Propagation of Beta/Gamma Rhythms in 1 the Cortico-Basal Ganglia Circuits of the 2 Parkinsonian Rat

21 Some motor impairments associated with Parkinson’s disease are thought to arise from pathological 22 activity in the neuronal networks formed by the basal ganglia (BG) and motor cortex. To evaluate 23 several hypotheses proposed to explain the emergence of pathological oscillations in Parkinsonism, 24 we investigated changes to the directed connectivity in these networks following dopamine depletion.

. CC-BY-NC-ND 4.0 International license under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available The copyright holder for this preprint (which was this version posted September 3, 2017. ; https://doi.org/10.1101/180455 doi: bioRxiv preprint Non-parametric Directionality (NPD) 1 We next investigated directed connectivity between recorded regions. The results of the analysis using 2 the NPD measure are presented in figure 5. The iCOH and the sum of the non-instantaneous parts 3 (forward and backward) of the NPD are similar, and both methods reveal similar patterns of 4 connectivity (data not shown). Analysis of the instantaneous NPD in isolation demonstrated the 5 existence of high amplitude, wide-band interactions that were similar to those found with magnitude 6 squared coherence (data not shown), and are likely due to zero-phase field spread of activity between 7 recordings. Analyses of directional interactions of the LFPs and ECoG hereon will use the forward 8 and backward components of the NPD to discern directional connectivity between structures. 9 We observed that directional interactions of low beta-band activity are predominant in the direction Hz, P=0.018) which would suggest that STN feedback to M2 is strengthened in the dopamine-1 4 depleted state. In the case of the STN/GPe circuit, and unlike iCOH, the non-instantaneous 1 5 components of NPD do not show 6-OHDA related increases in beta coupling in either direction for 1 6 the lesioned rats. Rather, NPD suggests a directional asymmetry in activity in the high beta/gamma 1 7 band with forward connections from GPe → STN connection far stronger than in the reverse direction 1 8 (cluster statistics testing differences between forward and backward spectra in the 6-OHDA 1 9 recordings: 4-43 Hz, P<0.001). Notably we see a feedback in the STN → STR that is only present in 2 0 the lesion condition, a feature that will be relevant with respect to results later in this article.

1
The pattern of activity in the high beta/gamma range between cortical and subcortical regions 2 2 appeared to be principally cortically leading with interactions in the 20-40 Hz range being most back to M2, are weaker than the forward connections, but are still present for striatal and pallidal 2 7 feedbacks to M2 (first column, row 2 and 4, figure 5). Again, there is a clear peak in the high beta 2 8 NPD from STN to STR in the lesioned animals although a dependence on dopamine was not seen to 2 9 be significant when cluster statistics were applied. Finding of NPD from STN to STR does not accord 3 0 the canonical circuit ( Figure 1) but may instead imply feedback to striatum via subcortical thalamo-. CC-BY-NC-ND 4.0 International license under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available The copyright holder for this preprint (which was this version posted September 3, 2017. ; https://doi.org/10.1101/180455 doi: bioRxiv preprint employed cluster statistics to determine significant differences between the non-conditioned NPD 1 spectra (presented in the previous section) and the conditioned variant shown in this section of the 2 results.

3
Conditioning the NPD using Local Field Potentials Recorded from the STN 4 We first conducted a partialization of the NPD estimate using LFPs recorded from within the STN 5 (figure 6). Conditioning with signals from the STN does not remove beta connectivity between the  P<0.001) or GPe (14-21 Hz, P<0.001). We also found in the dopamine-depleted state that there was  Notably we observed some separation in the effects of the conditioning between the control and lesion conditioning a tighter band corresponding to low beta. Lesioned animals only showed reductions at 2 2 higher frequencies (~24-45 Hz, high beta/low gamma) and only between GPe and STR. We observed 2 3 that conditioning of the NPD with the STN signal acted to significantly reduce interactions between 2 4 STR and GPe in both the forward (23-49 Hz, P<0.001) and reverse (27-49 Hz, P=0.001) directions.

5
Conditioning the NPD using Local Field Potentials Recorded from the GPe 2 6 Next, we performed the NPD analysis of recorded signals but this time conditioning the interactions 2 7 with LFPs recorded from within the GPe (figure 7). We found that the conditioning had the effect of 2 8 reducing NPD estimates in 6 out of 6 possible connections in the controls and 3 out of 6 in the 6-Hz, P=0.022). Furthermore, we found that activity in this feedback connection was significantly 1 increased in the controls when compared to the 6-OHDA lesioned animals (35-41 Hz, P=0.002).

Conditioning NPD Using Field Potentials Recorded from M2
3 The final analyses utilized ECoG signals recorded from the M2 to condition the NPD estimates 4 (results in figure 9). We found that the NPD estimates conditioned on M2 are greatly flattened and 1 1 first step towards verifying the plausibility of this hypothesis involves determining whether there is 1 2 indeed functional connectivity between STN and M2 in the beta band, and whether this occurs  has demonstrated that there is a beta peak in the directed coherence in the low beta range in the state, any residual beta band activity is routed via STR, whilst the hyper-direct pathway is relatively 2 6 quiescent. These findings support the idea that the dopamine-depleted state is associated with a 2 7 strengthening of the hyper-direct pathway and its subsequent feedback from STN to cortex.

8
We also provide evidence of multiple coexisting pathways for propagation of beta oscillations.

9
Notably, it was found that conditioning of the NPD with LFPs recorded from the STN (figure 6) does 3 0 not act to remove the 6-OHDA lesion associated beta NPD in the structures 'upstream' of the STN 3 1 (i.e. the STR and GPe). NPD in the low beta range is significant in both the forward and backward  resonance. This is summarised as follows: i) Comparison of forward and backward NPD for hypothesis that the STN/GPe circuit acts as a pacemaker that is in turn tuned by abnormal striatal 2 6 drive, but by some nonlinear transformation that is not discernible using the conditioned NPD. Taken 2 7 together we argue these findings give evidence against pathological beta rhythms arising from the 2 8 pacemaker-like activity of the STN/GPe sub-circuit. It has been proposed that aberrant cortico-striatal activity is involved in the emergence of pathological With conditioning, we find that the beta peak in the STR → GPe connection is removed in its entirety 2 when conditioning the NPD with the cortical signal (figure 9) in both the control and lesion 3 recordings (interestingly however, the STR → STN beta NPD does remain) suggesting that beta band 4 correlations are at least in part explained by cortical activity. Importantly, we found that conditioning 5 the STR → STN connection with the LFP recorded from GPe (figure 7) acts to significantly reduce 6 the NPD in both control and lesion experiments. This would suggest that beta flow is mediated via 7 GPe in agreement with the canonical view of the indirect pathway demonstrating the ability of partial 8 NPD to discern hierarchical connectivity. Further to this, the finding that upstream GPe/STR beta 9 NPD is susceptible to conditioning with LFPs from STN (figure 6) in dopamine-intact animals may 1 0 reinforce the idea that in the healthy network, routing of beta occurs predominantly through the 1 1 indirect pathway, whilst in the pathological case we see beta that is persistent and with activity that is 1 2 at least in part independent of that recorded at the STN. The presence of high beta/gamma oscillations in the subcortical network has been noted by a number that previously reported in anaesthetised rats (Magill et al. 2004;Sharott et al. 2005bSharott et al. , 2009). Activity 2 0 in the awake rats, during movement has also been reported, albeit at slightly higher frequencies  Results from analyses which used iCOH to investigate non-zero lag correlations between BG 2 5 structures and the cortex suggested that gamma interactions are routed in a way that bypasses STR as 2 6 a gamma peak is absent in the M2 (figure 4). The hyper-direct pathway is the other principal source of 2 7 cortical input to the BG, therefore the marked weakness of gamma interaction in the M2/STR when 2 8 compared to the M2/STN iCOH spectra may imply that the hyper-direct pathway is responsible for 2 9 gamma input to the network.

0
However, whilst there is a large peak in the high-beta/low-gamma band NPD for the M2 → STN lesion animals. This in combination with the evidence provided for hypothesis 5 suggests that high 1 beta/gamma originates at either STR or GPe and then propagates to downstream structures. Backward 2 gamma interactions from GPe to STR are apparent in the NPD conditioned on M2 or STN, suggesting 3 the STR signal is the result of local propagation of a gamma signal from GPe. From the canonical 4 view of the circuit it is not clear how gamma passes upstream from GPe. However, a substantial 5 proportion of GPe neurons that innervate the striatum have been shown to exist, with one GPe cell 6 type (arkypallidal neurons) projecting exclusively to striatum (Mallet et al., 2012;Abdi et al., 2015; . CC-BY-NC-ND 4.0 International license under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available The copyright holder for this preprint (which was this version posted September 3, 2017. ; https://doi.org/10.1101/180455 doi: bioRxiv preprint and intra-telencephalic projections, respectively) and so are likely to show a degree of independence 1 from one another. The data presented here also suggest a second pathway upstream of STN involving 2 the STR that is most evident in the recordings from control rats. We suggest that both pathways 3 contain signals shared by activity measured in the cortical ECoG as conditioning of the NPD acts to 4 remove beta peaks from the majority of connections that were analysed, leaving just beta coherence at Taken together, the analyses presented here speak to the existence of a high beta/low gamma rhythm have demonstrated an increase in activity in this band during movement suggesting that activity at 3 0 these frequencies in the M2 and SNr is prokinetic (Brazhnik et al. 2012). Our data would suggest that 3 1 high beta/gamma activity in the normal network is predominantly driven by the cortex as evidenced STN. However, studies by Zold and colleagues has demonstrated that oscillatory activity >20 Hz in 1 corticostriatal afferents is not effectively transmitted (Zold et al. 2012).
2 Furthermore, following partialization some interactions involving STN do remain. In particular we 3 provide evidence for a significant strengthening of feedback from STN to STR in the lesioned animals 4 in the high beta/gamma band. We speculate that this signal is facilitated through the strengthening of The use of partial coherence for inferring neural connectivity is not in itself a novel approach completely capture the activity going through the proposed pathway. This however is unlikely to be 1 8 completely the case due to the finite sampling of the structures afforded from the use of electrodes.

9
However, the large number of channels for recording ensure that multiple samples are obtained from 2 0 each structure. In the data presented here, subcortical structures were recorded from between 2-8 2 1 different channels which were all used to conditioning the estimate of directed coherence. It must also 2 2 be noted that this limitation is likely to apply most to the larger structures that were analysed, namely 2 3 the motor cortex and striatum, whereas recordings from the smaller sized STN is more likely to 2 4 capture the majority of activity. This must be considered when interpreting conditioning of the NPD 2 5 with respect to STR signals. It could be the case that M1 → STN connectivity remains as a result of 2 6 incomplete sampling of striatal signals.

7
Inference of Connectivity from Non-Spiking Brain Activity 2 8 This study focuses upon conclusions that are drawn from mesoscale recordings of brain activity as 2 9 measured either in the ECoG or LFPs. Transmission of information in the brain is facilitated by 3 0 axonal propagation of action potentials that are not explicitly captured in these signals. LFPs and firing by neurons in these structures (Mallet et al. 2008a(Mallet et al. , 2008b. Furthermore, we provide evidence 1 for the existence of temporally lagged correlations between rhythmic local field potentials recorded 2 between distinct regions of the cortico-BG network that imply causation from one signal to another, a 3 phenomenon that would itself not be possible without the transmission of action potentials. Future 4 work will require the investigation to determine whether directional interactions are ascertainable