Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
1
TITLE: Organellular imaging in vivo reveals a depletion of endoplasmic reticular calcium during 1
post-ictal cortical spreading depolarization 2
3
AUTHORS: Matthew A. Stern1, Eric R. Cole1,2, Claire-Anne Gutekunst1, Jenny J. Yang3, Ken 4
Berglund1* and Robert E. Gross1† 5
6
AFFILIATIONS: 7
1Department of Neurosurgery, Emory University School of Medicine, Atlanta, GA, United States 8
2Coulter Department of Biomedical Engineering, Emory University and Georgia Institute of 9
Technology, Atlanta, GA, United States 10
3Department of Chemistry, Center for Diagnostics and Therapeutics, Advanced Translational 11
Imaging Facility, Georgia State University, Atlanta, GA, United States 12
*Corresponding author: Ken Berglund, ken.berglund@emory.edu 13
†Present Address: Department of Neurological Surgery, Rutgers Robert Wood Johnson Medical 14
School, New Brunswick, NJ, United States 15
16
ABSTRACT 17
During cortical spreading depolarization (CSD), neurons exhibit a dramatic increase in cytosolic 18
calcium, which may be integral to CSD-mediated seizure termination. This calcium increase 19
greatly exceeds that during seizures, suggesting the calcium source may not be solely extracellular. 20
Thus, we sought to determine if the endoplasmic reticulum (ER), the largest intracellular calcium 21
store, is involved. We developed a two-photon calcium imaging paradigm to simultaneously 22
record the cytosol and ER during seizures in awake mice. Paired with direct current recording, we 23
reveal that CSD can manifest as a slow post-ictal cytosolic calcium wave with a concomitant 24
depletion of ER calcium that is spatiotemporally consistent with a calcium-induced calcium 25
release. Importantly, we observed both naturally occurring and electrically induced CSD 26
suppressed post-ictal epileptiform activity. Collectively, this work links ER dynamics to CSD, 27
which serves as an innate process for seizure suppression and a potential mechanism underlying 28
therapeutic electrical stimulation for epilepsy. 29
30
MAIN TEXT 31
32
INTRODUCTION 33
34
The endoplasmic reticulum (ER), a major calcium reservoir of the cell, is critically involved in 35
essential physiological processes. The perturbation of ER homeostatic regulation in turn has 36
serious pathophysiologic implications1. This is especially the case for the nervous system, where 37
intra- and inter-cellular communication is tightly regulated through the ER calcium, including 38
synaptic transmission, transcriptional regulation and plasticity2-4. Deciphering calcium signaling 39
dynamics at a spatiotemporal level during aberrant activity could inform our understanding of 40
disease mechanisms and their downstream sequelae. Calcium imaging of the ER has been enabled 41
through the development of various ER targeted dyes and genetically encoded calcium indicators5,
42
6, but these have hitherto not been applied in vivo to vertebrates. Thus, we devised an approach for 43
concurrent in vivo calcium imaging of the cytosol and ER. We then applied this approach to 44
elucidate the difference between two related neurological phenomena both causing high 45
intracellular calcium: seizures and cortical spreading depolarization (CSD)7. 46
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
2
CSD was first documented in 1944, when Aristides Leão reported his first series of 47
electrocorticography recordings of slow propagating waves of depression in rabbit cortex, an 48
investigation he began originally to induce epileptiform discharges8, 9. He would later go on to 49
characterize these traveling waves as large scale depolarizing events, exhausting the tissue into a 50
state of depression, hence the term CSD10. While CSD is largely considered to be pathologic across 51
a wide array of diseases11-15, its relationship with seizures appears to be more complicated 16. 52
Observations in animal models17-20 and patients21, 22 have demonstrated a strong association 53
between the two events with CSD occurring often at the end of seizures. Indeed, the ionic shifts 54
that occur during seizures are conditions that parallel those of CSD. Large extracellular deposition 55
of glutamate and potassium7, 23 and the shrinking of extracellular space24, 25 create a hyperexcitable 56
state for neurons that can beget CSD, in a feedforward amplificatory fashion26. The subsequent 57
impact of CSD in seizures is both deleterious and protective, being implicated as an underlying 58
cause of sudden unexplained death in epilepsy (SUDEP)27, as well as a mechanism of seizure 59
termination17. Thus, having a better understanding of the interplay between these two neurologic 60
phenomena could be of vital importance for both the suppression of seizures, as well as prevention 61
of one of epilepsy’s most feared complications. 62
In our previous studies using in vivo two-photon calcium imaging in awake mice to 63
evaluate seizure dynamics28, we observed slow propagating calcium waves at the end of seizures, 64
which corresponded with an absence of high frequency neuronal firing in electrocorticography 65
(EEG). We hypothesized that these may be CSD, as large calcium transients are a known 66
occurrence during spreading depolarizations29, 30 and have been observed as waves31 with similar 67
spatiotemporal properties. As the increase in intracellular calcium during CSD exceeds that during 68
seizures7, extracellular influx is unlikely to be the only contributing source of calcium, raising the 69
possibility of ER involvement. Furthermore, since elevated calcium levels are hypothesized to 70
mediate seizure termination32, this dramatic calcium rise may underly the purported seizure 71
suppressive effects of CSD. 72
To investigate the mechanisms underlying these stark intracellular increases in calcium, 73
we generated an adeno-associated viral (AAV) vector to transduce neurons with two calcium 74
indicators at the same time, each targeted to a separate intracellular compartment. In the cytosol 75
we express a yellow derivative of the GCaMP family, XCaMP-Y33, and in the ER lumen we 76
express a red indicator, RCatchER34. RCatchER is a low affinity (Kd = ~400 µM) calcium indicator 77
protein with fast kinetics (<1 ms) and a fluorescence intensity that varies linearly with calcium 78
concentration, making it an optimal indicator for ER calcium. We then intravitally recorded 79
generalized seizures35 in awake mice, concurrently with EEG and direct current (DC) recordings, 80
with DC being the gold standard for confirmation of CSD. 81
In the present study we determine that these slow propagating post-ictal calcium waves are 82
in fact CSDs. We next show that CSD is marked by a stark depletion of ER calcium not occurring 83
during normal or epileptiform activity. Our spatiotemporal analysis at a cellular level indicates that 84
this depletion is consistent with a calcium-induced calcium release (CICR). In addition, we observe 85
that depletion of ER calcium also occurs during CSD evoked by electrical stimulation. Finally, we 86
provide causal evidence that naturally occurring CSD is associated with a suppression of post-ictal 87
epileptiform activity and this same suppression can be achieved through the biologically similar 88
electrically evoked CSD. Collectively this work offers new insight into the biological 89
underpinnings of CSD and its potential utility for seizure control. 90
91
RESULTS 92
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
3
93
Intravital imaging of cytosol and ER calcium stores in awake mice 94
To study intracellular calcium dynamics in vivo with high spatiotemporal resolution, we developed 95
a recombinant AAV to express two genetically encoded calcium indicators (GECIs) of different 96
colors pan-neuronally through the human synapsin I (hSynI) promoter36. The yellow XCaMP-Y33 97
and the red-shifted RCatchER34 GECIs were separated by the self-cleaving P2A peptide to obtain 98
similar expression levels (Fig. 1a). While the XCaMP-Y will be localized to the cytosol, the 99
RCatchER includes calreticulin and KDEL sequences, for targeting and retention in the ER lumen, 100
respectively. We confirmed this expression pattern by immunohistochemistry (Fig. 1b). These 101
indicators were selected for their ability to be simultaneously excited with a single wavelength of 102
light between 1000 and 1040 nm (Fig. 1c). Emission was bandpass filtered to isolate each’s distinct 103
signal, thus enabling simultaneous calcium imaging in the two subcellular compartments with 104
single cell resolution. We stereotaxically injected this AAV into the motor cortex of mice and 105
installed chronic cranial windows with head plates to facilitate repeated imaging within subjects. 106
We then performed awake, head-fixed imaging in cortical layers 2/3 (Fig. 1d) to record cytosolic 107
and ER calcium changes across hundreds of neurons in the field (Fig. 1e). During spontaneous 108
locomotion activity, we observed transient increases in cytosolic calcium, classically serving as a 109
proxy for neuronal activity/action potentials, while the ER calcium remained relatively stable (Fig. 110
1f-i) in contrast to the subsequent recordings during pathologic activity. 111
112
113
114
Figure 1. Dual-color two-photon simultaneous intravital calcium imaging of the cytosol and 115
ER during neuronal activity in awake mice 116
(a) Schematic of AAV cassette encoding the XCaMP-Y and RCatchER GECIs (left) enabling 117
mutually exclusive expression in cytosol and ER respectively (right). (b) Immunohistochemistry 118
(confocal) demonstrating localization of XCaMP-Y and RCatchER to the cytosol and ER 119
respectively, with RCatchER but not XCaMP-Y colocalized with the ER marker SERCA. 120
Reconstructions of Z axis at crosshairs presented (right). (c) Overlapping two-photon excitation 121
spectra of the XCaMP-Y and RCatchER captured in vivo (N=1 mouse; n=298 neurons) with the 122
wavelengths used for simultaneous activation highlighted in grey. (d) Illustration of awake in vivo 123
recording set-up of a head-fixed-mouse poised on air-suspended chamber. (e) Representative field 124
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
4
of view (cortical layer 2/3; depth: 200 mm from pial surface) showing time-averaged projections 125
of cells expressing both XCaMP-Y (left, green channel) and RCatchER (right, red channel). (f) 126
Representative individual cell normalized calcium fluorescence traces (DF/F0) of XCaMP-Y 127
(green) and RCatchER (magenta), with detected spike times indicated (blue). (g) Spike raster and 128
corresponding histogram (1s bin width) of spontaneous firing detected in XCaMP-Y during 5 min 129
of baseline recording (N=1 subject; n=254 neurons). (h) Average spike trace relative to indexed 130
spike time, presented with pooled standard deviation of signal (shaded area). 131
132
CSD observed at generalized seizure termination is marked by a unique depletion of ER 133
calcium 134
For recording the calcium dynamics during seizures and purported CSD, we coupled our dual-135
color two-photon imaging with simultaneous EEG and DC recording (Fig. 2a). We recorded EEG 136
through chronically implanted electrodes and DC through a glass microelectrode attached to a DC 137
amplifier. To enable this, we fabricated multipaned concentric glass windows with silicone access 138
ports for repeated access to the brain (Fig. 2a and Supplementary Fig. 1). We induced epileptiform 139
activity by subcutaneous administration of pentylenetetrazol (PTZ; s.c. 40-50 mg/kg; N=9 140
subjects, 30 recordings; Fig. 2a-c). We recorded 23 generalized seizures (Fig. 2b), 4 of which were 141
fatal (analyzed separately in a subsequent section). For the non-fatal seizures, we observed pre-142
ictal spikes (PIS) across all three recording modalities (Fig. 2d, e), beginning within a few minutes 143
of PTZ injection (234±29 s (mean ± SE); N=19 seizures, 8 subjects). Typically, within 10 minutes 144
(471±53 s), we observed a seizure (length: 21±3 s) on EEG followed by a quiescent post-ictal 145
period. Concurrent calcium transients were observed through the cytosolic calcium indicator, 146
while ER calcium stayed relatively stable. Soon after termination of a seizure, we sometimes 147
observed a large and sustained increase in cytosolic calcium (23±5 s; N=6 seizures, 5 subjects), 148
concomitant with a negative DC shift, a hallmark of CSD (17.03±2.85 mV; N=4 seizures, 4 149
subjects with DC recording). This calcium change was comparable to that observed during seizures 150
(p=0.688, generalized linear mixed effects model (GLME), N=6 seizures, 5 subjects, 116-371 151
cells/recording), and greater than that occurring during PIS (p=2.87´10-21). We also observed a 152
concurrent large, rapid and sustained depletion of ER calcium which was significantly larger than 153
calcium changes occurring during the seizure or PIS (p=1.88´10-10 (CSD to seizure), p=7.01´10-
154
17 (CSD to PIS); Fig. 2c-e and Supplementary Fig. 2). PTZ administration did not always induce 155
seizures and CSD (Fig. 2b): it could also induce seizures without CSD, or epileptiform spiking 156
(spike-wave discharges, SWDs) that did not progress to seizure (sub-generalized). We did not 157
observe a comparable change in ER calcium during these events. The large increase in cytosolic 158
calcium and ER depletion was specific to CSD and not a general post-ictal phenomenon, with 159
statistically larger changes in post-ictal calcium during CSD as compared to the same post-ictal 160
period in those seizures without CSD (p= 0.0273 (cytosol), p=9.21´10-21 (ER), GLME, N=17 161
seizures (11 without, 6 with CSD), 8 subjects, 43-371 cells/recording; Fig. 2f). Taken together, we 162
confirmed that the slow propagating cytosolic calcium waves observed at the end of seizures 163
corresponded to CSD with the iconic DC shifts, and a depletion of ER calcium stores. 164
165
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
5
166
167
Figure 2. CSD is associated with generalized seizure termination and is marked by a unique 168
depletion of ER calcium stores 169
(a) Illustration depicting the awake head fixed two-photon imaging along with simultaneous EEG 170
and DC recording during PTZ-induced seizures (left). Image of cranial window with glass 171
electrode inserted through silicone access port (white arrow, right). (b) Pie charts representing 172
proportion of PTZ injections (N=30) that resulted in generalized seizures (N=23) (left) and, of 173
those generalized seizures, the proportions that occurred with (N=6) or without CSD (N=13) or 174
that were fatal (N=4) (right). (c) Representative frames by channel during a PTZ-induced 175
generalized seizure recording depicting the cytosolic (top row) and ER (bottom row) calcium 176
changes during a PIS (i), the seizure (ii) and the CSD (iii: wavefront invasion, iv: CSD after full 177
invasion). Note depletion (left side) of calcium in the recruited area during CSD invasion. (d) 178
Mean population calcium fluorescence (XCaMP-Y: green, RCatchER: magenta) with 179
synchronized EEG (grey) and DC (black) recordings during three separate recording sessions 180
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
6
within the same subject depicting sub-generalized epileptiform activity and seizures with and 181
without CSD. Corresponding rasters of individual cell calcium transients are presented below (y-182
axis: neurons ordered from left to right across the field). (e) Group level analysis of the average 183
individual cell changes in calcium during each event phase (PIS, seizure and CSD) across each 184
recording type (sub-generalized: N=4 subjects, 7 recordings; seizure without CSD: N=6 subjects, 185
12 recordings; and seizure with CSD: N=5 subjects, 6 recordings; n=43-413 cells per recording). 186
(f) Average post-ictal calcium changes are presented for the seizure recordings with and without 187
CSD. N=17 seizures (11 without, 6 with CSD), 8 subjects, 43-371 cells/recording. The effect of 188
event on calcium levels are modeled using GLME for e and f. Means with pooled standard error 189
are presented in all bar plots. *p<0.05, **p<0.01, ***p<0.001. 190
191
ER calcium depletion during seizure CSD is consistent with a CICR 192
Having established the sustained post-ictal calcium depletion in the ER to be specific to CSD, we 193
next sought to characterize the spatiotemporal features of these changes. Both cytosolic and ER 194
calcium changes appeared as a wavefront of propagation (Fig. 3a; Supplementary Movie 1). To 195
determine the timing of the changes across the individual cells, we leveraged our slope integral 196
feature detection approach for seizure traveling waves28 to find the times individual cells were 197
recruited to the CSD in XCaMP-Y and the time of their ER depletion in RCatchER (Fig. 3b). We 198
generated a colormap depicting relative recruitment times within the channels in the field of view 199
(Fig. 3c). Wavefronts in both channels appeared to move in the same direction, with the wave in 200
the ER slightly delayed. 201
We performed a spatial linear regression37, 38 on the recruitment times to determine the 202
wavefront vectors of propagation. We present this first as a polar plot, where the arrows illustrate 203
the direction and speed of the regressed vector of propagation, indicating the overlap in vectors 204
between channels (Fig. 3d and Supplementary Fig. 3). We found that the cytosolic increase and 205
ER depletion propagated in a contiguous fashion, with comparable velocities to each other (Fig. 206
3d-f). These velocities are consistent with typical CSD velocity7, 31. We observed that both 207
cytosolic and ER calcium changes were delayed relative to the CSD DC shift onset (cytosolic: 208
11.35±0.35 s; ER: 11.77±0.50 s; mean ± SE; N=4 seizures, 4 subjects, 142-371 cells/recording). 209
This is concordant with the DC recording site being about 1 mm lateral to the field of view and 210
thus earlier in the wavefronts’ paths of propagation medially, given the determined velocities of 211
propagation. We next plotted the recruitment times within each channel relative to their projected 212
position along the determined axis of propagation in the cytosolic channel (Fig. 3g). This analysis 213
demonstrated a consistently delayed ER depletion along the same axis of propagation as the 214
cytosolic increase. ER depletion was also found to follow the cytosolic increase within each cell 215
(Fig. 3h), with a significant delay of about 1 second (0.79±0.42 s, p=3.37´10-8, GLME, N=6 216
seizures, 5 subjects, 116-371 cells/recording). 217
Interestingly, the loss of calcium from the cytosol and return of calcium to the ER also 218
followed a wavefront pattern, occurring in roughly the same direction as the invasion, although 219
with a slower speed (Fig. 3b-g, i; offset). Unlike the initial change in calcium during the CSD 220
invasion, this subsequent change at the end of CSD was more gradual, making the slope integral 221
feature and maximum slope more difficult to discern at the individual cell level. Therefore, we 222
chose to use the point of maximal concavity, the elbow, being the point when a change started. We 223
again found a delay in recruitment relative to the DC shift, consistent with the slower propagation 224
speed of this change (cytosolic: 24.55±0.90 s; ER: 28.15±1.21 s). Computing the time delay 225
between these onsets and offset events within cells, we found lengths of change of calcium 226
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
7
comparable between the cytosol and ER (Fig. 3j) and consistent with the average duration of the 227
DC shift (38.93±4.88 s, N=4 seizures, 4 subjects). This short delay in ER calcium release following 228
the same spatiotemporal pattern at the cytosolic increase during CSD, with comparable velocity 229
and duration, suggests a CICR is occurring. 230
231
232
233
Figure 3. Spatiotemporal dynamics of subcellular compartment calcium changes during 234
CSD 235
(a) Time lapse images depicting the represented CSD wavefront invasion calcium changes. Circles 236
indicate regions of interest (ROIs) of the representative traces in b. Field orientation is indicated 237
(A: anterior, L: lateral, M: medial, P: posterior). (b) Representative individual cell calcium traces 238
of XCaMP-Y (green) and RCatchER (magenta), during generalized seizures occurring with post-239
ictal CSD along with concurrent EEG (grey) and DC (black) traces. The detected recruitment times 240
to the CSD invasion wavefront are indicated in orange and subsequent calcium changes at the 241
offset of the CSD in blue. (c) ROI colormap of determined recruitment times for identified cells 242
during the onset and offset of the representative CSD in panel a, with color corresponding to 243
recruitment time. (d) Polar plot of the CSD onset and offset propagation vectors modeled by 244
applying spatial linear regression to the neuronal recruitment times shown in panel c, showing 245
wavefront direction (vector angle) and propagation speed (vector magnitude). (e) Average speed 246
of CSD propagation at the onset and offset of the event by channel with standard error. Each CSD 247
is depicted using a unique color matched across all the panels in this figure. (f) Absolute difference 248
in direction between the XCaMP-Y and RCatchER vectors at the onset and offset of CSD. (g) 249
Individual cell recruitment times by channel of the CSD onset and offset projected on to the 250
propagation axis of XCaMP-Y. (h) Mean ER calcium change latency relative to the cytosolic 251
calcium change within cell during CSD invasion across all recordings with pooled standard error 252
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
8
(left) and the distribution of these latencies for each recording (right). The effect of channel on 253
recruitment times used to compute these latencies are modeled using GLME. (i) Absolute 254
difference in direction between the onset and offset vectors of the CSD within channel. (j) Average 255
duration of CSD within cell by channel across recordings with pooled standard error (left) and the 256
distribution of the event durations for each recording (right). For group level analysis (e, g, h-j) 257
N=6 seizures across 5 subjects with n=116-371 cells per recording. *p<0.05, **p<0.01, 258
***p<0.001. 259
260
Occasionally the seizure induced was fatal for the subject (Fig. 2b). For these, a terminal spreading 261
depolarization (TSD) was observed following the seizure during isoelectric EEG, with a negative 262
DC shift (-8.84±0.67 s; mean ± SE; N=4 seizures, 4 subjects) that did not recover (Fig. 4a). 263
Additionally, the initial spatiotemporal changes in calcium observed upon TSD invasion were 264
similar to those observed with CSD in the non-fatal seizures. The waves propagated with similar 265
speeds and direction (Fig. 4b-e), with the depletion of ER calcium following the cytosolic increase, 266
albeit with a slightly longer delay (Fig. 4f, g), and smaller magnitude of cytosolic calcium change 267
(Fig. 4h). Notably, the calcium changes did not return to their baseline values, with cytosolic 268
calcium remaining high and ER depleted, consistent with cell death. 269
270
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
9
271
272
Figure 4. TSD observed during fatal seizures demonstrate a permanent increase in cytosolic 273
calcium and depletion of ER calcium stores. 274
(a) Mean population calcium fluorescence (XCaMP-Y: green, RCatchER: magenta) with 275
synchronized EEG (grey) and DC (black) recordings during a fatal generalized seizure occurring 276
with TSD from a representative subject. Corresponding rasters of individual cell calcium transients 277
(y-axis: neurons ordered from left to right across the field) are presented below. (b) ROI colormap 278
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
10
of determined recruitment times for identified cells during TSD, with color corresponding to 279
recruitment time. (c) Polar plot of the TSD propagation vectors by channel, showing wavefront 280
direction (vector angle) and propagation speed (vector magnitude). (d) Average speed of TSD 281
propagation by channel with standard error. Each TSD is depicted using a unique color matched 282
with panels e and g in this figure. (e) Absolute difference in direction between the XCaMP-Y and 283
RCatchER vectors during TSD. (f) Individual cell recruitment times by channel of the TSD 284
projected on to the propagation axis of XCaMP-Y. (g) Mean ER calcium change latency relative 285
to the cytosolic calcium change within cell during TSD across all recordings with pooled standard 286
error (left) and the distribution of these latencies for each recording (right). The effect of channel 287
on recruitment times used to compute these latencies are modeled using GLME (p=6.25´10-11, 288
N=4 recordings, 4 subjects with N=35-289 cell/recording). (h) Average individual cell changes in 289
calcium by compartment during TSD across recordings presented with pooled standard error. The 290
effect of event on calcium level are modeled using GLME (cytosol: p=0.234, ER: p=4.81´10-4, 291
N=3 recordings, 3 subjects with N=35-289 cells/recording). For all group level analysis (d, e, g, 292
h) N=7 recordings across 7 subjects with n=82-418 cells per recording. *p<0.05, **p<0.01, 293
**p<0.001. 294
295
ER calcium depletion is a conserved feature across multiple types of CSD 296
We next sought to determine if these calcium changes were specific to CSD in the context of 297
seizures or if they were a property of CSD itself. For this we used an electrical stimulation model 298
of CSD39 where we stimulated through two epidural electrodes chronically implanted adjacent to 299
the recording field, while conducting the same imaging, EEG and DC recording paradigm we 300
employed earlier (Fig. 5a). By applying bipolar stimulation between the two electrodes (2 kHz, 301
100 µA, square wave, 10 s) we were able to reliably induce CSD (Supplementary Movie 2). The 302
DC shift during electrically evoked CSD was comparable to that during PTZ-induced CSD (Fig. 303
5b; -19.28±1.97 µV; mean ± SE; N=7 recordings, 7 subjects). Similarly, during the electrically 304
induced CSD we observed changes in calcium that paralleled PTZ-induced CSD, both in terms of 305
magnitude (Fig. 5C) and spatiotemporal pattern (Fig. 5d, e), albeit with the CSD propagating at a 306
faster speed, radially from the electrode pair. The cytosolic and ER calcium changes occurred with 307
the same velocities (Fig, 5f, g), with the ER depletion following the cytosolic increase (Fig 5h, i). 308
We observed a delay in the calcium changes relative to the DC shift consistent both with the 309
position of the imaging field relative to the simulating electrodes, and with the speed of 310
propagation (cytosol: 2.74±0.30 s; ER: 3.03±0.41 s). The durations of the calcium changes (Fig. 311
5j) were again consistent with the average duration of the DC shift (40.53±6.03 s, N=6 recordings, 312
6 subjects). In addition to the high frequency stimulation of long duration (2 kHz for 10 s), we 313
were able to induce CSD with short trains of stimulation of lower frequency, as typically used in 314
responsive neurostimulation (RNS) devices (Supplementary Fig. 4; N=3 subjects, 2 trials per 315
subject; 250-750uA [charge density within an order of magnitude of RNS range], 200 Hz, biphasic, 316
160 µs pulse width, five 100-ms trains with 5 s inter-train interval)40, 41. Thus, the intracellular 317
calcium dynamics observed during seizure associated CSD were also found with electrically 318
induced CSD, suggesting that the depletion of ER calcium is conserved across CSD. 319
320
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
11
321
322
Figure 5. Spatiotemporal subcellular calcium dynamics during stimulation-induced CSD 323
demonstrate ER depletion 324
(a) Illustration depicting the awake head fixed two-photon imaging along with simultaneous EEG 325
and DC recording during stimulation-induced CSD. Expanded is an image of a cranial window 326
with access port and electrodes for stimulation and EEG. (b) Mean population calcium 327
fluorescence (XCaMP-Y: green, RCatchER: magenta) with synchronized EEG (grey) and DC 328
(black) recordings during stimulation-induced (10 s) CSD from a representative subject. Note 329
stimulation artifact at 30 s. Corresponding rasters of individual cell calcium transients (y-axis: 330
neurons ordered from left to right across the field) are presented below. (c) Average individual cell 331
changes in calcium by compartment during CSD across recordings presented with pooled standard 332
error. (d) ROI colormap of determined recruitment times for identified cells during CSD, with 333
color corresponding to recruitment time. (e) Polar plot of the CSD propagation vectors by channel, 334
showing wavefront direction (vector angle) and propagation speed (vector magnitude). (f) Average 335
speed of CSD propagation by channel with standard error. Each CSD is depicted using a unique 336
color matched across all the panels in this figure. (g) Absolute difference in direction between the 337
XCaMP-Y and RCatchER vectors during CSD. (h) Individual cell recruitment times by channel 338
of the CSD projected on to the propagation axis of XCaMP-Y. (i) Mean ER calcium change latency 339
relative to the cytosolic calcium change within cell during CSD across all recordings with pooled 340
standard error (left) and the distribution of these latencies for each recording (right). The effect of 341
channel on the recruitment times use to compute these latencies are modeled using GLME 342
(p=4.63´10-8). (j) Average duration of CSD within cell by channel across recordings with pooled 343
standard error (left) and the distribution of the event lengths for each recording (right). For all 344
group level analysis (c, f, g, i, j) N=7 recordings across 7 subjects with n=82-418 cells per 345
recording. *p<0.05, **p<0.01, ***p<0.001. 346
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
12
347
Post-ictal activity decreases after CSD 348
CSD has been previously reported to arrest epileptiform activity in rodents17. We next sought to 349
see if that observation held true for our generalized seizures. For this, we quantified spike wave 350
discharges (SWDs) typically observed before and after seizures using our EEG recordings (Fig. 351
6a). We found that the SWD rate significantly decreased post-ictally when a seizure was followed 352
by CSD, but not when a seizure occurred without a CSD (Fig. 6b; without CSD: p=0.383, N=8 353
seizures, 5 subjects; with CSD: *p=0.031, N=6 seizures, 5 subjects; Wilcoxon sign-rank test), 354
suggesting a negative correlation between CSD and post-ictal epileptiform activity. To further test 355
if CSD is sufficient to suppress epileptiform activity, we induced CSD using electrical stimulation 356
5 minutes into the post-ictal phase in a subset of seizures unaccompanied by a CSD (Fig. 6c). We 357
found that, while the post-ictal SWD rate before electrical stimulation was not significantly 358
different from the pre-ictal period (Fig. 6d; Friedman test post-hoc comparison; p=0.7593), 359
following electrical stimulation evoked CSD, the SWD rate was significantly decreased (Friedman 360
test with post-hoc comparison: *p=0.0356, n=4 seizures, 3 subjects). An innate gradual decline in 361
post-ictal SWD rate could not account for this observed difference. We compared the recordings 362
in the same time periods (5-10 minutes post seizure) without electrical stimulation and CSD and 363
found there was no significant difference in SWD rate between the pre-ictal and the post-ictal 364
period (p=0.7422; n=8 seizures, Wilcoxon sign-rank test). Taken together, CSD, whether it occurs 365
naturally or evoked by electrical stimulation, has potential to diminish post-ictal epileptiform 366
activity. 367
368
369
370
Figure 6. CSD is associated with suppression of epileptiform activity. 371
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
13
(a) Representative EEG (grey), DC (black) and SWD rates (blue) of the pre- and post-ictal periods 372
for seizures occurring with and without CSD. SWD rate was not calculated during seizure (dark 373
grey box). (b) Box plot comparing the SWD rate between the pre- and post-ictal periods with and 374
without CSD. Wilcoxon sign-rank test was used. Without CSD: p=0.313, N=8 seizures, 5 subjects; 375
With CSD: *p=0.031, N=6 seizures, 5 subjects. (c) Representative EEG (grey), DC (black) and 376
SWD rate (blue) of the pre-ictal and post-ictal periods, during a seizure without naturally occurring 377
CSD, where a CSD was electrically induced post-ictally. The post-ictal period is further delineated 378
as before (pre-stim) and after stimulation (post-stim). SWD rate was not calculated during seizure 379
(dark grey box) or stimulation (light red box). (d) Box plot comparing the SWD rate between the 380
pre- and post-ictal periods (both pre-stim and post-stim) during seizures without naturally 381
occurring CSD, where a CSD was electrically induced. Friedman test with post-hoc comparison 382
was used. *p=0.0356, N=4 seizures, 3 subjects. 383
384
DISCUSSION 385
386
In this study we introduce our XCaMP-Y-P2A-RCatchER imaging construct, for simultaneous in 387
vivo two-photon calcium imaging in the cytosol and ER (Fig. 1). RCatchER has previously only 388
been used in vitro34. Here we expanded the utility of RCatchER by incorporating it into this multi-389
compartment in vivo imaging approach, a novel vertebrate intravital application of calcium 390
imaging to the ER. The ability to capture rapid ER dynamics in awake animals opens vast 391
possibilities for investigators, made all the more accessible through our single AAV design for 392
multicompartment two-color imaging. We envision the usefulness of our construct extending 393
across the biological sciences, from immunology for the study of lymphocyte activation42, to 394
cardiology for examining cardiomyocyte contraction43. 395
To demonstrate the utility of in vivo RCatchER, we apply this paradigm to a rodent seizure 396
model, which enabled us to uncover ER calcium dynamics unique to CSD. Principally, we observe 397
a depletion of ER calcium occurring during post-ictal CSD (Fig. 2) and electrically induced CSD 398
(Fig. 5) that does not occur during seizures themselves. Depletion of ER calcium was delayed by 399
a few seconds relative to a cytosolic calcium increase, suggestive of CICR (Fig. 3). We observed 400
comparative delays in depletion of ER calcium in TSD (Fig. 4) as well as in electrically evoked 401
CSD (Fig. 5). We also present further evidence of the influence of CSD on seizures, with a focus 402
on the post-ictal suppression of epileptiform activity that correlates with CSD occurrence, which 403
was also recapitulated through electrically evoked CSD (Fig. 6). This suggests post-ictal 404
suppression through CSD may serve as an innate therapeutic mechanism17 and also raises the 405
possibility of CSD as a therapeutic electrical stimulation etiology. 406
Given the potentially fast intracellular dynamics at play during CSD, we selected 407
RCatchER for our paradigm to maximize our temporal resolution, while also permitting 408
simultaneous cytosolic imaging. The vast majority of calcium indicator proteins, including 409
XCaMP-Y, are based on EF-hand motif calcium binding domains, as is the case with the 410
calmodulin of GCaMP44, which require cooperative binding and consequently exhibit non-linear 411
fluorescence dynamics33, 45. However, RCatchER has a unique calcium sensing mechanism, where 412
a single calcium binding site was engineered on the surface of the scaffold of a red fluorescent 413
protein, mApple34. This enables calcium ion binding with 1:1 stoichiometry without cooperativity, 414
whereby its change in fluorescence is not limited by a slow conformational change. Consequently, 415
RCatchER is an exceptionally fast acting calcium sensor, whose dissociation kinetics exceed the 416
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
14
temporal resolution of stopped-flow fluorescence measurements, while maintaining fluorescence 417
outputs linear to the calcium levels. This ideally positions RCatchER to capture ER dynamics. 418
Combining our dual-color imaging approach with simultaneous EEG and DC recording we 419
were able to capture single-cell neural activity and calcium homeostasis dynamics across hundreds 420
of neurons in awake mice during PTZ-induced seizures and subsequent slow propagating calcium 421
waves. It is known that CSD can follow or interrupt seizures17, 21. Large calcium increases 422
propagating as traveling waves have also been observed during CSD31, 46. Similar calcium 423
increases47, and traveling waves48-50 have been observed following seizures. While some have 424
inferred that these calcium waves are therefore indicative of CSD, here we offer definitive proof 425
of this association by corroborating these calcium waves with DC recordings. 426
An association between seizure and CSD occurrence has been demonstrated since early 427
investigations into CSD. Studies from the 1950s in anesthetized rabbit cortex showed that 428
electrically induced after-discharges, as well as PTZ-induced epileptiform activity could be 429
followed by slow potential changes and CSD51, 52. The authors hypothesized that the SD was 430
serving to arrest the seizures. They also noticed that the intensity of the epileptiform activity was 431
typically less in tissue that had experienced a prior CSD, indicative of a potential protective 432
consequence of CSD. These ideas are furthered by a PTZ kindling study in rats, where the 433
investigators found that with kindling the occurrence of CSD decreased while the occurrence of 434
epileptiform activity increased20. They, too, postulated that the CSD was arresting the seizure, but 435
added that as kindling progressed, between the evolution to a more gradual seizure onset and 436
upregulation of potassium reuptake mechanisms, the increase in extracellular potassium became 437
less abrupt and thus the probability of CSD occurrence decreased, in turn arresting fewer seizures. 438
They also noted a decrease in interictal spiking following CSD, a finding consistent with our results 439
here. Induction of CSD has also been demonstrated to suppress spike wave discharges53, 54 and 440
seizures17 in animal models, findings concordant with our demonstration of electrically induced 441
CSD decreasing post-ictal spiking (Fig. 6). 442
The large rise in intracellular calcium we observed during seizures and CSD could itself 443
have implications for seizure termination. During seizures, membrane depolarization opens 444
voltage gated calcium channels and releases magnesium block of calcium-permeable NMDA 445
glutamate receptors (NMDARs), which - coupled with excessive extracellular accumulation of 446
glutamate - causes a large and rapid influx of calcium. This results in acidification of the 447
intracellular compartment through the exchange of calcium and protons across the Ca2+/H+ 448
ATPase. This acidification, in turn, can lead to decreased conductance of voltage and ligand gated 449
channels23. As such, the acidification during a seizure has been hypothesized to promote seizure 450
termination. Additionally, excessive activation of ATPase would contribute to the depletion of 451
ATP, another potential component of seizure termination. Furthermore, the increase in 452
intracellular calcium will lead to excessive neurotransmitter release and eventual depletion of 453
synaptic vesicles, another potential factor in seizure termination32. The same processes occur 454
during CSD55, 56, although to a much greater extent. The intracellular calcium concentration is 455
estimated to rise to 6-25 µM during CSD7, 57, an order of magnitude greater than the rise of calcium 456
during seizures (700 nM)23, 58. In this study, we observed a significant release of ER calcium during 457
CSD, but not during seizures (Fig. 2). It is possible that this large ER calcium release specific to 458
CSD is contributing to the higher cytosolic calcium concentration compared to that during a 459
seizure, which could be a factor in the anti-seizure effect of CSD17. 460
Our finding that the depletion of ER stores follows the increase in cytosolic calcium, 461
presumably through voltage gated calcium channels and NMDARs, is suggestive of a CICR 462
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
15
occurring within neurons. While RCatchER and XCaMP-Y differ vastly in their kinetics, such a 463
difference in the sensors in and of themselves cannot explain the delay, with the kinetics of 464
RCatchER far exceeding those of XCaMP-Y. Calcium release from ER stores is mediated through 465
two receptor families, ryanodine receptors (RyR) and inositol (1,4,5)-triphosphate (IP3) receptors. 466
While IP3 is implicated in a variety of cell signaling cascades, RyR activation is more specific3. 467
The RyR1 isoform, which has minimal expression in nervous tissue, demonstrates voltage 468
dependence, mediated by a mechanical interaction with voltage dependent calcium channels. 469
RyR2, which is the isoform predominantly expressed in the cortex, is activated through calcium 470
influx and is voltage independent59, 60. The RyR response to calcium is biphasic, where they are 471
activated at around ~1 µM of calcium and are subsequently inactivated when calcium exceeds ~1 472
mM61-63. The calcium dependence of RyR activation may explain our finding that depletion of ER 473
calcium was small and insignificant during seizures, perhaps because intracellular calcium does 474
not reach RyR activation levels, whereas in CSD, it exceeds this threshold. Furthermore, voltage 475
dependent activation occurs faster (~2 ms) than the CICR, and while the activation of the RyR 476
channel is relatively fast (<10 ms) in the immediate presence of sufficiently high calcium, the 477
widespread induction of CICR is slower, being dependent on the rate of calcium influx and 478
diffusion through the cytosol2, 63. IP3 mediated calcium release can also occur over a similar 479
timeframe. As calcium is a co-agonist for the IP3 receptor and considered the driving force behind 480
larger concerted calcium releases, such as the one we observe here, this would also support the 481
CICR hypothesis, albeit by an alternate pathway64. Thus, the relative recruitment dynamics we 482
observe during CSD are temporally corroborative with a CICR, although further investigations are 483
needed to identify the precise receptors at play. 484
Following ER depletion, store operated calcium entry (SOCE) can occur through calcium 485
release activated channels (CRACs) at the cell plasma membrane, including the ORAI1/STIM1 486
complex, further increasing cytosolic calcium65-67. While NMDARs have been considered the 487
primary route of entry contributing to the large increase in cytosolic calcium during CSD68, SOCE 488
could be contributing to the sustained increase in cytosolic calcium observed in CSD. 489
Gain of function mutations causing ‘leaky’ RyR2 have been linked to SUDEP, with knock-490
in studies of the same mutated receptors in rodents demonstrating decreased threshold for seizures 491
and CSD69. While the effects of this mutation could be upstream of CSD by promoting cortical 492
excitability, it could also be directly impacting the generation of CSD, particularly if CICR or 493
SOCE are central to the CSD mechanism rather than only downstream consequences. 494
While we were able to make a comparison of the timing of cytosolic increase to ER calcium 495
depletion at the start of CSD, it was more difficult to make such a comparison for the offset of the 496
CSD. This is primarily due to the gradual and often smaller calcium changes occurring at the end 497
of the CSD, leading to less precision in the selection of an offset time. However, this variability is 498
small when compared with the length of CSD and with the offset event duration across the 499
population and thus does not dramatically impact the spatial regression for these slow traveling 500
waves. Mechanistically, irrespective of the imprecision in measurement, the timing of these two 501
events are within an acceptable range for the expected timing of cytosolic calcium return to the 502
ER through the sarco(endo)plasmic reticulum ATPase (SERCA) pump70. 503
It is important to note that other ER dynamics are occurring during CSD aside from the 504
observed calcium depletion that could have implications for our findings. ER is known to undergo 505
morphological change during CSD: fission and fragmentation of ER occurs in a calcium 506
calmodulin-dependent kinase II (CaMKII)-dependent manner46, which can potentially contribute 507
to the RCatchER signal we measured in this study. However, we deduce that the ER calcium signal 508
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
16
we measured in this study was not significantly affected by the morphological change of ER due 509
to the slow nature of ER fission and fusion. The RCatchER signal decreased within a second of 510
the increase in cytosolic calcium (Fig. 3), whereas the ultrastructural changes in the ER were found 511
to occur several seconds after the calcium influx, indicating that the ER calcium release precedes 512
the fission. Additionally, the RCatchER signal recovered within a minute following the depletion 513
of ER calcium (Fig. 3), while the ER fusion should take several minutes to restore continuity of 514
ER. Even if ER is fragmented into beaded structures, that by itself should not hinder the calcium 515
dependence of RCatchER fluorescence. Furthermore, the ER beading occurs predominantly in the 516
neuropil, while our imaging analysis primarily focused on somata. Given the calcium dependence 517
of CaMKII, perhaps the release of ER calcium stores we describe here facilitates the fission event. 518
Another consideration for our finding is the potential impact of intracellular pH on 519
RCatchER’s excitability. While intracellular acidification does occur during seizures, the time 520
course is a gradual change throughout the seizure and continuing during the post-ictal phase71, 521
rather than an abrupt drop post-ictally. During CSD acidification also occurs although there is a 522
delay in the drop in pH relative to the DC shift and the pH decrease is sustained for longer than 523
the DC shift72-74. Therefore, the dynamics we observe here are not temporally concordant with the 524
pH changes observed. Furthermore, the ER is well buffered, having little change in pH during 525
calcium release and experiences much smaller magnitude pH changes than the cytosol during 526
intracellular acidification75. 527
TSD is an anoxic variant of CSD occurring during death76, 77, known to have slightly 528
different mechanisms underlying its calcium dynamics. However, we still observed a depletion of 529
ER calcium stores following a large cytosolic increase, albeit more delayed. Under severe hypoxia, 530
hyperpolarization as a nonspreading depression occurs in the brain, observed as an isoelectric 531
EEG, preserving the ATP stores necessary for recovery. If, however, the hypoxic conditions last 532
more than a few minutes, the ATP stores become depleted, the ion gradients across the membranes 533
break down, and a TSD is, in turn, initiated7, 55. While calcium and NMDAR have been found to 534
be necessary for CSD occurrence in normoxic conditions, in the context of hypoxia, CSD can 535
occur in the absence of extracellular calcium and NMDAR function57, 78, 79. We observed such 536
TSD calcium waves after about a minute of isoelectric EEG during fatal seizure recordings (Fig. 537
4). While the propagation patterns were similar to CSDs in surviving animals with respect to 538
overlapping direction and speed, the delay in the ER depletion following the increase in cytosolic 539
calcium was slightly longer during TSD. The increase in cytosolic calcium fluorescence was 540
smaller during TSD than other CSD, perhaps contributing to this longer delay in a CICR. 541
Consistent with cell death and ATP depletion, the cytosolic calcium increase was sustained, and 542
ER calcium was not restored. While cells can typically tolerate the length of depolarization and 543
elevated calcium during a seizure or normoxic CSD55, with some neuroprotective effect even 544
having been demonstrated for the intracellular calcium concentrations reached during seizures80,
545
81, the levels experienced during CSD when sustained, as is the case in anoxia and TSD, are 546
generally regarded to be toxic3, 68. 547
Our findings support the plausibility of calcium homeostasis dysregulation during CSD. 548
However, further investigation is needed to determine the necessity of CICR in CSD and its seizure 549
suppressive effect. If CICR proves to be mechanistically involved, evidence of its exact mediators, 550
be it RyR or IP3 dependent, could provide us with new targets for modulation, informing our anti-551
epileptic arsenal. While CSD may be beneficial in the context of widespread aberrant brain activity 552
during a seizure, it can most certainly have pathologic consequences. Being able to mimic the anti-553
seizure effect of CSD while avoiding its toxic side effects could offer great therapeutic potential. 554
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
17
Alternatively, if CICR rather contributes to the prolonged depolarization and toxic consequences 555
of CSD, prevention of this calcium store depletion during CSD could also be of benefit. 556
The clinical implications of our investigation afford insight not only to potential 557
neuromodulatory targets, but also to hypothetical mechanisms underlying established therapeutics, 558
namely electrical stimulation used for seizure control. Notably, we were able to induce CSD using 559
different stimulation parameters, including ones paralleling typical RNS settings. Indeed, repeated 560
cortical stimulation has been shown to cause increases in extracellular potassium, likely 561
contributing to CSD induction82. The clinical benefits of RNS are likely multifaceted. In large part 562
the decreased seizure incidence is hypothesized to be mediated by neuroplasticity83, rather than 563
direct arrest of seizures, as seizures in most patients are not terminated by RNS, and the majority 564
of stimulation occurs interictally. However, in those patients for whom seizures are arrested, these 565
results raise the possibility that CSD could be occurring and contributing to the suppressive effect. 566
567
MATERIALS AND METHODS 568
569
Molecular Biology 570
An AAV2 transfer plasmid containing the XCaMP-Y-P2A-RCatchER cassette was generated 571
through standard molecular biology (restriction enzyme [RE] digestion, ligation, transformation, 572
and plasmid purification). The RCatchER cassette in the pcDNA3.1 vector34 was cut with NheI 573
and EcoRV and ligated into the NheI and blunted HindIII RE sites in an AAV2 transfer vector 574
with the human synapsin I (hSynI) promoter (Addgene plasmid number: 100843), resulting in 575
pAAV2/hSynI-RCatchER. The coding sequence of XCaMP-Y33 (GenBank accession number: 576
MK770163.1) was fully synthesized through a commercial service with an NheI RE recognition 577
site added at the 5’ end (gBlocks gene fragments, Integrated DNA Technologies). The NheI and 578
ClaI fragment of XCaMP-Y was ligated together with the ClaI and BamHI fragment of separately 579
synthetized P2A sequence into the respective RE sites upstream of RCatchER, resulting in 580
pAAV2/hSynI-XCaMP-Y-P2A-RCatchER. The correct sequence was confirmed through RE 581
analysis, as well as Sanger and next-generation sequencing (Eton Biosciences and Plasmidsaurus, 582
respectively). Endotoxin-free DNA was obtained through a commercial midiprep plasmid 583
purification kit (NucleoBond Xtra Midi EF, Takara Bio USA). 584
585
Viral Vector Production 586
We produced a recombinant adeno-associated viral (rAAV) vector pseudotyped with AAV9 capsid 587
protein in-house following a modified standard procedure84. In brief, we transfected human 588
embryonic kidney 293FT cells (Invitrogen) with three plasmids (helper [Addgene plasmid number: 589
112867], AAV2/9 rep/cap [Addgene plasmid number: 112865], and pAAV2/hSynI-XCaMP-Y-590
P2A-RCatchER) at 1:1:1 molar ratio using the calcium-phosphate method. We then harvested the 591
rAAV, purified through iodixanol gradient ultracentrifugation, and concentrated in Dulbecco’s 592
phosphate-buffered saline (D8537, Millipore-Sigma) supplemented with 0.001% (v/v) Pluronic F-593
68 (Millipore-Sigma). We aliquoted and stored it at -80°C until surgery. Using quantitative PCR, 594
we determined the titer to be 4×1014 viral genomes/mL. 595
596
Cranial Window Fabrication 597
To generate windows for chronic imaging with the capability for repeated access via a pulled glass 598
micropipette electrode for DC recording, we adapted protocols for concentric window design85 599
with single plane windows along with silicone access ports86, 87. We used two smaller inner layer 600
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
18
coverslips and one larger outer layer coverslip (3 mm and 5 mm diameter, respectively, #1 601
thickness, Warner Instruments), and generated a 0.7 mm diameter access port placed 0.75 mm 602
from the center (Supplementary Fig. 1a). When deciding on diameter of the window it is important 603
to consider the planned angle of penetration for the recording electrode, along with the thickness 604
of the electrode and depth of the access port so as to ensure the window has a large enough diameter 605
to accommodate these constraints (Supplementary Fig. 1b). We began by first etching the holes in 606
each glass layer at the correct coordinates to ensure the holes alignment (0.75 mm from the side of 607
the 3mm glass and 1.75 mm from the edge of the 5mm glass) when the glass layers are stacked 608
concentrically. For etching we suspended the glass using a spring hinged clip to apply light, but 609
sufficient pressure to the edges of the cover glass. To protect the glass from damage, we covered 610
the clip with heat shrink wrap. Using a conical sharp tip fine grit stone grinding burr (CA1063, 611
Minimo Precision Instruments and Tools) with a dental handpiece at a medium speed (~6000-7000 612
rpm), we slowly hand dry etched each cover glass at a 45o angle, moving halfway through each 613
piece of glass from each side to meet in the center, with applying dust free air routinely to clear 614
away the silicone dust during etching. Upon meeting in the center, we moved the burr into a 615
vertical position (perpendicular to the cover glass) to round out and straighten the beveled edges 616
from both sides of the glass as needed to minimize imaging artifacts. We then cleaned each cover 617
glass with lens paper and compressed air. 618
We next assembled the cover glass using optical adhesive (Optical Adhesive 71, Norland), 619
on a Styrofoam platform with vertical pins (000 insect pins; 0.5 mm shaft diameter) to thread the 620
holes for cover glass alignment. We adhered the layers serially using a small amount of adhesive 621
(just enough to spread to the edge of the 3mm glass), first the 5mm to a 3mm and then the second 622
3mm to the first, with interleaved UV curing (365 nm; wavelength of peak absorbance) for 30 min. 623
We then cured overnight (>12 h) at 50oC. The optical adhesive must be uniformly distributed 624
between articulating surfaces to prevent thin film interference. 625
The final step is to prepare the silicone membrane by filling the access port with optically 626
transparent silicone (Sylgard 184, Dow Corning), selected because it can be cured at room 627
temperature given the initial limited temperature range tolerance of the optical adhesive (-15-628
60oC). We prepared the silicone at a 10:1 mixture by weight of base to catalyst. Next, we increased 629
the viscosity of the silicone through heating to facilitate easier application to the port using heat 630
gun at 150-200oC for about a minute. Then using a 30-gauge insulin needle we applied a tiny drop 631
of the prepared silicone to fill the access port of the windows while suspending them in mid-air by 632
their edges with light tension to prevent wicking. We then transferred the suspended windows to 633
a vacuum chamber and cured at room temperature for two days to prevent bubble formation in the 634
silicone. We then followed this with a one-day cure at 50oC. Curing at a higher temperature 635
increases the silicone’s strength (higher shear modulus), while curing at a lower temperature 636
increases its elasticity (lower strain at break)88. We designed this approach to ensure a balance of 637
strength and elasticity such that the membrane will not deform under increased intracranial 638
pressure (strength), while at the same time will properly re-seal upon needle withdrawal 639
(elasticity). We stored these windows at room temperature. For a full cure we waited an additional 640
four days and implanted them within a few months, before the silicone dried out, becoming more 641
brittle and losing its elasticity. 642
643
Stereotaxic Surgery 644
All procedures involving live animals were conducted with approval from and in accordance with 645
Emory University’s Institutional Animal Care and Use Committee. Adapting standard protocols 646
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
19
for concentric cranial window implantation85, 89, we performed stereotaxic cranial window surgery 647
on adult (³ 90 day-old) albino C57BL6/N male mice (B6N-Tyrc-Brd/BrdCrCrl, Charles River, 648
Strain Code 493) concurrently with intracortical delivery of rAAV, electrode placement for 649
recording and stimulation, and headplates affixation. In brief, we secured the mice in a stereotaxic 650
frame and maintained them under anesthesia (1.5% isoflurane balanced in oxygen (1 L/min)). We 651
then performed a 3 mm craniotomy over the primary motor cortex. Two injections of rAAV (500 652
nL each; 2 nL/s) were performed through a pulled glass capillary tube (Nanoject 3.0, Drummond) 653
at 300 µm and 600 µm deep to the pial surface (0.30 mm anterior and 1.75 mm lateral to Bregma)90,
654
91. We then epidurally placed a thin polyamide insulated tungsten wire electrode with exposed tip 655
(125 µm; P1Technologies) at the posteromedial edge of the craniotomy, along with ipsilateral 656
reference and contralateral ground stainless steel screw electrodes (E363/96/1.6/SPC, 657
P1Technologies) in the skull over the cerebellum (0.7 mm burr holes). If the mouse was to be used 658
for stimulation as well, we also placed two additional wire electrodes (same material and size) 659
epidurally about half a millimeter apart along the posterolateral edge of the craniotomy. We 660
prefabricated all electrodes with gold pins to facilitate easy attachment to a recording preamplifier 661
or stimulus isolator. We next placed a cranial window to plug the craniotomy and affixed the 662
window and electrodes to the skull using dental acrylic (C&B Metabond, Parkell) along with a 663
stainless steel headplate (Models 3 and 4; Neurotar). We closed the skin to headplate using tissue 664
adhesive (Vetbond, 3M). 665
666
Two-Photon Imaging and Electrophysiology 667
We performed resonant scanning (30 Hz, 512x512 pixels) two-photon imaging on mice during 668
acute induced seizures. We used a two-photon microscope (HyperScope, Scientifica) equipped 669
with a pulsed tunable infrared laser system (InSight X3, Spectra-Physics) and controller software 670
(ScanImage, Vidrio Technologies). We selected 1000-1010 nm wavelengths for excitation (Fig. 671
1C) and separated the emissions by a dichroic mirror (565LP, Chroma) with band pass filters 672
(ET525/50m-2p and ET620/60m-2p, Chroma), collecting the light using GaAsP and multi-alkali 673
red-shifted photomultiplier tube, respectively. Beginning one month following surgery to allow 674
for adequate GECI expression, we head-fixed the mice in a carbon fiber airlifted chamber (Mobile 675
HomeCage, Neurotar), positioned under a long working distance 16x objective (water-immersion, 676
N.A. 0.80, Nikon). We connected the EEG electrodes to an AC preamplifier and a data acquisition 677
system (Sirenia, Pinnacle Technologies). For DC recording, on the day of a recording session, a 678
reference electrode (1-mm diameter Ag/AgCl pellet, Model E205, Harvard Apparatus) was placed 679
in the nuchal muscle of the mouse under isoflurane anesthesia. A micromanipulator (Kopf 680
Instruments) was affixed to the crossbar of the headplate holder. We stereotaxically inserted a 681
long-shank pulled glass electrode (1 mm diameter with ~70 µm at the tip, 1 – 3 MΩ when filled 682
with the following solution: 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 683
20 mM d(+)-glucose, 2 mM CaCl2, and 1.3 mM MgCl2) through the access port at a 45o angle into 684
the cortex to a depth of 100-300 µm below the pial surface. DC signals were recorded through a 685
patch clamp amplifier, a digital data acquisition system, and software (MultiClamp 700B, Digidata 686
1550B, and pClamp, respectively, Molecular Devices). Both EEG and DC signals were recorded 687
at 2 kHz, with a 0.5-300 Hz bandpass filter for the EEG and with a 500 Hz low-pass filter for the 688
DC recordings. EEG was continuously recorded while DC recording and two-photon imaging were 689
triggered by a TTL pulse generated in the EEG recording system immediately before either PTZ 690
injection or electrical stimulation. 691
692
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
20
Acute Seizure Model 693
We subcutaneously injected PTZ (40-50 mg/kg; P6500, Millipore-Sigma; sterile saline diluent) 694
and recorded for 20-45 min depending on the course of the seizure. All but one PTZ injection 695
resulted in at most one generalized seizure per recording session. We performed multiple seizure 696
experiments within the same subject provided the seizures were not fatal, with sessions separated 697
by at least a week to circumvent the effects of kindling92. 698
699
Electrical Stimulation 700
We performed electrical stimulation using a waveform and function generator (EDU33210A 701
Keysight, USA) to drive a stimulus isolator (DS4 or DS5, Digitimer, UK) attached to the 702
chronically implanted electrodes. Typical stimulation parameters were 50% duty cycle bipolar 703
pulses with amplitudes of ±200 µA at 2 kHz for 10 s. We stimulated the mice 30-40 s following 704
the start of image acquisition and would acquire 10-15 min of data depending on the course of the 705
CSD. To evaluate the threshold for inducing CSD with RNS style parameters we created 706
waveforms in MATLAB and then used the generator to drive them. We began stimulation at the 707
lowest setting and proceeded to increase current with each subsequent stimulation periods (25, 50, 708
100, 250, 500 and 750 µA) until CSD was induced, interwoven by 2-minute washout periods. 709
Following CSD we had a 20 min washout before the trial was repeated within the same subject. 710
For post-ictal simulation-induced CSD experiments we performed electrical stimulation 711
using the same approach as above in a subset of animals during seizures occurring without CSD. 712
In these recordings we waited 5 min following the end of the seizure to ensure a CSD did not occur 713
naturally and to acquire a baseline post-ictal spiking. We then stimulated to induced CSD and 714
recorded for at least an additional 5 min. 715
716
Histology 717
Following experiments, all subjects underwent transcardial perfusion (4% paraformaldehyde 718
(PFA) in phosphate buffered saline (PBS), 4oC; 32% PFA stock solution, Electron Microscopy 719
Solutions, cat no. 15714) and their brain tissue was extracted. The brains were further fixed 720
overnight in 4% PFA (4oC) and then cryoprotected for 36 hours in 30% sucrose (in PBS, 4oC). 721
The tissue was serially sectioned using a freezing microtome (Spencer Lens Co. equipped with a 722
Physitemp BFS-40MPA Controller and platform) at 40 µm thickness (coronal) and stored in PBS 723
at 4oC. 724
For triple immunofluorescence, free-floating sections were rinsed in PBS, blocked in PBS 725
solution containing 4% normal donkey serum (NDS), 4% normal goat serum (NGS) and 0.1% 726
Triton-X for 30 minutes at room temperature. After rinses in PBS, sections were then incubated 727
overnight at 4ºC with a combination of chicken anti-GFP (1:100, GFP-1020, Aves Labs, Davis, 728
CA, USA), mouse-anti-mCherry (1:200, AE002, Abclonal, Woburn, MA, USA), and rabbit anti-729
SERCA2 (1:50, A1097, Abclonal, Woburn, MA, USA) in PBS containing 2% NDS and 2% NGS. 730
Sections were rinsed in PBS and incubated in a PBS solution containing Alexa Fluor 488-731
conjugated goat anti-Chicken IgG (1:1000, Invitrogen A-11039, ThermoFisher Scientific, 732
Waltham, MA, USA), Alexa Fluor 594-conjugated donkey anti-mouse IgG (1:1000, 715-585-150, 733
Jackson Immunoresearch Laboratories, West grove, PA, USA) and Alexa Fluor 647-conjugated 734
goat anti-rabbit IgG (1:500, 111-607-008, Jackson Immunoresearch Laboratories, West grove, PA, 735
USA) secondary antibodies and 2% NDS/2% NGS for 1 hour at room temperature. Sections were 736
rinsed with PBS, then mounted on glass slides, dried and cover slipped with hard set mounting 737
medium containing the nuclear marker, DAPI (Vectashield, H-1500, Vector Laboratories, 738
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
21
Burlingame, CA, USA). Images were captured using a Leica SP8 upright confocal microscope 739
and LASX software. Immunofluorescence image processing, including projections and orthogonal 740
view generation, was performed using FIJI93. 741
742
Image Processing 743
We performed imaging pre-processing using the Suite2P software package94 with integrated 744
Cellpose95, performing motion registration, region of interest (ROI) detection and calcium 745
transient extraction of soma and surrounding neuropil. We additionally manually curated the 746
candidate ROIs to ensure they were from cell bodies and not overlapping with any major blood 747
vessels. We background subtracted the raw fluorescence traces taking the global minima of the 748
neuropil transients for each recording as a proxy for the background signal. We also generated 749
clean somatic signal by subtracting 70% of the surrounding neuropil signal from the somatic ROI 750
signal to remove out-of-plane contamination94 and adjust for photobleaching. We normalized the 751
traces as DF/F0, using the mean of the first 30 s of recording as baseline fluorescence (F0). Finally, 752
we double-reverse filtered the traces using an adaptive filtering routine (1 Hz, Butterworth lowpass 753
filter, order 3 to 5) that preserves the phase of the signal prior to processing. 754
755
Data Analysis 756
For baseline recording spike detection we employed a peak detection heuristic to select all 757
transients at least four standard deviations above baseline. We computed the average spike trace 758
with pooled standard deviation, aligning all the detected transients by their point of maximum 759
slope. The change in calcium level during each spike was taken as the difference between the 760
average signal during half-second windows before and after the recruitment time. 761
For recruitment time detection during seizures, we used our previously reported automated 762
approach28. In brief, the approach uses the mean population calcium trace along with EEG (in this 763
study we also used DC) to find PIS, seizure and CSD onsets in the recordings and defines detection 764
windows around these events. It then searches the individual traces within these windows for local 765
maxima in the integral of their slope, indicative of rapid and sustained increases in cytosolic 766
calcium that occur during recruitment to these seizure related events. These increases must exceed 767
an event and channel specific threshold above baseline for the cell to be considered recruited to 768
the event. All individual cell recruitments are then indexed by the point of steepest slope, being 769
generally accepted as a recruitment time for seizures96, 97, and the point least likely to be perturbed 770
by filtering. For processing the RCatchER signal, we inverted the transients, enabling our 771
algorithm, originally designed to find increases in cytosolic calcium, to find the decreases in signal. 772
With the majority of recordings and events we used the clean somatic traces for detection. We also 773
adapted this method to determine recruitment times of CSD offset. However, due to lower signal-774
to-noise ratio during the offset of CSD and during TSD, we needed to use the somatic traces before 775
neuropil subtraction for event detection. Additionally, for determining the offset of CSD a few 776
modifications to the approach were implemented given the gradual rather than rapid nature of the 777
change being measured. Namely we filtered the traces to 0.05-0.1 Hz and searched for the local 778
minima/maxima of concavity to estimate the beginning of the offset calcium change. 779
To determine the magnitude of calcium changes in individual cells during seizures and 780
CSD (including TSD), we computed the difference between the average signal during ten and five 781
second windows, respectively, before and after the recruitment times. For the calcium change 782
during the recruitment to PIS, we also computed the difference in calcium levels before and during 783
the spike. However, we found that using the average signal diminished the amplitude of change 784
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
22
too much and using only the maxima left the values too subject to shot noise. Therefore, we opted 785
to further filter the traces to 0.5 Hz before taking the maxima during a one second window before 786
the spike to compare with a two second window encompassing the full spike. To determine the 787
impact of event (PIS, seizure or CSD) on calcium change we model the calcium changes using a 788
GLME with effects coding and with each recording session modeled as a random effect. Then to 789
evaluate the if there was a sustained post-ictal calcium change in an unbiased way, we determined 790
the average post ictal calcium signal in each cell over a sliding 15 s window for each seizure 791
recording. Then using the same GLME approach we modeled the impact of CSD on the post-ictal 792
calcium. 793
For determining vectors of propagation, as we previously demonstrated28 we applied a 794
spatial linear regression algorithm, with L1 regularization (originally developed to model interictal 795
events in human intracranial data)37, 38, to our data, using the positions of the cells withing the field 796
and the determined event recruitment times. We only considered statistically significant vectors 797
(p<0.05; although for nearly all of the vectors p<0.001) in our analysis, where p-values were 798
computed by comparing model residuals to spatially shuffled data sets, as in prior work 37. Speed 799
and direction of propagation were taken from these vectors for comparisons within and across 800
channels. The same recruitment times used to determine these vectors were also used to compare 801
the relative timings of recruitment to these events across channels, modeling the effect of channel 802
on recruitment time using a GLME with effects coding and with each recording session modeled 803
as a random effect. 804
For computing the DC shift onset and offset we first smoothed the DC trace using a rolling 805
average, downsampling and interpolation. We took the local minima near the DC fall and local 806
maxima near the DC rise of the second derivative of the trace as the onset and offset times, 807
respectively. We used these times to compute the DC shift period lengths and latencies with respect 808
to the calcium changes. 809
We produced all movies and time-lapse representative frames from the motion registered 810
image stacks output of Suite2P processed using FIJI93. We filtered the stacks using a 3D gaussian 811
filter (X-Y: 0.5 standard deviation (SD); time: 1 SD) and down sampled to 6 Hz (movie) or 1 Hz 812
(representative frames) using bilinear interpolation. 813
To examine the SWD rates in EEG, we used a bandpower-based threshold detection 814
method28, 98 to find all EEG spikes in a recording. We then divided the recordings into the specific 815
periods we wished to compare. We used the same pre-ictal period as in the calcium data. We then 816
used Welch’s power spectral density to compute a spectrogram and used specific frequency 817
features to define the boundaries of the post ictal time periods being compared. The post-ictal 818
period began at the end of the seizure, defined as the point when the total power (<100 Hz) fell 819
below 5% of the maximum power achieved during the seizure. For recordings without post-ictal 820
stimulation, we ended the period 5 min later, being equivalent to when we would induce CSD in 821
the post ictal stimulation recordings. For recordings with post-ictal stimulation, we defined the end 822
of the post-ictal/pre-stimulation period as the point where a stimulation artifact EEG power (50-823
70 Hz & 170-190 Hz) crossed 50% of its maximum power. The post-ictal/post-stimulation period 824
began when the stimulation artifact power fell below 50% of its maximum and ended 5 min later. 825
We computed SWD rates by dividing the totals spike counts during each of these periods by their 826
length of time. We then computed moving-average SWD rate curves (Fig. 6 a-b) by convolving a 827
30 s-wide Gaussian window with a binarized array of detected spike times at the original signal 828
sampling rate. Statistics were performed using non-parametric pairwise tests, namely the Wilcoxon 829
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
23
sign-rank test for comparisons of two groups and the Friedman test for comparisons of more than 830
two groups. 831
832
Two-photon Excitation Spectra 833
For determining the excitation spectra of the two indicators, we imaged a mouse prepared for in 834
vivo imaging, expressing both XCaMP-Y and RCatchER, using the same approach as above, 835
except with galvo scanning (1.07 Hz, 512x512 pixels). We collected images across a series of 836
wavelengths (800-1250 nm; 10 nm interval) in both channels simultaneously (10 frames per 837
wavelength; saved as a time averaged projection). The laser attenuation was calibrated to maintain 838
constant power at the sample across the spectra, adjusting for wavelength-dependent laser output 839
and attenuation variance. Laser power and PMT gains were calibrated to prevent saturation of 840
PMTs at peak excitation values (XCaMP-Y: 970nm, RCatchER: 1100 nm), while ensuring 841
sufficient signal could be observed at 1010 nm. We recorded pre- and post-recording power 842
measurements and calibration frames to verify that power did not attenuate over the course of the 843
experiment, photobleaching did not occur and there was no degradation of the photodiodes. For 844
processing, we concatenated the image series using FIJI and performed ROI detection and transient 845
extraction using Suite2P. We background subtracted the traces (including removing 846
autofluorescence contamination, likely due to the older age of the mouse used), normalized these 847
to maximum power and performed a cubic interpolation between the discrete emission values to 848
produce the spectra curves. 849
850
REFERENCES 851
852
1. Mekahli, D., Bultynck, G., Parys, J., De Smedt, H. & Missiaen, L. Endoplasmic-853
Reticulum Calcium Depletion and Disease. Cold Spring Harbor Perspectives in Biology 3 854
(2011). 855
2. Verkhratsky, A. & Shmigol, A. Calcium-induced calcium release in neurones. Cell 856
Calcium 19, 1-14 (1996). 857
3. Ghosh, A. & Greenberg, M.E. Calcium signaling in neurons - molecular mechanisms and 858
cellular consequences. Science 268, 239-247 (1995). 859
4. Berridge, M. Neuronal calcium signaling. Neuron 21, 13-26 (1998). 860
5. Reddish, F., Miller, C., Gorkhali, R. & Yang, J. Calcium Dynamics Mediated by the 861
Endoplasmic/Sarcoplasmic Reticulum and Related Diseases. International Journal of Molecular 862
Sciences 18 (2017). 863
6. Suzuki, J., Kanemaru, K. & Iino, M. Genetically Encoded Fluorescent Indicators for 864
Organellar Calcium Imaging. Biophysical Journal 111, 1119-1131 (2016). 865
7. Dreier, J.P. & Reiffurth, C. The Stroke-Migraine Depolarization Continuum. Neuron 86, 866
902-922 (2015). 867
8. Leao, A.A.P. Spreading depression of activity in the cerebral cortex. Journal of 868
Neurophysiology 7, 359-390 (1944). 869
9. Somjen, G.G. Aristides Leao's discovery of cortical spreading depression. Journal of 870
Neurophysiology 94, 2-4 (2005). 871
10. Leao, A.A.P. Further Observations on the Spreading Depression of Activity in the 872
Cerebral Cortex. Journal of Neurophysiology 10, 409-414 (1947). 873
11. Hills, K.E., Kostarelos, K. & Wykes, R.C. Converging Mechanisms of Epileptogenesis 874
and Their Insight in Glioblastoma. Frontiers in Molecular Neuroscience 15, 23 (2022). 875
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
24
12. Dohmen, C., et al. Spreading depolarizations occur in human ischemic stroke with high 876
incidence. Annals of Neurology 63, 720-728 (2008). 877
13. Lauritzen, M., et al. Clinical relevance of cortical spreading depression in neurological 878
disorders: migraine, malignant stroke, subarachnoid and intracranial hemorrhage, and traumatic 879
brain injury. Journal of Cerebral Blood Flow and Metabolism 31, 17-35 (2011). 880
14. Strong, A.J., et al. Spreading and synchronous depressions of cortical activity in acutely 881
injured human brain. Stroke 33, 2738-2743 (2002). 882
15. Curry, R.N., et al. Glioma epileptiform activity and progression are driven by IGSF3-883
mediated potassium dysregulation. Neuron 111, 682-+ (2023). 884
16. Kramer, D.R., Fujii, T., Ohiorhenuan, I. & Liu, C.Y. Interplay between Cortical 885
Spreading Depolarization and Seizures. Stereotactic and Functional Neurosurgery 95, 1-5 886
(2017). 887
17. Tamim, I., et al. Spreading depression as an innate antiseizure mechanism. Nature 888
Communications 12, 1-15 (2021). 889
18. Calia, A.B., et al. Full-bandwidth electrophysiology of seizures and epileptiform activity 890
enabled by flexible graphene microtransistor depth neural probes. Nature Nanotechnology 17, 891
301-309 (2022). 892
19. Aiba, I., Ning, Y. & Noebels, J.L. A hyperthermic seizure unleashes a surge of spreading 893
depolarizations in Scn1a-deficient mice. Jci Insight 8, 19 (2023). 894
20. Koroleva, V.I., Vinogradova, L.V. & Bures, J. Reduced incidence of cortical spreading 895
depression in the course of pentylenetetrazol kindling in rats. Brain Research 608, 107-114 896
(1993). 897
21. Dreier, J.P., et al. Spreading convulsions, spreading depolarization and epileptogenesis in 898
human cerebral cortex. Brain 135, 259-275 (2012). 899
22. Fabricius, M., et al. Association of seizures with cortical spreading depression and peri-900
infarct depolarisations in the acutely injured human brain. Clinical Neurophysiology 119, 1973-901
1984 (2008). 902
23. Raimondo, J.V., Burman, R.J., Katz, A.A. & Akerman, C.J. Ion dynamics during 903
seizures. Frontiers in Cellular Neuroscience 9, 14 (2015). 904
24. Traynelis, S.F. & Dingledine, R. Role of Extracellular-Space in Hyperosmotic 905
Suppression of Potassium-Induced Electrographic Seizures. Journal of Neurophysiology 61, 927-906
938 (1989). 907
25. McBain, C.J., Traynelis, S.F. & Dingledine, R. Regional Variation of Extracellular-Space 908
in the Hippocampus. Science 249, 674-677 (1990). 909
26. Somjen, G.G. Ions in the Brain: Normal Function, Seizures, and Stroke (Oxford 910
University Press, New York, 2004). 911
27. Aiba, I. & Noebels, J.L. Spreading depolarization in the brainstem mediates sudden 912
cardiorespiratory arrest in mouse SUDEP models. Science Translational Medicine 7, 9 (2015). 913
28. Stern, M.A., Cole, E.R., Gross, R.E. & Berglund, K. Seizure event detection using 914
intravital two-photon calcium imaging data. Neurophotonics 11, 024202 (2024). 915
29. Basarsky, T.A., Duffy, S.N., Andrew, R.D. & MacVicar, B.A. Imaging spreading 916
depression and associated intracellular calcium waves in brain slices. Journal of Neuroscience 917
18, 7189-7199 (1998). 918
30. Nedergaard, M., Cooper, A.J.L. & Goldman, S.A. Gap-Junctions are Required for the 919
Propagation of Spreading Depression. Journal of Neurobiology 28, 433-444 (1995). 920
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
25
31. Enger, R., et al. Dynamics of Ionic Shifts in Cortical Spreading Depression. Cerebral 921
Cortex 25, 4469-4476 (2015). 922
32. Lado, F.A. & Moshé, S.L. How do seizures stop? Epilepsia 49, 1651-1664 (2008). 923
33. Inoue, M., et al. Rational Engineering of XCaMPs, a Multicolor GECI Suite for In Vivo 924
Imaging of Complex Brain Circuit Dynamics. Cell 177, 1346-1360 (2019). 925
34. Deng, X.N., et al. Tuning Protein Dynamics to Sense Rapid Endoplasmic-Reticulum 926
Calcium Dynamics. Angewandte Chemie-International Edition 60, 23289-23298 (2021). 927
35. André, V., Pineau, N., Motte, J.E., Marescaux, C. & Nehlig, A. Mapping of neuronal 928
networks underlying generalized seizures induced by increasing doses of pentylenetetrazol in the 929
immature and adult rat:: a c-Fos immunohistochemical study. European Journal of Neuroscience 930
10, 2094-2106 (1998). 931
36. Nathanson, J.L., Yanagawa, Y., Obata, K. & Callaway, E.M. Preferential labeling of 932
inhibitory and excitatory cortical neurons by endogenous tropism of adeno-associated virus and 933
lentivirus vectors. Neuroscience 161, 441-450 (2009). 934
37. Liou, J.Y., et al. Multivariate regression methods for estimating velocity of ictal 935
discharges from human microelectrode recordings. Journal of Neural Engineering 14, 11 (2017). 936
38. Smith, E.H., et al. Human interictal epileptiform discharges are bidirectional traveling 937
waves echoing ictal discharges. Elife 11, 20 (2022). 938
39. Marshall, W.H. Spreading cortical depression of Leao. Physiological Reviews 39, 239-939
279 (1959). 940
40. Heck, C.N., et al. Two-year seizure reduction in adults with medically intractable partial 941
onset epilepsy treated with responsive neurostimulation: Final results of the RNS System Pivotal 942
trial. Epilepsia 55, 432-441 (2014). 943
41. Geller, E.B., et al. Brain-responsive neurostimulation in patients with medically 944
intractable mesial temporal lobe epilepsy. Epilepsia 58, 994-1004 (2017). 945
42. Hogan, P.G., Lewis, R.S. & Rao, A. Molecular Basis of Calcium Signaling in 946
Lymphocytes: STIM and ORAI. in Annual Review of Immunology, Vol 28 (ed. W.E. Paul, D.R. 947
Littman & W.M. Yokoyama) 491-533 (Annual Reviews, Palo Alto, 2010). 948
43. Gilbert, G., et al. Calcium Signaling in Cardiomyocyte Function. Cold Spring Harbor 949
Perspectives in Biology 12, 29 (2020). 950
44. Akerboom, J., et al. Crystal Structures of the GCaMP Calcium Sensor Reveal the 951
Mechanism of Fluorescence Signal Change and Aid Rational Design. Journal of Biological 952
Chemistry 284, 6455-6464 (2009). 953
45. Ali, F. & Kwan, A.C. Interpreting in vivo calcium signals from neuronal cell bodies, 954
axons, and dendrites: a review. Neurophotonics 7, 17 (2020). 955
46. Kucharz, K. & Lauritzen, M. CaMKII-dependent endoplasmic reticulum fission by 956
whisker stimulation and during cortical spreading depolarization. Brain 141, 1049-1062 (2018). 957
47. Khoshkhoo, S., Vogt, D. & Sohal, V.S. Dynamic, Cell-Type-Specific Roles for 958
GABAergic Interneurons in a Mouse Model of Optogenetically Inducible Seizures. Neuron 93, 959
291-298 (2017). 960
48. Tran, C.H., et al. Interneuron Desynchronization Precedes Seizures in a Mouse Model of 961
Dravet Syndrome. Journal of Neuroscience 40, 2764-2775 (2020). 962
49. Farrell, J.S., et al. In vivo assessment of mechanisms underlying the neurovascular basis 963
of postictal amnesia. Scientific Reports 10, 13 (2020). 964
50. Heuser, K., et al. Ca2+ Signals in Astrocytes Facilitate Spread of Epileptiform Activity. 965
Cerebral Cortex 28, 4036-4048 (2018). 966
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
26
51. Van Harreveld, A. & Stamm, J.S. Consequences of cortical convulsive activity in rabbit. 967
Journal of Neurophysiology 17, 505-520 (1954). 968
52. Van Harreveld, A. & Stamm, J.S. Cortical response to metrazol and sensory stimulation 969
in the rabbit. Electroencephalography and Clinical Neurophysiology 7, 363-370 (1955). 970
53. Vergnes, M. & Marescaux, C. Cortical and thalamic lesions in rats with genetic absence 971
epilepsy. Journal of Neural Transmission-General Section, 71-83 (1992). 972
54. Vinogradova, L., Kuznetsova, G. & Coenen, A. Audiogenic seizures associated with a 973
cortical spreading depression wave suppress spike-wave discharges in rats. Physiology & 974
Behavior 86, 554-558 (2005). 975
55. Somjen, G.G. Mechanisms of spreading depression and hypoxic spreading depression-976
like depolarization. Physiological Reviews 81, 1065-1096 (2001). 977
56. Wei, Y., Ullah, G. & Schiff, S. Unification of Neuronal Spikes, Seizures, and Spreading 978
Depression. Journal of Neuroscience 34, 11733-11743 (2014). 979
57. Dietz, R.M., Weiss, J.H. & Shuttleworth, C.W. Zn2+ influx is critical for some forms of 980
spreading depression in brain slices. Journal of Neuroscience 28, 8014-8024 (2008). 981
58. Pal, S., Sombati, S., Limbrick, D.D. & DeLorenzo, R.J. In vitro status epilepticus causes 982
sustained elevation of intracellular calcium levels in hippocampal neurons. Brain Research 851, 983
20-31 (1999). 984
59. Sham, J.S.K., Cleemann, L. & Morad, M. Functional coupling of Ca2+ channels and 985
ryanodine receptors in cardiac myocytes. Proceedings of the National Academy of Sciences of 986
the United States of America 92, 121-125 (1995). 987
60. Furuichi, T., et al. Multiple types of ryanodine receptor Ca2+ release channels are 988
differentially expressed in rabbit brain. Journal of Neuroscience 14, 4794-4805 (1994). 989
61. Lanner, J.T., Georgiou, D.K., Joshi, A.D. & Hamilton, S.L. Ryanodine Receptors: 990
Structure, Expression, Molecular Details, and Function in Calcium Release. Cold Spring Harbor 991
Perspectives in Biology 2, 21 (2010). 992
62. Copello, J.A., Barg, S., Onoue, H. & Fleischer, S. Heterogeneity of Ca2+ gating of 993
skeletal muscle and cardiac ryanodine receptors. Biophysical Journal 73, 141-156 (1997). 994
63. Fill, M. & Copello, J.A. Ryanodine receptor calcium release channels. Physiological 995
Reviews 82, 893-922 (2002). 996
64. Foskett, J.K., White, C., Cheung, K.H. & Mak, D.O.D. Inositol trisphosphate receptor 997
Ca2+ release channels. Physiological Reviews 87, 593-658 (2007). 998
65. Giorgi, C., Marchi, S. & Pinton, P. The machineries, regulation and cellular functions of 999
mitochondrial calcium. Nature Reviews Molecular Cell Biology 19, 713-730 (2018). 1000
66. Zhang, S.Y.L., et al. STIM1 is a Ca2+ sensor that activates CRAC channels and migrates 1001
from the Ca2+ store to the plasma membrane. Nature 437, 902-905 (2005). 1002
67. Park, C.Y., et al. STIM1 Clusters and Activates CRAC Channels via Direct Binding of a 1003
Cytosolic Domain to Orai1. Cell 136, 876-890 (2009). 1004
68. Pietrobon, D. & Moskowitz, M.A. Chaos and commotion in the wake of cortical 1005
spreading depression and spreading depolarizations. Nature Reviews Neuroscience 15, 379-393 1006
(2014). 1007
69. Aiba, I., Wehrens, X.H.T. & Noebels, J.L. Leaky RyR2 channels unleash a brainstem 1008
spreading depolarization mechanism of sudden cardiac death. Proceedings of the National 1009
Academy of Sciences of the United States of America 113, E4895-E4903 (2016). 1010
70. Wuytack, F., Raeymaekers, L. & Missiaen, L. Molecular physiology of the SERCA and 1011
SPCA pumps. Cell Calcium 32, 279-305 (2002). 1012
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
27
71. Sato, S.S., et al. Simultaneous two-photon imaging of intracellular chloride concentration 1013
and pH in mouse pyramidal neurons in vivo. Proceedings of the National Academy of Sciences of 1014
the United States of America 114, E8770-E8779 (2017). 1015
72. Sun, X.L., et al. Simultaneous monitoring of intracellular pH changes and hemodynamic 1016
response during cortical spreading depression by fluorescence-corrected multimodal optical 1017
imaging. Neuroimage 57, 873-884 (2011). 1018
73. Menyhárt, A., et al. Age or ischemia uncouples the blood flow response, tissue acidosis, 1019
and direct current potential signature of spreading depolarization in the rat brain. American 1020
Journal of Physiology-Heart and Circulatory Physiology 313, H328-H337 (2017). 1021
74. Menyhárt, A., et al. Spreading depolarization remarkably exacerbates ischemia-induced 1022
tissue acidosis in the young and aged rat brain. Scientific Reports 7, 13 (2017). 1023
75. Kim, J.H., et al. Noninvasive measurement of the pH of the endoplasmic reticulum at rest 1024
and during calcium release. Proceedings of the National Academy of Sciences of the United 1025
States of America 95, 2997-3002 (1998). 1026
76. Dreier, J.P., et al. Terminal spreading depolarization and electrical silence in death of 1027
human cerebral cortex. Annals of Neurology 83, 295-310 (2018). 1028
77. Carlson, A.P., et al. Terminal spreading depolarizations causing electrocortical silencing 1029
prior to clinical brain death: case report. Journal of Neurosurgery 131, 1773-1779 (2019). 1030
78. Hernandez-Caceres, J., Macias-Gonzalez, R., Brozek, G. & Bures, J. Systemic ketamine 1031
blocks cortical spreading depression but does not delay the onset of terminal anoxia 1032
depolarization in rats. Brain Research 437, 360-364 (1987). 1033
79. Müller, M. & Somjen, G.G. Inhibition of major cationic inward currents prevents 1034
spreading depression-like hypoxic depolarization in rat hippocampal tissue slices. Brain 1035
Research 812, 1-13 (1998). 1036
80. Collins, F., Schmidt, M.F., Guthrie, P.B. & Kater, S.B. Sustained increase in intracellular 1037
calcium promotes neuronal survival. Journal of Neuroscience 11, 2582-2587 (1991). 1038
81. Franklin, J.L. & Johnson, E.M. Suppression of programmed neuronal death by sustained 1039
elevation of cytoplasmic calcium. Trends in Neurosciences 15, 501-508 (1992). 1040
82. Heinemann, U. & Lux, H.D. Ceiling of stimulus induced rises in extracellular potassium 1041
concentration in the cerebral-cortex of cat. Brain Research 120, 231-249 (1977). 1042
83. Anderson, D., et al. Closed-loop stimulation in periods with less epileptiform activity 1043
drives improved epilepsy outcomes. Brain 147, 521-531 (2024). 1044
84. Huang, X.P., et al. AAV2 production with optimized N/P ratio and PEI-mediated 1045
transfection results in low toxicity and high titer for in vitro and in vivo applications. J. Virol. 1046
Methods 193, 270-277 (2013). 1047
85. Goldey, G.J., et al. Removable cranial windows for long-term imaging in awake mice. 1048
Nature Protocols 9, 2515-2538 (2014). 1049
86. Roome, C.J. & Kuhn, B. Chronic cranial window with access port for repeated cellular 1050
manipulations, drug application, and electrophysiology. Frontiers in Cellular Neuroscience 8, 8 1051
(2014). 1052
87. Roome, C.J. & Kuhn, B. Voltage Imaging with ANNINE Dyes and Two-Photon 1053
Microscopy. in Multiphoton Microscopy (ed. E. Hartveit) 297-334 (Humana Press Inc, Totowa, 1054
2019). 1055
88. Moucka, R., Sedlacík, M., Osicka, J. & Pata, V. Mechanical properties of bulk Sylgard 1056
184 and its extension with silicone oil. Scientific Reports 11, 9 (2021). 1057
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
28
89. Stern, M.A., et al. Applications of Bioluminescence-Optogenetics in Rodent Models. 1058
Methods Mol Biol 2525, 347-363 (2022). 1059
90. Jiang, S., et al. Automated, highly reproducible, wide-field, light-based cortical mapping 1060
method using a commercial stereo microscope and its applications. Biomed. Opt. Express 7, 1061
3478-3490 (2016). 1062
91. Ferezou, I., et al. Spatiotemporal dynamics of cortical sensorimotor integration in 1063
behaving mice. Neuron 56, 907-923 (2007). 1064
92. Mason, C.R. & Cooper, R.M. Permanent change in convulsive threshold in normal and 1065
brain-damaged rats with repeated small doses of pentylenetetrazol. Epilepsia 13, 663-& (1972). 1066
93. Schindelin, J., et al. Fiji: an open-source platform for biological-image analysis. Nature 1067
Methods 9, 676-682 (2012). 1068
94. Pachitariu, M., et al. Suite2p: beyond 10,000 neurons with standard two-photon 1069
microscopy. bioRxiv, 061507 (2017). 1070
95. Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm 1071
for cellular segmentation. Nature Methods 18, 100-106 (2021). 1072
96. Wenzel, M., Hamm, J.P., Peterka, D.S. & Yuste, R. Reliable and Elastic Propagation of 1073
Cortical Seizures In Vivo. Cell Reports 19, 2681-2693 (2017). 1074
97. Somarowthu, A., Goff, K.M. & Goldberg, E.M. Two-photon calcium imaging of seizures 1075
in awake, head-fixed mice. Cell Calcium 96, 8 (2021). 1076
98. Rolston, J.D., Laxpati, N.G., Gutekunst, C.-A., Potter, S.M. & Gross, R.E. Spontaneous 1077
and evoked high-frequency oscillations in the tetanus toxin model of epilepsy. Epilepsia 51, 1078
2289-2296 (2010). 1079
1080
Acknowledgments 1081
We thank Thomas Eggers for his assistance with generating stimulation waveforms, Henry Skelton 1082
for his assistance with confocal imaging, Alexandra Nazzari for her comments on the abstract and 1083
Bona Kim for her illustration work. This work was supported by funding from the NIH 1084
[F31NS115479 (MAS), R21NS112948 (REG), S10OD021773 (KB)] and the Mirowski Family 1085
Foundation (REG). 1086
1087
Author contributions: 1088
Conceptualization: MAS, KB, REG; Methodology: MAS, KB; Software: MAS, ERC; Validation: 1089
MAS, KB, ERC; Formal analysis: MAS, ERC, KB; Investigation: MAS, KB, CAG; Resources: 1090
REG, KB, JJY, CAG, MAS; Data curation: MAS, ERC, KB; Writing—original draft: MAS; 1091
Writing—review and editing: MAS, KB, ERC, CAG, JJY, REG; Visualization: MAS, KB, ERC, 1092
CAG; Supervision: REG, KB; Project administration: MAS; Funding acquisition: REG, KB, MAS 1093
1094
Competing interests: JJY is the shareholder of InLighta Biosciences and is a named inventor on 1095
an issued patent (US10371708) for R-CatchER. REG has received research support and personal 1096
fees outside the submitted work from NeuroPace, Inc., owner of the RNS® system. The terms of 1097
these arrangement have been reviewed and approved by Emory University and Georgia State 1098
University, in accordance with their conflict-of-interest policies. All other authors declare they 1099
have no competing interests. 1100
1101
Data and materials availability: The viral vector plasmids generated from this project will be 1102
made available to researchers upon request through a material transfer agreement. All data needed 1103
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
29
to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Material. 1104
Code and summary data needed to replicate figures in the paper are publicly archived at Zenodo 1105
[repository pending, will be updated prior to publication]. Updated versions of the code will be 1106
available at the GitHub repository: https://github.com/Stern-MA/RCatchER_CSD. 1107
.CC-BY-NC-ND 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted September 24, 2024. ; https://doi.org/10.1101/2024.09.21.614252doi: bioRxiv preprint
