Imaging zebrafish neural tube development and single cell tracks of different fates.

(A) Schematic of ventral neural tube anatomy and pattern in zebrafish (partially adapted from Xiong et al., 2013). MFP, medial floor plate; LFP, lateral floor plate; pMN, motor neuron progenitors (MN, motor neurons); p2, V2 interneuron progenitors (V2, V2 interneurons).

(B) Labeling (at 8-16 cell stage) and imaging (at neural plate, neural keel and neural tube stages) scheme for acquiring three channels at the same time. The 405 laser in the first scan line excites the BFP cell tracking marker as well as converting Kaede(Green) to Kaede(Red). NC, notochord.

(C) Example data. Small circles on the image show manual segmentations generated on cells using GoFigure2 software. D-V, dorsal-ventral; A-P, anterior-posterior. Scale bar: 20µm.

(D) A pMN track in a mem-ebfp2, ptch2:kaede/olig2:gfp dataset. Circles in the cells show a 2D view of the spherical segmentations generated using GoFigure2 software.

(E) A LFP track in a h2b-ebfp2, ptch2:kaede/nkx2.2:mgfp dataset. Arrow indicates a sister cell of the centered track. Note that while this transgenic reporter has a membrane tag, the spherical segmentation is still capable of measuring the much weaker cytoplasmic fluorescent signal. See also Movie S1.

Dynamics and heterogeneity of Shh response and fate reporters in average cells and single cells.

(A-D) Average and single (A’-D’) tracks from a ptch2:kaede/olig2:gfp movie focusing on the anterior spinal cord (indicated by the schematic, A-B), and the posterior spinal cord (C-D), respectively. Colored shades around the average tracks represent ±S.D. Individual ptch2:kaede traces (A’,C’) show significant overlap between fates. Note that intensity units are not comparable between different datasets (e.g., between A and C) due to variation in expression, sample depth, mounting position, and imaging settings. The average tracks appear noisier before 10hpf and after 15hpf due to some single cell tracks not being long enough temporally (e.g., a tracked cell leaves the field of view) therefore fewer data points are available for averaging at the ends. LFP, Lateral floor plate cells; pMN, motor neuron progenitors. See also Movies S3,4. (E) Example single tracks that show very similar ptch2:kaede response dynamics but very different olig2:gfp expression, and vice versa. The schematic shows the axis ranges and labels, and legends for the two reporter traces.

Estimation of dynamic reporter activity from fluorescence intensity measurements

(A) Example raw tracks (a pMN [green], a LFP [brown] and a More Dorsal cell [blue] each). The tracks were trimmed to contain only data points between 10 and 15 hpf as this time window is best covered in the tracks. Before 10 hpf the signal is low and by 15 hpf most progenitors have become specified (Xiong et al., 2013).

(B) Tracks in (A) after multiple filter corrections as described in the text. Main trends over long time windows (>0.5 hours) are preserved while short fluctuations are removed by a moving-average smoothing operation.

(C) Rate of change in intensity by the first time derivative of (B).

(D) Modeled signaling level (ptch2:kaede transcription rate) of the same tracks using equation (3). Note the different peak height and duration of different cells. Varying the half-life value between 1 to 5 hours does not substantially change these results (data not shown).

Estimation of fluorescent reporter mRNA half-life.

(A) Heat-shock pulse induction of GFP and timelapse imaging (schematic). 24hpf tg(hsp70l:EGFP) embryos were shocked for 30 minutes in pre-warmed 37°C egg water in a heat block. Imaging began 2 hrs post heat-shock (hph, when the fluorescence was first detectable).

(B) Cell tracks of hsp70l:EGFP to validate equation (3) and estimate mRNA stability. Assuming that the heat-shock between t=0 and t=0.5 causes a transient pulse of mRNA production, the intensity track then covers the simple decay period of mRNA (as no more mRNA is produced after the pulse). The decay is exponential according to equation (3) and governed by the mRNA half-life. The fluorescent intensity increase rate then should correspondingly decrease exponentially towards a plateau value, predicting a linear relationship between time and - ln(1-I/Imax), in which I is raw intensity and Imax is the plateau value of I in later times (when no more mRNA is left and new fluorescent protein production stops). This predicted linear relationship is found for the tracks suggesting the simple model correctly describes the process. The slope of the lines (∼0.4) suggests the mRNA half-life of hsp70l:egfp to be ∼1.7 hours. As I approaches Imax, the quantity of ln(1-I/Imax) becomes sensitive to small fluctuations in I, resulting in the noisy spikes seen after 6 hours. This portion of the data was not used in the fitting.

(C) Estimation of Kaede mRNA stability using in situ hybridization. CyA, cyclopamine, an antagonist of Shh signaling that represses ptch2 promoter output. When soaked in CyA (CyA ctrl), the embryos exhibit a low basal level of Kaede expression in the neural tube. When CyA was added around 9-somite stage and embryos followed over time, the high level of Kaede mRNA as seen in the DMSO control was not maintained but decayed towards the basal level. In the trunk/tail neural tube (but not in the head), the level became similar to basal at +7.5 hours by comparing the CyA and CyA ctrl images. This result suggests Kaede mRNA is around for more than 6 hours after inhibition of signaling, with a half-life of about 3 hours. Scale bar: 400µm.

K-means clustering and temporal re-alignment analysis of tracks across datasets.

(A) K-means clustering combining smoothened tracks from 4 datasets combining ptch2:kaede with different reporters (olig2:gfp, imaged anteriorly: ‘Olig2-Ant’ or posteriorly: ‘Olig2-Post’; mnx1:gfp (‘Mnx’); shh:gfp (‘Shh-2’)).

(B) Fate contribution of clusters in (A). Heatmap shows, for each fate, the fraction of cells in different clusters. Black indicates no cells.

(C) Average ptch2:kaede dynamics by cell fate and dataset from processed tracks.

(D) Similar to (C), realigned starting times by the first time the modeled ptch2 promoter output exceeds a threshold of 0.025 of the normalized activity.

Correlations between position and fate choice in single cells.

(A) Ranking of lateral-medial/dorsal-ventral (LMDV) and anterior-posterior (AP) positions as a control for the correlational analysis. Data from an anterior movie and a posterior movie are compared (schematics). The black lines on the flag mark thresholds that best separate different fates. See descriptions in the text and also methods.

(B) Average position tracks. Note the range of anterior progenitors vs. posterior ones over the same time window, indicating different tissue geometry and cell dynamics along the body axis. LMDV: Lateral Medial-Dorsal Ventral distances. Error bars are S.D.

Correlations between different metrics of Shh response and fate choice in single cells.

(A) Shh response dynamics metrics. Under the modeled activity of the ptch2:kaede promoter, Response time is a count of time points at which ptch2:kaede promoter is “on” (>0.05 A.U./hr2). Average response is the average promoter activity across the whole time window. Maximum response is the highest promoter activity found in the time window.

(B) Flag correlation results of the 3 metrics in (A) (for which half-life=3 hours was used) over different Kaede mRNA half-life values. Note that response time and average response parameters switch predictive power as Kaede mRNA becomes stable. Maximum transient response level stays best over wide ranges of mRNA stability. The conclusion in (A) is therefore robust to this estimated parameter (Figure 4).

(C) Flag correlation results of the 3 parameters in (A) (where smoothing window=3 time points and iteration number=2) over different smoothing window (x axis) and iteration numbers (dark,1;light,2;lighter,3). The conclusion in (A) is therefore robust to this analysis, despite the sensitivity of equation (3) to the smoothing parameters.

(D) Summary of correlation percentages for posterior vs. anterior tracks for different morphogen interpretation models. In addition to metrics mentioned in the main text, cell speeds were also plotted here. The speeds were calculated from the positions of the cells in the track for 2-hour window centered on the early (11.5hpf) and late (14.5hpf) time points.