black curves: blank measurements prior to protein addition, consisting of vesicles (125 μM total accessible lipid) in assay buffer. The peak at ~510 nm in these traces represents direct excitation of the dans fluorophore at 284 nm. Thick gray curves: measurements after addition of 1 μM C2A domain. Dashed black curves: measurements after addition of 8 mM IP₆. Protein-lipid binding is evidenced by increased dans emission at ~510 nm due to FRET upon protein addition. Intensities in each panel are normalized to the maximum dans emission intensity of the sample prior to protein addition.

In order to identify the lipid components responsible for C2A association with PM(−)PIP_x vesicles, simpler lipid compositions were tested. FRET was observed upon addition of C2A to vesicles containing the anionic background lipids phosphatidylserine (PS) and phosphatidylinositol (PI) along with PC (Figure 2A), but not to vesicles containing only the zwitterionic components of our PM(−)PIP_x mixture (Figure 2B). Binding was restored upon replacement of PS and PI with phosphatidylglycerol (PG), an anionic lipid that is not normally present in mammalian plasma membranes (van Meer et al., 2008) (Supplementary data, Figure S1), suggesting that C2A binds anionic background lipids such as these through nonspecific interactions.

Figure 2. Granuphilin C2A interaction with background anionic lipids.

Fluorescence emission spectra are shown as described in the legend to Figure 1, using vesicles composed of (A) PC/PS/PI/dPE and (B) PM(−)PIP_x/PS/PI (compositions given in Table 1).

In contrast to C2A, no FRET was observed upon addition of 1 μM C2B domain to PM(−)PIP_x, indicating binding is too weak to measure with this lipid composition (Figure 3A). However, significant interaction was observed when PI(4,5)P₂ was included in the liposomes (Figure 3B). For both domains, the FRET efficiency was identical whether measured in the presence of 100 μM EDTA or 1 mM Ca²⁺ (data not shown), consistent with previous reports that granuphilin C2 domains bind anionic lipids in a Ca²⁺-independent manner (Wang et al., 1999; Yu et al., 2007).

Figure 3. Qualitative FRET measurement of granuphilin C2B binding to lipid vesicles.

Fluorescence emission spectra are shown as described in the legend to Figure 1, using 1 μM C2B domain and vesicles composed of (A) PM(−)PIP_x and (B) PM(+)PIP₂ (compositions given in Table 1).

IP₆ inhibits lipid binding by both granuphilin C2 domains. IP₆ is a soluble, fully phosphorylated analogue of PIP_x lipid headgroups, which competitively inhibits membrane binding by protein domains that target PIP_x lipids (Corbin et al., 2004; Landgraf et al., 2008; Takeuchi et al., 1997). For both granuphilin C2 domains and with all lipid compositions tested, protein-to-membrane FRET was completely reversed by the addition of 8 mM IP₆ as judged by dansyl emission at 512 nm (Figures 1-3). IP₆ addition also resulted in a measurable increase in the intrinsic Trp emission of the C2A domain even in the absence of membrane binding (Figure 1C), an effect that was not observed for C2B. Such a Trp fluorescence increase has been previously observed with PIP_x-binding pleckstrin homology domains (Corbin et al., 2004; Landgraf et al., 2008) and may be due to a change in the microenvironment around one or both of the Trp residues in granuphilin C2A upon binding IP₆.

3.2 Quantifying C2A affinity for IP₆ and IP₃

Qualitative measurements such as those shown in Figures 1-3 do not distinguish between tight and moderate lipid binding affinities, as any interaction in the micromolar range or tighter produces nearly complete binding. However, the lipid affinity of a tightly bound protein can be determined quantitatively using a competitive binding assay if an inhibitor is available with known affinity for the protein (Corbin et al., 2004). Thus, the affinity of C2A for IP₆ was first measured based on the intrinsic Trp fluorescence increase, in order that this domain's affinity for target vesicles could be subsequently quantified. IP₆ titration data fit well to a single-site binding model with an inhibitor equilibrium dissociation constant (K_I) of 1.8 ± 0.1 μM (Figure 4A). Similar measurement of C2A affinity for IP₃ reveals considerably weaker binding, with a K_I of 80 ± 30 μM (Figure 4B). This difference stands in contrast to the pleckstrin homology domain from phospholipase C-δ₁, which binds preferentially to PI(4,5)P₂ over other PIP_x species and exhibits a higher affinity towards IP₃ compared to IP₆ (Kavran et al., 1998). For the present study, the higher-affinity inhibitor IP₆ was used for subsequent competition experiments.

Figure 4. Equilibrium fluorescence measurement of granuphilin C2A domain binding to (A) IP₆ and (B) IP₃.

Trp fluorescence intensity (excitation 284 nm, emission 330 nm) was monitored upon titration of ligand into solutions of 0.2 μM C2A in assay buffer. The fluorescence increase was fit to eq. 1 (solid curves) and intensity values were normalized as described in Methods (section 2.4). Best-fit values are given in section 3.2. Error bars show standard deviation of three independent replicate experiments; where not visible, error bars are smaller than the data symbols.

3.3 C2A competition binding measurements

Competition binding measurements with IP₆ show that PI(4,5)P₂ enhances the affinity of granuphilin C2A for liposomes ~40-fold. As shown in Figure 5A and summarized in Table 2, IP₆ was titrated into solutions containing C2A domain bound to vesicles. A simple competitive inhibition model is sufficient to describe the IP₆ titration data. The inhibitor concentration at which the amount of vesicle-bound protein was half its initial value (IC₅₀) increased from 37 ± 6 μM to 1.6 ± 0.2 mM upon inclusion of 2% PI(4,5)P₂ (Table 2). These data, together with the measured affinity of C2A for IP₆ (Figure 4A), can be used to calculate a mole-fraction partition coefficient, K_x, describing the liposomal membrane partitioning equilibrium for each lipid composition (White and Wimley, 1999). This value is (9 ± 2) × 10⁶ for PM(−)PIP_x vesicles and increases ~40-fold to (390 ± 70) × 10⁶ for PM(+)PIP₂. If the C2A-membrane interaction is approximated as a bimolecular equilibrium between the protein and PI(4,5)P₂ headgroups rather than a membrane/aqueous phase partitioning, the same data yield an apparent equilibrium dissociation constant, K_d,PIP2, of 2.3 ± 0.4 nM (Table 2). Replacement of PI(4,5)P₂ with 2% PI(3,4,5)P₃ in target liposomes leads to similarly tight binding, with an affinity that differs from PM(+)PIP₂ vesicles by less than a factor of two (Table 2). This result is consistent with earlier reports of low PIP_x selectivity by this domain using less quantitative methods (Galvez-Santisteban et al., 2012; Wang et al., 2013). Overall, it is clear that granuphilin C2A binds with high affinity to liposomes containing physiological mixtures of background anionic lipids and either PI(4,5)P₂ or PI(3,4,5)P₃.

Figure 5. IP₆ competition measurement of granuphilin C2 domain vesicle binding.

Protein-to-membrane FRET was monitored (excitation 284 nm, emission 512 nm) of (A) 1 μM C2A or (B) 1 μM C2B in the presence of PM(−)PIP_x (filled symbols) or PM(+)PIP₂ vesicles (open symbols) upon titration of IP₆. The resultant FRET decrease was fit to eq. 2 (solid curves). Best-fit values are given in Table 2. Error bars are standard deviation of three independent replicate experiments, and where not visible are smaller than the data symbols. Inset to panel A: view expanded to show values at lower IP₆ concentrations.

Table 2.

Parameters from IP₆ competition measurements

Domain	Target vesicles	IC50 (IP6), μMa	Kx × 10−6b	Apparent Kd,PIPx, nMc	ΔG°x, kcal/mol	ΔG°PIPx, kcal/mol
C2A	PM(+)PIP2	1600 ± 200	390 ± 70	2.3 ± 0.4	−11.7 ± 0.1	−11.8 ± 0.1
C2A	PM(+)PIP3	1100 ± 100	270 ± 40	3.3 ± 0.5	−11.5 ± 0.1	−11.6 ± 0.1
C2A	PM(−)PIPx	37 ± 6	9 ± 2	N.D.d	−9.5 ± 0.1	N.D.d
C2A	PC/PS/dPE	10 ± 1	2.1 ± 0.4	N.D.d	−8.6 ± 0.1	N.D.d
C2A	PC/PI/dPE	5 ± 1	0.8 ± 0.1	N.D.d	−8.1 ± 0.1	N.D.d
C2A	PC/PG/dPE	11 ± 1	2.2 ± 0.3	N.D.d	−8.6 ± 0.1	N.D.d
C2B	PM(+)PIP2	21 ± 3	N.D.e	N.D.e	N.D.e	N.D.e

^aIC₅₀ values and errors are average and standard deviation of 3 or more independent replicate experiments.

^bMole-fraction partition coefficient, calculated using eq. 3.

^cCalculated using eq. 4.

^dApparent K_d,PIPx not calculated for vesicles lacking PIP_x.

^eIP₆ binding does not alter the intrinsic Trp fluorescence of C2B, therefore its K_I was not measured using this method.

In order to quantitatively test for specificity in C2A interactions with background anionic lipids, IP₆ competition titrations were conducted with this domain and simple 3:1 lipid mixtures of PC and either PS, PI, or PG. These titrations reveal similar affinities of the C2A domain for these three simple lipid compositions, indicating that interactions with background anionic lipids are nonspecific (Supplementary data, Figure S2). K_x values for these lipid compositions range from 8 × 10⁵ to 2 × 10⁶, lower by factors of 4-10 than for PM(−)PIP_x (Table 2). Such lower affinity for simple lipid mixtures versus physiological mixtures with similar anionic content has also been observed for other PIP_x-targeting domains, although the underlying reasons for this effect are not clear (Corbin et al., 2004).

IP₆ also competes with the C2B domain for lipid binding, with an IC₅₀ of 21 ± 3 μM (Figure 5B, Table 2). IP₆ binding does not significantly alter the intrinsic Trp fluorescence of C2B (data not shown), preventing direct measurement of C2B – IP₆ affinity using this approach.

3.4 Kinetics of membrane association and dissociation

FRET-based kinetic measurements of protein-lipid association and dissociation further demonstrate high-affinity binding. Stopped-flow association measurements (Figure 6A) generally fit well to single-exponential profiles with rate constants (k_obs) listed in Table 3. Rates of stochastic dissociation were measured by rapidly mixing protein-membrane complexes with an excess of unlabeled liposomes (Figure 6B). The decay constant of the resulting approach to equilibrium is equivalent to the rate constant for protein-membrane dissociation, k_off; these values are listed in Table 3.

Figure 6. Kinetic measurement of granuphilin C2 domain membrane interaction.

A: Association kinetics. Protein (0.3 μM) was rapidly mixed with PM(+)PIP₂ vesicles (75 μM total accessible lipid after mixing) in a stopped-flow fluorescence spectrometer and the increase in dans emission was monitored (excitation 284 nm). Single-exponential fits are shown (black curves, eq. 6). B: Dissociation kinetics. Protein (0.3 μM) was pre-bound to PM(+)PIP₂ vesicles (75 μM total accessible lipid, all concentrations after mixing) and rapidly mixed with excess PM(+)PIP₂ vesicles lacking dansyl-PE (300 μM total accessible lipid). Single-exponential fits are shown (black curves, eq. 5). Data points are averages from 9-12 mixing events. The smaller signal-to-noise ratios for C2B are due to the smaller amplitude of its FRET change upon membrane binding (compare Figure 3B to Figure 1B).

Table 3.

Measured rate constants from kinetic experiments

Domain	Target vesicles	koff, s−1a	kobs for association, s−1a,b	kon/koff × 10−6 expressed as partition coefficientc	koff/kon, nMc expressed as affinity for PIP2
C2A	PM(−)PIPx	13 ± 1	190 ± 30	10 ± 2	N.D.d
C2A	PM(+)PIP2	0.48 ± 0.01	150 ± 20	230 ± 40	4.3 ± 0.7
C2B	PM(+)PIP2	6.9 ± 0.1	54 ± 8	5 ± 1	200 ± 40

^aValues and errors shown are average and standard deviation of fits to 3 or more replicate measurements.

^bMeasured using 75 μM total accessible lipid.

^cValues of k_on were calculated using eqs. 7 and 8.

^dNot calculated for vesicles lacking PIP₂.

The observed association rate constant, k_obs, is related to the intrinsic association and dissociation rate constants, k_on and k_off, via eq. 7 (if the interaction is modeled as a phase partitioning) or eq. 8 (if the interaction is modeled as a bimolecular binding equilibrium). The ratios of these rate constants for C2A interacting with both PM(+)PIP₂ and PM(−)PIP_x vesicles agree within a factor of two of the respective equilibrium constants (K_x and K_d,PIP2) measured using the IP₆ competition assay (Tables 2-3). Kinetic measurements with the C2B domain suggest a ~50-fold weaker interaction relative to C2A with PM(+)PIP₂, due to a ~15-fold faster k_off and a ~3-fold slower k_on (Figure 6, Table 3). Overall, these two complementary approaches demonstrate a high-affinity interaction between the C2A domain and vesicles containing PI(4,5)P₂ and background anionic lipids.

4. Discussion

C2 domains are essential components of regulated exocytosis. Deletion of both C2 domains from granuphilin has been shown to produce localization in the cytosol or on vesicles rather than on the plasma membrane, as well as reduced vesicle–plasma membrane docking (Galvez-Santisteban et al., 2012; Tsuboi and Fukuda, 2006), suggesting that these domains play a key role in granuphilin plasma membrane targeting. Here we report the following new observations for C2 domain interaction with artificial vesicles composed of physiological lipid mixtures: (i) granuphilin C2A binds strongly to vesicles containing PI(4,5)P₂ or PI(3,4,5)P₃ and the background anionic lipids PS and PI, with apparent PIP_x affinities in the 2-5 nM range; (ii) C2A binds ~40-fold less strongly to otherwise identical vesicles lacking PIP_x; (iii) C2B interacts with vesicles containing PI(4,5)P₂ with sub-micromolar affinity, but not with vesicles lacking PIP_x; and (iv) IP₆ binds with low micromolar affinity and competes with lipid binding for both domains.

The granuphilin C2A domain interacts nonspecifically with monoanionic background lipids including PS, PI, and PG, and binds with much stronger affinity to vesicles that include polyanionic lipids such as PI(4,5)P₂ or PI(3,4,5)P₃ (Table 2). Furthermore, this domain binds IP₃ ~40-fold less tightly than the more anionic IP₆. Granuphilin C2A possesses a net charge near +8 at neutral pH (Olsson et al., 2011), and these observations suggest that electrostatics may play a dominant role in IP_x and PIP_x binding. Our results indicate that this domain binds PI(4,5)P₂ and PI(3,4,5)P₃ with similar affinity, and a previous study using immobilized lipids also showed little PIP_x selectivity for both C2A and C2B (Galvez-Santisteban et al., 2012). It has been suggested that promiscuous PIP_x-binding domains are likely regulated by the most abundant PIP_x species in their target membrane (Kavran et al., 1998). Granuphilin targets the plasma membrane, in part through interaction of its linker domain with Munc18/syntaxin (Tsuboi and Fukuda, 2006), and PI(4,5)P₂ is the most abundant PIP_x species in this compartment (Balla, 2005; Di Paolo and De Camilli, 2006; Lemmon, 2008). Our data suggest that C2 domain lipid binding may provide a significant contribution to the affinity of this protein's membrane interaction.

Available structural data for C2A are consistent with membrane binding via electrostatic interaction with PIP_x and background anionic lipids via one or more binding sites. Poisson-Boltzmann mapping reveals a large positively charged surface on this domain (Figure 7). This surface includes a canonical PIP_x binding motif, K(K/R)KTXXK(K/R), homologous to synaptotagmin 1 C2B (Figure 7, black arrow) (Galvez-Santisteban et al., 2012) as well as a cleft composed of basic residues in the β2-β3 and β6-β7 loops (Figure 7, green arrow). The latter cleft is equivalent to the Ca²⁺ binding site of synaptotagmin C2 domains, but granuphilin C2A contains only one of the five Ca²⁺-chelating Asp residues conserved among synaptotagmin isoforms (Bhalla et al., 2008). Further studies are needed to discern which site(s) on this domain are involved in binding lipids and/or IP₆.

Figure 7. Electrostatic surface representation of granuphilin C2A.

Surface electrostatic potentials are shown colored from -3 (red) to +3 (blue). Green arrow: putative binding site consisting of residues in the β2-β-3 and β6-β7 loops, which correspond to the Ca²⁺ binding loops of synaptotagmin C2 domains. Black arrow: putative binding site including residues K410, R411, and K412, homologous to the PIP₂ binding site of synaptotagmin 1 C2B (Radhakrishnan et al., 2009). Potential map calculated using the APBS plugin for Pymol.

Equilibrium fluorescence data for C2A – IP₆ binding and competition with lipids both fit well to simple hyperbolic functions, suggesting that IP₆ binds this domain at a single site and thereby blocks lipid interaction. Our data are consistent with the presence of one or more lipid binding sites, provided a single bound IP₆ is sufficient to displace the domain from the membrane. This assertion seems reasonable from an electrostatic standpoint, as binding of one IP₆ molecule should neutralize the net charge of the C2A domain and disrupt nonspecific electrostatic interactions. However, membrane interaction involving multiple lipid ligands may not be represented appropriately by a bimolecular dissociation equilibrium constant (K_d,PIPx). Therefore, we also report mole fraction partition coefficients (K_x), which do not assume protein lipid binding stoichiometry, in addition to the more familiar K_d. Both measured values reflect high-affinity association of C2A with PM(+)PIP₂ and PM(+)PIP₃ vesicles, and correspond to G° values for binding of approximately -12 kcal/mol (Table 2).

Strong interaction is further supported by kinetic measurements of protein-lipid association and dissociation. The ratios of measured rate constants for C2A association with both PM(+)PIP₂ and PM(−)PIP_x vesicles are in close agreement with the affinities reported from equilibrium competition measurements. The single-exponential kinetic profiles indicate that membrane association and dissociation for both domains likely proceed via a single rate-limiting step, but do not report on stoichiometry of lipid interaction. Notably, the dissociation rate constant of 0.48 s⁻¹ for C2A is extremely slow for a membrane-targeting C2 domain, consistent with its high affinity (Brandt et al., 2012; Corbin et al., 2007; Hui et al., 2005; Nalefski et al., 1997). Interestingly, when we attempted to measure membrane dissociation by rapidly adding high concentrations of IP₆ (8 mM), an approach used with other PIP_x-binding domains (Corbin et al., 2004; Knight and Falke, 2009), dissociation kinetics were biexponential and ~100-fold faster than those measured by the approach to equilibrium method (Supplementary data, Figure S3). The origins of this difference are currently under investigation; however, the good agreement between our IP₆-independent kinetic measurements (Table 3) and equilibrium IP₆ competition measurements (Table 2) suggests that high concentrations of IP₆ may alter the kinetic pathway of dissociation without affecting the equilibrium thermodynamics of the system.

While affinity of C2B for PM(+)PIP₂ vesicles cannot be calculated from our equilibrium competition measurements due to lack of a direct fluorescent reporter for C2B – IP₆ binding, the equilibrium constant can be calculated from kinetic parameters. The measured k_off/k_on ratio yields an apparent K_d,PIP2 of 200 ± 40 nM, weaker than C2A but still significantly strong for a domain that is tethered in close proximity to target lipids (Shao et al., 2008). Previous studies have shown that splice variants of granuphilin with and without the C2B domain exhibit similar cellular behavior; however, the presence of this domain could conceivably modulate sensitivity to target lipids or IP_x species that may regulate membrane binding in vivo (Torii et al., 2004; Tsuboi and Fukuda, 2006). Intriguingly, substitution of k_off/k_on in the place of K_d,PIP2 in eq. 4 yields an equilibrium constant for C2B – IP₆ interaction of 1.9 ± 0.5 μM, remarkably similar to the K_I of 1.8 ± 0.1 μM measured directly for C2A. Further studies are needed to clarify the potential role(s) of IP₃ and IP₆ binding to C2A and C2B in granuphilin function.

Finally, the competitive removal of granuphilin C2 domain-lipid binding by IP₆ suggests a possible mechanism of granuphilin release from the plasma membrane during exocytosis. IP₆ concentrations in insulin-secreting cells increase from ~40 μM to ~60 μM during exocytosis (Berggren and Barker, 2008; Larsson et al., 1997; Li et al., 1992). While our data suggest these concentrations are insufficient to remove C2A from membranes containing PI(4,5)P₂, the presence of 40-60 μM IP₆ [along with possibly other IP_x species (Illies et al., 2007)] could potentially compete with C2B–membrane binding and C2A binding to PIP_x-free membranes. Thus, it is possible that granuphilin C2 domains could be removed from their target lipids during insulin secretion by simultaneous increases in cellular IP_x production and localized decreases in membrane PI(4,5)P₂ content due to phospholipase C (PLC) activation (Efanov et al., 1997; Thore et al., 2007). The generation of IP₃ by PLC would also contribute to this dissociation, although IP₃ binds C2A less tightly than IP₆. As noted previously, deletion of both C2 domains results in the loss of both plasma membrane localization and inhibitory activity, suggesting that interaction with Munc18/syntaxin alone is insufficient for proper localization (Galvez-Santisteban et al., 2012; Tsuboi and Fukuda, 2006). Thus, it is feasible that signaling-induced dissociation of these C2 domains from the plasma membrane could facilitate release from protein binding partners and removal of the “brake” on exocytosis. Clearly, further studies in secretory systems are needed in order to test these hypotheses and further define the roles of granuphilin C2 domains in secretory vesicle docking and fusion.

Highlights.

Both granuphilin C2 domains bind lipids present in the plasma membrane: PS and PIP₂

C2A binds liposomes containing PIP₂ with high affinity (2-5 nM)

C2B binds liposomes containing PIP₂ with moderate affinity (~200 nM)

The signaling molecule IP₆ competes with anionic lipids for binding both domains

Abbreviations

C2 domain

domain with homology to the second conserved region of protein kinase C

PI(4,5)P₂ or PIP₂

phosphatidylinositol-(4,5)-bisphosphate

PI(3,4,5)P₃ or PIP₃

phosphatidylinositol-(3,4,5)-trisphosphate

PS

phosphatidylserine

PI

phosphatidylinositol

PC

phosphatidylcholine

PE

phosphatidylethanolamine

PG

phosphatidylglycerol

SM

sphingomyelin

CH

cholesterol

dPE or dansyl-PE

dansyl phosphatidylethanolamine

PIP_x

phosphatidylinositol phosphate

IP₆

inositol-(1,2,3,4,5,6)-hexakisphosphate

IP₃

inositol-(1,4,5)-trisphosphate

IP_x

inositol polyphosphate

Slp

synaptotagmin-like protein

EDTA

ethylenediamine tetraacetic acid

FRET

fluorescence resonance energy transfer

SUV

small unilamellar vesicle