An Optogenetic Demonstration of Motor Modularity in the Mammalian Spinal Cord

Abstract

Motor modules are neural entities hypothesized to be building blocks of movement construction. How motor modules are underpinned by neural circuits has remained obscured. As a first step towards dissecting these circuits, we optogenetically evoked motor outputs from the lumbosacral spinal cord of two strains of transgenic mice - the Chat, with channelrhodopsin (ChR2) expressed in motoneurons, and the Thy1, expressed in putatively excitatory neurons. Motor output was represented as a spatial field of isometric ankle force. We found that Thy1 force fields were more complex and diverse in structure than Chat fields: the Thy1 fields comprised mostly non-parallel vectors while the Chat fields, mostly parallel vectors. In both, most fields elicited by co-stimulation of two laser beams were well explained by linear combination of the separately-evoked fields. We interpreted the Thy1 force fields as representations of spinal motor modules. Our comparison of the Chat and Thy1 fields allowed us to conclude, with reasonable certainty, that the structure of neuromotor modules originates from excitatory spinal interneurons. Our results not only demonstrate, for the first time using optogenetics, how the spinal modules follow linearity in their combinations, but also provide a reference against which future optogenetic studies of modularity can be compared.

Figures

Figure 1. Experimental setup and design.

(A) Immunostaining of Chat spinal slices indicates that the range of ChR2 expression (green) included neurons in the ventral horn (laminae VIII and IX) and the intermediolateral nucleus (IML), neurons that were presumably cholinergic. (B) Immunostaining of Thy1 slices indicates ChR2 expression through most of the spinal grey matter, with expression being the strongest in the intermediate zone (lamina VII) where interneurons monosynaptically connected to the motoneurons of different muscles are most dense. (C) Example time traces of ankle force optogenetically elicited from the lumbosacral spinal cord of Chat (top panel) and Thy1 (bottom panel) mice. The traces of absolute force magnitude (defined over the X-Y plane) shown are averages across data evoked from multiple ankle locations in the workspace (N = 15 for Chat; N = 20 for Thy1), but from stimulations applied to the same spinal locus. Blue shadings indicate the time windows in which optical stimulations were applied. (D) The Chat mouse exhibited a shorter delay of force emergence after optical stimulation onset as compared with the Thy1 mouse (*U-test, p < 0.01; mean ± SD). (E) The motor output for any spinal stimulation locus was represented as a spatial force field. Isometric hind-limb forces optogenetically evoked at any spinal locus were recorded as the ankle was moved to different workspace locations by an automatic servo mechanism. (F) Examples of spatial force fields recorded from the setup depicted in E, derived from the Chat (left panel) and Thy1 (right panel) mouse strains. (G) In general, the Chat force fields tended to be composed of parallel vectors, while the Thy1 force fields, non-parallel vectors. This difference in structure is supported by the result that the variability of force-vector direction within each field, quantified by the circular standard deviation of the vectors’ azimuth angles, was much smaller in the Chat-fields than in the Thy1-fields (*, U-test, p < 0.01; mean ± SE).

Figure 2. Topographic organization of spinal force fields in the Chat mouse.

(A) We summarized force fields obtained from 5 mice, elicited using a single laser at the lowest power used at each locus (N = 30). Each force field is represented in the figure by the azimuth angle of the vector sum of all vectors in the field (y-axis), and by the spinal stimulation locus located with respect to the anterior edge of the T12 vertebra (x-axis). The Chat force fields were grouped into 6 clusters by our procedure based on k-means and the silhouette values. Individual force-field examples for all clusters are shown (Ex. 1 to Ex. 6). For each cluster, dotted lines mark the cluster mean ± SD along both the horizontal and vertical directions. (B) The average force fields for the 6 clusters shown in (A).

Figure 3. Topographic organization of spinal force fields in the Thy1 mouse.

(A) We summarized force fields obtained from 10 mice, elicited using a single laser at the lowest power used at each locus (N = 87). Each force field is represented in the figure by the azimuth angle of the vector sum of all vectors in the field (y-axis), and by the spinal stimulation locus located with respect to the anterior edge of the T12 vertebra (x-axis). The Thy1 force fields were grouped into 5 clusters by our procedure based on k-means and the silhouette values. Individual force-field examples for all clusters are shown (Ex. 1 to Ex. 5). For each cluster, dotted lines mark the cluster mean ± SD along both the horizontal and vertical directions. (B) The average force fields for the 5 clusters shown in (A). (C) The percentage overlap between the spatial ranges of stimulation loci of every pair of clusters was calculated for each animal, and averaged across animals. These percentage values are shown here as a heat map (blue = 0 overlap; red = 100% overlap). For example, the colour shown in the box at row 2, column 1 denotes the percentage of the range of cluster 2 that overlapped with the range of cluster 1. Black means there was insufficient or no data for the cluster pair for this evaluation.

Figure 4. Equilibrium points (EP) in force fields evoked from the Chat and Thy1 animals.

(A) Among all the Thy1 force fields we obtained using a single laser (N = 212 from 10 mice), EPs and “virtual” EPs located just outside the workspace were detected in 71 of them (33.5%). This figure summarizes the spatial distribution of the Thy1 EPs and virtual EPs, with the EPs from each mouse (whose name is prefixed by the letter “T”) denoted by a unique symbol. The workspace of the hind-limb is demarcated by dotted lines. It is visually apparent that there were 2 spatial clusters of EPs within the workspace, one located antero-dorsally, and another, postero-ventrally. Four specific examples of force fields with EPs are also shown. In each, the EP is marked with a black dot. (B) Among all the Chat force fields we obtained using a single laser (N = 93 from 5 mice, with 2 mice producing no force field with EP), EPs and virtual EPs located just outside the workspace were detected in 19 of them (20.4%). The EPs and virtual EPs from each mouse (whose name is prefixed by the letters “CT”) are denoted by a unique symbol. All Chat EPs were spatially circumscribed within 2 small regions near the workspace boundary, and for each, a specific example of force field with EP is shown. In both panels (A,B) since many EPs detected were located at the same nodes within our 8 × 8 force-vector grid, to visually clarify the spatial density of the EPs, we have added a very small Gaussian noise to both the x- and y-coordinates of the EPs when producing this figure.

Figure 5. Optogenetically evoked fields in both Thy1 and Chat displayed linear combination.

(A) To assess whether the property of linear combination holds for force fields optogenetically derived, we elicited force fields with co-stimulation from two laser beams, placed at separate spinal loci. (B) An example from a Chat mouse showing how the force field produced by co-stimulation (rightmost panel) can be reconstructed by linearly combining the force fields separately evoked at the individual loci (laser 1 and laser 2). The coefficients for this linear combination, shown to the left of the individual fields, were found by linear regression. The locations of the spinal stimulation loci, measured in mm with respect to the anterior edge of the T12 vertebra, are shown on top of the individual force fields. Similarity between the reconstructed and co-stimulation force fields was quantified using equation (5) stated in Methods. (C,D) Two examples from two different Thy1 mice demonstrating how spinal force fields could be linearly combined to explain fields derived from co-stimulation. (E,F) Similarity between the reconstructed and co-stimulation force fields was plotted against the coefficient ratio, defined as the smaller of the two combination coefficients divided by the larger coefficient, for the Chat (panel E) and Thy1 (panel F) mice, respectively. A coefficient ratio of 0 (x) implies that a “winner-take-all” model produces a better fit. Importantly, considering the cases with coefficient ratio > 0 (•), the average similarity values for the Chat and Thy1 were not significantly different from each other (p > 0.05). Also, in both mouse strains, when the coefficient ratio was > 0, similarity was not dependent on the coefficient ratio (p > 0.05), further suggesting the generality of the linear-combination model of force-field modules. In Thy1 (panel F), when coefficient ratio = 0, some co-stimulation fields were well explained by the winner (red x), but others were explained poorly (green x). Horizontal dotted lines mark mean ± SD of the similarity values for cases with coefficient ratio > 0. In (F) two specific data points with a coefficient ratio of 0 (marked with arrows) are shown in Fig. 6 as examples of winner-take-all.

Figure 6. Examples of Thy1 co-stimulation fields better explained by a winner-take-all model.

(A) In this example, our regression analysis found that a coefficient ratio of 0 was the best fit; but the “winner” field and the co-stimulation field matched poorly (similarity of 0.435). (B) An example in which the “winner” field and the co-stimulation field matched extremely well (similarity of 0.925).