Figure 8—figure supplement 1.

Figure 8—figure supplement 1. Our Granger causality pipeline revealed revealed right-to-left information flow among motor-correlated hindbrain neurons in larval zebrafish performing optomotor (OMR) responses. (A) Two-photon laser scanning calcium imaging in vivo was acquired at 5.81Hz in the hindbrain of Tg(elavl3:GCaMP5G) transgenic larvae performing OMR (data from Severi et al., 2018). Forty-one (red and blue) out of 139 neurons from a reference plane of a larval zebrafish hindbrain were motor-correlated and selected for subsequent analysis. Neurons located in the medial area (red, N = 20 neurons) are in the location of the V2a reticulospinal stripe (Kinkhabwala et al., 2011), distinct from other motor-correlated neurons (blue, N = 21). The activity of all other neurons (gray, N = 98 neurons) were not correlated to the motor output. (B) Sampled calcium traces of the three categories of neurons mentioned above were ordered by decreasing signal-to-noise ratio (SNR) from top to bottom. The moving grading (top trace) was presented to the right of the fish from tail to head to induce swimming, depicted as the tail angle (bottom trace, with a zoom over 4s showing two subsequent bouts, positive tail angles correspond to left bends). (C) The distribution of the tail angles for turns (tail angle α with 45^∘< |α|< 90^∘ was biased to the left.(D) GC matrix (left) and map of information flow (right, the arrows indicate the directionality of the GC links) for the naive BVGC on medial motor-correlated neurons. Of the 380 possible pairs, 351 were found to have a significant drive, suggesting that the naive BVGC algorithm is too permissive to false positive links, and justifying an adapted pipeline to remove spurious connections. (E) The neuronal drive, calculated as the sum of the strength of all outgoing GC links for each neuron, was correlated to the SNR of the calcium traces, especially high for medial neurons (r = 0.69, compared to r =0.45 across all 139 neurons). (F) Customization of the threshold and normalization of the GC values reduced the number of significant GC links and revealed numerous commissural links that were both rostrocaudal and caudorostral between medial motor-correlated neurons. (G) The new neuronal drive was less correlated with the SNR (from 0.69 to 0.49 for medial neurons). (H) Due to the high correlation in the selected neurons and the relatively high number of neurons (20), the new MVGC pipeline highlighted only the strongest GC links. (I) The density distribution of the significant GC link angles from the new MVGC analyses revealed a biased right-to-left information flow. The directionality of the information flow is consistent with the presence of the stimulus on the right side of the fish and the fish swimming forward with a bias towards the left (C, right). These results were conserved across planes of the same fish as well as across fish (Figure 4).