New method to record visual response of spines without BAP signals.

a. A gene encoding the inhibitory optogenetic protein SwiChR++ is introduced by AAV into layer 2/3 pyramidal neurons in mouse V1. b. GCaMP6s is used for the functional spine imaging. Cre-loxP system is used for the sparse expression of GCaMP6s by diluting the Cre-expressing vector. SwiChR++ is used to inhibit the somatic activity. A mixture of three types of AAV is injected to V1 A neuron that expresses both GCaMP6s and mCherry which is the fluorescence marker for the expression of SwiChR++ is shown. c. Before photoinhibition, soma shows orientation-selective responses (p value for the response = 2.01E-8 and Max ratio change=0.6057, see Methods for the definition). d. After photoinhibition, the visual response of soma is significantly abolished (p-value for the response = 0.097 and Max ratio change=0.0129, see Methods for the definition).

Comparison between the subtraction and photoinhibition methods for the visually responsive spine activity.

The preferred orientation of the soma of this example neuron is horizontal (coded yellow). a. Left: An example image of averaged dendritic branch. Middle: An orientation map of the same dendritic branch without BAP inhibition. Right: An orientation map of the same dendritic branch after BAP inhibition. b. Polar plots of visual responses from orientation-selective spines. All the spines show similar orientation preference between BAP-subtracted and BAP-inhibited methods. c. Cumulative plots of dF/F of all the orientation-selective spines with preferred orientation different from the soma of without subtraction, BAP-subtracted and BAP-inhibited methods. There is no statistically significant difference among different methods (p=0.86, without subtraction vs. BAP-inhibited; p=0.94, BAP-subtracted vs. BAP-inhibited; p=0.92, without subtraction vs. BAP-subtracted; paired-sample t-test). d. Cumulative plots of dF/F of all the orientation-selective spines with preferred orientation same as the soma of BAP-subtracted and BAP-inhibited methods. There is no statistically significant difference between two methods (p=0.45 paired-sample t-test). e, f Bar plots of difference of preferred orientation (ΔOri) between BAP-subtracted and BAP-inhibited spines with preferred orientation different from the soma (e) and same as the soma (f). g. Cumulative plots of gOSI of all the orientation-selective spines with preferred orientation different from the soma of BAP-subtracted and BAP-inhibited methods. There is no statistically significant difference between two-methods (p=0.29 paired-sample t-test). h. Cumulative plots of gOSI of all the orientation-selective spines with preferred orientation same as the soma of BAP-subtracted and BAP-inhibited methods. There is no statistically significant difference between two-methods (p=0.50 paired-sample t-test).

Functional spine map of a layer 2/3 pyramidal neuron.

a. A functional spine map from ∼1,000 spines recorded from the basal dendrites of neuron 1. Orientation-selective spines are illustrated by different colors, and uncolored spines are either visually unresponsive or visually responsive but nonselective. b. A spine orientation map from branch 11 of neuron 1 (see Table 1). Among nineteen spines recorded from this branch, eight spines selectively responded to orientation stimulation. c. Polar plots of orientation-selective spines in branch 11.

The preferred orientation of the soma can be predicted from the number of orientation-selective spines.

a. Distributions of preferred orientations (left) and directions (right) of spines in neuron 1. The peaks of the distribution match the preferred orientation and direction of the soma (red arrow). b. The distribution of the difference in preferred orientation between each spine and soma (ΔOrientation) of neuron 1. About half of spines are tuned similarly to the soma. c. Averaged data from 6 neurons shows the same trends as neuron 1 data.

Summary of recorded spines

Summation of calcium time courses of spines can predict the preferred orientation of the soma, but with the broad tuning curves.

a. The time course of the soma (upper panel; average of 10 trials) and the average of the spine signals (lower panel; average of 10 trials from 982 spines). The OSI and preferred orientation are shown for both datasets. Polar plots suggest the broader tuning of the average of the spine signals than the tuning of the soma. b. The tuning curve of the soma (upper panel; average of 10 trials) and the average of the spine tuning curves (lower panel; average of 10 trials from 982 spines). Polar plots suggest the broader tuning of the average of the spine tuning curves than the tuning curve of the soma. Although the OSI from the average of the spine signal and tuning curve is lower than that of the soma, the preferred orientation can be effectively predicted in both cases.

Functional clustering at the spine-pair level

The difference in preferred orientation (ΔOrientation) are plotted as a function of the distance between spines on each dendritic branch and the actual data (a, b, black line) are compared with the shuffled data (a, b, blue line). In the shuffled data, all the pines were shuffled, while maintaining the original spine locations and the same proportion of orientation-tuned spines. Shuffling was done either across cell (a, blue line) or within each dendritic branch (b, blue line). ΔOrientation between neighboring spines (< 3μm) of actual data is significantly smaller than that of shuffled data across a cell (a, p = 0.023 Mann–Whitney U-test with the Bonferroni correction) and smaller but not significantly different than that shuffled within a branch (b, p=0.83 Mann–Whitney U-test with the Bonferroni correction).

The prediction of somatic tuning was improved by the cluster model on the dendrites and the threshold model at the soma

a. The number of responsive spines is plotted as a function of the number of spines in a cluster. The black line shows number of spines responding to any orientation, and the red line shows number of spines responsive to the preferred orientation of the soma. The blue line represents the ratio between the red and black lines. b. The CI between the actual and predicted tuning curves of the soma plotted as a function of the number of spines in a cluster. Polar plots of the actual and predicted tuning curves of the soma are shown. c. The OSI plotted as a function of the number of spines in a cluster. d. The number of responsive spines is plotted as a function of the number of spines for the threshold. Black line shows number of spines responding to any orientation, and red line shows number of spines responsive to the preferred orientation of the soma. The blue line represents the ratio between the red and black lines. e. The CI between the actual and predicted tuning curves of the soma plotted as a function of the number of spines for the threshold. Polar plots of the predicted tuning curve of the soma are shown. f. The OSI plotted as a function of the number of spines used as a threshold.

The combined model further improves soma tuning.

a. A combined model is applied to neuron 5 (see Table 1). CIs are calculated at various combinations of values for the number of spines (1-10 spines) in a cluster and the number of spines (0-70 spines) used as the threshold, and CI values are shown as a heat map. b. Another example of a combined model in neuron 2. c. Population averaged data from six neurons. Heat maps are calculated for each neuron and averaged.

Analysis of dendritic branches with cluster of spines responding to the similar orientations.

a. Two examples of neuron showing the segments with cluster of spines responding to the same orientation as the soma preferred orientation (red) and different orientation from the soma preferred orientation (yellow). Clustered segments are distributed spatially uniform manner on the dendritic field in both neurons. b. Proportion of clustered segments (red and yellow) and non-clustered segments (blue). Among clustered segments (21.9%), the segments of spines responding to the same orientation as the soma preferred orientation (red) is twice as much as the segments of spines responding to the different orientation to the soma (yellow). c. Proportion of spines on the clustered (red and yellow) and non-clustered (blue) dendritic segments. Among spines on the clustered segments, the spines responding to the same orientation as the soma preferred orientation (red) is three times as much as the spines responding to the different orientation to the soma preferred orientation (yellow). d, e. Distribution profiles of clustered segments on the dendritic field are shown. Number of clustered segments is plotted as a function of the distance from the soma (d) and the branch order (e). f. Prediction of soma tuning using spine responses on the clustered dendrites. Two examples from the same neurons shown in (a). upper plot: Actual soma tuning. middle plot: Predicted tuning using spine responses on the clustered segments with the same preferred orientation as the soma. lower plot: Predicted tuning using spine responses on the clustered segments with both the same and the different preferred orientation to the soma. Better prediction was obtained when using the both clustered segments.