Diacylglycerol-dependent hexamers of the SNARE-assembling chaperone Munc13-1 cooperatively bind vesicles

Significance Munc13-1 is a molecular chaperon that facilitates recruitment and docking of synaptic vesicles at the active zone of the synapse. Using model membranes and statistical modeling, we show that Munc13-1 forms mainly hexamers on diacylglycerol rich microdomains on lipid membranes. These hexamers act cooperatively to capture vesicles from a solution. Statistics show that each hexamer binds to one vesicle. Disruption of the interactions at Munc13-1 hexagonal interface by point mutations based on crystallographic data alters its oligomerization state and hexamers are no longer observed. Furthermore, the mutant oligomer loses its cooperativity in binding vesicles. Our study suggests that the Munc13-1 hexamers observed on lipid bilayers resemble the hexagons revealed by crystallography.


Models of Munc13-1 binding to DAG microdomains
In the following models we will assume that the DAG microdomains are of uniform area which is reasonable considering the measured area distribution (see main text).
1. Isolated Munc13-1 that bind DAG microdomains independently In this first model, we will assume that the clusters are just an optical bias due to the colocalization of isolated monomeric Munc13-1 molecules in the same DAG microdomain.According to the non-cooperativity of the Munc13-1 binding, the distribution of the number of Munc13-1 monomers, i.e., the apparent cluster copy number in this model, should follow a Poisson probability: where  is the number of monomers and 〈〉 the mean apparent cluster copy number.〈〉 cannot be directly measured experimentally because the domains without any Munc13-1 are not visible, i.e.,  0 is unknown.Hence, the observed copy number probability is: Eq. 3 is valid only for  1.The mean of the observed cluster copy number distribution is therefore: (S3) As expected, when 〈〉 is large, it can very well be approximated by 〈〉 because there is hardly any DAG microdomain without Munc13-1 cluster.Experimentally, we found 〈〉 〈〉 4.8 Munc13-1 per cluster.The predicted and observed cluster copy numbers are displayed in Fig. 2 and S2.They clearly do not match indicating that this model is not appropriate here.

Mixture of isolated Munc13-1 and K-mers
In this second model, we assume that Munc13-1 is present in two forms on the bilayer: monomers and K-mers.These two populations behave independently of each other and are randomly distributed among the cluster.Hence, each would present a copy number distribution following the Poisson probability, exactly as in the first model: where  (resp.) is the number of monomers (resp.K-mers) and 〈〉 (resp.〈〉) the mean number of monomers (resp.K-mers) per cluster.
In principle, each cluster is a mix of monomers and pre-determined oligomers.In case of hexamers, the distribution of monomers and hexamers is exemplified in Supporting Table S1.Hence the probability distribution   of the cluster at size N, i.e., with N Munc13-1 molecules in the cluster, is: 〈〉 can be estimated exactly as in the first model by from the observed mean copy number obtained by only considering clusters with 1 to  1 copies of Munc13-1, 〈〉 , by rewriting Eq. ( 3): 〈〉 is then obtained numerically from Eq. S6.
Finally, 〈〉 is well approximated by: where <k> is the mean number of K-mers per cluster, and 〈〉 is obtained exactly as in the first model.
Once 〈〉 and 〈〉 are established the probability distribution described by Eq.S5 is completely defined.When the K-mers are hexamers, the probability distribution obtained from Eq. S5 with these values for 〈〉 and 〈ℎ〉 is displayed in Fig. 2 and matches very well the experiment cluster copy number distribution.
Another prediction from this model is that the fraction of DAG microdomains without any Munc13-1 is:  0  〈 〉  〈 〉 = 0.06 = 6% (S7) Experimentally we did not find any Munc13-1 molecule in 9% of DAG microdomains which is reasonably close to the predicted value considering the simplicity of the model.

Statistical comparison of the predicted and observed distribution of oligomers
To determine what oligomer seems to describe best the observed cluster size distribution, we use a parameter to test the oligomerization degree, O, that quantitatively compares the predicted and observed histograms for each -mer  ∑ (S8)  was chosen as the largest cluster observed experimentally, 18.   in the numerator is used to reduce the impact of the error on the less frequent cluster sizes and     in the denominator directly represents the difference between the observation and the prediction.A larger  indicates a better approximation of the experimental distribution.For instance, in Fig. 2E (wild-type) and 3G (mutant), the best approximations are respectively obtained for hexamers and tetramers.Table S1: the table is an example of the different possibilities to obtain clusters of various copy numbers by combining monomers and hexamers.For instance, a cluster with 10 Munc13-1 can be due to 10 monomers or 4 monomers and 1 hexamer.