Molecular recording of calcium concentration via calcium-dependent protein proximity labeling

Calcium ions serve as key intracellular signals. Local, transient increases in calcium concentrations can activate calcium sensor proteins that in turn trigger downstream effectors. In neurons, calcium transients play a central role in regulating neurotransmitter release and synaptic plasticity. However, it is challenging to capture the molecular events associated with these localized and ephemeral calcium signals. Here we present an engineered biotin ligase that combines the power of genetically encoded calcium indicators with protein proximity labeling. The enzyme, Cal-ID, biotinylates nearby proteins within minutes in response to elevated local calcium levels. The biotinylated proteins can be identified via mass spectrometry and visualized using microscopy. In neurons, Cal-ID labeling is triggered by neuronal activity, leading to prominent protein biotinylation that enables transcription-independent activity mapping in the brain. Cal-ID produces a biochemical record of calcium signaling and neuronal activity with high spatial resolution and molecular specificity.


Introduction
Calcium ions (Ca 2+ ) are universal second messengers in eukaryotic cells. Ca 2+ play essential roles in cellular physiology and Ca 2+ homeostasis is critical to cell viability. Cytosolic Ca 2+ levels are typically maintained at ~100 nM, which is more than 10,000-fold lower than the extracellular Ca 2+ levels 1 . However, cytosolic Ca 2+ can increase rapidly due to release from intracellular stores or influx through plasma membrane channels. The resulting Ca 2+ transients activate Ca 2+ sensor proteins, which often bind to effector proteins in a Ca 2+ -dependent manner to exert regulatory function. Ca 2+ signaling is often localized to subcellular microdomains that form near the sites of Ca 2+ influx as well, and thus can control both the position and timing of downstream Ca 2+ signaling features prominently in neuronal physiology, where tight spatial and temporal control of signaling events is critical. Ca 2+ regulates neurotransmitter release at the presynaptic active zone, and neurotransmitter receptors in turn allow Ca 2+ influx into the postsynaptic region; additionally, G-protein-coupled neurotransmitter receptors trigger endoplasmic reticulum (ER)-mediated Ca 2+ signaling 3 . Local Ca 2+ dynamics play a key role in synaptic plasticity, and activity-dependent gene expression requires Ca 2+ signaling to the nucleus 4 .
Indeed, Ca 2+ dynamics can serve as a valuable proxy of neuronal activity. Voltage-gated Ca 2+ channels allow Ca 2+ influx from the extracellular space when the plasma membrane depolarizes 5 . Therefore, chemical and genetic tools to monitor Ca 2+ levels, such as Fura and Fluo dyes, and genetically encoded Ca 2+ indicators such as GCaMPs, have been extensively utilized to report on neuronal activity in real time 6 . These tools, which enabled researchers to track activities of many neurons via live imaging, have become a landmark technological advance in neuroscience. However, they do not provide information about the molecular events occurring at the site of Ca 2+ signaling events.
In the last decade, proximity labeling has emerged as a new approach for surveying macromolecular interaction and localization in living cells 7 . In proximity labeling, an engineered enzyme covalently modifies nearby proteins, often by generating diffusible but short-lived reactive compounds. For example, the E. coli Bifunctional ligase/repressor (BirA) has been adapted to promiscuously biotinylate proximal proteins by releasing an unstable intermediate 8,9 . Biotinylated proteins can then be identified via mass spectrometry, providing a powerful workflow for an unbiased, high-resolution mapping of protein-protein interaction.
Importantly, proximity labeling captures these interactions in their biological context. We thus hypothesized that if proximal protein labeling could be done in a Ca 2+ -dependent manner-with a labeling enzyme activated by elevated local Ca 2+ levels-it would enable proteomic investigation of Ca 2+ signaling microdomains.
In this study, we engineered a biotin ligase that switches conformation between inactive and active states depending on its Ca 2+ -binding status. This enzyme, named Cal-ID, biotinylates nearby proteins when local Ca 2+ levels are elevated. We expressed Cal-ID in HEK293T cells and investigated biotinylated proteins, allowing us to identify CEP131 and ASPM as primary components of the centrosomal Ca 2+ signaling microdomain during mitosis. The localization of Ca 2+ signaling microdomains and Cal-ID labeling exceeds the resolution of light microscopy, so we employed expansion microscopy technology to visualize ER Ca 2+ release hotspots on the ER membrane. We also applied Cal-ID to mouse primary cortical neurons and found that plasma membrane Ca 2+ ATPases (PMCA) 1 and 2 occupy the most prominent Ca 2+ signaling microdomains. Cal-ID expressed in the mouse brain allowed us to distinguish activated neurons during an initial phase of kainic acid-induced seizure. These results highlight the value of Cal-ID as a novel and powerful tool that provides a new way to trace molecular events associated with local Ca 2+ concentration in cells, including Ca 2+ influx associated with neuronal activity.

Development of Cal-ID, a Ca 2+ -dependent, promiscuous biotinylation ligase
The genetically encoded Ca 2+ indicator GCaMP functions via a Ca 2+ -dependent conformational change that allows reversible intramolecular protein complementation of split green fluorescent protein (GFP). We envisioned that Ca 2+ -dependent complementation of a split proximity labeling enzyme, instead of split GFP, would enable proximity protein labeling in response to elevated local Ca 2+ levels. We chose the BioID labeling enzyme, which is an R118G mutant of the BirA biotin ligase, because it does not require any toxic or synthetic substrates and is proven to work in living mouse brain [10][11][12] . Based on the configuration of the genetically encoded calcium indicator GCaMP6s, we fused the Ca 2+ sensor calmodulin and the Ca 2+dependent calmodulin-binding peptide RS20 with circularly permuted, split BioID 13-15 (Extended Data Fig. 1a). We expressed these fusions in HEK293T cells and found that a variant split between T195 and G196 has increased levels of promiscuous biotinylation when cytosolic Ca 2+ levels were elevated by treatment with the sarco/endoplasmic reticulum Ca 2+ -ATPase (SERCA) inhibitor thapsigargin (Extended Data Figs. 1b and 1c). Of note, this T195/G196 split site has not been tested in previous studies of split BirA variants [14][15][16][17] . Since this engineered enzyme successfully responded to elevated cytosolic Ca 2+ levels in living cells, we named it Cal-ID (Calcium-dependent BioID).
Because Cal-ID labeling is irreversible, background levels of Ca 2+ -independent activity pose a more serious challenge than background signal in fluorescence-based Ca 2+ indicators.
Biotinylated proteins may accumulate over extended labeling periods, or in response to high Cal-ID expression levels. Therefore, we set out to suppress activity of Cal-ID in the low-Ca 2+ state and to enhance its signal-to-noise ratio. First, since a highly active, promiscuous BirA derivative called TurboID has been developed, we introduced the TurboID mutations to Cal-ID 8 .
Next, we designed an inhibitory peptide, AviTag*-a non-biotinylatable and truncated variant of the wild-type BirA substrate peptide AviTag-that should block the enzyme active site. We fused this pseudosubstrate peptide to the N-terminus of Cal-ID, where it inhibits the enzyme in the absence of Ca 2+ , yet is sequestered in response to Ca 2+ -dependent calmodulin-RS20 binding (Fig. 1a). Indeed, we found that the N-terminal AviTag* fusion increased the Ca 2+ responsiveness of Cal-ID during an extended 1-hour incubation in high extracellular biotin (Extended Data Fig. 2a). Last, a recent study reported a new split site, L73/G74, for proteinfragment complementation of TurboID 16 . We revised Cal-ID with this split site since the new split site variant shows prominent biotinylation only after a 15-min incubation (Extended Data We then tested Ca 2+ -dependent proximal biotinylation by this revised Cal-ID using various modes of cytosolic Ca 2+ induction. In order to mimic the activity-dependent Ca 2+ influx in neurons, we expressed the L-type voltage-gated calcium channel (L-VGCC) pore-forming alpha 1C subunit (CaV1.2) in HEK293T cells and showed that KCl depolarization triggers Cal-ID biotinylation (Fig. 1c). We also showed that Ca 2+ influx mediated by the Ca 2+ ionophore ionomycin also activates Cal-ID, similar to the effects of thapsigargin (Fig. 1d). Beyond these artificial stimuli, we observed that replenishing serum in the cell culture media also activates Cal-ID (Fig. 1d), suggesting that the activation of signaling cascades, which ultimately trigger ERmediated Ca 2+ release, represents a physiological stimulus that activates Cal-ID.
Cal-ID revealed cell cycle-dependent Ca 2+ microdomains at centrosomes Because TurboID has much higher activity than Cal-ID, it is not an ideal point of comparison for Cal-ID labeling experiments. We thus constructed cpTurboID, a circularly permuted TurboID without calmodulin or RS20, as a Ca 2+ -independent control enzyme with labeling activity more similar to that of Cal-ID (Fig. 1b). cpTurboID is activated by spontaneous refolding, thereby providing a better control for the background activity of Cal-ID. It is also expected to reduce the risks of cellular stress from depleting cellular biotin over extended incubation time, which occurs with highly reactive TurboID 8 . To compare these enzymes at defined expression levels, we generated stable HEK293T cell lines with tetracycline-inducible Cal-ID or cpTurboID expression via Sleeping Beauty transposition and performed biotinylation experiments 18,19 . We found that cpTurboID has stronger yet comparable biotinylation levels to Cal-ID when they are expressed at similar levels and incubated for the same amount of time (Fig. 2a). In addition, we found that treatment of BAPTA-AM, a cell membrane-permeable form of the Ca 2+ chelator BAPTA, reduces Cal-ID-mediated biotinylation but does not affect cpTurboID-mediated biotinylation in cells (Extended Data Fig. 2c). These results demonstrate that cpTurboID is a suitable Ca 2+ -independent control for Cal-ID.
We next visualized the subcellular labeling pattern in Cal-ID-expressing cells using streptavidinconjugated fluorophores and compared them with cpTurboID-expressing cells. cpTurboID biotinylation was observed throughout the cytosol, matching our expectations for a free and cytosolic promiscuous biotin ligase (Fig. 2b). In contrast, many Cal-ID-expressing cells showed distinctive, cell cycle-associated subcellular labeling patterns. In particular, mitotic Cal-IDexpressing cells showed strong biotinylation at the centrosome, which was visualized by γtubulin (Figs. 2b; see arrows, and 2b). Indeed, the importance of a Ca 2+ microdomain in mitotic spindle formation is well established, and the elevated local Ca 2+ concentration at the centrosome has been observed via genetically encoded Ca 2+ indicators 20-23 . By detecting these well-studied Ca 2+ microdomains, our results demonstrate that Cal-ID can detect physiologically relevant Ca 2+ signaling events and demonstrate the spatial resolution of Ca 2+ -dependent, proximal protein labeling by Cal-ID.

Quantitative proteomics identified Ca 2+ signaling constituents in living cells
To investigate the distinct Cal-ID biotinylation patterns observed by microscopy in more detail, we wished to identify Cal-ID target proteins by proteomic mass spectrometry. Proximity labeling produces complex proteomic samples, but we and others have shown that ratiometric comparison of biotinylated samples can reveal specific patterns of biotinylation enrichment in the cell [24][25][26] . We thus carried out proximity labeling in Cal-ID-and cpTurboID-expressing cells and prepared samples for quantitative mass spectrometry with tandem mass tag (TMT) labeling for a ratiometric comparison of biotinylated proteins (Fig. 2d).
We detected 964 proteins overall, and biological replicates showed high quantitative reproducibility in both labeling enzymes (r 2 > 0.98; Extended Data Figs. 2d and 2e). We then compared cpTurboID-and Cal-ID-enriched proteins and identified 29 proteins that were significantly enriched in Cal-ID relative to cpTurboID samples at 5% false discovery rate (FDR) and log2 fold change > 0.5 cutoff (Fig. 2e, Supplementary Table 1). Cal-ID-enriched proteins are a subset of overall proteins labeled by cpTurboID, and so this analysis was designed to discover the most significant Cal-ID-specific labeling, but not to build an exhaustive list of the Ca 2+associated proteome. The dramatic enrichment of calmodulin presumably reflects Cal-ID selfbiotinylation and recovery of tryptic fragments from this fusion protein. Due to its high sequence similarity, we cannot rule out the possibility that endogenous calmodulin is a major target of Cal-ID biotinylation, and so it could have contributed to the high enrichment of calmodulin as well. The second most enriched protein is CEP131, a centrosomal component protein. In addition to CEP131, ASPM, which is known to be associated with mitotic spindle regulation 27 , is also significantly enriched by Cal-ID labeling 27 . This result is consistent with the robust Cal-ID biotinylation at centrosomes in mitotic cells. In addition, we found that 5 out of component of the mitotic spindle, and β-Tubulin staining near the centrosome overlaps with fluorescent streptavidin staining (Extended Data Fig. 3a). Beyond CEP131, ASPM, and tubulins, Cal-ID enriches several known Ca 2+ signaling-associated proteins. These labeled proteins include IQGAP1, a Ras GTPase-like protein with IQ domains that bind to calmodulin, and MCU, a mitochondrial Ca 2+ transporter involved in mitochondrial Ca 2+ uptake 28,29 . We also saw enrichment of MYO6 and MYL6 myosin, which requires Ca 2+ for their motor function. Last, we identified FKBP7, an ER chaperone that is thought to bind to Ca 2+ because its mouse ortholog has Ca 2+ -binding affinity 30 .
Since several Cal-ID target proteins identified by mass spectrometry have known subcellular localization, we investigated their potential co-localization with Cal-ID biotinylation. We identified Cal-ID enrichment of two nucleolar proteins, SURF6 and ZCCHC17. We immunostained SURF6, which showed higher enrichment than ZCCHC17, and found that SURF6 was strictly localized to the nucleolus in non-mitotic cells, although streptavidin signal was relatively weak in the nucleolus (Extended Data Fig. 3b; top panel). However, during mitosis, SURF6 can be found in the cytosol and along the spindle fiber, suggesting that SURF6 biotinylation is mostly occurring during mitosis (Extended Data Fig. 3b; bottom panel). While the function of human SURF6 in mitosis is not clear, yeast orthologs of SURF6 are known to be involved in spindle formation, which potentially explains SURF6 localization and biotinylation during mitosis 31 . In addition, when we performed immunostaining of FKBP7, an ER protein, we were able to find a subset of cells that show ER biotinylation patterns overlapping with FKBP7 (Extended Data Fig. 3c). It is possible that these cells have active signaling cascades triggering Ca 2+ release from the ER. In summary, these results collectively show that Cal-ID combined with ratiometric quantitative proteomics can identify physiologically relevant Ca 2+ signaling microdomains in living cells.

ER Ca 2+ release hotspots visualized by expansion microscopy
Having characterized the strongest biotinylation targets of free cytosolic Cal-ID, we sought to test whether Cal-ID can be localized to an organelle in order to record local Ca 2+ concentration.
Particularly, we showed that serum replenishment activates Cal-ID potentially through ER Ca 2+ release downstream of growth factor signaling (Fig. 1d). Thus, we hypothesized that Cal-ID targeted to the outer membrane of the ER would record Ca 2+ concentration microdomains formed by ER Ca 2+ release following serum replenishment. We generated ER outer membrane- We wished to visualize ER Ca 2+ release by super-resolution microscopy as the size of Ca 2+ microdomains can fall below the Abbe diffraction limit for light microscopy 2,3 . We reasoned that expansion microscopy (ExM), which achieves effectively higher resolution by physically expanding a biological specimen 33 , could be readily applied to Cal-ID biotinylated samples because the biotinylated proteins and the streptavidin-fluorophore conjugates can be preserved through the physical expansion process. We labeled sites of serum-induced ER Ca 2+ release in HEK293T cells using Cal-ID-ERM, expanded the cells by ~4.5× using the protein retention ExM (proExM) method 34,35 , and imaged the expanded sample with a confocal microscope (Fig. 3a). Imaging of biotinylated proteins using fluorophore-conjugated streptavidin visualized Ca 2+ concentration 'hotspots' near the ER 36 responding to serum activation, while the enzyme was expressed throughout the ER ( Fig. 3b; top panel). Higher magnification images revealed that these hotspots are overlapping with Cal-ID-ERM epitope V5 signals or located adjacent to the V5 signals (

Cal-ID is activated by neuronal activity
In neurons, cytosolic Ca 2+ influx from the extracellular space and release of Ca 2+ from intracellular storage are key events in the regulation of neuronal function. Neuronal activity is tightly associated with Ca 2+ influx through voltage-gated Ca 2+ channels 37 . In addition, synaptic Ca 2+ influx through neurotransmitter receptors, like NMDA receptors, is important for synaptic plasticity 37 . We therefore set out to express Cal-ID in mouse primary neurons and test whether Cal-ID biotinylation is increased by neuronal activity. We delivered Cal-ID or cpTurboID to mouse primary cortical neurons by lentiviral transduction (Fig. 4a). To test how Cal-ID responds to neuronal activity, we silenced or activated the neuronal cultures and compared the overall Cal-ID biotinylation levels via western blot. We silenced neuronal cultures with NBQX and APV, which are antagonists of AMPA and NMDA glutamate receptors, respectively. For activation, we first silenced the culture and then treated with CaCl2 to induce neuronal activity 38 . We found that the extent of Cal-ID mediated biotinylation tracked with neuronal activity in the culture, while cpTurboID-mediated biotinylation was unaffected by activity levels (Fig. 4b; see arrow).
We next performed Cal-ID biotinylation in primary hippocampal neurons and visualized biotinylated proteins using streptavidin-conjugated fluorophores. We again found that activated neuronal cultures showed higher overall biotinylation levels than silenced cultures (Fig. 4c). In order to investigate cell-to-cell variation in Cal-ID labeling within a single culture and correlate it with neuronal activity, we generated a Cal-ID-P2A-mScarlet-I fusion that synthesizes mScarlet-I co-translationally with Cal-ID and splits the two proteins into separate polypeptides by means of the P2A sequence. After Cal-ID labeling, we stained neurons to visualize a classic marker of neuronal activity, phospho-rpS6 (Ser235/236), along with biotinylation 39 . We found that mouse primary hippocampal neurons expressing Cal-ID showed a strong positive correlation between streptavidin and phospho-rpS6 signal intensity (Figs. 4d and 4e). The expression levels of the Cal-ID protein, monitored by mScarlet-I, did not show a similar correlation with phospho-rpS6, so the difference in biotinylation reflects greater Cal-ID activity in neurons with high phospho-rpS6 rather than greater abundance of the enzyme (Extended Data Fig. 4a). Similarly, we also found a positive correlation between phospho-CaMKII and streptavidin signals but not from phospho-CaMKII and mScarlet-I signals from mouse hippocampal neurons (Extended Data Figs. 4b and 4c). The results indicate that Cal-ID labeling provides a biochemical strategy to study molecular events associated with neuronal activity that triggers intracellular Ca 2+ influx. We next sought to identify Cal-ID enriched proteins through a ratiometric proteomic comparison with cpTurboID. As in the experiment conducted with HEK293T cells, we prepared samples from mouse primary cortical neurons for quantitative mass spectrometry via TMT labeling. We captured 669 proteins in this labeling experiment, with high reproducibility (r 2 > 0.92) between biological replicates in both samples (Extended Data Figs. 6a and 6b). Similar to our analysis of the HEK293T mass spectrometry results, we again directly compared cpTurboID and Cal-ID enriched proteins to identify top targets of cytosolic Cal-ID in neurons and identified 6 proteins which were robustly biotinylated by Cal-ID with a 5% FDR (Fig. 5a). Calmodulin is again the protein most enriched by Cal-ID labeling, reflecting Cal-ID self-biotinylation and potential biotinylation of endogenous calmodulin as discussed above. We found two plasma membrane Ca 2+ ATPases (PMCA1 and PMCA2; Atp2b1 and Atp2b2, respectively) from the list of 6 significantly enriched proteins. PMCAs play a role in maintaining Ca 2+ homeostasis by pumping Ca 2+ out from the cytoplasm 40,41 . The C-terminus of PMCAs is known to have multiple Ca 2+ signaling-associated domains including calmodulin binding, high affinity Ca 2+ binding for allosteric regulation, and a protein kinase C (PKC) target site 42,43 . In addition to PMCAs, we identified wolframin (Wfs1), an ER membrane protein known to be involved in Ca 2+ homeostasis, as a target of Cal-ID. Mutations in Wfs1 cause Wolfram syndrome, which is characterized by juvenile-onset insulin-dependent diabetes mellitus with optic atrophy, deafness, dementia, and psychiatric illness 44,45 . While the exact molecular mechanisms are unclear, it is thought to regulate ER Ca 2+ levels in the context of ER-mitochondria crosstalk 46,47 .
These top Cal-ID targets identified by mass spectrometry demonstrated the capacity of Cal-ID to detect Ca 2+ signaling microdomain-associated proteins in neurons.
Proximity ligation assay provides selective imaging of Cal-ID targets As PMCAs were identified as the top Cal-ID target proteins, we hypothesized that biotinylation of PMCAs may have contributed to the distinct streptavidin-fluorophore staining patterns in neurons expressing Cal-ID (Extended Data Fig. 4b). Therefore, we sought to perform immunostaining of PMCAs and check their co-localization with streptavidin signals. However, the broad plasma membrane localization and high abundance of the PMCAs, as well as the presence of endogenous biotin carrier proteins, posed a challenge to the interpretation of colocalization. We therefore turned to proximity ligation assays (PLA), which rely on very close (< 40 nm) co-localization of two different antibodies to create fluorescent signal (Fig. 5b) 48 . PLA between antibodies for biotin and the target protein provides a complementary method to confirm biotinylation of a specific target protein by Cal-ID and resolve its location more clearly.
To confirm the applicability of PLA to visualize biotinylation in neurons, we first examined selfbiotinylation of Cal-ID and cpTurboID enzymes using anti-biotin antibodies along with anti-V5 antibodies against the V5 epitope tag on the biotin ligases. Consistent with the previous streptavidin imaging and western blot results, we found an even starker difference in biotinylation between naïve and stimulated neurons expressing Cal-ID, but not from cpTurboIDexpressing neurons (Extended Data Figs. 6c and 6d). Next, we performed PLA with biotin and PMCA antibodies. We found that biotin-PLA signals from both PMCA1 and PMCA2 were significantly decreased when neuronal activity was suppressed (Figs. 5c to 5f). By focusing on biotin derived from proximity labeling and excluding endogenous biotin, these PLA results verify that Cal-ID biotinylation is increased by neuronal activity.
Cal-ID enables transcription-independent in vivo neuronal activity mapping Brain-wide activity mapping strategies typically rely on reporter expression that requires transcription 49 . Our imaging and mass spectrometry results from neuronal culture revealed that Cal-ID robustly biotinylates plasma membrane proteins in active neurons, suggesting that Cal-ID biotinylation can be used for transcription-independent in vivo activity mapping in the brain.
Biotin readily crosses the blood-brain barrier, and BirA-based proximal biotinylation in the brain has been reported with exogenous supplementation of biotin 12,50-52 . However, we noted that previous studies performed multiple rounds of biotin administration over several days to obtain biotinylated proteins.
Because multi-day labeling would not provide the high temporal resolution needed for brain activity mapping, we first tested whether a single biotin injection can produce detectable levels of biotinylated proteins in vivo. We expressed the constitutive cpTurboID labeling enzyme in the mouse brain by injecting adeno-associated virus (AAV) encoding cpTurboID into the cortex of neonatal mice at postnatal day 0 or 1 (P0/1). At 4 weeks of age, we intraperitoneally (i.p.) injected biotin and allowed biotinylation to proceed for 1 to 3 hours before sacrificing animals.
Robust protein biotinylation was detected from all timepoints, including 1 hour labeling (Extended Data Fig. 6e), indicating that biotinylation began within 1 hour of biotin i.p. injection.
We then tested whether Cal-ID labeling and biotin imaging can demarcate activated neurons.
After delivering Cal-ID by intracortical, unilateral AAV injection of P0/1 pups and allowing animals to mature for six weeks, we tested broad neuronal activation in the kainic acid-induced seizure model 53 . We performed biotin injection (i.p.), waited for 45 minutes to ensure that exogenous biotin reached the cortex, and then injected saline or kainic acid (i.p.) for the control or seizure conditions, respectively. We then monitored the mice for 45 minutes after the saline/kainic acid injection, ensured seizure behaviors (modified Racine scale 1 or higher 54 ) in the seizure condition, euthanized the animals, and prepared brain tissue for fluorescence microscopy (Fig. 6a). We found prominent neuronal biotinylation in the cortex of kainic acidinjected mice, but not from the cortex of saline-injected mice nor the hemisphere without AAV injection (Fig. 6b, Extended Data Figs. 7a and 7b), showing that Cal-ID neuronal biotinylation can effectively report on induced activity in the brain. We quantified the streptavidin signal intensity in Cal-ID-expressing neurons across 6 kainic acid-injected mice and found that it was significantly higher than the signal seen in Cal-ID-expressing neurons across 5 saline-injected mice (Figs. 6c and 6d). These in vivo results show that Cal-ID biotinylation offers a new way to map brain activity that is independent of transcription-based reporter expression and can be triggered by biotin injection.

Discussion
Ca 2+ signaling is fundamental to cell physiology and widely conserved across eukaryotes. Most Ca 2+ signals start from a transient increase of Ca 2+ at a specific cellular region 2 . While Ca 2+ diffuses rapidly, elevated local Ca 2+ concentrations can be captured by Ca 2+ sensor proteins and activate downstream signaling cascades. In this study, we exploited the function of the Ca 2+ sensor protein calmodulin to control the activity of a promiscuous biotin ligase in a Ca 2+dependent manner. Since BirA-mediated biotinylation has a short labeling radius (~10 nm), this strategy enables us to biochemically label proteins near the sites where local Ca 2+ levels are elevated 7 (Extended Data Fig. 8). Increased local Ca 2+ occupies Ca 2+ binding sites at calmodulin, shifting Cal-ID to the active, folded state via calmodulin-RS20 binding, thereby leading to proximal biotinylation. Once local Ca 2+ levels drop to resting state, calmodulin will release Ca 2+ , disrupting the interaction with RS20 and allowing the Cal-ID to return to an inactive conformation. This conformational cycle places a permanent covalent mark on proximal proteins in response to transient changes in local Ca 2+ concentration.
Cal-ID labeling in HEK293T cells highlighted Ca 2+ signaling microdomains at the mitotic centrosome. The necessity of local Ca 2+ signaling for mitotic spindle formation has been known for decades, which aligns well with our observation of robust Cal-ID biotinylation at the centrosome. It is noteworthy that, while multiple pericentriolar proteins such as pericentrin (PNCT) and AKAP450 contain potential calmodulin-binding motifs, the primary proteins constituting centrosomal Ca 2+ signaling microdomains have not been identified 55,56 . Our Cal-ID mass spectrometry results implicated CEP131 and ASPM-both of which contain a calmodulinbinding IQ domain-as key components for centrosomal Ca 2+ signaling. Potential Ca 2+ -dependent roles of CEP131 and ASPM in mitotic spindle regulation could expand our knowledge of Ca 2+ signaling events at the centrosome.
Ca 2+ transients on the outer surface of the ER, which have been seen in many different cell types, often show 'hotspots' of Ca 2+ release 2 . These hotspots are typically generated at ERplasma membrane or ER-mitochondria junctions, and they are thought to be the sites of local Ca 2+ signaling 36,57 . In this study, we recorded ER Ca 2+ release hotspots using Cal-ID-ERM and visualized them at super-resolution with ExM technology. With 20 minutes of recording, we were able to show strongly biotinylated proteins near the ER membrane and around some IP3R Ca 2+ channels, and potential diffusion or transport of these proteins from the original biotinylated sites. We demonstrated that the combination of these two technologies synergizes powerfully; ExM enables visualization of subcellular Ca 2+ microdomains recorded by Cal-ID at super-resolution with relatively high accessibility, as imaging can be performed with conventional diffraction-limited microscopes.
In neurons, Ca 2+ influx is associated with neuronal activity and plays an essential role in many regulatory processes 4 . At synapses, Ca 2+ influx via neurotransmitter receptors, Ca 2+ channels, and release of ER-stored Ca 2+ is a central signaling mechanism for local synaptic regulation on both short (milliseconds) and long (hours) timescales 58 . Ca 2+ signaling back to the nucleus is essential for activity-dependent gene expression 59 . We therefore expect that identifying Ca 2+ microdomain constituents via Cal-ID will advance our understanding of the molecular physiology of the neuron. In primary cortical neurons, Cal-ID labeling enriches Atp2b1 (PMCA1) and Atp2b2 (PMCA2)-major Ca 2+ pumps which are responsible for extruding cytosolic Ca 2+ .
The high abundance of PMCAs in neurons and their affinity to calmodulin may contribute to their Cal-ID labeling. However, there are other abundant neuronal proteins with known calmodulin-binding motifs, such as voltage-gated Ca 2+ channels and ryanodine receptors, which drive major Ca 2+ influx into cytoplasm in neurons 37 . Therefore, it is plausible to speculate that high local Ca 2+ concentrations and Ca 2+ signaling microdomains are maintained around PMCAs, since active export is relatively slower than Ca 2+ influx through channels. In addition, our results helped to speculate the subcellular location of fluorescent signals generated by genetically encoded Ca 2+ indicators that incorporate calmodulin, such as GCaMPs. Further investigation of the Ca 2+ signaling events around PMCAs is warranted to expand our understanding on the formation and regulation of Ca 2+ microdomains in neurons.
With prominent plasma membrane biotinylation that facilitates visualization of Cal-ID-labeled neurons, Cal-ID labeling and imaging could be especially advantageous in neuroscience as an approach to track neuronal activity. Cal-ID biotinylation can be temporally controlled through biotin delivery and reaches detectable levels within 1 hour of labeling. We showed that Cal-ID biotinylation in the mouse cortex was elevated during induced activity caused by kainic acid treatment, leading to prominent and non-uniform labeling of individual neurons. These in vivo results show that Cal-ID provides a new strategy for brain activity mapping by attaching covalent biochemical marks to nearby proteins in response to Ca 2+ influx, instead of expressing reporter transcripts or proteins under control of the Ca 2+ signaling cascade. The nature of proximity labeling gives Cal-ID the potential to record neuronal activity at a subcellular level-perhaps even at individual synapses-across the whole brain. Furthermore, proteomic analysis of biotinylated proteins will enable identification of proteins associated with local neuronal activity in the brain. Of note, this approach may offer distinctive advantages over other tools such as CaMPARI, a photoactivatable Ca 2+ indicator 60,61 . While CaMPARI can 'mark' neuronal activity at a great temporal resolution, the need for light activation limits the maximum area of recording. However, enzymatic labeling by Cal-ID will be much slower than the millisecond timescale of CaMPARI, which must be considered when interpreting these data. is possible that Cal-ID localization is affected by the inclusion of calmodulin, and proteins with calmodulin-binding motifs are more easily targeted for biotinylation. Since many calmodulinbinding partners have Ca 2+ -dependent binding affinity, and most calmodulin-associated molecular processes are involved in Ca 2+ -dependent regulation, we believe that the possibility of identifying false positives due to calmodulin-binding affinity is low 63 . Nonetheless, we suggest that Cal-ID labeling results should be interpreted with due consideration for the existence of calmodulin-binding motifs.
In this study, we presented Cal-ID, an engineered proximal protein ligase that is activated by Ca 2+ -dependent refolding. Cal-ID functions as a molecular recorder of Ca 2+ microdomains, thereby providing a unique way to investigate Ca 2+ signaling events in cells. We showed that Cal-ID-labeled proteins can be analyzed in many different ways by employing technologies such as mass spectrometry, proximity ligation assay, and expansion microscopy. We also demonstrated the in vivo capability of Cal-ID by successfully labeling neurons in the activated mouse brain. Thus, we suggest that Cal-ID has great application potential in multiple fields of biology in which cellular activation accompanied by Ca 2+ influx is important.

Methods
All animal protocols are in accordance with the regulations of University of California, Berkeley Animal Care and Use Committee and the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. Animals were housed in a 12-hour reverse dark/light cycle with free access to water and food.

Cloning
For cloning, PCR primers were designed to have overhangs, PCR reactions were performed with Q5 polymerase (New England BioLabs, NEB), and the fragments were assembled using HiFi (NEB) or Gibson assembly (NEB). Ligated plasmid products were introduced by heat shock transformation into competent Stbl3 (Invitrogen). The cloning products were confirmed using Sanger sequencing. Calmodulin and RS20 sequences were derived from GCaMP6s 13 . All biotin ligase variants were derived from E. coli biotin protein ligase, from BioID or TurboID 8,9 . HEK293T Cell Culture and Generation of Stable Cell Lines HEK293T cells were maintained with DMEM with GlutaMax (Gibco) supplemented with 10% tetracycline-free FBS (Gibco). Cal-ID and cpTurboID transgenes were cloned into the pSBtet-Hyg After that, 1 × Duolink® Ligation buffer was prepared, ligase was added following 1:40 ratio, mixed well and added to the coverslips. The ligation reaction was performed for 30 min at 37 °C in a humidity chamber. Following ligation, coverslips were washed twice with Wash A buffer, and 1 × Amplification reaction mix with 1:80 enzyme:buffer ratio was prepared and added. The amplification reaction was performed for 100 min at 37 °C in a humidity chamber. Final wash was performed with 1 × Wash B buffer twice for 10 min then 0.01 × Wash B buffer once for 1 min. The coverslips were mounted using VECTASHIELD PLUS® (Vector Labs) or Duolink® DAPI mounting solution provided with the kit.

Image Analysis
For streptavidin and P-S6/P-CaMKII correlation, PLA, brain slice analyses, ImageJ (1.53f51)/FIJI was used 68 . Cell bodies of the mScarlet-I or V5-expressing neurons were selected via Threshold, Watershed, and Analyze Particles tools with 50 -infinity size selection in ImageJ. Cells on the edges were excluded. After selecting soma regions, target mean intensity values were extracted from the selections and transferred to R. The results were analyzed in R and visualized using ggplot2 (3.0.0) 69 . The heatmap was generated by Calibration Bar tool in ImageJ. Movie was generated using IMARIS (Oxford Instruments, version 10).

Subcellular Fractionation
Samples were kept on ice during the procedure and centrifugations and incubations were conducted at 4 °C. All solutions contained phosphatase inhibitors (Sigma-Aldrich PhosSTOP TM ), protease inhibitors (Roche cOmplete TM ), 5 mM sodium pyrophosphate, 1 mM EDTA, and 1 mM EGTA. Cultured cortical neurons were collected by scraping and homogenizing by passage through a 26G needle 12 times, in homogenization buffer (320 mM sucrose, 10 mM HEPES).
The homogenate was centrifuged at 800 g for 10 min to obtain the post-nuclear pelleted fraction 1 (P1) and supernatant fraction 1 (S1). S1 was further centrifuged at 15,000 g for 20 min to yield P2 (membranous fraction) and S2 (cytosolic fraction). P2 was resuspended in Milli-Q® water, adjusted to 4 mM HEPES (pH 7.4), and incubated with agitation for 30 min before centrifugation at 25,000 g for 20 min to obtain LP1 (mitochondria, pre-and postsynaptic membranes) and LS2 (synaptosomal cytosolic fraction). LP1 was resuspended in 50 mM HEPES (pH 7.4), mixed with an equal volume of 1 % Triton X-100, and incubated with agitation for 8 min. Lastly, centrifugation at 25,000 g for 20 min yielded the presynaptic membranes (supernatant) and the postsynaptic density (pellet).
Intracranial neonatal mouse injections CD1 neonatal mice (P0) were cryo-anesthetized by placing on ice for ~2-3 min. When the animal was fully anesthetized, confirmed by lack of movement, it was gently placed in a head mold. Each pup received a total of 700 nL of 1:1 saline-diluted AAV9 (HHMI Viral Tools, Cal-ID: 7. 6×10 12 GC/mL, cpTurboID: 5.