Figure 3. Cluster analysis of V1^R firing pattern at E12.5.

(A, insets) Cluster analysis of embryonic V1^R firing pattern was performed using three parameters that describe the firing pattern during a 2 s suprathreshold depolarizing pulses: the mean of the half-amplitude event duration (mean ½Ad), the coefficient of variation of ½ Ad (CV ½Ad) allowing to quantify the action potential variation within a train (CV was set to 0 when the number of spikes evoked by a depolarizing pulse was ≤3) and the duration ratio ddr = Σ½ Ad/Pw, obtained by dividing the sum of ½ Ad by the pulse duration Pw, that indicates the total time spent in the depolarized state. For example, ddr = 1 when a plateau potential (PP) lasts as long as the depolarizing pulse. Conversely, its value is low when the depolarizing pulse evokes a single AP only. (A) Dendrogram for complete linkage hierarchical clustering of 164 embryonic V1^R (N = 140) according to the values of log mean ½Ad, CV ½Ad, and log ddr. The colored matrix below the dendrogram shows the variations of these three parameters for all the cells in the clusters (colored trees) extracted from the dendrogram. (B) The number of clusters was determined by analyzing the distribution of silhouette width values (see Materials and methods). The box plots show the distribution of silhouette width values when the number of clusters k varies from 2 to 12. The mean silhouette width values (red diamond-shaped points) attained their maximum when the estimated cluster number was 5. (C) 3D plot showing cluster distribution of embryonic V1^R according to log mean ½Ad, CV ½Ad, and log ddr. Each cluster corresponds to a particular firing pattern as illustrated in (D). V1^R that cannot sustain repetitive firing of APs (1–3 AP/pulse only, gray, single spiking [SS]), V1^R that can fire tonically (red, repetitive spiking [RS]), V1^R with a firing pattern characterized by a mix of APs and relatively short PPs (dark purple, mixed event short PP [ME short PP]), V1^R with a firing pattern characterized by a mix of APs and relatively long plateau potentials (light purple, mixed event long PP [ME long PP]), and V1^R with evoked PPs only (blue, PP). The arrow in (C) indicates three misclassified V1^R that could not sustain repetitive firing although they were assigned to the cluster of repetitively firing V1^R (see text).

Figure 3—figure supplement 1. Distributions of log half-amplitude event duration (log ½Ad), coefficient of variation of ½ Ad (CV ½Ad), and log depolarizing duration ratio (log ddr) values related to the cluster analysis of embryonic V1^R firing patterns.

(A1) Histogram of log mean ½Ad for the whole V1^R population at E12.5 (n = 164; bin width 0.1). The histogram was well fitted by the sum of three Gaussian curves with means and SDs of 1.135, 2.046, and 2.84, and 0.316, 0.181, and 0.21, respectively. (A2) Histogram of the values of log mean ½Ad sorted after cluster analysis showing single spiking (SS) V1^R (gray), repetitive spiking (RS) V1^R (red), mixed events (ME) V1^R with short plateau potentials (ME short PP V1^R, light purple), ME V1^R with long plateau potentials (ME long PP V1^R, dark purple), and PP V1^R (blue). log mean ½Ad was significantly different between SS V1^R, PP V1^R, the whole ME V1^R population (ME_s and ME_l V1^R), and PP V1^R (Kruskal–Wallis test p<0.0001; SS V1^R versus RS V1^R, p<0.0001; SS V1^R versus ME V1^R, p<0.0001; SS V1^R versus PP V1^R, p<0.0001; RS V1^R versus ME V1^R, p=0.0004; RS V1^R versus PP V1^R, p<0.0001; ME V1^R versus PP V1^R, p=0.018; SS V1^R n = 46, RS V1^R n = 69, ME_s V1^R n = 9, ME_l V1^R n = 4, PP V1^R n = 35). (B1) Histogram of CV ½Ad for the whole V1^R population at E12.5 (n = 164; bin width 5%). Note that a large population of V1^R had zero CV ½Ad (n = 83). The histogram for CV ½Ad ≠ 0 was fitted by the sum of three Gaussian curves with means and SDs of 23.4, 68.4, and 117 (%) and 8.9, 6.8, and 4.1, respectively. (B2) Histograms of the values of CV ½Ad sorted after cluster analysis showing SS V1^R (black), RS V1^R (red), ME_s V1^R (light purple), ME_l V1^R (dark purple), and PP V1^R. CV ½Ad was not significantly different between SS V1^R and PP V1^R (CV ½Ad of SS V1^R and PP V1^R = 0.682% and 0%, respectively: only one of the 46 SS V1^R displayed 3 PA and had a CV ½Ad of 31.37). CV ½Ad was significantly different between RS V1^R and the whole ME V1^R population and also between SS V1^R or PP V1^R and RS V1^R or ME V1^R (Kruskal–Wallis test p<0.0001; SS V1^R versus RS V1^R, p<0.0001; SS V1^R versus ME V1^R, p<0.0001; SS V1^R versus PP V1^R, p=0.846; RS V1^R versus ME V1^R, p=0.0003; RS V1^R versus PP V1^R, p<0.0001; ME V1^R versus PP V1^R, p<0.0001). (C1) Histogram of log ddr (sum of ½Ad divided by pulse duration) for the whole V1^R population at E12.5 (n = 164; bin width 0.2). The histogram was fitted by the sum of two Gaussian curves with means and SDs of −2.51 and −0.851, and 0.2 and 0.46, respectively. (C2) Histograms of the values of log ddr sorted after cluster analysis showing SS V1^R (black), RS V1^R (red), ME_s V1^R (light purple), ME_l V1^R (dark purple), and PP V1^R. log (ddr) was not significantly different between ME V1^R and PP V1^R, while it was significantly different between SS V1^R and RS V1^R, SS V1^R and the whole ME V1^R population, SS V1^R and PP V1^R, RS V1^R and the whole ME V1^R population, RS V1^R and PP V1^R (Kruskal–Wallis test p<0.0001; SS V1^R versus RS V1^R, p<0.0001; SS V1^R versus ME V1^R, p<0.0001; SS V1^R versus PP V1^R, p<0.0001; RS V1^R versus ME V1^R, p<0.0001; RS V1^R versus PP V1^R, p<0.0001; ME V1^R versus PP V1^R, p=0.977). ME_s V1^R and ME_l V1^R differed only by their CV ½Ad (Mann–Whitney test, log mean ½Ad for ME_s V1^R versus log mean ½Ad for ME_l V1^R, p=0.26; CV ½Ad for ME_s V1^R versus CV ½Ad ME_l V1^R, p=0.0028 and log ddr for ME_s V1^R versus log ddr for ME_l V1^R, p=0.1483). It is noteworthy that the distribution of the values of each metric was multimodal, thus indicating that each of them could partially discriminate different groups of embryonic V1^R according to their firing pattern.