Astrocyte-mediated transduction of muscle fiber contractions synchronizes hippocampal neuronal network development

Exercise supports brain health in part through enhancing hippocampal function. The leading hypothesis is that muscles release factors when they contract (e.g., lactate, myokines, growth factors) that enter circulation and reach the brain where they enhance plasticity (e.g., increase neurogenesis and synaptogenesis). However, it remains unknown how the muscle signals are transduced by the hippocampal cells to modulate network activity and synaptic development. Thus, we established an in vitro model in which the media from contracting primary muscle cells (CM) is applied to developing primary hippocampal cell cultures on a microelectrode array. We found that the hippocampal neuronal network matures more rapidly (as indicated by synapse development and synchronous neuronal activity) when exposed to CM than regular media (RM). This was accompanied by a fourfold increase in the proliferation of astrocytes. Further, experiments established that the astrocytes release factors that inhibit neuronal excitability and facilitate network development. Results provide new insight into how exercise may support hippocampal function through regulating astrocyte proliferation and subsequent taming of neuronal activity into an integrated network.


33
Exercise is a highly effective strategy for maintaining cognitive health throughout life, even when initiated at 34 late stages in life [1][2][3]. Many studies have shown robust long-term changes in the hippocampus from increased 35 physical activity, such as increased adult hippocampal neurogenesis, synaptogenesis and enlarged hippocampal 36 volume which likely support enhanced cognition [3][4][5][6][7]. However, the mechanisms by which exercise produces 37 such dramatic changes in the hippocampus remain elusive. Uncovering the mechanisms that are responsible for 38 enlarging the hippocampus and enhancing its function could be used to reverse-engineer treatments for 39 cognitive pathologies that result in a diminished size and function of the hippocampus, such as Alzheimer's 40 disease, stress, depression, anxiety, PTSD, Cushing's disease, epilepsy, and normal aging [8]. 41 Cumulative research over the past few decades has suggested that factors released from contracting muscles 42 (such as lactate [9], growth factors [10,11], trophic factors [12], and myokines [13,14]) provide crucial signals 43 that support enhanced plasticity [15]. However, how muscle factors affect hippocampal cells is still being 44 worked out. Recently, we found that repeated electrical contractions of the hindlimb muscles of anesthetized 45 mice in a pattern that produced endurance adaptations in the muscles (40 reps, twice a week for 8 weeks) caused 46 increased numbers of new astrocytes in the hippocampus and enlarged the volume of the dentate gyrus by 47 approximately 10% [16]. This suggests astrocytes are sensitive to muscle factors and proliferate when they 48 detect muscle factors in the blood. Given the role that astrocytes play in forming the blood-brain barrier, they 49 are well situated to transduce signals from the blood into the brain. 50 One way to study the interactions between contracting muscle cells and hippocampal cells including neurons 51 and astrocytes is to isolate the cells and perform experiments in vitro. For example, previous in vitro studies 52 found that muscle conditioned media attracted neurites of spinal cord motor neurons to form neuro-muscular 53 junctions [17]. Along this line, our lab has been examining cross-talk between muscles and neurons in vitro. We 54 recently found that when media from contracting muscle fibers derived from a C2C12 mouse myoblast cell line 55 is applied to neuronal cultures derived from a mouse embryonic stem cell line plated on a micro-electrode array, 56 it enhanced overall neural firing rates of the neurons [18]. 57 To further explore how factors from contracting muscles might influence hippocampal cells, we developed an in 58 vitro preparation in which primary mouse skeletal muscle cells are plated on a functionalized substrate. The 59 myoblasts develop bundles of myotubes and begin to contract spontaneously. We then take the media 60 surrounding the contracting muscles (conditioned media, CM) and apply that media to in vitro primary 61 hippocampal cell cultures that include neurons and astrocytes. The objectives of this study were to determine 62 whether CM influences the function and maturation of hippocampal neuronal networks, and to investigate the 63 consisting of DPBS, 5% goat serum (Sigma), and 1% bovine serum albumin (Sigma) overnight at 4 °C. The 128 samples were treated with primary antibodies overnight at 4 °C, secondary antibodies for 2 hours, and DAPI 129 Synapse detection using double-fluorescent label method 139 To detect and measure synapses and filamentous actin at presynaptic terminals, the double-fluorescent label 140 method was used by following protocol with modifications [26]. Briefly, synapses or F-actin were double-141 labeled with two different antibodies at different channels. One set of puncta from one channel and the other set 142 from the other channel were colocalized, then verified as synapses. The total intensity of synapse and F-actin 143 was measured from integrated image planes after the colocalization process. The analysis was performed by 144

146
Calcium imaging was performed using Cal-590-AM (AAT Bioquest, 20510) by the manufacturer's protocols. 147 Briefly, samples were incubated with DMEM, 0.04% Pluronic F-127 (Sigma), and 5 µM Cal-590-AM for an 148 hour at 37°C. After washout, samples were supplemented with DMEM with no phenol red to reduce 149 background noise. The dye-loaded cells were excited at 574 nm and imaging was performed at the frame rate of 150 12 fps. For quantitative analysis, the average fluorescence intensity of selected regions of interest was 151 calculated. Then the trace of fluorescent dynamics was calculated as Python (3.9.7), and MATLAB. Raw data were filtered using a 2nd order Butterworth high pass filter with 200 164 Hz cutoff frequency. Action potentials were detected as spikes by a threshold of 4 × standard deviations for both 165 rising and falling edge from the noise magnitude distribution. Spikes were only detected by active electrodes 166 which were defined by electrodes containing at least 5 spikes/min. The parameters of burst detection are as 167 follows: Maximum interval to start burst: 50 ms, maximum interval to end burst: 50 ms, minimum interval 168 between bursts: 100 ms, minimum duration of burst: 50 ms, minimum number of spikes in burst: 4. 169

Synchrony index
170 The synchrony of spike trains between electrodes was assessed through cross-correlation for discrete functions 171 where and are spike trains consisting of 0 (no spike) and 1 (spike), and are the th and th electrode, and 176 is the th discrete time. ̅ = 0 and ̅ = 1 represent completely asynchronous and synchronous, respectively. 177

Statistical analysis
178 SAS (9.4) and R (4.0.3) were used for statistical analysis. p < 0.05 was considered statistically significant. Data 179 were considered normally distributed when the absolute value of the skewness and kurtosis was less than 1 and 180 2, respectively. In the case of non-normal distribution, a power transform was used to transform data to meet the 181 normality conditions. Actin intensity, muscle contraction amplitudes, calcium signal, and astrocyte number in 182 response to the glia inhibitor were evaluated by two-sample t-test (RM vs. CM, control vs. BTS 10 µM, and 183 control vs. glia inhibitor). Synapse number, vesicle accumulation, and astrocyte number were evaluated by two-184 way ANOVA with day (day 2 vs. day 9) and treatment (CM vs. RM) as factors. The MEA outcomes of the BTS 185 study were analyzed using repeated measures three-way ANOVA with cohort as a blocking variable, day (day 2 186 to 9) entered as a within-subjects, muscle treatment (2 levels: RM vs. CM) as a between-subjects factor, and 187 drug (2 levels: control vs. BTS) entered as a between-subjects factor. Similarly, the MEA outcomes of the glia 188 reduced study were analyzed using repeated measures three-way ANOVA with cohort as a blocking variable, 189 day (day 2 to 9) entered as a within-subjects, muscle treatment (2 levels: RM vs. CM) as a between-subjects 190 factor, and astrocyte composition (3 levels: presence vs. absence vs. absence with astrocyte releasate (Ast-RM  191 and Ast-CM)) as a between-subjects factor. For the burst rate (BTS and glia reduced study) and synchrony 192 index (glia reduced study), aligned rank transform was used for the non-parametric test since data were 193 considered non-normally distributed. Post-hoc pairwise differences between means were performed using 194 Fisher's least significant difference test. 195

196
Contracting muscle conditioned media enhances neuronal activity 197 measured by microelectrode arrays 198 Consistent with our previous MEA study with C2C12 mouse myoblast cell line and mouse embryonic stem cell-199 derived neuronal culture [18], CM from primary skeletal muscle cells increased spike and burst rates of primary 200 hippocampal neurons across days (Fig 1A and 1B). The general pattern of development of spike trains over time 201 in RM was consistent with other studies using primary hippocampal cells and primary sensory neurons at a 202 similar cell seeding density [28,29]. Significant differences in spike rates were observed between days (F7, 80 = 203 19.4, p < 0.001) and between RM versus CM treatments (F1, 80 = 202.2, p < 0.001). The interaction between day 204 and treatment was also significant (F7, 80 = 23.7, p < 0.001). Similar to spike rate, burst rate also showed 205 significant effects of day (F7, 72 = 10.5, p < 0.001), treatment (F1, 72 = 38.3, p < 0.001), and interaction between 206 the two (F7, 72 = 21.1, p < 0.001). The interactions were caused by a different pattern of results for earlier time-207 points (days 2-7), as compared to later time-points (days 8 and 9). At the early time-points, CM had a higher 208 spike and burst rates, but at later time-points the difference in them between RM and CM was reduced, absent, 209 or reversed. 210 and CM with BTS (bottom right) on day 9. Each line and the red box represent a single firing, and synchronous 219 burst, respectively. Scale bar: 2 s. 220 Similar to spike rate and burst rate, CM also caused neurons to fire more synchronously as compared to RM 221 ( Fig 1C). A two-way ANOVA showed a significant effect of day (F7, 75 = 175.7, p < 0.001), treatment (F1, 75 = 222 4083.4, p < 0.001) and the interaction between the two (F7, 75 = 231.1, p < 0.001). However, unlike spike rate 223 and burst rate which showed greater differences between CM and RM at initial time-points than later time-224 points, the synchrony index showed the reverse pattern, with greater differences at the later time-points and no 225 difference at the early time-points when little synchronous firing occurred. 226 Having shown that CM increases spike rate, burst rate and synchronous firing of primary hippocampal neurons 227 in culture, we next wanted to evaluate whether contraction of the muscles was necessary for the MEA effects. 228 The alternative is that muscle cells release neuro-active factors regardless of whether they are contracting. 229 Hence, we repeated the experiment except we treated the muscle cells with a contraction inhibitor before 230 collecting the media. We used a known skeletal muscle myosin II inhibitor, N-benzyl-p-toluene sulphonamide 231 (BTS). BTS weakens myosin's interaction with F-actin and ex vivo studies found 10 μM of BTS suppresses 232 force production by 60% [23,24]. Consistent with these results, the amplitudes of muscle contraction were 233 reduced by 69% with BTS in vitro (t4= 20.7, p < 0.001; Fig 1D). Moreover, we performed calcium imaging of 234 the skeletal muscles in vitro (see details in Materials and methods). Calcium dynamics in muscle cultures show 235 a 73% reduction between control and 10 μM of BTS (t4= 4.9, p =0.0083; Fig 1E). Widefield (S1 and S2 videos) 236 and calcium imaging (S3 and S4 videos) of skeletal muscles are available as supplementary materials. 237 BTS prevented CM from increasing spike and burst rate, but had no effect on baseline spike and burst rate in 238 RM. This suggests muscle cell contractions are required for CM to increase spike and burst rate.  Fig 2B). On day 2, F-actin was not detected in 286 either CM or RM so results are not shown. However, filamentous actin was detected on day 9 in both groups, 287 and average F-actin intensity in CM was higher by 48% (t10= 2.9, p = 0.0152; Fig 2C). 288 Muscle conditioned media from contracting muscles induces astrocyte 289 proliferation 290 The numbers of astrocytes significantly increased by tenfold from day 2 to day 9 collapsed across RM and CM. 291 CM consistently displayed fourfold greater numbers of astrocytes than RM on both days (Fig. 2D) Astrocytes regulate neuronal activity in vitro 297 To determine the role of astrocytes in the increased spike rate observed in hippocampal primary cultures 298 exposed to CM versus RM, we repeated the experiment in cultures with reduced astrocyte populations. To 299 remove astrocytes from the primary hippocampal cell culture, we applied a glia inhibitor following established 300 protocols [25] (see details in Materials and methods). Consistent with previous accounts, this resulted in an 81% 301 reduction in the number of astrocytes in the culture (t14= 5.70, p < 0.001) (Fig 3A). Confocal images of 302 astrocyte populations in control and astrocyte reduced culture are shown (Fig 3B). 303 between these factors were also significant (p < 0.001). The presence or absence of astrocytes factor includes 3 314 levels, presence, absence, and absence but with the releasate from astrocytes added back in (Ast-RM and Ast-315 CM groups; see statistical methods). Post-hoc tests indicated that CM increased spike rate relative to RM in 316 normal hippocampal cultures consistent with the previous result (p < 0.001), but also in cultures with reduced 317 astrocytes (p < 0.001). To our surprise, we found that a reduction of astrocytes increased spike rate in both RM 318 and CM (p < 0.001), and the increase in CM with reduced astrocytes was an order of magnitude higher 319 compared to the unaltered culture in CM (Fig 3C). Moreover, the increase in spike rate in CM relative to RM 320 was greater for cultures with reduced astrocytes as compared to unaltered cultures as reflected by the significant 321 interaction between muscle media and presence/absence of astrocytes (F2, 53 = 49.8, p < 0.001). Taken together, 322 these results suggest astrocytes inhibit neuronal activity and CM increases their inhibitory function to counteract 323 the excitatory effect of CM on neuronal activity. 324 To determine whether astrocytes mediate their inhibitory function through releasing factors into the media or 325 whether they need to be physically present to exert their inhibitory effect, we included the Ast-RM and Ast-CM 326 treatments. In these treatments, astrocytes were removed from the hippocampal culture but media from intact 327 hippocampal cultures with astrocytes was added back in after being exposed to either RM or CM. RM and Ast-328 CM reduced spike rate relative to RM and CM when the hippocampal cultures were deprived of astrocytes. 329 Moreover, the spike rate in the Ast-groups was similar to when the astrocytes are physically present in intact 330 primary hippocampal cultures. This is supported by non-significant post-hoc test between groups where 331 astrocytes were physically present versus absent but releasate added back in (p = 0.0537). Further, comparisons 332 between normal cultures exposed to CM and astrocyte-deprived cultures exposed to Ast-CM showed no 333 significant difference (p = 0.9597). Likewise, normal cultures exposed to RM showed no difference from 334 astrocyte deprived cultures with Ast-RM (p = 0.68). These results suggest that astrocytes mediate their 335 inhibitory effect through releasing factors into the media and do not need to be physically present in the culture 336 to exert their influence. 337 338

339
Here we establish for the first time an in vitro platform to explore interactions between contracting primary 340 muscle cells and primary hippocampal cells. One of the leading hypothesized mechanisms for pro-cognitive 341 effects of exercise is that muscle contractions release factors that cross into the brain where they directly 342 influence hippocampal cells involved in cognition [7,10,14]. This hypothesis is supported by our recent 343 discovery that muscle contractions alone, through electrical stimulation of the sciatic nerve in anesthetized 344 mice, are capable of increasing the generation of new astrocytes in the hippocampus and increasing the volume 345 of the dentate gyrus [16]. The in vitro model developed herein adds to this literature by identifying a novel 346 mechanism by which muscle cells may communicate with hippocampal cells. Muscle cells release factors that 347 cause hippocampal neurons to become excited and hippocampal astrocytes to proliferate faster. The expanded 348 astrocytes play a role in regulating neuronal excitability. Together this leads to a network that has overall 349 greater excitability than in absence of the muscle signals, but also greater inhibition from astrocytes. The 350 astrocytes thus tame the increased electrical activation of the circuit from the muscle factors in a way that leads 351 to selective strengthening of coordinated activation patterns between neurons. 352 A key finding was that muscle contractions are necessary for CM to influence spike rate and burst rate in the 353 hippocampal cultures. When muscles were prevented from contracting by administering BTS, CM no longer 354 produced the increased effects on spike and burst rate (Fig 1A and 1B). This adds important validity to the 355 model since exercise involves mechanical forces and the hypothesis is that muscle contractions release factors 356 that they otherwise would not release to communicate their status of engaging in physical activity to the 357 hippocampus. We were able to make this conclusion because the effect of BTS on spike and burst rate was 358 specific to CM, it had no impact when administered in RM, (i.e., BTS-RM spike and burst rate was similar to 359 RM, but BTS-CM showed reduced spike and burst rate relative to CM). However, this was not true for the 360 synchrony index where BTS appeared to directly eliminate synchronous firing of neurons whether in RM or 361 CM ( Fig 1C). Thus, we cannot be certain that the muscle contractions are necessary for the effect of CM on 362 enhancing synchronous firing of action potentials. A method is needed that can prevent the muscle cells from 363 contracting that does not directly interfere with any of the MEA outcomes. 364 In the context of whole organismal exercise, muscles communicate with hippocampal neurons while 365 hippocampal neurons are involved in the sensorimotor processing in the brain that occurs during physical 366 activity [15,35]. Indeed, acute activation of the hippocampus is strongly correlated with running speed and 367 repeated exercise training increases adult hippocampal neurogenesis and astrogliogenesis [3,5-7]. Together with 368 the in vitro data collected herein, the results suggest that muscle contractions contribute to the plasticity in the 369 hippocampus by responding with signals that increase the number of new astrocytes to counterbalance the 370 excitation that is likely intrinsic to the hippocampus involved in the sensorimotor response to physical activity. 371 Possibly related to the excitation of the hippocampus, whole organismal exercise produces a microenvironment 372 in the hippocampus that is conducive for neurogenesis and synaptogenesis. Consistent with recent reports, 373 results from the present study suggest that signals from muscle contractions likely contribute to the 374 synaptogenic microenvironment. The fact that CM increased maturation of the hippocampal network and 375 increased the formation of mature synapses is consistent with such a role. 376 It is notable that while CM greatly increased astrogliogenesis, it did not increase neurogenesis in the primary 377 hippocampal cultures. This could be because the stem cell environment was disrupted during the in vitro 378 preparation. On the other hand, the lack of an increase in neurogenesis is consistent with the in vivo study in 379 which the contribution of muscle contractions was isolated. As mentioned earlier, this study found that repeated 380 electrical contractions of muscles while the animal is anesthetized are capable of increasing astrogliogenesis but 381 not neurogenesis in the hippocampus [16]. Taken together, these results strongly suggest that additional factors 382 besides the muscle releaseate are necessary to increase neurogenesis. This does not rule out the possibility that 383 increased astrocytes and a microenvironment conducive for synaptogenesis may represent support systems 384 signaled by the muscles in anticipation of neurogenesis which usually accompanies whole organismal exercise. 385 Increased astrocytes could be key to how CM increases synaptogenesis and inhibits neuronal excitability. 386 Astrocytes are well-known homeostatic regulators of neuronal activity. They directly modulate the ratio of 387 excitatory and inhibitory synapses and neurotransmitter concentrations such as GABA and glutamate based on 388 environmental needs. An in vitro study found the appearance of GABAergic and glutamatergic synapses by 24 389 hours when embryonic rat ventral spinal neurons were cultured on astrocytes as compared to 4 and 7 days, 390 respectively in the neuron-only culture [36]. Furthermore, when astrocyte-conditioned media was supplemented 391 in astrocyte-deprived situations, increases in GABAergic synapses, axon length [37], and receptors [38] were 392 detected. Thus, astrocytes appear to release factors that increase GABAergic synapses and do not need to be 393 physically adjacent to neurons to exert their inhibitory influence. This is consistent with their role in our study 394 where media from neuronal cultures with astrocytes was capable of recapitulating the inhibitory effect of 395 astrocytes in a neuronal culture without astrocytes physically present (Fig 3C). It is well established that 396 astrocytes and astrocyte proliferation occur in response to epilepsy, and the evidence suggests astrocytes release 397 factors such as gliotransmitters and tumor necrosis factor-alpha (TNF-α) that inhibit neuronal excitability and 398 are protective against excitotoxicity [39]. 399 In addition to increasing the inhibitory function of astrocytes, CM also increased the maturation of the 400 hippocampal network as reflected in the MEA data and the maturation of synapses. In this study, we used 401 vesicle clustering and F-actin accumulation to quantify mature synapses. implies that CM enhances the maturation of functional synapses with greater capacities to transport vesicles 409 upon action potentials. This could explain why neuronal cultures plated on MEA exposed to CM displayed 410 greater levels of synchronized firings of action potentials than RM because there were more mature synapses. 411 The increased number of astrocytes in the neuronal cultures exposed to CM may have contributed to the 412 increased maturation of the hippocampal network by strengthening specific synapses and weakening others by 413 pruning and inhibition. Mature synapses appeared earlier and were pruned earlier in cultures with more 414 astrocytes as a consequence of exposure to CM. Whereas synapses continued to increase from day 2 to day 9 in 415 RM, in CM they reached their peak around day 2 and were already in decline by day 9. The in vitro model is only useful to the extent that it reflects features of the whole-organismal phenomenon. 420 Whole organismal exercise increases synaptogenesis and astrogliogenesis in the hippocampus and the in vitro 421 model displays these features. It is not intended to represent the entirety of exercise's impact on the 422 hippocampus. For example, increased excitability (spike rate, burst rate, synchrony index) of the primary 423 hippocampal network as measured by the MEA has no direct meaning for how neurons behave in the actual 424 hippocampus during exercise. There is no "appropriate" level of excitability, and higher or lower excitability is 425 not any "better" than the other as there is no benchmark for successful performance in a dish. The value of the 426 reduced approach is thus not for recapitulating the patterns of neuronal activity and circuit dynamics, but rather 427 to isolate the effect of muscle factors on basic properties of hippocampal cells. 428 Before conducting the present studies, we knew astrocytes increased in the hippocampus in response to 429 exercise, and that muscle contraction alone was capable of recapitulating this effect. We observed the same 430 phenomenon of increased astrocytes in the in vitro model, which justifies its use for exploring the role of 431 increased astrocytes in the hippocampal response to muscle contractions. Because of the in vitro model, we now 432 have a hypothesis for why astrocytes are responsive to muscle factors, they play an inhibitory role in taming the 433 excitability of neurons that occurs in parallel with muscle contractions. 434 Future studies will build on the in vitro platform to explore potential reciprocal communication between muscle 435 cells and hippocampal cells through a co-culture with shared media exchange. We are also interested in using 436 the platform to explore the potential mechanism by which CM causes astrocytes to proliferate and hippocampal 437 networks to mature faster. Finally, we are interested in identifying the bio-active factors released from the 438 contracting muscles that influence the hippocampal cultures. In the future, such information could be used to 439 reverse engineer treatments to recapitulate pro-cognitive effects of exercise in the absence of physical activity. 440 441 Acknowledgements 442 We