The L-type Voltage-Gated Calcium Channel co-localizes with Syntaxin 1A in nano-clusters at the plasma membrane

The secretory signal elicited by membrane depolarization traverses from the Ca2+-bound α11.2 pore-forming subunit of the L-type Ca2+-channel (Cav1.2) to syntaxin 1 A (Sx1A) via an intra-membrane signaling mechanism. Here, we report the use of two-color Photo-Activated-Localization-Microscopy (PALM) to determine the relation between Cav1.2 and Sx1A in single-molecule detail. We observed nanoscale co-clusters of PAmCherry-tagged Sx1A and Dronpa-tagged α11.2 at a ~1:1 ratio. PAmCherry-tagged Sx1AC145A, or PAmCherry-tagged Sx2, an inactive Cav1.2 modulator, in which Cys145 is a Ser residue, showed no co-clustering. These results are consistent with the crucial role of the single cytosolic Sx1ACys145 in clustering with Cav1.2. Cav1.2 and the functionally inactive transmembrane-domain double mutant Sx1AC271V/C272V engendered clusters with a ~2:1 ratio. A higher extent of co-clustering, which coincides with compromised depolarization-evoked transmitter-release, was observed also by oxidation of Sx1ACys271 and Cys272. Our super-resolution-imaging results set the stage for studying co-clustering of the channel with other exocytotic proteins at a single-molecule level.


Data analyses
PALM rendering -PALM images were analysed using the NSTORM module in NIS-Elements (Nikon) or the published ThunderSTORM software 1 to identify peaks and group them into functions that reflect the positions of single molecules 2 . The peak grouping procedure used a distance threshold and a temporal gap to account for possible molecular blinking 3 . The temporal gap was determined for each fluorophore separately in order to minimize over-counting of molecules. Rendering of individual molecules used intensities that correspond to the probability density values of their fitted Gaussians. Considering the maximal probability density values detected in the field set these intensities. Three dimensional PALM was conducted and analysed using the astigmatism method 4 . Calibration was conducting using 100nm Tetraspec fluorescent beads (Invitrogen). Three-dimensional decoding was performed using Nikon NSTORM software.

Second order statistics and pair correlation function
In this study, we used multiple statistical tools to describe the self-and co-clustering of molecules, as detected by PALM. Self-clustering of a single species can be analysed using second order statistics (e.g. univariate pair-correlation function, or PCF) or using a distance-dependent clustering. Co-clustering can be analysed using bivariate second-order statistics (e.g. bivariate PCF, or BPCF). We briefly describe these approaches below, while more details can be found in a recent review 5 . (Fig. S1a) -For two point patterns that represent two different populations, the bivariate PCF, g12(r), is defined as follows 6 :

Second order statistics
(1) g12(r)= [The number of points of pattern 2 at distance r from an arbitrary point of pattern 1] Where is the mean density of points of pattern 2 6 . Following a similar notation to Wiegand and Moloney 6 , a bivariate PCF can be calculated for a pixelated image using the following definitions: where, is the ring with radius r and width w centred on the k'th point of type 2 (here points of type 2 are simply type i molecules, or S2, as defined above). ni is the total number of points of type i in the study region of area A. The operator Pnts [Sj,X] counts the points of type j, namely Sj, in region X. The operator Area counts the number of cells in the region X. The related Wiegand-Moloney's O12(r) function is defined by 6 : .
For a single population pattern, the univariate PCF is defined as: g(r)≡ g11(r), in analogy to the g12(r) function.
Molecular clustering (e.g. Fig. 1c,f) -An alternative approach to the second order statistics is to identify clusters of proteins and then study various properties of these clusters. Since all proteins in our sample are equal, we perform non-hierarchical clustering using a nearest-neighbor distance 7 , and with an a-priori unknown number of clusters.
Molecular co-clustering analysis (Fig. 5c,d) -For the content analysis of co-clusters by two species, we first generated shrunk pictures (each pixel is 100x100 nm containing multiple green and red dots, i.e. molecular positions, in the corresponding area). We defined the clusters' pixels as having at least 50 red dots (i.e., 50/0.01 µm2) and 5 green dots (5/0.01 µm 2 ). Then we generated a binary picture with pixels filtered according to the above criteria. In this binary picture, we found connected components using the Matlab function 'regionprops'. This step yielded the clusters to be analyzed further for counting their number of incorporated green and red dots.
First we got cluster's shrunk picture (now, each pixel is 10x10 nm Null models -For interpretation of the statistical analyses above, it is useful to compare them to different models: The model of Complete Spatial Randomness (CSR; e.g. in Fig 1b,e) serves to quantify the deviation of a point pattern from randomly distribution, where points are distributed according to the Poisson process. The assumption is that there is no interaction between the points of the pattern and, thus, the points have a constant density as the first order statistics over the study region. The resultant PCF equals 1, i.e. g(r)=1, regardless of r.

Considering two different species, two orthogonal processes would result in a No
Interaction (NI) model (Fig. S1a). Here, the model results in flat PCFs with a value of 1, i.e. g12(r) =1, regardless of r. This indicates no interaction (i.e. no spatial correlation) between the species. Prior knowledge on the physical binding of the two species (e.g. from biochemical assays) can then help to interpret the studied interactions as physical binding events of the species.

Comparing bivariate PCFs of multiple cells
The RL model is individual to each cell and there is no simple way to compare the bivariate PCFs (BPCFs) from multiple cells. Here, we apply two complementary ways to compare and average BPCFs computed for multiple cells: (i) the extent of mixing (EOM), and (ii) the standardized bivariate PCF (SBPCF). Fig. S1c) -Previously, we have introduced a measure we termed the 'Extent of Mixing' (EOM) 8,9 . As a first step, this measure is computed individually for each cell as follows:

(i) The extent of mixing (EOM;
( 3) where is the BPCF for cell i, and is the average of 19 simulated BPCFs due to the RL model for cell i. Next, the EOM can be readily averaged for multiple (N) cells: The errors are computed as: The EOM typically yields PCF values that range between 0 (for the NI model) and 1 (for the RL model). Thus, it provides an intuitive measure of the extent of interaction between two species. Note that this measure is normalized, and thus, absolute correlation values cannot be compared.
where, is the set of all simulated for cell i (i.e. the previous notation in Eqs. 1,2 is now replaced with the more elaborate term ). We further define: