Inhibitory interneurons distribute widely across the mouse thalamus and form ontogenetic spatial clusters

The proportion and distribution of local inhibitory neurons (interneurons) in the thalamus varies widely across mammals. This is reflected in the structure of thalamic local circuits, which is more complex in primates compared to smaller-brained mammals like rodents. An increase in the number of thalamic interneurons could arise from addition of novel interneuron types or from elaboration of a plesiomorphic ontogenetic program, common to all mammals. The former has been proposed for the human brain, with migration of interneurons from the ventral telencephalon into higher order thalamus as one of its unique features (Letinic and Rakic, 2001). Here, we identify a larger than expected complexity and distribution of interneurons across the mouse thalamus. All thalamic interneurons can be traced back to two developmental programs: one specified in the midbrain and the other in the forebrain. Interneurons migrate to functionally distinct thalamic nuclei, where the midbrain-derived cells populate the sensory thalamus, and forebrain-generated interneurons only the higher order regions. The latter interneuron type may be homologous to the one previously considered to be human-specific, while we also observe that markers for the midbrain-born class are abundantly expressed in the primate thalamus. These data therefore point to a shared ontogenetic organization of thalamic interneurons across mammals.


31
The thalamus is a forebrain structure that develops from the diencephalic prosomere 2 (p2)   Science. Allen Cell Types Database. Available from: celltypes.brain-map.org) and DropViz

30
Unexpectedly however, 22.1±4.0% of the total GABAergic population in TC regions did not 31 express Sox14 (Fig. 1B,C; 3Bii), and these cells appeared spatially largely non-overlapping 32 with the Sox14 + class (Fig. 1C,D). In particular, we observed the largest proportion of Sox14 -    20µm-thick coronal section, displayed as a z-projection and showing distribution of reconstructions of the Sox14 GFP/+ thalami ( Fig. 1C; 2B,C). The data best fit two spatial 1 clusters, as assessed from the silhouette score (Fig. 2Bii Fig. 2B,D). The two thalamic molecular GABAergic groups 5 therefore occupy their own respective spatial clusters, with the Sox14cells located more 6 rostrally and medially compared to the Sox14 + interneurons. shown. Scale bars, 500µm. C. Performance of unsupervised k-Means algorithm in 1 identifying thalamic interneuron spatial clusters from the P14 Sox14 GFP/+ data (n=3 brains, 2 see also Fig. 1) as measured by the silhouette score, which varies with number of clusters 3 (k). We choose k=2 as this point has the highest score. D. Proportion of Sox14 + and Sox14 -4 GABAergic cells in each spatial cluster, averaged over three brains (mean±SE).

6
To independently confirm our findings and control for potential effects of looking at a juvenile 7 age (P14), we also mapped anatomical distribution of all Gad1 + and Chrna6 + cells across the 8 adult mouse TC nuclei at P56, using the Allen Mouse Brain Atlas (© 2004 Allen Institute for 9 Brain Science. Allen Mouse Brain Atlas. Available from: mouse.brain-map.org; Lein et al.,

16
Mouse thalamus therefore exhibits wider interneuron distribution and diversity than has been 17 previously reported, with at least two molecularly and spatially distinct classes. The largest 18 interneuron class, which is distributed across FO and HO sensory nuclei including the dLGN, can be defined as Sox14 + . Conversely, the smaller Sox14 -GABAergic population is found   of Sox14 + Gad1 + and Sox14 -Gad1 + cells across TC regions in P14 Sox14 GFP/GFP relative to 5 P14 Sox14 GFP/+ data (mean±SE, n=3 brains/genotype). There is a significant reduction in the 6 Sox14 + Gad1 + population (p=2.7×10 -4 , two-sample two-tailed t-test), but no statistically 7 significant difference in the size of the Sox14 -Gad1 + group (p=0.4, two-sample two-tailed t-

15
We therefore conclude that the Sox14 + thalamic interneurons derive from the midbrain, and   the Sox14 + interneuron precursors migrating from the midbrain into the first and higher order 24 sensory TC regions (Fig. 5A). Previously, dLGN interneurons were found to populate this looked at the numbers and migratory morphology of GFP + (i.e. En1 + ) cells in the thalamus at 1 E16.5, E17.5, P0.5, P1.5 and P2.5. We focused on the dLGN, LP and VP, but left out the 2 PO and MG, due to low overall numbers of interneurons in these two regions in the 3 juvenile/adult mouse thalamus (Fig. 1, Supp. Fig. 1).

4
At E16.5 no GFP + cells were present in the thalamus. From E17.5 to P2.5 their numbers 5 progressively increased in all of the regions analysed (Fig. 5A

11
To infer their direction of migration, we determined the leading process orientation of 12 migrating GFP + cells along all three dimensions (ventro-dorsal, latero-medial, caudo-rostral;

19
Heat maps indicate that at a population level (integrated across dimensions), GFP + cells 20 migrate into the dLGN, LP and VP in a caudo-rostral and dorso-ventral direction (Fig. 5D), 21 consistent with the position of the thalamus in the brain relative to their midbrain origin.
However, GFP + precursors in the dLGN and LP have a dominant medio-lateral orientation, 23 while those in the VP an opposite, latero-medial orientation, as can also be seen from polar 24 histograms (Supp. Fig. 2C). This suggests that midbrain-derived interneuron precursors 25 enter TC regions simultaneously in two distinct streams, one migrating rostro-ventro-laterally

18
Altogether, we therefore conclude that the Sox14 -Pvalb + thalamic interneurons originate from

25
Confocal z-stacks covering the extent of the thalamus across all axes (caudo-rostral, ventro-1 were then viewed with the Neurolucida software. TC nuclei were identified from the DAPI 2 counterstain, using cytoarchitectonically recognizable structures, such as the dLGN, the 3 habenular complex, the TRN, the anterior pretectum and the fasciculus retroflexus (fr), as 4 landmarks for orientation and reference. The cells of interest (Table 3) were assigned to TC 5 regions by comparing the sections to the Allen Brain Reference Atlas and annotated and 6 counted manually. For each brain, only one hemisphere was analysed (chosen in a 7 randomized way). For experiments using Gad1 + and Chrna6 + in situ hybridization data from 8 the Allen Mouse Brain Atlas resource (© 2004 Allen Institute for Brain Science. Allen Mouse 9 Brain Atlas. Available from: mouse.brain-map.org; Lein et al., 2006), all images of P56 10 C57BL/6J coronal brain sections containing the thalamus were downloaded for each gene

11
(every 8 th 25µm-thick section, sampling every 200µm across the thalamus), and analysed in 12 the same way as described above.

14
3D reconstructions of cell (Table 3) distributions across thalamic regions were generated for 15 each brain separately using the Neurolucida software (MBF Bioscience), from the acquired 16 confocal z-stacks or Allen Mouse Brain Atlas in situ hybridization data as described above.

17
For each image the outline of the thalamus and the surrounding structures was manually 18 traced using the 'contour' function and the cells were annotated with the 'marker' function,

19
placed at the centre of the soma. Traced images were then aligned in sequential rostro-

15
We also performed k-Means clustering on the 3D distribution of Gad1 + cells obtained from in 16 situ hybridisation data from the Allen Mouse Brain Atlas.

48
Thalamic nuclei convey diverse contextual information to layer 1 of visual cortex.
related regions based on differential sensitivity to engrailed proteins.    distribution of (i) Gad1 + and (ii) Chrna6 + cells. In (i), k-Means clustering was applied to the 6 data using k=2 (highest silhouette score, 0.512); the resulting spatial clusters are shown as a 7 z-and y-projection and colour-coded. One dot represents one neuron. ISH data was 8 downloaded from the Allen Mouse Brain Atlas (© 2004 Allen Institute for Brain Science.