Different states of synaptotagmin regulate evoked versus spontaneous release

The tandem C2-domains of synaptotagmin 1 (syt) function as Ca2+-binding modules that trigger exocytosis; in the absence of Ca2+, syt inhibits spontaneous release. Here, we used proline linkers to constrain and alter the relative orientation of these C2-domains. Short poly-proline helices have a period of three, so large changes in the relative disposition of the C2-domains result from changing the length of the poly-proline linker by a single residue. The length of the linker was varied one residue at a time, revealing a periodicity of three for the ability of the linker mutants to interact with anionic phospholipids and drive evoked synaptic transmission; syt efficiently drove exocytosis when its tandem C2-domains pointed in the same direction. Analysis of spontaneous release revealed a reciprocal relationship between the activation and clamping activities of the linker mutants. Hence, different structural states of syt underlie the control of distinct forms of synaptic transmission.


Supplementary Figure 3.
Proline linker mutant forms of syt are properly targeted to synapses. (a) Full-length WT or proline linker mutant forms of syt were expressed in syt KO hippocampal neurons via lentiviral infection. Neurons were immunostained at 13 ~ 15 DIV using anti-syt (green) and anti-physin (red) antibodies. Shown are representative confocal images. Scale bar = 10 μm. (b) The degree of colocalization was quantified and plotted; all syt constructs were efficiently targeted to synapses. Data are represented as mean ± SEM. For each condition, n = 6, two independent cultures from independent litters of mice were examined, and three independent regions from each coverslip were analyzed. No significance was detected using one-way ANOVA, F = 0.5024, p = 0.8271. Figure 4. All of the proline linker mutant forms of syt fully rescued the size of the RRP in syt KO neurons. (a) Representative EPSCs, elicited by perfusion of hypertonic sucrose (500 mM), were recorded from WT, syt KO, and syt KO neurons expressing full-length WT or proline linker mutant forms of syt; the black bars indicate the application of sucrose. (b) RRP size was quantified by integrating the sucrosedriven EPSCs. All of the proline linker mutants rescued the size of the RRP to a similar extent as WT syt. Data are represented as mean ± SEM. *** p < 0.001 versus WT, oneway ANOVA followed by Tukey's multiple comparisons test. For each condition, data were collected from 15~24 cells in a total of 6~8 coverslips, where 2 coverslips was obtained from each of 3~4 independent litters of mice. Recording were made from 2~5 cells per coverslip. The number of independent litters, N, and the number of cells, n, are indicated in the bar graph as N/n. Results from one-way ANOVA analysis are provided in Supplementary Table 7 Free energy surface for a tetra-peptide model for the linkage between C2A and a polyproline segment: Leu-Gln-Pro-Pro. (d) Free energy surface for a tetra-peptide model for the linkage between C2B and a poly-proline segment: Leu-Gln-Pro-Pro. (e) A snapshot from the metadynamics simulation illustrates the steric interaction between Leu and Gln residues at the linkage between C2A and the poly-proline segment that controls the relatively fixed orientation of the poly-proline relative to the C2A domain. (f) Same as in panel e, but focused on the linkage between C2B and the poly-proline segment. Figure 8. Orientation of the Ca 2+ binding loops in the C2-domains. (a) Schematic illustration of the α angles that represent the orientations of the Ca 2+ binding loops with respect to the nPro linker. C2A is green, C2B is blue, and the polyproline linker is gray. (b) Schematic illustration of the β angles that represent the relative orientation of the Ca 2+ binding loops in the two C2-domains. The arrows illustrate the orientations of the Ca 2+ binding loops. (c) The C2A α angles in the four C2A-9Pro-C2B models (see text and Supplementary Table 8 during MD simulations. (d) The C2B α angles in the four C2A-9Pro-C2B models during MD simulations. Figure 9. Properties of the four C2A-9Pro-C2B models during MD simulations. The four rows summarize results for the (a) 1a-1b, (b) 1a-2b, (c) 2a-1b and (d) 2a-2b C2A-9Pro-C2B models, respectively. For each model, the left panel shows the time dependence of the collective variables used in the metadynamics simulations for C2A-9Pro and C2B-9Pro; the right panel illustrates the starting and final structures of each C2A-9Pro-C2B simulation. C2A is green, C2B is blue, and the poly-proline linker is gray.

Supplementary Notes 1
To determine the relative orientation of the tandem C2-domains of syt, when connected by a poly-proline rod, and to determine whether these domains are constrained, two types of molecular simulations were carried out. First, metadynamics simulations were used to compute the free energy surfaces for the rotation of the poly-proline rod with respect to each individual C2-domain. This information was then used to construct models for the tandem C2-domains connected by poly-proline linkers of nine, ten and eleven residues, and to compute the relative orientation of the tandem domains.

Metadynamics simulations for individual C2-domains
As shown in Supplementary Fig. 7, the (ζ,ψ) free energy map for C2A-9Pro has two dominant basins with ζ~0 (basin 2a) and 180° (basin 1a); they are fairly close in free energy (~2 kcal/mol) but are separated by sufficiently high (~15.3 kcal/mol) barriers. These conformers are more stable, compared to the basins with ψ~-50°, due to the hydrophobic packing of the connecting Leu residue against the nearby Gln (Supplementary Fig. 7e; in fact, the free energy map is qualitatively very similar to that computed for the isolated tetra-peptide that corresponds to the C2A-9Pro linkage (Leu-Gln-Pro-Pro, Supplementary Fig. 7c). Similarly, as shown in Supplementary Fig. 7b, the (ζ,ω) free energy map for C2B-9Pro is also highly similar to that for the isolated tetrapeptide that corresponds to the C2B-9Pro linkage (Pro-Pro-Leu-Gly, Supplementary Fig.  7d). The dominant basins (1b and 2b) have ζ~120°, to avoid the steric collision between Leu side chain and the proline rings ( Supplementary Fig. 7f); the free energy difference between the two basins is rather large (~7 kcal/mol) since ω~180° is the preferred conformation in the peptide.

Conformer 1a-2b represents the energetically most stable C2A-9Pro-C2B model
The free energy maps indicate that for both C2-domain-9Pro linkages, the relative orientation of the poly-proline segment and the C2-domain is relatively rigid. Therefore, it is possible to build C2A-9Pro-C2B models by combining the low freeenergy conformers for C2A-9Pro and C2B-9Pro; taking two dominant conformers for each (1a/2a and 1b/2b), we arrived at four possible models for C2A-9Pro-C2B. These four models were each simulated for ~70 ns using NPT molecular dynamics ( Supplementary Fig. 9).
With the assumption that there is minimal interaction between the two C2domains when separated by a poly-proline rod, the relative stabilities of the four models can be estimated by combining the free energies of different C2-domain-linker conformers (see Supplementary Table 8. The results indicate that only two C2A-C2B conformations, 1a-2b ( Supplementary Fig. 7b) and 2a-2b (Supplementary Fig. 7d) are relevant at RT, with the former being the dominant configuration with a population of about 98%. We note that the approximation of minimal domain-domain interaction in C2A-9Pro-C2B appears a valid one since the two domains indeed remain far apart in all four relaxed models (see Supplementary Fig. 9); although the minimal domain-domain distances (data not shown) may reach ~5-10 Å during some segments of the trajectories, persistent inter-domain contacts were not observed. Moreover, the internal structures of the two domains in the C2A-9Pro-C2B simulations remain very similar to those in the isolated C2-domain-9Pro simulations, again supporting the assumption of minimal domain interactions. Finally, the collective variables chosen in the metadynamics simulations remain close to their expected values during all C2A-9Pro-C2B simulations (Supplementary Fig. 9 first column), confirming that these dihedral angles remain the key variables for characterizing the relative poly-proline/C2 orientations; one C2-domain is not strongly perturbed by the presence of the other C2domain. The fact that these dihedral angles remain largely constant during the simulations again highlight the relative rigidity of the poly-proline/C2 connection; the C2domains, however, have sufficient time during the simulation to equilibrate their local interactions as indicated by the changes in the orientation of the Ca 2+ -binding loops relative to the poly-pro linker (see beginning and final structures shown in Supplementary Fig. 9 and the time dependence of the alpha angles in Supplementary  Fig. 8). Therefore, although a more precise estimate of the populations of these conformers could be obtained with metadynamics simulations for the C2A-9Pro-C2B system in the presence of both calcium ions and multi-component lipid membrane, which would be highly demanding computationally (in addition to the concern of using a non-polarizable force field for calcium-protein interactions 1 ), the computational results presented herein provide adequate support for 1a-2b being the dominant conformation for the tandem C2-domains connected with a poly-proline rod, especially viewed together with results from penetration assay in this study and the NMR data from Ref. 2 . In subsequent MD simulations for C2A-nPro-C2B as n is varied from 9 to 11, only the 1a-2b conformer is considered. The observation that the 2a-2b conformation is estimated to be only ~2 kcal/mol higher in free energy is consistent with the experimental observation that an alternative C2A-C2B orientation is likely to play a different physiological role during exocytosis).