In vivo imaging of the coupling between neuronal and CREB activity in the mouse brain

Summary

Sensory experiences exert long-term modifications of neuronal circuits by modulating activity-dependent transcription programs, which are vital for regulation of long-term synaptic plasticity and memory. However, it has not been possible to precisely determine the interaction between neuronal activity patterns and transcription factor activity. Here we present a technique using 2-photon fluorescence lifetime imaging (2pFLIM) with new FRET biosensors to chronically image in vivo signaling of CREB, an activity-dependent transcription factor important for synaptic plasticity, at single cell resolution. Simultaneous imaging of red-shifted CREB sensor and GCaMP permitted exploration of how experience shapes the interplay between CREB and neuronal activity in the neocortex of awake mice. Dark-rearing increased the sensitivity of CREB activity to Ca²⁺ elevations, and elongated the duration of CREB activation to more than 24 h in the visual cortex. This technique will allow researchers to unravel the transcriptional dynamics underlying experience-dependent plasticity in the brain.

eTOC blurb

Laviv et al developed new biosensors that allow researchers to image plasticity-regulating activity of a transcription factor CREB in single cells in the brain of living mice.

Introduction

Cortical circuits are responsible for integrating sensory information from the external environment. While neuronal activity operates on millisecond timescale, sensory experience can drive plasticity over days (Holtmaat and Caroni, 2016). One major signaling pathway mediating long-term plasticity is via the conversion of specific neuronal activity patterns to gene transcription programs (West et al., 2002; Yap and Greenberg, 2018). Activity-dependent transcription factors play key roles in decoding synaptic activity and translating it into long-term signaling (Alberini and Kandel, 2015; Cohen et al., 2015; Kandel, 2012). To better understand the mechanisms mediating transcription-mediated circuit plasticity during animal’s learning, it is essential to simultaneously monitor neuronal activity and transcription in a large ensemble of neurons in vivo. Recent improvements of genetically encoded calcium indicators (GECI) (Chen et al., 2013) have enabled researchers to image neuronal activity with cellular and sub-cellular resolution in vivo in behaving animals (Cichon and Gan, 2015; Peters et al., 2014). However, transcription factor activity during learning has mainly been assessed through immunostaining in fixed brain tissue (Alberini, 2009), and thus studying dynamics of transcription factor in neurons of behaving animals has not been possible.

Among activity-dependent transcription factors, cAMP response-element binding protein (CREB) has long been implicated in synaptic plasticity, learning and memory (Bourtchuladze et al., 1994; Han et al., 2007; Pittenger et al., 2002; Silva et al., 1998). CREB binds to multiple recognition sites across the genome (Conkright et al., 2003) in response to increased intracellular Ca²⁺ elevation– in synapses (Bito et al., 1996; Wheeler et al., 2012), cell body (Hardingham et al., 2001) and the nucleus (Dudek and Fields, 2002) and regulates transcription of a wide array of genes (Lonze and Ginty, 2002). Plasticity-inducing stimuli in just a few dendritic spines can lead to CREB activation, suggesting high sensitivity of the CREB signaling pathway (Zhai et al., 2013). Therefore, CREB is considered a primary molecular junction between synaptic inputs and long-term gene-transcription-dependent plasticity (Kandel, 2012). CREB is activated by phosphorylation of Serine 133 (S133), a target of Ca²⁺-dependent signaling such as protein kinase A (PKA) and Ca²⁺/calmodulin-dependent kinases (CaMK) (Gonzalez and Montminy, 1989; Yamamoto et al., 1988). Immunostaining of phosphor-S133 of CREB has been widely used as a marker of CREB activation in vivo (Barco and Marie, 2011). However, to probe how sensory experience shapes CREB dynamics in individual cells in vivo, it is imperative to use an approach which allows for live cell imaging.

Fluorescence resonance energy transfer (FRET) biosensors have revolutionized the investigation of intracellular signaling dynamics in live cells (Miyawaki, 2003). Two-photon fluorescent lifetime imaging (2pFLIM) has been extensively used for imaging FRET in light scattering brain tissue. 2pFLIM signals are relatively insensitive to fluorescence fluctuations, local fluorophore concentration, light scattering and movement artifacts (Díaz-García et al., 2017; Yasuda, 2006). Previously, FRET sensors for CREB activity have been developed based on small peptides derived from the CREB kinase inducible domain (KID) and CREB binding domain (KIX) of CREB binding protein (CBP) (Friedrich et al., 2010; Kobrinsky et al., 2003; Spotts et al., 2002). However, since these sensors are based on synthetic peptides derived from CREB, their regulation and phosphorylation can be different from full-length CREB, and this may affect signal specificity. Furthermore, these sensors were designed for intensity ratiometric imaging, which suffers from wavelength dependent light scattering (Díaz-García et al., 2017; Yasuda, 2006). Overall, the utilization and interpretation of previous sensors’ signal will be complicated especially during in vivo imaging.

Here, we developed sensitive and specific CREB biosensors which report S133 dependent activation of full-length CREB and optimized for in vivo imaging. We demonstrate that changes in CREB activity in single neurons of layer 2/3 somatosensory cortex can be monitored over days during an enriched environment paradigm. Importantly, we demonstrate the utilization of a red-shifted CREB biosensor for simultaneous in-vivo two-photon imaging of CREB activity with neuronal Ca²⁺ dynamics by combining our red CREB sensor with GCaMP6s, a green Ca²⁺ sensor (Chen et al., 2013). Using this approach, we examine the interplay between CREB signaling and neuronal activity in individual neurons in the primary visual cortex. We find that dark rearing, a paradigm inducing adult cortical plasticity, greatly enhances the sensory evoked magnitude and duration of CREB activation. We find that this experience dependent modulation of CREB activity was not associated with increased neuronal calcium, but rather a shift in the coupling between CREB and neuronal activity. Overall, monitoring of activity-dependent molecular signaling using 2pFLIM allows for in vivo interrogation of neuronal signaling and provides a means to dissect the molecular dynamics governing experience dependent plasticity in brain circuits.

Results

Design and validation of a CREB FLIM biosensor

To measure CREB activation in living cells, we developed a sensor to monitor phosphorylation at S133, the main phosphorylation site necessary for CREB signaling (Parker et al., 1996; Yamamoto et al., 1988). The sensor measures the binding of phosphorylated CREB with KIX peptide, which specifically increases its binding affinity to CREB only when it is phosphorylated at S133 (Chrivia et al., 1993; Radhakrishnan et al., 1997). By tagging full length CREB with a donor fluorophore and KIX with acceptor fluorophores, the activation of CREB is reflected as an increase in FRET and mirrors endogenous CREB activity since it retains its transcriptional activity (Chao et al., 2002). While CREB activity is additionally regulated by phosphorylation of S142 and S143 (Gau et al., 2002; Kornhauser et al., 2002), CREB-KIX interaction, and thus our sensor signal, predominantly depends on S133 phosphorylation (Mayr et al., 2001). We have developed two versions of the sensor, a green version (G-CREB), composed of mEGFP-CREB and mCherry-KIX-mCherry (mCh-KIX-mCh), and a red version (R-CREB), composed of mCyRFP2-CREB and mMaroon1-KIX-mMaroon1 (Fig. 1A). mCyRFP2 is a new fluorescent protein, which has improved photostability compared with mCyRFP1, and similar long-strokes shift as well as a mono-exponential fluorescent lifetime decay (Fig. S1). mMaroon1 is a far-red florescent protein (Bajar et al., 2016) that can be used as a highly efficient FRET acceptor for mCyRFP2 (Fig. S1).

Figure 1: Design and validation of a CREB FLIM biosensor.

(A) Schematic design of the CREB FLIM sensor for G-CREB and R-CREB.

(B-C) Representative images of pseudo-colored FLIM before and 40 min after induction of CREB activation by elevation of cAMP (25 μM forskolin and 100 μM IBMX), in HEK-293 cells transfected with G-CREB donor or S133A mutant (B) or with R-CREB and S133A mutant (C). Scale bar, 20 μm.

(D) Average time course of ΔBF in HEK cells following forskolin/IBMX application for G-CREB and S133A sensors.

(E) Same as (C) but for R-CREB sensor.

(F) Quantification of mean ΔBF following 30 – 45 minutes from forskolin/IBMX application in HEK cells (G-CREB: 0.135 ± 0.008 and 0.0186 ± 0.005 for G-CREB and S133A respectively, n=28 for each, p < 0.0001, R-CREB: 0.09 ± 0.004 and 0.013 ± 0.005 for R-CREB and S133A respectively, n = 26 and 32 cells, ****p < 0.0001, two-tailed unpaired t-test.

(G) Average time course of ΔBF of G-CREB in HEK cells following forskolin application (25 μM), followed by response reversal following application of NKY 80 (100 μM), adenylyl cyclase inhibitor. Red and blue lines represent single exponential fits to increase or decrees in binding, half-time of 11.7 and 22.8 min, respectively.

(H) Top: Hippocampal CA1 cell co-expressing R-CREB (magenta) and GCaMP6s (green) merged intensity image, before and after NMDA (20μM) application. Bottom: pseudo-colored FLIM before and 5 minutes after NMDA application. Scale bar- 20μm.

(I) Average time course (blue) and individual cell traces (grey) of ΔBF of R-CREB sensor following NMDA application. Black line is single exponential fit to increase in binding, half-time is 2.1 minutes.

(J) Top: experimental setup for depolarization induced CREB/Ca²⁺ measurement. Bottom: CA1 cell expressing GCaMP/R-CREB before and during stimuli. Scale bar- 20 μm.

(K) ΔF/F₀ for GCaMP6s response during stimulation, n=16 cells.

(L) Lifetime pseudo-colored images for same cell as in (J), for R-CREB before and after stimuli. Scale bar, 20 μm.

(M) Average time-course of R-CREB ΔBF following depolarization, red line is exponential fit. Half-life is 34.2 s, n = 16/16 cells/slices.

**** p < 0.0001, two tailed unpaired t-test, Error bars indicate SEM for all panels.

To measure FRET changes, we used 2pFLIM (Yasuda et al., 2006), a quantitative approach providing robust signals for longitudinal in vivo imaging (Díaz-García et al., 2017; Ma et al., 2018). Following expression of sensors in cell lines, we measured the binding fraction (BF) of G-CREB or R-CREB following stimulation of cells with application of the cocktail of forskolin and IBMX, which increases intracellular cAMP level and activates CREB. This resulted in a robust and sustained increase in the BF of CREB sensor (Fig. 1B–C). To validate the specificity of this FRET increase, we mutated S133 to alanine (S133A), preventing its phosphorylation. This abolished the increase in BF following the same stimulation (changes in BF were 0.135 ± 0.008 and 0.0186 ± 0.005 for G-CREB and S133A, 0.09 ± 0.004 and 0.013 ± 0.005 for R-CREB, respectively, Fig. 1B–F). Furthermore, we could reverse the increase in sensor activity following forskolin application using an adenylyl cyclase inhibitor (NKY 80, Fig. 1G, Fig. S2A, half-time for activation and reversal are 11.7 and 22.8 min, respectively), suggesting specificity and reversibility of the sensor to CREB activity. To further examine CREB sensor activity in neurons, we transfected hippocampal organotypic neurons with G-CREB (Fig. S2B–E). Application of forskolin increased G-CREB BF in neurons and this increase depended on S133 phosphorylation since introduction of S133A mutation in CREB abolished forskolin induced BF increase (Fig. S2C–E). To measure CREB activity following Ca²⁺ elevation, we applied NMDA for 2 minutes at near-physiological temperature (35°C) to hippocampa l neurons co-expressing R-CREB and GCaMP6s (Fig.1H–I). We observed rapid CREB activation which plateaued within 5 minutes following Ca²⁺ influx monitored by GCaMP6s (Fig. 1H–I, half-time of 2.1 minutes). To measure the kinetics of our CREB sensor in slices, we electrically stimulated the Schaffer collateral axons near a CA1 pyramidal cell which expressed both GCaMP and R-CREB (Fig. 1J). We simultaneously monitored Ca²⁺ elevations and R-CREB responses during and following delivery of 350 pulses at 10Hz (Fig. 1K). BF of R-CREB showed a rapid increase following stimuli, with a half-life of 34 s (Fig. 1K–L), and decayed with a half-life of 9.6 min (Fig. S2F). These measurements are in general agreement with rapid (seconds) depolarization induced CREB phosphorylation at S133 (Cohen et al., 2016, 2018) and slower (minutes) dephosphorylation (Bito et al., 1996), measured by immunohistochemistry in dissociated neurons, although on-kinetics is slower than previously reported value (8 s) (Cohen et al., 2016). Together, our results demonstrate that our CREB sensors are well suited to provide specific detection of S133 dependent CREB activity in living cells.

In vivo two-photon FLIM

To image CREB activity in a population of neurons, we generated AAVs to express G-CREB under the neuronal synapsin-1 (SYN) promoter. Three weeks after viral injection, we observed nuclear-localized mEGFP-CREB positive cells co-localized with mCh-KIX-mCh cells (Fig. 2A, Fig. S2F–G). Previous studies have revealed that CREB overexpression using herpes simplex virus (HSV) can cause an increase in the neuronal excitability (Kim et al., 2014; Zhou et al., 2009). We therefore performed whole-cell patch clamp recordings of cortical neurons infected with AAV encoding G-CREB or CyRFP1 (control) to examine electrophysiological properties. We did not find changes in synaptic transmission, firing rates, sag, after-hyperpolatization (AHP) and spike properties of neurons expressing G-CREB compared to control neurons in acute cortical slices (Fig. 2B–C, Fig. S3A–D). One possible reason for this discrepancy could be the relative lower expression levels induced by AAV compared to HSV. We quantified levels of AAV driven G-CREB or R-CREB using immunohistochemistry against total CREB levels, which revealed moderate levels of overexpression (~2 fold, Fig. S3E–F) in contrast to HSV mediated expression which results in ~10 fold overexpression of CREB per cell (Kim et al., 2014).

Figure 2: In vivo two-photon FLIM of CREB activity in L2/3 cortical cells.

(A) Representative in vivo image of G-CREB expression in L2/3 cells in the somatosensory cortex, for mEGFP-CREB (green), mCh-KIX-mCh (magenta) and merged overlay. Scale bar, 100 μm.

(B) Representative traces of whole cell current clamp recordings from L2/3 pyramidal neurons in acute coronal cortical slices expressing control AAV (CyRFP) or G-CREB sensor at three different current steps (−100, 0 and 200 pA).

(C) Mean number of spikes evoked by increasing depolarizing current steps. n=17/7 and 23/9 (cells/animals) for G-CREB sensor and control cells respectively, p = 0.78, two-tailed unpaired t-test.

(D) Pseudo-colored FLIM images of L2/3 cells expressing either WT G-CREB or S133A mutant. Scale bar is 50 μm.

(E) Comparison of in vivo BF values for G-CREB and S133A CREB sensor. Average BF is 0.18 ± 0.003 and 0.108 ± 0.0009, n = 250/4 and 212/4 (cells/animals) for G-CREB and S133A, respectively (****p < 0.0001, two-tailed unpaired t-test).

(F) Correlation between mEGFP-CREB photon number and BF for the same cells in (E), Spearman correlation values are 0.11 (p = 0.09) for WT and 0.11 (p = 0.1) for S133A, respectively.

(G) Representative pseudo-colored in vivo FLIM images of CREB activity in the same cells over 2 days for WT and S133A sensor, scale bar 20 μm.

(H) Quantification of changes in BF over 2 imaging session, n=164/3 and 124/3 cells/animals for WT and S133A sensor, respectively.

(I) Normalized ΔBF for same population of cells as in (H), average change is 0.003±0.02 and 0.001±0.001 for WT and S133A, respectively.

(J) Representative images of pseudo-colored FLIM before and 40 min after acute induction of CREB activation in vivo by inclusion of 1 mM Forskolin in ACSF through a hole drilled in cranial window coverglass, scale bar, 50 μm.

(K) Average time course of mean ΔBF in cortical neurons in vivo following forskolin application for G-CREB and S133A sensor.

(L) Quantification of mean ΔBF following 30–45 minutes from forskolin application, 0.117 ± 0.008, n=41 for G-CREB and 0.021 ± 0.006, n=34 for WT and S133A from 3 and 2 different animals, p < 0.0001, two-tailed unpaired t-test.

****p < 0.0001, Error bars indicate SEM for all panels.

We used 2pFLIM to measure the fluorescence lifetime of individual L2/3 cell nuclei under light anesthesia (methods section). 2pFLIM provides robust signals, independent of the target depth and intensity (Fig. S4). To evaluate the capability of our approach to quantitatively measure CREB activity in single cells over days, we performed detailed analysis of signal-to-noise characteristics of our sensor in vivo, by comparing sensor lifetime and photon collection duration (~1–60 sec, Fig. S5A). We found that frame-to-frame variation of the S133A mutant sensor, which should have minimum biological variation, decreased with increased measurement duration, following the theoretical curve of shot noise (Fig. S5B). When we averaged signal over 40–80 s (290–580 frames at 7.8 Hz imaging), similar to the time constant of the sensor (Fig. 1), the noise levels are < 1% binding fraction. While wildtype G-CREB sensor showed similar frame-to-frame variation as S133A, average BF as well as cell-to-cell variations were much larger (Fig. 2D–E, Fig. S5C). The Number of photons averaged did not correlate with BF of G-CREB and S133A (Fig 2F), suggesting that sensor readout is relatively independent of expression level. In vivo 2pFLIM allowed us to repeatedly measure fluorescent lifetime of G-CREB in the same cells over days (Fig 2G). We measured BF from the same identified cells in two consecutive imaging sessions over the course of 24h (Fig. 2G). On average, BF did not significantly change over days for both S133A and WT sensors, but the G-CREB sensor showed a higher degree of day to day cellular variability than S133A sensor (Fig. 2H–I, Fig. S5B–C). To directly drive CREB activity in vivo, we topically applied forskolin onto a hole in the imaging window, which resulted in a rapid increase in BF, which depended on intact S133 (binding fraction change of 0.117 ± 0.008 and 0.021 ± 0.006 for G-CREB and S133A CREB, respectively, Fig. 2J–L). Collectively, our results demonstrate that G-CREB can report S133 dependent CREB activity in vivo.

CREB dynamics in the somatosensory cortex following sensory enrichment

It has been known that environmental enrichment drives cortical plasticity in somatosensory areas (Bengoetxea et al., 2012). Numerous signaling molecules, including CREB, were shown to be modulated under this condition (van Praag et al., 2000; Rampon et al., 2000). Previous measurements of intracellular signaling during environmental enrichment have mainly used immunostaining in fixed brain tissue and thus it has not been possible to investigate the temporal association between signaling dynamics in single cells and sensory enrichment. Therefore, we monitored CREB activity in the same population of somatosensory L2/3 cells using 2pFLIM over multiple days while mice experienced an enriched environment (Fig. 3A, see Methods).

Figure 3: CREB dynamics in the somatosensory cortex following sensory enrichment.

(A) Experimental design for monitoring the effect of enriched environment on CREB activity in the somatosensory cortex. HC: home cage, EE: enriched environment.

(B) Representative pseudo-colored FLIM images of the CREB sensor in the same population of cells over 3 days interval during imaging sessions in HC₁, HC₂ and EE₃.

(C) Quantification of ΔBF in single cells to HC₁ for HC₁-HC₂-EE, n = 83/3 (cells/animals). One way ANOVA followed by Tukey’s multiple comparisons test.

(D) Alternate experimental design for monitoring effect of enriched environment on ongoing CREB activity.

(E) Representative pseudo-colored images of FLIM of CREB sensor in the same population of cells over 3 days interval during imaging sessions in HC₁, EE₂ and HC₃.

(F) Quantification of ΔBF in single cells over days for HC₁-EE₂-HC₃, n = 83/3 (cells/animals). One way ANOVA followed by Tukey’s multiple comparisons test.

(G) Same as E only for CREB sensor with S133A mutation. Scale bar, 50μm for all images.

(H) Same as (F) only for G-CREB sensor with the S133A mutation, n = 87/4 (cells/animals). One way ANOVA followed by Tukey’s multiple comparisons test.

(I) ΔBF normalized to CREB activity in HC in the first imaging session. Average change for G-CREB of 0.006±0.003, 0.053±0.04, 0.045±0.004 and 0.005±0.003 for HC₁-HC₂, HC₁-EE₃, HC₁-EE₂ and HC₁-HC₃ respectively, for S133A 0.0002±0.001 and −0.0007±0.001 for HC₁-EE₂ and HC₁-HC₃ respectively. One way ANOVA followed by Sidak’s multiple comparisons test. n.s, p>0.05, **** p < 0.0001. Error bars represent SEM.

Mice kept in their home cage (HC) for a period of 3 days with minimal sensory stimuli show little change in overall CREB activity during this period (HC₁-HC₂, Fig. 3A–C). We then exposed the same mice to an enriched sensory environment, which included housing in larger cages containing various objects. Object type and location were changed daily for a period of 3 days before a third imaging session (EE₃) (Fig. 3A–C). Subsequent imaging of the same population of cells revealed a significant increase in overall CREB activity following enriched environment (changes in average BF across cells of 0.053 ± 0.004 and 0.006 ± 0.003 for HC₁-EE₃ and HC₁-HC₂ respectively, Fig. 3B–C, 3I). To rule out non-specific effects, we performed chronic imaging of G-CREB in mice which were transferred from HC after first imaging session to EE before a second imaging session. Again, we could reliably detect an increase in overall CREB activity following EE stimuli (change of 0.045 ± 0.004 for HC₁-EE₂, Fig. 3D–F, 3I). Interestingly, subsequent transfer of mice back to HC environment for 3 days resulted in a reduction of CREB activity levels to the baseline level (change of 0.005 ± 0.003 between HC₁-HC₃, Fig. 3D–F, 3I). In contrast, in vivo imaging of S133A CREB sensor did not show appreciable changes in BF following EE (change of 0.0002±0.001 and −0.0007±0.001 for HC₁-EE₂ and HC₁-HC₃ respectively, Fig. 3G–I). These results demonstrate that in vivo 2pFLIM imaging of CREB sensors allows for single cell monitoring of CREB activity in response to alteration of sensory experience.

Simultaneous CREB and GCaMP two-photon imaging

To fully understand the coupling between neuronal activity and transcription, it is crucial to simultaneously record ongoing CREB activity together with neuronal activity. R-CREB sensor can be simultaneously and separately imaged alongside the highly sensitive GFP based GECI, GCaMP6 (Chen et al., 2013). This combination permitted spectral separation of GCaMP6s and R-CREB under 920nm 2p excitation in hippocampal slices (Fig. 4A, Fig. S6A–C). We then tested simultaneous imaging of R-CREB and GCaMP in vivo. Cells in L2/3 motor cortex were infected with AAV encoding these sensors for in vivo 2pFLIM, while mice were head-fixed on a custom-built running disk with adjustable counterweights (Prevedel et al., 2016) (Fig. 4B–E). Following a habituation period, mice could run on the disc during minutes-long imaging sessions. The improved properties of mCyRFP2 allowed for continuous imaging of R-CREB sensor alongside GCaMP6s in vivo at 8-Hz frame rate for more than 20 min with minimal photo-bleaching (single cell fluorescence of 102.6±0.4%, 104.9±1.0% and 91.37±1.2% after 2, 5 and 20 minutes of imaging, Fig. 4B–C,). GCaMP6s calcium transients and R-CREB FLIM signals within individual cells exhibited no detectable cross-talk during imaging sessions (Fig. 4D–E). Importantly, simultaneous imaging enabled us to directly compare in vivo Ca²⁺ dynamics of L2/3 cells expressing both R-CREB and GCaMP6s to neighboring cells expressing only GCaMP6s. We observed no difference in Ca²⁺ signals between cells with and without R-CREB expression in awake mice (Fig. S6D), corroborating our electrophysiological measurements (Fig. 2B–C, S3A–D). This combination also enabled us to image CREB-Ca²⁺ interplay in L2/3 visual cortex of awake mice during visual stimulation. Again, spectral separation permitted detection of FLIM signal in the red emission channel alongside visually evoked calcium transients, with no detectable cross-talk (Fig. 4F–G). Altogether, this approach enables precise measurements of the coupling between neuronal activity and transcription factor activity in vivo in awake animals.

Figure 4: Simultaneous in vivo imaging of red-shifted CREB sensor and GCaMP6.

(A) Emission spectra of mCyRFP2 and mEGFP with dotted lines indicating green and red band-pass filter ranges.

(B) Example image of mCyRFP2-CREB field of view during in vivo imaging session for the 1^st, 5^th and 20^th minute of an imaging session during consecutive 7.8Hz frame imaging. Scale bar, 100 μm.

(C) Normalized changes in mean fluorescence (black line) and individual cells (grey lines) across imaging sessions, n = 87/3 (cells/animals). Average fluorescence after 1, 5 and 20 min was 102.6±0.4, 104.9±1.0 and 91.37±1.2%, respectively.. p = 0.155 and p = 0.0001 for comparison of 1 and 5 minutes and 1 and 20 minutes, One way ANOVA followed by Dunnett’s multiple comparisons test.

(D) R-CREB (magenta) GCaMP6s (green) intensity images during in vivo imaging session in the motor cortex. Scale bar, 50 μm.

(E) Traces during a 5 min long imaging session showing activity profiles of cells marked in (D) for GCaMP6s transients alongside lifetime measured in red-channel.

(F) R-CREB (magenta) GCaMP6s (green) intensity images during in vivo imaging session in the visual cortex. Scale bar, 50 μm.

(G) Traces during a 50 seconds long imaging session showing visual evoked (blue vertical line) activity profiles of cells marked in (F) for GCaMP6s transients alongside lifetime measured in red-channel.. Errors represent SEM.

Experience dependent CREB dynamics in the primary visual cortex

Following validation of simultaneous imaging of CREB and Ca²⁺ activity in vivo, we set out to quantify how these intracellular signals are coupled in L2/3 visual cortex and how this coupling depends on sensory experience. In the visual cortex, dark rearing (DR) of adult animals induces changes in activity-dependent signaling that leads to the regulation of gene transcription (Tropea et al., 2006). Previous studies have found that dark rearing induces a shift in plasticity state, through a reduced and increased threshold for NMDA-dependent LTP and LTD, respectively (Philpot et al., 2003), correlated with increased ocular dominance plasticity (He et al., 2006). Thus, we postulated that the relationship between Ca²⁺ activity and CREB signaling in the visual cortex could be shaped by DR. We placed adult (>P45) mice in the dark for 7 days and examined Ca²⁺ and CREB activity in vivo in L2/3 visual cortical neurons in awake mice before, during, and after repeated presentations of a natural movie sequence (Fig. 5A). Visual stimulation was performed in the dark so sensory-experience was precisely controlled. We found that, on average, L2/3 cortical neurons of DR mice displayed a robust increase in CREB activity over the 30 min of visual stimulation (Fig. 5B–D, “DR, +stimuli”). This elevated CREB activity persisted for at least 24h (Fig. 5B–D). In contrast, control DR mice, who did not receive visual stimulation, displayed no average change in CREB activity (Fig. 5B–D, “DR, −stimuli”). In naïve mice (housed in normal dark-light settings) CREB activity increased in response to visual stimulation, but at a significantly lower level compared to DR mice and CREB activity returned to the baseline after 24h (Fig. 5B–D, “Naïve, +stimuli”, orange). In naïve animals, a small population of cells did show high activation of CREB at 30 min (BF > 0.05) (Fig. 5C: middle, light orange). However, these cells showed the reversal of BF at 24h, suggesting that the persistence of CREB activation is driven by a shift in plasticity state following DR. In addition, repeated measurements of R-CREB BF before and 7 days after DR revealed no change in basal BF (Fig. S6E), and there was no significant difference in BF between DR and naïve mice prior to visual stimulation (Fig. S6F). These results suggest that both the amplitude and persistence of cellular CREB activity in response to visual stimulation is regulated by sensory experience.

Figure 5: In vivo imaging of experience dependent CREB activity following sensory stimulation.

(A) Illustration of the experimental paradigm: in vivo imaging of R-CREB/GCaMP6s activity in awake head-restrained mice during visual stimulation, either following normal rearing (naïve) or following dark rearing (DR).

(B) Representative pseudo-colored FLIM images of R-CREB activity. Top: DR mouse without visual stimulation, middle: naïve mouse with visual stimulation, and bottom: DR mouse with visual stimulation. Images are show CREB activity before, 0.5 h, and 24 h after visual stimulation. Scale bar: 50 μm.

(C) Time course of ΔBF in single cells BF across the 3 groups. Grey lines denote individual cells and colored thick lines denote average changes. Thick lines with light color denote average trace of cells with high BF (> 0.05) at 22 – 30 min. Statistics are for ΔBF averaged over 22 – 30 min following visual stimuli and at 24 h, compared with baseline: p = 0.98 and p = 0.11 for DR, −stimuli (n = 133/4; cells/animals); p < 0.0001 and p = 0.08 for Naïve, + stimuli (n = 181/4); p < 0.0001 and p < 0.0001 for DR, +stimuli (n = 185/4), for 0.5h and 24h, respectively. One way ANOVA followed by Dunnett’s multiple comparison test.

(D) Comparison of ΔBF across groups DR, −stimuli: −0.0006 ± 0.002 and −0.005±0.003; naïve, + stimuli: 0.035±0.003 and 0.006±0.003, DR, +stimuli: 0.062±0.003 and 0.06±0.003 for 0.5h and 24hr respectively. p < 0.0001 for all comparisons except between DR no stimuli and Naive at 24 h, p = 0.072. One way ANOVA followed by Tukey’s multiple comparison test. n.s. p > 0.05, ****p < 0.0001. Error bars represent SEM.

Since CREB is known to be activated by Ca²⁺ elevation (Sheng et al., 1991), the observed changes by DR could be caused by differences in visual activity in individual cells. To address this, we examined Ca²⁺ and CREB activity during visual stimulation in cells co-expressing GCaMP6s and R-CREB. As demonstrated in two example cells (Fig. 6A–B), individual presentations of the natural movie sequence evoked selective Ca²⁺ responses. Since CREB is likely an integrator of somatic Ca²⁺ influx (Hardingham et al., 2001), we compared cumulative sum of Ca²⁺ with changes in CREB activity (Fig. 6C). Despite slightly higher accumulation of Ca²⁺ activity in naïve animals (integrated Ca²⁺: median ƩΔF/F = 3.63 and 4.16, respectively, p < 0.001, Mann-Whitney test), cells in DR animals displayed larger increases in CREB BF than cells in naïve animals in response to similar levels of integrated Ca²⁺ (Fig. 6C, S6G–H). In addition, we did not find any difference in cumulative Ca²⁺ during visual stimulation in cells co-expressing GCaMP and R-CREB to cells expressing GCaMP alone in the same fields of view (Fig. S6I). We then grouped cells based on their CREB BF levels at 30 min of visual stimulation into high and low activity groups. Cell populations with high changes in CREB BF (> 0.05) exhibited a similar trend in the relationship between CREB activity and integrated Ca²⁺ during the first 30 min of stimuli, for both DR and naïve animals (Fig. 6D). However, CREB activity in DR mice but not in naïve mice maintained that level of activity 24h later, irrespective of the degree of activation during the visual stimulation (Fig. 6D, right). Consistent with the sustained activity of CREB in DR mice, increase in CREB BF following 30 min of visual stimuli was strongly correlated with CREB BF levels 24 h later in DR mice (Spearman’s r = 0.45, p < 0.001), although we also observed a smaller, significant correlation in in the naïve group (Spearman’s r = 0.30, p = 0.002) (Fig. 6E). These data suggest that the coupling between neuronal and CREB activity is modified by sensory experience, and thus provide a potential mechanism for experience dependent shifts in plasiticty state of visual cortical circuits.

Figure 6: Interplay between cumulated Ca²⁺ and CREB dynamics.

(A) Representative Ca²⁺ traces of a cell from a DR animal. Shown are ΔF/F₀ responses during individual trials of a natural movie (30 trials total, 7.8Hz imaging frame rate), averaged over trials ΔF/F₀ activity (bottom left). Summed Ca²⁺ (∑ΔF/F₀) and CREB ΔBF across trials are also shown for this cell (middle). The trial-by-trial increase in CREB ΔBF is plotted with the corresponding summed Ca²⁺ (∑ΔF/F₀) (right). Data points are pseudo-colored by trial number.

(B) Same as in (A) for a cell from a naïve animal.

(C) Relationship between CREB ΔBF and summed Ca²⁺ across cells in DR (red) and naïve (grey) mice. Data points are mean and standard error. n = 61/4 and 114/4 (cells/animals) for DR and naïve, respectively. Also shown is average ΔBF after 30 min of visual stimulation and 24 h later (right) for cells with high visually-evoked Ca²⁺ activity (ƩΔF/F₀ > 4). n = 17/4 and 57/4 (cells/animals) for DR and naïve, respectively.

(D) Relationship between CREB ΔBF and summed Ca²⁺ for cells with high (solid lines, closed circles) and low (dashed lines, open circles) CREB ΔBF (>0.05 or <0.05 at ΔBF30 min, respectively). Also shown is average ΔBF after 30 min of visual stimulation and 24 h later (right) for each of these groups for cells with high visually-evoked Ca²⁺ activity (ƩΔF/F₀ > 4). n = 9 and 8 for low and high CREB in DR mice, and 45 and 12 (cells) for low and high CREB ΔBF in naïve animals, respectively).

(E) Correlation between levels of CREB ΔBF at 30 min and 24 h for cells from DR (red circles, r = 0.45, p < 0.001, n = 56) and naïve (grey circles, r = 0.2930 p = 0.002, n = 87/4, cells/animals) animals.

Discussion

Here we report the development and implementation of in vivo 2pFLIM for chronic monitoring of neuronal transcription factor activity in mouse cortex. Using a novel CREB sensor, we demonstrated that individual cells in the somatosensory cortex undergo dynamic changes in CREB activity over days, positively associated with sensory enrichment. We engineered a red-shifted CREB sensor, which can be combined with green GECIs, to provide simultaneous optical readout of transcription factor activity and calcium activity in awake animals. This enabled us to establish coupling between CREB signaling and neuronal activity in vivo. This approach could be generalized to utilize various biosensors for a wide range of signaling molecules important for synaptic plasticity.

New CREB sensor measures S133 dependent CREB activity

Several sensors for CREB have been previously developed. All of them are based on the interaction of KIX and KID, and thus, these sensors likely report the activity of kinases that can phosphorylate KID peptide in the nucleus (like PKA and CaMKIV (Shaywitz and Greenberg, 1999)). In contrast, our sensors directly monitor the binding of full-length CREB with KIX via S133 phosphorylation and thus these sensors would be regulated similarly to endogenous CREB. In this study, we characterized detection of rapid CREB activity using our sensor following electrical stimulation in hippocampal organotypic slices (half-life of 34 s). Previous immunofluorescence measurements in dissociated neuronal cultures have indicated a faster time-course for endogenous CREB S133 phosphorylation (~8 seconds (Cohen et al., 2016)).Therefore, future utilization of the CREB sensor for precise kinetic analysis will necessitate caution and validation with endogenous CREB signaling readout. Another potential limitation of our CREB sensor is that, since it uses an active transcription factor, it may interfere with endogenous signaling pathways. Indeed, previous studies have found that strong overexpression of CREB resulted in increased neuronal excitability (Kim et al., 2014; Zhou et al., 2009) and could further shift behavioral performance of animals in learning paradigms (Han et al., 2007; Kida et al., 2002; Lisman et al., 2018). In this study we used relatively mild promoter (SYN) and AAVs producing a ~2 fold increased expression level (Fig. S3E–F). Under these conditions, we did not observe any detectable effects of CREB overexpression in slices, by comparing cellular electrophysiological properties (Fig. 2 and S3A–D), and in vivo by measuring spontaneous and evoked calcium transients in cells expressing CREB senor (Fig. S6D, S6I). This is also consistent with a study showing that CREB overexpression with AAV does not alter excitability in adult mouse hippocampus (Yu et al., 2017). Future utilization of recent CRISPR-Cas9 dependent knock-in strategies (Mikuni et al., 2016; Nishiyama et al., 2017) could replace overexpression of biosensors with targeted knock-in of sensors based on endogenous signaling proteins.

CREB dynamics during experience-dependent plasticity

Ongoing changes in the pattern and rate of neuronal activity are manifested in a wide range of activity dependent gene expression profiles (Flavell and Greenberg, 2008). In particular, previous studies have examined how changes in neuronal activity in vivo and in vitro can lead to CREB mediated gene transcription using immunohistochemical analysis of S133 phospho-CREB (Alberini, 2009). In dissociated neuronal cultures, sustained synaptic activity can induce CaMKII-calmodulin dependent CREB phosphorylation (Bito et al., 1996; Ma et al., 2014). In hippocampal slices, CREB in pyramidal neurons can be activated by a burst of somatic action potentials without synaptic input (Dudek and Fields, 2002), or LTP-inducing stimuli in a few dendritic spines (Zhai et al., 2013). In addition, CREB activity can be detected in various brain regions following learning paradigms or sensory stimuli (Ginty et al., 1993; Han et al., 2007; Liu and Graybiel, 1996; Moore et al., 1996; Tropea et al., 2006). However, immunohistochemistry allows for analysis at one time-point, making it difficult to determine temporal dynamics of CREB during specific behavior, as well as variability across cells or within animals.

The benefit of our in vivo imaging technique is highlighted by our demonstration of CREB dynamics in the somatosensory cortex during enriched sensory environment over days (Fig. 3) and simultaneous imaging of R-CREB and GCaMP in awake animals following and during sensory manipulation in the visual cortex (Figs. 4, 5). Our technique thus enables us to access to the coupling between CREB activation and neuronal plasticity at the single cell level. Previously, live imaging revealed increased transcription and translation of the immediate early genes c-Fos and Arc during experience-dependent plasticity (Barth et al., 2004; Cao et al., 2015; Kawashima et al., 2014; Wang et al., 2006). While these genes are considered to be regulated by CREB, the temporal coupling between CREB activation and expression of these target genes in single cell level is not well understood. Future utilization of simultaneous imaging of CREB activity and IEG expression may help to elucidate experience dependent transcription-translation interplay.

Visual cortical experience-dependent plasticity

Structural and functional synaptic changes follow sensory deprivation in the primary visual cortex of young and adult mice ( Hengen et al., 2013; Hensch, 2005; Scholl et al., 2017; Hofer et al., 2006; Keck et al., 2013; Rose et al., 2016). While the CREB signaling pathway is thought to be involved in visual cortical plasticity (Mower et al., 2002; Pham et al., 1999; Pulimood et al., 2017; Suzuki et al., 2004), it has been difficult to attribute differences in CREB signaling within single cells and their activity profiles. Furthermore, while a general relationship between neuronal activity and CREB activity has been described in vitro (Bito et al., 1996; Fields et al., 1997), it has been difficult to precisely determine how this interplay is manifested in cortical circuits in vivo, especially during sensory manipulation.

Our in vivo imaging approach revealed a dynamic regulation of CREB activity in the visual cortex: dark rearing for 7 days dramatically increases visually evoked CREB activity following minutes, and maintains its elevated activity levels for a period of at least one day. Interestingly, this increase was not associated with an appreciably increased calcium dynamics, and instead, modified the CREB readout or sensitivity to integrated Ca²⁺ signaling. Other potential mechanisms for this modulation could arise from neuromodulators or neurotrophic BDNF-TrkB signaling, both previously associated with experience dependent plasticity in the visual cortex (Huang et al., 1999; Yaeger et al., 2019). Future studies could determine whether this experience-dependent shift represents a general phenomenon of cortical plasticity and identify the underlying mechanisms.

By establishing an experimental approach to visualize the spatiotemporal dynamics of CREB activity combined with Ca²⁺ activity in vivo, we have correlated functional neuronal responses with molecular signaling dynamics at the microcircuit level. In vivo 2pFLIM of molecular signaling in the brain alongside neuronal activity will allow researchers to investigate the mechanisms underlying experience dependent plasticity within functional neuronal circuits in the brain.

STAR methods

Lead contact and materials availability

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ryohei Yasuda (Ryohei.yasuda@mpfi.org). Plasmids generated in this study are available upon request and will be deposited to Addgene.

Experimental Model and Subject Details

Mice

All experimental procedures were approved by the Max Planck Florida Institute for Neuroscience Institutional Animal Care and Use Committee and were performed in accordance with guidelines from the US NIH. For in utero electroporation: Swiss Webster pregnant dams were obtained from Charles River Laboratories and were used in E14.5–15.5. Following electroporation, both male and female offspring mice were used for cranial window surgeries at p45–60. For all other experiments, male C57BL/6 mice were purchased from Charles River Laboratories and used for viral injections and cranial windows surgeries at P45-P80 for in vivo imaging and at p21–28 for acute slice experiments.

Methods details

DNA plasmids and AAVs

For in vivo FLIM characterization, plasmids under CAG promoter were used to express mEGFP, EGFP with A206K mutation (Zacharias et al., 2002), and mEGFP-mCherry (Murakoshi et al., 2008). For construction of CREB sensor, EGFP followed by a short linker of PTPTPT was fused to the N terminus of CREB (from RSV CREB, a gift from Marc Montminy, Addgene plasmid # 22394). The KIX sequence (amino acids 591–662, generous gift from Oliver Griesbeck) was flanked between mCherry sequences and separated by linkers, as follows: mCherry-SGLRSRA-KIX domain-SAVDGTAGPGSG-mCherry. For cell line experiments, both donor and acceptor were cloned into pEGFP-C1 under the CMV promoter. For expression and production of AAVs, both donor and acceptor were cloned into AAV plasmid backbone under the human synapsin-1 promoter. For red-shifted R-CREB sensor, mEGFP was replaced with mCyRFP2 and mCherry was replaced with mMaroon1 (Bajar et al., 2016). Mutant CREB donor was assembled by a single point mutation in the CREB reading sequence resulting in S133A mutation in CREB. All cloning procedures were performed using Gibson assembly cloning kits (NEB) and mutations were performed with the Q5 mutagenesis kit (NEB). AAVs for CREB sensor (serotype 9) were packaged in the UNC vector core. AAVs (serotype 9) encoding for hSyn-GCaMP6s (a gift from The Genetically Encoded Neuronal Indicator and Effector Project (GENIE) & Douglas Kim) were purchased from Upen viral core and Addgene.

Cell culture experiments

HEK293T cells (GE Dharmacon, Fischer Scientific) were cultured in DMEM supplemented with 10% FBS at 37 °C in 5% CO2 and tr ansfected with plasmids using Lipofectamine 2000 (Invitrogen). For G-CREB, mEGFP-CREB/mCh-KIX-mCh were transfected in a ratio of 1:3, and R-CREB, mCyRFP2-CREB/mMaroon1-KIX-mMaroon1 were transfected in a ratio of 1:2. S133 donor mutants were used at same ratios as WT donor. Imaging was performed 24–48 h following transfection. HEK Cells were used as an expression platform only, and were not rigorously tested for potential contamination from other cell lines.

Organotypic hippocampal slice cultures

Hippocampal slices were prepared from postnatal 4- to 6-day-old C57BL/6 mice, as described previously (Stoppini et al., 1991). In brief, 350 μm-thick hippocampal slices were dissected using a tissue chopper. Slices were placed on Millicell membranes (Millipore) in culture medium containing minimal essential medium (Life Technologies), 20% horse serum, 1 mM L-glutamine, 1 mM CaCl2, 2mM MgSO4, 12.9 mM D-glucose, 5.2 mM NaHCO3, 30 mM Hepes, 0.075% ascorbic acid, 1 μg/mL insulin, which was exchanged every other day. Neurons were either transfected via gene gun (Laviv et al., 2016) or infected by pressure injection of AAV mixtures using picrospritzer (100–200nl) and imaged at DIV 13–18.

In utero electroporation

In utero electroporation was performed as previously reported (Saito, 2006). In brief, Swiss Webster E14.5–15.5 timed pregnant dames (Charles River Laboratories) were anesthetized with ~2% isoflurane and administered 0.1 mg buprenorphine SR (ZooPharm) for analgesia, uterine horns were exposed though an abdominal incision, and the right lateral ventricle of each embryo was injected with plasmids encoding either CAG-mEGFP or CAG-mEGFP-mCherry, at concentration of 1 μg/μl with 0.01% Fast Green dye (Sigma-Aldrich). Five electrical pulses (40 V, 50-ms duration, 1 Hz) were delivered using a NEPA21 electroporator (NEPAGENE) with electrodes directed to the motor cortex.

AAV injection and cranial window surgeries

Mice were deeply anesthetized with isoflurane for induction (3–5%) and maintenance (2%). Mice were administered 1 μg/g buprenorphine SR for analgesia, placed in a stereotaxic frame, and administered 0.2 mg/kg dexamethasone and 5 mg/kg carprofen to prevent edema and inflammation. Following removal of skin and skull exposure, a 2.5–3mm circular craniotomy was performed centered over intended imaging site using a dental drill. The site of injection was determined using stereotactic coordinates, based on mouse brain atlas for the somatosensory (1.4 mm posterior and 3.3mm lateral from bregma), motor (1mm posterior and 1.5 mm lateral from bregma) and visual (0.3 mm anterior and 2.5 mm lateral from lambda) cortex. The pipette was lowered to approximately 300 μm below the pial surface, and AAVs were slowly pressure injected over 5 min using a Picospritzer (Parker) at a rate of 0.3Hz with pulses of 20–100 msec pulse duration. Total volume of injection was 100–350 nl. For G-CREB sensor, we injected a mixture of mEGFP-CREB/mCh-KIX-mCh at 2×10¹² and 8×10¹² viral genome (vg)/ml. For R-CREB/GCaMP6s we used a mixture of mCyRFP2-CREB, mMaroon-KIX-mMaroon1, and GCaMP6s at concentration of 1×10¹³, 3×10¹³, and 1×10¹² vg/ml respectively. Following injections, the glass pipette was left in place for 5 minutes and then slowly removed. The skull was sealed using a 2.3 mm, No.1 circular coverglass glued on a 5mm circular coverglass. The coverglass was then cemented to skull along with a head plate to secure the head during imaging using dental cement (C and B Metabond, Parkell). Mice were left to recover and were used for in vivo imaging 21–30 days following surgery. For chronic measurements, mice that showed signs of occlusion of window or tissue damage during the course of imaging experiments were excluded from analysis.

2pFLIM of organotypic cultures and cell lines

FLIM imaging of HEK cells and hippocampal slice cultures were performed using custom 2p microscopy as previously described (Laviv et al., 2016). Chameleon Ti:sapphire laser (Coherent) was used for excitation at 920 nm. Emission was collected with a 60× 1.0 NA objective (Olympus), divided with a 565-nm dichoic mirror (Chroma) and detected with two PMTs with low transfer time spread (H7422–40p, Hamamatsu) placed after wavelength filters (et520/60–2p for green and et620/60–2p for red, Chroma). PMT voltage was set to 820 V. Average laser intensity for power was set at 1.5–2.0 mW, as measured under the objective. Imaging was performed at room temperature (25°c), except for measurements of kinetics with NM DA and electrical stimulation, which were carried out at 35–36°c. Dual color fluorescenc e lifetime images were obtained using two time-correlated single-photon counting board (Time Harp 260, Picoquant) controlled with custom software written in C#.

Slice electrophysiology

For the characterization of electrophysiological properties of G-CREB expressing cells, C57BL/6 mice (p21-p28) were injected with an AAV mix for G-CREB sensor (mEGFP-CREB and mCh-KIX-mCh) and a mixture of FLEX-CyRFP and Cre on the right and left hemispheres respectively. Following 2–3 weeks of expression, animals were sedated by isoflurane inhalation, and perfused intracardially with ice-cold choline chloride solution (in mM: choline chloride 124, KCl 2.5, NaHCO₃ 26, MgCl₂ 3.3, NaH2PO₄ 1.2, Glucose 10 and CaCl₂ 0.5; pH 7.4, equilibrated with 95%O₂/5%CO₂). Brains were then removed and placed in the same chilled choline chloride solution and coronal acute slices of 400μm from left and right hemispheres were collected and placed in oxygenated (95%O₂/5%CO₂) ACSF (in mM: NaCl 127, KCl 2.5, Glucose 10,NaHCO3 25, NaH2PO4 1.25, MgCl2 2, CaCl₂ 2) at 32°C for 1h and then maintained at RT for the rest of the experiment. Cortical infected pyramidal neurons were visualized using epifluorescent illumination. Whole cell current clamp recordings were obtained using a Multiclamp 700B amplifier. Patch pipettes (3–6 ΩM) were filled with a K Gluconate solution (in mM: K gluconate 130, Na phosphocreatine 10, MgCl2 4, NaATP 4, MgGTP 0.3, L- Ascorbic acid 3, HEPES 10. pH 7.4, 310 mOsm). Experiments were performed at room temperature (~23°C) and slices were perfused with o xygenated ACSF. Recordings were digitized at 10 KHz and filtered at 2 KHz. All data was acquired and analyzed with a custom software written in Matlab.

Electrical stimulation in organotypic slices

Following transfection of R-CREB/GCaMP6s, labeled CA1 cells were stimulated by placing a bipolar extracellular electrode near Schaffer collateral axons, with a 350 pulse 10Hz stimuli train. Simultaneous imaging of GCaMP/CREB was performed at 0.78Hz. for on-rate kinetics measurements. Extracellular oxygenated ACSF contained 2mM Ca²⁺ 2mM Mg²⁺ and was maintained at 35–36°c.

In vivo imaging

In vivo 2pFLIM was performed using a custom 2-photon microscope, with a Chameleon Ti:sapphire laser (Coherent) tuned to 920–940 nm. The laser was modulated by Pockel cells (Conoptics, Model 350–80LA). Microscope was constructed with one pair of 5mm Galvo mirrors (Cambridge Technologies) coupled with a scan lens (Thorlabs, LSM04-BB) and tube lens (Thorlabs, TTL200-B). Emission was collected with a 20×1.0 NA objective (Olympus) or a 16×0.8 NA (Nikon), divided with a 565-nm dichroic mirror (Chroma) and detected with two PMTs with low transfer time spread (H7422–40p, Hamamatsu) placed after wavelength filters (et520/60–2p for green and et620/60–2p for red, Chroma). Excitation power was set at 10–40 mW, as measured under the objective. Two color FLIM data was acquired via a Time-Correlated Single Photon Counting board (Time Harp 260, Picoquant) with the temporal resolution of 200 ps using custom software written with C# (source code in https://github.com/ryoheiyasuda/FLIMage_public). For CREB activity imaging in anesthetized mice, images (128 × 128 pixels) were acquired at the frame rate of 7.8 Hz and summed over 600 frames. For dual imaging of CREB activity (R-CREB) and calcium (GCaMP6s) in awake mice, images (128 × 128 pixels) were collected at the frame rate of 7.8Hz and summed over 400 frames. Both R-CREB and GCaMP signals are acquired as fluorescence lifetime images, but GCaMP6s data was analyzed only for intensity.

Forskolin application during in vivo imaging

Following 2–3 weeks of cranial window and injection of G-CREB sensor AAV mix, mice were anesthetized with isoflurane and head-fixed with head-plate. A small hole was drilled in the side of the coverglass with a dental drill. Mice were then imaged with 2pFLIM with ACSF solution as immersion fluid, and following 5–10 minutes of baseline imaging, ACSF was replaced with ACSF containing 1mM forskolin, continuing to image same field of view for 30–40 min.

Immunohistochemistry

Adult mice were deeply anesthetized by intraperitoneal injection of Ketamine (10 mg/ml)/ Xylazine(1 mg/ml) and intracardially perfused with saline, then 4% paraformaldehyde in 0.1 M phosphate buffer (PB). Whole brains were postfixed for 2 h in the same fixative at 4°C, and PBS overnight. 50 μm thick coronal sections were cut by vibratome (VT1200S, Leica). The sections were permeabilized in 0.5% TritonX-100/PBS for 10 min at room temperature, and blocked in blocking buffer (4% BSA or 5% normal goat serum (NGS) in 0.1% TritonX-100/ 0.01% NaN₃/ PBS). The sections were reacted with primary antibodies at 4°C overnight, i ncubated with secondary antibodies for 2 h at room temperature, and finally mounted in Fluoromount-G (Southern Biotech). The confocal images were captured with a laser scanning microscope system (LSM880, Zeiss). The following antibodies were used: CREB (48H2) Rabbit mAb (Cell Signaling, #9197), 1:800 in 5% NGS blocking buffer; ATTO647-conjugated RFP-Booster (Chromotek, rba647n-10), 1:600 in 4% BSA blocking buffer; Alexa Fluor 488-conjugated goat anti-rabbit IgG (Invitrogen, A-11034), 1:1000 and Alexa Fluor 405-conjugated goat anti-rabbit IgG (Invitrogen, A-31556), 1:1000 in 5% NGS blocking buffer.

Enriched environment

For experiments involving enriched environment, mice were first individually housed in conventional mouse cage, containing only nesting material. To increase sensory enrichment, mice were transferred to a larger size cage which contained various object such as different nesting materials, plastic nests and tunnels, running wheel and metal chains which were hanging from cage wire. Objects were altered on a daily basis for a period of 3 days to maintain sensory novelty.

Dark Rearing experiments

For dark rearing experiments, 2–3 weeks following viral injection of AAVs encoding R-CREB and GCaMP6s in primary visual cortex1, mice were individually housed in complete darkness in environmental chambers, with air, food and water access. Mice were checked daily with IR illumination and night vision goggles for general health. Following 7 days of dark rearing, mice were transferred in the dark and head-fixed under the microscope for the 1^st imaging session. Following 5 minutes of baseline recording of CREB and GCaMP6s activity, mice were exposed to visual stimuli displayed in a monitor situated 15 cm from mice. The stimuli consisted of a short natural movie sequence, lasting 30s. We measured GCaMP6s responses using the green channel, while measuring R-CREB FLIM signal for 20 s following each movie sequence. This stimulation pattern was repeated for 30 min. In the naïve animal group, we performed the same visual stimulation paradigm while housing mice in normal light-dark cycle housing conditions.

Mice were returned to their respective housing condition, and were imaged 24h later to determine CREB activity levels, which were averaged over a 5 min of imaging session in the dark.

FLIM analysis

FLIM analysis was performed with a custom software written in C# (source code: https://github.com/ryoheiyasuda/FLIMage_public). To measure fluorescence lifetime, we fit fluorescence lifetime decay curve F(t) with a monoexponential or biexponential function convolved with the Gaussian pulse response function

$$A(t)=A_0{\mathrm{\Sigma }}_i{P}_{i}H\left(t,t_0,{\tau }_i,{\tau }_G\right)$$

where P_i is the fractional population with the decay time constant of τ_i, and H(t) is an exponential function convolved with the Gaussian instrument response function (IRF),

$$H\left(t,t_0,t_i,t_G\right)=\frac{1}{2}\exp \left(\frac{{\tau }_G^2}{2{\tau }_i^2}-\frac{t-t_0}{{\tau }_i}\right)\mathrm{erfc}\left(\frac{{\tau }_G^2-{\tau }_i\left(t-t_0\right)}{2{\tau }_i{\tau }_G}\right),$$

in which τ_G is the width of the Gaussian pulse response function, t₀ is the time offset, and erfc is the error function. A₀ is the initial fluorescence before convolution. Weighted residuals were calculated as

$$E(t)={(F(t)-A(t))}^2/F(t).$$

Fitting was performed by minimizing the summed error δ² ₌ Σ_tE(t) for parameters t₀, τ_i (i = 1,2) and τ_G.

P_DA was referred to as ‘binding fraction’ if measuring binding. To generate a fluorescence lifetime image, we calculated the averaged fluorescence lifetime (τ_m) by the mean photon arrival time subtracted by t₀ in each pixel as

$${\tau }_m={\mathrm{\Sigma }}_{t}tF(t)/{\mathrm{\Sigma }}_{t}F(t)-t_0,$$

We generate a binding fraction image (a map of P_DA) by numerically solving the relationship between τ_m and P_DA with equations

$${\tau }_m={\mathrm{\Sigma }}_{t}tA\left(t,P_{DA}\right)\text{d}t/{\mathrm{\Sigma }}_{t}A\left(t,P_{\text{DA}}\right)-t_0$$

and

$$A\left(t,P_{\text{DA}}\right)=A_0\left[\left(1-P_{\text{DA}}\right)H\left(t,t_0,{\tau }_{\text{D}},{\tau }_{\text{G}}\right)+P_{\text{DA}}H\left(t,t_0,{\tau }_{\text{DA}},{\tau }_{\text{G}}\right)\right],$$

where t₀ and τ_G are obtained from a curve fitting to the fluorescence lifetime decay averaged over all ROIs in an image. τ_D and τ_DA were obtained from donor alone or CREB sensor expressed in HEK-293 cells in a separated experiments.

For analysis of CREB activity in vivo, we drew ROIs to positive nuclei in the donor channel and quantified τ_m and P_DA. Cells were chosen for analysis based on photon counting from the cell (>50,000 photons).

The theoretical noise of the binding fraction (P_DA) due to photon shot noise can be calculated for very short IRF (τ_G = 0), where τ_m is approximately:

$${\tau }_m=\frac{\left(1-p_{\mathrm{DA}}\right){\tau }_D^2+p_{\mathrm{DA}}{\tau }_{AD}^2}{\left(1-p_{\mathrm{DA}}\right){\tau }_D+p_{\mathrm{DA}}{\tau }_{AD}}.$$ Eq. 1

Since the relationship between the shot-noise of τ_m (δτ_m) and photon counting (N) is known to be $\delta {\tau }_m={\tau }_m/\sqrt{N}$ (Yasuda et al., 2006), the theoretical noise of the binding fraction (P_DA) due to shot noise can be calculated by taking differential of Eq. 1 as:

$$\delta P_{\text{DA}}=\frac{\left[1-p_{\mathrm{DA}}\left(1-r^2\right)\right]\left[1-p_{\mathrm{DA}}(1-r)\right]}{r(1-r)}\frac{1}{\sqrt{N}},$$

where r = τ_AD/τ_D.

For chronic images of the same cells, imaging field of view was re-identified using blood vessel topography and fluorescence in green/red channels. For anesthetized imaging multiple fields of view were imaged whereas for awake imaging one field of view was imaged and analyzed per animal. For dual GCaMP-CREB imaging, we drew ROIs on cells expressing both GCaMP and CREB and segments of 200–400 frames at 7.8 Hz repetition were summed for FLIM analysis. MATLAB or C# code for FLIM analysis will be provided upon request.

GCaMP analysis

Images were corrected for in-plane motion using a correlation-based approach (MATLAB). ROI drawing was performed in ImageJ (Schindelin et al., 2012)A. ROI’s were circular or drawn using custom software (Cell Magic Wand). Fluorescence time-courses were computed as the mean of all pixels within the ROI at each time point and were extracted using Miji (Sage et al., 2012). For each imaging session, fluorescence time courses were computed as changes in fluorescence, ΔF, relative to the baseline fluorescence, F₀, which was computed as a 300 point (2.34 s) median filtered fluorescence trace. Fluorescence signals were sometimes contaminated by surrounding neuropil, so we used a computational subtraction procedure to isolate cell calcium signals as follows: (1) perform a robust fit (MATLAB) of ΔF/F₀ against neuropil ΔF/F₀ (25 pixel radius around ROI center) and (2) subtract a scaled version of the neuropil ΔF/F₀, where the scaling factor equals the slope from the robust fit. Cells with a correlation >0.20 between residual ΔF/F₀ and neuropil ΔF/F₀ were excluded from further analysis. Summed Ca²⁺ (∑ΔF/F₀) has computed as the sum (across stimulus trials) of the mean ΔF/F₀ during visual stimulation.

Quantification and Statistical Analysis

Statistical tests of student t test and one or two way ANOVA as indicated in figure legends. GraphPad Prism and MATLAB were used for statistical analysis. Data is presented throughout paper as mean and errors represent SEM, n represents number of cells/animals or cells/slices as described in figure legends. For all statistical tests *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.001 were considered significant. For comparisons of CREB and CREB-Ca²⁺ relationships, we used bootstrapped principle components analysis (Sokal and Rohlf, 1995). No statistical methods were used to predetermine sample size.

Data and Code Availability

The data supporting the current study are available from the corresponding author on request. Code and software for analysis of FLIM data is available on Yasuda lab Github: https://github.com/ryoheiyasuda/FLIMage_public.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CREB (48H2)	Cell signaling	Cat #9197 RRID-AB_331277
ATTO647-conjugated RFP-Booster	Chromotek	Cat #rba647n-10
Bacterial and Virus Strains
AAV-pSyn-mEGFP-CREB (AAV-9)	This study	N/A
AAV-pSyn-mEGFP-CREB-S133A (AAV-9)	This study	N/A
AAV-pSyn-mCyRFP2-CREB (AAV-9)	This study	N/A
AAV-pSyn-mCherry-KIX-mCherry (AAV-9)	This study	N/A
AAV-pSyn-mMaroon1-KIX-mMaroon1 (AAV-9)	This study	N/A
AAV-pSyn-GCaMP6s	(Chen et al., 2013)	Addgene_100843
Chemicals, Peptides, and Recombinant Proteins
Forskolin	Tocris	Cat #1099
IBMX	Tocris	Cat #2845
NKY-80	Tocris	Cat #5071
Lipofectamine 2000	Thermo Fisher Scientific	Cat # 11668030
Rimadyl (Carpofen)	Zoetis	10000319
Dexamethasone	Phoenix	Cat # 24305
C & B METABOND	Parkell	N/A
Buprenorphine SR	ZooPharm	N/A
Experimental Models: Cell Lines
HEK 293 cells	GE Dharmacon, Fischer Scientific	Cat # HCL4517
Experimental Models: Organisms/Strains
Mouse: Swiss Webster	Charles River Laboratories	CRL:024
Mouse: C57BL/6J	Charles River Laboratories	CRL:027
Recombinant DNA
pCAG-mEGFP	This study	N/A
pCAG-mEGFP-mCherry	(Murakoshi et al., 2008)	N/A
pCMV-mEGFP-CREB	This study	N/A
pCMV-mCyRFP2-CREB	This study	N/A
pCMV-mCherry-KIX-mCherry	This study	N/A
pCMV-mMaroon1-KIX-mMaroon1	This study	N/A
pCAG-GCaMP6s	(Chen et al., 2013)	Addgene_100844
Software and Algorithms
Graph Pad Prism 8	Graph Pad	https://www.graphpad.com/scientific-software/prism/
MATLAB	Mathworks	https://www.mathworks.com/products/matlab.html
Image J	(Schindelin et al., 2012)	https://imagej.nih.gov/ij/
Miji	(Sage et al., 2012)	https://imagej.net/ImageJ
FLIMage	This study	Yasuda lab. https://github.com/ryoheiyasuda/FLIMage_public

Highlights.

• New CREB biosensors (green and red) for CREB activity.

• Chronic in vivo imaging of CREB activity in L2/3 cortical cells over days.

• Simultaneous imaging of CREB and Ca²⁺ in awake mice.

• Interplay between CREB-Ca⁺² is modulated by sensory experience.