ArticlePDF Available

Abstract and Figures

Central nervous system myelination increases action potential conduction velocity. However, it is unclear how myelination is coordinated to ensure the temporally precise arrival of action potentials and facilitate information processing within cortical and associative circuits. Here, we show that myelin sheaths, supported by mature oligodendrocytes, remain plastic in the adult mouse brain and undergo subtle structural modifications to influence action potential conduction velocity. Repetitive transcranial magnetic stimulation and spatial learning, two stimuli that modify neuronal activity, alter the length of the nodes of Ranvier and the size of the periaxonal space within active brain regions. This change in the axon-glial configuration is independent of oligodendrogenesis and robustly alters action potential conduction velocity. Because aptitude in the spatial learning task was found to correlate with action potential conduction velocity in the fimbria-fornix pathway, modifying the axon-glial configuration may be a mechanism that facilitates learning in the adult mouse brain.
iTBS shortens nodes of Ranvier (A-D) Confocal images of nodes of Ranvier (Na v 1.6; red) and paranodes (CASPR; green) in M1 (A and B) and CC (C and D) after 14 days of sham stimulation or iTBS. (E and F) Cumulative M1 node length distribution (656 sham nodes, black circles; 802 iTBS nodes, gray triangles; K-S test, K-S D = 0.19, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and average M1 node length per sham (white bars) and iTBS (gray bars) mouse (F) (n = 4 mice per group; t test, t = 2.95, p = 0.02). (G and H) Cumulative CC node length distribution (G) (867 sham nodes; 700 iTBS nodes; K-S D = 0.13, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and average CC node length per sham and iTBS mouse (H) (n = 4 mice per group; t test, t = 2.50, p = 0.04). (I and J) Cumulative M1 node length distribution (I) (452 sham nodes; 576 iTBS nodes; K-S D = 0.070, p = 0.15; inset: violin plot of node length, MWU test, p = 0.35) and average M1 node length per sham and iTBS mouse (J) (n = 3 per group; t test, t = 0.70, p = 0.52) 7 days after cessation of stimulation (14+7 days). (K and L) Cumulative CC node length distribution (K) (587 sham nodes; 696 iTBS nodes; K-S D = 0.16, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and average CC node length per sham and iTBS-treated mouse (L) (n = 3 per group; t test, t = 4.2, p = 0.01) 7 days after cessation of stimulation (14+7 days). *p < 0.05, ****p < 0.0001. Violin plots show the median (solid line) and interquartile range (dashed lines). Bars show mean ± SD. Scale bars represent 1 mm. See also Figure S2.
… 
Nodes between mature internodes are plastic (A and B) Mature node of Ranvier (Na v 1.6; red) flanked by CASPR + paranodes (blue) and mGFP + internodes (green) in M1 of Plp-CreER::Tau-mGFP mice following sham treatment (A) or iTBS treatment (B). (C and D) Cumulative mature node length distribution in M1 (C) (273 sham mature nodes, black circles; 325 iTBS mature nodes, gray triangles; K-S D = 0.11, p = 0.03; inset: violin plot of node length, MWU test, p = 0.0019) and average mature M1 node length per sham (white bars) and iTBS (gray bars) animal (D) (n = 3 per group; t test, t = 2.52, p = 0.056). (E and F) Cumulative mature node length distribution in CC (E) (495 sham mature nodes; 435 iTBS mature nodes; K-S D = 0.22, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and average mature M1 node length per individual sham-stimulated and iTBS mouse (F) (n = 3 per group; t test, t = 2.94, p = 0.04). (G and H) Mature node of Ranvier (Na v 1.6; red) flanked by CASPR + paranodes (blue) and mGFP + internodes (green) in the fimbria of Plp-CreER::Tau-mGFP in mice that underwent NL (G) or L (H) in the RAM. (I and J) Cumulative mature node length distribution in the fimbria (I) (319 NL mature nodes, black circles; 453 L mature nodes, blue diamonds; K-S D = 0.25, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and average mature node length per individual NL (white bars) and L (blue bars) mouse (J) (n = 3 per group; t test, t = 2.75, p = 0.051). # p < 0.06, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Violin plots show the median (solid line) and interquartile range (dashed lines). Bars show mean ± SD. Scale bars represent 1 mm.
… 
iTBS and spatial learning alter the size of the periaxonal space (A and B) TEM images of axons within CC of a sham-stimulated mouse (A) and iTBS mouse (B). (C) Violin plot of g-ratio (323 sham axons, white; 301 iTBS axons, gray; MWU test, p < 0.0001). (D) Average g-ratio per individual sham and iTBS animal (t test, t = 2.98, p = 0.04). (E) Axonal diameter versus g-ratio for axons in CC of sham-stimulated (black circles; n = 323 axons) and iTBS (gray triangles; n = 301 axons) mice (K-S test for g-ratio, K-S D = 0.18, p < 0.0001). (F) Violin plot of the number of myelin wraps per axon (104 sham axons; 125 iTBS axons; MWU test, p = 0.23). (G) Average number of myelin wraps per individual sham and iTBS animal (t test, t = 0.54, p = 0.61). (H and I) High-magnification TEM image of a myelinated (M) axon in CC of a sham-stimulated mouse (H) and iTBS mouse (I). White arrows, major dense line. Blue shading, periaxonal space. (J) Violin plot of periaxonal space width (75 sham axons; 60 iTBS axons; MWU test, p < 0.0001). (K) Average periaxonal space width per individual sham and iTBS animal (t test, t = 2.84, p = 0.04). (L) Axon diameter versus g-ratio for axons in the fimbria of NL (black circles) or L (blue diamonds) mice (209 NL axons; 374 L axons; K-S test for gratio, K-S D = 0.24, p < 0.0001). (M) Violin plot of g-ratio (209 NL axons, white; 374 L axons, blue; MWU test, p < 0.0001). (N) Average g-ratio per individual NL and L animal (t test, t = 2.64, p = 0.057). (O and P) TEM image of a M axon in the fimbria of a NL mouse (O) and L mouse (P). (Q) Violin plot of the number of myelin wraps per axon (55 NL axons; 55 L axons; MWU test, p = 0.10). (R) Average number of myelin wraps per individual NL and L animal (t test, t = 0.98, p = 0.38). (S) Violin plots of periaxonal space width (36 NL axons; 38 L axons; MWU test, p = 0.0023). (T) Average periaxonal space width per individual NL and L animal (t test, t = 2.78, p = 0.04). Capped lines, single myelin wrap. # p < 0.06, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Violin plots show the median (solid line) and interquartile range (dashed lines). Bars show mean ± SD for n = 3 animals per group. Scale bars represent 200 nm (A and B), 25 nm (H and I), or 50 nm (O and P).
… 
Content may be subject to copyright.
Article
Periaxonal and nodal plasticities modulate action
potential conduction in the adult mouse brain
Graphical abstract
Highlights
dAltering neuronal activity reversibly modifies node of Ranvier
length
dNeuronal activity alters periaxonal space width in the adult
mouse brain
dNodal and periaxonal plasticities synergistically modulate
action potential speed
dThe level of skill acquisition with learning correlates with
action potential speed
Authors
Carlie L. Cullen, Renee E. Pepper,
Mackenzie T. Clutterbuck, ...,
Jennifer Rodger, Renaud B. Jolivet,
Kaylene M. Young
Correspondence
kaylene.young@utas.edu.au
In Brief
Cullen et al. show that mature brain cells,
called oligodendrocytes, undergo
ultrastructural changes during learning.
They lengthen the nodes of Ranvier and
compress the periaxonal space to speed
up action potential conduction. They also
report that faster information transfer
speeds correlate with greater skill
acquisition during learning.
Cullen et al., 2021, Cell Reports 34, 108641
January 19, 2021 ª2020 The Authors.
https://doi.org/10.1016/j.celrep.2020.108641 ll
Article
Periaxonal and nodal plasticities modulate action
potential conduction in the adult mouse brain
Carlie L. Cullen,
1,8
Renee E. Pepper,
1,8
Mackenzie T. Clutterbuck,
1
Kimberley A. Pitman,
1
Viola Oorschot,
2,7
Loic Auderset,
1
Alexander D. Tang,
3
Georg Ramm,
2
Ben Emery,
4
Jennifer Rodger,
3,5
Renaud B. Jolivet,
6
and Kaylene M. Young
1,9,
*
1
Menzies Institute for Medical Research, University of Tasmania, Hobart, TAS 7000, Australia
2
Ramaciotti Centre for Cryo-Electron Microscopy, Monash University, Melbourne, VIC 3800, Australia
3
Experimental and Regenerative Neuroscience, School of Biological Sciences, University of Western Australia, Perth, WA 6009, Australia
4
Jungers Center for Neurosciences Research, Department of Neurology, Oregon Health and Science University, Portland, OR 97239-3098,
USA
5
Perron Institute for Neurological and Translational Research, Perth, WA 6009, Australia
6
De
´partement de Physique Nucle
´aire et Corpusculaire, University of Geneva, 1211 Geneva 4, Switzerland
7
Present address: European Molecular Biology Laboratory, Electron Microscopy Core Facility, Heidelberg, Germany
8
These authors contributed equally
9
Lead Contact
*Correspondence: kaylene.young@utas.edu.au
https://doi.org/10.1016/j.celrep.2020.108641
SUMMARY
Central nervous system myelination increases action potential conduction velocity. However, it is unclear
how myelination is coordinated to ensure the temporally precise arrival of action potentials and facilitate in-
formation processing within cortical and associative circuits. Here, we show that myelin sheaths, supported
by mature oligodendrocytes, remain plastic in the adult mouse brain and undergo subtle structural modifica-
tions to influence action potential conduction velocity. Repetitive transcranial magnetic stimulation and
spatial learning, two stimuli that modify neuronal activity, alter the length of the nodes of Ranvier and the
size of the periaxonal space within active brain regions. This change in the axon-glial configuration is inde-
pendent of oligodendrogenesis and robustly alters action potential conduction velocity. Because aptitude
in the spatial learning task was found to correlate with action potential conduction velocity in the fimbria-
fornix pathway, modifying the axon-glial configuration may be a mechanism that facilitates learning in the
adult mouse brain.
INTRODUCTION
Within the central nervous system (CNS), oligodendrocytes (OLs)
elaborate myelin internodes to facilitate the rapid and saltatory
conduction of action potentials and provide vital trophic support
to axons via the periaxonal space (Simons and Nave, 2015). The
speed of action potential conduction depends on the molecular
and structural parameters of the axon, including its diameter; the
presence, length, and thickness of myelin internodes; the length
of the nodes of Ranvier; and the relative density of ion channels
clustered at the nodes and paranodes (Arancibia-Ca
´rcamo et al.,
2017;Ford et al., 2015;Freeman et al., 2016;Halter and Clark,
1993;Seidl, 2014;Young et al., 2013). Axial conduction through
the periaxonal space is also important for the saltatory propaga-
tion of action potentials (Cohen et al., 2020). These parameters
must be carefully controlled, because even seemingly small
changes in conduction velocity (CV) have the potential to alter
spike-time arrival and prevent input synchrony-dependent facil-
itation (Zbili et al., 2020). In addition, it has been proposed that
CV influences brain wave synchrony (Pajevic et al., 2014). If
this is the case, myelination must precisely establish the nodes
of Ranvier and the periaxonal space or provide some degree of
structural plasticity to this system.
Oligodendrogenesis and myelination commence late in devel-
opment (Jakovcevski et al., 2009;Kessaris et al., 2006;Lu et al.,
2002) and peak before adolescence. However, new OLs are
generated throughout life (Dimou et al., 2008;Hill et al., 2018;
Hughes et al., 2018;Rivers et al., 2008;Yeung et al., 2014;Young
et al., 2013) to replace dying cells (Koenning et al., 2012;Yeung
et al., 2014) and add myelin to previously unmyelinated (UM) or
partially myelinated axons (Hill et al., 2018). The process of mye-
lination is significantly influenced by experience, because social
isolation early in life reduces the addition of myelin to the prefron-
tal cortex (Liu et al., 2012;Makinodan et al., 2012), and reducing
visual input (monocular deprivation) shortens the myelin inter-
nodes elaborated by OLs in the affected optic nerve (Etxeberria
et al., 2016;Osanai et al., 2018). Conversely, in the adult brain,
increasing neuronal activity, either by direct neuronal stimulation
or through learning a new skill or task, promotes oligodendro-
genesis and myelination of the activated circuits (Cullen et al.,
Cell Reports 34, 108641, January 19, 2021 ª2020 The Authors. 1
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
ll
OPEN ACCESS
2019;Gibson et al., 2014;Li et al., 2010;McKenzie et al., 2014;
Mitew et al., 2018;Sampaio-Baptista et al., 2013;Steadman
et al., 2020).
In the developing zebrafish spinal cord, the extension and
retraction of internodes is regulated by distinct patterns of cal-
cium activity within the nascent myelin sheaths, and this can
be partially regulated by neuronal activity (Baraban et al., 2018;
Krasnow et al., 2018). In the mammalian brain, periodic and mito-
chondria-derived calcium transients can be detected within
myelin sheaths that increase in frequency at the peak of cortical
myelination and during remyelination in the adult mouse brain
(Battefeld et al., 2019). Even after maturation, OLs retain some
capacity for internode remodeling, with a subset of internodes
extending or retracting over time (Hill et al., 2018;Hughes
et al., 2018). This capacity is perhaps best highlighted by the
extension of established myelin sheaths to occupy an adjacent
segment of recently demyelinated axon, following the ablation
of a single myelinating OL in the zebrafish spinal cord (Auer
et al., 2018). However, the extent to which mature OLs remodel
their internodes in the healthy brain, in response to specific phys-
iological stimuli such as altered neuronal activity, has not been
explored.
Herein, we show that modulating neuronal activity, either arti-
ficially by delivering low-intensity repetitive transcranial mag-
netic stimulation (LI-rTMS) or physiologically through learning,
has no net effect on gross internode length but induces adaptive
changes to the axo-myelinic ultrastructure to alter action poten-
tial CV in the adult mouse brain. Because the performance of in-
dividual mice in the spatial learning radial arm maze (RAM) task
correlated with CV along myelinated (M) fimbria-fornix axons,
these ultrastructural changes may facilitate learning.
RESULTS
The gross myelinating morphology of mature OLs does
not change with iTBS
We have previously shown that non-invasive LI-rTMS, delivered
in an intermittent theta burst stimulation (iTBS) pattern, promotes
the survival and maturation of new OLs within the primary motor
cortex (M1) (Cullen et al., 2019). To determine whether the non-
invasive stimulation of M1 could induce adaptive changes in
myelinating OLs, we labeled a subset of mature cortical OLs
by giving a single dose of tamoxifen to adult (postnatal day [P]
83) Plp-CreER::Tau-mGFP transgenic mice one week before
commencing LI-rTMS (see STAR Methods). Following 14 days
of sham stimulation or iTBS, we analyzed the morphology of
myelinating OLs within M1 (Figures 1A and 1B) and the underly-
ing corpus callosum (CC) (Figures 1C and 1D) that were labelled
with a membrane-targeted form of green fluorescent protein
(mGFP). More specifically, we measured the length of mGFP
+
in-
ternodes that were flanked on each end by contactin-associated
protein (CASPR)
+
paranodes (Figures 1C and 1D). We found that
internodes elaborated by pre-existing mGFP
+
M1 OLs were
shorter (5–118 mm, range; 30.28 ±0.77, mean ±SD) (Figures
1E and 1F) than those elaborated in CC (~10–104 mm, range;
50.68 ±1.02, mean ±SD) (Figures 1G and 1H) but that iTBS
did not alter the average length (Figures 1F and 1H) or length dis-
tribution (Figures 1E and 1G) of internodes in either region. The
density of mGFP
+
internodes within CC (Figure S1) made it
impossible to attribute internodes to any single cell within this re-
gion. However, we were able to determine the number of inter-
nodes maintained by individual OLs within M1 and found that
this was also unchanged by iTBS (sham: 29 ±3, iTBS: 28 ±2,
mean ±SEM; t test, p = 0.62; n = 13 and 12 cells, respectively),
indicating that iTBS does not lead to detectable changes in the
gross myelinating morphology of mature OLs.
iTBS shortens nodes of Ranvier
At the end of each internode, anchoring proteins expressed by
the myelin loops, such as neurofascin 155, interact with the
axonal proteins contactin and CASPR to form paranodes (Bhat
et al., 2001;Charles et al., 2002;Klingseisen et al., 2019;Peles
et al., 1997;Sherman et al., 2005). These paranodal junctions
maintain voltage-gated sodium channels (Na
v
1.6) at the nodes
of Ranvier (Freeman et al., 2015,2016;Suzuki et al., 2004). It
has been suggested that node length is plastic and may change
in response to altered neuronal activity to fine-tune action poten-
tial propagation in accordance with information processing
needs (Arancibia-Ca
´rcamo et al., 2017;Ford et al., 2015); how-
ever, this has not yet been confirmed experimentally.
To determine whether iTBS affects specific axonal domains,
we immunolabeled coronal brain sections from iTBS and sham-
stimulated animals to visualize paranodes (CASPR) and nodes
of Ranvier (Na
v
1.6) (Figures 2A–2D). By identifying regions of
dense Na
v
1.6 staining that were clearly flanked by abutting
CASPR
+
paranodes, we quantified the length of individual no-
deswithinM1(Figures 2A, 2B, and 2E) and the underlying CC
(Figures 2C, 2D, and 2F). We found that 14 days of iTBS treat-
ment shifted node length distribution toward shorter nodes
within both M1 (Figure 2E) and CC (Figure 2G), which corre-
sponded to an ~19% reduction in average node length in M1
(Figure 2F) and an ~16% decrease in CC (Figure 2H). Nodal
shortening was not accompanied by a change in paranode
length (or length distribution) (Figures S2A–S2D). To confirm
the effect of iTBS on node length, we also performed high-res-
olution stimulated emission depletion (STED) microscopy to
visualize callosal nodes (Na
v
1.6) and paranodes (CASPR) in a
separate cohort of animals following 14 days of iTBS or sham
stimulation (Figures S2E and S2F). Consistent with our initial
observations, we found that iTBS shifted the node length distri-
bution toward shorter nodes (Figure S2G), which corresponded
to a reduction of ~18% in average node length per mouse within
CC (Figure S2H).
To evaluate the impact of treatment duration on node length,
we compared the effect of delivering a sham stimulation or
iTBS for 7, 14, or 28 days (Figure S2). 7 days of iTBS was insuf-
ficient to alter average node length in M1 (Figure S2I) or CC (Fig-
ure S2J). By contrast, 28 days of iTBS shortened average node
length by ~19% in M1 (Figure S2I) and ~18% in CC (Figure S2J),
which is equivalent to 14 days of stimulation. Therefore, in
response to iTBS, nodes shorten and then remain short while
stimulation is maintained. To determine whether ceasing iTBS
results in node length reverting to that measured in sham-stimu-
lated mice, we delivered 14 days of sham or iTBS and then
ceased stimulation for 7 days (14+7 days) before measuring
the length of individual nodes. At this time point, the average
2Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
length and length distribution of M1 nodes was equivalent in
sham and iTBS mice (Figures 2I and 2J), suggesting that this
form of plasticity is reversible. Within CC of 14+7 day iTBS
mice, node length distribution remained significantly shifted to-
ward shorter nodes (Figure 2K), corresponding to an ~20%
decrease in average node length (Figure 2L), perhaps suggesting
that nodal changes are more long-lived in the white matter.
Spatial learning lengthens nodes of Ranvier
Learning physiologically modulates neuronal activity (Benche-
nane et al., 2010;Dupret et al., 2010,2013;Negro
´n-Oyarzo
et al., 2018). Spatial learning enhances activity within the hippo-
campal-cortical network and is accompanied by an increase in
OL addition to this network (Steadman et al., 2020), whereas
motor learning is accompanied by a similar increase in oligo-
Figure 1. iTBS does not alter OL gross myelinating morphology
(A and B) Compressed confocal z stack of a mGFP
+
(green) OL in the
primary motor cortex (M1) of Plp-CreER::Tau-mGFP transgenic mice
after 14 days of sham stimulation (A) or iTBS (B). Inset, mGFP
+
cell
bodies show co-labeling for CC1 (red) and OLIG2 (blue).
(C and D) Single mGFP
+
internodes flanked on either end by CASPR
+
paranodes (blue) and dense Na
v
1.6 staining at the node of Ranvier (red)
in the corpus callosum (CC) of sham-stimulated mice (C) or iTBS mice
(D).
(E and F) Cumulative M1 mGFP
+
internode length distribution for sham
(black circles) and iTBS (gray triangles) mice (E) (744 sham internodes;
733 iTBS internodes; Kolmogorov-Smirnov [K-S] test, K-S D = 0.049, p =
0.32; inset violin plot of internode length, Mann-Whitney U [MWU] test,
p = 0.63) and the average M1 internode length per sham (white bars) and
iTBS (gray bars) mouse (F) (n = 3 mice per group; t test, t = 0.31, p = 0.77).
(G and H) Cumulative CC mGFP
+
internode length distribution (G) (n =
144 sham internodes; n = 166 iTBS internodes; K-S test, K-S D = 0.08,
p = 0.63; inset violin plot of internode length, MWU test, p = 0.57) and
average CC internode length per sham and iTBS mouse (H) (n = 3 mice
per group; t test, t = 0.55, p = 0.61).
Arrowheads indicate the end of an internode. Violin plots show the
median (solid line) and interquartile range (dashed lines). Bars show
mean ±SD. Scale bars represent 20 mm.
See also Figure S1.
dendrogenesis and myelination in CC (Sampaio-Baptista
et al., 2013;Xiao et al., 2016). To explore the possibility
that mature OLs also respond to learning (L), we adminis-
tered tamoxifen to P60 Plp-CreER::Tau-mGFP transgenic
mice and then trained these mice (from P74) to learn a hip-
pocampal-dependent RAM task (STAR Methods;Figures
S3A and S3C). The subset of Plp-CreER::Tau-mGFP trans-
genic mice exposed to the maze, but not trained to learn
the location of food rewards, are called no-learning (NL)
controls (Figure S3B). Following 14 days of NL or L, we
analyzed the morphology of mGFP-labeled myelinating
OLs within the hippocampal fimbria (Figures 3A and 3B),
a major white matter tract that connects both hippocampi
with subcortical and cortical regions, including the thal-
amus and prefrontal cortex (Jin and Maren, 2015;Wyss
et al., 1980). Because of the density of mGFP
+
internodes
withinthefimbria(Figure S1C),itwasnotpossibletoreli-
ably attribute internodes to a single cell; however, by
measuring the length of individual mGFP
+
internodes, flanked
by CASPR
+
paranodes (Figure 3C), we determined that inter-
nodes within the fimbria were ~5–109 mminlength(inset,Fig-
ure 3D) and that spatial learning had no effect on the average
length or length distribution of internodes in this region (Figures
3D and 3E).
To determine whether RAM learning induced nodal plasticity,
we quantified the length of nodes (Na
v
1.6) and paranodes
(CASPR) (Figures 3F and 3G) within the fimbria of NL and L
mice. Unlike iTBS, which shortened nodes, spatial learning pro-
duced a shift in node length distribution toward longer nodes
(Figure 3H) and produced a corresponding ~16% increase in
average node length (Figure 3I). However, like iTBS, spatial
learning did not alter the average length or length distribution
of paranodes (Figures S3D and S3E).
Cell Reports 34, 108641, January 19, 2021 3
Article
ll
OPEN ACCESS
Nodal plasticity does not require adult
oligodendrogenesis
The addition of new OLs can influence node length in the adult
mouse brain (Schneider et al., 2016), and the administration of
tamoxifen to P60 Pdgfra-CreER
TM
::Rosa26-YFP mice, to allow
lineage tracing of oligodendrocyte progenitor cells (OPCs) during
RAM training, revealed that RAM learning increases the number
of new OLs detected in the fimbria (NL: 76 ±4 YFP
+
PDGFRa--
neg OLIG2
+
OLs/mm
2
, L: 115 ±23 YFP
+
PDGFRa-neg OLIG2
+
OLs/mm
2
[mean of n = 3 mice per group ±SD]; two-tailed un-
paired t test, t = 2.78, df = 4, p = 0.04). Because previous reports
indicate that preventing adult oligodendrogenesis simulta-
neously lengthens nodes and their flanking paranodes
(Schneider et al., 2016), RAM learning, which increases oligo-
dendrogenesis, should shorten nodes and paranodes. Instead,
RAM learning lengthened the nodes of Ranvier and had no effect
on paranode length, suggesting that adult oligodendrogenesis
cannot account for nodal plasticity in this context.
Figure 2. iTBS shortens nodes of Ranvier
(A–D) Confocal images of nodes of Ranvier (Na
v
1.6; red) and paranodes (CASPR; green) in M1 (A and B) and CC (C and D) after 14 days of sham stimulation or
iTBS.
(E and F) Cumulative M1 node length distribution (656 sham nodes, black circles; 802 iTBS nodes, gray triangles; K-S test, K-S D = 0.19, p < 0.0001; inset: violin
plot of node length, MWU test, p < 0.0001) and average M1 node length per sham (white bars) and iTBS (gray bars) mouse (F) (n = 4 mi ce per group; t test, t = 2.95,
p = 0.02).
(G and H) Cumulative CC node length distribution (G) (867 sham nodes; 700 iTBS nodes; K-S D = 0.13, p < 0.0001; inset: violin plot of node length, MWU test, p<
0.0001) and average CC node length per sham and iTBS mouse (H) (n = 4 mice per group; t test, t = 2.50, p = 0.04).
(I and J) Cumulative M1 node length distribution (I) (452 sham nodes; 576 iTBS nodes; K-S D = 0.070, p = 0.15; inset: violin plot of node length, MWU test, p = 0.35)
and average M1 node length per sham and iTBS mouse (J) (n = 3 per group; t test, t = 0.70, p = 0.52) 7 days after cessation of stimulation (14+7 days).
(K and L) Cumulative CC node length distribution (K) (587 sham nodes; 696 iTBS nodes; K-S D = 0.16, p < 0.0001; inset: violin plot of node leng th,MWU test, p<
0.0001) and average CC node length per sham and iTBS-treated mouse (L) (n = 3 per group; t test, t = 4.2, p = 0.01) 7 days after cessation of stimulation
(14+7 days).
*p < 0.05, ****p < 0.0001. Violin plots show the median (solid line) and interquartile range (dashed lines). Bars show mean ±SD. Scale bars represent 1 mm.
See also Figure S2.
4Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
To determine whether nodal plasticity can occur in the
absence of oligodendrogenesis, tamoxifen was administered
to P76 Pdgfra-CreER
TM
::Rosa26-YFP::Myrf
fl/fl
(Myrf
fl/fl
) mice
to fluorescently label OPCs and conditionally delete myelin
regulatory factor (Myrf), a transcription factor essential for
OL maturation and myelin maintenance (Emery et al., 2009).
The conditional deletion of Myrf from adult OPCs reduced
OL addition in M1 (p < 0.05) and CC (p < 0.001) by >60%
within 30 days of tamoxifen delivery (Figures S4A–S4F). How-
ever, when sham stimulation or iTBS was initiated 14 days af-
ter tamoxifen delivery (P76+14) and the nodes of Ranvier
(Na
v
1.6) were imaged in M1 at P76+28 (Figures S4Gand
S4H), iTBS again shifted the node length distribution toward
shorter nodes (Figure S4I, K-S test, p < 0.0001) and
decreased the average node length by ~16% (Figure S4J, t
test,p=0.04).WithinCCofMyrf
fl/fl
mice, iTBS also shortened
nodes (Figure S4K, K-S test, p < 0.0001; Figure S4L, t test, p =
0.01). Similarly, when P60+14 Myrf
fl/fl
mice underwent 14 days
of RAM training, length distribution of the node of Ranvier
(Na
v
1.6) was again shifted toward longer nodes in mice that
learned (L mice) compared with those that did not (NL con-
trols) (Figures S4M–S4O, K-S test, p < 0.0001), and this corre-
sponded to an ~24% increase in average node length (Fig-
ure S4P, t test, p = 0.004). These data indicate that the
nodal plasticity induced by iTBS or spatial learning cannot
be entirely attributed to new OL addition.
To explore the possibility that nodal plasticity is instead medi-
ated by mature, pre-existing OLs, we used Plp-CreER::Tau-
mGFP transgenic mice to selectively identify nodes of Ranvier
(Na
v
1.6) that were already formed before iTBS or sham stimula-
tion. These nodes were identified as being flanked by mGFP
+
pre-existing internodes (Figures 4A and 4B). Within M1, iTBS
shortened the mature nodes (Figure 4C) and tended to decrease
the average length of the mature nodes in each mouse by ~18%
(Figure 4D, p = 0.056). Within CC, iTBS also shifted mature node
length distribution toward shorter nodes (Figure 4E), which cor-
responded to an ~23% decrease in average node length (Fig-
ure 4F), suggesting that node shortening is facilitated by mature
OLs.
To determine whether the node lengthening induced by spatial
learning is also facilitated by mature OLs, we similarly analyzed
nodes flanked by mGFP
+
pre-existing internodes in the fimbria
of NL (Figure 4G) and L (Figure 4H) Plp-CreER::Tau-mGFP trans-
genic mice. Consistent with our earlier observations, learning
shifted the length distribution of mature nodes toward longer no-
des (Figure 4I) and increased the average length of mature nodes
by ~17% (Figure 4J, p = 0.051). These data suggest that mature
OLs facilitate activity-induced nodal plasticity through the subtle
expansion or retraction of existing myelin internodes.
Figure 3. Spatial learning lengthens nodes of Ranvier
(A and B) Compressed confocal z stack of a mGFP
+
(green) OLs in the hip-
pocampal fimbria of Plp-CreER::Tau-mGFP transgenic mice after 14 days of
no-learning (NL) (A) or learning (L) (B) in the radial arm maze.
(C) Example of a single mGFP
+
fimbria internode flanked on either end by
CASPR
+
paranodes (blue) and dense Na
v
1.6 staining (red) at the node of
Ranvier.
(D and E) Cumulative internode length distribution for mGFP
+
internodes of
NL (black circles) and L (blue diamonds) mice (D) (n = 532 NL; n = 685 L; K-S
test, K-S D = 0.054, p = 0.34; inset: violin plot of internode length, MWU test,
p = 0.15) and average fimbria internode length per animal (E) in NL mice
(white bars) and L mice (blue bars) (n = 3 mice per group; t test, t = 0.053,
p = 0.96).
(F and G) Confocal images of nodes of Ranvier (Na
v
1.6; red) and paranodes
(CASPR; green) in the fimbria of NL mice (F) and L mice (G).
(H and I) Cumulative node length distribution in the fimbria (H) (448 NL nodes;
520 L nodes; K-S D = 0.25, p < 0.0001; inset: violin plot of node length, MWU
test, p < 0.0001) and average fimbria node length per animal (I) in NL and L
mice (n = 4 per group; t test, t = 2.72, p = 0.03).
*p < 0.05, ****p < 0.0001. Violin plots show the median (solid line) and inter-
quartile range (dashed lines). Bars show mean ±SD. Scale bars represent
15 mm (A–C) and 1 mm (F and G).
See also Figures S1 and S3.
Cell Reports 34, 108641, January 19, 2021 5
Article
ll
OPEN ACCESS
iTBS and spatial learning alter the size of the periaxonal
space
For internodes to encroach on or retract from the nodes of Ranv-
ier, the thickness or structure of myelin must be modified. For
example, the sustained activation of extracellular signal-regu-
lated kinases 1 and 2 (ERK1/2) in mature OLs induces a marked
decrease in node length in CC of adult mice by increasing myelin
thickness (Jeffries et al., 2016), and in the optic nerve, the
thrombin-dependent detachment of paranodal loops leads to
thinner myelin (fewer myelin wraps) and longer nodes (Dutta
et al., 2018). Therefore, we performed transmission electron mi-
croscopy (TEM) to quantify the g-ratio [axon diameter/(axon +
myelin sheath diameter)] of M axons in CC of adult sham and
iTBS-treated mice (Figures 5A–5K). We found that iTBS resulted
in a shift in the distribution of axonal g-ratio measurements to-
ward smaller values (Figures 5C and 5E), which corresponded
to an ~7% reduction in the average g-ratio measured for M
axons in iTBS mice compared with controls (Figure 5D).
Because axon diameter was similar between treatment
groups (iTBS: 0.53 ±0.17 mm, sham: 0.58 ±0.16 mm; t test,
p = 0.09), the decreased g-ratio could reflect an increase in
myelin thickness; however, when we quantified the number of
myelin wraps (lamellae) around individual axons in the CC, we
found that the number of myelin wraps was not affected by treat-
ment (Figures 5D–5I). Similarly, the average thickness (period-
icity) of each wrap (the distance between each major dense
line) (capped lines in Figures 5E and 5G) (sham: 8.3 ±0.3 nm,
iTBS: 8.6 ±0.2 nm, mean ±SD; n = 3 per group; t test, t =
1.27, p = 0.27) and the cross-sectional area of the inner tongue
process (sham: 11,287 ±2,266 nm
2
, iTBS: 13,124 ±
3,179 nm
2
, mean ±SD; n = 3 per group; t test, t = 0.81, p =
0.46) were unchanged by iTBS. However, the width of the
fluid-filled space that exists between the axon and the internode,
known as the periaxonal space, increased by ~47% following
iTBS (Figures 5J and 5K), perhaps suggesting that the myelin
sheath is pushed outward or reconfigured away from the axon.
This was not the result of a generalized or widespread change
in ion balance and osmosis resulting from iTBS, because no ef-
fect was seen on the size of M1 neuronal somata (NEUN
+
MAP2
+
; sham-stimulated: 150.4 ±8.6 mm
2
, iTBS: 150.6 ±
7.6 mm
2
, mean ±SD; n = 4 mice per group; t test, t = 0.04, df =
6, p = 0.96) or callosal OL somata (mGFP
+
; sham-stimulated:
38.5 ±2.2 mm
2
, iTBS: 38.1 ±3.7 mm
2
, mean ±SD; n = 3 mice
per group; t test, t = 0.15, df = 4, p = 0.88).
Remarkably, spatial learning had the opposite effect on myelin
ultrastructure in the fimbria (Figures 5L–5T). Learning produced a
Figure 4. Nodes between mature internodes are plastic
(A and B) Mature node of Ranvier (Na
v
1.6; red) flanked by CASPR
+
paranodes
(blue) and mGFP
+
internodes (green) in M1 of Plp-CreER::Tau-mGFP mice
following sham treatment (A) or iTBS treatment (B).
(C and D) Cumulative mature node length distribution in M1 (C) (273 sham
mature nodes, black circles; 325 iTBS mature nodes, gray triangles; K-S D =
0.11, p = 0.03; inset: violin plot of node length, MWU test, p = 0.0019) and
average mature M1 node length per sham (white bars) and iTBS (gray bars)
animal (D) (n = 3 per group; t test, t = 2.52, p = 0.056).
(E and F) Cumulative mature node length distribution in CC (E) (495 sham
mature nodes; 435 iTBS mature nodes; K-S D = 0.22, p < 0.0001; inset: violin
plot of node length, MWU test, p < 0.0001) and average mature M1 node length
per individual sham-stimulated and iTBS mouse (F) (n = 3 per group; t test, t =
2.94, p = 0.04).
(G and H) Mature node of Ranvier (Na
v
1.6; red) flanked by CASPR
+
paranodes
(blue) and mGFP
+
internodes (green) in the fimbria of Plp-CreER::Tau-mGFP in
mice that underwent NL (G) or L (H) in the RAM.
(I and J) Cumulative mature node length distribution in the fimbria (I) (319 NL
mature nodes, black circles; 453 L mature nodes, blue diamonds; K-S D =
0.25, p < 0.0001; inset: violin plot of node length, MWU test, p < 0.0001) and
average mature node length per individual NL (white bars) and L (blue bars)
mouse (J) (n = 3 per group; t test, t = 2.75, p = 0.051).
#
p < 0.06, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Violin plots show
the median (solid line) and interquartile range (dashed lines). Bars show mean
±SD. Scale bars represent 1 mm.
See also Figure S4.
6Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
significant shift toward larger g-ratio values for M axons in the
fimbria (Figures 5L and 5M), which corresponded to an ~9% in-
crease in the average g-ratio (Figure 5N, p = 0.057). This increase
in g-ratio could not be explained by a change in axon diameter
Figure 5. iTBS and spatial learning alter the
size of the periaxonal space
(A and B) TEM images of axons within CC of a
sham-stimulated mouse (A) and iTBS mouse (B).
(C) Violin plot of g-ratio (323 sham axons, white;
301 iTBS axons, gray; MWU test, p < 0.0001).
(D) Average g-ratio per individual sham and iTBS
animal (t test, t = 2.98, p = 0.04).
(E) Axonal diameter versus g-ratio for axons in CC
of sham-stimulated (black circles; n = 323 axons)
and iTBS (gray triangles; n = 301 axons) mice (K-S
test for g-ratio, K-S D = 0.18, p < 0.0001).
(F) Violin plot of the number of myelin wraps per
axon (104 sham axons; 125 iTBS axons; MWU
test, p = 0.23).
(G) Average number of myelin wraps per individual
sham and iTBS animal (t test, t = 0.54, p = 0.61).
(H and I) High-magnification TEM image of a
myelinated (M) axon in CC of a sham-stimulated
mouse (H) and iTBS mouse (I). White arrows, major
dense line. Blue shading, periaxonal space.
(J) Violin plot of periaxonal space width (75 sham
axons; 60 iTBS axons; MWU test, p < 0.0001).
(K) Average periaxonal space width per individual
sham and iTBS animal (t test, t = 2.84, p = 0.04).
(L) Axon diameter versus g-ratio for axons in the
fimbria of NL (black circles) or L (blue diamonds)
mice (209 NL axons; 374 L axons; K-S test for g-
ratio, K-S D = 0.24, p < 0.0001).
(M) Violin plot of g-ratio (209 NL axons, white;
374 L axons, blue; MWU test, p < 0.0001).
(N) Average g-ratio per individual NL and L animal
(t test, t = 2.64, p = 0.057).
(O and P) TEM image of a M axon in the fimbria of a
NL mouse (O) and L mouse (P).
(Q) Violin plot of the number of myelin wraps per
axon (55 NL axons; 55 L axons; MWU test, p =
0.10).
(R) Average number of myelin wraps per individual
NL and L animal (t test, t = 0.98, p = 0.38).
(S) Violin plots of periaxonal space width (36 NL
axons; 38 L axons; MWU test, p = 0.0023).
(T) Average periaxonal space width per individual
NL and L animal (t test, t = 2.78, p = 0.04).
Capped lines, single myelin wrap.
#
p < 0.06, *p <
0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Violin
plots show the median (solid line) and interquartile
range (dashed lines). Bars show mean ±SD for n =
3 animals per group. Scale bars represent 200 nm
(A and B), 25 nm (H and I), or 50 nm (O and P).
(NL: 0.61 ±0.01 mm, L: 0.56 ±0.11 mm,
mean ±SD; n = 3 per group; t test, t =
0.87, p = 0.42) or the number of myelin
wraps elaborated (Figures 5O–5R), but it
was associated with an ~29% decrease
in the average width of the periaxonal
space (Figures 5S and 5T). These data
suggest that modulating neuronal activity
can induce adaptive changes in myelin ultrastructure to alter the
length of the node of Ranvier and the width of the periaxonal
space, two of the axon-glial parameters that strongly influence
action potential CV.
Cell Reports 34, 108641, January 19, 2021 7
Article
ll
OPEN ACCESS
The size of the periaxonal space modulates CV
The frequency and composition of nodal domains can markedly
affect action potential propagation along an axon (Ford et al.,
2015;Freeman et al., 2016;Schneider et al., 2016), and node
length is predicted to exert a strong influence on CV (Aranci-
bia-Ca
´rcamo et al., 2017;Halter and Clark, 1993). Within the
rat cortex, the estimated density of Na
v
1.6 channels at the
node is relatively stable across nodes of various lengths, and
this constant density of ion channels was predicted to ensure a
concave relationship between node length and speed of con-
duction, such that increasing node length increased CV to a
point, after which very long nodes (like very short nodes) caused
a reduction in CV (Arancibia-Ca
´rcamo et al., 2017). Furthermore,
the inclusion or omission of periaxonal space width could result
in new internodes increasing or decreasing the simulated CV of
an axon, respectively (Young et al., 2013). More recently, it
was reported that electrical conductance within the periaxonal
space facilitates the saltatory propagation of action potentials
and that changing the size of this space would theoretically alter
CV (Cohen et al., 2020).
To determine how changes in node length and periaxonal ul-
trastructure might influence CV, we performed computational
simulations of action potential propagation in M callosal axons
by adapting the mathematical model developed by Richardson
and colleagues (Bakiri et al., 2011;Richardson et al., 2000)(Fig-
ures 6A and 6B; see STAR Methods and Figure S5). We initially
explored how the size of the periaxonal space might affect CV
(Figures 6C and 6D), and in line with Cohen et al. (2020), we found
that simulated CV is up to 3.5 times slower when the periaxonal
space width is set to 20 nm rather than 0 nm (Figure 6C), which
equates to a conduction delay of up to ~9 ms (at 21C) over a dis-
tance of 1 cm (Figure 6D).
By initially configuring the model parameters so that the peri-
axonal space was uniform under the internode and paranode
(Figure 6E) and setting all other parameters to match those
measured under sham conditions (Figure 6F; Table S1), we ob-
tained a theoretical CV of 1.18 m/s at 21C (1.91 m/s at 37C)
(Figure S5). Reducing the node length to that measured after
iTBS (Figure 2;Table S1) slowed action potential propagation
by ~2.3%, whereas altering the myelin sheath parameters
(measured increase in periaxonal space and associated
decrease in g-ratio = altered myelin) (Figure 5;Table S1) exerted
a greater influence on CV, effectively slowing propagation by
~8.6% (Figure 6F). We found that modifying both sets of param-
eters—i.e., reducing node length and increasing the periaxonal
space width (altered myelin) to reflect the ultrastructural changes
observed following iTBS (Figures 2 and 5;Table S1)—had an ad-
ditive effect, slowing CV by ~10.9% (to 1.05 m/s) (Figure 6F). This
effect was exaggerated at 37C, with iTBS inducing a theoretical
~12.3% reduction in CV (Figure S5D).
We then changed the model parameters so that the periaxonal
space under the internode could change but did not exceed 3 nm
at the paranode (Figure 6G), which is the middle of the size range
that has been reported elsewhere (Nans et al., 2011;Rosenbluth,
1995;Waxman et al., 1995). In this scenario, the simulated CV for
sham-stimulated axons was slightly faster: 1.34 m/s at 21C(Fig-
ure 6H; 37C data in Figure S5F). Reducing node length still
slowed CV by ~2.3%, but the effect of altering the periaxonal
space (altered myelin) was diminished, slowing conduction by
only ~3.7%, such that the combined effect of reduced node
length + altered myelin (iTBS) was also weaker, inducing a theo-
retical ~6% reduction in CV (Figure 6H; ~5.7% reduction at 37C
in Figure S5F). In a third model, we simulated a scenario in which
the periaxonal space was narrower, by half, at the paranode but
changed proportionally with the internodal periaxonal space
(Figure 6I). In this scenario, the simulated CV using sham param-
eters was 1.32 m/s (Figure 6J) at 21C (37C data in Figure S5H),
but the reduction in CV induced by altering the periaxonal space
width (altered myelin; ~9.2%) or by mimicking the iTBS condition
(~11.5%, Figure 6J; 12.5% at 37CinFigure S5H) was compara-
ble to the case in which periaxonal space width was consistent at
the internode and paranode (Figures 6E and 6F).
Finally, we modeled action potential conduction in the fimbria
using a parameter set matching the experimental data obtained
from NL mice (Figures 6K and 6L; Table S1). The simulated CV
was 0.95 m/s (Figure 6L) at 21C (37C data in Figure S5L), which
is far slower than in CC but is within the range of measured CVs
for this region (Corcoba et al., 2015;Jones et al., 1999).
Increasing node length by ~30%, to reflect the change produced
by RAM learning (Figure 3), increased the simulated CV by 8.9%,
while decreasing the periaxonal space width (Figure 5;Table S1)
increased CV by 7.3%. Again, implementing all changes
measured following learning (Figures 3 and 5;Table S1) had an
additive effect, increasing CV by 16.6% (Figure 6L). At 37C,
this is predicted to correspond to a 21.6% increase in CV
following learning (Figure S5L). These data suggest that adaptive
structural changes at nodes of Ranvier and the periaxonal space
act in concert to slow down or speed up action potential conduc-
tion along an axon.
iTBS and spatial learning have opposing effects on
action potential CV
Because our model predicted that iTBS would slow action po-
tential CV along CC axons, we performed ex vivo field potential
recordings of compound action potentials (CAPs) in CC of
sham and iTBS mice (Figure 7A). The average CV for M axons
in CC of sham-stimulated mice was 1.27 ±0.19 m/s, and this
was reduced by ~18% following iTBS (1.04 ±0.19 m/s) (Fig-
ure 7B). The amplitude of the M axon component of the CAP
also increased by ~40% following iTBS (Figure 7C) and the
half-width decreased by ~9% (Figure 7D), suggesting that a
greater number of action potentials arrived simultaneously at
the recording electrode. The UM axon component was unaltered
by iTBS (Figures 7A–7D). Slowing M axon conduction to this de-
gree did not influence motor coordination, because forelimb
swing time, measured during treadmill running (Figure 7E), and
the average number of foot-slip errors made crossing a ledged
beam (Figure 7F) were unchanged. However, we have previously
shown that delivering iTBS to the motor cortex of adult mice can
subtly enhance fine-motor-skill learning (Tang et al., 2018), sug-
gesting that modulating CV may facilitate fine-motor-skill
acquisition.
To test our in silico prediction that RAM learning would in-
crease CV in the fimbria, we performed ex vivo CAP recordings
at 21C within the fimbria-fornix pathway of NL and L mice (Fig-
ure 7G). Average CV along M axons in the fimbria of NL mice was
8Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
0.91 ±0.3 m/s (Figure 7H) and learning increased CV by ~46%
(1.3 ±0.39 m/s) (Figure 7H). When recordings were made over
a larger distance, it was also possible to measure the CV of a
distinct population of fast, M axons that conduct at 1.4 ±
0.15 m/s in NL mice (n = 5). Learning again increased CV, this
time by ~19% (1.6 ±0.18 m/s; n = 6; t test, t = 2.594, df = 9,
p = 0.02), but it did not alter the peak amplitude (Figure 7I) or
half-width (Figure 7J) of the M axon CAP or affect the CV of
UM axons (Figures 7H–7J).
Because mice have varying levels of aptitude for RAM
learning, we determined how well each mouse performed by
subtracting the number of errors made in the first training session
from the number of errors made in the last training session. We
found that the level of improvement in the RAM task correlated
with the M axon CAP CV in the fimbria-fornix pathway (Figure 7K).
These data suggest that subtle changes to the ultrastructure of
established myelin sheaths may modulate CV to facilitate
learning.
Figure 6. The size of the periaxonal space modulates CV
(A) Action potentials simulated at consecutive callosal nodes at 21C.
(B) Extended time course of action potentials generated at 21C, highlighting a slight depolarizing after-potential (DAP) and a hyperpolarizing after-potential
(HAP).
(C) Simulated CV of callosal axons relative to periaxonal space width (psw)at37
C (magenta) and 21C (gray). The dashed line indicates average psw following
sham stimulation (psw = 6.477 nm; 1.18 m/s at 21C, 1.91 m/s at 37C), and insets show action potential waveforms at the extremities of the tested range (psw =
0 or 20 nm) at 21C and 37C.
(D) Conduction delay over 1 cm relative to psw at 21C and 37C (8.5 ms at 21C and 5.2 ms at 37C for psw = 6.477 nm).
(E) Schematic showing model parameters in which psw at the paranode is uniform to internode psw.
(F) Predicted CV of a sham-stimulated axon (white) versus an axon with the node length shortened, the periaxonal space widened, or both (iTBS) using the model
depicted in (E).
(G) Schematic showing model parameters in which psw at the paranode is set to %3 nm.
(H) Predicted CV of a sham-stimulated axon versus an axon with the node length shortened, the periaxonal space widened, or both (iTBS) using the model
depicted in (G).
(I) Schematic showing model parameters in which psw at the paranode is set to half of the internode psw.
(J) Predicted CV of a sham-stimulated axon versus an axon with the node length shortened, the periaxonal space widened, or both (iTBS) using the model
depicted in (I).
(K) Action potentials simulated at consecutive nodes within the fimbria at 21C.
(L) Predicted CV of a NL control axon within the fimbria (white) versus an axon with the nodes lengthened, the psw narrowed, or both (L).
See also Figure S5 and Table S1.
Cell Reports 34, 108641, January 19, 2021 9
Article
ll
OPEN ACCESS
DISCUSSION
Plasticity of the node of Ranvier
Central myelination is adaptive (for reviews, see Almeida and
Lyons, 2017;Bechler et al., 2018;Mount and Monje, 2017;Pep-
per et al., 2018), and it has been proposed that changing the
pattern of myelination within a circuit could profoundly affect
neural processing (Pajevic et al., 2014). This area of research
has largely focused on the addition of new myelin over time
and in response to altered neuronal activity (Cullen et al., 2019;
Gibson et al., 2014;Li et al., 2010;McKenzie et al., 2014;Mitew
et al., 2018;Sampaio-Baptista et al., 2013;Xiao et al., 2016;
Figure 7. iTBS and spatial learning modu-
late CV
(A) Compound action potential (CAP) recorded in
CC of a sham mouse (black) or iTBS mouse (gray ).
(B) CV of myelinated (M) and unmyelinated (UM)
axons in sham-stimulated mice (white bars, black
circles; n = 7) or iTBS mice (gray bars and tri-
angles; n = 7). 2-way ANOVA: interaction F(1,24) =
3.37, p = 0.078; axon population F(1,24) = 211.1,
p < 0.0001; treatment F(1,24) = 5.81, p = 0.023.
(C) Peak amplitude of M and UM axons. 2-way
ANOVA: interaction F(1,24) = 1.25, p = 0.27; axon
population F(1,24) = 68.10, p < 0.0001; treatment
F(1,24) = 6.75, p = 0.015.
(D) Half-width of CAP peak corresponding to M
and UM axons. 2-way ANOVA: interaction
F(1,24) = 0.13, p = 0.71; axon population F(1,24) =
4.63, p = 0.04; treatment F(1,24) = 10.22, p =
0.003.
(E) Average forelimb swing time during treadmill
running at 28 cm/s for sham-stimulated mice (n =
6) and iTBS mice (n = 6) after 0, 7, and 14 days of
stimulation. Restricted maximum likelihood
(REML) mixed-effect model: interaction F(2,18) =
0.079, p = 0.92; day F(1.37,12.36) = 10.62, p =
0.004; treatment F(1,10) = 1.64, p = 0.22.
(F) Average foot slips made by sham-stimulated
and iTBS mice during a ledged-beam task
following 0, 7, and 14 days of stimulation.
Repeated-measure (RM) 2-way ANOVA: interac-
tion F(2,20) = 0.006, p = 0.99; day F(1.91, 19.14) =
7.45, p = 0.004; treatment F(1,10) = 0.098, p =
0.76.
(G) CAP recorded in the fimbria of mice that un-
derwent NL (black) or L (blue) in the RAM.
(H) CV of M and UM axons in NL mice (white bars,
black circles; n = 6) and L mice (blue bars and
diamonds; n = 7). 2-way ANOVA: interaction
F(1,22) = 4.35, p = 0.048; axon population F(1,22) =
63.74, p < 0.0001; training paradigm F(1,22) =
4.46, p = 0.046.
(I) Peak amplitude of M and UM axons. 2-way
ANOVA: interaction F(1,22) = 0.18, p = 0.67; axon
population F(1,22) = 36.17, p < 0.0001; training
paradigm F(1,22) = 2.72, p = 0.11.
(J) Half-width of M and UM axon CAP peaks. 2-
way ANOVA: interaction F(1,22) = 0.81, p = 0.37;
axon population F(1,22) = 0.24, p = 0.62; training
paradigm F(1,22) = 0.03, p = 0.85.
(K) Reduction in RAM errors made by individual
mice (n = 7) from the first training trial to the last
training trial versus measured CV of M axons.
Simple linear regression: F(1,5) = 8.33, p = 0.032,
R
2
= 0.6251, equation Y = 22.03*X 0.52.
Bars show mean ±SD. Line graphs show mean ±
SEM (E and F). The linear regression graph (K)
shows the line of best fit (dashed line) and 95%
confidence bands (dotted lines). *p < 0.05 by
Bonferroni’s post-test.
10 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
Young et al., 2013), but it is equally important to know whether
existing OLs adapt to altered neuronal activity and whether this
affects action potential propagation.
Once formed, mature myelinating OLs are long lived (Tripathi
et al., 2017), and most internodes they support are stable over
time. However, these mature cells retain some capacity to adjust
internode length (Auer et al., 2018;Hill et al., 2018). We report
that neither iTBS nor spatial learning induces the overt extension
or retraction of mature internodes (Figures 1 and 3), which is
consistent with sensory enrichment or deprivation failing to alter
the existing myelin sheath length within the somatosensory cor-
tex of adult mice (Hughes et al., 2018). However, both iTBS and
spatial learning produce a marked change in node length (Fig-
ures 2,3, and 4). Pathological node lengthening (Hinman et al.,
2015;Howell et al., 2006;O’Hare Doig et al., 2017;Reimer
et al., 2011) is associated with myelin loss (Howell et al., 2006)
or paranodal pathology (Huff et al., 2011;Reimer et al., 2011).
We found no evidence of myelin pathology following iTBS or
learning, and the adaptive changes in node length occurred
without a change in paranode length, suggesting that activity-
induced nodal plasticity involves a subtle expansion or contrac-
tion of the existing myelin internodes without an accompanying
change in the number of myelin cytoplasmic loops. Previous
studies have suggested that node size could be physiologically
modulated as node diameter increases along gerbil globular bu-
shy cell axons, as they approach the calyx of Held (Ford et al.,
2015), and node length varies more significantly between CC
axons in the rat brain than it does along the length of individual
axons (Arancibia-Ca
´rcamo et al., 2017). Although further
research is required to uncover the precise mechanics that allow
the node of Ranvier to be shortened and lengthened, our data
suggest that neuronal activity is a primary driver of nodal plas-
ticity in the adult mouse brain.
Plasticity of the periaxonal space
The direct optogenetic stimulation of layer V pyramidal neu-
rons in the premotor cortex (Gibson et al., 2014), the pharma-
cogenetic stimulation of layer 2/3 pyramidal neurons in the pri-
mary somatosensory cortex (Mitew et al., 2018), and the
optogenetic inhibition of parvalbumin
+
interneurons in the
anterior cingulate cortex (Piscopo et al., 2018) decrease the
g-ratio of axons within the underlying CC. iTBS also reduces
the g-ratio of CC axons; however, this is not the result of
increased myelin thickness or decreased axon diameter but
is instead associated with a marked increase in the periaxonal
space width (Figure 5). Conversely, spatial learning instead in-
creases the g-ratio of axons within the hippocampal fimbria,
but this is also associated with a change (decrease) in the
size of the periaxonal space (Figure 5).
The periaxonal space is diminished in mice deficient in myelin-
associated glycoprotein (MAG) (Georgiou et al., 2004;Li et al.,
1994;Quarles, 2007), a protein expressed specifically in the peri-
axonal myelin membrane (Pronker et al., 2016), and vacuoles
form within this space along optic nerve axons following con-
nexin 32/47 knockout (Menichella et al., 2006). A marked patho-
logical swelling of the periaxonal space has also been observed
following acute ischemic white matter injury and demyelination
(Aboul-Enein et al., 2003). We did not observe vacuolization, or
an uneven separation of the myelin from the axons, suggesting
that iTBS- or learning-induced modulation of the periaxonal
space is not pathological but rather a physiological adaptation
to altered neuronal activity. Although the mechanism underlying
this physiological plasticity remains unknown, it may involve
MAG or connexin 32/47, be initiated by neurotransmitter recep-
tor activation on the myelin sheath, or result from a change in the
movement of ions and metabolites into the periaxonal space in
response to neuronal activity (reviewed in Micu et al., 2018).
Nodal and periaxonal plasticities modulate CV
The periaxonal space has been hypothesized to act as a recep-
tacle for calcium, glutamate, and potassium released by the
axon during action potential propagation, as well as OL-derived
lactate and pyruvate, which are then shuttled through monocar-
boxylate transporters to provide metabolic support to the axon
(reviewed in Micu et al., 2018). It is also an important regulatory
element for action potential conduction (Cohen et al., 2020). The
periaxonal space is typically narrower at the paranode than un-
der the internode (Nans et al., 2011;Rosenbluth, 1995;Waxman
et al., 1995), and the capacity for ultrastructural adaptation of the
periaxonal space to influence CV likely depends on its size at
both locations. We were unable to measure periaxonal space
width at the paranode but incorporated three possible paranodal
scenarios into our modeling: (1) the periaxonal space width is
uniform under the internode and paranode and changes at
both locations, (2) the periaxonal space is narrower under the
paranode and does not exceed 3 nm, or (3) the periaxonal space
is narrower at the paranode but changes proportionally with the
internodal periaxonal space (Figure 6;Figure S5). In all scenarios,
adjusting the node length and the internodal periaxonal space
width to match the experimental values obtained following
iTBS consistently slowed the simulated CVs compared with
the sham condition (Figure 6;Figure S5), although the effect
was not as strong when the paranodal periaxonal space width
remained constant.
When we performed computational modeling to assess the
impact that each ultrastructural adaptation measured could
have on CV, we found that the iTBS-induced decrease in node
length was predicted to slow CV, whereas the learning-induced
increase in node length was predicted to increase CV (Figure 6;
Figure S5). These data are consistent with a theoretical inverse U
relationship between CV and node length in cortical axons, in
which nodal sodium channel density has been shown to remain
relatively constant across nodes of differing lengths (Arancibia-
Ca
´rcamo et al., 2017). In the context of iTBS, our computational
modeling suggests that expanding the periaxonal space has a
larger effect on action potential slowing than shortening the
node length has; however, following spatial learning, increasing
node length and decreasing the periaxonal space width sped
up action potential conduction to a similar extent. From the
data presented, it is not possible to deduce how or even if a
change in node length is mechanistically coupled with a change
in periaxonal space width. However, our observation that iTBS
and RAM learning modified the length of the node of Ranvier
and the width of the periaxonal space in a way that ensured a
synergistic effect, rather than an opposing effect, on CV makes
this an intriguing possibility.
Cell Reports 34, 108641, January 19, 2021 11
Article
ll
OPEN ACCESS
Regulating CV is important for network function
The dynamic modulation of CV could be an adaptive mechanism
to ensure the coincident arrival of action potentials at postsyn-
aptic targets. Indeed, when we performed ex vivo CAP record-
ings in CC, we found that iTBS not only slowed CAP CV but
also increased the peak amplitude of the CAP response and
decreased the half-width, indicating a more synchronized arrival
of action potentials at the recording electrode (Figure 7). A larger
periaxonal space likely increases the current flowing out of the
axon at the internodes and eventually out from underneath the
myelin sheath at the node of Ranvier, possibly increasing
coupling between neighboring nodes. In other contexts,
increased coupling can reduce action potential CV and promote
action potential synchronization (Katz and Schmitt, 1940;
Schmidt and Kno
¨sche, 2019). By contrast, RAM learning
increased M axon CAP CV in the fimbria-fornix pathway without
altering the peak amplitude or half-width of the CAP response,
suggesting that spatial learning acted to increase CV without
influencing relative transit times.
At the network level, adjusting action potential arrival times at
postsynaptic neurons could influence the probability of postsyn-
aptic firing and determine whether potentiation or depression is
induced (Feldman, 2012;Markram et al., 2012). Recent evidence
suggests layer V pyramidal and CA3 neurons are highly sensitive
to the degree of input synchrony, with asynchronous inputs (even
5 ms apart) producing an action potential of reduced amplitude
and reducing the level of spontaneous activity subsequently
measured (Zbili et al., 2020). Both theoretical and experimental
data have shown that the precise regulation of action potential
CV and the firing frequency of neuronal populations is also impor-
tant for synchronous or time-locked brain wave oscillations (Kato
et al., 2020;Noori et al., 2020;Pajevic et al., 2014;Steadman
et al., 2020), which are associated with various cognitive func-
tions, including selective attention, information processing, sen-
sory gating of information, learning, memory formation, and con-
sciousness (Ainsworth et al., 2012;Burgess et al., 2007;Buzsaki,
2006;Sirota et al., 2008). Slowing CV by applying iTBS did not
lead to overt changes in motor coordination (Figure 7), but it is
interesting to speculate that it may facilitate the iTBS-mediated
improvement in fine-motor-skill acquisition (Tang et al., 2018).
Following RAM learning, we found that the CV of M axons in the
fimbria-fornix pathway of individual mice associated with their
degree of improvement (learning) (Figure 7), suggesting that
adaptive changes in CV may facilitate learning. During spatial
learning, oligodendrogenesis increases the coordinated coupling
of activity in the hippocampus (CA1) and medial prefrontal cortex
to enable memory consolidation (Steadman et al., 2020). Nodal
and periaxonal plasticities may be vital homeostatic mechanisms
that work in concert with de novo myelination to refine action po-
tential CV in response to altered neural circuit activity and in turn
influence information coding in the CNS.
STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
dKEY RESOURCES TABLE
dRESOURCE AVAILABILITY
BLead contact
BMaterials availability
BData and code availability
dEXPERIMENTAL MODEL AND SUBJECT DETAILS
dMETHOD DETAILS
BTransgenic lineage tracing and gene deletion
BLow intensity repetitive transcranial magnetic stimula-
tion
BDigiGait
TM
gait analysis
BLedge beam task
BSpatial learning
BTissue preparation and immunohistochemistry
BConfocal microscopy and image quantification
BStimulated emission depletion (STED) microscopy
BTransmission electron microscopy
BConduction velocity modeling
BCompound action potential recordings
dQUANTIFICATION AND STATISTICAL ANALYSIS
SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at https://doi.org/10.1016/j.
celrep.2020.108641.
ACKNOWLEDGMENTS
We thank Dr. Rowan Tweedale (the University of Queensland) for constructive
feedback on the manuscript. We thank Dr. Lee Cossell and Prof. David Attwell
(University College London) for advice on computational modeling and Dr.
Carola Thoni and Shane Rix (Lastek: Photonics Technology Solutions,
Australia) for assistance with STED imaging. This research was supported
by grants from the National Health and Medical Research Council of Australia
(NHMRC) (1077792 and 1139041), MS Research Australia (11-014, 16-105,
and 17-007), the Australian Research Council (DP180101494), the Medical
Research Future Fund (EPCD08), the Swiss National Science Foundation
(31003A_170079), and the National MS Society. Fellowships were awarded
to C.L.C. (MS Research Australia and the Penn Foundation, 15-054), K.A.P.
(NHMRC, 1139180), B.E. (NHMRC, 1032833; Warren endowed professorship
in neuroscience research), K.M.Y. (NHMRC, 1045240; MS Research Australia/
Macquarie Group Foundation, 17-0223), and J.R. (NHMRC, 1002258; the
Perron Institute for Neurological and Translational Science; and MS Western
Australia). Scholarships were awarded to L.A. (Australian Postgraduate Award)
and to M.T.C. and R.E.P. (Menzies Institute for Medical Research, University of
Tasmania).
AUTHOR CONTRIBUTIONS
C.L.C., K.M.Y., R.B.J., R.E.P., K.A.P., J.R., and B.E. developed the project and
wrote the manuscript. C.L.C., R.E.P., K.A.P., M.T.C., L.A., V.O., R.B.J., and
K.M.Y. carried out the experiments. K.M.Y., J.R., C.L.C., B.E., and R.B.J. ob-
tained the funding. C.L.C., R.E.P., M.T.C., and R.B.J. performed the statistical
analyses and generated the figures. K.M.Y., A.D.T., G.R., J.R., and C.L.C. pro-
vided supervision.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: May 22, 2020
Revised: November 18, 2020
Accepted: December 21, 2020
Published: January 19, 2021
12 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
REFERENCES
Aboul-Enein, F., Rauschka, H., Kornek, B., Stadelmann, C., Stefferl, A., Br
uck,
W., Lucchinetti, C., Schmidbauer, M., Jellinger, K., and Lassmann, H. (2003).
Preferential loss of myelin-associated glycoprotein reflects hypoxia-like white
matter damage in stroke and inflammatory brain diseases. J. Neuropathol.
Exp. Neurol. 62, 25–33.
Ainsworth, M., Lee, S., Cunningham, M.O., Traub, R.D., Kopell, N.J., and Whit-
tington, M.A. (2012). Rates and rhythms: a synergistic view of frequency and
temporal coding in neuronal networks. Neuron 75, 572–583.
Almeida, R.G., and Lyons, D.A. (2017). On myelinated axon plasticity and
neuronal circuit formation and function. J. Neurosci. 37, 10023–10034.
Arancibia-Ca
´rcamo, I.L., Ford, M.C., Cossell, L., Ishida, K., Tohyama, K., and
Attwell, D. (2017). Node of Ranvier length as a potential regulator of myelinated
axon conduction speed. eLife 6, e23329.
Auer, F., Vagionitis, S., and Czopka, T. (2018). Evidence for myelin sheath re-
modeling in the CNS revealed by in vivo imaging. Curr. Biol. 28, 549–559.e3.
Bakiri, Y., Ka
´rado
´ttir, R., Cossell, L., and Attwell, D. (2011). Morphological and
electrical properties of oligodendrocytes in the white matter of the corpus cal-
losum and cerebellum. J. Physiol. 589, 559–573.
Baraban, M., Koudelka, S., and Lyons, D.A. (2018). Ca
2+
activity signatures of
myelin sheath formation and growth in vivo. Nat. Neurosci. 21, 19–23.
Battefeld, A., Popovic, M.A., de Vries, S.I., and Kole, M.H.P. (2019). High-fre-
quency microdomain Ca2+ transients and waves during early myelin internode
remodeling. Cell Rep. 26, 182–191.e5.
Bechler, M.E., Swire, M., and Ffrench-Constant, C. (2018). Intrinsic and adap-
tive myelination—A sequential mechanism for smart wiring in the brain. Dev.
Neurobiol. 78, 68–79.
Benchenane, K., Peyrache, A., Khamassi, M., Tierney, P.L., Gioanni, Y., Batta-
glia, F.P., and Wiener, S.I. (2010). Coherent theta oscillations and reorganiza-
tion of spike timing in the hippocampal-prefrontal network upon learning.
Neuron 66, 921–936.
Bhat, M.A., Rios, J.C., Lu, Y., Garcia-Fresco, G.P., Ching, W., St Martin, M., Li,
J., Einheber, S., Chesler, M., Rosenbluth, J., et al. (2001). Axon-glia interac-
tions and the domain organization of myelinated axons requires neurexin IV/
Caspr/Paranodin. Neuron 30, 369–383.
Burgess, N., Barry, C., and O’Keefe, J. (2007). An oscillatory interference
model of grid cell firing. Hippocampus 17, 801–812.
Buzsaki, G. (2006). Rhythms of the brain (Oxford University Press).
Charles, P., Tait, S., Faivre-Sarrailh, C., Barbin, G., Gunn-Moore, F., Deni-
senko-Nehrbass, N., Guennoc, A.-M., Girault, J.-A., Brophy, P.J., and Lu-
betzki, C. (2002). Neurofascin is a glial receptor for the paranodin/Caspr-con-
tactin axonal complex at the axoglial junction. Curr. Biol. 12, 217–220.
Cohen, C.C.H., Popovic, M.A., Klooster, J., Weil, M.-T., Mo
¨bius, W., Nave, K.-
A., and Kole, M.H.P. (2020). Saltatory conduction along myelinated axons in-
volves a periaxonal nanocircuit. Cell 180, 311–322.e15.
Corcoba, A., Steullet, P., Duarte, J.M.N., Van de Looij, Y., Monin, A., Cuenod,
M., Gruetter, R., and Do, K.Q. (2015). Glutathione deficit affects the integrity
and function of the fimbria/fornix and anterior commissure in mice: Relevance
for schizophrenia. Int. J. Neuropsychopharmacol. 19, pyv110.
Crawford, D.K., Mangiardi, M., and Tiwari-Woodruff, S.K. (2009). Assaying the
functional effects of demyelination and remyelination: revisiting field potent ial
recordings. J. Neurosci. Methods 182, 25–33.
Cullen, C.L., Senesi, M., Tang, A.D., Clutterbuck, M.T., Auderset , L., O’Rourke,
M.E., Rodger, J., and Young, K.M. (2019). Low-intensity transcranial magnetic
stimulation promotes the survival and maturation of newborn oligodendro-
cytes in the adult mouse brain. Glia 67, 1462–1477.
Dimou, L., Simon, C., Kirchhoff, F., Takebayashi, H., and Go
¨tz, M. (2008).
Progeny of Olig2-expressing progenitors in the gray and white matter of the
adult mouse cerebral cortex. J. Neurosci. 28, 10434–10442.
Doerflinger, N.H., Macklin, W.B., and Popko, B. (2003). Inducible site-specific
recombination in myelinating cells. Genesis 35, 63–72.
Dupret, D., O’Neill, J., Pleydell-Bouverie, B., and Csicsvari, J. (2010). The reor-
ganization and reactivation of hippocampal maps predict spatial memory per-
formance. Nat. Neurosci. 13, 995–1002.
Dupret, D., O’Neill, J., and Csicsvari, J. (2013). Dynamic reconfiguration of hip-
pocampal interneuron circuits during spatial learning. Neuron 78, 166–180.
Dutta, D.J., Woo, D.H., Lee, P.R., Pajevic, S., Bukalo, O., Huffman, W.C.,
Wake, H., Basser, P.J., SheikhBahaei, S., Lazarevic, V., et al. (2018). Regula-
tion of myelin structure and conduction velocity by perinodal astrocytes. Proc.
Natl. Acad. Sci. USA 115, 11832–11837.
Emery, B., Agalliu, D., Cahoy, J.D., Watkins, T.A., Dugas, J.C., Mulinyawe,
S.B., Ibrahim, A., Ligon, K.L., Rowitch, D.H., and Barres, B.A. (2009). Myelin
gene regulatory factor is a critical transcriptional regulator required for CNS
myelination. Cell 138, 172–185.
Etxeberria, A., Hokanson, K.C., Dao, D.Q., Mayoral, S.R., Mei, F., Redmond,
S.A., Ullian, E.M., and Chan, J.R. (2016). Dynamic modulation of myelination
in response to visual stimuli alters optic nerve conduction velocity.
J. Neurosci. 36, 6937–6948.
Feldman, D.E. (2012). The spike-timing dependence of plasticity. Neuron 75,
556–571.
Ford, M.C., Alexandrova, O., Cossell, L., Stange-Marten, A., Sinclair, J., Kopp-
Scheinpflug, C., Pecka, M., Attwell, D., and Grothe, B. (2015). Tuning of Ranv-
ier node and internode properties in myelinated axons to adjust action poten-
tial timing. Nat. Commun. 6, 8073.
Freeman, S.A., Desmazie
`res, A., Simonnet, J., Gatta, M., Pfeiffer, F., Aigrot,
M.S., Rappeneau, Q., Guerreiro, S., Michel, P.P., Yanagawa, Y., et al.
(2015). Acceleration of conduction velocity linked to clustering of nodal com-
ponents precedes myelination. Proc. Natl. Acad. Sci. USA 112, E321–E328.
Freeman, S.A., Desmazie
`res, A., Fricker, D., Lubetzki, C., and Sol-Foulon, N.
(2016). Mechanisms of sodium channel clustering and its influence on axonal
impulse conduction. Cell. Mol. Life Sci. 73, 723–735.
Georgiou, J., Tropak, M.B., and Roder, J.C. (2004). Myelin-associated glyco-
protein gene. In Myelin biology and disorders, R.A. Lazzarini, J.W. Griffin, H.
Lassman, K.-A. Nave, R. Miller, and B.D. Trapp, eds. (Elsevier), pp. 421–467.
Gibson, E.M., Purger, D., Mount, C.W., Goldstein, A.K., Lin, G.L., Wood, L.S.,
Inema, I., Miller, S.E., Bieri, G., Zuchero, J.B., et al. (2014). Neuronal activity
promotes oligodendrogenesis and adaptive myelination in the mammalian
brain. Science 344, 1252304.
Halter, J.A., and Clark, J.W., Jr. (1993). The influence of nodal constriction on
conduction velocity in myelinated nerve fibers. Neuroreport 4, 89–92.
Hill, R.A., Li, A.M., and Grutzendler, J. (2018). Lifelong cortical myelin plasticity
and age-related degeneration in the live mammalian brain. Nat. Neurosci. 21,
683–695.
Hinman, J.D., Lee, M.D., Tung, S., Vinters, H.V., and Carmichael, S.T. (2015).
Molecular disorganization of axons adjacent to human lacunar infarcts. Brain
138, 736–745.
Hippenmeyer, S., Vrieseling, E., Sigrist, M., Portmann, T., Laengle, C., Ladle,
D.R., and Arber, S. (2005). A developmental switch in the response of DRG
neurons to ETS transcription factor signaling. PLoS Biol. 3, e159.
Howell, O.W., Palser, A., Polito, A., Melrose, S., Zonta, B., Scheiermann, C.,
Vora, A.J., Brophy, P.J., and Reynolds, R. (2006). Disruption of neurofascin
localization reveals early changes preceding demyelination and remyelination
in multiple sclerosis. Brain 129, 3173–3185.
Huff, T.B., Shi, Y., Sun, W., Wu, W., Shi, R., and Cheng, J.-X. (2011). Real-time
CARS imaging reveals a calpain-dependent pathway for paranodal myelin
retraction during high-frequency stimulation. PLoS ONE 6, e17176.
Hughes, E.G., Orthmann-Murphy, J.L., Langseth, A.J., and Bergles, D.E.
(2018). Myelin remodeling through experience-dependent oligodendrogenesis
in the adult somatosensory cortex. Nat. Neurosci. 21, 696–706.
Jakovcevski, I., Filipovic, R., Mo, Z., Rakic, S., and Zecevic, N. (2009). Oligo-
dendrocyte development and the onset of myelination in the human fetal brain.
Front. Neuroanat. 3,5.
Jeffries, M.A., Urbanek, K., Torres, L., Wendell, S.G., Rubio, M.E., and Fyffe-
Maricich, S.L. (2016). Erk1/2 activation in preexisting oligodendrocytes of adult
Cell Reports 34, 108641, January 19, 2021 13
Article
ll
OPEN ACCESS
mice drives new myelin synthesis and enhanced CNS function. J. Neurosci.
36, 9186–9200.
Jin, J., and Maren, S. (2015). Prefrontal-hippocampal interactions in memory
and emotion. Front. Syst. Neurosci. 9, 170.
Jones, G.A., Norris, S.K., and Henderson, Z. (1999). Conduction velocities and
membrane properties of different classes of rat septohippocampal neurons re-
corded in vitro. J. Physiol. 517, 867–877.
Kang, S.H., Fukaya, M., Yang, J.K., Rothstein, J.D., and Bergles, D.E. (2010).
NG2+ CNS glial progenitors remain committed to the oligodend rocyte lineage
in postnatal life and following neurodegeneration. Neuron 68, 668–681.
Kato, D., Wake, H., Lee, P.R., Tachibana, Y., Ono, R., Sugio, S., Tsuji, Y., Ta-
naka, Y.H., Tanaka, Y.R., Masamizu, Y., et al. (2020). Motor learning requires
myelination to reduce asynchrony and spontaneity in neural activity. Glia 68,
193–210.
Katz, B., and Schmitt, O.H. (1940). Electric interaction between two adjacent
nerve fibres. J. Physiol. 97, 471–488.
Kessaris, N., Fogarty, M., Iannarelli, P., Grist, M., Wegner, M., and Richardson,
W.D. (2006). Competing waves of oligodendrocytes in the forebrain and post-
natal elimination of an embryonic lineage. Nat. Neurosci. 9, 173–179.
Klingseisen, A., Ristoiu, A.-M., Kegel, L., Sherman, D.L., Rubio-Brotons, M.,
Almeida, R.G., Koudelka, S., Benito-Kwiecinski, S.K., Poole, R.J., Brophy,
P.J., and Lyons, D.A. (2019). Oligodendrocyte neurofascin independently reg-
ulates both myelin targeting and sheath growth in the CNS. Dev. Cell 51, 730–
744.e6.
Koenning, M., Jackson, S., Hay, C.M., Faux, C., Kilpatrick, T.J., Willingham,
M., and Emery, B. (2012). Myelin gene regulatory factor is required for mainte-
nance of myelin and mature oligodendrocyte identity in the adult CNS.
J. Neurosci. 32, 12528–12542.
Krasnow, A.M., Ford, M.C., Valdivia, L.E., Wilson, S.W., and Attw ell, D. (2018).
Regulation of developing myelin sheath elongation by oligodendrocyte cal-
cium transients in vivo. Nat. Neurosci. 21, 24–28.
Li, C., Tropak, M.B., Gerlai, R., Clapoff, S., Abramow-Newerly, W., Trapp, B.,
Peterson, A., and Roder, J. (1994). Myelination in the absence of myelin-asso-
ciated glycoprotein. Nature 369, 747–750.
Li, Q., Brus-Ramer, M., Martin, J.H., and McDonald, J.W. (2010). Electrical
stimulation of the medullary pyramid promotes proliferation and differentiation
of oligodendrocyte progenitor cells in the corticospinal tract of the adult rat.
Neurosci. Lett. 479, 128–133.
Liu, J., Dietz, K., DeLoyht, J.M., Pedre, X., Kelkar, D., Kaur, J., Vialou, V., Lobo,
M.K., Dietz, D.M., Nestler, E.J., et al. (2012). Impaired adult myelination in the
prefrontal cortex of socially isolated mice. Nat. Neurosci. 15, 1621–1623.
Lu, Q.R., Sun, T., Zhu, Z., Ma, N., Garcia, M., Stiles, C.D., and Rowitch, D.H.
(2002). Common developmental requirement for Olig function indicates a mo-
tor neuron/oligodendrocyte connection. Cell 109, 75–86.
Makinodan, M., Rosen, K.M., Ito, S., and Corfas, G. (2012). A critical period for
social experience-dependent oligodendrocyte maturation and myelination.
Science 337, 1357–1360.
Markram, H., Gerstner, W., and Sjo
¨stro
¨m, P.J. (2012). Spike-timing-dependent
plasticity: a comprehensive overview. Front. Synaptic Neurosci. 4,2.
McIntyre, C.C., and Grill, W.M. (2002). Extracellular stimulation of central neu-
rons: influence of stimulus waveform and frequency on neuronal output.
J. Neurophysiol. 88, 1592–1604.
McKenzie, I.A., Ohayon, D., Li, H., de Faria, J.P., Emery, B., Tohyama, K., and
Richardson, W.D. (2014). Motor skill learning requires active central myelina-
tion. Science 346, 318–322.
Menichella, D.M., Majdan, M., Awatramani, R., Goodenough, D.A., Sirkowski,
E., Scherer, S.S., and Paul, D.L. (2006). Genetic and physiological evidence
that oligodendrocyte gap junctions contribute to spatial buffering of potassium
released during neuronal activity. J. Neurosci. 26, 10984–10991.
Micu, I., Plemel, J.R., Caprariello, A.V., Nave, K.-A., and Stys, P.K. (2018). Axo-
myelinic neurotransmission: a novel mode of cell signalling in the central ner-
vous system. Nat. Rev. Neurosci. 19, 49–58.
Mitew, S., Gobius, I., Fenlon, L.R., McDougall, S.J., Hawkes, D., Xing, Y.L., Bu-
jalka, H., Gundlach, A.L., Richards, L.J., Kilpatrick, T.J., et al. (2018). Pharma-
cogenetic stimulation of neuronal activity increases myelination in an axon-
specific manner. Nat. Commun. 9, 306.
Mount, C.W., and Monje, M. (2017). Wrapped to adapt: Experience-dependent
myelination. Neuron 95, 743–756.
Nans, A., Einheber, S., Salzer, J.L., and Stokes, D.L. (2011). Electron tomogra-
phy of paranodal septate-like junctions and the associated axonal and glial cy-
toskeletons in the central nervous system. J. Neurosci. Res. 89, 310–319.
Negro
´n-Oyarzo, I., Espinosa, N., Aguilar-Rivera, M., Fuenzalida, M., Aboitiz,
F., and Fuentealba, P. (2018). Coordinated prefrontal-hippocampal activity
and navigation strategy-related prefrontal firing during spatial memory forma-
tion. Proc. Natl. Acad. Sci. USA 115, 7123–7128.
Noori, R., Park, D., Griffiths, J.D., Bells, S., Frankland, P.W., Mabbott, D., and
Lefebvre, J. (2020). Activity-dependent myelination: A glial mechanism of
oscillatory self-organization in large-scale brain networks. Proc. Natl. Acad.
Sci. USA 117, 13227–13237.
O’Hare Doig, R.L., Chiha, W., Giacci, M.K., Yates, N.J., Bartlett, C.A., Smith,
N.M., Hodgetts, S.I., Harvey, A.R., and Fitzgerald, M. (2017). Specific ion chan-
nels contribute to key elements of pathology during secondary degeneration
following neurotrauma. BMC Neurosci. 18,62.
Osanai, Y., Shimizu, T., Mori, T., Hatanaka, N., Kimori, Y., Kobayashi, K.,
Koyama, S., Yoshimura, Y., Nambu, A., and Ikenaka, K. (2018). Length of
myelin internodes of individual oligodendrocytes is controlled by microenvi-
ronment influenced by normal and input-deprived axonal activities in sensory
deprived mouse models. Glia 66, 2514–2525.
Pajevic, S., Basser, P.J., and Fields, R.D. (2014). Role of myelin plasticity in os-
cillations and synchrony of neuronal activity. Neuroscience 276, 135–147.
Peles, E., Nativ, M., Lustig, M., Grumet, M., Schilling, J., Martinez, R.,
Plowman, G.D., and Schlessinger, J. (1997). Identification of a novel contac-
tin-associated transmembrane receptor with multiple domains implicated in
protein-protein interactions. EMBO J. 16, 978–988.
Pepper, R.E., Pitman, K.A., Cullen, C.L., and Young, K.M. (2018). How do cells
of the oligodendrocyte lineage affect neuronal circuits to influence motor func-
tion, memory and mood? Front. Cell. Neurosci. 12, 399.
Piscopo, D.M., Weible, A.P., Rothbart, M.K., Posner, M.I., and Niell, C.M.
(2018). Changes in white matter in mice resulting from low-frequency brain
stimulation. Proc. Natl. Acad. Sci. USA 115, E6339–E6346.
Pronker, M.F., Lemstra, S., Snijder, J., Heck, A.J.R., Thies-Weesie, D.M.E.,
Pasterkamp, R.J., and Janssen, B.J.C. (2016). Structural basis of myelin-asso-
ciated glycoprotein adhesion and signalling. Nat. Commun. 7, 13584.
Quarles, R.H. (2007). Myelin-associated glycoprotein (MAG): past, present and
beyond. J. Neurochem. 100, 1431–1448.
Reimer, M.M., McQueen, J., Searcy, L., Scullion, G., Zonta, B., Desmazieres,
A., Holland, P.R., Smith, J., Gliddon, C., Wood, E.R., et al. (2011). Rapid
disruption of axon-glial integrity in response to mild cerebral hypoperfusion.
J. Neurosci. 31, 18185–18194.
Richardson, A.G., McIntyre, C.C., and Grill, W.M. (2000). Modelling the effects
of electric fields on nerve fibres: influence of the myelin sheath. Med. Biol. Eng.
Comput. 38, 438–446.
Rivers, L.E., Young, K.M., Rizzi, M., Jamen, F., Psachoulia, K., Wade, A., Kes-
saris, N., and Richardson, W.D. (2008). PDGFRA/NG2 glia generate myelinat-
ing oligodendrocytes and piriform projection neurons in adult mice. Nat. Neu-
rosci. 11, 1392–1401.
Rosenbluth, J. (1995). Pathology of demyelinated and dysmyelinated axons . In
The Axon Structure, Function and Pathophysiology, S.G. Waxman, J.D. Koc-
sis, and P.K. Stys, eds. (Oxford University Press).
Sampaio-Baptista, C., Khrapitchev, A.A., Foxley, S., Schlagheck, T., Scholz,
J., Jbabdi, S., DeLuca, G.C., Miller, K.L., Taylor, A., Thomas, N., et al.
(2013). Motor skill learning induces changes in white matter microstructure
and myelination. J. Neurosci. 33, 19499–19503.
Schmidt, H., and Kno
¨sche, T.R. (2019). Action potential propagation and syn-
chronisation in myelinated axons. PLoS Comput. Biol. 15, e1007004.
14 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
Schneider, S., Gruart, A., Grade, S., Zhang, Y., Kro
¨ger, S., Kirchhoff, F., Ei-
chele, G., Delgado Garcı
´a, J.M., and Dimou, L. (2016). Decrease in newly
generated oligodendrocytes leads to motor dysfunctions and changed myelin
structures that can be rescued by transplanted cells. Glia 64, 2201–2218.
Seidl, A.H. (2014). Regulation of conduction time along axons. Neuroscience
276, 126–134.
Sherman, D.L., Tait, S., Melrose, S., Johnson, R., Zonta, B., Court, F.A.,
Macklin, W.B., Meek, S., Smith, A.J.H., Cottrell, D.F., and Brophy, P.J.
(2005). Neurofascins are required to establish axonal domains for saltatory
conduction. Neuron 48, 737–742.
Simons, M., and Nave, K.A. (2015). Oligodendrocytes: Myelination and axonal
support. Cold Spring Harb. Perspect. Biol. 8, a020479.
Sirota, A., Montgomery, S., Fujisawa, S., Isomura, Y., Zugaro, M., and Buzsa
´ki,
G. (2008). Entrainment of neocortical neurons and gamma oscillations by the
hippocampal theta rhythm. Neuron 60, 683–697.
Srinivas, S., Watanabe, T., Lin, C.-S., William, C.M., Tanabe, Y., Jessell, T.M.,
and Costantini, F. (2001). Cre reporter strains produced by targeted insertion
of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1,4.
Steadman, P.E., Xia, F., Ahmed, M., Mocle, A.J., Penning, A.R.A., Geraghty,
A.C., Steenland, H.W., Monje, M., Josselyn, S.A., and Frankland, P.W.
(2020). Disruption of oligodendrogenesis impairs memory consolidation in
adult mice. Neuron 105, 150–164.e6.
Suzuki, A., Hoshi, T., Ishibashi, T., Hayashi, A., Yamaguchi, Y., and Baba, H.
(2004). Paranodal axoglial junction is required for the maintenance of the
Nav1.6-type sodium channel in the node of Ranvier in the optic nerves but
not in peripheral nerve fibers in the sulfatide-deficient mice. Glia 46, 274–283.
Tang, A.D., Hong, I., Boddington, L.J., Garrett, A.R., Etherington, S., Reynolds,
J.N., and Rodger, J. (2016a). Low-intensity repetitive magnetic stimulation
lowers action potential threshold and increases spike firing in layer 5 pyramidal
neurons in vitro. Neuroscience 335, 64–71.
Tang, A.D., Lowe, A.S., Garrett, A.R., Woodward, R., Bennett, W., Canty, A.J.,
Garry, M.I., Hinder, M.R., Summers, J.J., Gersner, R., et al. (2016b). Construc-
tion and evaluation of rodent-specific rtms coils. Front. Neural Circuits 10,47.
Tang, A.D., Bennett, W., Hadrill, C., Collins, J., Fulopova, B., Wills, K., Bindoff,
A., Puri, R., Garry, M.I., Hinder, M.R., et al. (2018). Low intensity repetitive
transcranial magnetic stimulation modulates skilled motor learning in adult
mice. Sci. Rep. 8, 4016.
Tripathi, R.B., Jackiewicz, M., McKenzie, I.A., Kougioumtzidou, E., Grist, M.,
and Richardson, W.D. (2017). Remarkable stability of myelinating oligodendro-
cytes in mice. Cell Rep. 21, 316–323.
Waxman, S.G., Kocsis, J.D., and Stys, P.K. (1995). The axon: Structure, func-
tion and pathophysiology (Oxford University Press).
Wyss, J.M., Swanson, L.W., and Cowan, W.M. (1980). The organization of the
fimbria, dorsal fornix and ventral hippocampal commissure in the rat. Anat.
Embryol. (Berl.) 158, 303–316.
Xiao, L., Ohayon, D., McKenzie, I.A., Sinclair-Wilson, A., Wright, J.L., Fudge,
A.D., Emery, B., Li, H., and Richardson, W.D. (2016). Rapid production of
new oligodendrocytes is required in the earliest stages of motor-skill learning.
Nat. Neurosci. 19, 1210–1217.
Yeung, M.S., Zdunek, S., Bergmann, O., Bernard, S., Salehpour, M., Alkass,
K., Perl, S., Tisdale, J., Possnert, G., Brundin, L., et al. (2014). Dynamics of
oligodendrocyte generation and myelination in the human brain. Cell 159,
766–774.
Young, K.M., Psachoulia, K., Tripathi, R.B., Dunn, S.-J., Cossell, L., Attwell, D.,
Tohyama, K., and Richardson, W.D. (2013). Oligodendrocyte dynamics in the
healthy adult CNS: evidence for myelin remodeling. Neuron 77, 873–885.
Zbili, M., Rama, S., Yger, P., Inglebert, Y., Boumedine-Guignon, N., Fronzaroli-
Moliniere, L., Brette, R., Russier, M., and Debanne, D. (2020). Axonal Na
+
channels detect and transmit levels of input synchrony in local brain circuits.
Sci. Adv. 6, eaay4313.
Cell Reports 34, 108641, January 19, 2021 15
Article
ll
OPEN ACCESS
STAR+METHODS
KEY RESOURCES TABLE
REAGENT or RESOURCE SOURCE IDENTIFIER
Antibodies
Rat (IgG2a) monoclonal (GF090R) anti-GFP Nacalai Tesque, inc. Cat# 04404-84; RRID: AB_10013361
Mouse (IgG2b) monoclonal anti-APC (CC-1) Millipore Cat# OP80; RRID: AB_2057371
Mouse (IgG1) monoclonal anti-CASPR
(Clone K65/35)
UC Davis/NIH NeuroMab Facility Cat# 75-001; RRID: AB_2083496
Mouse (IgG1) monoclonal anti-NeuN Millipore Cat# MAB337; RRID: AB_2298772
Goat (IgG) polyclonal anti-mouse PDGFRaR & D Systems Cat# AF1062; RRID: AB_2236897
Rabbit polyclonal anti-OLIG2 Millipore Cat# AB9610; RRID: AB_570666
Rabbit polyclonal anti-Na
v
1.6 Alamone labs Cat# ASC-009; RRID: AB_2040202
Rabbit polyclonal anti-MAP2 Millipore Cat# AB5622; RRID: AB_91939
Donkey anti-rat Alexa Fluor 488 Thermo Fisher Scientific Cat# A-21208; RRID: AB_2535794
Donkey anti-mouse Alexa Fluor-568 Thermo Fisher Scientific Cat# A10037; RRID: AB_2534013
Donkey anti-goat Alexa Fluor-568 Thermo Fisher Scientific Cat# A-11057; RRID: AB_2534104
Donkey anti-rabbit Alexa Fluor 647 Thermo Fisher Scientific Cat# A-31573; RRID: AB_2536183
Goat anti-mouse STAR RED Abberior Cat# STRED-1001-500UG; RRID: n/a
Goat anti-rabbit STAR Orange Abberior Cat# STORANGE-1002-500UG; RRID: n/a
Hoechst 33342 Invitrogen Cat# H1399; RRID: n/a
Deposited Data
CV model This paper https://github.com/JolivetLab.
Experimental Models: Organisms/Strains
Mouse / C57BL/6J The Jackson Laboratories IMSR Cat# JAX:000664; RRID:
IMSR_JAX:000664
Mouse / Plp-CreER
T
The Jackson Laboratories Cat# JAX:005975; RRID:
IMSR_JAX:005975
Mouse / Tau-mGFP The Jackson Laboratories Cat# JAX:021162; RRID:
IMSR_JAX:021162
Mouse / Myrf floxed The Jackson Laboratories Cat# JAX:010607; RRID:
IMSR_JAX:010607
Mouse / Pdgfra-CreER
TM
The Jackson Laboratories Cat# JAX:018280; RRID:
IMSR_JAX:018280
Mouse / Rosa26-YFP The Jackson Laboratories Cat# JAX:006148; RRID:
IMSR_JAX:006148
Oligonucleotides
Cre 50CAGGT CTCAG GAGCT ATGTC
CAATT TACTG ACCGTA
Integrated DNA Technologies n/a
Cre 30GGTGT TATAAG CAATCC CCAGAA Integrated DNA Technologies n/a
GFP 50CCCTG AAGTTC ATCTG CACCAC Integrated DNA Technologies n/a
GFP 30TTCTC GTTGG GGTCT TTGCTC Integrated DNA Technologies n/a
Rosa26 wildtype 50AAAGT CGCTC TGAGT
TGTTAT
Integrated DNA Technologies n/a
Rosa26 wildtype 30GGAGC GGGAG
AAATG GATATG
Integrated DNA Technologies n/a
Rosa26 YFP 50GCGAA GAGTT TGTCC
TCAACC
Integrated DNA Technologies n/a
Myrf 50AGGAG TGTTG TGGGA AGTGG Integrated DNA Technologies n/a
Myrf 30CCCAG GCTGA AGATG GAATA Integrated DNA Technologies n/a
(Continued on next page)
e1 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
RESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kaylene
Young (Kaylene.Young@utas.edu.au).
Materials availability
This study did not generate new unique reagents.
Data and code availability
All data generated and analyzed for this study is included in the manuscript. Data and code for computational modeling are available
from GitHub (https://github.com/JolivetLab).
EXPERIMENTAL MODEL AND SUBJECT DETAILS
All animal experiments were approved by the University of Tasmania Animal Ethics Committee and carried out in accordance with the
Australiancode of practice for the care anduse of animals for scientific purposes. All wild-type and transgenic mice were maintained on a
C57BL/6J background. Heterozygous Plp-CreER transgenic mice (Doerflinger et al., 2003) were crossed with heterozygous Tau-lox-
STOP-lox-mGFP-IRES-NLS-LacZ-pA (Tau-mGFP) Cre-sensitive reporter mice (Hippenmeyer et al., 2005) to generate double heterozy-
gous offspring for the fluorescent labeling and tracing of OLs. Heterozygous Pdgfra-CreER
TM
transgenic mice (Kang et al., 2010)were
crossed with homozygous Rosa26-YFP Cre-sensitive reporter mice (Srinivas et al., 2001) to generate double heterozygous offspring for
the fluorescent labelingand tracing of OPCs and the newborn cells they produce. Homozygous Myrf loxP-flanked exon 8 mice (Myrf
fl/fl
)
(Emery et al., 2009) were crossed with heterozygous Pdgfra-CreER
TM
transgenic mice (Kang et al., 2010)orhomozygousRosa26-YFP
Cre-sensitive reporter mice (Srinivas et al., 2001)toproducePdgfra-CreER
TM
:: Myrf
fl/fl
and Myrf
fl/fl
:: Rosa26-YFPoffspring, respectively.
These offspring were then intercrossed to generate Pdgfra-CreER
TM
:: Rosa26-YFP:: Myrf
fl/fl
(Myrf
fl/fl
)andRosa26-YFP:: Myrf
fl/fl
(control)
mice for experiments. Male and femalemice were housed in same sex groups (2-4per cage), in individually ventilated cages(Optimice)
on a 12 h light cycle (twilight phase starts06:30, full lights on07:00) at 21 ±2Cwithad libitum access to standard rodentchow (Barrastoc
rat and mousepellets) and water. Experimentalmice (P60-P90) weighed18-35 g at the start of experiments andwere randomly assigned
to each treatment, but care was taken to ensure littermates were represented across treatment groups.
METHOD DETAILS
Transgenic lineage tracing and gene deletion
Cre,Rosa26-YFP and Tau-mGFP transgenes were detected by PCR as described by Cullen et al. (2019), and the Myrf
floxed
gene was
detected as described by Emery et al. (2009). In brief, genomic DNA (gDNA) was extracted from ear biopsies by ethanol precipitation
and PCR was performed using 50-100ng of gDNA with the following primer combinations: Cre 50CAGGT CTCAG GAGCT ATGTC
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Software and Algorithms
Fiji (ImageJ) https://imagej.nih.gov/ij/ RRID: SCR_002285
GraphPad Prism (version 8) GraphPad Software, Inc. RRID: SCR_002798
pClamp (version 11) Molecular Devices RRID: SCR_011323
MATLAB Mathworks RRID: SCR_001622
Volocity 3D Image Analysis Software Perkin Elmer RRID: SCR_002668
Other
Multi-Maze modular system (RAM) Ugo Basile Cat# 41500
Kepco Bop 100-4M programmable power
supply
TMG test equipment n/a
BenchLink waveform builder Agilent Technologies Part # 33521A
GM08 Gauss Meter, DC and 15Hz - 10kHz Hirst Magnetic Instruments n/a
Scientifica SliceScope Pro 1000
electrophysiology rig
Scientifica Ltd. n/a
Axopatch 200B patch clamp amplifier Axon Instruments n/a
Iso-Stim 01D stimulus isolator NPI Electronic n/a
DigiGait
TM
gait analysis system Mouse Specifics, Inc. n/a
Cell Reports 34, 108641, January 19, 2021 e2
Article
ll
OPEN ACCESS
CAATT TACTG ACCGTA and Cre 30GGTGT TATAAG CAATCC CCAGAA; GFP 50CCCTG AAGTTC ATCTG CACCAC and GFP 30
TTCTC GTTGG GGTCT TTGCTC; Rosa26 wild-type 50AAAGT CGCTC TGAGT TGTTAT, Rosa26 wild-type 30GGAGC GGGAG
AAATG GATATG and Rosa26 YFP 50GCGAA GAGTT TGTCC TCAACC; Myrf 50AGGAG TGTTG TGGGA AGTGG and Myrf 30CCCAG
GCTGA AGATG GAATA.
To activate Cre-recombinase in OPCs and induce targeted DNA recombination, Tamoxifen was dissolved in corn oil (40mg/ml) by
sonication at 21C for 2 h and administered to adult mice (P60-P83) by oral gavage at a dose of 300mg tamoxifen/kg body weight
daily for four consecutive days. Plp-CreER:: Tau-mGFP mice were given a single dose of 50mg/kg, 100mg/kg or 300mg/kg body
weight to enable clear visualization of individual mGFP
+
internodes (Figure S1).
Low intensity repetitive transcranial magnetic stimulation
Low intensity repetitive transcranial magnetic stimulation (Li-rTMS) was delivered as per Cullen et al. (2019). Briefly, 600 pulses of
intermittent theta burst stimulation (iTBS; 192 s) was delivered using a custom made 120mT circular coil designed for rodent stim-
ulation (8mm outer diameter, iron core) (Tang et al., 2016a,2016b). Stimulation parameters were controlled by a waveform generator
(Agilent Technologies) connected to a bipolar voltage programmable power supply (KEPCO BOP 100-4M, TMG test equipment). Ex-
periments were conducted at 100% maximum power output (100V) using custom monophasic waveforms (400ms rise time; Agilent
Benchlink Waveform Builder). Mice were restrained using plastic body-contour shape restraint cones (0.5mm thick; Able Scientific).
The coil was manually held over the midline of the head with the back of the coil positioned in line with the front of the ears (~Bregma
3.0). Sham mice were positioned under the coil for 192 s (as per iTBS), but no current was passed through the coil. Stimulation was
carried out once daily, at the same time, for 7, 14 or 28 consecutive days. Li-rTMS did not elicit observable behavioral changes in the
mice during or immediately after stimulation.
DigiGait
TM
gait analysis
Gait analysis was performed using the DigiGait
TM
treadmill imaging system (Mouse Specifics, Inc). Prior to, and again after 7 and
14 days of iTBS or sham stimulation, mice were habituated to the treadmill enclosure for 5 min before the treadmill was turned on
and the speed of the transparent belt increased from 10cm/s to 28cm/s over 2 min. Short ~10 s videos were recorded from under-
neath the mice as they ran at a belt speed of 28cm/s. The 10 s videos were cut to a length of ~3-4 s in which mice were running straight
without obvious acceleration or deceleration, by an experimenter blind to treatment group. The short video clips were analyzed using
the semi-automated DigiGait
TM
analysis software. Each digital analysis output was then cross checked for processing errors (e.g., a
forepaw mistakenly labeled as a hind paw) and corrected, if required, before data were exported for statistical analysis.
Ledge beam task
Motor balance and coordination was assessed using the ledged beam task prior to, and after 7 and 14 days of iTBS or sham-stim-
ulation. Mice were placed at the lower (wider) end of a plexiglass beam that tapered from 3.5cm to 0.5cm over a length of 50 cm and
was placed at an incline of 30 degrees. A peripheral ledge (0.5cm wide) ran along each side of the beam, 1cm below the top. After an
initial training session, mice were video recorded as they ran up the beam and into their home cage, located at the narrow end. These
videos were manually scored for foot slips by an experienced experimenter blind to the treatment groups. A foot slip was scored
when a mouse placed any paw on the peripheral ledge instead of the central beam.
Spatial learning
To induce spatial learning, adult (P60) male and female Plp-CreER:: Tau-mGFP or littermate control mice were administered a single
dose of tamoxifen (50mg/kg), then handled daily for two weeks before being trained in an 8-arm radial arm maze (RAM) task (Figure S3)
over 14 days. 5 days prior to RAM training, non-learning and learning mice had their access to normal mouse chow restricted to 6h per
day but were given food rewards (Froot Loopspieces) in their home cage. This food restriction protocol ensured that mice were main-
tained at ~90% of their free feeding body weight and were motivated to seek out and consume the food rewards when available.
The RAM was carried out using the multi-maze system for mice (Ugo-Basile) in a radial 8-arm configuration with spatial cues placed
on each of the surrounding walls, ~30cm above the maze. RAM training consisted of two phases - a familiarization phase (days 1-3)
and a learning phase (days 4-14) (Figure S3), and each mouse underwent 3 trials per day, with 60 min between each trial. During the
familiarization phase, all arms of the RAM were closed off and an individual mouse was placed in the octagonal center of the maze
with a single Froot Loop(cut into 8 approximately equal sized pieces) for 10 min. No-learning control mice were returned to the
familiarization phase conditions for the 14 days of the task, to ensure that they were subjected to the same environment and an equiv-
alent level of handling and that they also received Froot Loops, but did not learn the RAM task (Figure S3A). Learning mice pro-
ceeded to undertake a learning phase, in which the 8 pieces of Froot Loopwere distributed, so that one piece was placed at
the end of each arm of the RAM. An individual mouse was placed in the center of the maze and could explore the maze for
10 min (Figure S3B). Over the next 11 days, the mice learned that each arm contained a single food reward and that repeat entries
would not result in another reward. Therefore, repeated entries into an arm in which the food reward had already been consumed was
counted as an error, and the average number of errors made per trial was quantified as a measure of learning (Figure S3C). 24 h after
the final trial, mice were either perfusion fixed and their brains collected and prepared for either fluorescent or transmission electron
microscopy, or were used to generate acute brain slices for compound action potential recordings.
e3 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
Tissue preparation and immunohistochemistry
Mice were perfusion-fixed with 4% paraformaldehyde (PFA; Sigma) (w/v) in phosphate buffered saline (PBS). Brains were cut into 2mm-
thick coronal slices using a 1mm brain matrix (Kent Scientific) before being post-fixed in 4% PFA at 21C for 90 min. Tissue was cryo-
protected overnight in 20% (w/v) sucrose (Sigma) in PBS and snap frozen in OCT (ThermoFisher) for storage at 80C. 30mmcoronal
brain cryosections, containing the primary motor cortex and underlying corpus callosum (~Bregma +0.5) or the dorsal region of the
fimbria (~Bregma 1.5), were collected and processed as floating sections (Cullen et al., 2019). Primary and secondary antibodies
were diluted in PBS blocking solution [0.1% (v/v) Triton X-100 and 10% fetal calf serum in PBS] and applied to sections overnight at
4C, unless staining involved the use of mouse anti-CC1 (1:100 Calbiochem),in which case antibodies werediluted in Tris bufferedsaline
(TBS) blocking solution [0.1% (v/v) Triton X-100 and 10% fetal calf serum in TBS]. Primary antibodies included goat anti-PDGFRa(1:200;
R&D Systems), rabbit anti-OLIG2 (1:400 Millipore), rat anti-GFP (1:2000; Nacalai Tesque), rabbit anti-NaV
1.6
(1:500 Alomone Labs),
mouse anti-CASPR (Clone K65/35; 1:200 NeuroMab), mouse anti-NeuN (1:200 Millipore) and rabbit anti-MAP2 (1:1000 Millipore). Sec-
ondary antibodies, which were conjugated to AlexaFluor 488, 568 or 647 (Invitrogen) were donkey anti-goat (1:1000), donkey anti-
rabbit (1:1000), donkey anti-mouse (1:1000), and donkey anti-rat (1:500). Nuclei were labeled using Hoechst 33342 (1:1000; Invitrogen).
Confocal microscopy and image quantification
Confocal images were collected using an UltraView Nikon Ti Microscope with Volocity Software (Perkin Elmer). High-magnification
images (z-spacing of 0.5-2mm) were collected using standard excitation and emission filters for DAPI, FITC (AlexaFluor-488), TRITC
(AlexaFluor-568) and CY5 (AlexaFluor-647), then stitched together to make a composite image of a defined region of interest. To
quantify internode number and length for OLs within the primary motor cortex (M1), high-magnification images (40x objective)
were collected through individual mGFP-labeled cortical OLs (0.5mm z-steps) that had a visible cell body. To quantify internodes
in the corpus callosum (CC) and hippocampal fimbria, high-magnification images (60x objective) were collected (0.5mm z-steps)
and used to identify individual mGFP-labeled internodes that were flanked by CASPR
+
paranodes. To measure node of Ranvier
(Na
v
1.6) and paranode (CASPR) length, high-magnification (100x) single z-plane confocal images were collected from M1, the CC
and the fimbria. Node and paranode lengths were only measured when a node and its flanking paranodes were intact within the single
z-plane. To measure neuronal soma size within M1, high-magnification (40x objective) confocal images (0.5mm z-steps) were
collected from 5 fields of view and used to identify and measure NeuN
+
soma that were enveloped by a clear MAP2
+
ring. For quan-
tification of cell number, low-magnification (20x objective) confocal z stacks (2 mm spacing) were collected through M1, CC or the
hippocampal fimbria and stitched together to make a composite image of a defined region of interest. All image analysis was carried
out using ImageJ (NIH) by a researcher blind to experimental treatment.
Stimulated emission depletion (STED) microscopy
30mm coronal cryosections containing the CC (~Bregma +0.5) were collected and prepared as floating sections. Rabbit anti-Na
v
1.6
(1:500 Alomone Labs) and mouse anti-CASPR (Clone K65/35; 1:200 NeuroMab) primary antibodies were diluted in PBS blocking so-
lution [0.1% (v/v) Triton X-100 and 10% fetal calf serum in PBS] and applied to sections overnight at 4C. The sections were washed in
PBS (3 310 min) before overnight application (4C) of goat anti-mouse STAR Red (1:500, Abberior) and goat anti-rabbit STAR Or-
ange (1:500, Abberior) secondary antibodies. The sections were mounted in antifade liquid mounting media (Abberior) and covered in
a 170 mm thick glass coverslip (ProSciTech, cat # EMS72291-06).
STED imaging was performed using a two-color Abberior STEDYCON system (Abberior Instruments GmbH) attached to a Nikon
NiE confocal microscope equipped with 405nm, 488 nm, 561 nm, and 640 nm pulsed excitation lasers, a pulsed 775 nm STED laser
and a 100x oil immersion objective lens (N.A 1.4). Images were acquired using Abberior STEDYCON smart control software. For all
images the pixel size and dwell time were kept consistent at 20nm and 10ms, respectively. 561nm and 640nm excitation lasers were
set to 10% power but STED laser power was optimally set to 100% (STAR orange, 561nm) or 56.2% (STAR red, 640nm). Single z-
plane STED images were collected from the CC to enable the precise, high-resolution visualization of nodes of Ranvier (Na
v
1.6) and
their abutting paranodes (CASPR).
Transmission electron microscopy
Following 14 days of iTBS, sham stimulation or RAM training, P105 or P89 mice were perfused with Karnovsky’s fixative (2.5%
glutaraldehyde, 2% PFA, 0.25mM CaCl
2
,0.5mMMgCl
2
in 0.1M sodium cacodylate buffer). Brains were cut into 1mm-thick coronal
slices using a 1mm brain matrix (Kent Scientific) and post-fixed in Karnovsky’s fixative for 2h at 21C. The tissue blocks were rinsed
and stored in 0.1M sodium cacodylate buffer overnight. The medial part of the CC (~Bregma +0.5 to 0.5) or hippocampal fimbria
(~Bregma 1.0 to 2.0) was dissected and incubated in 1% osmium tetroxide / 1.5% potassium ferricyanide [OsO
4
/
K
3
Fe(III)(CN)
6
] in 0.1M sodium cacodylate buffer in the dark for 2h at 4C, before being dehydrated in ethanol and propylene oxide,
and embedded in Epon812 resin. Ultrathin 70nm sections were cut using a Leica Ultra-cut UCT7 and stained with uranyl acetate
and lead citrate. High resolution electron microscopy imaging was done at 80kV on a JEOL 1400-Flash (CC) or a Hitachi HT7700
(fimbria) transmission electron microscope. Sectioning, imaging, and image analysis was carried out by an experimenter blind to
the treatment group.
Image analysis was carried out using ImageJ (NIH). The proportion of myelinated axons and the g-ratio of myelinated axons [axon
diameter / (axon + myelin diameter)] were measured from at least 100 axons from 5 images per animal. The number of myelin wraps
Cell Reports 34, 108641, January 19, 2021 e4
Article
ll
OPEN ACCESS
was quantified by counting major dense lines for a minimum of 15 transected axons per mouse, from n = 3 mice per treatment group.
The cross-sectional area of the adaxonal inner tongue membrane was quantified by tracing the outer edge of the inner tongue using
the polygon selection tool in ImageJ. The average thickness of myelin wraps per axon, and the width of the periaxonal space were
measured from ultra-high-magnification (3120-200k) images of axons ensheathed by compact myelin a minimum of five wraps thick
(R10 transected axons per mouse, from n = 3 mice per treatment group). To ensure unbiased axon sampling and measurements of
the periaxonal space width, high-magnification images were collected by an experimenter blind to treatment condition. These coded
images were then analyzed by a second experimenter and decoded after quantification was complete.
Conduction velocity modeling
In order to evaluate the effect on action potential propagation of experimentally observed changes in node length and myelin structure,
i.e., periaxonal space width + associated change in g-ratio, [data derived from the population mean from n = 3 animals for iTBS exper-
iments (Figures 1,2,and5;Figure S2) and n = 3-4 animals for RAM experiments (Figures 3 and 5;Figure S3)], we further adapted the
mathematical model of action potential propagation in myelinated axons proposed by Richardson and colleagues (model ‘C’, their
figure 1; Bakiri et al., 2011;Richardson et al., 2000). A recent MATLAB (The MathWorks) implementation of that model by Cossell
and colleagues can be downloaded from GitHub (https://github.com/AttwellLab/MyelinatedAxonModel)(Arancibia-Ca
´rcamo et al.,
2017;Bakiriet al., 2011;Ford et al., 2015;Younget al., 2013). That package was downloaded in June 2018 and run on MATLAB R2016b.
The mathematical description of ion channels at nodes of Ranvier follows the Hodgkin-Huxley formalism. Briefly, nodes express
three types of ion channels, a fast sodium channel ifast
Na responsible for the initiation of action potentials, a persistent sodium
channel ipersistent
Na , and a slow potassium channel islow
Kresponsible for the termination of action potentials. The kinetics of the three
currents is derived from McIntyre and Grill (2002). Briefly, ifast
Na is written:
ifast
Na =gNaf m3hðVENaÞ(1)
with gNaf the current conductance, Vthe membrane voltage at the node, ENa =60mV the reversal potential for sodium ions, and mand
hsome gating variables. Following the Hodgkin-Huxley formalism, each gating variable xin the model follows the generic equation:
dx
dt =axð1xÞbxx(2)
with aand bsome functions of V. For ifast
Na ,aand bare given by:
am=6:57 V+20:4
1eV+20:4
10:3
(3)
bm=0:304 V+25:7
1eV+25:7
9:16
(4)
ah=0:34 V+114
1eV+114
11
(5)
bh=12:6
1+eV+31:8
13:4
:(6)
The persistent sodium current ipersistent
Na is given by:
ipersistent
Na =gNap p3ðVENaÞ(7)
with:
ap=0:0353 V+27
1eV+27
10:2
(8)
bp=0:000883 V+34
1eðV+34Þ=10Þ:(9)
The slow potassium current islow
Kis given by:
islow
K=gKsðVEKÞ(10)
e5 Cell Reports 34, 108641, January 19, 2021
Article
ll
OPEN ACCESS
with EK=90mV the reversal potential for potassium ions and:
as=0:3
1+eV+53
5
(11)
bs=0:03
1+eðV+90Þ:(12)
Finally, all membranes contain a leak current given by:
iL=gLðVELÞ:(13)
The Q10 for gates m,h,p, and sare 2.2, 2.9, 2.2 and 3.0 respectively, described at 36C [note that this has the effect of slowing down
the kinetics of gates m,h, and pwith respect to McIntyre and Grill (2002)]. All the other parameters of the model are given in Table S1.
The same framework was used to simulate both corpus callosum and fimbria axons but with different parameter sets (see Table S1).
All numerical values were set to the average values measured for the sham or no-learning experimental groups, with the exception of
parameters that were significantly different following iTBS or learning – in these cases the sham, iTBS, no-learning or learning values
were directly entered into the model (see Table S1). Parameters that were not obtained from our experimental data were adapted
from Arancibia-Ca
´rcamo et al. (2017). Myelin thickness was automatically calculated using:
Myelin thickness =ðd=gd2pswÞ=2 (14)
with dthe axon diameter, gthe g-ratio and psw the periaxonal space width. Myelin lamella periodicity was taken as myelin thickness
divided by 6.5 so that the number of wraps is 7 in all conditions as observed experimentally (see Figure 5), assuming that the extra-
cellular space between myelin lamellae comprises part of the periodicity, and to account for the fact that there is no extracellular
space contributing to the total width of the myelin on the most external lamella. Note that as the number of wraps does not change
between any of the conditions, changes in myelin thickness reflect the addition – or subtraction – of cytoplasmic space between lipid
bilayers. Adding or removing cytoplasmic space filled with intracellular solution between lipid bilayers does not significantly affect the
resistance or capacitance across myelin wraps, and thus does not affect CV (see Figure S5O).
Unless stated otherwise, simulations were run using a time step of 0.1ms and 51 nodes. Internode segments were chosen to be <
1mm (0.98mm; n = 52 segments per internode) and we verified that this was sufficient to reach convergence for the CV over the whole
range of simulated axons (Figure S5). Action potentials were triggered by a square pulse of 0.5nA lasting 10ms. Ion channels at jux-
taparanodes were not modeled, as is common in the field. When altering the length of the node of Ranvier (see below), the density of
ion channels at the node was taken to be constant (see Arancibia-Ca
´rcamo et al., 2017, for a systematic discussion of how this affects
action potential CV).
To evaluate individually the effect of a node length reduction or a change in the myelin sheath (i.e., increase in periaxonal space
width + accompanying decrease in g-ratio) on CV, we initially ran four sets of simulations. First, we used a parameter set matching
the observations obtained in the sham condition (column ‘Sham’ in Table S1). We then ran the same simulations after reducing node
lengths (‘Short nodes’). Third, we ran simulations modifying the myelin sheath but keeping node length as per the sham condition
(‘Alt. myelin’). Finally, we ran a simulation with a fourth set of parameters implementing both of the experimental changes observed
following iTBS, i.e., a reduction in node length and altered myelin sheath (change in periaxonal space + corresponding change in g-
ratio). Simulations were run at 21C and at 37C(Figure 6;Figure S5). We additionally investigated three different scenarios for con-
duction at the paranode. The periaxonal space width at the paranode was taken to be either [i] equal to the periaxonal space width in
the internode; [ii] equal to the periaxonal space width under the internode if that is less than 3nm, but to be at most 3nm otherwise; or
[iii] equal to half the periaxonal space width in the internode (Figure 6;Figure S5). Unless otherwise specified, scenario [i] is in use.
Each paranode was taken to be 2 segments long (1.96mm long).
To evaluate the functional consequence of myelin alterations, we additionally ran a set of simulations varying the periaxonal
space width from 0 to 20nm at both 21C and 37C(Figure 6). These simulations show that at 37C, the periaxonal space can
shift action potential CV between 4.36 m/s (psw = 0nm) and 1.25 m/s (psw = 20nm; Figure 6). These numbers illustrate how potent
and elegant this mechanism is, as it can speed up or slow down action potential conduction by a factor of 3.5 by making minor
adjustments to the structure of myelinated axons. The functional consequences of this change to propagation speed at 37Cisto
alter the arrival time of action potentials by 6ms over a distance of 1cm (Figure 6), enough to alter learning via spike-timing depen-
dent plasticity for instance.
Finally, Cohen et al. (2020) recently reported different values for the axonal and periaxonal space resistivities. In particular, they
reported that the axonal resistivity is about three times larger than the periaxonal space resistivity. In order to evaluate how this would
impact our results, we simulated conduction by cortical axons (see Figure 6;Figure S5) using the resistivity values reported by Cohen
et al. (2020) (axonal resistivity = 1.5 U$m; periaxonal space resistivity = 0.54 U$m), and adapting the conductance of some ionic
channels at nodes of Ranvier to match action potential conduction velocities to those reported in Figure 6 and S5 (Figure S6;
g
Naf
= 550 mS/mm
2
,g
K
= 24 mS/mm
2
). With these new parameters, we observe a slight reduction of the effects reported in Figures
6and S5, but still find that adjusting the periaxonal space width effectively modulates action potential CV (Figure S6).