Dissociation Between Neuronal and Astrocytic Calcium Activity in Response to Locomotion in Mice

Abstract Locomotion triggers a coordinated response of both neurons and astrocytes in the brain. Here we performed calcium (Ca2+) imaging of these two cell types in the somatosensory cortex in head-fixed mice moving on the airlifted platform. Ca2+ activity in astrocytes significantly increased during locomotion from a low quiescence level. Ca2+ signals first appeared in the distal processes and then propagated to astrocytic somata, where it became significantly larger and exhibited oscillatory behaviour. Thus, astrocytic soma operates as both integrator and amplifier of Ca2+ signal. In neurons, Ca2+ activity was pronounced in quiescent periods and further increased during locomotion. Neuronal Ca2+ concentration ([Ca2+]i) rose almost immediately following the onset of locomotion, whereas astrocytic Ca2+ signals lagged by several seconds. Such a long lag suggests that astrocytic [Ca2+]i elevations are unlikely to be triggered by the activity of synapses among local neurons. Ca2+ responses to pairs of consecutive episodes of locomotion did not significantly differ in neurons, while were significantly diminished in response to the second locomotion in astrocytes. Such astrocytic refractoriness may arise from distinct mechanisms underlying Ca2+ signal generation. In neurons, the bulk of Ca2+ enters through the Ca2+ channels in the plasma membrane allowing for steady-level Ca2+ elevations in repetitive runs. Astrocytic Ca2+ responses originate from the intracellular stores, the depletion of which affects subsequent Ca2+ signals. Functionally, neuronal Ca2+ response reflects sensory input processed by neurons. Astrocytic Ca2+ dynamics is likely to provide metabolic and homeostatic support within the brain active milieu.


Introduction
Brain function depends on closely coordinated interactions between different cell types, including neurons, glial cells, cells of the blood vessels, and noncellular elements such as extracellular space and extracellular matrix. This system of brain elements and their interactions is called the brain acti v e milieu. 1 , 2 The activity of electrically excitable neurons and their synaptic communication have been routinely recorded with electrodebased methods, demonstrating neuronal contribution to major brain functions: locomotion, 3 sensor y acti vity, 4 vision, 5 learning, and memor y. 6 Eclecticall y nonexcita b le elements of the brain acti v e milieu, including neuroglia, cannot be fully assessed with electrode-based techniques and require alternative approaches such as real-time optical imaging.
Optical methods allow monitoring of various types of cellular activity in the brain that include ionic signaling, 7-10 metabolic processes, 11 , 12 cellular morphology, and structure of extracellular space. [13][14][15] Ca 2 + imaging is the most popular: Multiple genetically encoded sensors were developed to monitor [Ca 2 + ] i in specific cell types. [16][17][18] Two-photon microscopy provides cellular and subcellular resolution for imaging with genetically encoded sensors in vi v o . [19][20][21] Ho w ever, combining this tec hnique with animal behavior studies is difficult because the animal is kept in a head-fixed restrained condition throughout the experiment. Thus, experiments have been routinely performed on treadmills and air-floating balls that keep animals in a fixed position while allowing a certain type of locomotion. 22 , 23 These methods can be combined with a virtual reality environment to design various behavior al par adigms. 24 , 25 However, whether virtual r eality faithfull y r e pr esents the r eal world as an experimental animal percei v es it r emains an open question. In addition, a head-restricted animal cannot turn and al w ays runs in a single dir ection on tr eadmills. The air-floating ball allows an animal to turn. Yet, the animal's posture differs from that of an animal running on a flat surface: It r esemb les climbing an object, with the tail and the hindlimbs positioned below the head and forelimbs.
Using a flat airlifted platform rectifies the limitations of the a bov e systems for 2-photon imaging in awake animals. This technique was used recently to monitor [Ca 2 + ] i dynamics in neurons and microglia in mice. 26 , 27 However, the other elements of the brain acti v e milieu hav e not been systematicall y studied in this experimental paradigm. Here, we compared the Ca 2 + activity of astrocytes and neurons of mice navigating an airlifted platform.

Animals
All pr ocedur es wer e performed in accordance with the local guidelines for animal care (The Finnish Act on Animal Experimentation), FELASA ethical recommendations, NIH Guide for the Care and Use of La borator y Animals, and the animal ethics and welfare committee of Jiaxing Uni v ersity. The C57BL/6 male ( N = 6) mice were used in 2-photon imaging experiments. Mice were kept under standard housing conditions with a 12-h light/dark cycle (lights on at 9 am ) with access to water and food ad libitum. Each animal underwent surgery at the age of 12-16 wk, followed by habituation to the experimental setup and imaging sessions. All experiments were carried out during the light period between 10 am and 5 pm . temperature at 37.0 • C, a heating pad was used. An eye ointment was applied to the eyes to keep them moist during surgery. The head skin was removed, and a 3-mm cr aniotom y a bov e the somatosensory cortex was performed using a dental drill (RWD, Shenzhen, China). Throughout the procedure, the skull w as ke pt moist with applications of sterile saline. The ster eotaxic injection of AAV5-gfaABC1D-cyto-GCaMP6f (Addgene viral pr e p #52925-AAV5), titer 10 11 GC/mL, was performed to get an astr ocyte-specific expr ession of geneticall y encoded Ca 2 + indicator GCaMP6f in the somatosensory cortex (AP −2.3, ML + 0.5, DV −0.6). To obtain expression in neurons, we used AAV2/9-CaMKII-GCaMP6f, titer 10 12 GC/mL (VectorBuilder, Guangzhou, China). The total volume of 500 nL was injected via a glass pipette of 5 μL volume (Drummond, USA) using the automatic pump at the rate of 2 nL/s (Ultra Micro Pump III; World Precision Instruments, Sarasota, FL, USA). After each viral injection, the pipette tip was held for 10 min to ensure proper AAV diffusion in the region of interest. Next, 4-mm round cover glass (Warner Instruments, USA) was sealed to the skull with biocompatible adhesive (Vetbond, World Precision Instruments, Sarasota, FL, USA) and uni v ersal r esin cement Rel yX (3M, China). A lightweight stainless steel head plate (model 2, Neurotar, Helsinki, Finland) was mounted on the head with dental acrylic (Super-Bond C&B, Sun Medical, Japan) to restrain a mouse in the experimental setup under the objecti v e of a 2-photon microscope (Femtonics, Budapest, Hungary). Dexamethasone (2 mg/kg) w as administer ed to mice by subcutaneous injection to reduce the surgery-induced inflammation.

Animal Habituation and Training
Befor e the ima ging sessions, mice wer e handled and ha bituated to the experimental setup. Habituation started 3-4 wk after the surgery when the animals fully recovered. Awake mice were head-fixed under a microscope using a Mobile HomeCage device (Neurotar, Helsinki, Finland). Mice heads were secured in the clamping mechanism, and the device was connected to a pressurized air sour ce . Once the carbon fiber cage began to float, animals could move it with their paws. Animals were habituated to head fixation in four to six 2-h training sessions r e peated once per day.

Ca 2 + Imaging
After habituation and training, mice were used for Ca 2 + imaging. Multiphoton fluorescence microscope (Femtonics, Budapest, Hungary) with a 16 × NA 0.8 water-immersion objective (CFI75 LWD 16X W, Nikon, Tok yo, J apan) or with a 20 × NA 0.8 waterimmersion objecti v e (XLUMPLFLN20XW, Ol ympus, Tok yo, J apan) was used for Ca 2 + imaging in vivo. Ti:Sapphire laser (Chameleon Ultra, Coherent, USA) with an excitation w av elength of 920 nm was used to excite the fluorescence of GCaMP6f. Fluorescence signal from a brain area of 600 × 600 or 175 × 175 μm 2 (512 × 512 pixels) was recorded in resonant scanning mode at 30 frames/s. The signal w as filter ed with a 520/60 nm bandpass filter (Semroc k, Roc hester, NY, USA) and then detected with a GaAsP photomultiplier (H11706P-40, Hamamatsu, Japan). Simultaneously, autofluor escence w as filter ed with a 650/100 nm bandpass filter and detected with a second identical photomultiplier. Several imaging sessions of 10 min were carried out for each mouse. In parallel with Ca 2 + imaging, the animal movements on the platform wer e monitor ed with locomotion-tr ac king softw ar e (Neurotar, Helsinki, Finland). All sessions were video recorded in infrared light. The experimental session did not exceed 2 h per mouse.

Pr ocessing Ca 2 + Ima ging Data
A few pr ocedur es and basic estimated parameters were the same for astrocytic and neuronal Ca 2 + data. At the same time, some extr acted c har acteristics differed to matc h the disparities in the responses of the 2 cell types. Preprocessing steps were identical for both astrocytic and neuronal data. Most of the ima ge pr ocessing w as done using r outines fr om Ima ge-funcut, uCats, and other softw ar e dev eloped in the gr oup using the common scientific Python libr ary stac k (SciPy, Scikit-learn, Scikitimage). 28 , 29 Preprocessing: Temporal Binning, Motion Correction, and Denoising The raw imaging data recorded at 30 Hz were binned along a time dimension by summing intensity values in nonoverlapping batches of 5 frames before further processing to increase the signal-to-noise ratio. Next, rigid-body motion correction was applied using an autofluorescence signal recorded in the red channel as a morphology marker. For a mor e sta b le shift estimation, the autofluorescence signal was adaptively denoised by approximating eac h fr ame by the first 50 spatial principal components, smoothed by a Gaussian filter with σ = 1 pixel. The spatial displacements estimated for the autofluorescence channel have then been applied to the green channel containing GCaMP6f fluorescence.
After shift correction, the data were further denoised using the algorithm developed in the lab. In brief, the data were Anscombe transformed to stabilize variance, and then the r ecording w as di vided into ov erlapping spatial windows (patches), where the data are approximated by truncated SVD with a smoothing filter applied to the spatial components, which is then followed by inverse SVD and Anscombe transforms, av era ging the appr oximations in the ov erlapping parts of the patches.

Calculation of F/F and Acti v e Area
Slowl y v ar ying fluor escent baseline lev el F 0 w as determined as a lower envelope running through local minima of the fluorescence signal in each pixel, smoothed by a Gaussian filter with σ = 15 s. The resulting baselines were negatively biased, and to correct this bias, the baseline signals in each pixel were then shifted by a constant parameter minimizing the absolute value of the mode of the residuals after baseline subtraction, assuming that most frequent fluctuations are small and are due to noise, while true Ca 2 + ev ents ar e sparse. Further anal ysis is done on r elati v e fluor escence changes F/F = (F −F 0 )/F 0 × 100%. An arbitrarily chosen threshold of 15% increase over the baseline ( F/F ≥ 15%) was selected as it provided the most visuall y informati v e r esults. Clusters of suprathreshold pixels larger than 16 pixels within a fr ame w er e marked as an "acti v e segment" if they pr oduced ov erlapping structur es in at least 3 consecuti v e frames. The fractional area of such acti v e segments in each frame normalized to the total stained area within the field of view was named "acti v e ar ea" and used to c har acterize the collecti v e Ca 2 + activity.

Analysis Specific to Astrocytes
Indi vidual astr ocytic domains wer e visuall y identified and manually traced. Inside the domain territory, the soma was likewise outlined. The rest of the domain was called processes.
F/F timecourses were plotted for both processes and soma. Then, the timecourses were normalized to their peak values. The latency between soma and processes was estimated as the difference between the times when F/F reached 15% of the maximal value. The prominence of F/F oscillations was estimated by av era ging w av elet power in the interval when the signal was above 15% of the peak value in the frequency range 0.1-0.3 Hz (continuous wavelet transform, Morlet w av elet).

Analysis Specific to Neurons
Because neuronal Ca 2 + data displayed a high spatially localized spontaneous activity, w e employ ed a slightly modified constr ained non-ne gati v e matrix decomposition (CNMF) appr oach to visualize this activity during locomotion. 30 CNMF is initially optimized for extracting Ca 2 + dynamics from cell bodies, but lifting temporal dynamics and spatial localization constraints can allow for the segmentation of sparse components. The algorithm was applied in overlapping 100 × 100 patches with 12 components per patch to solve for. A second run of the algorithm on the entire frame with the patch-based components as the initial condition was then applied to refine the segmentation. The resulting spatial components could still contain irregular low-amplitude shapes, which were clipped by using Li's auto threshold algorithm on nonzero pixels in each component. 31 The resulting solution produced many spatial segments with similar dynamics, further merged by clustering their corresponding dynamics (agglomerative clustering with Ward linkage) to produce 12 final spatial regions of interest (ROIs).

Sta tistical Anal ysis
Data ar e pr esented as median [25th-75th percentiles]. The sample number ( n ) indicates the number of recordings, periods of locomotion or quiescence, or cells and is specified for each experiment. Astrocytic Ca 2 + activity was recorded in 2 mice and neuronal in 4 mice. Statistical analysis was conducted using Python. For comparison of astrocytic and neuronal Ca 2 + activity ( F/F and acti v e ar ea) during the periods of quiescence and locomotion, the nonparametric Mann-Whitney U -test was used. For other comparisons, linear mixed-effects models were applied. 32 The experimenters were not blinded to the experimental conditions, and no randomization was performed. All statistical details of the experiments ar e pr ovided in the main text and figure legends. P < .05 w as consider ed statisticall y significant.

Ca 2 + Activity in Cortical Astrocytes Is Linked to Locomotion but Not Location
The somatosensory cortex area of mice containing projections of the left forelimb was injected with an AAV for expr ession of geneticall y encoded Ca 2 + sensor GCaMP6f under either astroc ytic (AAV5-gfaABC1D-c yto-GCaMP6f) or neuronal (AAV2/9-CamKII-GCaMP6f) promotors ( Figure 1 A). The cranial glass window was implanted above the site of injection. A lightweight stainless steel head plate was glued above the windo w. After reco very from the surgery, the animals were left for 3-4 wk in individual cages for GCaMP6f expression. After handling the animals and their habituation to the airlifted platform, Ca 2 + imaging experiments were performed (Video S1).
First, the imaging was done on astrocytes expressing GCaMP6f in cortical layer 1 ( Figure 1 B; Video S2). The animal locomotion episodes alternated with quiescence periods ( Figure 1 C and D). During quiescent periods, Ca 2 + activity appeared as small, scattered events. 10 Jointly, these events cover ed onl y a tiny frame ar ea (mean acti v e ar ea per period of quiescence: 1.0 [0.8-1.7]% of frame; n = 21) and had low F/F (mean F/F per period of quiescence: 1.4 [1.0-1.6]%; n = 21). Consistent with previous reports, the locomotion was accompanied by a rise in astrocytic [Ca 2 + ] i . 20 , 33 , 34 We observed an increase in both the acti v e ar ea (mean acti v e ar ea per e pisode of locomotion: 20.2 [10.7-36.1]% of frame; n = 29; P < .001, mixed-effects model) and F/F (mean F/F per episode of locomotion: 9.9 [6.3-19.2]%; n = 29; P < .001, mixed-effects model). Sometimes, the animal r an sever al times with short intervals of quiescence. Astrocytic [Ca 2 + ] i increased during each running episode, but this increase became smaller in subsequent episodes.
Astr ocytic Ca 2 + ima ging w as combined with animal tr ac king ( Figure 1 E). The mouse ran for several seconds in each episode and moved around the entire territory of the platform during the experiment. However, the animal pr eferr ed to r est (quiescence period) at the platform's peripher y. Ca 2 + acti vity followed the locomotion pattern without an obvious link to the animal's location.

Integr a ti v e Function of Astrocytes
Spontaneous Ca 2 + events are primarily localized to distal astrocytic processes. 35 , 36 We observed that during locomotion, [Ca 2 + ] i increased in the entire astrocytic domain, including soma. This observ ation pr ompts a question of whether Ca 2 + activity initiated in soma subsequentl y di v erges to astr ocytic pr ocesses or [Ca 2 + ] i rises in distal pr ocesses pr opa gate tow ard the soma. We   These results suggest that Ca 2 + signals start in astrocytic pr ocesses and conv erge into soma during animal locomotion. Astrocytic soma inte gr ates the Ca 2 + signal, amplifies it, and generates Ca 2 + oscillations.

Neuronal Ca 2 + Response to Locomotion
Neur onal Ca 2 + acti vity w as r ecorded in cortical layer 1 ( Figure 3 A). In this layer, we rar el y saw somata of the neurons but observed multiple processes. Some of these processes can be visually identified as dendrites with spines, and others cannot be classified into any specific type (dendrites or axons). Many neur onal pr ocesses exhibited high Ca 2 + acti vity during quiescent periods ( Figure 3 B and C; Video S3). Thus, unlike astrocytes, the neurons were active during locomotion and quiescent periods.
During quiescent periods, the mean acti v e ar ea w as 6.0 To identify how differ ent neur onal compartments contribute to Ca 2 + dynamics during quiescent periods and locomotion, we applied a constrained non-negati v e matrix factorization (CNMF) approach implemented in CaImAn, open-source softw ar e for Ca 2 + ima ging anal ysis. 37 With CNMF followed by clustering, we identified neuronal structures activ ated synchr onousl y within indi vidual r ecordings ( Figur e 3 D). These structures represented separate neuronal processes. We cannot confidently conclude whether these processes belong to the same neuron; thus, we considered them activity units. We analyzed each activity unit by plotting the corresponding [Ca 2 + ] i timecourses ( Figure 3 E). Such anal ysis r ev ealed 2 types of units, which had either high activity or low activity during quiescent periods. In the low activity, units [Ca 2 + ] i increased during locomotion. In the high activity, units [Ca 2 + ] i did not increase and in some cases decreased. Also, we noticed that the [Ca 2 + ] i increased during locomotion in the regions where we could not unequi v ocall y identify neur onal structur es. These regions may correspond to the out-of-focus or dim structures.

Response to Locomotion
Next, we compared the parameters of Ca 2 + responses to locomotion between astrocytes and neurons. First, we measured mean F/F during the quiescent period (Q) and locomotion (L). The Q/L ratio of F/F w as significantl y higher in neurons than in astr ocytes (in neur ons: 0. These findings demonstrate that neuronal Ca 2 + activity relia b l y incr eases with ev er y locomotion e pisode . In contr ast, astrocytic Ca 2 + signal has refractoriness that prevents Ca 2 + r esponse fr om dev eloping during closel y timed locomotion episodes.

The Distinct Timecourses of Neuronal and Astrocytic Ca 2 + Response to Locomotion
Cortical neur ons r ecei v e sensor y input fr om thalamic n uclei. During locomotion, neuronal firing triggers Ca 2 + entry through v olta ge-gated Ca 2 + channels. We observed that overall neuronal Ca 2 + activity in the somatosensory cortex increased during animal locomotion. However, neurons remained active in quiescent periods between episodes of locomotion. In some neuronal units, activity during quiescence was higher than that during locomotion.
Local [Ca 2 + ] i elevations in astrocytic leaflets can be mediated by Ca 2 + entry through the plasma membrane (through ionotr opic r ece ptors or r ev ersed Na + /Ca 2 + exchanger), wher eas large Ca 2 + transients in soma and branches are mediated by Ca 2 + r elease fr om endogenous stor es-endoplasmic r eticulum and mitochondria. [38][39][40] During animal quiescence, astrocytes generated small, scattered Ca 2 + events. These events can be either spontaneous or triggered by local synaptic activity. During locomotion, astrocytic Ca 2 + activity substantially increased. Relati v e to Ca 2 + acti vity during the quiescent period, this increase w as significantl y larger in astr ocytes than in neur ons. Mor eov er, astr ocytic [Ca 2 + ] i elev ation w as delayed for several seconds after the beginning of the locomotion episode and neuronal [Ca 2 + ] i elevation. These findings suggest that the activity of the local neuronal network may not trigger the astrocytic Ca 2 + signal. Most likely, astrocytic response originates from modulatory subcortical projections, such as noradrenergic input from the locus coeruleus targeting adrenergic r ece ptors a bundantl y expr essed in matur e astr ocytes. [41][42][43] Delayed astr ocytic Ca 2 + response also raises a question of its physiological relevance. This delay suggests that astrocytes do not respond to sensory signals in real time, hence are involved in the r eal-time pr ocessing of sensor y information. On the other hand, astrocytic Ca 2 + activity may be inv olv ed in modulation of synaptic plasticity, 44 regulation of vascular tone, 45 and metabolic activation of astrocytes 46 that occur with a dela y. F or example, synaptic plasticity and memory consolidation may happen after sensory information is processed, filter ed, and compar ed to pr evious memories in the brain. The same applies to vasodilatation and metabolic activation of astrocytes, which can represent a systemic response to the increased energy demand of the brain following robust neuronal activity during locomotion. 47

A Refractory Period of Ca 2 + Activity in Astrocytes
Another example of dissociation between Ca 2 + activity in the local neuronal network and astrocytes is highlighted by distinct changes in the PRR. When an animal ran twice with a short interval ( < 30 s), the neuronal network r elia b l y r esponded to each locomotion episode. Astrocytes responded with a large Ca 2 + transient only to the first run. The response to the second run was significantly diminished. Most likely, it is related to the mechanisms of Ca 2 + signal generation in these 2 cell types. In neurons, the bulk of Ca 2 + enters through ligand-or voltagegated channels of the plasma membr ane . 48 , 49 Because the transmembr ane Ca 2 + gr adient is quic kl y r estor ed, neur ons ar e r eady to respond to the second run. Global astrocytic Ca 2 + transients de pend on Ca 2 + r elease thr ough inositol 1,4,5-trisphosphate r ece ptors (InsP 3 R) fr om the endoplasmic r eticulum, which, upon str ong stim ulation, gets de pleted, while InsP 3 R become inacti v ated. 39 , 40 , 50 It takes substantially more time to restore the ability to generate Ca 2 + signals for astrocytes than for neurons. What is the physiological r elev ance of such a r efractor y period? Astrocytic Ca 2 + stimulates oxidative phosphorylation. 51 However, due to the sparse distribution of cytochromes in the astr ocytic electr on tr ansport c hain, there is also a high probability of r eacti v e oxygen species (R OS) generation in r esponse to [Ca 2 + ] i elevation in astrocytes. 52 , 53 In moderate quantities, (B and C) Q/L ratio, calculated for Ca 2 + activity in cortical neurons ("N," n = 8 recordings) and astrocytes ("A," n = 7 recordings; * * * P < .001 and * * P = .001 for F/F (B) and active ar ea (C), r especti v el y, Mann-Whitney U -test). (D) Cr oss-corr elation gr aphs betw een F/F and animal speed for neurons (left) and astrocytes (right) during an episode of locomotion. (E) Peak correlation coefficients for neurons ("N," n = 8 recordings) and astrocytes ("A," n = 7 recordings; P = .87, mixed-effects model) for the analysis presented at (D). (F) Lag of cross-correlation peak in neurons ("N," n = 8 recordings) and astrocytes ("A," n = 7 recordings; * * * P < .001, mixed-effects model). (G) The timecourses demonstr ating c hanges in F/F ( top ) and acti v e ar ea (middle) of neur onal (left) and astr ocytic (right) Ca 2 + acti vity during pairs of locomotion episodes. (Bottom) The timecourse of animal speed. The formula is for PRR calculation where < Ca 2 + response 1 > is the mean Ca 2 + response to the first locomotion in 1 recording; < Ca 2 + response 2 > is the same for the second locomotion. (H and I) PPRs for F/F ( * * P = .002, mixed-effects model) and acti v e ar ea ( * P = .014, mixed-effects model) in neurons ("N," n = 7 paired runs) and astrocytes ("A," n = 14 paired runs).
ROS play a signaling role and promote synaptic plasticity in the brain. 54 In high quantities, they cause oxidati v e str ess and cell damage. Therefore, the astrocytic refractory period may pr eserv e R OS's beneficial effects while pr ev enting cell damage.

Integr a ti v e Function of Astrocytes
At the subcellular level, [Ca 2 + ] i did not increase simultaneously in the entire astrocytic morphological domain during locomotion. [Ca 2 + ] i elevation started in the distal processes and then pr opa gated into the soma. This finding is consistent with a higher pr oba bility for Ca 2 + activity generation in distal astr ocytic pr ocesses. 35 , 55 This phenomenon may be explained by the specific localization of synapses near distal astrocytic pr ocesses. Howev er, a r ecent r e port demonstrated a r elati v el y even distribution of synapses near distal processes and astrocytic soma. 56 An alternati v e explanation is that distal astrocytic pr ocesses hav e a higher surface-to-v olume ratio (SVR). 36 High SVR increases the amplitudes of [Ca 2 + ] i elevations induced by Ca 2 + entr y thr ough the plasma membrane and consequently incr eases the pr oba bility of Ca 2 + -de pendent Ca 2 + r elease fr om endogenous stores.
Our observations also suggest that astrocytic soma operates as an inte gr ator of Ca 2 + acti vity starting in the pr ocesses. Ca 2 + signal inte gr ation in astrocytic soma is reminiscent of the integr ation of membr ane v olta g e chang es from dendritic inputs by soma in neur ons. 57 Howev er, in the case of neurons, somatic inte gr ation is translated into a pattern of action potentials. What is the functional r elev ance of Ca 2 + activity inte gr ation in astrocytes? One possibility is an amplification of the Ca 2 + signals. The Ca 2 + response in soma was several fold higher than that in the pr ocesses. Differ ent lev els of [Ca 2 + ] i r egulate distinct molecular pathways. 58 Thus, Ca 2 + amplification may target specific cellular functions. For example, somatic [Ca 2 + ] i elevation can boost ATP pr oduction by pr omoting the acti vity of the tricarboxylic acid cycle enzymes, the proteins of the electron transport chain, and the ATP synthase. 59 , 60 Such enhanced metabolic response of astrocyte can be the physiological outcome of Ca 2 + inte gr ation, homologues to, and yet distinct from, synaptic inte gr ation and action potential generation in neurons.

Ca 2 + Oscillations in Astrocytic Soma
In addition, Ca 2 + inte gr ation in astrocytic soma triggers [Ca 2 + ] i oscillations. [Ca 2 + ] i oscillations in astrocytes were routinely r ecorded in primar y cultur es and in slices in r esponse to pharmacological stimulation. [61][62][63][64] Mathematical models of astrocytic [Ca 2 + ] i dynamics also predict oscillatory activity due to positive and negati v e feedbacks. 65 , 66 Although astrocytes, in principle, should possess mechanisms for [Ca 2 + ] i oscillations, oscillatory behavior was not reported in vivo. A possible reason is that Ca 2 + acti vity in astr ocytes in vi v o is defined by inputs fr om many elements of the brain acti v e milieu (local neuronal network, neur omodulator y pr ojections, b lood v essels, etc.), which dri v e the sequence of astrocytic Ca 2 + events. Ho wever, astroc yte oscillatory mechanisms can re-emerge during somatic [Ca 2 + ] i elevations. These oscillations originate specifically in the soma and fade a wa y when pr opa gating tow ard the astr ocytic peripher y.
[Ca 2 + ] i oscillations r equir e [Ca 2 + ] i to exceed a certain threshold. Large [Ca 2 + ] i elevations and well-expressed Ca 2 + stores in the soma provide optimal conditions for [Ca 2 + ] i oscillations. 65 Possib l y, the information is encoded in the frequency of [Ca 2 + ] i oscillations that depends on the strength of the external signal and intrinsic properties of an individual astrocyte. The oscillatory Ca 2 + signal is decoded by the enzymes containing Ca 2 + binding motifs with a distinct affinity that can regulate their phosphorylation and activate specific cellular programs. 67

Conclusion
We demonstrate that Ca 2 + dynamics in astrocytes does not follow neuronal Ca 2 + activity that accompanies locomotion, indicating that astrocytic [Ca 2 + ] i is controlled by pathways distinct fr om sensor y inputs that acti v ate neur onal networks. While neur onal [Ca 2 + ] i faithfull y follows e pisodes of animal locomotion, astrocytic Ca 2 + signals develop on a different timescale, compatible with distinct roles astrocytes play in sustaining nervous tissue and supporting neuronal information processing and stor age .