Synapse-specific opioid modulation of thalamo-cortico-striatal circuits

Essential revisions:

The following points must be addressed before final acceptance of the manuscript. Most of these concerns can be addressed with additional analysis, modifications of the figures and extension of the discussion. For your understanding of these essential revisions, the full reviews are appended below.

1) In Figure 3E,F, it is unclear what proportion of MSNs exhibited an EPSC when AAV2-retro-Cre was injected into the AAC and AAV-DiO-ChR2 was injected into the MThal. Further describing the proportion of cells in the DMS exhibited EPSCs would help show the distribution of cells that receive input from mThal cells that also project to the ACC. A scatter plot of the responders/non-responders of MSNs that exhibited EPSCs would be informative.

We agree with the reviewers that the question of how frequently ACC-projecting MThal inputs synapse to the MSNs in the DMS is an interesting one but might not be possible to address with the limitation of existing reagents and methods. This is mainly due to two factors, both of which affect the interpretation of non-responders. (1) The existing retrograde viruses, including rAAV2-retro we used here, have been improved in terms of label efficiency but are far from 100%. Therefore, we cannot virally infect all axons within the ACC that arise from the MThal by our viral combination of rAAV2-retro expressing Cre-recombinase in the ACC and AAV2 expressing DIO-ChR2-EYFP in the MThal. As a result, if an MSN does not exhibit an EPSC upon light stimulation, it does not necessarily mean that it is not receiving ACC-projecting MThal inputs. (2). Whether an EPSC, thus a synaptic connection, can be detected also depends on the expression level of ChR2, which can vary across animals, and is a known caveat of the optogenetic tools. The smaller EPSCs we observed in these experiments (Figure 3F) relative to other experiments (Figure 1C and Figure 2B) is likely due to the low efficiency of the retrograde Cre-recombinase expression in the MThal cells which resulted in much lower expression level of ChR2 in the MThal compared to that expression observed in other experiments.

In summary, given the limitation of the current technology and intersectional nature of this experiment, in our system, the lack of EPSCs does not lead to the conclusion of lack of bona fide synaptic connections. Therefore, we feel that in this case, the scatter plot will not address the question the reviewer is interested in. In the meantime, we now emphasize in the main text that what we can claim is the existence of the ACC-projecting MThal neurons which also project to the MSN in the DMS, and this projection is sensitive to enkephalin (subsection “Thalamostriatal and thalamocortical projections can arise from the same medial thalamic neuronal population”).

2) There is concern about the lack of reversal (due to no application of antagonist) or application of antagonist that does not reverse the opioid inhibition of the synaptic response/firing of the cell. In several experiments neurons are included in the analysis where the opioid inhibition is not reversed by the antagonist (e.g. figure 4C,F; Figure 4—figure supplement 1C,F, Figure 5C). There also appear to be cells that have a data point for the agonist but without a connected data point for the antagonist. For example, Figure 4—figure supplement 1, Figure 4C,F, Figure 5. Of particular concern is the data shown in Figure 4F which is pivotal to the interpretation of the data presented in this study. The effect of DPDPE on the IPSC seems very variable. In some cells the agonist inhibited the response and this was sometimes reversed, sometimes no antagonist response is shown (for largest inhibition) and in some cases the antagonist produced a large overshoot in the size of the synaptic current. There only appears to be one cell where δ receptor activation reduces the size of the IPSC and this is reversed to baseline by the antagonist. How could these changes in synaptic responses be due to activation of the relevant opioid receptor if they were not reversed by the relevant antagonist? If there is a biological basis for it, rather than experimental, it would be interesting to see it discussed.

We believe that the variability in effect of DPDPE on the IPSCs is a mix of experimental limitations in this biological system. First, the IPSCs are generally large and disynaptic or polysynaptic meaning small changes in excitability can have large effects. If the IPSCs are polysynaptic, inhibiting the PV cells would lead to disinhibition of other pyramidal cells which may in turn excite more PV cells as a negative feedback. Because a small number of PV cells densely innervate pyramidal cells, adding or losing the input of a single PV cell can have a large effect on IPSC amplitude particularly when the frequency of optical stimulation is low and the number of trials that are averaged is limited.

We have established standards for recording quality and health of cells. In some experiments, cell health or recording criteria did not hold long enough for a full set of pharmacology experiments, e.g., only agonists were added and no antagonists were added. We still include those cells with the part that all criteria were met. That is why some data points did not have antagonist data (these data points lack a gray dashed line). For this reason, we were very careful with our statistical analyses and utilized steps to specifically deal with the unmatched data points. As stated in subsection “Quantification and Statistical Analysis”, “The Skillings-mack test is a non-parametric test, which allows for missing data points (unbalanced design).”

The lack of reversal in some examples may be due to incomplete reversal as a result of competition between DPDPE and naltrindole. Additionally, desensitization or rundown of the channelrhodopsin currents in the MThal to PV synapses may also contribute to the non-full reversal with antagonists. In combination with the notion above that small changes in the EPSP amplitude evoked in putative PV neurons could bring the PV cells below action potential threshold, a small change in channel rundown can have a major effect in this experimental setup.

In addition, these experiments involve stepping the membrane potential between -60 mV and +5 mV. There was potentially some contamination of the IPSCs with EPSCs as a small inward current is noticeable in the example traces in Figure 4E. Small drift in the actual holding potential or changes in space clamping over the experiment could also result in slight changes of the relative contributions of EPSCs and IPSCs, and thus the peak amplitude of the IPSCs measured.

In Figure 5, multiple possibilities also exist for some non-full reversal effects. We believe that the reversal of DPDPE inhibition by naltrindole is slow, and therefore may not have been completely reversed in a timeframe relevant for these experiments. This could be due to kinetics of competition binding, or the inhibition may be mediated in part by kinase activity which can have effects after G-protein signaling is terminated. In addition, some other forms of plasticity could be induced by DOR activation which would be slow to reverse on the timescale of these experiments.

We do want to experimentally address whether the issue is due to non-specific effects of the pharmacological reagents we used. The rationale is that one potential possibility of the partial irreversibility is because DPDPE may modulate synaptic transmission via DOR-independent pathway, therefore DOR antagonist naltrindole would not able to reverse the DPDPE-mediated modulation of IPSCs. Therefore, we have performed an additional experiment which is to repeat the experiments shown in Figure 4D-F, except that we applied DOR-antagonist naltrindole first, followed by DOR-agonist DPDPE. AAV2-CAG-ChR2(H134R)-TdTomato virus was injected in the MThal, and feed-forward inhibitory inputs to L5 pyramidal neurons the ACC were evaluated (Figure 4—figure supplement 3). We recorded optically-evoked IPSCs (oIPSCs), and first perfused DOR-antagonist naltrindole followed by DOR-agonist DPDPE. Naltrindole pre-treatment prevented the DPDPE-mediated modulation of oIPSC, as observed in Figure 4 and Figure 5, (Figure 4—figure supplement 3), indicating that DPDPE modulates GABAergic neurotransmission via a DOR-dependent pathway.

3) The authors suggest that overall activation of the striatum will determine whether the person experiences the affective aspect of pain (Introduction). Is there evidence that activation of the striatum, regardless of whether this comes from the thalamus or cortex, is aversive? Or do the different inputs to the striatum result in a different experience of the aversive component of pain?

We thank the reviewers for pointing this out. Our original rationale was that since both increased ACC and medial thalamic activities are known to be associated with pain perception (e.g., Casey et al., 1994; Davis et al., 1997; Peyron et al., 1999, Peyron et al., 2000), and these two regions most strongly converge in the DMMS (Hunnicutt et al., 2016), we predicted that decreased activity of these terminals in the DMS would be involved in pain relief, though this has not been directly demonstrated to our knowledge. We have now re-worded this part to reflect that MOR and DOR activation is predicted to modulate pain responses in opposite ways (Introduction, “Thus, MOR and DOR activation are predicted to play opposite roles in pain processing mediated by this thalamo-cortico-striatal circuit”), and we think this is better reflective of our data and current knowledge.

4) It is surprising that, given the substantial circuit differences from D1 and D2 spiny projection neuron activation, that D1 and D2 SPNs were not identified. Although the differential effect of MOR agonist alleviates this concern, it is nonetheless preferable to have an independent measure that each laser activates distinct pathways. When activating the 470 laser and the 625 laser simultaneously, the currents should sum linearly if they are from distinct inputs.

We agree with the reviewers that investigating cell type-specific effects onto D1 and D2 MSNs would be interesting. So far, under our experimental conditions, we have not observed obvious heterogeneous responses, at least not a bi-modal distribution as one might expect if there are major differences in D1 versus D2 responses, of opioid inhibition of MThal inputs to the MSNs, indicating that inputs to both D1 and D2 MSNs might be modulated in a similar fashion (Author response image 1). Previous work comparing the excitatory input kinetics between D1 and D2 MSNs suggest that the risetime of mEPSCs are slower in D2 MSNs (Al-muhtasib et al., 2018). If the DAMGO effect on EPSC amplitude is different between D1 versus D2 MSNs, a correlation is expected between baseline EPSC risetime and DAMGO effect on EPSC amplitude in our data. Under current conditions, no significant correlation between DAMGO effect on EPSC amplitude and baseline EPSC risetime obtained from all MSNs is observed (Author response image 1B). Furthermore, comparison of two putative MSN subpopulations, defined using a cluster analysis among all recorded MSNs based on the baseline EPSC kinetics (Figure 4—figure supplement 3C), yielded non-significantly different distributions of DAMGO effect on EPSC (Figure 4—figure supplement 3D). Together, this suggests that so far, there is no evidence that observed mu-opioid mediated suppression of thalamostriatal inputs is dependent on the identity of the recorded MSN. We think the systematic investigation of cell-type specific response is beyond the scope of current manuscript but an interesting question for future investigation. We have now added a notion that we have not observed any evidence for opioids’ modulation of the thalamic inputs that suggest potential D1 and D2 MSN specific effects (subsection “Opioid modulation of the thalamo-cortico-striatal circuitry”).

Author response image 1

Download asset Open asset

Regarding the dual channelrhodopsin activation experiments, we addressed the potential cross-activation of Chrimson with 470 nm excitation in Figure 1—figure supplement 1D-E by expressing each channelrhodopsin variant separately in individual animals, and stimulated with both the 470 nm and 625 nm excitation. From these data the degree of cross-activation was estimated to be very low (approximately 4% for Chrimson by 470 nm, and < 1% for ChR2 and CsChR by 625 nm). We feel that the cross-activation data together with the data demonstrating a clear inhibition of the 470 nm-evoked currents by DAMGO with no effects on the 625 nm-evoked currents should alleviate the concern of unintended activation with 625 nm light. To further verify our results and conclusions that specific opioid receptor types modulate distinct inputs to the striatum we performed independent control experiments as shown in Figure 1—figure supplement 2. In these experiments individual projection origins, i.e. ACC, PFC, AMThal, and MThal, were injected with only one ChR2 variant, CsChR, and subsequently tested their sensitivity to opioid modulation. We confirmed that DAMGO specifically modulated the thalamostriatal projections, and therefore that these projections represent functionally and pharmacologically distinct inputs compared to the corticostriatal projections. Thus, we feel confident that there was very little cross talk and any potential cross-activation did not change any interpretations of the data.

We did not attempt the additivity analysis suggested due to the fact that the activation of Chrimson by 470 nm light is still not 0 (as mentioned above, 4%; Figure 1—figure supplement 1). While the low intensity of 470 nm stimulation we used may not evoke large EPSCs, we cannot exclude the possibility that 470 nm and 625 nm light together lead to non-linear effects for reasons beyond our current understanding of these tools.

5) Does Figure 6B show action potentials? These must be escaped action currents. Nevertheless, they appear to have half-widths of 3 ms. Using escaped action currents as a measure of firing rate may be affected by the "voltage command" even when not broken in to the whole-cell mode.

Figure 6B shows action currents recorded in cell-attached mode. Voltage clamp mode setting on the amplifier was used and the voltage command was set and adjusted when needed during the experiment, to obtain an amplifier current of 0 pA. This approach ensures that the spike activity of the recorded neuron was not influenced by the cell-attached mode settings, since no current is flowing between the cell and electrode (Perkins, 2006). The firing rate therefore should not be affected by the voltage command. In addition, quantification of the action potential widths (width at half-max action current amplitude, Author response image 2) during baseline and agonist applications (DAMGO and DPDPE) were combined, showing that the mean action potential width is 1.36 ms, consistent with the reported action potential widths in L5 pyramidal neurons in the mouse ACC (1.13 ms, Lee et al., 2007) and in L5 pyramidal neurons in the neocortex (2.08 ms, see neuronelectro.org). In conclusion, the recorded currents resemble action potentials with expected kinetics, which remain stable during the recording and therefore, we do not have evidence that ‘voltage command’ in cell-attached mode affect our results.

Author response image 2

Download asset Open asset

6) Do the specific MThal inputs to ACC synapse on the ACC projections to DMS?

To answer the question, we have performed additional experiments to record from the L5 pyramidal neurons in the ACC that project to the DMS to test whether they receive MThal inputs (Author response image 3). To label the DMS-projecting L5 pyramidal neurons in the ACC (Author response image 3A-B), we injected retrograde beads in the DMS; and to optically stimulate the MThal inputs, we injected AAV2-CAG-ChR2(H134R)-tdTomato in the MThal of the same animal (Author response image 3A). From all recorded DMS-projecting L5 pyramidal neurons, ~91.7% (11 out of 12 neurons) received MThal inputs (Author response image 3C) with a charge transfer and onset value within the range of EPSCs reported in the main text (Author response image 3C-D, see also Figure 4—figure supplement 2C and Figure 7—figure supplement 1A). These results indicate that indeed, MThal inputs to ACC synapse onto the DMS-projecting neurons in the ACC.

Author response image 3

Download asset Open asset

7) Although different opioids may have differing effects, it is difficult to discern what the endogenous effect of opioids may be on the circuit. MOR activation on MThal to ACC pyramidal pathway causes a decrease in EPSC amplitude and a decrease in IPSC amplitude. Meanwhile, DOR activation causes a decrease in IPSC amplitude in ACC pyramidal neurons. What is the net effect on firing?

We agree with the reviewers that the endogenous effect of opioids is an import question and a future direction to take. However, addressing the endogenous effects of opioid receptor activation is currently difficult and is beyond the scope of the current work. The main obstacle for addressing this question is the ability to reliably release and quantify the release of endogenous opioids, e.g., enkephalin or dynorphin in mouse brain slice. We did attempt to trigger release of enkephalin in transgenic mouse lines, which expressed channelrhodopsin in putative enkephalin-releasing cells (PENK-IRES2-Cre;Ai32 mice). Unfortunately, it was difficult to establish the reliable stimulation protocols for enkephalin release due to lack of reliable ways to measure enkephalin release. For the same reason, it is difficult to pharmacologically address the net effects on firing without knowing whether/how the endogenous opioid receptor is activated. These pharmacological agents, e.g. enkephalins, endorphins and dynorphins, have differential affinities for mu, δ and kappa opioid receptors, though the affinity differences can be minimal between mu and δ receptors (e.g. Raynor et al., 1994). Differences in affinity, the sites of action, and the concentration of endogenous agonists will together determine the net effect of firing in the ACC. Based on our findings, our prediction is that at modest concentrations of endogenous opioid receptor ligands with highest affinities for MORs preferably reduce ACC firing, and therefore suppress both direct thalamostriatal and the indirect thalamo-cortico-striatal pathways, whereas those endogenous ligands with highest affinities for DORs, at relatively selective concentrations increase ACC firing. Because IPSCs are sensitive to both mu and δ agonists, low to moderate concentrations of non-selective agonists would be predicted to impact feed-forward inhibition more profoundly and have a net excitatory effect, while higher concentration would greatly diminish EPSCs producing a net inhibitory effect. The future work that measures the endogenous ligand release sites and concentrations will provide key information for answering these questions. Since the direct testing of these predictions is not feasible with our current means, and the related topic might be work of its own in the future, we have not included major discussions but have added a short one to this point (subsection “Opposing effects of opioid subtypes on circuitry modulation”).