Microtubule polarity determines the lineage of embryonic neural precursor in zebrafish spinal cord

The phenomenal diversity of neuronal types in the central nervous system is achieved in part by the asymmetric division of neural precursors. In zebrafish neural precursors, asymmetric dispatch of Sara endosomes (with its Notch signaling cargo) functions as fate determinant which mediates asymmetric division. Here, we found two distinct pools of neural precursors based on Sara endosome inheritance and spindle-microtubule enrichment. Symmetric or asymmetric levels of spindle-microtubules drive differently Sara endosomes inheritance and predict neural precursor lineage. We uncover that CAMSAP2a/CAMSAP3a and KIF16Ba govern microtubule asymmetry and endosome motility, unveiling the heterogeneity of neural precursors. Using a plethora of physical and cell biological assays, we determined the physical parameters and molecular mechanisms behind microtubule asymmetries and biased endosome motility. Evolutionarily, the values of those parameters explain why all sensory organ precursor cells are asymmetric in flies while, in zebrafish spinal cord, two populations of neural precursors (symmetric vs asymmetric) are possible.


Figure S1 :
Figure S1: Clustering methodology and Par3 asymmetry.a-d GMM (a-b) or K-mean clustering (c-d) of Sara endosome ratio as a function of spindle-MT enrichment (n=62 NPs).a, c Sara endosome ratio and spindle-MT measured in pole B. b, d Sara endosome ratio and spindle-MT measured in the pole having more Sara endosome.e, Sum z-projection showing spindle-MTs (GFP-DCX, green) and Par3 (Par3-mCherry, magenta).LUT shows respective densities.Relative pole B percentage of spindle-MT enrichment and Par3 ratio are indicated.Dashed lines, cell contours.Bar, 5µm.f, GMM clustering of relative pole B Par3 ratio as a function of spindle-MT enrichment (n=21 NPs, DI=80.9%).Two clusters are found for symmetric (blue) and asymmetric (red) NPs.Note that in this report we focus on cortical Par3, and do not consider the cytoplasmic pool as studied in 17 .g, Relative pole B Sara endosome ratio as a function of spindle-MT enrichment for control (grey, n=61 NPs) and Par3 MO (green, n=14 NPs) datasets.Par3 MO dataset can be clustered in two NP pools (DI=85.7%).h, Histogram of spindle-MT enrichment measured in pole B of Par3 MO NPs (n=25).Above, data are clustered (GMM clustering analysis) into two groups (DI=92.0%),symmetric (blue dots) and asymmetric (red dots) NPs.i, Percentage of asymmetric NPs (according to spindle-MT enrichment) for all photoconverted NPs (n=32), n•p (n=15), n•n (n=11) or p•p lineages (n=6).*, p< 0.05; **, p<0.01.Non indicated comparison, non-significant (N.S).Chi-square tests, 95% confidence.

Figure S2 :
Figure S2: Space and time normalization of Sara endosome tracking.a, Scheme of a dividing NP indicating α, ƴ and positions β (centrosomes and spindle center, respectively).Grey region, central spindle area with antiparallel MT array.Dash box, Sara endosome tracking area.Green, spindle-MTs and their orientation (+ and -ends).b, Scheme of Sara endosome spatial registration.E1, Sara endosome position.Red arrows, orthogonal projection of E1 location on the line connecting either centrosome α or ƴ with the spindle center β set as origin.The orthogonal length between E1 and its projection is set as new y coordinate and the length between E1 projection and β is set as new x coordinate.The angle ƴ ̂ is used for time normalization of Sara endosome tracks.c, Methodology of time registration with registered t=0s set as the time preceding ƴ ̂ important decrease (>10°) (red dot).d-f, Traces of angles ƴ ̂ (d-e; n=23 NPs) or their mean (f) as a function of non-registered time (d) or registered time (e-f).

Figure S3 :
Figure S3: Density plots and randomized dataset .a, Histogram showing the frequency of Sara endosome number (n) per bin (bin Δx=0.5µm, 20 bins per time point, see methods) at registered time t=0s from asymmetric heatmap dataset (f).Black line, fit of expected Poisson statistics around λ(t=0s) (mean number of endosomes per bin).Below, custom density LUT based on √λ scale units showing high number of endosomes (red), number of endosomes close to λ (white) and low number of

Figure S4 :
Figure S4: NP clusters merge after cytokinesis and methodology of acquisition for physical parameters of Sara endosome motility.a-b, Histograms of Sara endosome ratio measured in the pole having more Sara endosomes for late cytokinesis (a; n=29 NPs) or Kressmann et al. 1 (b; n=45 NPs) datasets.Above, clustering of the data merges (green dots), and NPs cannot be clustered in symmetric/asymmetric pools as in Fig. 1d.c, Individual MSD traces as a function of delay for each Sara endosome track in combined dataset (n=51 NPs, 337 tracks).d, Weighted average MSD as a function of delay for control combined dataset (blue line).Blue dashed line, quadratic fit.Black line, linear fit.Value of diffusion () calculated from the quadratic fit (R²=0.99,95% confidence, see methods) is indicated.Shade, SEM.e-f, Histograms of the distribution of duration of segments on transport (e; n=51 segments) and diffusive state (f; n=1009 segments) (see methods).Black line, exponential fit.Estimation of  off and  on  from the exponential fits and their correlation coefficient (R², 95% confidence) are indicated.

Figure S5 :
Figure S5: Rescue, MSD, Heatmaps and ANOVA for KIF16Ba MO dataset: a, Relative pole B Sara endosome ratio as a function of spindle-MT enrichment for control (grey, n=61 NPs), KIF16Ba MO (red, n=18 NPs) and rescue of KIF16Ba MO (blue, n=17 NPs) datasets.b, Individual MSD traces as a function of delay for each Sara endosome tracks in KIF16Ba MO dataset (n=18 NPs, 79 tracks).c, Weighted average MSD as a function of delay for control KIF16Ba MO dataset (red line).Red dashed line, linear fit.Black line, quadratic fit.Value of  calculated from the linear fit (R²=0.98,95% confidence, see methods) is indicated.Shade, SEM.d-e, Spatio-temporal density plot of Sara endosome binned number as a function of registered time for KIF16Ba MO dataset as in Supplementary Fig. 3c, e-f (d) and Fig. 3b-c (e).

Figure S6 :
Figure S6: CAMSAP domains, morphants and mutants, and GFP-SARA CRISPR Knock-in: a, Scheme of SMART predicted domains for the different CAMSAP proteins found in zebrafish and comparison with fly Patronin (see methods for NCBI sequence IDs).CH domain (Red), Coil Coiled domains (green) and CKK domain (blue) are displayed from N-terminal (left) to C-terminal (right) of the corresponding protein with the relative number of base pairs indicated.b-c, Maximal z-projection showing CAMSAP3a (mScarlet-CAMSAP3a, magenta) and microtubules (GFP-DCX, green) in 24hpf zebrafish spinal cord (b)

Figure S7 :
Figure S7: Heatmaps and ANOVA for CAMSAP2a -/-; CAMSAP3a MO dataset: a-b, Spatio-temporal density plot of Sara endosome binned number as a function of registered time for CAMSAP2a -/-; CAMSAP3a MO dataset as in Supplementary Fig. 3c,e-f (a) and Fig. 3b-c (b) NPs (n=17 NPs, 1523 endosomes).ANOVA comparison of Sara endosome mean densities as a function of registered time between cell center and cell sides for CAMSAP2a -/-; CAMSAP3a MO dataset as in Fig. 3d-f.

Figure S8 :
Figure S8: Heatmaps and ANOVA for Dlic1 MO dataset: a-b, Spatio-temporal density plot of Sara endosome binned number as a function of registered time for Dlic MO dataset as in Supplementary Fig. 3c,e-f (a) and Fig. 3b-c (b) NPs (n=11 NPs, 2137 endosomes).c, ANOVA comparison of Sara endosome mean densities as a function of registered time between cell center and cell sides for Dlic MO dataset as in Fig. 3d-f.