Ultrastructure of synaptic connectivity within sub-regions of the SCN revealed by 1 genetically encoded EM tag and SBEM 2

13 The suprachiasmatic nucleus (SCN) in the hypothalamus of the vertebrate brain is the 14 central pacemaker regulating circadian rhythmicity throughout the body. The SCN 15 receives photic information through melanopsin-expressing retinal ganglion cells 16 (mRGC) to synchronize the body with environmental light cycles. Determining how 17 these inputs fit into the network of synaptic connections on and between SCN neurons 18 is key to impelling our understanding of the regulation of the circadian clock by light and 19 unraveling the relevant local circuits within the SCN. To map these connections, we 20 used a newly-developed Cre-dependant electron microscopy reporter, APEX2, to label 21 mitochondria of mRGC axons, and serial blockface scanning electron microscopy to 22 resolve the fine structure of mRGC in 3D volumes of the SCN. The maps thus created 23 provide a first draft of the patterns of connectomic organization of SCN in the core and 24 the shell, composed of different neuronal subtypes, and here shown to differ with regard 25 to the patterning of their mRGC input as the shell receives denser mRGCs synaptic 26 inputs compared to the core. This challenges the presently held view that photic 27 information coming directly from the retina is mainly integrated by the core region of the SCN.


INTRODUCTION
In mammals, circadian organization of physiology, behavior, and metabolism are 34 necessary for a healthy lifespan. Although circadian rhythms are cell-autonomous and 35 are found in almost every cell, the ventral hypothalamic brain region suprachiasmatic  Together, these results bring us better insights into the complexity of SCN 115 connectivity and the regional specialization of photic information integration. They also 116 constitute a framework to understand the age and disease-dependent deterioration of 117 circadian organization.  In addition to the ultrastructural features of cells in the SCN, we also found 170 axonal processes that contained darkly stained mitochondria, which are most likely from 171 the mRGCs. As an initial random characterization of these axons with densely labeled 172 mitochondria, we divided the volume on a 5 x 5 grid that resulted in 25 boxes. Out of the 173 25 boxes, 9 were selected and 5 to 10 axons per box containing darkly stained 174 mitochondria were traced and segmented for their entire length contained within the 175 volume. None of these axons traced back to a soma within the imaged volume, and 176 were not myelinated, as expected for mRGC axons expressing the APEX2 tag. All of 177 them contained more than one labeled mitochondrion. However, we found that among 178 axons containing labeled mitochondria nearly 25.45% of axons were recognized and 179 counted in the SCN shell and 22.92% of axons in the SCN core had at least one 180 unlabeled mitochondrion. In other words, based on mitochondria staining in one section 181 alone, we cannot conclusively classify an axon as being an mRGC or non-mRGC axon, 182 and therefore, for all subsequent analyses, we traced any axon and checked for labeled 183 mitochondria to classify them as mRGC axon. The unlabeled axons were likely of local 184 origin constituting intra-SCN network or afferent projections from non-retinal sources. In order to evaluate the cell characteristics in the core and shell of the SCN, we 189 first attempted to segment all cells whose nuclei are visible in the respective volumes 190 (Figure 2A). We found a slightly higher density of neurons in the shell compared to the 191 core (core: 53.3 neurons per 100,000 µm³; shell = 67.5 per 100,000 µm³; Figure 2B). 192 We traced and reconstructed the nuclei and computed their volume. We found relatively 193 larger nuclei in the shell (core = 318.53 ± 7.35 µm 3 ; shell = 346.45 ± 8.53 µm 3 ; p<0.05; 194 Figure 2C). We marked the astrocytes based on their characteristic features and found 195 a similar density in both SCN regions (core: 6.7 per 100,000 µm 3 ; shell: 5.7 per 100,000 196 µm 3 , Figure 2B). Too few astrocyte nuclei were fully inside the volumes for a meaningful 197 comparison of their volume. 198 We comprehensively segmented the processes of the neurons whose soma was each volume. Most SCN neurons are either bipolar or multipolar, indicating a regional 203 specialization, with more bipolar neurons in the shell (core: 34.6%; shell: 60%) and 204 more multipolar neurons in the core (core: 61.5%; shell: 35%, Figure 2E).

205
The number of dendrites extending from the SCN neurons reflects the amount of 206 axodendritic input on a macroscale and the diversity with which the dendrites project to 207 their own region compared to other regions. We randomly selected 200 dendrites, fully 208 skeletonized them, and quantified which were proximal dendrites, defined as dendrites 209 from a neuron whose soma is within our collected volume. Out of the dendrites we were  Figure 2F).

215
To further characterize the neurons in both regions, we compared the density of 216 nucleoli, a sign of high protein synthesis activity in the nucleus, and stigmoid bodies, a 217 cytoplasmic structure specific to hypothalamic regions of the brain. We found a similar 218 density of nucleoli in neurons in the core (1.44 ± 0.15 nucleoli/nuclei) and in the shell 219 (1.15 ± 0.06 nucleoli/nuclei, p=0.219; Supp Figure 1A). In the core, there is an 220 estimation of 10 stigmoid bodies per 100,000 µm³, and 28 stigmoid bodies per 100,000 221 µm³ in the shell (Supp Figure 1B).  Neurites that presented the morphological characteristics of neurons (no post-synaptic 231 synapses, small constant diameter) were excluded from the analyses. We found a total 232 of 411 synapses in the core and 371 in the shell, contacting the soma or proximal 233 dendrites ( Figure 3A). We found a significantly higher density of synapses on soma in 234 the shell (core = 8.38 ± 1.90 synapses; shell = 11.10 ± 1.56; p=0.026; Figure 3B) as well 235 as for proximal dendrites (core = 13.65 ± 1.39 synapses/100µm; shell = 20.28 ± 174 236 synapses/100µm; p<0.001; Figure 3C). We then refined the analysis by using the APEX2 labeling to classify the 240 synapses as being formed with mRGCs (with APEX2 labeled mitochondria) or non 241 mRGCs axons. We found that mRGC synaptic density in the shell is significantly higher 242 than in the core, on both soma (core = 3.58 ± 0.68 synapses; shell = 6.70 ± 0.72; 243 p=0.003; Figure 3D) and proximal dendrites (core = 10.01± 1.48 synapses/100µm; shell 244 = 23.52 ± 2.51 synapses/100µm; p<0.001; Figure 3E). On the contrary, we found no     The contact between the two dendrites is usually directly on the dendrite shaft. One 271 occurrence of a DDCS on a dendritic spine has been observed in our volume. The      b. Characteristics of mRGC and non-mRGC boutons. 378 We then randomly selected 100 mRGC and non-mRGC boutons and fully 379 segment them and their content ( Figure 6B). We did not observe a difference in volume 380 of mRGC boutons (core = 0.83 ± 0.06 µm 3 ; shell 0.88 ± 0.05 µm 3 ; p = 0.12; Figure 6C Figure 6D).

386
In addition, we noticed that only a small fraction of mRGCs boutons presented dendritic 387 intrusions (~2%), a structure that increases the contact surface between pre-and post-388 synaptic elements, while a significant percentage of non-retinal boutons present 389 dendritic intrusions (core = 35%; shell = 44 %; Figure 6E). 2.17 ± 0.19 µm 3 ; p<0.001; Figure 7F).       Güldner & Wolff, 1974) and were observed to 544 be uncommon (Moore & Bernstein, 1989). Prior to that, the discovery of this type of 545 synapse between mitral and granule cells in the mammalian olfactory bulb started a 546 growing discussion of the function of DDCS (Rall et al., 1966;Shepherd, 2009). Even 547 though extensive work was done to study DDCS in the olfactory system, the DDCS in 548 SCN was barely investigated since it was found.

549
In a previous study, we observed that dendrites with DDCS form a network in the 550 core SCN (Kim et al., 2019). We confirmed the presence of a dense network of DDCS-551 positive dendrites is present in the core and, although less dense, in the shell. The In addition to the connectivity variation between both regions of the SCN, this 581 dataset allowed us to identify specific ultrastructures that cannot be observed otherwise.

582
In particular, we were able to identify the presence of stigmoid bodies (SB) in the mouse 583 SCN, with a higher density in the shell compared to the core. SB was described as a 584 structure similar to the nucleolus located in the cytoplasm of neurons. While its exact 585 function still remains unclear, it was speculated to be involved in the aromatization of 586 androgens to estrogens, since the use of the antibody against the placental antigen 587 associated with aromatase P-450 reveals that the SBs are located in sex-steroid-588 sensitive peripheral tissues such as ovary and testis (Santolaya, 1973;Shinoda et al., 589 1993). These neurons are in multiple brain regions including the hypothalamus, 590 thalamus, amygdala, septum, hippocampus, colliculi, and brainstem (Santolaya, 1973).  We observed a relatively low number of myelinated axons compared to 604 unmyelinated ones. They appear to be passing through the SCN and made no synaptic 605 connections with the local dendrites or somas in our volume, let alone terminate in it.

606
The core has a higher density of myelinated axons than the shell, matching previous     675 To analyze these datasets, we used the publicly available software package   Counting of nuclei, astrocytes, stigmoid bodies, nucleoli 710 To count the nuclei, astrocytes, stigmoid bodies, and nucleoli in the two data sets, 711 a 5x5 grid was created. The entirety of the data set was scanned systematically by 712 starting from one corner and going through all the created grids and marking each of the 713 nuclei, astrocytes, stigmoid bodies, and nucleoli with a different color. This process was 714 done until all the areas in the volume sets of the ventral and dorsal SCN were covered.