Identification of SCN target neurons in the midline hypothalamus and ventral thalamus
To obtain a relatively unbiased picture of how SCN electrophysiological output influences neuronal activity in downstream target nuclei, we first evaluated responses to electrical stimulation of the SCN region. Accordingly, we performed multielectrode (64 channel) recordings from the SPZ, PVN and ventral thalamus (and in a few cases also the SCN) while delivering intermittent current pulses via a concentric stimulating electrode centred between ventromedial portions of the paired SCN (Fig. 1a, Additional file 1: Fig. S1a). This position was chosen to allow us to stimulate the SCN in both hemispheres simultaneously, while minimising the risk of stimulating neurons located in adjacent parts of the hypothalamus. To facilitate subsequent optogenetic circuit mapping, experiments were performed in slices derived from mice where cre-dependent channelrhodopsin 2 was directed to VIP-expressing (VIP+/cre; Ai32+/−, n = 20) or all GABAergic cells (GAD2+/cre; Ai32+/−, n = 40), as well as cre-negative Ai32+/− littermates (n = 13) [18, 19, 29,30,31,32]. Since there were no overt differences in the prevalence or nature of responses to electrical stimulation of the SCN identified across these genotypes, we combined the relevant data for initial analysis.
To confirm that our experimental approach effectively activated SCN neurons, we first analysed data from a subset cells recorded within the SCN itself (n = 23) or nearby regions of the SPZ (n = 59 cells within 600 μm of the stimulation site). Of note, a substantial proportion of SCN neurons (n = 17/23) exhibited robust increases in firing following local electrical stimulation that, in all but 1 cell, persisted following treatment with 20 μm (+)-bicuculline and 50 μm D-AP5/20 μm CNQX (BIC+iGluX) to block fast ionotropic GABA and glutamate receptors (Additional file 2: Fig. S2a-c). By contrast, no cells outside the SCN responded under these conditions. Thus, our electrical stimulation approach provided widespread and selective activation of SCN output, engaging ~ 70% cells within a radius of 325 μm from centre of the stimulation site (Additional file 2: Fig. S2d).
We next then examined in more detail the response of extra-SCN cells to electrical stimulation of the SCN region. From the experiments outlined above, we were able to isolate the activities of 666 individual neurons in total across the SPZ, PVN and ventral thalamus. The proportions of responding neurons varied strongly as a function of distance from the SCN, with relatively high proportions of responding neurons in SPZ and lower proportions at the more dorsal regions we recorded from (Fig. 1a). Consistent with previous investigations of SCN output pathways in the rat [25, 27], among those responding neurons, we found subsets of cells that exhibited reproducible decreases (n = 54) or increases (n = 59) in firing following electrical stimulation of the SCN. Although mouse SCN neurons are considered to be primarily, if not exclusively, GABAergic [8, 22], electrical stimulation of the SCN region has previously been reported to engage glutamatergic output [33]. Accordingly, we probed the neurochemical basis of the responses identified here by bath applying BIC followed by co-application with ionotropic glutamate receptor antagonists (iGlu-X), as above. On the basis of changes in response to SCN electrical stimulation and following antagonist treatments, this allowed us to identify five basic classes of target neurons (Fig. 1a–g).
Among those cells that reduced spike output following electrical stimulation under baseline conditions, in the majority of cases (n = 36/54), BIC treatment completely abolished responses, confirming a GABAergic origin (Fig. 1c, h; termed ‘GABA inhibited’). Interestingly, however, in the remaining subset (n = 18/54), inhibitory responses were replaced by robust increases in firing following BIC treatment which were subsequently abolished under BIC+iGlu-X (Fig. 1d, i). Hence, electrical stimulation drove both GABAergic and glutamatergic input to such cells with ‘mixed’ responses, albeit with inhibitory GABAergic responses dominating under baseline conditions. Both types of inhibitory cells exhibited similar response kinetics (Fig. 1b) and their prevalence decreased as a function of distance from the SCN (Fig. 1a; χ2 test, P < 0.001), in keeping with previously reported variations in the density of SCN output projections [8, 11].
By contrast to the above, few of the cells that exhibited excitatory responses to electrical stimulation under baseline conditions were sensitive to BIC (Fig. 1e, j; n = 12/59 cells). Interestingly, across the modest number of ‘GABA activated’ cells identified here, we found response latencies were significantly slower than for GABA inhibited cells (Fig. 1b, Sidak’s post-test, P = 0.006). This suggests such responses may arise via a multisynaptic mechanism (e.g. due to a decrease in inhibitory input from ‘GABA inhibited’ cells), consistent with our identification of cells with similar properties following optogenetic stimulation of SCN VIP cells [19]. However, in the majority of cases where we observed excitatory responses (n = 47/59), these persisted under BIC but were abolished under BIC+iGlu-X, indicating a glutamatergic origin (Fig. 1f, k; termed ‘Glu. activated’). Of note, BIC treatment also revealed an additional population of cells (n = 13) that exhibited glutamatergic excitatory responses but entirely lacked any overt response to electrical stimulation under baseline conditions (Fig. 1g, l; termed ‘Glu. Disinhibited’). Across these groups of cells with excitatory glutamatergic responses, kinetics were broadly similar to those with inhibitory responses (Fig. 1b), and these cells were more commonly found in the SPZ/PVN than ventral thalamic regions (n = 65/532 vs. 7/134 respectively, Fisher’s exact test, P = 0.02).
To account for the possibility that the prevalence or nature of target cell responses to SCN stimulation varied in a circadian manner; subsets of the experiments described above were performed in slices prepared either at the beginning (ZT1-3) or end (ZT10-12) of the animals light phase (n = 29 and n = 44 slices, respectively). While most of the response types described above were found in similar proportions between these two time-points ‘Glu. activated’ cells were more commonly identified during early day recordings (Additional file 3: Fig. S3a). We did not detect any significant variation in stimulus-evoked responses or baseline firing rates under the various treatment conditions for any cell type (Additional file 3: Fig. S3b-f). We did, however, note a trend towards reduced spontaneous firing rates and response amplitudes (prior to antagonist treatment) for GABA inhibited cells recorded during early day vs. early night (Additional file 3: Fig. S3b). This observation would be consistent with the notion that the daytime increase in spontaneous GABAergic output from the SCN drives circadian variation in the activity of these GABA inhibited cells (investigated in more detail below).
Our finding that electrical stimulation evokes GABAergic, primarily inhibitory, responses in a subset of neurons across the SPZ, PVN and ventral thalamus is in keeping with known SCN neurochemistry [8, 10, 22]; however, the presence of many cells exhibiting glutamatergic responses is surprising. Indeed, while there is certainly prior evidence that stimulation of the SCN region can evoke glutamatergic responses [25, 27, 33], clear anatomical evidence for a population of glutamatergic cells in the mouse SCN is lacking [10]. Alternatively then, glutamatergic responses could instead originate with stimulation of glutamatergic neurons found lateral to the SCN [10] or via retinal afferents that pass through the nucleus [34,35,36].
To evaluate the latter possibility, we next performed a similar set of recordings to those described above but with the stimulating electrode placed on one of the optic nerves (Fig. 2a, Additional file 1: Fig. S1b). Across 9 slices (prepared during early day), we identified 10 cells that exhibited excitatory responses and 3 cells that displayed stimulus-driven decreases in firing in response to optic nerve stimulation. These responses were primarily observed at recording sites contralateral to the stimulated optic nerve (8/10 excitatory and 2/3 inhibitory), in line with prior anatomical observations of retinofugal projections [34,35,36]. Excitatory responses were reliably blocked by iGlu-X treatment (both alone and in combination with BIC) but were not blocked by BIC alone (Fig. 2b, c). Moreover, the proportion of cells displaying these glutamatergic activations was not statistically different to that observed following SCN stimulation (including both Glu. activated and disinhibited subtypes) in slices prepared at the same time of day (Fig. 2a; Fisher’s exact test, P = 0.07). By contrast, the proportion of cells exhibiting inhibitory responses following optic nerve stimulation was significantly less than we encountered following SCN stimulation (Fig. 2a, Fisher’s exact test, P = 0.02), nor did we observe any cells exhibiting GABAergic excitations. For one of the three neurons that was inhibited following optic nerve stimulation, the decrease in firing was blocked by iGlu-X treatment alone, suggesting that response was secondary to glutamatergic stimulation of the SCN (Additional file 4: Fig. S4a,c). The same was never observed for cells showing inhibitory responses following SCN stimulation (Additional file 4: Fig. S4b,d; n = 7 GABA inhibited and n = 4 mixed tested). It remains unclear whether responses of the remaining two neurons exhibiting GABA-mediated inhibitions following optic nerve stimulation reflected some minimal direct activation of the SCN or, perhaps, a role for the recently discovered subset of GABAergic intrinsically photosensitive retinal ganglion cells (ipRGCs) [37]. In either case, since optic nerve stimulation readily evokes excitatory glutamatergic responses in downstream neurons, we conclude that the occurrence of such responses following SCN stimulation at least primarily reflects stimulation of the retinal afferents that pass through the SCN on route to those regions [34,35,36].
We previously reported that optogenetic stimulation of SCN VIP cells drives inhibitory (GABA-mediated) responses in a subset of neurons across the SPZ, PVN and ventral thalamus [19]. To determine the extent to which the inhibitory responses observed here reflect activation of GABAergic output from VIP neurons, in a subset of experiments (n = 20 VIP+/cre;Ai32+/− slices), we combined SCN electrical stimulation with optogenetic VIP-cell stimulation (Fig. 3a). As reported previously, wide-field blue light flashes applied over the SCN drove robust inhibitory responses in a subset of neurons in downstream target regions (Fig. 3a, n = 8) that were abolished under BIC treatment (Fig. 3b, c). This population of VIP-innervated (VIPIn) cells only partially overlapped with those exhibiting inhibitory responses to electrical stimulation. Indeed, only 2 of 12 cells identified in these experiments with inhibitory responses and 0 of 6 cells with mixed responses to electrical stimulation also showed inhibitory responses to optogenetic stimulation (Fig. 3a). The optogenetic approach employed here robustly activates VIP cells throughout the SCN [19]; therefore, a substantial proportion of the SCN output revealed by electrical stimulation must come from non-VIP cells. Interestingly, however, we also found a number of VIPIn cells (n = 6/8) that did not display any overt responses to SCN electrical stimulation. Thus, the electrical stimulation approach employed here also, to some extent, underestimates the true proportion of downstream target neurons receiving SCN input.
To probe this possibility further, we next performed similar experiments in slices from animals where ChR2 expression is directed to GAD2-expressing cells, thereby targeting all GABAergic neurons throughout the SCN (GAD2+/cre; Ai32+/−). In pilot experiments, we found that wide-field optogenetic stimulation (delivered as above for VIP cells) resulted in direct activation of many neurons outside the SCN. Accordingly, to selectively target just SCN GABAergic neurons, we delivered local illumination within one SCN hemisphere via a finer (105 μm core) fibre attached to a penetrating recording electrode while recording from target neurons of the same hemisphere via a second electrode (Additional file 5: Fig. S5a; n = 16 slices). As expected, the majority of SCN neurons detected in these experiments (n = 18/21) exhibited robust excitatory responses to optogenetic stimulation that were unaffected by treatment with ionotropic glutamate and GABA receptor blockers (Fig. S5b,c). We also found a small subset of cells in downstream target regions (n = 5/115) that exhibited inhibitory, GABA-mediated responses, to optogenetic stimulation (Additional file 5: Fig. S5b,c). By comparison with selective stimulation of SCN VIP cells, a nominally greater proportion of these ‘GABAIn’ cells also exhibited inhibitory responses to electrical stimulation of the SCN (n = 3/5). However, this approach also failed to evoke detectable responses in a subset of neurons that exhibited GABA-mediated responses to more conventional stimulation (n = 5/8 cells), in keeping with the more local nature of the optogenetic stimulus.
Since electrical stimulation drove more widespread SCN output than targeted optogenetic stimulation, in subsequent experiments, we instead employed optogenetic approaches in GAD2+/cre; Ai32+/− slices for identification of the neurochemical phenotype of SCN-target cells rather than driving downstream responses. Accordingly, in a final subset of recordings, we used wide field optogenetic stimulation of the SPZ, PVN and ventral thalamus alongside electrical stimulation of the SCN region to define which of those responding neurons were themselves GABAergic (Fig. 4a; n = 24 GAD2+/cre; Ai32+/− slices). Consistent with the known neuroanatomy of the target regions [10], the majority of cells recorded from the SPZ and ventral portions of the PVN (~ 80%; Fig. 4a) exhibited robust activations following optogenetic stimulation that persisted in the presence of BIC+iGlu-X, indicating they were GABAergic (GAD2-expresing). By contrast, cells that lacked excitatory responses to optogenetic stimulation were less commonly encountered in recordings from more dorsal sites (Fig. 4a; PVN and, especially, the ventral thalamus) consistent with the greater proportion of glutamatergic neurons found in those regions [10]. Most notably, however, among cells that responded to electrical stimulation of the SCN region, all but one, also exhibited robust direct optogenetic activation (n = 8/9 GABA inhibited, Fig. 4b; n = 4/4 Mixed, Fig. 4c; n = 6/6 GABA activated, Fig. 4d; n = 3/3 Glu. Activated, Fig. 4e). Indeed, the proportions of neurons exhibiting responses to SCN electrical stimulation were significantly greater among optogenetically responsive vs. non-responsive cells (Fig. 4a; χ2 test P = 0.03). Thus, SCN (and retinal) output to the recorded regions almost exclusively targets GABAergic neurons.
Circadian rhythmicity in hypothalamic and ventral thalamic targets of SCN and retinal input
Having identified neurons across the SPZ, PVN and ventral thalamus receiving input from the SCN and/or retina, we next investigated in more detail whether such influences were associated with specific circadian patterns in spontaneous neural activity in the recipient neurons. To obtain stable long term (> 26 h) recordings from individual target neurons, we performed perforated multi-electrode array (pMEA) recording in VIP+/cre; Ai32+/− slices (n = 15 prepared during early day and n = 18 during late day). We then combined both electrical and optogenetic stimulation of the SCN region to distinguish VIPIn cells from those cells that received inhibitory input from other classes of SCN neurons (Fig. 5a).
Consistent with our experiments using penetrating electrodes, electrical stimulation of the SCN region reliably identified subsets of target neurons displaying inhibitory (n = 54/740) and excitatory responses (n = 34/740). So as not to disrupt ongoing rhythmic activity during these long-term recordings, we did not apply antagonist treatments to determine the neurochemical basis of such responses here. Nonetheless, based on the data presented above, the former group (presumably including both GABA inhibited and Mixed subtypes) reflects those cells receiving strong GABAergic input from the SCN, while the latter group will primarily comprise those receiving excitatory glutamatergic input from the retina. We suspect this includes ‘Glu. Disinhibited’ cells since these longitudinal recordings revealed that synaptically driven responses were not always evident for the full duration of our recordings, despite continued spontaneous firing (Additional file 6: Fig. S6a). Across the populations of cells that were activated or inhibited by SCN electrical stimulation, response amplitudes did not vary in a consistent manner as a function of time of day or time since start of recording, however (Additional file 6: Fig. S6b).
We next examined the spontaneous firing profiles of neurons responsive to electrical stimulation of the SCN region (Fig. 5b,c). The majority of both inhibited (n = 45/54) and activated neurons (n = 25/35) displayed evidence of circadian variation in firing (see methods), as did cells that lacked responses to electrical stimulation (n = 566/651; Fig. 5d, χ2 test, P = 0.75). Further, among rhythmic cells in each class, the duration of the high firing epoch, 24 h mean and peak-trough amplitude of the firing rate rhythm were all statistically equivalent (Fig. 5e). Importantly, however, while the basic circadian waveforms of neurons in each class were virtually identical (Fig. 5f), the timing of peak firing showed substantial variation across the groups (Fig. 5g).
Among cells exhibiting inhibitory responses to electrical stimulation there was a striking absence of peak firing across the mid-late projected day (Fig. 5g; χ2 test, P = 0.01). The timing of peak firing was not overtly associated with anatomical location (Additional file 6: Fig. S6c) nor with time since start of recording (Additional file 6: Fig. S6e-f; χ2 test, P = 0.65). Thus, cells that receive inhibitory input from the SCN exhibit a broad but non-random distribution of phases, with peak firing during the night or early projected day. Analysis of the phase distributions of cells across slice preparations produced equivalent results, highlighting a significant variation when the phase of peak firing was expressed relative to prior LD cycle but not time since start of recording (Additional file 4: Fig. S6d,g).
By contrast, neither cells that were activated by electrical stimulation of the SCN region nor those that were unresponsive exhibited this property. Instead, the phases of peak firing for the latter two groups were almost evenly distributed across the day and night (Fig. 5f; Additional file 6: Fig. S6d). Further analysis indicated that for unresponsive (but not activated cells), the phasing of cellular rhythms exhibited a weak but significant clustering relative to time since start of recording (Additional file 6: Fig. S6e-g). Thus, it appears that that the acute slice recording procedure either induces or re-sets neuronal rhythms in a subset of neurons across the PVN, SPZ and ventral thalamus that lack input from the SCN or retina.
We next employed optogenetic stimulation to discriminate which of the recorded neurons received input from SCN VIP cells. As expected, a subset of neurons exhibited inhibitory responses to optogenetic stimulation (n = 18 VIPIn cells) that partially overlapped with those inhibited following electrical stimulation (Fig. 5a, Fig. S6a). However, the majority of cells that were inhibited by SCN electrical stimulation were non-VIPIn (n = 43/54). Nevertheless, in line with our characterisation of inhibited (primarily non-VIPIn) cell populations above and previous analyses of VIPIn cells [19], VIPIn and non-VIPIn electrically inhibited cells were very similar (Fig. 5h–l). Hence, both groups exhibited similar basic circadian properties (Fig. 5h–j) and exhibited a broad but non-random distribution of phases (Fig. 5k, Additional file 6: Fig. S6d-g). We did, however, find that peak-trough amplitudes were significantly higher among VIPIn cells (Fig. 5i), consistent with our previous data indicating that SCN VIP cells exhibit particularly robust circadian firing rate rhythms [19].
Reliance of hypothalamic and thalamic neuronal rhythms on SCN output
Our data indicate that inhibitory SCN output influences neuronal activity rhythms in a subset of neurons across the SPZ, PVN and ventral thalamus. However, we also identify many other neurons in those target regions that exhibit circadian modulations in firing, but lack observable responses to electrical/optogenetic stimulation of the SCN. To better understand the potential reliance of these oscillations on rhythmic output from the SCN, we next prepared slices equivalent to those used above but with the SCN region removed by scalpel cut prior to pMEA recording (Fig. 6a). We then compared the resulting data to that from equivalent recordings in slices containing the SCN (i.e. cells contributing to Fig. 5 regardless of response to stimulation).
We were still able to identify many neurons across the SPZ, PVN and ventral thalamus (n = 51/78, 58/84 and 75/120 respectively) that displayed evidence of circadian rhythmicity in the complete absence of the SCN (Fig. 6c–e). However, the proportion of neurons that passed our criteria for rhythmicity was significantly reduced in each region, relative to slice recordings with the SCN intact (Fig. 6a; χ2 tests, all P < 0.001). As expected from our analysis of cells that lack detectable SCN input (Fig. 5g and Additional file 6: S6d-g), across populations of cells from each anatomical region, the phase of peak firing was similar in SCN-removed and intact slices, being more closely related to time since the start of recording rather than projected time of day, particularly for SPZ cells (Additional file 7: Fig. S7a-c). We did, however, observe subtle changes in the nature of the rhythmic activity in the absence of the SCN. Specifically, while overall mean firing rates and rhythm amplitude were similar to those observed in SCN containing slices (Additional file 7: Fig. S7d-e), the circadian waveforms of neurons were broader in slices that lacked the SCN (Fig. 6b, f–h). This implies a further role for SCN-derived signals in shaping the circadian waveforms of at least some of those neurons that do not absolutely rely on the central clock to rhythms in spontaneous firing.
Collectively then, these data identify distinct classes of oscillatory neurons across the major SCN target regions in the SPZ, PVN and ventral thalamus. A subset of neurons rely on intact SCN projections to sustain circadian variation in firing ex vivo. By contrast, a more numerous group of cells express at least transient neuronal rhythmicity which does not require input from the SCN, and whose phase is seemingly unrelated to prior light history, suggesting rhythms in of at least some of this this latter group are induced or reset ex vivo.