Engineering and biochemical characterization of FRCaMPi

The NCaMP7-like topological design [10], which involves inserting the sensing domains within the FP, has recently resurged in the field of fluorescent biosensors and resulted in the successful construction of several green ncp indicators for glutamate and calcium. Those examples include ncpiGluSnFRs [19] for glutamate, ultrahigh-affinity GECIs [21] for detecting low nanomolar Ca²⁺ levels (<100 nM), bioluminescent Ca²⁺ indicator based on ncpGCaMP6s [20], and mNeonGreen-based Ca²⁺ indicators of superior sensitivity [9,10]. Recently, FRCaMP was introduced as the first red GECI engineered based on the calcium-binding domain CaM and the M13 peptide from a non-metazoan origin, possessing a similar dynamic range but a greater sensitivity than R-GECO1 [24]. In addition, FRCaMP did not interact with the cytosolic environment of the mammalian cells. Motivated by these advancements, we initially explored the benefits of an altered topology in FRCaMP. To alter the topology of FRCaMP, we genetically connected the terminal of M13 and CaM with a flexible linker (GGSGGSGSGS) and reintroduced the original termini of mApple protein (Fig 1A). We then termed this ncpFRCaMP as FRCaMP with the inverted topology (FRCaMPi; i stands for “inverted”). Consistent with previous observations of the biochemical properties of ncp indicators, in vitro characterization revealed that FRCaMPi binds Ca²⁺ with a dissociation constant (K_d) that is approximately 3-fold lower (81 ± 2 nM) compared to FRCaMP (214 ± 6 nM; S3 Table and S1D Fig). Similarly, FRCaMPi inherited the outstanding dynamic range (16.3 ± 0.3-fold) from FRCaMP (16.4 ± 0.7). The Hill coefficient for FRCaMPi was slightly higher (3.1 ± 0.2) compared to that of FRCaMP (2.5 ± 0.2) (S3 Table). In addition, the biophysical properties of FRCaMPi, such as fluorescence spectra, quantum yield, extinction coefficient, dynamic range, and pK_a, remained very similar to the parental FRCaMP (S1B and S1C Fig and S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Design of FRCaMPi and characterization in HeLa cells.

(A) FRCaMPi sensor is created by topological inversion of FRCaMP. Upper panel, a diagram showing the design of FRCaMP and FRCaMPi; the lower panel predicted 3D protein structures of indicators75 and a table of amino acid sequences of the linkers. (B) Basal brightness and photobleaching curves of red GECIs characterized in transfect HeLa cells by side-by-side comparison. For the plasmids used to transfect HeLa cells, a nuclear exclusion sequence (NES) was added to each red GECI and mGreenLantern reference was attached via a P2A peptide to express red and green FP in near equimolar ratio. *NES was not added to XCaMP-R due to the presence of F2A sequence C-terminal reported to have nuclear exclusion. The basal fluorescence of red GECIs (F_red) was normalized against the fluorescence of mGreenLantern (F_green). Fluorescence trace of FRCaMPi, jRGECO1a, XCaMP-R, and K-GECO1 over 10-min photobleaching, normalized to the mean brightness of jRGECO1a in the first frame. (C) Representative images showing the expression of mGreenLantern and FRCaMPi in HeLa cells. (D) Representative fluorescence trace induced by histamine (50 μM) and ionomycin (5 μM) in HeLa cells expressing K-GECO1, XCaMP-R, jRGECO1a, and FRCaMPi. (E) ΔF_max/F_min of FRCaMPi compared to other red GECIs during the period of pharmaceutically induced Ca²⁺ response (K-GECO1: n = 53 cells from 4 wells; XCaMP-R: n = 71 cells from 5 wells; jRGECO1a: n = 50 cells from 4 wells; FRCaMPi: n = 50 cells from 4 wells, two independent cultures). Kruskal–Wallis multiple-comparison test: p < 1 × 10–15. Pairwise Dunn’s comparison test with FRCaMPi: p = 5.92e–5 (jRGECO1a), p = 9.73e–7 (XCaMP-R) and p < 1e–15 (K-GECO1). (F) Scatter plot of ΔF_max/F_min against basal brightness of individual cells. Statistics were the same as (E). The quantitative data presented in this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003048.g001

The increased calcium binding affinity and preserved key features from FRCaMP, such as large dynamic range, led us to compare the performance of FRCaMPi directly with FRCaMP in cultured neurons using external electrical bursts at 87 Hz of 300 APs to evoke Ca²⁺ responses. Notably, FRCaMPi demonstrated a peak fluorescence response that was 2.36 times as high as that observed with FRCaMP (S2B Fig). We found that the decay half kinetics is slightly slower for FRCaMPi in comparison to FRCaMP (S2C Fig). Sensors featured by a higher calcium binding affinity are known to buffer a greater quantity of Ca²⁺ and could result in heightened sensitivity but prolonged decay kinetics [2,3,6,8,21]. This was consistent with our observation in FRCaMPi. In conclusion, topological inversion of FRCaMP resulted in a new sensor with over twice the sensitivity of its precursor, thus representing a promising candidate for further applications in neurons both in vitro and in vivo.

Characterization of FRCaMPi in mammalian cell cultures

Since FRCaMPi demonstrated even higher sensitivity in the initial in vitro characterization, we decided to benchmark FRCaMPi performance against other known state-of-the-art red calcium indicators, including X-CaMPR [16], K-GECO1 [15], jRGECO1a [13] (Table 1), characterized by high sensitivity in the cytosol of mammalian cells. To demonstrate the utility of FRCaMPi in mammalian cells, we first characterized the indicator in cultured HeLa cells. By expressing a bicistronic vector, the basal brightness of red GECI was reported as the fluorescence ratio of the indicator fluorescence to the fluorescence of a reference protein, mGreenLantern (Fig 1B and 1C). A similar brightness level was found between FRCaMPi and jRGECO1a and the brightness level of K-GECO1 was slightly higher (1.2-fold) compared to FRCaMPi (Fig 1B). Under continuous wide-field illumination at 7.4 mW/mm² (about 4–5 times higher than typically used for FRCaMPi imaging in cultured cells), the photobleaching rate of FRCaMPi was about 2- and 3.6-folds faster than those of jRGECO1a and K-GECO1, respectively. Nevertheless, we observed a quicker signal plateau of FRCaMPi after 3 min with eventually a brightness level similar to jRGECO1a at 10 min of bleaching time (Fig 1B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. In vitro biophysical properties of red GECIs.

https://doi.org/10.1371/journal.pbio.3003048.t001

To assess the capability of FRCaMPi to detect Ca²⁺ signals in HeLa cells, we recorded the fluorescent responses to cytoplasmic Ca²⁺ transients elicited by consecutive application of histamine followed by ionomycin. During pharmacologically induced Ca²⁺ changes, the maximal fluorescence response (ΔF_max/F_min) of HeLa cells expressing FRCaMPi was at least 2-fold to the response of XCaMP-R and jRGECO1a and was more than 4 folds higher than that for K-GECO1 (Fig 1D and 1E and S1 Movie). To this point, we noted that FRCaMPi exhibited a larger dynamic range without sacrificing a baseline fluorescence. In support of this notion, we also found that in pharmaceutically induced Ca²⁺ responses, cells expressing FRCaMPi showed matched basal fluorescence levels of those expressing jRGECO1a still exhibited a distribution over higher dynamic range, indicating a higher maximal brightness of FRCaMPi that can possibly achieve (Fig 1F).

Next, we systematically and quantitatively assessed ΔF/F₀, SNR, and kinetics of response of FRCaMPi in cultured mouse hippocampal neurons, adapting an electrical stimulation protocol that was well established for other GECIs, including GCaMPs, jRGECO1a, and NIR-GECO1 [2,5,13] (Figs 2A and S3A-S3C). Field stimuli (45–50 V, 83 Hz, 1 ms) were delivered in trains of 1, 2, 3, 5, 10, 20, 40, 80, and 160 pulses to transduced mouse hippocampal neurons and 5 min of illumination were applied prior to trace recording to remove initial bleaching to equalize the effect of bleaching to all indicators (S4E Fig). Fluorescence changes detected from cell bodies show that FRCaMPi possesses a 50–60% larger peak ΔF/F₀ compared to jRGECO1a and K-GECO1 at 1 pulse (Fig 2B and 2E and Table 2). As the number of electrical pulses increased, the peak difference of FRCaMPi to K-GECO1 and XCaMP-R became more distinct (2.77 and 3.34 folds, respectively) and stayed approximately 1.5 times as high as jRGECO1a (Fig 2B, 2C and 1E and Table 2). GECIs with a high calcium affinity and Hill coefficient tend to saturate more easily at smaller spikes [6]. However, despite possessing a low K_d and a Hill coefficient larger than 3, fluorescence responses of FRCaMPi showed fair linearity on the log scale with no notable saturation up to 160 pulses (Fig 2D and 2E). This behavior could be attributed to the extensive dynamic range of the indicator, as supported by experiments in HeLa cell calcium responses. The baseline brightness of FRCaMPi in primary mouse hippocampal neurons observed was not significantly different from that of jRGECO1a but higher than that of X-CaMPR and K-GECO1 (S4C Fig). Given a similar level of baseline brightness but more rapid bleaching, the peak SNR of FRCaMPi remains comparable to jRGECO1a after 5 min of illumination (Fig 2F and Table 2), which could be attributed by the larger sensitivity (i.e., peak ΔF/F₀ in response to pulse trains) of the indicator. The rise kinetics of FRCaMPi appeared similar to those of other indicators (Fig 2G and Table 2), while the decay kinetics of FRCaMPi was slower, with the 2.8-fold difference to jRGECO1a decreased over the increased number of pulses (Fig 2H and Table 2). Altogether, FRCaMPi faithfully reports Ca²⁺ dynamics in mammalian cell cultures with greater dynamics and sensitivity observed than existing red indicators of high sensitivity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Properties of red GECIs measured in dissociated neurons.

https://doi.org/10.1371/journal.pbio.3003048.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Characterization of FRCaMPi in primary mouse hippocampal neurons.

(A) FRCaMPi sensor is created by topological inversion of FRCaMP. Upper panel, a diagram showing the design of FRCaMP and FRCaMPi; the lower panel predicted 3D protein structures of indicators 75 and a table of amino acid sequences of the linkers. Representative images of a neuron co-expressing FRCaMPi (TRITC) and emiRFP670 (NIR) (top) and schematics of field stimulation protocol (n = 5 neurons from two independent cultures). Scale bar, 50 µm. (B–D) Averaged fluorescence traces generated in response to (B) one pulse, (C) 10 pulses, and (D) 160 pulses for tested GECIs. Traces are shown as mean ± s.e.m, indicated by solid lines and shaded area, respectively (K-GECO1, n = 54 cells from 3 coverslips, 2 independent cultures; XCaMP-R, n = 24 cells from 2 coverslips, 2 independent cultures; jRGECO1a, n = 65 cells from 4 coverslips, 3 independent cultures; FRCaMPi, n = 56 cells from 6 coverslips, 3 independent cultures). Graph inset displays boxplot of peak ΔF/F₀ of each GECI to applied action potentials. (E) Peak ΔF/F₀ of FRCaMPi compared to other red GECIs as a function of pulses. Descriptive statistics here and for panels g,h,i same as in C–E. (F) Normalized SNR of FRCaMPi compared to other red GECIs as a function of pulses. Normalized to the mean value of jRGECO1a at each used pulse number. (G) The half the rise time of red GECIs as a function of pulses. (I) Half decay time of red GECIs as a function of pulses. Graph inset displays boxplot of half rise/decay time of each GECI after 10 action potentials. Box plots are used here (insets in panels C, D, E, H, I). Box indicates the median and 25–75th percentile range, and the whiskers represent 1.5 times the interquartile range. Line plots are used here (E–H) and throughout the paper (dot, mean; error bars, s.e.m.) See S4 Table for detailed statistics and exact p values. The quantitative data presented in this figure can be found in S1 Data. Created in BioRender, https://BioRender.com/e01g235.

https://doi.org/10.1371/journal.pbio.3003048.g002

Red genetically encoded calcium indicators (GECIs) are known to have significantly compromised functional performance in vivo, particularly during long-term expression. Therefore, before proceeding with further development, we assessed the fluorescence response of FRCaMPi expressed in the central amygdala (CeA) neurons using fiber photometry (S5A-C Fig). Notably, even after three months of expression, FRCaMPi exhibited reproducible and reliable responses during licking behavior and air-puff stimulus (S5D and S5E Fig). The combined in vitro and in vivo results thus encouraged us to optimize the probe further for its soma-targeted version.

Development and characterization of soma-targeted FRCaMPi

A recent effort in developing soma-localized green GECIs [22,23] has inspired us to explore the possibility of engineering a competent red soma-localized counterpart. To achieve this, we appended C-terminus of FRCaMPi with the four peptides known to facilitate somatic localization and assessed them in cultured neurons for fluorescence changes (peak ΔF/F₀), SNR in response to evoked extracellular stimulation, and cytotoxicity (assessed as total number of functional cells obtained from transfected neuronal culture; S6 Fig). Based on the screening, we chose to use the RPL10-tagged FRCaMPi for further characterization and named the selected variant SomaFRCaMPi (Fig 3A-3C). The RPL10 peptide was previously used to generate the soma-localized versions of the GCaMP6 sensors, which is a component of the 60S ribosome subunit [18]. To understand whether the SomaFRCaMPi is a competitive soma-targeting indicator in red, we generated fusions of K-GECO1, XCaMP-R, and jRGECO1a with the same tag and linker for comparison and referred to them as SomaK-GECO1, SomaXCaMP-R, and SomajRGECO1a (S4B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. SomaFRCaMPi engineering and characterization via electric field stimulation in primary mouse hippocampal neurons.

(A, B) Schematics of expression cassette and neurons showing the engineering of soma-localized red GECIs by tethering red GECI to ribosome via rpl10a peptide. (C) Representative images of a neuron co-expressing SomaFRCaMPi (TRITC) and emiRFP670 (NIR) (n = 5 from two independent neuronal cultures). Scale bar, 50 µm. (D–G) Plot of fluorescence intensity along neurites, normalized to the intensity at the soma, from neurons expressing either (D) K-GECO1 or SomaK-GECO1, (E) XCaMP-R or SomaXCaMP-R, (F) jRGECO1a or SomajRGECO1a, (G) FRCaMPi or SomaFRCaMPi (K-GECO1: n = 6 neurites from 5 cells; SomaK-GECO1: n = 7 neurites from 5 cells; XCaMP-R: n = 6 neurites from 5 cells; SomaXCaMP-R: n = 6 neurites from 4 cells; jRGECO1a: n = 6 neurites from 5 cells; SomajRGECO1a: n = 6 neurites from 5 cells; FRCaMPi: n = 7 neurites from 6 cells; SomaFRCaMPi: n = 6 neurites from 6 cells). An unpaired student t-test was used. Dots, individual neurites; solid line, exponential fitting curve. The data of the non-targeted indicator was colored in gray to aid visualization. (H) Baseline brightness of soma-localized red GECIs characterized in primary hippocampal neurons by side-by-side comparison (SomaK-GECO1: n = 90 cells from 5 wells; SomaXCaMP-R: n = 84 cells from 4 wells; SomajRGECO1a: n = 85 cells from 5 wells; SomaFRCaMPi: n = 101 cells from 5 wells, two independent cultures). Kruskal–Wallis ANOVA was used followed by a post hoc Dunn’s multiple comparison test. Dot plot is used here and throughout the paper (dots, individual neurons; horizontal line, median; whiskers, interquartile range). Data is normalized to the median of the SomajRGECO1a values. (I–K) Averaged ΔF/F Ca²⁺ transients to one pulse (I), 10 pulses (J), and 160 pulses (K) by GECIs. All averaged traces are shown as mean ± s.e.m, indicated by solid lines and shaded area, respectively. Graph inset displays boxplot of peak ΔF/F₀ of each GECI to applied action potentials. Statistics in C–E are the same as in F. (L) Peak ΔF/F₀ of SomaFRCaMPi compared to other red GECIs as a function of pulses. (M) SNR of FRCaMPi compared to other red GECIs as a function of pulses. (N) Peak ΔF/F₀ of SomaFRCaMPi compared to that of FRCaMPi as a function of pulses. Statistics in o for SomaFRCaMPi and FRCaMPi are the same as in J and Fig 1F, respectively. (O) Half the rise time of red GECIs as a function of pulses. (P) Half decay time of red GECIs as a function of pulses (SomaK-GECO1: n = 61 cells from 4 coverslips, 3 independent cultures; SomaXCaMP-R: n = 64 cells from 4 coverslips, 3 independent cultures; SomajRGECO1a: n = 100 cells from 6 coverslips, 4 independent cultures; SomaFRCaMPi: n = 148 cells from 7 coverslips, 4 independent cultures). Graph inset displays boxplot of half rise/decay time of each GECI after 10 action potentials. Box indicates the median and 25–75th percentile range, and the whiskers represent 1.5 times the interquartile range. See S5 Table for detailed statistics and exact p values. The quantitative data presented in this figure can be found in S1 Data. Created in BioRender, https://BioRender.com/e01g235.

https://doi.org/10.1371/journal.pbio.3003048.g003

First, we characterized the subcellular localization of all generated constructs compared to the corresponding non-targeted GECIs in cultured mouse neurons. The ribo-tagged indicators exhibited a much steeper fluorescence decrease along the neurites compared to non-targeted versions. Specifically, ribo-tagged indicators have fluorescence dropped to 36.8% of soma value (we use e^–1-based function for fitting and quantification here, which is often used for characterizing the dropping speed of traits exhibiting an exponential decay) at a distance that was only 5-times shorter than that of the non-targeted version, indicating a strong elimination of expression from neurites by ribo-tagging (Fig 3D-3G). Previous studies have suggested that ribo-tagging could decrease the baseline fluorescence, resulting in dimmer constructs [22,25]. Similarly, we found that ribo-tagging resulted in a decreased baseline fluorescence of all tested red indicators in neuronal culture, with the extent differed for each indicator. Pairwise comparison of corresponding soma-targeted and non-targeted biosensors revealed that soma localization reduced brightness by 7.38 folds for SomajRGECO1a and by 4.85 folds for SomaXCaMP-R while SomaFRCaMPi and SomaK-GECO1 received much lower decreases in baseline (2.24 and 1.47 folds, respectively; S4D Fig). Consequently, SomaFRCaMPi demonstrated a baseline 2.64- and 2.16-fold brighter than that of SomajRGECO1a and SomaXCaMP-R, respectively (Fig 3H).

We then performed functional characterization of SomaFRCaMPi by adapting electrical field stimulation, the same as was applied to non-targeted indicators described above. Similar to the characterization of non-targeted indicators in neurons, SomaFRCaMPi exhibited the highest sensitivity (peak ΔF/F₀ = 1.74 at 1 pulse) among all tested indicators (Fig 3I-3K). Furthermore, we found that the peak ΔF/F₀ of SomaFRCaMPi preserved a degree of monotonical increase to the growing number of pulses similar to FRCaMPi with no obvious saturation detected up to 160 pulses (Fig 3I). The peak response of SomaFRCaMPi was more pronounced relative to non-targeted GECIs in comparison, showing the largest peak ΔF/F₀ in all the numbers of stimuli applied and approximately 2-fold peak difference response to SomajRGECO1a at numbers of pulses in the range from 40 to 160 (Fig 3I and Table 2). Notably, despite the initial bleaching, the SNR of SomaFRCaMPi was still 90–180% higher compared to all other indicators at 1 pulse and stayed at least twice as high as that of SomajRGECO1a (Fig 3M and Table 2) across all assessed pulse numbers, in consonance with the large peak ΔF/F₀ and a higher baseline observed for SomaFRCaMPi (Fig 3H and 3I). In comparison to the non-targeted indicators, SomaFRCaMPi also exhibited 22–112% higher peak ΔF/F₀ relative to that of FRCaMPi between pulse numbers from 1 to 80 (Fig 3N). However, in other soma-localized indicators, we observed either a stronger decrease in baseline brightness or a significant drop in sensitivity such as for SomaK-GECO1 (S4E and S4H Fig). SomaFRCaMPi shares a similar half-rise kinetics to that of SomaXCaMP-R and SomaK-GECO1 (Fig 3O and Table 2) and showed a faster half-rise time than SomajRGECO1a. The decay kinetics of SomaFRCaMPi remained slower but showed a lesser degree of difference to that of other soma-localized indicators compared to the case in non-targeted GECIs (compare results in Figs 2H and 3P, and see Table 2 for the extracted numerical values). Those results demonstrate the high sensitivity and superior capability of SomaFRCaMPi to report calcium signals in neuronal culture. We thus excluded SomaXCaMP-R, SomaK-GECO1a, and their non-targeted versions from further characterization in vivo, referring to their much lower dynamic range and SNR in neuronal culture.

Characterization of SomaFRCaMPi in zebrafish larvae

The transparent nature of zebrafish larvae [26] makes imaging of the entire fish brain possible with spatial precision and cellular resolution for studying the function of neuronal circuits involved in behaviors [27,28]. Similar to the application in mouse studies, RPL10 peptide has been used in zebrafish as a tool for cell type-specific transcriptomic profiling [29,30], but has not yet tethered to reporting neuronal activity in vivo. We, therefore, investigated whether SomaFRCaMPi with ribo-tag motif could also be used in zebrafish for functional imaging, and if so, whether this would have any advantages over non-target sensors in fish. In this regard, we generated transgenic lines that pan-neuronally express jRGECO1a, FRCaMPi, and their soma-targeted variants SomajRGECO1a and SomaFRCaMPi followed by a side-by-side comparison of their performance in zebrafish larvae. However, during establishing stable lines, we observed that SomajRGECO1a in zebrafish was significantly dimmer compared to other indicators in the F1 generation. Specifically, the baseline brightness of SomajRGECO1a was still approximately three times lower than that of SomaFRCaMPi, despite the implementation of a 2.0-fold higher laser power for SomajRGECO1a (S7A Fig). Additionally, the fluorescence response of SomajRGECO1a upon stimulation was limited, despite acquired using a laser power (0.74 mW/mm²) 10 times to that for the other three indicators (0.076 mW/mm²; S7 Fig and S3 Movie). These factors complicated the selection of SomajRGECO1a-expressing zebrafish for breeding and the establishment of stable transgenic lines for sensors’ assessment in F3 generation. Consequently, we decided to exclude SomajRGECO1a from further breeding efforts.

We first characterized the expression of indicators in larvae aged at 6 days post-fertilization (dpf) in F3 generation. The baseline brightness of the three indicators in fish larvae was comparable, with jRGECO1a being 28% and 17% brighter than FRCaMPi and SomaFRCaMPi, respectively (S8A Fig). We observed that in SomaFRCaMPi-expressing larvae, cell bodies could be more easily identified with minimal neuropil fluorescence found in the regions enriched with neuropils (Fig 4A) and in stratified neuropils in cell-body region (e.g., the tectal periventricular cells; S8B Fig). A quantification of soma-to-neuropil fluorescence in optical tectum further shows that fish larvae expressing SomaFRCaMPi exhibit a soma-to-neuropil fluorescence ratio more than twice to that of fish expressing non-targeted indicators, implicating a restriction of SomaFRCaMPi expression to neuronal soma (Fig 4B) in fish. With clearer cell bodies observed in SomaFRCaMPi-expressing larvae, we next investigated whether ribo-tagging facilitates the identification of neuronal soma in fish larvae by automated segmentation of maximum projection images using Cellpose [31]. Larvae expressing SomaFRCaMPi resolved hundreds to thousands more segmented regions of interest (ROIs), including the neuronal cell bodies obscured by neuropil structures such as those in the optical tectum, resulting in at least two times as high as those in larvae expressing FRCaMPi (Fig 4C and 4D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Expression and performance of jRGECO1a, FRCaMPi, and SomaFRCaMPi in reporting neuronal activities in zebrafish larvae.

(A) Representative single plane confocal images showing fore-, mid-, and hindbrain regions of live zebrafish larva expressing jRGECO1a, FRCaMPi, and SomaFRCaMPi in neurons driven by the HuC promoter (n = 10 fish at 6 dpf for each). (B) Representative single-plane confocal image of optical tectum showing the segmentation of somatic and neuropil fluorescence (top) and plotted soma/neuropil ratio of fluorescence (bottom) (jRGECO1a: 2,597 cells from 6 fish; FRCaMPi: n = 1,826 cells from 6 fish; SomaFRCaMPi: n = 2,884 cells from 6 fish). (C) Representative max projection confocal images (top) and segmentation masks (bottom) by Cellpose (n = 14 fish for jRGECO1a, n = 16 fish for FRCaMPi, n = 23 fish for SomaFRCaMPi). (D) Number of ROIs detected per fish larva in Cellpose (n = 14 fish for jRGECO1a, n = 16 fish for FRCaMPi, n = 23 fish for SomaFRCaMPi). (E) Schematics of looming stimulation set up in which black expanding circular shadows projected in front of the larval zebrafish embedded in agarose gel. (F) Representative confocal images showing fluorescence response of optical tectum before and during looming stimulation periods (n = 3 fish each). (G) Representative single trial fluorescence traces of soma and neuropil region of red GECI-expressing neurons during looming stimulation shown as ticks (n = 18 cells from 3 fish each). (H–K) Peak ΔF/F₀ (H), peak SNR at neuronal soma (I), soma-to-neuropil peak ΔF/F₀ ratio (J), and soma-to-neuropil peak SNR ratio (K) of the fluorescence response during looming stimulation (n = 18 cells from 3 fish each). One-way ANOVA test was used, followed by a post hoc Tukey test. The Shapiro–Wilk test was used for the normal distribution test. Scale bar, 50 μm. In dot plots, dots are individual data points; the line denotes the median. Box indicates the median and 25–75th percentile range, and the whiskers represent 1.5 times the interquartile range. The quantitative data presented in this figure can be found in S1 Data. Created in BioRender, https://BioRender.com/q72h354.

https://doi.org/10.1371/journal.pbio.3003048.g004

To understand the dynamic performance of SomaFRCaMPi in zebrafish compared to non-targeted versions, we first examined fluorescence response in fish delivering external electrical stimulation pulse to larvae (S8C Fig), which evoke brain-wide and highly reproducible responses that facilitate benchmarking the sensors side-by-side (1 s duration, 60 V). Neuronal responses from cell bodies in superior medulla oblongata were recorded. We observed the highest peak ΔF/F₀ and SNR attributed to SomaFRCaMPi, followed by FRCaMPi (S8E Fig). All three indicators possess a similar half-rise kinetics, while SomaFRCaMPi and FRCaMPi activities share a slower half decay kinetics compared to that of jRGECO1a (S8F Fig).

We then examined the feasibility of SomaFRCaMPi to report Ca²⁺-induced fluorescence response in larvae, presenting the larvae with looming visual stimuli, which can be considered as more biologically relevant compared to external electrical shock (Fig 4E). Neuronal activity was recorded from optical tectum where neurons were found to be robustly activated due to inputs from retinal ganglion cells (RGC) during looming [32]. Looming induced clear and reproducible responses in neurons expressing SomaFRCaMPi, FRCaMPi, and jRGECO1a (Fig 4F and 4G). Similarly to the findings from external electrical stimulation, we found that tectal cell bodies expressing SomaFRCaMPi demonstrate the highest peak ΔF/F₀ of 325% upon stimulation. More significantly, the peak SNR at cell bodies for SomaFRCaMPi larvae was 70% higher that of jRGECO1a. FRCaMPi larvae further showed 16% higher peak ΔF/F₀ (Fig 4H) and 24% larger SNR (Fig 4I) than those of jRGECO1a. Notably, we observed a much weaker fluorescence response in tectal neuropil in fish expressing SomaFRCaMPi (Fig 4F and 4G) and, accordingly, much higher folds of soma-to-neuropil ratio in peak fluorescence response (Fig 4J) and SNR (Fig 4K) for SomaFRCaMPi fish. The pattern of both half rise and decay kinetics between the three indicators are similar to the observation in electrically evoked neuronal activities (S8G Fig). Together, those results demonstrate the high structural and functional compatibility of ribo-tagged FRCaMPi to zebrafish, highlighting the facilitation of reporting somatic neuronal activity in fish by SomaFRCaMPi and demonstrating the superior sensitivity of SomaFRCaMPi in fish.

Two-photon calcium imaging of neuronal populations in mouse brain

Two-photon (2P) microscopy offers exceptional optical sectioning and deep tissue imaging capabilities, enabling the acquisition of high-contrast, cell-resolution functional images [33,34]. This makes it particularly suitable for in vivo imaging of scattering tissues, such as the mouse brain. For in vivo mouse experiments, we utilized 2P imaging to evaluate the calcium dynamics of FRCaMPi and SomaFRCaMPi. Using AAVs, we expressed jRGECO1a, FRCaMPi or their soma-targeting variants into the primary visual cortex and recorded activities from L2/3 neurons of awake mice in response to drifting gratings presented in eight orientations (Fig 5A). Consistent with previous reports of 2P imaging of ribo-tagged GCaMPs [22,25], we first found that the fluorescence of SomaFRCaMPi was confined to neuronal cell bodies whereas the expression of FRCaMPi and jRGECO1a contributed to a much broader neuropil fluorescence surrounding (Fig 5B). Under identical experimental conditions, the expression of SomajRGECO1a resulted in many fluorescent foci with nuclear filling, and no fluorescence dynamics was detected during the recording of spontaneous activity (n = 3 mice; Fig 5B). Therefore, we did not examine SomajRGECO1a further.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. In vivo neural population imaging in mice V1 cortex.

(A) Schematics of the imaging experiment with drifting grating visual stimulation. (B) Representative single-plane two-photon images of V1 neurons expressing FRCaMPi, SomaFRCaMPi, jRGECO1a, and SomajRGECO1a (FRCaMPi: n = 20 FOVs from 5 mice; SomaFRCaMPi: n = 16 FOVs from 4 mice; jRGECO1a: n = 20 FOVs from 5 mice; SomajRGECO1a: n = 14 FOVs from 5 mice). (C) Trial averaged response amplitudes of single responsive neurons in V1 for different angles of drifting gratings indicated by red arrows. (D) Box plot of fraction of recorded V1 neurons responding to moving grating stimuli (jRGECO1a, n = 12 FOVs from 5 mice; FRCaMPi, n = 12 FOVs from 5 mice; SomaFRCaMPi, n = 9 FOVs from 3 mice) (E). Cumulative distribution function of neural response (ΔF/F₀) to the preferred stimuli for V1 layer 2/3 (L2/3) neurons (jRGECO1a: n = 438 cells from 5 mice; FRCaMPi: n = 609 cells from 5 mice; SomaFRCaMPi: n = 930 cells from 3 mouse). Two-sample Kolmogorov–Smirnov test was used. Central line inside the box, median; box, interquartile range; whiskers, maximum and minimum; outliers, individual data points. One-way ANOVA test was used followed by Tukey’s multiple comparison. (F and G). Same as (D) and (E) but for V1 layer 5 (L5) neurons. n = 23 cells, 4 FOVs from 2 mice (jRGECO1a), 23 cells, 4 FOVs from 2 mice (FRCaMPi), and 15 cells, 4 FOVs from 2 mice (SomaFRCaMPi). Two-sample Kolmogorov–Smirnov test was used. (H) Spontaneous calcium activity of V1 neurons expressing jRGECO1a, FRCaMPi, and SomaFRCaMPi from representative sessions. Calcium traces of either uncorrected (left), neuropil signal (middle), or corrected with surrounding neuropil signal subtraction (right). (I) Correlation coefficients of spontaneous calcium activities between pairs of V1 L2/3 neurons plotted against the centroid distance (jRGECO1a: n = 82,765 neuron pairs from 4 mice; FRCaMPi: n = 310,392 neuron pairs from 4 mice; SomaFRCaMPi: n = 50,355 neuron pairs from 4 mice). (J) Correlation coefficients of spontaneous calcium activities between pairs of V1 L2/3 neurons that are within 100 μm for data shown in (H). (K and L) Same as (H) and (I), but for V1 L5 neuron pairs (jRGECO1a: n = 24,220 neuron pairs from 3 mice; FRCaMPi: n = 114,782 neuron pairs from 3 mice; SomaFRCaMPi: n = 1,721 neuron pairs from 3 mice). The quantitative data presented in this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003048.g005

The basal fluorescence of FRCaMPi in 2P imaging was 70% of that of jRGECO1a, while SomaFRCaMPi had approximately half the basal fluorescence of FRCaMPi (S9A Fig). This level of decreased baseline brightness in SomaFRCaMPi compared to FRCaMPi corresponded to that found in mouse neuronal culture (S4D Fig). Despite the twice lower basal fluorescence of SomaFRCaMPi, we used the same imaging parameter for fair comparison among all indicators to record visual stimulus-evoked activity. We found that the L2/3 neurons expressing all three indicators exhibited specific response selectivity to grating orientation (Fig 5C). For kinetics, we applied a sampling frequency (33 Hz) that was similarly used in GCaMP8 characterization [8] and observed 1.4-fold slower half-decay kinetics in FRCaMPi (0.32 s) compared to jRGECO1a (0.23 s). Strikingly, the half-decay time of SomaFRCaMPi (0.19 s) is similar to jRGECO1a and 41% faster than that of FRCaMPi (S9B ang S9C Fig). Meanwhile, the total number of responsive neurons extracted by SomaFRCaMPi from a single field of view (FOV) in L2/3 (930 neurons/16 FOVs = 58.125 neurons/FOV) was twice that obtained with FRCaMPi (30.45 neurons/FOV) and 2.6 times that achieved with jRGECO1a (21.9 neurons/FOV) from multiple mice. To further validate that the improvement in neuronal segmentation can be achieved in brain regions with different neuroanatomy, we imaged the parabrachial nucleus in the brainstem, where the neuronal population density is much higher than in the cortex (2.73 versus 0.94 neurons per gram of brain tissue) [35]. We conducted imaging across a large volume with a field of view of 0.8 × 1.0 mm², comprising four z-stacks with no overlapping neurons (S10A and S10B Fig). SomaFRCaMPi enabled automatic segmentation of twice as many neuronal ROIs per field of view compared to FRCaMPi (S10C Fig) and reported activity from thousands more neurons when datasets were combined (S10D Fig). These findings demonstrated that SomaFRCaMPi facilitates improved segmentation and allows for high-throughput recording of neuronal activities in two-photon imaging modalities from brain regions with varying neuronal density levels. We then quantified and compared the sensitivity of the three GECIs by adapting two established metrics, the proportion of responsive neurons detected and the cumulative distribution of peak ΔF/F₀ from each neuron [4]. The fraction of responsive cells from recordings of the FRCaMPi- and SomaFRCaMPi-expressing mice was 1.21 to 1.52-fold higher than those of jRGECO1a (Fig 5D). In parallel, the cumulative distribution of peak ΔF/F₀ of FRCaMPi and SomaFRCaMPi was substantially shifted over larger values, indicating FRCaMPi and SomaFRCaMPi being more sensitive than jRGECO1a (Fig 5E). The median peak ΔF/F₀ amplitude for FRCaMPi and SomaFRCaMPi was more than 177% to that of jRGECO1a. Consecutively, the fraction of responsive cells of SomaFRCaMPi is higher than FRCaMPi (median 29 and 23%, respectively; Fig 5D) and we also observed a slightly right-shifted distribution of peak ΔF/F₀ of SomaFRCaMPi relative to that of FRCaMPi (median 1.75 and 1.60, respectively; Fig 5E). The observed higher sensitivity level for SomaFRCaMPi during 2P-imaging remained consistent with that observed in neuronal cultures and zebrafish. In addition, FRCaMPi and SomaFRCaMPi also showed higher-value distribution of SNR in L2/3 neurons compared to jRGECO1a (S9D Fig). Notably, the distribution of SNR values for SomaFRCaMPi (median 7.59) is only 7% less to that of FRCaMPi (median 8.17, p = 0.08) despite of a lowered baseline brightness by half.

To further assess the dynamics and sensitivity of our red indicators during deep tissue imaging in vivo, we recorded visual stimulus-evoked activities from L5 neurons (Fig 5F and 5G). We observed a very similar pattern in the peak ΔF/F₀ distribution for the three indicators to those seen in L2/3 (Fig 5G). The median peak ΔF/F₀ amplitudes for FRCaMPi and SomaFRCaMPi remained 43–68% higher than that of jRGECO1a. Notably, the fraction of responsive cells in recordings from mice expressing SomaFRCaMPi (median 19%) was almost three times that of FRCaMPi and jRGECO1a (medians 6.6% and 6.7%, respectively; Fig 5F). Meanwhile, in L5 neurons, we found that the SNR of jRGECO1a, FRCaMPi, and SomaFRCaMPi were converging. Nonetheless, FRCaMPi exhibited the highest SNR, followed by SomaFRCaMPi and jRGECO1a (medians 6.57, 6.43, and 6.35, respectively; S9E Fig), similar to observations in L2/3. These results demonstrate that the indicators are capable of reporting neuronal activities from deeper cortical layers under two-photon (2P) imaging, albeit with lower sensitivity compared to 2P imaging in L2/3 neurons.

Given the outstanding peak ΔF/F₀ and peak SNR of SomaFRCaMPi found in visual cortex, we extended our investigation to its long-term expression performance, specifically analyzing Ca²⁺ spikes from spontaneous activity recordings. In response to prevalent concerns in the field regarding nuclear filling [2] and reduced indicator responses during prolonged expression [2,13], we sought to address these issues with SomaFRCaMPi. Notably, neither SomaFRCaMPi nor jRGECO1a exhibited significant nuclear filling after 4 months of expression in V1 L2/3, with only minimal nuclear filling observed (S9H Fig). At 6 weeks of expression, SomaFRCaMPi demonstrated 46% higher peak ΔF/F₀ compared to jRGECO1a and a similar SNR in spontaneous spikes (S9I-S9K Fig). In the same mice after 4 months of expression, both indicators demonstrated a drop of sensitivity with around a 30% drop of peak ΔF/F₀ (S9I and S9J Fig) compared to that of 6 weeks. However, the decline in peak SNR for SomaFRCaMPi was only 4.5% (median: 8.87 at 6 weeks versus 8.48 at 4 months; p = 8.9e–27), approximately half the reduction observed in jRGECO1a (−8.7%, median: 8.87 at 6 weeks versus 8.48 at 4 months; p = 4.5e–38) (S9K Fig). Consequently, even after 4 months of expression, SomaFRCaMPi maintained 38% higher peak ΔF/F₀ and a slightly higher SNR compared to jRGECO1a. These results suggest that SomaFRCaMPi can reliably report neuronal activity over extended periods with less degradation in SNR, highlighting its potential for stable calcium imaging in the long term.

In conjunction with findings from the study on ribo-tagged GCaMPs [22,25], we found highly synchronized neuropil fluorescence in recordings of FRCaMPi- and jRGECO1a-expressing mice but not in that for SomaFRCaMPi (Fig 5H). To understand the extent to which SomaFRCaMPi reduces neuropil contamination in 2P imaging, we analyzed distance-wise correlation of activities between neurons from the recording of spontaneous events in V1. Using raw imaging datasets for analysis, we initially observed a correlation level of around 0.45 in activity between neurons expressing FRCaMPi (mean 0.43) and jRGECO1a (mean 0.46), which decreased to 0.13 (same for FRCaMPi and jRGECO1a) with increasing distance to 600 µm. This correlation was 2- to 7-fold lower in neurons expressing SomaFRCaMPi compared to both FRCaMPi and jRGECO1a (Fig 5I). We then implemented a standard algorithmic approach to subtract neuropil signals from neuronal traces as previously suggested [22,36]. This adjustment resulted in a more than a 3-fold reduction in signal correlation between neurons expressing either jRGECO1a or FRCaMPi. We further quantified correlation coefficient of neuron pairs within 100 µm (Fig 5J). Post-subtraction, the signal correlation levels of jRGECO1a (mean 0.099) and FRCaMPi (mean 0.093) were slightly lower than those of uncorrected SomaFRCaMPi (mean 0.13) but remained within a very similar degree of low correlation. Neurons expressing SomaFRCaMPi also exhibited a slight drop in correlation following signal correction. The mean correlation coefficient of corrected signals was still more than 50% lower than those of jRGECO1a and FRCaMPi after correction. Moreover, the pattern of comparison in correlation was found similar in neuronal activity from L5 (Fig 5K and 5L). Therefore, SomaFRCaMPi substantially reduces the neuropil-induced correlation between cells in 2P imaging experiments with superior sensitivity levels, demonstrating its capacity as a competent soma-localized GECI in red for 2P imaging.

To further evaluate the sensitivity of SomaFRCaMPi compared to top-performing green sensors in the field, we co-expressed RiboL1-GCaMP8s and SomaFRCaMPi in the nucleus of the solitary tract (NTS) of the brainstem. RiboL1-GCaMP8s was chosen as the most sensitive soma-GECIs reported to date as it has been found to exhibit a similar fluorescence response in vivo to non-targeted jGCaMP8s, one of the most sensitive green calcium indicators to date (see Fig S5 in the reference [25]). Neuronal activity was examined using dual-color 2P calcium imaging in response to multiple internal organ stimuli (Fig 6A and 6B). Neurons co-expressing green and red calcium probes showed selective and dose-dependent functional responses to increasing gastric distension, characterized by a growing number of responding neurons and elevated fluorescence response amplitudes (Fig 6C and 6D), validating the functionality of the NTS calcium imaging system. Next, we compared the peak responses from NTS neurons that co-expressed both indicators. We found that the overall peak ΔF/F level of SomaFRCaMPi was approximately 33% less than that of RiboL1-GCaMP8s, as demonstrated by least-squares fitting (Fig 6E). However, we observed an overall lower noise in the red channel (Fig 6D). Specifically, examination of the baseline noise level in SomaFRCaMPi showed it to be approximately 32% lower than that of RiboL1-GCaMP8s (Fig 6F), resulting in an almost identical Z-score between the two probes (Fig 6G). These results indicate that SomaFRCaMPi exhibits lower imaging noise and comparable sensitivity to a state-of-the-art green soma-GECI.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Dual-color in vivo calcium imaging of NTS tuning during gastric distension.

(A) Schematics depicting NTS two-photon calcium imaging setup during gastric distension. (B) Representative two-photon NTS images showing the co-expressing of RiboL1-jGCaMP8s and SomaFRCaMPi with the shared ROIs identified from Cellpose segmentation. Scale bar, 50 μm. (C and D) Heat maps depicting time-resolved responses to a series of gastric distensions from NTS neurons co-expressing Ribo-GCaMP8s (C) and SomaFRCaMPi (D) from the image in (B) (329 neurons). (E) Scatter plot showing the peak ΔF/F of green (Ribo-GCaMP8s) and red (SomaFRCaMPi) calcium responses from NTS neurons co-expressing both indicators during gastric distension. (F–G) Similar to (E), but for noise evaluation (F) and peak Z-score analysis (G). n = 1,580 neurons from 2 mice. The grey diagonal dash represents a slope equal to 1, and the black dashed line denotes the linear fit of the data. The quantitative data presented in this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003048.g006

Wide-field primary neocortical imaging of awake mice using SomaFRCaMPi

Provided with the faithful neuropil elimination and the enhanced sensitivity of SomaFRCaMPi under 2P microscopy, we next investigated the benefits of using SomaFRCaMPi in awake behaving mice using one-photon widefield microscopy, which is a more accessible and less complex system compared to 2P setups. Challenges in neuron detection and deteriorated somatic signals [37] were often observed in one-photon widefield or micro-endoscopic imaging of the mouse brain due to tissue scattering and the strong out-of-focus background fluorescence [38,39]. Restricting the expression of indicators to neuronal somas, in turn, may help to directly enhance the clarity of somatic calcium signaling and improve signal accuracy in wide-field imaging [23] (S1 Text). So far, optical recording with single-cell resolution using either one-photon micro-endoscopic or widefield imaging modality was dominated by green GECIs [40–45], while few studies have reported on the extraction of single-neuron activity using red GECIs using wide-field imaging [46]. Meanwhile, those studies were often facilitated by achieving optimal neuronal labeling density [40–42,44–46] and/or special optics such as holographic illumination patterns [43,46] to reduce background and improve cell signal detection (S1 Text). Provided with the superior sensitivity of our red probes in zebrafish and mice, we, therefore, investigated whether SomaFRCaMPi could report population neuronal activity with single-cell resolution and improved accuracy in mouse cortex under a widefield imaging setup.

For this experiment, we pan-cortically expressed either SomaFRCaMPi or FRCaMPi in the mouse cortex (S11A and S11B Fig) and installed a polymer cranial window that spans the dorsal cortex of mice [47]. We then head-fixed the mouse and imaged neuronal activity using a wide-field stereoscope [48] (Fig 7A). By tuning the stereoscope to the highest spatial resolution achievable for our setup (zoom body set to 6.3×), we achieved a millimeter-sized FOV for populational imaging of neurons. A green indicator, GRAB-5HT3.0, was co-expressed to help assure pan-neuronal expression pattern and enhance visualization of the vasculature morphology as a reference for locating specific brain regions at higher magnifications (Fig 7B and S2 Movie). We chose to image the primary somatosensory (S1) and visual cortex (V1), which are considered to be two of the densest cortical regions, with 110,000 ± 1,000 and 164,000 ± 50,000 neurons per mm³, respectively [49]. Thus, obtaining quality somatic signals would be, in principle, more challenging in those regions compared to others due to a stronger labeling density in the background. To ensure a fair comparison between non-targeted and soma-localized versions of FRCaMPi, and to evaluate the contribution of neuropil contamination to distance-dependent neuronal activity correlation, neuronal activity was recorded only while the mouse was resting on the treadmill. This approach was taken to prevent additional input from correlated neuronal activities in the somatosensory and visual cortex that could arise due to locomotion, as suggested by previous studies [50–52]. As a result, we successfully detected Ca²⁺ spikes from neuronal soma in both S1 and V1 of brains expressing FRCaMPi and SomaFRCaMPi (Fig 7C and 7D). We identified many more spiking neurons by simple visual assessment of the raw datasets (S5 Movie). In parallel, we observed a consistently high labeling density of 80–90% of cortical neurons estimated at the imaging site using post hoc histological analysis (S11C Fig). Further analysis of the number of ROIs segmented by CNMF-E [38] confirmed that the number of ROIs identified per FOV for SomaFRCaMPi in S1 and V1 was 3.3 and 2.9 times larger than that for FRCaMPi, respectively (Fig 7E). Consistent with observation in mouse neuronal culture and in mouse in vivo under 2P, the baseline fluorescence of SomaFRCaMPi neurons in mouse cortex was about 23% and 57% dimmer than that of FRCaMPi ones under the same imaging parameters in S1 and V1, respectively (Fig 7F). The profiling of neuronal spikes showed similar peak ΔF/F₀ between FRCaMPi- and SomaFRCaMPi-expressing mice in S1 and 2-fold peak ΔF/F₀ for SomaFRCaMPi spikes in V1 (Fig 7G), consistent with the observation in 2P mouse imaging. However, the peak SNR of SomaFRCaMPi spikes from both brain regions were higher than FRCaMPi ones, with 10% and 40% higher peak SNR observed for SomaFRCaMPi spikes in S1 and V1, respectively (Fig 7H). This indicates that the expression of SomaFRCaMPi reduces the overall background noise to a level that achieves a higher SNR despite owning a lower brightness. The decay kinetics remained very similar between SomaFRCaMPi and FRCaMPi obtained from individual spike waveforms (Fig 7I). Meanwhile, SomaFRCaMPi neurons also reported approximately 50% more calcium events than those of FRCaMPi in FOV from both brain regions (Fig 7J).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Recording of FRCaMPi and SomaFRCaMPi dynamics in

S1 and V1 using wide-field microscopy. (A) Schematics of the wide-field one photon imaging experiment in awake resting mice. (B) Representative macroscopic images showing pan-cortical expression of gGRAB_5-HT3.0 excited by 470 nm (left) and SomaFRCaMPi excited by 565 nm (right) (n = 4 mice). (C) Representative images (the standard deviation of images from one imaging session) for the S1 or V1 region in either SomaFRCaMPi- or FRCaMPi-expressing mice (top) and the corresponding correlation image by CNMF-E (bottom) (n = 4 mice each). (D) Top, representative calcium fluorescence traces with calcium events identified for FRCaMPi and SomaFRCaMPi from the experiments shown in (C) (n = 249 neurons from 5 FRCaMPi mice and n = 1,081 neurons from 5 SomaFRCaMPi mice). Bottom, aligned calcium events displayed as mean and S.D. (E) Number of ROIs detected per FOV by CNMF-E (S1: n = 4 FOVs from 4 FRCaMPi mice, n = 6 FOVs from SomaFRCaMPi 4 mice; V1: n = 3 FOVs from 3 FRCaMPi mice and n = 5 FOVs from 5 SomaFRCaMPi mice). In dot plots, line denotes median. (F–H) Baseline brightness (F), peak ΔF/F₀ (G), and peak SNR (H) of SomaFRCaMPi compared to FRCaMPi in S1 and V1 (S1: n = 161 neurons from 4 FRCaMPi mice, n = 1,143 neurons from 4 SomaFRCaMPi mice. V1: n = 563 neurons from 3 FRCaMPi mice, n = 1,653 neurons from 5 SomaFRCaMPi mice). (I) The decay time of SomaFRCaMPi compared to FRCaMPi in S1 and V1 (S1: n = 56 neurons from 4 FRCaMPi mice, n = 595 neurons from 4 SomaFRCaMPi mice. V1: n = 343 neurons from 3 FRCaMPi mice, n = 1,247 neurons from 5 SomaFRCaMPi mice). (J) Spike rate of SomaFRCaMPi compared to FRCaMPi in S1 and V1 (S1: n = 37 neurons from 4 FRCaMPi mice, n = 364 neurons from 4 SomaFRCaMPi mice. V1: n = 212 neurons from 3 FRCaMPi mice, n = 717 neurons from 5 SomaFRCaMPi mice). (K) Pearson correlation coefficient of calcium dynamics plotted against distance for different cell pairs from S1 region of mice expressing FRCaMPi (n = 3,410 cell pairs from 5 mice) or SomaFRCaMPi (n = 136,576 cell pairs from 4 mice). (L) As in K but for V1 (n = 99,451 cell pairs from 3 FRCaMPi mice and n = 296,784 cell pairs from 5 SomaFRCaMPi mice). (M) Pearson correlation coefficient of calcium dynamics between pairs of neurons within 800 μm. Statistics are the same as in K and L. Mann–Whitney U test. Scale bar, 50 μm. Bar chart is shown as mean ± s.e.m. Box indicates the median and 25–75th percentile range, and the whiskers represent 1.5 times the interquartile range. Spike rate and Pearson correlation coefficient are plotted as bars to aid the demonstration of data distribution. See S6 Table for more details. The quantitative data presented in this figure can be found in S1 Data. Created in BioRender,.https://BioRender.com/t42r557.

https://doi.org/10.1371/journal.pbio.3003048.g007

We next measured the distance-wise correlation of neuronal activity from neuron pairs within the FOV of SomaFRCaMPi versus FRCaMPi mice. To achieve this, we analyzed the Pearson correlation coefficient (PCC) of raw data traces extracted from ROIs as previously suggested [22,23]. In S1 and V1 of mice expressing either of the indicators, we observed correlation coefficients of 0.73–0.99 between cell pairs, with the drop of correlation found for cell pairs at increasing distances over the range of 0–800 µm. The magnitude of correlation was much higher than that observed in 2P imaging experiments, and the correlation difference between FRCaMPi and SomaFRCaMPi was also much higher. As suggested by Chen and colleagues [22], optical scattering and out-of-focus fluorescence likely contribute to the overall blurriness observed in wide-field imaging and, thus, dominate high correlation by noise. Despite presenting a much denser spiking population and a higher spiking rate on top (Fig 7C, 7E and 7J), SomaFRCaMPi still exhibited a lower correlation coefficient across the broad range of distances with a steeper decrease of coefficient between neuronal pairs, ending up at a level approximately 20% to 30% less than FRCaMPi at 800 µm (Fig 7K-M). We further analyzed the mean correlation coefficient for distance ranges: 0–200, 200–400, and 400–800 µm and found correlations at different distance ranges were all significantly lower in SomaFRCaMPi-expressing mice compared to FRCaMPi ones (S11D Fig). To summarize, in both S1 and V1 regions, SomaFRCaMPi reduced contamination to a less signal correlation, enabled detection of more neurons with a higher number of neuronal spikes, and increased SNR using the widefield setup. These results collectively underscore the efficacy of SomaFRCaMPi for one-photon widefield imaging, demonstrating its advanced ability as a red soma-localized indicator to report more somatic signals with higher precision in this noisy imaging modality.

Molecular cloning

The FRCaMPi gene was generated using PCR with overlapping fragments [69] using Fw-LSSmOrange-BglII, RSP-EcoRI-r, CaMSp-M13, CaMSp-M13-r, and RSPi-EcoRI-r primers listed in S2 Table. The FRCaMPi gene was cloned into the pBAD/HisB-TorA-mTagBFP plasmid at BglII/EcoRI restriction sites using the Fw-LSSmOrange-BglII/RSPi-EcoRI-r primers to express FRCaMPi variant in the periplasm of bacterial cells. To construct the pAAV-CAG-NES-FRCaMPi plasmid, the FRCaMPi gene was PCR-amplified as the BglII-EcoRI fragment using the Fw-LSSmOrange-BglII/RSPi-EcoRI-r primers and swapped with the mCherry gene in the pAAV-CAG-NES-mCherry vector.

Synthetic DNA oligonucleotides used for cloning were synthesized by Tsingke Biotechnology. PrimeStar Max master mix (Takara) or 2×Hieff PCR Master Mix (Yeasen) was used for high-fidelity polymerase chain reaction (PCR) amplifications. Restriction endonucleases were purchased from New England BioLabs and used according to the manufacturer′s protocols. DNA ligations were performed using T4 DNA ligase (New England BioLabs) or the NovoRec plus one step cloning kit (Novoprotein). The ligation products were chemically transformed into the TOP10 E. coli strain (Biomed) and cultured according to the standard protocols. Sequencing of bacterial colonies and purified plasmids was performed using Sanger sequencing (Zhejiang Youkang Biological Technology).

Small-scale isolation of plasmid DNA was performed with commercially available Mini-Prep kits (Qiagen or Tiangen). The genes of red GECIs were de novo synthesized by Tsingke Biotechnology or Shanghai Generay Biotech based on the cDNA sequences reported in the original publications. Oligonucleotides for cloning were purchased from Tsingke Biotechnology or Zhejiang Youkang Biological Technology (S2 Table). To clone pAAV-CAG-NES-FRCaMPi-P2A-mGreenLantern, pAAV-CAG-XCaMP-R-P2A-mGreenLantern and pAAV-CAG-NES-jRGECO1a-P2A-mGreenLantern the complementary DNA of mGreenLantern was first PCR amplified with SpeI/EcoRI flanking sites and swapped with the EGFP gene in the pAAV-CAG-miRFP670–2-P2A-GFP plasmid (WeKwikGene plasmid #0000007). Then, the cDNA of NES-FRCaMPi, NES-jRGECO1a or XCaMP-R was PCR amplified with KpnI/ AgeI flanking sites and swapped with the miRFP670−2 gene in the pAAV-CAG-miRFP670–2-P2A-GFP. To clone pAAV-CAG-NES-K-GECO1-P2A-mGreenLantern and the cDNA of K-GECO1 was PCR amplified with BglII/ AgeI flanking site and swapped with the FRCaMPi gene in the pAAV-CAG-NES-FRCaMPi-P2A-mGreenLantern plasmid. We fused nuclear export sequence (NES) to the N-termini of red GECIs except for XCAMP-R where the F2A sequence appears to be an NES. To clone pAAV-Syn-NES-Red-GECI, the cDNA of NES-jRGECO1a, NES-KGECO1 or XCaMP-R was PCR amplified with BamHI/EcoRI flanking site and swapped with the CoChR-GFP gene in the pAAV-Syn–CoChR–GFP (Addgene plasmid #59,070). From the peptide screening result, we choose to use ribo-tag to engineering a soma-localized version of FRCaMPi strictly localized to soma. Originally inspired by the SomaGCaMP paper [23], we aimed to facilitate protein folding by adding a flexible linker (GGSGGSGGT)3 from SomaGCaMP6f2 to the C-terminus of FRCaMPi. SomaFRCaMPi was then generated by linking RPL10. To clone pAAV-Syn-NES-FRCaMPi-RPL10, the cDNA fragment of NES-FRCaMPi and RPL10 was PCR amplified with overlapping fragments using Syn-NES-F, FRCaMPi-R, Linker27-F and WPRE- Linker27-R primers in S2 Table. To clone pAAV-Syn-NES-K-GECO1-RPL10, pAAV-Syn-XCaMP-R-RPL10 and pAAV-Syn-NES-jRGECO1a-RPL10, the cDNA fragment of NES-K-GECO1, XCaMP-R or NES-jRGECO1a and RPL10 was PCR amplified with overlapping fragments using Syn-NES-F, Syn-XCaMPR-F, Syn-NES-jRGECO1a-F, Linker27-XCaMPR-R, Linker27-jRGECO1a-R, Linker27-K-GECO1-R and WPRE-Linker27-R primers in S2 Table.

Protein purification and characterization

Proteins were expressed and purified as described in [70]. All experiments for protein characterization were performed at ambient temperature of 21–23°C.

The extinction coefficient values for the FRCaMPi indicator in a Ca²⁺-saturated state or apo-state were calculated in 30 mM HEPES buffer, pH 7.2 (adjusted by KOH), 100 mM KCl (buffer A), supplemented with 5 mM CaCl₂ or 10 mM EDTA, respectively, using the alkaline denaturation method and assuming that the mCherry-like red chromophore had an extinction coefficient of 38,000 M^–1 cm^–1 at 455 nm in 1 M NaOH [14]. The absorption spectra were registered using a NanoDrop 2000c Spectrophotometer (Thermo Scientific, DE, USA)

To determine the quantum yield of FRCaMPi excited at 540 nm in a Ca²⁺ saturated state or apo-state, the integrated fluorescence values (in the range of 550–750 nm) were measured in buffer A supplemented with 5 mM CaCl₂ or 10 mM EDTA, respectively, and compared with the same values for the equally absorbing at 540 nm mCherry protein (quantum yield of 0.22 (ref. [14])). To determine the quantum yield of the purified FRCaMPi protein excited at 400 nm in the apo-state, the integrated fluorescence values (in the range of 410–750 nm) were measured in buffer A supplemented with 10 mM EDTA and compared with the same values for the mTagBFP2 protein equally absorbing at 400 nm (quantum yield of 0.64 (ref. [71])). The fluorescence spectra were acquired using a CM2,203 spectrofluorometer (Solar, Minsk, Belarus).

pH titrations for FRCaMPi (50 nM final concentration) in a Ca²⁺-saturated state or apo-state were performed in buffer: 30 mM citric acid, 30 mM borax, and 30 mM NaCl with pH values ranging from 3.0 to 11.0 (adjusted by HCl or NaOH), supplemented with 0.1 mM CaCl₂ or 0.1 mM EDTA, respectively, followed by incubation for 20 min, as described in [70]. Red fluorescence (Ex: 525 nm/Em: 580–640 nm) was registered using a 96-well ModulusTM II Microplate Reader (Turner Biosystems, Sunnyvale, CA, USA).

To determine the equilibrium calcium K_d, the fluorescence of the FRCaMPi protein (50 nM final concentration) was measured in a mixture of two stock buffers: 30 mM 3-(N-morpholino) propanesulfonic acid (MOPS), pH 7.2 (adjusted by KOH), 100 mM KCl supplemented with 10 mM EGTA or 10 mM Ca-EGTA (buffer C), as described previously [70]. Buffer B and C were mixed to 0, 0.25, 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 9, 9.5, 9.75, and 9, 875, and 10 mM Ca_total concentrations to obtain 2.1, 7.9, 16.3, 35.4, 79.2, 135.5, 208.4, 312.6, 468.9, 731.5, 1254.6, 2817.6, 5856.0, 11858.0, 21923.7, and 81,276 nM Ca_free concentrations (the free calcium concentrations were corrected by 1.084-fold according to the GCaMP6s indicator K_d at 144 nM (ref. [2])). To determine the equilibrium calcium K_d in the presence of 1 mM Mg²⁺ ions, the fluorescence of the FRCaMP protein (50 nM final concentration) was measured in a mixture of buffer B and C supplemented with 1 mM MgCl₂. The dilutions of the B and C buffers solutions were the same, and free calcium concentrations were considered as described above in the absence of Mg²⁺ ions. Red fluorescence (Ex: 525 nm/Em: 580–640 nm) was registered using a 96-well ModulusTM II Microplate Reader (Turner Biosystems, Sunnyvale, CA, USA).

Photobleaching experiments were performed with suspensions of purified proteins in mineral oil, as previously described [72]. Briefly, the photobleaching were measured using purified proteins dialyzed in buffer A supplemented with either 10 mM EDTA or 5 mM CaCl₂ at a 50 µM concentration in aqueous microdroplets in mineral oil using a Zeiss Axio Imager Z2 microscope (Zeiss, Germany) equipped with a 120 W mercury short-arc lamp (LEJ, Germany), a 63 × 1.4 NA oil immersion objective lens (PlanApo, Zeiss, Germany), a 550/25 BP excitation filter, a FT 570 beam splitter, and 605/70 BP emission filters for red channel and 470/40 BP excitation filter, a FT 495 beam splitter, and 525/50 BP emission filters for green channel. Light power densities (red channel 9 mW/cm² and green channel 16 mW/cm²) were measured at the rear focal plane of the objective lens using a PM100D power meter (ThorLabs, Germany) equipped with an S120VS sensor (ThorLabs, Germany). No corrections were applied to the experimental data.

Characterization in HeLa cell culture

HeLa (ATCC CCL-2) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin, and seeded in a 24-well glass-bottom plate (P24-0-N Cellvis) after Matrigel (356,235, BD Biosciences) coating. Cells at 80–90% confluency were transfected with 500 ng of DNA (pAAV-CAG-NES-RedGECI-P2A-mGreenLantern) using Hieff Trans Liposomal Transfection Reagent (Yeasen Biotechnology, 40802ES02) and cultured for. Before imaging, the medium was replaced with 0.5 mL Ca²⁺ and Mg²⁺ free Hanks balanced salt solution (HBSS). To quantify the brightness level of red GECIs in HeLa cells, images were acquired in both TRITC (excitation 561 nm at 1.53 mW/mm²) and FITC (excitation 488 nm at 2.5 mW/mm²) using a ×20 NA 0.75 objective (Nikon). The mean fluorescence intensity in the TRITC and FITC channels for ROIs were extracted and the red-to-green fluorescence ratios were calculated after background subtraction for each channel. To obtain photobleaching curve of red GECIs in HeLa cells, images were acquired using the TRITC channel at a laser power of 7.4 mW/mm².

To record histamine-induced Ca²⁺ oscillation of red GECIs, 5 mM histamine and 1 M CaCl₂ was added to create a final concentration of 50 μM and 6 mM of CaCl₂. After 10 min of imaging, 1 μl 2.5 mM ionomycin was added to make a final 5 μM ionomycin solution and cells were imaged for another 2–6 min to capture ionomycin-induced calcium influx. Image was acquired using TRITC (1.53 mW/mm²) with a ×20 NA 0.75 objective (Nikon). All images were acquired using Nikon Eclipse Ti inverted microscope, SPECTRA X light engine (Lumencor), and OrcaFlash4.0v2 camera (Hamamatsu), controlled by NIS-Elements AR software. The maximal dynamic response, ΔF_max/F_min, was defined as maximal fluorescence change during the period of ionomycin-induced calcium influx, where F_min is the minimal fluorescence value of a single frame before fluorescence increase and ΔF_max = F_max – F_min, where F_max is the maximal fluorescence value of a single frame during ionomycin-induced response. Data were analyzed offline using NIS Elements Advance Research software (versions 5.21.00 and 5.30.00), Excel (Microsoft), and GraphPad Prism 9.

Characterization in primary neuronal culture

In S2 Fig, dissociated neuronal cultures were isolated from the C57BL/6 mice at postnatal days 0–1 and were grown on a 24-well cell imaging black plate with a glass bottom coated with 0.001% polylisine; then, the tissue cultures were treated (Eppendorf, Germany) in Neurobasal Medium A (GIBCO, UK) supplemented with 2% B27 Supplement (GIBCO, UK), 0.5 mM glutamine (GIBCO, UK), 50 U/ml penicillin, and 50 µg/ml streptomycin (GIBCO, UK). The rAAV particles for neuronal transduction were purified from 10 to 150 cm dishes, as described in the original paper [73]. On the 4th day in vitro, neuronal cultures were transduced with a mixture of rAAV viral particles carrying AAV-CAG-NES-FRCaMPi and AAV-CAG-NES-NCaMP7. The cells were imaged using an Andor XDi Technology Revolution multi-point confocal system on DIV 16 (spontaneous activity at 37 °C, 5% carbon dioxide) and DIV 17–18 (electrical field stimulation at 21–23 ^oC).

Stimulation of neuronal cultures was performed using a self-built electrical system described earlier [73]. In this step, 300 voltage pulses of a 1 ms duration (0.5 ms negative phase, 0.5 ms interphase, and 0.5 ms positive phase) at an 87 Hz frequency with an amplitude of ±70 V were applied to the neuronal cultures in 24-well plates through iridium electrodes with a 5 mm gap. 10 µM cyanquixaline (6-cyano-7-nitroquinoxaline-2,3-dione) (CNQX) and 100 µM (2R)-amino-5-phosphonovaleric acid (APV or AP5) were added before stimulation to block spontaneous neuronal activity. To quantify the fluorescence intensity, the background noise determined from the adjacent cell-free area was subtracted from mean fluorescence intensity value for the cytosolic sub-region of the cell of the similar area. The rise and decay half-times were calculated as time difference between time point corresponding to the calcium spike maximum and time points at half-maximum at the left and right edges of the spike, respectively.

For functional profiling of cytosolic-expressing and soma-localized red GECIs (Figs 1, 2 and S4), cultured neurons were transduced with viruses at 5 DIV. Constructs used for transduction included: AAV2/9-Syn-K-GECO1-WPRE, AAV2/9-Syn-SomaK-GECO1-WPRE, AAV2/9-Syn-XCaMP-R-WPRE, AAV2/9-Syn-SomaXCaMP-R-WPRE, AAV2/9-Syn-jRGECO1a-WPRE, AAV2/9-Syn-SomaFRCaMPi-WPRE (Shanghai Sunbio Medical Biotechnology). All data presented were obtained from transduced neurons, except for those in Figs 1C, 2F, S4A and S4B, neurons were co-transfected with pAAV-Syn-Red-GECI-WPRE and pAAV-CAG-emiRFP670-P2A-GFP (WeKwikGene #0000006) by the calcium phosphate transfection method as used previously. To compare different red sensors side-by-side in neuronal culture, the volume of each AAV was diluted and adjusted to approximately 1 × 10¹³ vg/ml of AAV particles using titer number provided by the manufacture. All the viruses used were not undergone number of free-thaw cycles more than twice. To obtain a high density of transduced neurons, approximately 1 μl of diluted viruses were applied per well of the 24-well plate. On DIV 11–13, neurons were transferred to a stimulation chamber (Warner Instruments, RC-49MFSH), stimulated and imaged in extracellular buffer solution containing 150 mM NaCl, 5 mM KCl, 10 mM glucose, 10 mM HEPES, pH 7.4, 2 mM CaCl₂, 1 mM MgCl₂. Electrical pulses (83 Hz) of 40–50 V, 1-ms width were evoked by field stimulation with the stimulation chamber and a voltage amplifier (ATA2041, Aigtek). A total number of eight trains containing increased numbers of electrical stimuli (1, 2, 3, 5, 10, 20, 40, 80, 160) were evoked for each image of neuronal trace. Images were acquired at 20 Hz using a Nikon Eclipse Ti inverted microscope equipped with a ×20,0.75 NA objective (Nikon), TRITC (Ex: 561 nm, 1.53 mW/mm²) from SPECTRA X light engine (Lumencor), and OrcaFlash4.0v2 camera (Hamamatsu), controlled by NIS-Elements AR software.

For electrical stimulation experiments in Figs 1, 2 and S4, neurons from all wells were recorded for 5 min before stimulation to remove any potential bleaching effect. ROIs corresponding to the soma of responsive neurons was manually selected from each train of a field stimulation trace. In neuronal culture, peak ΔF/F₀ is defined as maximal fluorescence change of a spike train over F₀, where F₀ is the average of 1 s fluorescence baseline right before the stimulus. SNR is calculated as the maximal fluorescence change over the standard deviation of 1 s fluorescence baseline before the stimulus, reported as relative ratio. Neuronal ROIs for calculating peak ΔF/F₀ and SNR were selected independently for different analytical experiment. In particular, ROIs for peak ΔF/F₀ were selected from individual neurons with a single background whereas ROIs for peak SNR were selected from neurons with each paired to an adjacent background. To correct variation of fluorescence response observed in independent transduced cultures, the mean of peak ΔF/F₀ and SNR between those cultures were normalized. Fluorescence trains from ROI was excluded from dataset when fluorescence spikes failed recovery to baseline or were not evoked by electrical stimulation.

To obtain the kinetics of the indicator, epochs from ΔF/F₀ traces containing rise and decay stage of the trace were first manually extracted and fitted using a single exponential function using MATLAB to calculate their half kinetics. Half-rise (t_{1/2, rise}) and decay (t_{1/2, decay}) time were defined as the time to the half of peak ΔF/F₀ in the rise phase and the time from peak ΔF/F₀ to half of peak ΔF/F₀ in the decay phases, respectively, calculated by interpolation. Neuronal ROI was excluded when fluorescence spikes failed recovery to baseline or has a fitting coefficient less than 0.8. For neuronal brightness of red GECIs, only neuronal soma was used to extract mean fluorescence intensity of the ROI. For the bleaching analysis, neurons that exhibited fluorescence spikes and did not return to baseline levels by the end of the imaging session were excluded. Data were analyzed offline using NIS Elements Advance Research software (versions 5.21.00 and 5.30.00), MATLAB (version 2022a), Excel (Microsoft), and GraphPad Prism 9.

Animals

All animal maintenance and experimental procedures were conducted according to institutional and ethical guidelines for animal welfare, and all animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Westlake University, Hangzhou (animal protocol #KP-19–044) or Tsinghua University (animal protocol #15-GZC3.G23-1), Beijing, China, or the Institutional Animal Care and Use Committee at The Scripps Research Institute or the National Research Center “Kurchatov Institute” Committee on Animal Care (NG-1/109PR of 13 February 2020) and were done in accordance with the Russian Federation Order Requirements N 267 МЗ and the National Institutes of Health Guide for the Care and Use of Laboratory Animals. For all experiments involving mice throughout, the C57BL/6J strain (supplied by the animal facility of Westlake University or the laboratory animal research center at Tsinghua University or the Scripps Research Institute or the National Research Center “Kurchatov Institute”) was used regardless of sex.

Immunohistochemistry

Mice were deeply anesthetized with 1% pentobarbital sodium (10 μl/g) and perfused transcardially with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in PBS. Brains were kept in PBS containing 4% PFA overnight. Coronal slices (50 μm thick) were cut using a vibratome (VT1200 S, Leica Microsystems) and then incubated in a staining solution composed of 3% (v/v) molecular Nissl stain (Invitrogen, N21483) and 0.1% (v/v) DAPI in PBS for 15 min. The slices were then rinsed in PBS, briefly dried, and mounted on the glass slides using prolong diamond antifade (Invitrogen, P36961, USA). For quantification of labeling density, after in vivo experimentation, imaging site was marked on the surface of polymer window using fast green. After transcardial perfusion, the tissue corresponding to the imaging site was cut and marked using a fine surgical scalpel. At least two slices were imaged from each imaged brain region and for each indicator. Images were acquired using CSU-W1 SoRa imaging setup of Nikon Spinning-Disk Field Scanning Confocal System with 405, 561 and 647 nm excitation using a ×10 NA0.45 and ×20,0.80 objective lens (Nikon), controlled by NIS-Elements AR software.

For labeling density analysis, we first applied CLAHE [74] function in ImageJ to all channels to homogenize fluorescence intensity distribution and thereby to correct for possible bias in automatic segmentation due to locally dim or bright fluorescence. Then Cellpose [31] was used for segmentation with identical settings. The ROIs at the edge of image were removed. The labeling density was defined as the ratio between GECI-expressing and Nissl +ve ROIs.

Confocal imaging of zebrafish larvae

Two-photon calcium imaging of mouse parabrachial nucleus (PBN) neurons

Two-photon Ca²⁺ imaging of mouse V1 cortical neurons

Analysis of two-photon imaging of the mouse cortex.

Neuronal calcium signals were extracted by Suite2p (Python). ROI detection and signal extraction parameters were uniform for all experiments. And further analysis was performed using MATLAB. Neuropil signals were subtracted from soma signals, with an equation suggested by Suite2p: F_corrected = F_soma − 0.7 × F_neuropil. To screen stimulus response neurons, we first computed trial-average activity of all neurons in each grating directions. Trials were aligned to the stimulus onset. We selected neurons with activity that is significantly different from the baseline in at least one direction (by comparing activity before (−2–0 s) and after (0–2 s) onset, p < 0.05, Wilcoxon rank sum test). We than calculated ΔF/F₀ = (F − F₀)/F₀ for response neurons, where F is the instantaneous fluorescence signal and F₀ is the average fluorescence in the interval 1 s before the start of the visual stimulus. The baseline fluorescence values in extended Fig 7A were reported using fluorescence intensity at neuronal soma without neuropil correction. For each responsive cell, we defined the preferred stimulus as the stimulus that evoked the maximal ΔF/F₀ amplitude in trial averaged activity (peak values during the 2 s of stimulus presentation). The half-decay time was calculated as the time required for each trace to reach half of its peak value (baseline fluorescence subtracted). It was calculated by interpolation with fitted trial averaged activity using a single exponential function. Only cells for which the trial averaged maximal ΔF/F₀ amplitude was higher than four standard deviations above the baseline signal were included in the analysis. The fraction of cells detected as responsive was calculated as the number of significantly responsive cells over all the cells extracted by Suite2p. OSI of responsive neurons was calculated by comparing the response in the preferred orientation (R_pref) and that in the orthogonal orientation (R_orth).

To obtain pairwise Pearson correlation coefficients, neuronal calcium signals were extracted using Suite2p. The z-score was when computed for each neuronal ROI, both with and without neuropil correction. Pearson correlation coefficients were then calculated from the calcium traces of all possible pairs of neurons across different distances.

For the analysis of spontaneous Ca²⁺ dynamics in the primary visual cortex, images of mice expressing calcium indicators at 6 weeks and 4 months were processed using the CaImAn [56] software suite. Specifically, NormCorre [76] in the CaImAn library was used for piece-wise rigid motion correction of the images and the Constrained Non-negative Matrix Factorization (CNMF) [77] algorithm in CaImAn was used for neuronal signal extraction and spike deconvolution. The peak ΔF/F₀ was determined by identifying the maximum value of the ΔF/F₀ signal derived from the CNMF output, where F₀ was defined through a moving windows quantile approach. The peak SNR was computed as the ratio of peak ΔF/F₀ value to the standard deviation of the ΔF/F₀ signal. For S9I Fig, we use isolated single spikes, defined as those without other spikes occurring within 40 frames before or after the event.

Dual-color two-photon imaging of brainstem

One-photon wide-field Ca²⁺ imaging in awake mice

Analysis of one-photon in vivo calcium imaging data in awake mice.

Images from wide-field calcium imaging were processed using open-access code and methods in the following steps;

Denoising: publicly available Python code of SUPPORT [79] was used for denoising the images before signal extraction. Denoising was only performed onto images acquired from mice over 4 months to reduce the noise level dominated by tissue scattering and to aid somatic signal detection. SUPPORT was operated on Pytorch 1.12.1 and CUDA 11.3 with an NVIDIA RTX 3,090 GPU and an Intel Sliver 4212R CPU;

Identification of neuronal ROIs and extraction of calcium signals: publicly available MATLAB code of CNMF-E [38] was used for the identification of neurons in 1-photon calcium imaging data. Down-sampling by 4 in x and y dimension and patch-based processing were used to lower the memory requirements for the analysis. AR(1) model was used for the deconvolution of the calcium traces. Ring model was used for identifying the background. ROI detected within or at proximity of blood vessels were excluded. Images with fluorescence change due to significant motion artifacts were excluded;

Identification of spike events: spikes were detected by the CNMF-E algorithm with AR1 spike deconvolution model. To identify the calcium signal trace originated from a single spike, spike events where past and future spikes were not within 1 s were chosen. From the chosen spikes, average and standard deviations of those spike signals were calculated;

Calculation of pairwise Pearson correlation coefficients: for each ROI detected by CNMF-E algorithm, raw fluorescence traces were extracted from each neuron by calculating the average value under the ROI mask for each frame. Then, Pearson correlation coefficients were then calculated from the calcium traces of all possible pairs of neurons across different distances;

Calculation of Peak ΔF/F₀ and peak SNR: To calculate F0, B-spline fitting was applied to three frames selected from the fluorescence signal (F) prior to the onset of each identified spike. Peak ΔF/F₀ was then defined as the maximal fluorescence change over baseline identified for the calcium transients of each neuron. Peak SNR was defined as the maximal fluorescence change divided by the standard deviation of F₀ as follows.

Calculation of fluorescence decay time: Same as spike identification applied in the “Identification of spike events” above, single spike events where past and future spikes were not within 1 s were chosen to extract the ΔF/F₀ trace of each ROI and determine fluorescence decay time. F₀ was determined by B-spline fitting to the entire fluorescence trace of each ROI. Full decay time was then calculated using the time between the peak and the first time point that reached the negative ΔF/F₀ of the spike.