Phosphatidylinositol 4,5-Bisphosphate Increases Ca2+ Affinity of Synaptotagmin-1 by 40-fold*

Background: Synaptotagmin-1, a Ca2+ sensor of neuronal exocytosis, interacts with the anionic phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2). Results: Microscale thermophoresis shows that PIP2 binding to the polybasic patch of synaptotagmin-1 increases the Ca2+ affinity by >40-fold. Conclusion: PIP2 and Ca2+ binding to synaptotagmin-1 is strongly cooperative. Significance: Understanding the interplay between Ca2+, synaptotagmin-1, and PIP2 is crucial for our understanding of neurotransmitter release. Synaptotagmin-1 is the main Ca2+ sensor of neuronal exocytosis. It binds to both Ca2+ and the anionic phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2), but the precise cooperativity of this binding is still poorly understood. Here, we used microscale thermophoresis to quantify the cooperative binding of PIP2 and Ca2+ to synaptotagmin-1. We found that PIP2 bound to the well conserved polybasic patch of the C2B domain with an apparent dissociation constant of ∼20 μm. PIP2 binding reduced the apparent dissociation constant for Ca2+ from ∼250 to <5 μm. Thus, our data show that PIP2 makes synaptotagmin-1 >40-fold more sensitive to Ca2+. This interplay between Ca2+, synaptotagmin-1, and PIP2 is crucial for neurotransmitter release.

In the synaptic terminal, neurotransmitter release is mediated by fusion of synaptic vesicles with the plasma membrane. Fusion is triggered by a sudden increase in the cytoplasmic Ca 2ϩ concentration in response to membrane depolarization. The protein synaptotagmin-1 (together with synaptotagmin-2 and synaptotagmin-9) is the main Ca 2ϩ sensor of the fast phase of neuronal exocytosis (reviewed in Ref. 1). Synaptotagmin-1 contains a single transmembrane domain close to the N terminus, which anchors the protein to synaptic vesicles. The transmembrane domain is connected by a 61-residue unstructured linker to two C2 domains, C2A and C2B. The mechanism by which synaptotagmin-1 triggers membrane fusion is still debated, but structural rearrangements of the plasma membrane and/or interactions with SNARE proteins have been implicated (1).
Ca 2ϩ binding to synaptotagmin-1, originally demonstrated by equilibrium dialysis using native protein (2), has been characterized by isothermal titration calorimetry (3) and NMR (4 -6) using a soluble fragment containing both C2 domains (C2AB fragment, residues 97-421). The C2A domain binds to three Ca 2ϩ ions with affinities ranging from 50 M to 10 mM. The C2B domain binds two Ca 2ϩ ions, both with ϳ200 M affinity.
In the presence of Ca 2ϩ , the C2 domains of synaptotagmin-1 also bind to membranes containing anionic phospholipids, with little specificity for the phospholipid species (3, 6 -14). Interestingly, binding already occurs at Ca 2ϩ concentrations well below the Ca 2ϩ affinity of free synaptotagmin-1. Here, anionic phospholipid headgroups complement the Ca 2ϩ -binding sites, increasing the affinity of C2AB for Ca 2ϩ to ϳ5-100 M (3, 6 -8, 11, 13). In the absence of Ca 2ϩ , a conserved polybasic lysine patch located on the C2B domain can also bind to anionic lipids, and this binding is strongly preferential for the polyanionic phospholipid phosphatidylinositol 4,5-bisphosphate (PIP 2 ) 4 (3, 9 -14). Binding of PIP 2 to the polybasic patch might increase the Ca 2ϩ affinity (12), although this is still controversial (3) and has hitherto not been characterized in detail.
Experimentally, measuring synaptotagmin-1 binding to PIP 2 and/or Ca 2ϩ is not trivial. Isothermal titration calorimetry and NMR require high (100 M to 1 mM) concentrations of protein (3)(4)(5). Therefore, high affinities well below these concentrations cannot be accurately determined with these approaches. Binding of synaptotagmin to PIP 2 is often inferred from binding of the C2 domains to artificial membranes containing a defined fraction of PIP 2 (e.g. by FRET (3), pulldown assays (11,13), or density flotations (3,12)). However, it is difficult to quantitatively distinguish Ca 2ϩ from PIP 2 binding with these approaches. We have recently shown (10) that Ca 2ϩ binding to synaptotagmin-1 can be directly measured with a new technique called microscale thermophoresis (MST) (15,16). MST is based on the principle that molecules move along a tempera-ture gradient in a capillary (the Soret effect). Upon binding to Ca 2ϩ or PIP 2 , the surface properties of synaptotagmin-1 change, resulting in an altered thermophoretic behavior. In this study, we applied MST to study PIP 2 and Ca 2ϩ cooperative binding to synaptotagmin-1.

RESULTS
We performed MST measurements on the Alexa Fluor 488labeled C2AB fragment of synaptotagmin-1 (residues 97-421). With this technique, a glass capillary is filled with a dilute protein solution (50 nM). Fluorescence is then measured at a spot in the capillary that is heated with a focused IR laser beam. Heating (by ϳ5°C) results in the generation of a temperature gradient along the axis of the capillary (Fig. 1, A and B). The C2AB fragment thermodiffuses out of this heated spot (measured by fluorescence recording), resulting in a protein gradient that is reversed when the IR laser is switched off. The amount of fluorescence decrease at the heated spot (the MST signal) was changed in the presence of Ca 2ϩ , thus providing a direct readout of Ca 2ϩ binding to the C2AB fragment. Evidently, Ca 2ϩ binding alters the thermophoretic (i.e. surface, charge) properties and thereby the thermodiffusion of synaptotagmin (10). Varying the calcium concentration in the capillary thus allowed us to obtain a binding curve (Fig. 1C).
We fitted the binding curves with simple Michaelis-Menten kinetics assuming a single binding site (see "Experimental Procedures"). This model does not take into account binding of multiple Ca 2ϩ ions (or PIP 2 molecules; see below), and for some curves, this simplification may affect the quality of the fit. However, the overall quality of the data did not warrant fitting with a more sophisticated binding model. Thus, we could not differentiate between the different calcium-binding sites, and we report only the apparent dissociation constant (K Ca ).
C2AB bound to Ca 2ϩ with K Ca ϭ 221 Ϯ 23 M (n ϭ 3). Control experiments with Mg 2ϩ or a mutant with disrupted Ca 2ϩ binding (D178A/D230A/D232A/D309A/D363A/D365A, called C2a*b*) (3,10) showed that the change in the MST signal was indeed due to binding of Ca 2ϩ ions to the established binding sites in the C2 domains. Furthermore, the MST measurements were not affected by the presence of the dye because a similar binding constant of K Ca ϭ 206 Ϯ 40 M was obtained with the unlabeled C2AB fragment using the intrinsic tryptophan fluorescence as the readout (C2AB has three tryptophans) (Fig. 1D). We then set out to study the cooperativity of Ca 2ϩ and PIP 2 binding.
One of the main advantages of MST compared with alternative techniques for measuring Ca 2ϩ binding is the low concentration of protein that is required: measurements could be carried out with C2AB concentrations as low as 50 nM, which is 3-4 orders of magnitude below that reported for isothermal titration calorimetry (3) or NMR (4 -6). This low concentration allowed us to measure PIP 2 binding by adding PIP 2 directly to the capillary (Fig. 2D). Even PIP 2 isolated from porcine brain with long fatty acid acyl chains (dominant species C18:0 and C20:4) is water-soluble at concentrations up to ϳ9 mM and does poorly form micelles because of its high anionic charge (18).  (Fig. 2D). In this experiment, an excess of 1 mM Mg 2ϩ was present to suppress potential nonspecific interactions of Ca 2ϩ with PIP 2 or C2AB. At higher PIP 2 concentrations, the Ca 2ϩ affinity increased even further (to Ͼ40-fold; K Ca ϭ 3.3 Ϯ 1.3 M at 40 -80 M PIP 2 compared with 221 Ϯ 23 M without PIP 2 ) (Fig. 3, A-C). Accordingly, the addition of Ca 2ϩ progressively increased the binding affinity of C2AB for PIP 2 (from K PIP2 ϭ 20 Ϯ 5 M without Ca 2ϩ to Ͻ2 M at Ͼ20 M Ca 2ϩ ). This cooperativity is not specific for PIP 2 or the length of the acyl chains because another phosphoinositide (20 M phosphatidylinositol 3,5-bisphosphate) or short-chain PIP 2 (20 M 1,2-dioctanoyl-sn-glycero-3-phosphatidylinositol 4Ј,5Јbisphosphate; C8:0) also increased the apparent Ca 2ϩ affinity (K Ca ϭ 11 Ϯ 5 and 8 Ϯ 5 M, respectively). PIP 2 binding required the well conserved polybasic patch that is located on the C2B domain because removal of two lysines from this patch (K326A/K327A, the so-called KAKA mutant (12)) (Fig. 3, D-F, and Fig. 4, A and B) almost completely abolished PIP 2 -dependent MST changes, even at very high Ca 2ϩ concentrations. Accordingly, the apparent affinity for Ca 2ϩ was increased by only ϳ3-fold in the presence of 80 M PIP 2 (from K Ca ϭ 195 Ϯ 35 M to 61 Ϯ 11 M). Thus, we could detect only PIP 2 binding to the polybasic patch and did not observe PIP 2 binding via the Ca 2ϩ -binding sites on the C2A  (12). Compared with the wild type, the amplitude of the fluorescence changes of the KAKA mutant was reduced due to the altered thermophoretic properties that resulted from the substitution of charged residues. In E, the solid (no PIP 2 ) and dashed (80 M PIP 2 ) lines are fits with K Ca ϭ 195 and 61 M, respectively. Note that for the KAKA mutant, PIP 2 binding was dramatically reduced compared with the wild type. Each experiment was repeated at least twice; error bars show the range of data points. and C2B domains, in contrast to previous observations by us and others (3, 10 -12, 14). It is likely that, for the interaction of the Ca 2ϩ -binding pockets with the membrane, hydrophobic residues surrounding these pockets must insert into the membrane (6 -8, 11, 12, 14), although we cannot exclude that PIP 2 binding to the Ca 2ϩ sites is silent (i.e. does not change the MST signal). Nevertheless, the Ca 2ϩ -binding pocket of the C2B domain does affect PIP 2 binding to the polybasic patch because disruption of Ca 2ϩ binding to the C2B domain (D309A/ D363A/D365A, called C2Ab*) reduced the affinity for PIP 2 by ϳ4-fold (from K PIP2 ϭ 20.4 Ϯ 5.2 M to 70 Ϯ 24 M) (Fig. 4B).
We then performed MST experiments with mutants disrupted in Ca 2ϩ binding to the C2A domain (D178A/D230A/ D232A, called C2a*B). Surprisingly, only a small and insignificant PIP 2 -or Ca 2ϩ -dependent change in the MST signal of C2a*B was observed compared with the wild type (Fig. 4, A and  B). Accordingly, the combination of C2a*B with the KAKA mutation did not markedly differ from the KAKA mutant with all Ca 2ϩ -binding sites intact. Apparently, Ca 2ϩ binding to the C2A domain does not result in a detectable change in the thermophoretic properties of the C2AB fragment. In contrast, Ca 2ϩ binding could no longer be detected by MST upon disruption of the C2B domain. Thus, only Ca 2ϩ binding to the C2B domain seems to change the thermophoretic properties of the C2AB fragment, indicating that the calcium-dependent changes reported above are exclusively mediated by the C2B domain. Perhaps this selectivity is related to the thermodynamically divergent modes of Ca 2ϩ binding of synaptotagmin-1: Ca 2ϩ binding to the C2A domain is endothermic, and that to the C2B domain is exothermic (3). Finally, Ca 2ϩ concentrations above 100 M increased the apparent PIP 2 affinity of synaptotagmin-1 even when both Ca 2ϩ -binding sites were disrupted (double mutant C2a*b*) (Fig. 4B). This indicates that Ca 2ϩ was still able to bind to the double mutant at very high Ca 2ϩ concentrations in the presence of PIP 2 , perhaps by binding directly to PIP 2 (19,20).

DISCUSSION
In this work, we have shown that PIP 2 binds to the polybasic patch of the C2B domain of synaptotagmin-1, in agreement with earlier studies (10 -14, 21). PIP 2 binding to the polybasic patch increases the apparent affinity of the C2B domain for Ca 2ϩ by Ͼ40-fold. Conversely, Ca 2ϩ binding to the C2B domain increases the affinity for PIP 2 by Ͼ10-fold. Cooperative PIP 2 and Ca 2ϩ binding to synaptotagmin-1 has been observed previously (12). This cooperativity is probably not caused by complementation of the Ca 2ϩ -binding sites, as suggested earlier by us and others (3, 6 -8), because the polybasic patch and the Ca 2ϩ -binding sites are located quite far apart (Fig. 4C). Instead, PIP 2 may interact in a structurally less defined manner with the polybasic patch and other solvent-exposed basic residues (9,12), and this may increase the Ca 2ϩ affinity simply by charge screening. Alternatively, the polybasic patch may form a structurally defined complex with PIP 2 similar to the C2 domains of rabphilin-3A and PKC␣ (22)(23)(24)(25). In fact, cooperative PIP 2 and Ca 2ϩ binding has been observed for these C2 domains (22)(23)(24), very similar to our observations for the C2B domain. Moreover, the crystal structure of the C2B domain (26) can be superimposed with that of the PIP 2 -bound C2 domain of PKC␣ (25), rendering it likely that PIP 2 binds to the C2AB fragment of synaptotagmin-1 in a similar manner (Fig. 4, C and D). Thus, it is conceivable that such PIP 2 binding increases the Ca 2ϩ affinity via a conformational change. However, how PIP 2 and Ca 2ϩ precisely bind in a cooperative manner to synaptotagmin-1 remains to be elucidated.
Together, we conclude that PIP 2 binding to the polybasic patch of synaptotagmin-1 dramatically increases the Ca 2ϩ sensitivity. As discussed previously (12), this explains the reduced release probability of the KAKA mutant in hippocampal neurons (12,27) and in Drosophila (28). It also explains why in vivo already 10 M Ca 2ϩ is sufficient for physiological release of neurotransmitters in the calyx of Held (29). PIP 2 modulation of synaptotagmin-1 may well be of major physiological relevance when considering that PIP 2 is the predominant phospholipid species at the sites of docked vesicles in PC12 cells (30).
Finally, our work demonstrates the value of MST for measuring molecular interactions. Although we were unable to detect Ca 2ϩ binding to the C2A domain under our conditions, MST can be extremely sensitive and allows for monitoring medium and high affinity interactions with only picomoles of material. MST has the potential to complement the limited set of techniques available to measure Ca 2ϩ and PIP 2 binding to proteins under equilibrium conditions such as isothermal titration calorimetry and NMR.