Development of the Sectorial Arrangement of TRN
The TRN is comprised of nonoverlapping, modality-specific sectors that are defined by the source of their ascending thalamocortical (TC) and descending corticothalamic (CT) collateral inputs [3, 43, 49–51]. While TC and CT projections pass through TRN at perinatal ages, [52–57], it remains unclear whether they innervate the TRN in a segregated, modality-specific manner. To address this, we made injections of a retrograde tracer (CTB conjugated to either 488 or 647 Alexa dyes) into visual (Fig. 2, blue) or somatosensory cortex (Fig. 2, yellow) in WT mice at postnatal day 1 (n = 6). As early as P2, CTB injections into visual cortex (Fig. 2A) resulted in retrograde labeling of TC neurons in dLGN as well as the pulvinar, but not in ventrobasal complex (VB) (Fig. 2B). In all six cases, the labeled axons within TRN were restricted to the head (i.e., dorsal-caudal region) of the nucleus (Fig. 2C, see Fig. 1B). Moreover, we failed to detect any axonal labeling ventral to the apex of TRN, a boundary that delineates visual and nonvisual sectors of the TRN in the adult mouse [43]. A similar modality specific patterning was evident following CTB injections into somatosensory cortex (Fig. 2D). These injections labeled TC neurons that were restricted to somatosensory thalamic nuclei such as the ventroposteromedial (VPM) and ventroposterolateral (VPL) nuclei (i.e., ventrobasal complex (VB), Fig. 2E). In TRN, the axons from somatosensory nuclei were restricted to the ventro-medial region at or below the apex (Fig. 2F) adjacent to VB.
While TRN sectors are established at perinatal ages, it is unclear whether this arrangement is preserved in the absence of peripheral sensory input. To test for this, we made similar CTB injections in Math 5−/− mice, a mutant mouse that lacks > 95% of retinal ganglion cells and that is completely devoid of retinal input to the brain [31, 33, 37]. Similar to WT, injections into visual cortex of Math 5−/− mice (Figs. 2G and H) led to retrograde labeling in the head of TRN above the apex (Fig. 2I, blue, n = 3), while injections into somatosensory cortex (Figs. 2J and K) labeled axons in the ventromedial region of TRN at or below the apex (Fig. 2L, yellow, n = 2).
Taken together, these results reveal that the sensory specific organization of TRN is established by perinatal ages and that such organization is retained in the absence of sensory stimulation. These results also allowed for us to examine the formation of visual feedforward and feedback connections by targeting the dorsal caudal aspect (head) of TRN (see Fig. 1B).
dLGN innervation of TRN
To visualize excitatory feedforward projections from dLGN to TRN, we crossed corticotropin releasing hormone-Cre (CRH-Cre) mice with an Ai9 reporter line [43, 58, 59] in order to express tdTomato in TC neurons of 1st order thalamic nuclei such as the dLGN and VB complex [43, 44, 58]. Figure 3 provides examples of coronal sections of labeled TC axons at different postnatal ages in both WT (Fig. 3A) and Math 5−/− (Fig. 3B) mice. For both groups, as early as P0, TC axons coursed through TRN on a trajectory from the medial aspect of the nucleus to the lateral edge where they aggregate to form the internal capsule. At later ages, axons within TRN were arranged into fascicles forming a reticular configuration. This arrangement emerged as early as P7 and was readily apparent within the head of vTRN by P14.
It is important to note that the projections from dLGN to TRN involve fine caliber axon collaterals while thalamocortical axons are much larger and bundled into fascicles [3]. Indeed, these processes are difficult to resolve at the light microscopic level. Thus, to visualize terminal boutons of thalamocortical axon collaterals in TRN we used an antibody against a vesicular glutamate transporter 2 (vGluT2) [38–40, 60, 61]. Figure 4A illustrates the punctate vGluT2 labeling of TC axon collaterals (blue) in relation to TC tdTomato axonal labeling (purple) and TC somatic NeuN labeling (yellow) in the same coronal section through the TRN of an adult mouse.
To determine when TC axon collateral terminals appear in vTRN and whether they are affected by the removal of retinal signaling, we examined vGluT2 labeling in coronal sections from age-matched WT (Fig. 4B, top) and Math 5−/− (Fig. 4B, bottom) mice. To analyze the rate of TRN innervation by TC axon collaterals, we quantified the percentage of vTRN area covered by vGluT2 puncta across different postnatal ages and plotted the median values (Fig. 4C; [43]). Measurements were confined to the head of TRN above the apex (Fig. 1B.) In WT, the area covered by vGluT2 puncta increased with age (filled circles, n = 104 sections, Kruskal-Wallis: p < 0.0001). During the first postnatal week (P5-P7) vGluT2 labeling was sparse but showed a progressive increase, eventually covering nearly half of vTRN by P10 (P5 and P7 vs P10, Dunn’s multiple comparison test, p < 0.01) and then all of it by P21 (all comparisons between P10, P14 and P21, p < 0.01). At P21, thalamic innervation of visual TRN plateaued and did not differ from P28 (p = 0.3063). Thus, innervation of vTRN by TC axon collaterals occurred largely during the second postnatal week (P7-P14) and encompassed the entire visual sector by the end of the third postnatal week (P21). Similar to WT, there was an age-related increase in vGluT2 puncta in the vTRN of Math 5−/− mice (Fig. 4C, open circles, n = 72 sections, Kruskal-Wallis: p < 0.0001). Moreover, the age-related increase in vGluT2 labeling had a similar trajectory in WT and Math 5−/− mice (two-way ANOVA, Fgenotype (1,164) = 0.5503, p = 0.4593). Together, these data demonstrate that while TC axons course through TRN at birth, the appearance of axon collaterals increases during the second postnatal week and reaches adult-like levels by the end of the third postnatal week (P21). Furthermore, the absence of retinal inputs does not alter the time course of TC innervation of visual TRN.
TRN axon innervation of dLGN
To visualize inhibitory feedback projections from TRN to dLGN we used GAD65-EGFP mice [27, 35, 36], which labels TRN neurons with EGFP as early as P0 (Fig. 5A). We also crossed Math 5−/− mutants with GAD65-EGFP mice to assess whether the absence of retinal signaling disrupted the timing of feedback innervation.
Figure 5 provides coronal sections of dLGN from early postnatal GAD65 and GAD65 x Math 5−/− mice. In WT (Fig. 5B, top row), TRN fibers were absent from dLGN at birth, but began to innervate the nucleus at P2 from the medial-ventral border. By the end of the first postnatal week TRN input encompassed all of dLGN in a dense plexus of terminals. We analyzed the rate of TRN innervation by quantifying the spatial extent of terminals expressed as a percentage of dLGN area [29, 30] and plotted the median values (Fig. 5C) for WT and Math 5−/− groups. In WT (filled circles, n = 47 sections), TRN terminals arrived in dLGN at P2 and showed a progressive increase with age (Kruskal-Wallis, p < 0.0001). Most of the projections arrived between P2-4 and showed nearly a 3-fold increase over this period (median values, P2 = 22.40 vs P4 = 61.34). By P6, TRN innervation was complete, forming a dense but diffuse network that spanned the entire nucleus (median values, P6 = 83.87, P9 = 82.67, P18 = 86.08). In Math 5−/− (n = 59 sections). The absence of retinal signaling appeared to accelerate TRN innervation of dLGN (Fig. 5B bottom row). At P0, TRN axons could be seen innervating the ventral medial border, and by P2 they nearly encompassed all of dLGN with some axons extending close to the dorsal lateral border just beneath the optic tract. TRN terminals showed a two-fold increase in innervation between P0 and P2, (open circles n = 59) which rose steadily between P4-6 (median values: P0 = 26.18, P2 = 51.02, P4 = 61.28). This pattern of innervation was significantly different from WT mice showing greater innervation between P0-3 (multiple K-S tests, for P0, P2 and P3, D = 1.000, adjusted p-value = < 0.05). Together, these data demonstrate that TRN projections arrive in dLGN during the first postnatal week and that loss of retinal signaling accelerates innervation but does not seem to alter the overall density or pattern of innervation.
Development of feedforward synaptic connections between dLGN and TRN
We adopted an optogenetic approach to assess when functional feedforward connections between dLGN and TRN neurons emerge and to test whether this time course was influenced by the loss of retinal signaling. To photoactivate TC axon collateral terminals and record postsynaptic responses in TRN neurons, we used CRH-Cre x Ai32 (ChR2-EYFP) mice and crossed these onto a Math 5−/− background. To determine if ChR2 expression in TC neurons was sufficient to drive postsynaptic activity in TRN at early postnatal ages, we recorded light evoked activity directly from TC neurons. At P4, ChR2-EYFP was evident in TC neurons, and blue light stimulation with brief pulses (1ms) led to light evoked depolarizations and spike firing (Figure S1A-B).
To examine feedforward excitatory postsynaptic activity in TRN, we presented repetitive trains of blue light at different temporal frequencies (0.25Hz, 0.5Hz, 1Hz, 5Hz, 10Hz, 50Hz) and conducted voltage clamp recordings using a potassium-based internal solution while holding vTRN neurons at -70mV. We recorded from 186 WT and 114 Math 5−/− neurons between P5-28 that were verified to be in vTRN based on their biocytin filled reconstructions.
Examples of the light evoked responses to blue light pulse trains (1ms, 20 pulses) trains presented at 0.5Hz, 5Hz, and 10Hz are shown in Fig. 6. During the first postnatal week (P5-7), photoactivation of TC axon collaterals in visual TRN evoked little postsynaptic activity. EPSC activity was rare, of low amplitude and could only follow low rates of stimulation (e.g., 0.5Hz). Between the second and third postnatal weeks in both WT and Math 5−/− mice, the incidence of light evoked EPSC activity was more prevalent, greater in amplitude and capable of following higher rates of stimulation (e.g., 10 and 50Hz by week 3). A similar profile was observed during the fourth postnatal week with robust excitatory responses recorded throughout the stimulus train. Of notable significance was the emergence of synaptic depression when higher rates of stimulation were used. For example, by week 3, at 5 Hz and 50 Hz the amplitude of EPSCs began to attenuate after the initial pulse but then stabilized to values that were about half the amplitude of the first response. This form of synaptic depression can be quantified by generating paired pulse ratios where the amplitude of the Nth response within a stimulus train can be is divided by the initial response (EPSCn/EPSC1; see Fig. 7C-D).
A summary of these age-related changes for WT and Math 5−/− mice are shown in Fig. 7, where we plot the incidence of responsive neurons (Fig. 7A), the amplitude of the initial (and maximal) response (Fig. 7B), and the paired pulse ratios (Fig. 7C EPSCn/EPSC1; Fig. 7D EPSC10/ESPC1). For both WT and Math 5−/−, during the first postnatal week, only a few neurons showed weak light evoked responses (P5-7, WT: 3/18, 17%; Math 5−/−: 0/8, 0%). However, between P11-13, the incidence of light evoked EPSC activity showed about a three-fold increase (P11-13, WT: 25/38, 66%; Math 5−/−: 20/26, 77%). The incidence of light evoked activity continued to rise so that by P21 nearly all neurons tested were responsive (P19-21, WT: 28/31, 90%; Math 5−/−: 16/18, 89%). This progression was unaffected by the loss of retinal signaling with Math 5−/− mice showing a similar trajectory as WT (Fisher’s exact test, p > 0.05). Consistent with the increase in incidence, both groups showed an age-related increase in peak amplitude. Overall, amplitude increased with age (Fig. 7B, two-way ANOVA, Fage(2,138) = 5.743, p = 0.0040) and was unaffected by the loss of retinal signaling (Fgenotype(1,138) = 0.2572, p = 0.6128). In WT and Math 5−/−, peak amplitude increased during the second postnatal week but was stable between weeks 3 and 4 (Tukey’s multiple comparison test, week 2 vs week 3: p = 0.0174; week 2 vs week 4: p = 0.0096; week 3 vs week 4: p = 0.7982).
To estimate the degree of synaptic depression, we computed paired pulse ratios (Fig. 7C EPSCn/EPSC1; Fig. 7D EPSC10/ESPC1). A summary of these ratios for each stimulus pulse (EPSCn/EPSC1) for a 1 Hz train for both WT and Math 5−/− is shown in Fig. 7C. For both groups, paired pulse ratios exhibited a progressive decline after the initial response that stabilized by the 5th pulse (two-way ANOVAs with repeated measures, Fstimulus (1.415, 167.0) = 637.8, p < 0.0001). However, the magnitude of depression was significantly greater during the second postnatal week (red circles) than either the third (green squares) or fourth (blue triangles) weeks (two-way ANOVAs, WT: Fage(2,79) = 13.63, p < 0.0001, Math 5−/−: Fage(2,39) = 11.98, p < 0.0001; Tukey’s multiple comparison tests: week 2 vs week 3 or 4: p < 0.0001 in both WT and Math 5−/−).
To quantify the degree of synaptic depression at different temporal frequencies (0.25, 0.50, 1, 5, 10, and 50Hz) we calculated the paired pulse ratios between the 10th and 1st response (Fig. 7D). For both groups, there was an age dependent depression with higher temporal frequencies exhibiting the greatest decrements (two-way ANOVAs with Tukey’s multiple comparison testing, WT: Fweek(2,410) = 6.131 p = 0.0024, Ffrequency(5,410) = 92.47, p < 0.0001; Math 5−/−: Fweek(2,208) = 21.26, p < 0.0001, Ffrequency(5,208) = 115.1, p < 0.0001, for both genotypes, 0.25Hz or 0.5Hz vs all other stimulation rates: p < 0.0001, 10Hz vs 50Hz WT: p = 0.1822, Math 5−/−: p = 0.2735). Additionally, we noted that the degree of synaptic depression weakened with age with weeks 3 and 4 exhibiting larger paired pulse ratios than week 2 (WT: week 2 vs week 3: p = 0.0015, week 2 vs week 4: p = 0.0449, week 3 vs week 4: p = 0.8169; Math 5−/−: week 2 vs week 3 or 4: p < 0.0001, week 3 vs week 4: p = 0.4586)
Development of feedback synaptic connections between TRN and dLGN
To examine the emergence of feedback inhibitory activity from TRN to dLGN, we conducted whole cell recordings from SST-Cre x ChR2 mice and ones crossed on a Math 5−/− background. To determine if ChR2 expression in TRN neurons was sufficient to drive postsynaptic activity in dLGN at early postnatal ages, we recorded light evoked activity from vTRN neurons (supplemental Fig. 1C). As early as P2, ChR2-EYFP was evident in TRN and photoactivation with brief pulses (1ms) of blue light led to reliable depolarizations and spike firing (supplemental Fig. 1D).
To investigate TRN-mediated inhibition of TC neurons in dLGN, we conducted voltage clamp recordings using a cesium-based internal solution while holding neurons at 0 mV [62, 63]. We recorded 352 WT and 195 Math 5−/− neurons in dLGN between postnatal ages P2-P35.
Examples of the light evoked responses at different postnatal weeks to trains of blue light pulses (1ms) presented at different temporal frequencies (0.5Hz, 5Hz, and 10Hz) are shown in Fig. 8. In both WT and Math 5−/− dLGN, during the first postnatal week, light evoked inhibitory activity was weak and infrequent. When present, it was typically limited to the first pulse of a stimulus train. By the second postnatal week, inhibitory activity was stronger and more prevalent, but responses to repetitive stimulus trains were limited to low rates of stimulation (e.g., 0.5Hz, 1Hz and 5Hz). By the start of the third postnatal week, all TC neurons exhibited light evoked inhibition that was also accompanied by a form of synaptic depression. At temporal frequencies ≥ 5Hz, IPSC activity began to diminish after the initial pulse but stabilized midway through the stimulus train to values that were about half the amplitude of the first response.
A summary of these age-related inhibitory responses for WT and Math 5−/− mice are depicted in Fig. 9 where we plot the incidence of light evoked responses, (Fig. 9A), the amplitude of the initial evoked (and maximal) IPSC (Fig. 9B), and paired pulse ratios (Fig. 9C EPSCn/EPSC1; Fig. 9D EPSC10/ESPC1). For both groups, the incidence of light evoked responses followed a similar time course (Fisher’s exact test, p > 0.05 at all ages) with IPSC activity emerging as early as P4 (P4 WT: 5/14, 36%, Math 5−/−: 33%) and increasing rapidly thereafter so that by the end of the first postnatal week virtually all TC neurons exhibited responses (P7 WT: 18/21, 86%; Math 5−/−: 8/9, 89%, P10-12 WT: 18/18, 100%, Math 5−/−: 11/11, 100%; >P12 WT: 198/198, 100%, Math 5−/−: 116/121, 96%). Both groups showed a comparable age-related increase in IPSC amplitude (Fig. 9B; two-way ANOVA with Tukey multiple comparison testing, Fage(4,156) = 30.26, p < 0.0001; Fgenotype(1,156) = 0.1095, p = 0.7412; ). Between weeks 1–3, IPSC amplitude exhibited a progressive increase (WT or Math 5−/−: week 1 vs week 3, 4 or 5: p < 0.01, week 2 vs week 4 or 5: p < 0.05) that stabilized after week 3 (WT or Math 5−/−: week 3 vs week 4 or 5: p > 0.05).
For both groups, measurements of paired pulse ratios taken for a 1Hz stimulus train (Fig. 9C EPSCn/EPSC1) or across a range of temporal frequencies (0.25, 0.50, 1, 5, 10 and 50Hz; Fig. 9D) reflected a sustained synaptic depression. At 1 Hz, ratios computed for each pulse within the stimulus train show an initial reduction that remained constant throughout stimulation with the greatest attenuation seen prior to the third postnatal week (Fig. 9C; two-way ANOVA with repeated measures, Fage(7,115) = 11.31, p < 0.0001). However there was no difference between WT (solid symbols) and Math 5−/− (open symbols; two-way ANOVAs, week 2: Fgenotype (1,41) = 0.0003, p = 0.9866; week3: Fgenotype (1,32) = 0.0620, p = 0.8049; week 4: Fgenotype (1,26) = 2.911, p = 0.0999; week 5: Fgenotype (1,16) = 0.0048, p = 0.9455). When examined across a range of temporal frequencies, paired pulse ratios based on the 10th pulse revealed a sustained depression, with higher temporal frequencies exhibiting the largest decline (two-way ANOVAs WT: Ffrequency(5,231) = 60.54, p < 0.0001; Math 5−/−: Ffrequency(5,385) = 187.0, p < 0.0001; Tukey multiple comparison testing in both WT and Math 5−/−: 0.25Hz, 0.5Hz or 1Hz vs 5Hz, 10Hz or 50Hz: p < 0.0005). Such depression weakened with age (two-way ANOVAs WT: Fweek(4,231) = 24.03, p < 0.0001; Math 5−/−: Fweek(4, 385) = 61.23, p < 0.0001; within each stimulus and genotype, week 1 or week 2 vs weeks 3–5: p < 0.0001, week 4 vs week 5: p > 0.05) but still remained robust for both groups.
Finally, we assessed how the emergence of feedback inhibition from TRN influenced spike activity of dLGN TC neurons. We conducted current clamp recordings (WT n = 52, Math 5−/− n = 47 cells) and photoactivated TRN terminals while injecting a depolarizing current pulse to trigger a steady train spiking in TC neurons. Changes in spike firing were then calculated by comparing equivalent periods (0.5 s) of activity in the presence or absence of blue light stimulation (10 or 50 Hz).
Examples for WT (top) and Math 5−/− (bottom) neurons recorded at 2 and 4 weeks are shown in Fig. 10A in which a square current pulse (gray, 1500ms) evoked tonic firing before (left) and during blue light stimulation (500ms, blue) at 10Hz (middle) and 50Hz (right; see also supplemental Fig. 2 for recordings at week 1–4, and 6). For both groups, during postnatal weeks 1 and 2, activation of TRN terminals at 10Hz or 50Hz had little impact on spike firing. However, by week 3, TRN activation began to reduce TC spiking, and by week 4, it led to an almost complete suppression of activity. These data are summarized in Fig. 10B which plots the average firing rate (left) observed during control and photostimulation epochs. In the absence of photostimulation, both WT and Math 5−/− groups showed comparable firing rates (three-way ANOVA, mixed effects analysis of control epoch: Fweek(4,90) = 7.626, p < 0.0001, Fgenotype(1,90) = 0.06348, p = 0.8017). During photostimulation, there was an age-related suppression of TC spiking that emerged at week 3, continued to decline through week 4 and then led to a nearly complete suppression by week 6 (three-way ANOVA, mixed effects analysis: Fweek(4,90) = 13.73, p < 0.0001; Tukey multiple comparison testing: week 1 or 2 vs week 4 or 6: p < 0.0005). A similar pattern emerged when firing rates are converted to percent change (stim-control/control) in activity (Fig. 10B, right) with values approaching 100% reduction at 6 weeks of age (mean +/- SEM: WT 10Hz: 90.107 +/- 4.893, 50Hz: 92.755 +/- 3.916, Math 5−/− 10Hz: 87.517 +/- 5.035, 50Hz: 88.483 +/- 8.670).