Node of Ranvier remodeling in chronic psychosocial stress and anxiety

Differential expression of myelin-related genes and changes in myelin thickness have been demonstrated in mice after chronic psychosocial stress, a risk factor for anxiety disorders. To determine whether and how stress affects structural remodeling of nodes of Ranvier, another form of myelin plasticity, we developed a 3D reconstruction analysis of node morphology in C57BL/6NCrl and DBA/2NCrl mice. We identified strain-dependent effects of chronic social defeat stress on node morphology in the medial prefrontal cortex (mPFC) gray matter, including shortening of paranodes in C57BL/6NCrl stress-resilient and shortening of node gaps in DBA/2NCrl stress-susceptible mice compared to controls. Neuronal activity has been associated with changes in myelin thickness. To investigate whether neuronal activation is a mechanism influencing also node of Ranvier morphology, we used DREADDs to repeatedly activate the ventral hippocampus-to-mPFC pathway. We found reduced anxiety-like behavior and shortened paranodes specifically in stimulated, but not in the nearby non-stimulated axons. Altogether, our data demonstrate (1) nodal remodeling of the mPFC gray matter axons after chronic stress and (2) axon-specific regulation of paranodes in response to repeated neuronal activity in an anxiety-associated pathway. Nodal remodeling may thus contribute to aberrant circuit function associated with anxiety disorders.

arena with an empty Plexiglas cylinder on one side of the cage. The mouse's movements were tracked with Ethovision XT10 (Noldus Information Technology) for 150 s. The mouse was then returned to the home cage, and the arena cleaned.
Immediate after, for the social target trial, the same mouse was placed back into the arena for 150 s, with an unfamiliar CD1 in the cylinder. Time in the interaction zone (IZ) was measured for each trial. A social interaction ratio (time spent in the IZ with the social target present / time spent in the IZ with no target present, multiplied by 100) was calculated for each mouse. For CSDS-exposed animals, susceptible mice were defined as having SI ratios below a boundary defined as the strain-specific control mean score minus one standard deviation 2 . Other CSDS-exposed mice were considered resilient because their social interaction ratio was similar to controls.

RNA-Sequencing and differential gene expression analysis
We re-analyzed RNA sequencing data from brain samples of B6 and D2 mice after CDSD, published by us previously 2 (GEO accession GSE109315). Briefly, mice were sacrificed 6-8 days after the last CDSD session and RNA was extracted with TriReagent (Ambion). Sequencing libraries were prepared with ScriptSeq v2 RNAseq library preparation kit (Epicentre) and sequencing was performed on NextSeq500 (single-end 96 bp; Illumina). Differential expression analysis on voom normalized 3 gene expression values were performed using limma eBayes 4,5 , comparing resilient and susceptible mice to their same-strain controls. Here, we conducted gene set enrichment analysis (GSEA Desktop v4.1.0 6,7 ) using the differential expression results published in 2 . For the GSEA, we ranked the differential expression gene lists by interaction of p-value and fold change of the genes (logFC*p-value). We then analyzed these lists for enrichment of genes belonging to Gene Ontology (GO) terms of node of Ranvier (GO:all, N=40), node (GO:0033268, N=15), paranode (GO:0033270, N=11), juxtaparanodes (GO:0044224, N=10) and internode (GO:0033269, N=4).
The needle was left in place for another 3 minutes, and then slowly withdrawn (min. 3 minutes). The order of control and DREADD surgeries were balanced across and within days. For post-surgical analgesia, mice were given 5 mg/kg of carprofen subcutaneously before removing from anesthesia. Following surgery, mice were single housed and permitted to recover for 3-4 weeks before beginning behavioral tests, ensuring viral expression.

Experiment 2: DREADD -Behavioral testing and CNO injections
All behavioral tests (except for the SA test) were carried out at the start of the light phase. The mice were brought into the test room 30 minutes before each test. Light conditions for each test are detailed below, and a timeline for behavioral experiments can be seen in Figure 3. Cages were changed once a week, but never within 24h before a behavioral test. For each test the order of mice (control and DREADD) was randomized using random number generation, and the experimenters were blind to the condition of the animal. Mouse movement was recorded and tracked using Ethovision WT (v13, Noldus Technologies). CNO injections were continued once per day, between 8am -10am, for a total of 15 days. The injection order was varied by using one of four randomly generated order lists each day, and the mice were weighed every second day to ensure correct dosing and monitor wellbeing.

Effects of chronic DREADD activation on anxiety-like behavior (OFT, EZM2)
On day 13, prior to receiving CNO, we carried out the open field test (OFT). The light conditions were bright to ensure the anxiogenic nature of the test (290 lux). Each mouse was allowed to explore an arena (50 x 50 cm) for 5 minutes. We defined the center of the arena as 5 cm away from the walls at each point. The time the mice spent in the center vs periphery was computed. After the test, each mouse received an injection of CNO and was returned to their home cage.
On day 14, the EZM was repeated with slight modifications (EZM2). To enhance novelty and reduce habituation-induced lack of motivation to explore, we added fresh bedding material to the open zones (changed between each mouse). The apparatus, environmental conditions, and recorded parameters were the same as in EZM1.
After the test, the mice received a CNO injection.

Effects of chronic DREADD activation on social behavior (SA)
To test for effects of chronic vHPC-mPFC activation on social avoidance behavior, we performed the SA test at the end of the light phase of day 15. To avoid all acute effects of CNO, the mice did not receive an injection this morning. After acclimating to the test room, each mouse went through two trials of the SA test similarly as after CSDS (see above). Here, a naïve male wild-type B6 mouse was used as a social target. The time spent in the IZ and the social interaction (SI) ratio were computed.

Effects of an acute re-activation of the chronically activated projection on anxiety-like behavior (EPM).
To explore whether the chronic activation had affected the acute response to CNO, we performed an additional test of anxiety-like behavior following a priming injection.
The morning after the SA test (day 16) each mouse received an injection of CNO in the behavioral test room, and after 20-30 minutes they were tested in an elevated plus maze (EPM). The EPM measures anxiety-like behavior with similar parameters as the EZM, but the novel apparatus was expected to minimize habituation-related lack of exploratory drive. To start the test, mice were placed in the center of the apparatus and allowed to freely explore the two opposing closed arms, and the two opposing open arms. Time spent in each arm type was tracked, along with time spent in the center area (as a proxy for risk assessment behavior) 8 .

Imaging Experiment 1: Imaging was performed with ZEISS LSM 880 Confocal Laser
Scanning microscope with AiryScan (Zeiss). The distance from the bregma and the position of the ACC (layer 5/6) or forceps minor for each section was first determined at 10X magnification with a mouse brain atlas 9 . Nodes were then identified with a 63X oil objective in layers V/VI of the ACC and in the forceps minor. To image individual nodes within a field of view, a region around a node was cropped, and a zstack of the cropped region was acquired. Z-stacks were acquired at a resolution of 0.04 x 0.04 x 0.10 µm. Experiment 2: Imaging was performed as above but the hippocampal fimbria was first identified using 20X magnification. Paranodes, identified with 63X oil objective, overlapping with mCherry axons (mCherry+) as well as paranodes that did not colocalize with mCherry (mCherry-) were imaged within the hippocampal fimbria.

3D segmentation and morphometry of paranodes and juxtaparanodes
We developed an automated pipeline to segment and analyze the morphology of paranodes and juxtaparanodes, as well as to measure the length of nodes of Ranvier in the acquired 3D microscopy images. The pipeline initially segmented paranodes and juxtaparanodes applying geometric deformable models. However, because more than one pair of paranodes or juxtaparanodes were captured in the acquired images, the pipeline determined the main orientation of the segmented paranodes and juxtaparanodes and excluded those not along the main orientation, as shown in Figure S4a-f.
In more detail, we first applied a 3D median filter using a 5 x 5 x 3 sliding window to denoise the acquired 3D images of paranodes (red channel) and juxtaparanodes (green channel) separately to each channel. For segmentation we fused the medianfiltered red and green channels into a single channel 3D image, denoted as , by taking the maximum intensity value between the two channels at each voxel ( Figure   S4a). To segment the 3D image , first, we applied Frangi filtering 10 to to enhance its curvilinear structures, i.e., paranodes and juxtaparanodes, and suppress the background. Then, we thresholded the enhanced image to generate a 3D binary image used to initialize the Chan-Vese active surface model 11 . We used the implementation of the Chan-Vese model available in Matlab's Image Processing Toolbox (version 2018b). We set the parameters as follows: contraction bias was 0.1, smoothness factor was 0.1, and the maximum number of iterations was 100.
Applying the connected component analysis to the segmentation result, we generated a preliminary segmentation of paranodes and juxtaparanodes denoted as ( Figure S4b). To exclude segmented components other than the paranodes and juxtaparanodes of interest, we first generated a 2D maximum intensity projection of the label image along the direction of the focal plane, z-axis, as shown in Figure   S4c. We applied Hough transform 12 to the maximum projection image of to detect line segments in the image. We used the slope of the longest detected line segment, which represented the main orientation of the segmented paranodes and juxtaparanodes, to draw a line * that expanded to the image borders. The dashed line in Figure S4c shows * in the maximum projection image of associated with the main orientation of the segmented components. Then each segmented component was projected on * , and its projection length was measured (Figure S4d). The segmented components associated with the two longest projections, with nonintersecting projections, were selected as the final labels for the paranodes and juxtaparanodes of interest. For that, we first selected the longest projection and then the second longest projection that did not intersect with the longest projection.
Because we applied the segmentation on the fused image, we used the two final segmented components as the initialization surfaces to segment paranodes and juxtaparanodes on 3D median-filtered images separately, using the Chan-Vese model with the same parameter settings as described earlier. Figures S4e and f show the segmentation boundary of the paranode and juxtaparanode of interest in their corresponding channels.
We quantified morphological aspects of the segmented paranodes and juxtaparanodes in 3D following the approach in references 13,14 . We first extracted the skeleton of paranodes by applying a distance transform-based skeletonization method from 15 (Figure S4g1 and g2). With a plane perpendicular to the skeleton, we automatically extracted cross-sections along the length of segmented paranodes.
The cross-sectional morphology of paranodes was quantified by the equivalent diameter and the length of the minor and major axes of the fitted ellipse. Moreover, we measured the length of paranodes by measuring the arc length of the acquired skeletons, as shown in Figure S4g. The same procedures were applied to analyze the morphology of juxtaparanodes. Denote the set of voxel coordinates in two distinct paranodes by and . We measured the length of a node of Ranvier, Figure S4g2, by using a robust version of ( , ) = min ∈ min ∈ ( , ), where (. ) is the Euclidean distance between two points. Define the distance between the set of points and a point as ( , ) = min ∈ ( , ). Then, the robust distance between two paranodes was defined as follows: where 2 is the 2 nd percentile.

Statistical analysis
We assessed group differences in node and paranode morphology using a mixed model design, in which individual mice and paranode dependency (two paranodes originate from the same node are presumed to be non-independent) were treated as random factors and group (control, resilient and susceptible) and staining batch as fixed factors. We compared behavioral test differences between groups using an unpaired (two-tailed) Student's t test or a Mann-Whitney U test in case data were non-normally distributed. Two-way repeated ANOVA was used to analyze repeated testing in the EZM task. Mixed model analysis was performed using R (4.2.2) and other statistical analyses using Prism 8.   NR=node of Ranvier.