We next investigated whether the optogenetically enhanced sound-evoked activity of a small group of cells would cause activity changes in other non-stimulated cells. During holographic optogenetic stimulation of the targeted cells, the non-target, but sound-responsive cells (n=995 cells for 16 kHz target ensemble condition and n=675 cells for 54 kHz target ensemble condition), also changed their activity, showing either increased or decreased response amplitudes (Figure 2A bottom and C bottom).
If cortical networks rebalance their activity, we speculated that the increased tone-evoked activity in the targeted ensemble would lead to a decrease in tone-evoked activity in coupled ensembles. Such rebalancing would keep the activity within the cortical network stable. Moreover, given that we increased the activity to the preferred sound frequency, if this rebalancing happens, it should occur only for the distinct sound frequency related to the cell’s tuning property. For example, stimulation of a 16 kHz ensemble should cause a greater reduction in the 16 kHz tone response of non-targeted 16 kHz cells compared to their response to the 54 kHz tone.
To address these questions, we investigated whether increased activity in the targeted cells influenced the activity of non-target cells and how these changes were related to the tuning properties of the cells. We first confirmed that the overall population activity from sound-responsive cells, including both target and non-target cells, did not differ across conditions (control, 16 kHz, or 54 kHz target ensemble conditions; all P>0.05, Figure 3A). This suggests that non-target cells may adjust their activities during the target cell stimulation to maintain the global network activity level. To identify the activity changes based on functional characteristics of cells, we defined each sound-responsive cell’s frequency selectivity by computing a difference between response amplitude to 16 and 54 kHz from the baseline session (frequencypreference=(ΔF/F(16kHz))−(ΔF/F(54kHz))). We then divided these cells into either 16 kHz preferring or 54 kHz preferring groups, taking 0 (i.e., no preference) as a criterion (Figure 3B). Both subgroups exhibited stronger tone-evoked responses to their preferred frequency, independent of the condition (t(5700)=4.79, p<0.0001; Figure 3—figure supplement 1). This confirms that the criterion for cell group threshold is valid.
Non-target co-tuned cells show more decreased response amplitudes due to stimulation when synchronized with their preferred tones.
(A) Stimulation effect (ΔF/Fstim – Δ F/Fbaseline) in all sound-responsive cells, including both target and non-target cells, responding to either 16 kHz (blue) or 54 kHz (orange) pure tones, representing global activity changes due to the stimulation. No significant differences between stimulation conditions and responses to different frequencies were observed (all p>0.05). (B) Sub-categorization of cells based on the frequency selectivity for each target stimulation condition (left: 16 kHz stim, right: 54 kHz stim). Cells were first grouped into either 16 kHz preferring cells (blue) or 54 kHz preferring cells (orange). Within each cell group, cells were further subdivided into low-, mid-, and high-frequency selectivity categories based on their 33% quartile ranges. For visualization, frequency preference was log-transformed; original frequency selectivity distributions are shown in the upper insets. Vertical dashed lines indicate 33% quartile ranges. (C) Stimulation effect (ΔF/Fstim – Δ F/Fbaseline) in 16 kHz (blue) and 54 kHz (orange) preferring cells. Both cell groups show decreased amplitude to their preferred frequency regardless of conditions due to acoustic stimulus-specific adaptation. Only co-tuned cells (16 kHz preferring cells for 16 kHz stimulation or 54 kHz preferring cells for 54 kHz stimulation) show a further decrease in response amplitudes due to the stimulation, when the preferred pure tone (PT) frequency was synchronized. Error bars indicate SEM across FOVs (*p<0.0001). (D) Stimulation effect from the model prediction. Amplitude changes computed from simulated data by applying cell suppression to all cells (All supp.), random cells (Random supp.), or only co-tuned cells (Co-tuned supp.) were compared with real data. Only the Co-tuned supp. model showed a significant amplitude decrease for co-tuned neurons compared to non-co-tuned neurons, similar to the result from the real data (p<0.05; see text for more detail). (E) Post-stimulation effect (ΔF/Fpost – Δ F/Fstim) 16 kHz (blue) and 54 kHz (orange) preferring cells. No significant response amplitude changes were observed. Error bars indicate SEM across FOVs. (F) Response amplitude change based on the frequency selectivity category for each cell group (blue: 16 kHz preferring cells, orange: 54 kHz preferring cells). Significant response amplitude changes relative to the control condition were observed only for high-frequency selectivity category when target stimulated cells were co-tuned (*p<0.05).
We then focused on our main question by comparing the stimulation effect of the two target ensemble groups to the control condition to identify whether stimulation decreased the response of non-target co-tuned neurons. Neural activity in AC rapidly shows stimulus-specific adaptation to the repeated presentation of the stimulus (Ulanovsky et al., 2004; Yarden and Nelken, 2017; Malmierca et al., 2014), which can obscure stimulation-related changes. We thus used the response amplitude change between the baseline and the ‘stimulation’ control session as a representative threshold to test the effect of the stimulation. We once again used the difference in response amplitude between the baseline and stimulation sessions as the measure of the stimulation effect (ΔF/F(stimulationsession)−ΔF/F(baselinesession)). Neighboring cells within 20 μm from the target stimulation point were removed from the analysis since they could have been directly affected by the stimulation.
We compared the stimulation effect between non-target co-tuned and non-co-tuned cells across conditions (16 and 54 kHz target ensembles as well as control conditions) for different pure tone presentations. Since our primary interest was how non-target cells respond to increased activity in target ensembles, we focused on conditions where the pure tone frequency matched or did not match the tuning properties of the non-target cells. Since we stimulated during tone presentation the effects of the holographic stimulation and stimulus-specific adaptation co-occurred. To isolate these components, we used a linear mixed-effect model with cell group, condition, and pure tone frequency as fixed factors, and FOVs as a random factor. We then performed analysis of variance (ANOVA) on the model to assess the main effects and interactions.
A marginally significant main effect of the condition (F(2,37.1)=2.983, p=0.0628) on the response change in the stimulation session relative to the baseline session (i.e., stimulation effect) was observed, indicating that these changes may depend on the stimulation condition. We further analyzed the data to better understand how the different factors interacted in the response amplitude changes. A significant interaction between the pure tone frequency and cell group (F(1,4397.6)=186.967, p<0.0001) suggests that each cell group responded differently to the two pure tone frequencies. Specifically, the response amplitude decreases in the stimulation session relative to the baseline session were more pronounced for each cell group when the played pure tone matched to their tuning property. This interaction between pure tone frequency and cell group highlights the importance of frequency tuning in modulating response amplitudes. Such response amplitude decreases of non-target cells to their preferred pure tone presentation further align with the stimulus-specific adaptation due to repeatedly presented pure tones (Ulanovsky et al., 2004). Additionally, a significant three-way interaction across condition, cell groups, and pure tone frequency (F(2,4397.6)=3.517, p=0.0298) suggests the combined effects of the stimulation condition and the cell group on response amplitude depend on the pure tone frequency. The stimulation effect is not uniform across cell groups and depends heavily on the frequency, highlighting a complex interplay between the tuning property of cells, stimulation condition, and presented pure tone frequency.
Consequently, we analyzed post hoc comparisons estimated marginal means with contrasts, as our focus was how co-tuned cells change their responses due to the increased activity in the target cells along with the frequency of the presented pure tone.
For the 16 kHz preferring cell group (n=537), we observed a greater stimulation effect (i.e., decrease in response amplitude) for 16 kHz tone presentation when the 16 kHz target ensemble was stimulated compared to the control condition (t(124) = 3.114, p=0.0064). For all other pairs, no significant stimulation effect was observed. This suggests that non-target 16 kHz co-tuned cells reduce their response amplitudes when target ensembles share the same tuning property. Furthermore, such response change occurs only when they process their preferred frequency (Figure 3C, left).
We repeated the experiments and the analysis with 54 kHz cells as the target group. In general, we observed similar results. The stimulation effect was significantly more pronounced for 54 kHz tone presentation when 54 kHz target ensemble (n=359) was stimulated compared to the control condition (t(168) = 3.074, p=0.0069; all p-values were adjusted for multiple comparisons using the Tukey method). All other pairs did not show any stimulation effect (Figure 3C, right).
To further explore whether the stimulation effect could be explained by activity rebalancing within the co-tuned network, we implemented a simple model in which a suppression term was applied either to all non-target cells, randomly selected non-target cells, or specifically to non-target co-tuned cells. By comparing three different model outcomes and the real data, we observed a significant effect of the model type (F(3, 3343)=56.243, p<0.0001). Moreover, an interaction between the model type and cell groups was observed (F(3, 3343)=49.635, p<0.0001). Applying suppression to only non-target co-tuned cells during the stimulation session yielded a significant response amplitude decrease for co-tuned cells compared to non-co-tuned cells (F(1, 3343)=48.68, p<0.0001), which resembles the real data. In contrast, applying suppression to all non-target cells and random non-target cells led to similar amplitude changes in both co-tuned and non-co-tuned neurons (F(1, 3343)=0.01, p=0.925 for all suppression; F(1, 3343)=0.05, p=1), which was not observed in either the real data or the simulated data restricted to co-tuned cell suppression (Figure 3D). These results suggest that the target cell stimulation induces a selective activity suppression within the co-tuned network for processing their preferred frequency.
Together, these results indicate that the effect of holographic optogenetic stimulation depends not on the specific tuning of cells, but on the co-tuning between stimulated and non-stimulated neurons. Also, this effect is not driven solely by a few non-target cells with large response changes. Rather, the overall population of cells shows relative response changes due to the stimulation when synchronized with their preferred frequency.
Overall, these results further suggest that when neural activity is increased in a subset of target cells due to photostimulation in addition to the target sound presentation, other co-tuned cells selectively reduce their tone-evoked responses to their preferred tone presentation, indicating that the network rebalances to maintain network activity within a certain range.