The authors use two transgenic mouse lines D1-Cre and A2A-Cre with Cre-dependent Channelrhodopsin to express Channelrhodopsin in direct pathway MSNs or indirect pathways MSNs in dorsomedial striatum. They then record in striatum or Substantia Nigra pars reticularis (SNr) which is thought to subserve much of the functions of the primate GPi due to the small size of GPi in rodents. They find that activation of the direct pathways results in an increase in the velocity of the mice in an open field, and that activation of the indirect pathway results in a decrease in their velocity. In both mice they record from a small number of striatal cells that are activated by the blue light and are putative direct or indirect projection neurons. When recording the response of the SNr cells to the light in striatum, they find both excited and inhibited units in about equal proportions. In the direct pathway expressing mice, they find that the latency of the response of the inhibited units is faster than that of the excited units and that the activity of these units is a better predictor of when locomotor starts will occur pointing to the inhibition as the primary effect and the excitation as the secondary effect perhaps through disinhibition from the nearby inhibited units, interneurons, or network effects. In the indirect pathway expressing mice the latency of the response of the excited units is faster than of the inhibited units, and the activity of the excited units is a better predictor of when immobility starts will happen. These results generally support the classical model of the direct and indirect pathway as inhibiting and disinhibiting SNr, respectively, and thus activating or inhibiting motor programs, but show that a large proportion of SNr neurons behave in an opposing manner.
In many current optogenetic manipulation experiments targeting specific circuits, authors do not record from the relevant circuits, but instead are showing behavioral effects without addressing the complexity of the network effects that their manipulation may be producing. In this case, the authors specifically asked how their manipulation was affecting neuronal activity downstream the basal ganglia circuit and found a large proportion of units behaving in a manner that contradicted basic expectations (which is not surprising, but often not addressed). This also brings up the question of how these units are exerting the behavioral effects that they and others have observed after manipulating direct or indirect pathway neurons. A likely possibility is that speed is the critical factor and the units with the shorter latency responses are driving the behavioral response, consistent with their findings.
The behavior they use (locomotion in an open field) is not a behavior likely to be highly dependent on striatal circuits and velocity may not be good measure for the effects of their manipulations. There is virtually no discussion of that the animals actually were doing in the open field (grooming, sitting, sleeping, ect), so the reader gets a very poor sense of what effect the manipulations were actually having. It would be much more appealing to use a different behavioral task.
(EDIT: They do this in Fig. 5. Great!) Using mouse 'mean instantaneous open-field velocity' as a behavioral readout: not a good fit to the data. It would have been more informative to show the running-notRunning transition probabilities, and the running speeds in just the subset of trials where mice were running the whole time. Is this effect about speed, or likelihood of running?
Still in early days of in-vivo ChR2 labeling, I wish the standards were higher for claiming that evoked spikes and spontaneous spikes indeed come from the same neuron. Their plot in Fig 1 hides the complexity involved in this.
It is unclear why the ~5 mice they used per group along with 16 channel microwire arrays produced only ~27 units from the SNr. When these units are further broken down into groups the numbers are so small that it is hard to understand for example whether the baseline differences between the inhibited and excited units in the SNr are real. Similarly unclear why they only have ~6 responsive units from striatum, this is very low considering they had 2 mice and a 32 microwire array. They don't mention the total number units they recorded and the % of responsive striatal cells they got, nor do they show any examples of brain sections with the infected striatal area and fibers in SNr or pallidum.
"To best approximate physiological patterns of striatal activity, we used a constant illumination paradigm to avoid imposing patterned firing on stimulated neurons." This is a very strange thing to say! Stimuli are 100ms and 1s long - sharp onset, sharp offset, with a fiber-optic shaped activation spot. Patterned firing is certainly being imposed, physiological patterns are not well approximated.