Abstract
The cAMP-response element–binding protein (CREB) transcription factor controls the expression of the neuronal immediate early genes c-fos, Arc, and Bdnf and is essential for long-lasting synaptic plasticity underlying learning and memory. Despite this critical role, there is still ongoing debate regarding the synaptic excitation–transcription (E–T) coupling mechanisms mediating CREB activation in the nucleus. Here we employed optical uncaging of glutamate to mimic synaptic excitation of distal dendrites in conjunction with simultaneous imaging of intracellular Ca2+ dynamics and transcriptional reporter gene expression to elucidate CREB E–T coupling mechanisms in hippocampal neurons cultured from both male and female rats. Using this approach, we found that CREB-dependent transcription was engaged following dendritic stimulation of N-methyl-d-aspartate receptors (NMDARs) only when Ca2+ signals propagated to the soma via subsequent activation of L-type voltage–gated Ca2+ channels resulting in activation of extracellular signal-regulated kinase MAP kinase signaling to sustain CREB phosphorylation in the nucleus. In contrast, dendrite-restricted Ca2+ signals generated by NMDARs failed to stimulate CREB-dependent transcription. Furthermore, Ca2+-CaM-dependent kinase–mediated signaling pathways that may transiently contribute to CREB phosphorylation following stimulation were ultimately dispensable for downstream CREB-dependent transcription and c-Fos induction. These findings emphasize the essential role that L-type Ca2+ channels play in rapidly relaying signals over long distances from synapses located on distal dendrites to the nucleus to control gene expression.
Significance Statement
The transcription factor cAMP-response element–binding protein (CREB) controls gene expression required for long-lasting synaptic plasticity and learning and memory, yet the synapse-to-nucleus signaling mechanisms mediating CREB activation are still unclear. Using glutamate uncaging to mimic synaptic input to dendrites, this study shows that Ca2+ signals propagated to the soma by L-type Ca2+ channels engage the extracellular signal-regulated kinase (ERK) cascade to mediate CREB phosphorylation and CREB-dependent transcription. In contrast, dendrite-restricted Ca2+ signals generated primarily by N-methyl-d-aspartate receptors failed to effectively engage this signaling pathway or CREB-dependent transcription. In addition, we found that while ERK and CaMK pathways may both contribute to increased CREB phosphorylation immediately following neuronal stimulation, sustained ERK signaling to CREB was necessary to drive CREB-dependent transcription.
Introduction
The transcription factor cAMP-response element–binding protein (CREB) is required for the long-lasting maintenance of key forms of synaptic plasticity underlying learning and memory, such as the late-phase of long-term potentiation (LTP), through its regulation of immediate early genes (IEGs), including those encoding the transcription factor c-Fos, the synaptic protein Arc, and the neurotrophic factor BDNF (Sheng and Greenberg, 1990; Tao et al., 1998; West et al., 2001; Shepherd et al., 2006; Pettit et al., 2022; Ma et al., 2023). Despite the necessity of CREB in learning and memory, the synapse-to-nucleus excitation–transcription (E–T) coupling mechanisms mediating its activation are still contested.
Activation of CREB in the nucleus requires its phosphorylation on Ser133, which allows recruitment of transcriptional coactivators (Sheng et al., 1990, 1991; Chrivia et al., 1993; Kwok et al., 1994). In general, two major types of kinase pathways are implicated in mediating CREB phosphorylation in response to neuronal activity: those that signal through the extracellular signal-regulated kinase (ERK) MAP kinase pathway and those that converge on the Ca2+-CaM–dependent kinase IV (CaMKIV). Both types of E–T coupling pathways can be activated downstream of the second-messenger Ca2+; however, the source of Ca2+ can vary. Some prior studies indicate that dendritic N-methyl-d-aspartate receptor (NMDAR) Ca2+ influx is sufficient to activate ERK leading to its slow, long-distance translocation from dendrites to the nucleus to phosphorylate CREB (Karpova et al., 2013; Zhai et al., 2013; Melgarejo da Rosa et al., 2016). However, other studies show that Ca2+ influx through somatic CaV1 L-type voltage–gated Ca2+ channels (LTCC) can rapidly drive CREB-regulated gene transcription via either ERK (Rosen et al., 1994; Impey et al., 1998; Dolmetsch et al., 2001; Dudek and Fields, 2002; Servili et al., 2018, 2019) or CaMKIV signaling (Chawla and Bading, 2001; Wheeler et al., 2008). In addition, a number of studies propose Ca2+-CaM–dependent kinase II (CaMKII) as an intermediary between NMDARs or LTCCs and CREB (Wheeler et al., 2008, 2012; Li et al., 2016; Wang et al., 2017; Perfitt et al., 2020), with prior work supporting CaMKII signaling to both CaMKIV (Ma et al., 2014; Cohen et al., 2018) and ERK (Illario et al., 2003, 2005; Salzano et al., 2012; Melgarejo da Rosa et al., 2016). Finally, elevations in nuclear Ca2+ can directly activate CREB via CaMKIV (Hardingham et al., 2001; Limbäck-Stokin et al., 2004; Yu et al., 2017; Lobos et al., 2021).
While all the above CaMK and ERK signaling mechanisms may contribute to CREB phosphorylation, most prior studies did not make corresponding direct measurements of downstream CREB-dependent transcription nor conduct systematic analyses of which CREB signaling mechanisms may be engaged in response to different types of neuronal stimuli. Moreover, differences in the sources of Ca2+ mediating CREB activation may encode specific attributes of neuronal stimulation, with pathways downstream of dendritic NMDARs potentially relaying information about neuronal input in terms of the number of individual dendrites stimulated (Zhai et al., 2013). In contrast, pathways downstream of somatic LTCCs may reflect total neuronal output resulting from integration of all dendritic inputs. Thus, differences in the Ca2+ sources and signaling pathways engaged may allow CREB to encode different types of information. With the notable exception of single-cell stimulation by glutamate uncaging (GU) used in Zhai et al. (2013), previous studies mainly characterized CREB activation in response to global, extracellular field stimulation or bulk, chemical depolarization of neuronal populations. While foundational to our understanding of CREB regulation, population-based approaches are limited in their ability to directly correlate neuronal stimulus history with CREB-dependent transcription. To overcome these limitations, here we employed a single-neuron, optical stimulation approach to generate different patterns of neuronal activity while simultaneously observing both the spatiotemporal dynamics of the resulting Ca2+ signals and the output of a live-cell reporter of CREB-dependent transcription to determine which of the various proposed E–T coupling pathways are engaged.
Materials and Methods
Culture and transfection of hippocampal neurons
Hippocampi were isolated and dissociated using papain from Postnatal Day (P)0–P1 Sprague Dawley rats, both male and female without determination of sex, seeded on 1.0 cover glass precoated with poly-d-lysine at 125,000 cells per 18 mm coverslip and 350,000 cells per 25 mm coverslips in MEM (Invitrogen) supplemented with FBS, Pen/Strep, and l-glutamine, maintained at 37°C with 5% CO2. Neurons were grown in Neurobasal-A (NBA) supplemented with B27, GlutaMAX, and Pen/Strep starting at DIV 1, fed with half media exchanges approximately every 5 d and supplemented with antimitotics (uridine + fluoro-deoxyuridine) starting at DIV 5. Rat hippocampal neurons were transfected 48 h prior to imaging on DIV 13–15 using Lipofectamine 2000 (Invitrogen) and NBA without supplements. A 0.5–4 µg DNA was mixed with Lipofectamine 2000, 3 µl for 18 mm coverslips, and 5 µl in the case of 25 mm coverslips for 20 min prior to being added to cells for 90 min. Transfection media solution was then removed and returned to 50% conditioned media and 50% fresh supplemented NBA.
Plasmid DNA constructs
In order to construct an optimized CREB transcriptional reporter, the pGL4.29 cAMP-response element (CRE) hydro luciferase reporter was obtained, and the luciferase sequence was removed and replaced with a 2xsGFPNLS reporter from the previously generated nuclear factor of activated T cell (NFAT)-AP1 reporter (Wild et al., 2019). To this end, a BsiWI cut site was introduced to the C terminus of the pGL4.29 plasmid, and an EcoRI site was introduced to the C terminus of the sGFPx2 insert so that following PCR using NEB Phusion DNA polymerase, these EcoRI and BsiWI sites could be utilized to ligate PCR-generated plasmid fragments using NEB T4 ligase (see Table 1 for oligonucleotide primer sequences used in this cloning). The CREmut reporter was generated by introducing three subsequent mutagenesis reactions (see Table 1 for mutagenesis primer sequences), using NEB Phusion PCR reaction protocol to replace the CRE motif with a scrambled sequence not recognizable by any transcription factors, verified using JASPAR (Castro-Mondragon et al., 2022). MEK-Turq2 was generated from MEK1-GFP (Addgene plasmid 14746) using standard cloning techniques to transfer the MEK1 cDNA sequence into a mTurq2-N1 vector driven by a CMV promoter. YFP-ERK2 was cloned from GFP-ERK2 (Addgene plasmid 37145) into a sYFP-C1 vector containing a CMV promoter. The Ca2+ indicator jRGECO1a (Dana et al., 2016) was previously cloned into a pCAG vector backbone as described in Wild et al. (2019). See Table 2 for the source of all plasmids and other reagents used in this study.
Oligonucleotide primer sequences
Key reagents and resources
Confocal microscopy and MNI GU
Live imaging was performed on a 3i-Marianas/Zeiss Observer.Z1 inverted spinning disk confocal microscope as described previously (Wild et al., 2019), with a 40× or 63× oil immersion objective lens (1.4 NA). Data acquisition was performed with the SlideBook 6.0 software (Intelligent Imaging Innovations) and ImageJ software used for data analysis. Neurons were imaged in Mg2+-free Tyrode’s imaging solution pH 7.4 (135 mM NaCl, 5 mM KCl, 25 mM HEPES, 10 mM glucose) containing 3 mM CaCl2, 30 µM glycine, and 1 µM tetrodotoxin (TTX) along with experimental pharmacological inhibitors where applicable. Neurons were kept in the dark and incubated at 34°C during imaging. A 4-methoxy-7-nitroindolinly (MNI)-caged L-glutamate (Tocris Bioscience) was added to the neurons to a final concentration of 1.3 mM and 5 min of incubation to allow for diffusion of MNI throughout the chamber; an ROI of 1 × 1 µm was selected adjacent to a dendritic spine ∼150–175 µm from the soma. MNI glutamate was uncaged using a 405 nm laser with 1 ms dwell time (50–100 µW) with pulses at either 0.5 or 1 Hz as indicated.
Live-cell imaging of Ca2+ during GU
All experiments were performed using the spinning disk confocal microscope as described above with the CRE-GFP reporter imaged using 488 excitation, 525/50 emission, and jRGECO1a imaged with the 561 nm excitation, 617/73 emission. To define ROIs for measurement of integrated Ca2+ signals during GU, a single z-plane with the stimulated dendrite and soma both in focus was imaged using 561 nm excitation at 10 Hz with 40 ms exposure times. To calculate integrated ΔF/Fo Ca2+ signals, ROIs were defined within the stimulated dendrite, the soma, and a background area outside the neuron. The background was then subtracted from both the dendrite and soma values, the Fo value was calculated by finding a minimum average over 10 image frames, and ΔF/F0 was calculated. The integrated Ca2+ signals for each ROI were calculated using the summation of the entire time-lapse series.
Live-cell imaging of CRE-GFP reporter expression
Prior to CRE-GFP reporter experiments, TTX was added to culture media for 3 h to silence spontaneous neuronal activity. For CRE-GFP reporter imaging, 20 xy images were collected, centered around the nucleus, at step sizes of 0.5 µm to create a z-stack of 10 µm, using 488 nm excitation with 100 ms exposure times and 561 nm excitation with 40 ms exposure time. Z-stacks were collected prior to GU, immediately following GU, and then every 30 min following GU up to 120 min. To calculate reporter expression, the z-stack at the center of the nucleus was selected and intensity averaged with two surrounding z-planes, an ROI was drawn in the nucleus, identified by the lack of jRGECO1a expression as jRGECO1a is excluded from the nucleus due to the presence of an NES. Nuclear background subtracted values were normalized to averaged poststimulation background subtracted nuclear values.
Live-cell imaging of ERK translocation
Neurons were plated on 25 mm coverslips as described previously, at DIV 13–15; 48 h prior to imaging, cells were transfected as described previously with plasmids encoding MEK1-mTurq2 0.75 µg, YFP-ERK2 0.75 µg, and jRGECO1a 0.5 µg. Prior to ERK translocation assays, TTX was added to culture media 2 h prior to silence spontaneous neuronal activity. MEK1-mTurq2 was imaged using the 445 nm excitation and at 50 ms exposures, YFP-ERK2 was imaged using 515 nm excitation with 50 ms exposures, and jRGECO1a imaged using 561 nm excitation and 40 ms exposures. Prior to GU, 20 xy z-stack images were collected centered around the nucleus, and post-GU stimulation z-stack images were taken every 10 min. To quantify ERK and MEK localization, the z-plane in the center of the nucleus was selected, and the intensity was averaged with two surrounding z-planes. Using the absence of jRECO1a in the nucleus as a marker, ROIs were selected in the nucleus, cytoplasm, and background regions. MEK cytosolic expression and ERK cytosolic and nuclear expression were quantified prior to GU to ensure comparable expression levels and initial ERK localization. ERK nuclear localization was calculated by normalizing nuclear intensity at 10, 20, and 30 min post-GU to nuclear intensity immediately following GU.
Imaging spine size following 1 Hz GU
Neurons plated on 18 mm coverslips as described previously were transiently transfected with plasmids encoding iRFP702 1.5 µg, jRGECO1a 0.5 µg, and ER-moxGFP 0.75 µg 48 h prior to imaging. Cells were then imaged in Mg2+-free Tyrode’s imaging solution as described previously with or without 1 µM TatCN19o added. Dendritic spines were imaged using the 63× objective with 0.2 µm z-stack steps, using 640 nm excitation to image cytosolic fill iRFP702 and 488 nm excitation to image ER-moxGFP to identify ER expressing spines. Our group recently found that ER containing spines more reliably undergo sustained structural LTP (sLTP) following GU (Dittmer and Dell’Acqua, 2024); therefore only ER expressing spines were selected for GU uncaging and no stimulation controls to reduce the possibility of nonresponding spines. Spines were stimulated with a 1 min 1 Hz GU train, an established protocol to induce sLTP (Dittmer and Dell’Acqua, 2024), along with simultaneous imaging of Ca2+ using jRGECO1a to calculate integrated Ca2+ signal as described above. Z-stacks of iRFP filled spines were then acquired every 10 min following GU. Images were deconvoluted to the nearest neighbor using the SlideBook software and maximum projections used to make masks to calculate the spine head area.
Epifluorescence microscopy and quantification of phospho-CREB and c-Fos expression following GU
To detect pCREB and c-Fos expression following GU, neurons were plated on 18 mm coverslips, transfected with jREGECO1a at DIV 13–15 48 h prior to imaging and treated with TTX 2 h prior to GU. As above, single neurons were stimulated with GU according to experimental conditions described in the text and figure legends. For pCREB staining, neurons were fixed either immediately following GU stimulation or after 15 min in 4% PFA, 4% sucrose solution in PBS, and 50 mM HEPES solution at pH 7.4 for 10 min. In the case of c-Fos expression, after GU stimulation, cells were returned to culture media containing the same pharmacological agents present during stimulation for 2 h at 37°F and then fixed. For both pCREB and c-Fos staining, cells were permeabilized in 0.1% Triton X-100 in PBS for 10 min at RT. Cells were then blocked in 4% BSA solution in PBS for 2 h at RT, with 0.1% Triton X-100 also present in the case of c-Fos. Coverslips were then incubated with either rabbit anti-CREB phospho-Ser133 (1:500; Cell Signaling Technology) for 2 h in BSA blocking solution or rabbit anti-c-Fos (1:250; Synaptic Systems) in respective BSA solutions. Following 3 × 10 min washes in PBS on a rocker, coverslips were incubated in Alexa Flour 488 goat, anti-rabbit antibodies for 1 h. Coverslips were washed an additional 3× for 10 min in PBS and mounted on slides using ProLong Gold with DAPI (Invitrogen) to stain nuclei.
Cells were imaged on a Zeiss Axiovert 200 M inverted microscope with a Zeiss 40× objective in a similar manner to Martinez et al. (2024). Stimulated neurons were located using the Cy3/Texas Red filter to identify the jRGECO1a expressing neuron previously stimulated based on morphology. The 20 xy images were taken to obtain z-stacks centered around the nucleus at 0.5 µm steps using the DAPI, FITC/Alexa Fluor 488, and Cy3/Texas Red filter sets. pCREB staining was imaged using 200 ms exposures, and c-Fos staining was imaged using 100 ms exposures with the FITC/Alexa Fluor 488 filter set. pCREB and c-Fos nuclear expression was determined following z-stack deconvolution using “nearest-neighbor” algorithm in Slidebook 6.0, and the center z-plane of the nucleus was selected and then averaged with the two surrounding z-planes. The average intensity of the nucleus was quantified by drawing a ROI using the outline created by both the absence of jRGECO1a signal and the nuclear staining of DAPI. Background subtracted raw values were quantified in arbitrary fluorescence units (AFUs).
Experimental design and statistical analysis
All statistical analysis was performed using the Prism 10.0 (GraphPad) software. All data are normally distributed and shown as mean ± SEM unless otherwise indicated in figure legends. Normally distributed datasets were analyzed using Student’s t test with Welch’s posttest, ANOVA repeated measures with mixed-effect analysis for cells imaged over time and ANOVA with Kruskal–Wallis posttest for multiple comparisons where applicable. Non-normally distributed datasets are shown as median and interquartile range (IQR) as indicated and compared using nonparametric Mann–Whitney ranked sum test with median and IQRs graphed. Significance is reported as ∗p < 0.05, and data are expressed and graphed as mean ± SEM. Actual p values are reported when provided by the software. When actual p values are not provided by the software, only ∗p < 0.05, ∗∗p < 0.01, ∗∗p < 0.001, and ∗∗∗∗p < 0.0001 are reported.
Results
CREB-dependent transcription requires Ca2+ signal propagation to the soma
To investigate the somatodendritic Ca2+ sources and neuronal stimulation patterns that effectively engage CREB-dependent transcription, we utilized LTP-like GU protocols in dissociated rat hippocampal neurons to manipulate postsynaptic input to dendrites through direct activation of NMDARs in the presence of TTX to prevent recurrent network activity triggered by action potential firing (Matsuzaki et al., 2004; Zhai et al., 2013; Dittmer et al., 2017, 2019; Wild et al., 2019). Neurons were transiently cotransfected with the red fluorescent, genetically encoded Ca2+ indicator jRGECO1a to characterize Ca2+ dynamics and a new live-cell reporter we developed to allow direct observation of CREB-dependent transcription. To this end, an existing CREB-dependent luciferase reporter containing three CRE was modified to instead express a nuclear-localized, tandem SGFP2 reporter in response to CREB binding and transcriptional activation (CRE-2xsGFPnls; Fig. 1A). PEST degradation motifs were also incorporated to promote GFP turnover and limit basal reporter accumulation. Following a 3 h pretreatment with TTX to silence spontaneous neuronal activity and decrease basal GFP reporter expression, MNI glutamate was photo-uncaged using repetitive 405 nm laser excitation sequentially at six dendritic spines located on different dendritic branches (Fig. 1A), similar to as in some previous studies of E–T coupling (Zhai et al., 2013; Wild et al., 2019). The Ca2+ signals resulting from GU stimulation were imaged (Fig. 1B) and quantified (Fig. 1C), and then the nuclear GFP signal generated by the CRE-2sxGFPnls reporter was monitored in real time for 2 h after stimulation (Fig. 1D). In particular, we employed GU LTP induction protocols characterized in our previous studies of NFAT E–T coupling to either generate somatically propagating Ca2+ signals that closely resemble NMDAR-initiated dendritic plateau potentials/complex spikes using a 1 Hz GU train or local, dendritically restricted, NMDAR-mediated Ca2+ transients using a 0.5 Hz GU train (Fig. 1B; Wild et al., 2019). As observed in our prior work, these two NMDAR stimulation protocols generated comparable total, integrated dendritic Ca2+ signals, but only 1 Hz GU reliably triggered Ca2+ transient/spike propagation to the soma (Fig. 1C).
CREB-dependent transcriptional activation requires Ca2+ propagation to the soma. A, Experimental timeline for CRE reporter experiments (top), distribution of GU over six branches (bottom left), CRE reporter design (bottom right). B, Right, Representative images of jRGECO1a, pseudocolored, before and during either 1 Hz propagating (top) versus 0.5 Hz nonpropagating GU stimulation (bottom); scale bar, 20 µm. Left, Traces of jRGECO1a ΔF/F0 over time for a single dendritic spine/branch (site of GU stimulation indicated with +) during either 1 Hz (top) or 0.5 Hz (bottom) GU stimulation. See also Extended Data Movies 1-1 and 2-1. C, Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CRE reporter experiments in Figure 1D (1 Hz, n = 18 over 7 biological replicates, dendritic Ca2+ 7,781 ± 801, somatic Ca2+ 3,391 ± 311; 0.5 Hz, n = 12 over 8 biological replicates, dendritic Ca2+ 7,388 ± 977, somatic Ca2+ 222 ± 23; ns, p = 0.76; ****p < 0.0001 Student’s t test with Welch’s test). D, Left, CRE reporter expression following GU stimulation [1 Hz GU (black solid) and 0.5 Hz (blue solid), no stimulation/NS (dashed black)]. NS, n = 23 over 10 biological replicates, 0.89 ± 0.025; 1 Hz, n = 18 over 7 biological replicates, 1.52 ± 0.050; 0.5 Hz, n = 12 over 8 biological replicates, 0.93 ± 0.038; 1 Hz versus 0.5 Hz ***p = 0.0009, NS control versus 1 Hz ****p < 0.0001 ANOVA repeated measures with mixed-effect analysis. Right, Representative images of CRE reporter expression, pseudocolored with nucleus outlined, 0 and 90 min following 1 Hz stimulation (inset); scale bar, 10 µm. E, Left, CREmut reporter expression following somatically propagating 1 Hz GU (gray) versus 1 Hz CRE reporter expression as shown in Figure 1D. CREmut no stimulation/NS (dashed gray). CREmut 1 Hz, n = 11 over five biological replicates, 0.97 ± 0.047; CREmut NS, n = 10 over five biological replicates, 0.86 ± 0.057; CREmut 1 Hz versus CRE 1 Hz **p = 0.0014 ANOVA repeated measures with mixed-effect analysis. Right, CRE reporter expression following GU stimulation in the presence of 5 μg/mL actinomycin D (solid pink) or 1 Hz no actinomycin D (solid black) and no GU stimulation in the presence of actinomycin D (dashed pink) or no drug (dashed black) over 1 h post-GU; **p = 0.00466 ANOVA repeated measures with mixed-effect analysis; 1 Hz no drug, n = 6 over three biological replicates 1.83 ± 0.018; ns no drug, n = 7 over three biological replicates 0.95 ± 0.031; 1 Hz + actinomycin D n = 8 over three biological replicates 1.145 ± 0.008; ns + actinomycin D, n = 8 over three biological replicates 1.03 ± 0.013. F, Left, pCREB expression immediately following 1 Hz GU without or with caged MNI glutamate and 0.5 Hz dendrite-restricted stimulation, raw intensities shown (1 Hz no MNI, n = 13 over 8 biological replicates, 1,364 ± 234; 1 Hz with MNI, n = 22 over 9 biological replicates, 3,808 ± 411; 0.5 Hz, n = 8 over 5 biological replicates, 1,213 ± 350; ****p < 0.0001 t test with Welsh’s t test). Right, Representative images of stimulated cells identified poststaining by their morphology visible through jRGECO1a imaging (grayscale left) and nuclear pCREB staining (pseudocolored right) following either 1 Hz somatically propagating (top) or 0.5 Hz distally restricted stimulation (bottom). Nuclei identified by nuclear exclusion of jRGECO1a and DAPI staining shown in outlines, scale bar = 20 µm. G, Left: c-Fos expression 2 h following 1 Hz stimulation. One hertz GU without or with caged MNI glutamate and 0.5 Hz distally restricted stimulation, raw intensities, median, and IQR are graphed (1 Hz no MNI, n = 11 over 9 biological replicates, 460, 351-731; 1 Hz with MNI: n = 17 over 10 biological replicates, 1277, 731-1825; 0.5 Hz: n = 9 over 6 biological replicates, 355, 258-534; **p = 0.0024; ****p < 0.0001). Mann–Whitney ranked sum. Right, Representative images of stimulated cells identified poststaining by their morphology visible through jRGECO1a imaging (grayscale left) and nuclear c-Fos staining (pseudocolored right) following either 1 Hz somatically propagating (top) or 0.5 Hz distally restricted stimulation (bottom). Nuclei identified by nuclear exclusion of jRGECO1a and DAPI staining shown in outlines; scale bar, 20 µm. See also Extended Data Figure 1-1 related to Figure 1.
Figure 1-1
A. Left: Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CREmut experiments shown in Fig. 1E vs 1 Hz described in Fig 1C CREmut: dendritic Ca2+ 16656 ± 1692, somatic Ca2+ 6086 ± 694; ***p = 0.0003, **p = 0.0032, t-test with Welch’s test. Right: Quantification of dendritic and somatic Ca2+ signals quantified as AUC of jRGECO1a ΔF/F0 over all stimulated branches during 1 Hz vs 1 Hz + ActD experiments. 1Hz: dendritic Ca2+ 6821 ± 1919, somatic Ca2+ 5847 ± 932.3; 1 Hz + Act D: dendritic somatic Ca2+ 11908 ± 3269, somatic Ca2+ 4426 ± 876.5; t-test with Welch’s test B. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during pCREB experiments shown in Fig. 1F. 1 Hz no MNI: dendritic Ca2+ 1633 ± 235, somatic Ca2+ 334 ± 38; 1 Hz with MNI: dendritic Ca2+ 11105 ± 694.7, somatic Ca2+ 4920 ± 412; 0.5 Hz: dendritic Ca2+ 15745 ± 1377, somatic Ca2+ 656 ± 92; ****p < 0.0001, *p = 0.012, t-test with Welch’s test. C. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during c-Fos experiments shown in Fig. 1G. 1 Hz no MNI: dendritic Ca2+ 1819 ± 279, somatic Ca2+ 317 ± 27; 1 Hz with MNI: dendritic Ca2+ 13256 ± 853, somatic Ca2+ 5674 ± 359; 0.5 Hz: dendritic Ca2+ 14382 ± 1087, somatic Ca2+ 831 ± 153; ns p = 0.43, ****p < 0.0001, t-test with Welch’s test. Download Figure 1-1, TIF file.
Video of Ca2+ responses to 1 Hz stimulation as depicted in Fig 1B. Time 10x speed.
Video of Ca2+ responses to 0.5 Hz stimulation as depicted in Fig 1B. Time 10x speed.
Concomitant observation of CRE-GFP reporter expression revealed that only the somatically propagating Ca2+ stimuli generated by 1 Hz GU promoted CREB-dependent transcription (Fig. 1D), with CRE-GFP reporter expression being significantly induced by 1 Hz GU stimulation when compared with both 0.5 Hz GU stimulation and control conditions in the absence of stimulation, which both failed to increase CRE-GFP reporter expression over time (Fig. 1D). As additional controls, we firstly mutated the consensus CRE binding sites in the reporter and observed that this CREmut-2xGFPnls reporter was unresponsive to 1 Hz GU stimulation, and secondly, we showed that 1 Hz GU stimulated CRE reporter expression was also inhibited by the RNA polymerase inhibitor actinomycin D (Fig. 1B), which had no impact on somatic Ca2+ signal propagation (Extended Data Fig. 1-1A). Together these controls confirm the dependence of CRE-GFP reporter expression on CREB binding and transcriptional activation, respectively. Overall, these results indicate that CREB-dependent reporter transcription is rapidly initiated within <30 min after dendritic NMDAR stimulation but only if this stimulation results in somatic Ca2+ propagation.
Based on previous studies, CREB activation in response to dendrite-restricted, NMDAR-mediated Ca2+ stimuli might be expected to be delayed by 60 min or more in comparison with soma-propagating Ca2+ stimuli due to the time required for ERK activated locally in dendrites to translocate to the nucleus (Wiegert et al., 2007; Karpova et al., 2013; Zhai et al., 2013). However, even 90–120 min following 0.5 Hz GU, we observed no significant induction of CRE-GFP reporter activity (Fig. 1D). Likewise, in parallel experiments employing 1 Hz GU to induce dendrite-to-soma Ca2+ propagation (Extended Data Fig. 1-1B), we observed a rapid increase in CREB Ser133 phosphorylation detected by post hoc immunostaining when compared with mock-stimulation control conditions where 1 Hz laser excitation was delivered in the absence of MNI glutamate (Fig. 1F). In contrast, 0.5 Hz GU-induced dendrite–restricted Ca2+ signals (Extended Data Fig. 1-1B) failed to significantly increase pCREB in the nucleus above control levels (Fig. 1F). Finally, 1 Hz soma-propagating, but not 0.5 Hz dendrite-restricted, Ca2+ stimuli significantly induced c-Fos expression detected by immunostaining 2 h post-GU compared with mock-stimulated controls (Fig. 1G; Extended Data Fig. 1-1C). Thus, both upstream CREB phosphorylation and downstream CREB transcriptional activity induced by dendritic NMDAR stimulation, whether measured using our exogenous live-cell CRE-GFP reporter or endogenous c-Fos staining, corresponded with Ca2+ signal propagation to the soma but not with dendritic Ca2+ levels.
CREB-dependent transcription requires Ca2+ signal propagation to the soma carried by LTCCs
These findings for somatic Ca2+ driving CREB-dependent transcription conflict with some previous studies of CREB E–T coupling using GU stimulation of dendrites (Zhai et al., 2013) but are consistent with others using antidromic electrical stimulation to drive somatic action potential firing and LTCC opening (Dudek and Fields, 2002). In addition, our present findings for CREB are reminiscent of our prior findings using similar experimental approaches showing that NFAT E–T coupling in response to dendritic NMDAR stimulation required Ca2+ signal propagation to the soma (Wild et al., 2019). In these prior studies of NFAT, we found that LTCCs were responsible for dendrite-to-soma Ca2+ spike propagation downstream of NMDAR activation by GU even in the absence of action potential firing and independent of Ca2+ release from the ER (Wild et al., 2019). Notably, LTCCs are implicated in CREB E–T coupling and IEG regulation by a number studies, but the spatiotemporal requirements of LTCC Ca2+ signaling to CREB are still unclear (Murphy et al., 1991; Bading et al., 1993; Impey et al., 1996; Mermelstein et al., 2000; Dolmetsch et al., 2001; Dudek and Fields, 2002; Wheeler et al., 2008, 2012; Ma et al., 2014; Li et al., 2016; Wang et al., 2017; Servili et al., 2019; Ma et al., 2023). Accordingly, as in our previous work on NFAT, we confirmed here that somatic Ca2+ propagation elicited using 1 Hz GU was inhibited in the presence of the LTCC antagonist nimodipine, with no negative impacts on the magnitude of the resulting total, integrated dendritic Ca2+ signals (Fig. 2A,B). However, nimodipine fully prevented induction of CRE-GFP reporter expression following 1 Hz GU (Fig. 1B), once again demonstrating the necessity of LTCCs and somatic Ca2+ propagation for CREB-dependent transcription and indicating that the lack of CRE-GFP reporter induction in response to 0.5 Hz GU is related to the lack of Ca2+ propagation to the soma and not to the reduced frequency or intensity of dendritic NMDAR stimulation.
CREB-dependent transcription requires Ca2+ signal propagation to the soma carried by the LTCCs. A, Representative images of jRGECO1a, pseudocolored, before and during GU (indicated with +) 1 Hz + 5 µM nimodipine stimulation and traces of jRGECO1a ΔF/F0 over stimulation over a single branch during 1 Hz + 5 µM nimodipine stimulation (right); scale bar, 20 µm. See also Extended Data Movie 3-1. B, Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CRE reporter experiments in Figure 2C with 1 Hz as shown in Figure 1B; 1 Hz + Nim, n = 13 over six biological replicates, dendritic Ca2+ 12,024 ± 1296, somatic Ca2+ 564 ± 104; *p = 0.011; ****p < 0.0001 t test with Welch’s test. C, CRE reporter expression following 1 Hz GU + 5 µM nimodipine (solid red) versus 1 Hz GU (solid black) as shown in Figure 1D, no stimulation/NS with 5 µM nimodipine (dashed red). NS + Nim, n = 24 over six biological replicates, 0.791 ± 0.039; 1 Hz + Nim, n = 13 over six biological replicates, 0.831 ± 0.029; ****p < 0.0001 ANOVA repeated measures with mixed-effect analysis. D, Left, Quantification of dendritic and somatic Ca2+ influx quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during 1 Hz + Cd2+ experiments compared with 1 Hz stimuli as shown in Figure 1C (1 Hz + Cd2+ dendritic Ca2+ 16,076 ± 1202; somatic Ca2+ 841 ± 142; ****p < 0.0001; t test with Welch’s test). Right, CRE reporter expression following 1 Hz GU + 20–50 µM Cd2+ (solid purple) versus 1 Hz GU (solid black) as shown in Figure 1D, no stimulation with 20–50 µM Cd2+ (dashed purple; 1 Hz + Cd2+, n = 12 over 4 biological replicates, CRE 0.94 ± 0.12; Ns + Cd2+, n = 7 over 3 biological replicates, CRE 0.93 ± 0.025; 1 Hz vs 1 Hz + Cd2+; **p = 0.0024 ANOVA repeated measures with mixed-effect analysis). E, Left, pCREB expression immediately following 1 Hz GU + nimodipine and 1 Hz as shown in Figure 1F, raw intensities graphed in AFU (1 Hz + Nim: n = 10 over 6 biological replicates, 1,383 ± 293; ****p < 0.0001 t test with Welsh’s test). Right, Representative images of stimulated cells identified poststaining by their morphology visible through jRGECO1a imaging (grayscale left) and nuclear pCREB staining (pseudocolored right) following either 1 Hz somatically propagating (top) or 1 Hz stimulation + 5 µM nimodipine (bottom). Nuclei identified by nuclear exclusion of jRGECO1a and DAPI staining shown in outlines; scale bar, 20 µm. F, Left, c-Fos expression 2 h following 1 Hz stimulation + nimodipine versus 1 Hz GU as in Figure 1G, raw intensities in AFU, median and IQR are graphed (1 Hz + Nim, n = 10 over 5 biological replicates, 634, 358–914; **p = 0.0067 Mann–Whitney ranked sum). Right, Representative images of stimulated cells identified poststaining by their morphology visible through jRGECO1a imaging (grayscale, left) and nuclear c-Fos staining (pseudocolored, right) following either 1 Hz somatically propagating (top) or 1 Hz + nimodipine (bottom). Nuclei identified by nuclear exclusion of jRGECO1a and DAPI staining shown in outlines; scale bar, 20 µm. See also Extended Data Figure 2-1 related to Figure 2.
Figure 2-1
A. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during 1 Hz + Nim pCREB experiments shown in Fig 2E and 1 Hz as in 1F. 1 Hz + Nim: dendritic Ca2+ 15853 ± 1014, somatic Ca2+ 564 ± 53. ****p < 0.0001, **p = 0.0012, t-test with Welch’s test. B. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during c-Fos experiments shown in Fig 2F and 1 Hz as in Fig 1G. 1 Hz + Nim: dendritic Ca2+ 15144 ± 1774, somatic Ca2+ 619 ± 89. ns p = 0.36, ****p < 0.0001, t-test with Welch’s test. Download Figure 2-1, TIF file.
Video of Ca2+ responses to 1 Hz stimulation in 5 μM nimodipine as depicted in screen shots in Fig 2A. Time 10x speed.
Prior studies indicate that CaV1 subunit conformational changes elicited by the voltage-dependent activation of LTCCs may be sufficient for CREB activation when coupled with another source of Ca2+ influx, such as that provided by NMDARs (Li et al., 2016). Nimodipine blocks both CaV1 LTCC voltage-gated conformation changes and Ca2+ influx, while addition of Cd2+ allows the blocking of Ca2+ influx without inhibiting conformational changes. Even in the presence of nimodipine, we previously demonstrated that 1 Hz GU stimulation induces a prolonged NMDAR-mediated depolarization that can be measured at the soma; however, the fast voltage spikes corresponding to somatic Ca2+ signals generated by LTCCs are inhibited by nimodipine (Wild et al., 2019). Thus, in the absence of nimodipine but in the presence of Cd2+ instead, LTCCs in both the dendrites and the soma would be able to undergo conformational changes in response to NMDAR-mediated depolarization but not produce Ca2+ signals. However, under these conditions, we show that Cd2+ not only prevented 1 Hz GU-induced Ca2+ propagation to the soma as expected (Fig. 2D) but also blocked the induction of CRE-GFP reporter activity (Fig. 2D), once again reaffirming a dependence on somatic Ca2+. In addition, based on the similar effects of Cd2+ and nimodipine, the ability of nimodipine to prevent CRE-GFP reporter induction cannot be attributed solely to its ability to block LTCC conformational signaling. Corroborating this CRE-GFP reporter data, in parallel experiments both CREB phosphorylation (Fig. 2E) and c-Fos expression (Fig. 2F) were also inhibited by nimodipine inhibition of LTCCs that prevented somatic Ca2+ propagation (Extended Data Fig. 2-1A,B).
ERK activation is necessary for CREB-dependent transcription in response to dendrite-to-soma Ca2+ propagation
Several prior studies have implicated the Ras→Raf→MEK→ERK MAP kinase pathway in CREB E–T coupling (Rosen et al., 1994; Impey et al., 1998; Dolmetsch et al., 2001; Tyssowski et al., 2018). ERK-dependent CREB activation has been reported both downstream of LTCC activation (Dolmetsch et al., 2001; Wu et al., 2001a; Impey et al., 2002; Servili et al., 2018, 2019), including somatic LTCC Ca2+ influx (Dudek and Fields, 2002), and alternatively downstream of NMDAR Ca2+ influx restricted to dendrites, independent of LTCC Ca2+ influx (Zhai et al., 2013). Previous studies have also found conflicting effects of MEK–ERK inhibition on CREB phosphorylation and transcription. Phosphorylation of CREB following KCl depolarization has been observed in the presence of MEK inhibitors PD98059 (Ma et al., 2014) or U0126 (Wheeler et al., 2008) in several studies implicating CaMKII→CaMKIV signaling to CREB. Yet in other studies, PD98059 inhibited CREB phosphorylation (Wu et al., 2001a,b) and c-Fos expression (Zhai et al., 2013; Cohen et al., 2018). Thus, despite considerable attention, two major questions remain unanswered: (1) Under what conditions is ERK and/or CaMK activity necessary for CREB transcriptional activity? And (2) under these conditions does ERK and/or CaMK signaling to CREB originate in the dendrites or soma or possibly both?
To investigate the contribution of ERK signaling to CREB-dependent transcription, here we utilized the MEK inhibitor PD98059 to assess impacts on 1 Hz GU induction of CRE-GFP reporter activity, CREB phosphorylation, and c-Fos expression. Importantly, we found PD98059 inhibition of MEK→ERK signaling did not alter 1 Hz GU-induced Ca2+ signals either in dendrites or the soma in any of our experiments (Fig. 3A; Extended Data Fig. 3-1A,B). However, PD98059 caused a significant reduction in 1 Hz GU induction of CRE-GFP reporter activity (Fig. 3B). Interestingly, CREB Ser133 phosphorylation measured immediately following 1 Hz GU stimulation was not significantly inhibited by PD98059 (Fig. 3C); however, given the CRE-GFP reporter results, we further investigated this potential discrepancy between the ERK-dependency of downstream CREB transcriptional control and upstream CREB phosphorylation. Previous studies reported a rapid, initial phase of CREB phosphorylation immediately following brief KCl depolarization that is ERK-independent and then a second, more sustained phase of CREB phosphorylation 15–30 min later that is ERK-dependent (Impey et al., 1998, 2002; Wu et al., 2001b). Given these earlier studies indicating PD98059 inhibition of CREB phosphorylation may not manifest until 15–30 min post-KCl stimulation and that our sequential 1 Hz GU uncaging stimulation of six dendritic spine takes ∼10–12 min to complete, we independently assessed CREB Ser133 phosphorylation 15 min after 1 Hz GU stimulation and found that PD98059 now significantly impaired pCREB regulation (Fig. 3C). Furthermore, c-Fos induction following 1 Hz GU was also inhibited by PD98059 (Fig. 3D). Together these data show that while rapid, initial CREB phosphorylation following 1 Hz GU stimulation may be largely ERK-independent, more sustained pCREB expression as well as downstream CREB-dependent transcription is ERK-dependent. These results are also consistent with early reports that transient CREB phosphorylation following neuronal stimulation was not sufficient for transcriptional activity that instead requires prolonged pCREB elevation (Bito et al., 1996; Wu et al., 2001b; Moosmang et al., 2005).
Activation of ERK is necessary for CREB-dependent transcription in response to dendrite-to soma Ca2+ propagation. A, Left, Traces of jRGECO1a ΔF/F0 over stimulation over a single branch during 1 Hz + 25 µM PD98059 (left). Right, Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during 1 Hz + PD98059 experiments and 1 Hz as shown in Figure 1B (1 Hz + PD98059, n = 12 over 6 biological replicates, dendritic Ca2+ 9,855 ± 1549, somatic Ca2+ 4,019 ± 620). Dendritic Ca2+ ns, p = 0.25; somatic Ca2+ ns, p = 0.38; t test with Welch’s test. B, CRE reporter expression following 1 Hz GU + 25 µM PD98059 (solid teal), no stimulation + PD (dashed teal) versus 1 Hz GU (solid black) as shown in Figure 1D. Ns + PD, n = 9 over 5 biological replicates, 0.842 ± 0.036; 1 Hz + PD, n = 12 over 6 biological replicates, 1.00 ± 0.041. **p = 0.0012 ANOVA repeated measures with mixed-effect analysis. C, pCREB expression immediately following 1 Hz GU, as shown in Figure 1F, and 1 Hz GU + 25 µM PD98059, and 15 min following 1 Hz no MNI, 1 Hz GU, and 1 Hz GU + 25 µM PD98059, raw intensities shown (1 Hz GU + 25 µM PD98059, n = 11 over 5 biological replicates, 3,509 ± 382; 1 Hz no MNI 15 min post stim, n = 8 over 7 biological replicates, 910 ± 221; 1 Hz GU 15 min post stim, n = 10 over 7 biological replicates, 1,989 ± 422; 1 Hz GU + 25 µM PD98059 15 min post stim, n = 12 over 7 biological replicates, 978 ± 147). NS, p = 0.60, post 15 min no MNI versus 1 Hz *p = 0.041, post 15 min 1 Hz versus 1 Hz + PD *p = 0.044, t test with Welsh’s test. D, c-Fos expression 2 h following 1 Hz GU stimulation + 25 µM PD98059 versus 1 Hz GU as shown in Figure 1G, raw intensities, median, and IQR graphed (1 Hz + PD98059, n = 12 over 8 biological replicates, 692, 377-1157; *p = 0.024 Mann–Whitney ranked sum). E, YFP-ERK2 nuclear intensity normalized to poststimulation. ERK nuclear localization following 1 Hz somatically propagating stimulation (black), no stimulation/NS (dashed gray), 0.5 Hz dendrite-restricted stimulation (solid blue), 1 Hz + nimodipine (solid red) no stimulation/NS + nimodipine (dashed red), 1 Hz + PD98059 (solid teal), and NS + PD98059 (dashed teal) (1 Hz, n = 12 over 5 biological replicates, 1.44 ± 0.011; Ns, n = 20 over 7 biological replicates, 1.08 ± 0.028; 0.5 Hz, n = 7 over 4 biological replicates, 1.03 ± 0.014; 1 Hz + Nim, n = 9 over 5 biological replicates, 1.10 ± 0.012; NS+ Nim, n = 12 over 5 biological replicates, 1.07 ± 0.016; 1 Hz + PD, n = 7 over 3 biological replicates, 1.08 ± 0.01; NS + PD, n = 7 over 3 biological replicates, 1.09 ± 0.011; 1 Hz vs no stim ***p = 0.0003; 1 Hz vs 0.5 Hz *p = 0.0103; 1 Hz vs 1 Hz + Nim *p = 0.012; 1 Hz vs 1 Hz + PD *p = 0.021; ANOVA repeated measures with mixed-effect analysis). Inset YFP-ERK expression immediately and 10 min following GU either 1 Hz (left) or 0.5 Hz (right); scale bar, 10 µm. See also Extended Data Figure 3-1 related to Figure 3.
Figure 3-1
A. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during pCREB experiments shown in Fig. 3C. 1 Hz GU + 25 μM PD98059: dendritic Ca2+ 10918 ± 1188, somatic Ca2+ 4455 ± 339; 1 Hz no MNI 15 min post stim: dendritic Ca2+ 1942 ± 227, somatic Ca2+ 318 ± 69; 1 Hz GU 15 min post stim: dendritic Ca2+ 12574 ± 1447, somatic Ca2+ 5328 ± 674; 1 Hz GU + 25 μM PD98059 15 min post stim: dendritic Ca2+ 13457 ± 2141, somatic Ca2+ 5742 ± 409. t = 0 dendritic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.89; t = 0 somatic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.39. t = 15 dendritic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.75, t = 15 somatic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.61, ****p < 0.0001, t-test with Welch’s test. B. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during c-Fos experiments shown in Fig. 3D, 1 Hz + PD98089: dendritic Ca2+ 14677 ± 1717, somatic Ca2+ 4414 ± 598, Dendritic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.47, Somatic Ca2+ 1 Hz vs 1 Hz + PD ns p = 0.087, t-test with Welch’s test. C. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during ERK translocation experiments shown in Fig. 3E. 1 Hz: dendritic Ca2+ 10929 ± 1663, somatic Ca2+ 5249 ± 469; 0.5Hz: dendritic Ca2+ 11904 ± 1312, somatic Ca2+ 508 ± 83; 1 Hz + Nim: dendritic Ca2+ 20405 ± 2642, somatic Ca2+ 816 ± 89; 1 Hz + PD: dendritic Ca2+ 16752 ± 1284, somatic Ca2+ 7235 ± 1231. 1 Hz vs 0.5 Hz dendritic Ca2+ ns p = 0.65; 1 Hz vs 1 Hz + Nim dendritic Ca2+ **p = 0.0089; 1 Hz vs 1 Hz + PD dendritic Ca2+ * p = 0.021, somatic Ca2+ ns p = 0.17. ****p < 0.0001, t-test with Welch’s test. D. Pre stimulation YFP-ERK2 nuclear (left) and cytosolic (right) expression. Ns: nuclear - 13970 ± 2585; cytosolic -27963 ± 5128; 1 Hz: nuclear - 10704 ± 1769; cytosolic - 19969 ± 3612; 0.5Hz: nuclear - 8405 ± 3721; cytosolic - 19493 ± 9069; 1 Hz + Nim: nuclear - 5514 ± 2023; cytosolic - 12945 ± 5358; Ns + Nim: nuclear - 10862 ± 2476; cytosolic - 23829 ± 5462; 1 Hz + PD: nuclear 7606 ± 1637; cytosolic 15643 ± 3965; Ns + PD: nuclear 6751 ± 2193; cytosolic 13485 ± 5065; ANOVA with Kruskal-Wallis test. E. Pre stimulation cytosolic MEK1-mTurq2 expression. Ns: 23774 ± 4536; 1Hz: 16847 ± 3065; 0.5Hz: 16612 ± 7752; 1 Hz + Nim: 13422 ± 4773; Ns + Nim: 26098 ± 4727; 1 Hz + PD: 13077 ± 2377; Ns + PD: 15958 ± 4719; ANOVA with Kruskal-Wallis test. Download Figure 3-1, TIF file.
Given the ERK dependence of CREB regulation we observed in response to 1 Hz GU, we set out to examine upstream control of ERK activation. ERK activation following neuronal depolarization has previously been correlated with ERK translocation to the nucleus (Impey et al., 1998; Wiegert et al., 2007; Karpova et al., 2013; Zhai et al., 2013). To image ERK translocation, we transiently transfected YFP-labeled ERK2, along with mTurqoise2-labeled MEK1 consistent with previous studies (Wiegert et al., 2007; Zhai et al., 2013). Sequential stimulation of six dendritic branches by 1 Hz GU promoted ERK enrichment in the soma and translocation into the nucleus that was clearly visible by 10 min poststimulation, resulting in a significant increase in ERK nuclear localization that was maintained out to at least 30 min poststimulation (Fig. 3E). In contrast, ERK translocation was not observed following 0.5 Hz GU stimulation that generated dendrite-restricted Ca2+ signals or following 1 Hz GU in the presence of PD98059 to inhibit upstream MEK activity (Fig. 3E; Extended Data Fig. 3-1C). These observations as well as the lack of ERK translocation following 1 Hz GU in the presence of nimodipine (Fig. 3E), to prevent somatic Ca2+ propagation, suggests that ERK activation required LTCC-mediated somatic Ca2+ signaling. Notably, in our experiments examining ERK translocation to the nucleus, there were no significant differences between ERK2-YFP and MEK1-mTurqoise2 expression levels across conditions (Extended Data Fig. 3-1D–E). These results showing LTCC-dependent ERK translocation, along with the LTCC and ERK dependence of CRE-GFP reporter and c-Fos induction detailed above, indicate that ERK activated in response to somatic LTCC opening is controlling downstream CREB-dependent transcription.
CaMKII signaling is not necessary for CREB-dependent transcription in response to somatic Ca2+ propagation
Various mechanisms are suggested to link LTCC activation to downstream Ras→Raf→MEK→ERK signaling, including directly via LTCC-bound H-Ras and voltage-dependent conformational changes independent of Ca2+ influx (Servili et al., 2018; Servili et al., 2019). Another potential intermediary between LTCC Ca2+ influx and ERK activation is CaMKII, which can interact with and activate Raf-1 upstream of ERK in non-neuronal cells; however, this pathway has not yet been explicitly shown to regulate ERK in response to neuronal depolarization (Illario et al., 2003, 2005; Salzano et al., 2012). Previous studies also describe multiple pathways in which CaMKII activity may contribute to CREB activation independent of ERK signaling downstream of LTCCs (Bito et al., 1996; Wheeler et al., 2008, 2012; Wang et al., 2017). While many previous studies used the pan-CaMK inhibitor KN93 to investigate CaMKII contributions to CREB signaling (Bito et al., 1996; Wu et al., 2001b; Wheeler et al., 2008; Zhai et al., 2013), we found that use of KN93 in our 1 Hz GU paradigm significantly inhibited Ca2+ propagation to the soma (Extended Data Fig. 4-1A), which is consistent with previous reports that KN93 also directly inhibits LTCCs (Gao et al., 2006). Thus, any impacts of KN93 on CREB activation reported in previous studies are confounded by a combination of inhibitory impacts on CaMKII, on other CaMKs that are sensitive to KN93 (i.e., CaMKI and CaMKIV), and also on LTCC Ca2+ influx.
To investigate the possible role of CaMKII in mediating CREB-dependent transcription here, we instead utilized a more-specific CaMKII inhibitor TatCN19o derived from a naturally occurring inhibitor protein CaMKIIN, which unlike KN93 does not inhibit other CaMKs (Coultrap and Bayer, 2011) or strongly impair 1 Hz GU-induced somatic Ca2+ signals that depend on LTCCs (Fig. 4A; Extended Data Fig. 4-1B,C). In addition, TatCN19o and other related CaMKII peptide inhibitors prevent CaMKII binding to NMDARs (Coultrap and Bayer, 2011) and LTCCs (Wang et al., 2017) that is implicated in conformational signaling (Aow et al., 2015; Li et al., 2016); moreover TatCN19o does so at a high potency as previously investigated both in vitro (Coultrap and Bayer, 2011) and in vivo (Deng et al., 2017; Rumian et al., 2023). Yet, surprisingly, we found that TatCN19o did not prevent 1 Hz GU stimulation from inducing CRE-GFP reporter expression (Fig. 4B), CREB phosphorylation (Fig. 4C; Extended Data Fig. 4-1B), or c-Fos expression (Fig. 4D; Extended Data Fig. 4-1C). As an additional positive control for the efficacy of CaMKII inhibition by TatCN19o, we confirmed under our experimental conditions that 1 µM TatCN19o inhibited 1 Hz GU-induced dendritic spine size enlargement, which is a hallmark of sLTP (Extended Data Fig. 4-1D) known to require CaMKII activity (Stein and Zito, 2019). These findings together suggest that, in response to the stimulation paradigms employed here, surprisingly CaMKII does not appreciably contribute to activation of CREB-dependent transcription either dependent or independent of ERK.
CaMKII signaling is not necessary for CREB activation and transcription. A, Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CRE reporter experiments in Figure 4B for 1 Hz + TatCN190 vs 1 Hz as shown in Figure 1B (1 Hz + TatCN19o, n = 15 over 4 biological replicates, dendritic Ca2+ 14,359 ± 2124, somatic Ca2+ 7,021 ± 359; **p = 0.0096; ****p < 0.0001; t test with Welch’s test). B, CRE reporter expression following 1 Hz GU + 1 µM TatCN19o (solid orange), no stimulation/NS + TatCN19o (dashed orange) versus 1 Hz GU (solid black) as shown in Figure 1D. NS + TatCN19o, n = 10 over three biological replicates, 0.7781 ± 0.04830; 1 Hz + TatCN19o, n = 15 over four biological replicates, 1.301 ± 0.05769; ns, p = 0.19; ANOVA repeated measures with mixed-effect analysis, mean and SEM shown, 1 Hz versus 1 Hz + TatCN19o; NS, p = 0.19. C, pCREB expression immediately following 1 Hz GU as shown in Figure 1F and 1 Hz GU + 1 µM TatCN19o raw intensities in AFU graphed. TatCN19o 1 Hz, n = 10 over 7 biological replicates, 3,203 ± 399; ns p = 0.30; Welsh’s t test. D, c-Fos expression 2 h following 1 Hz stimulation + 1 µM TatCN19o versus 1 Hz GU as shown in Figure 1G, raw intensities in AFU, median and IQR graphed (1 Hz + TatCN19o, n = 11 over 6 biological replicates, 2084, 820-5204; ns, p = 0.40 Mann–Whitney ranked sum). See also Extended Data Figure 4-1 related to Figure 4.
Figure 4-1
A. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CRE reporter 1 Hz + 5 μM KN93 experiments. KN93 1Hz: n = 11 over 5 biological replicates, dendritic Ca2+ 10189 ± 885, somatic Ca2+ 948 ± 237, ns p = 0.055, ****p < 0.0001, t-test with Welch’s test. B. Quantification of dendritic and somatic Ca2+ influx quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during pCREB experiments shown in Fig. 4C. 1 Hz + 1 μM TatCN19o dendritic Ca2+ 12960 ± 1517, somatic Ca2+ 4061 ± 462, dendritic Ca2+ ns p = 0.29, somatic Ca2+ ns p = 0.18, t-test with Welch’s test. C. Quantification of dendritic and somatic Ca2+ influx quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during c-Fos experiments shown in Fig. 4D. 1 Hz + TatCN19o: dendritic Ca2+ 11971 ± 2055, somatic Ca2+ 4543 ± 376, ns p = 0.57, *p = 0.040, t-test with Welch’s test. D. Left: Quantification of change in dendritic spine area over time following 60, 1 Hz GU pulses adjacent to the spine in the presence or absence of 1 μM TatCN19o; 1Hz: n = 16 over 4 biological replicates, no stim: n = 18 over 4 biological replicates, 1 Hz + TatCN19o n = 11 over 4 biological replicates, no stim + TatCN19o n = 9 over 4 biological replicates; 1 Hz vs 1 Hz + TatCN19o: *p = 0.0322, **p = 0.0022 ANOVA Repeated Measures with mixed-effects analysis quantifying timepoints 50-60 min post GU. Center: quantification of Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 throughout each spine GU stimulation. Right: representative images of iRFP expressing spines pre GU and following GU, scale bar 1μm. Download Figure 4-1, TIF file.
CaMKK→CaMKIV signaling is also dispensable for CREB-dependent transcription in response to somatic Ca2+ propagation
In addition to ERK and MSK/RSK family kinases downstream of ERK (Xing et al., 1996; Sindreu et al., 2007), nuclear resident CaMKIV can phosphorylate CREB (Matthews et al., 1994; Sun et al., 1994; Bito et al., 1996), as well as the CREB coactivator CBP (Chawla et al., 1998; Impey et al., 2002). Depolarization-induced nuclear Ca2+ signals have been correlated with plasticity-related transcription (Yu et al., 2017; Lobos et al., 2021), presumably allowing for direct activation of CaM and CaMKIV in the nucleus. In sympathetic and cortical neurons, it has also been shown that following depolarization and LTCC Ca2+ influx, CaM is shuttled to the nucleus by CaMKIIγ to promote activation of CaMKIV leading to CREB phosphorylation (Deisseroth et al., 1998; Ma et al., 2014; Cohen et al., 2018). CaMKIV activation requires not only Ca2+-CaM binding but also phosphorylation by CaMKK, with the inhibition of nuclear CaM (Deisseroth et al., 1998; Limbäck-Stokin et al., 2004) or of CaMKK by STO609 (Ma et al., 2014) preventing CREB phosphorylation. We therefore investigated the contribution of the CaMKK→CaMKIV pathway using the CaMKK inhibitor STO609. Despite slightly reducing somatic Ca2+ influx in some (Fig. 5A) but not all experiments (Extended Data Figs. 5-1A,B), STO609 did not inhibit CRE-GFP reporter expression following 1 Hz GU (Fig. 5B). Interestingly, we actually observed increases in CREB phosphorylation (Fig. 5C; Extended Data Fig. 5-1A) and c-Fos expression (Fig. 5D; Extended Data Fig. 5-1B) following 1 Hz GU in the presence of STO609. These data, along with the insensitivity to CaMKII inhibition by TatCN19o shown in Figure 4, indicate that CREB E–T coupling mediated by LTCC somatic Ca2+ propagation is independent of the LTCC→CaMKII→CaMKK→CaMKIV pathway or even a CaMKK→CaMKIV pathway reliant on nuclear Ca2+.
CaMKK→CaMKIV signaling pathways are also dispensable for CREB-dependent transcription. A, Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during CRE reporter experiments in Figure 5B, 1 Hz + 5 µM STO609 and 1 Hz as shown Figure 1C (1 Hz + STO609, n = 14 over 4 biological replicates, dendritic Ca2+ 5,526 ± 690, somatic Ca2+ 2,141 ± 278; *p = 0.041; **p = 0.0055 t test with Welch’s test). B, CRE reporter expression following 1 Hz GU + 5 µM STO609 (solid gold), no stimulation/NS + 5 µM STO609 (dashed gold) versus 1 Hz GU (solid black) as shown in Figure 1D. NS + STO, n = 24 over five biological replicates, 0.86 ± 0.038; 1 Hz + STO, n = 14 over four biological replicates l, 1.43 ± 0.058. ns, p = 0.73 ANOVA repeated measures with mixed-effect analysis. C, pCREB expression immediately following 1 Hz GU as shown in Figure 1F and 1 Hz GU + 5 µM STO609 (n = 10 over 6 biological replicates), raw intensities shown (1 Hz + STO, n = 10 over 6 biological, 7,091 ± 1416; *p = 0.049 t test with Welsh’s t test). D, c-Fos expression 2 h following 1 Hz stimulation + 5 µM STO609 versus 1 Hz GU as shown in Figure 1G, raw intensities, median, and IQR shown (1 Hz + STO609, n = 12 over 7 biological replicates, 2469, 1667–3949; **p = 0.0093). Mann–Whitney ranked sum. See also Extended Data Figure 5-1 related to Figure 5.
Figure 5-1
A. Quantification of dendritic and somatic Ca2+ signals quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during pCREB experiments shown in Fig. 5C. 1 Hz + STO609: dendritic Ca2+ 18358 ± 4340, somatic Ca2+ 4464 ± 564. Dendritic Ca2+ ns p = 0.13, Somatic Ca2+ ns p = 0.52, t-test with Welch’s test. B. Quantification of dendritic and somatic Ca2+ influx quantified by AUC of jRGECO1a ΔF/F0 over all stimulated branches during c-Fos experiments shown in Fig. 5D. 1 Hz + STO: dendritic Ca2+ 12510 ± 1766, somatic Ca2+ 4880 ± 257, Dendritic Ca2+ ns p = 0.71, Somatic Ca2+ ns p = 0.084, t-test with Welch’s test. Download Figure 5-1, TIF file.
Discussion
Here we show that dendritic NMDAR stimulation engages CREB-dependent transcription downstream of subsequent dendrite-to-soma Ca2+ propagation that activates ERK signaling to CREB. We find that this Ca2+ signaling from dendrites to the soma is LTCC-dependent, with LTCC antagonism preventing somatic Ca2+ propagation, ERK translocation, CREB phosphorylation, CREB-dependent transcription, and c-Fos induction. Despite many reports in the literature of CaMKII and/or CaMKIV signaling to CREB, we observed that both CaMKII and CaMKIV activities were dispensable for CREB phosphorylation, as well as CREB-dependent transcription, when activated downstream of dendritic NMDAR stimulation. Importantly, our study was designed to directly compare these various upstream Ca2+ mobilization and downstream enzymatic mechanisms proposed to control CREB-dependent transcription to address contradictions in the literature, which are complicated by variations in the methodologies employed for neuronal stimulation and/or analysis of CREB activation. Many previous studies used bath chemical stimulations of the entire neuronal cultures instead of targeted, single-cell stimulation (Wu et al., 2001b; Wheeler et al., 2008, 2012; Li et al., 2016; Tyssowski et al., 2018). By taking advantage of GU to illicit either dendrite-restricted or soma-propagating Ca2+ stimuli, which we confirmed by Ca2+ imaging in every experiment, we revealed that somatic Ca2+ signaling is necessary for CREB-dependent transcription, which is in agreement with some prior studies of CREB E–T coupling in LTP using antidromic electrical stimulation to trigger somatic depolarization and LTCC opening (Dudek and Fields, 2002). As in our previous studies of NFAT E–T coupling (Wild et al., 2019), while 1 Hz GU stimulation of dendrites usually led to Ca2+ propagation to the soma and 0.5 Hz stimulation generally led to dendrite-restricted Ca2+ elevation, we had to exclude 10–15% of cells from our analyses because either 1 Hz stimuli did not propagate or 0.5 Hz stimuli led to somatic propagation (KHZ and MLD unpublished observations). In addition, we previously found that the probability of even low-frequency GU triggering dendrite-to-soma Ca2+ propagation dramatically increased the closer the site of uncaging was to the soma (Wild et al., 2019). Thus, such variable responses following GU highlights the necessity of monitoring Ca2+ when attempting to correlate the resulting patterns of neuronal activity with physiological responses, perhaps explaining differences in our conclusions from those of Zhai et al. where similar GU stimulation approaches were employed to activate CREB but Ca2+ responses were not simultaneously monitored (Zhai et al., 2013).
Our reporter assay monitoring the CREB transcriptional response in a single stimulated cell over time also has many advantages over using bath chemical stimulation or field stimulation, which requires fixing cells at various time points to take measurements of pCREB and c-Fos mRNA or protein expression normalized to separate nonstimulated cell population averages. For instance, while many studies have correlated pCREB expression with downstream IEG induction, the use of our CRE-GFP reporter in combination with pCREB staining confirmed previous observations that elevation of pCREB observed immediately following stimulation is not always correlated with sustained CREB-dependent transcription (Dolmetsch et al., 2001; Wu et al., 2001b). Moreover, our parallel results from measuring CRE-GFP reporter, pCREB, and c-Fos expression when combined indicate that the ERK dependence of c-Fos induction we observed is likely in large part due to ERK acting upstream of CREB rather than through parallel activation of other transcription factors, such as Elk-1 (Marais et al., 1993). In addition, our CRE-GFP reporter would have allowed detection of a delayed phase of CREB-dependent transcription proposed to be NMDAR dendritic Ca2+ and ERK-dependent but LTCC-independent (Zhai et al., 2013); however, we observed no second, later phase of reporter expression following dendrite-restricted Ca2+ signals induced using either 0.5 Hz GU or 1 Hz GU with nimodipine present.
While our findings clearly implicate MEK→ERK activation downstream of somatic LTCCs in CREB E–T coupling, future investigations are necessary to characterize upstream activation of Ras→Raf signaling, including identifying the Ras guanine-nucleotide exchange factor(s) involved and to elucidate the spatiotemporal dynamics of this pathway. For instance, are LTCC→ERK signaling intermediaries organized in the LTCC microdomain and if so by what anchoring interactions? Our current assessment also leaves open the possibility that other Ca2+ sources could still contribute to the ERK→CREB signaling readouts we monitored, but these contributions would be supplemental and not sufficient to drive CREB-dependent transcription. For instance, Ca2+ released from the ER in response to LTCC opening could be enhancing signaling by LTCCs in the soma, including signaling involving conformational changes (Li et al., 2016; Servili et al., 2018, 2019). It is also possible that in our experiments, some activated ERK travels from at least proximal dendrites to the soma (Karpova et al., 2013; Zhai et al., 2013) leading to the somatic ERK accumulation we observed after 1 Hz GU. However, our experiments clearly indicate that distal dendrite-restricted influx of Ca2+ alone with or without LTCC conformational signaling is not sufficient to significantly increase ERK translocation to the nucleus, pCREB levels, or downstream transcription detectable by either the CRE-GFP reporter or c-Fos expression.
It is also plausible that there is a small contribution of CaMKII to CREB activation in our experiments, especially given the nonsignificant trends toward decreased CREB phosphorylation and CRE-GFP reporter expression we observed in the presence of the CaMKII inhibitor TatCN19o. However, based on the necessity of ERK activity and the lack of attenuation of CREB-dependent transcription by the CaMKK inhibitor STO609, it is likely that any such contribution of CaMKII signaling is occurring upstream of ERK activation (Illario et al., 2003, 2005; Salzano et al., 2012), as opposed to downstream through activation of CaMKK→CaMKIV. It is also notable that several studies implicating CaMKII→CaMKIV signaling to CREB were performed in cortical (Li et al., 2016) and/or sympathetic neurons (Wheeler et al., 2012; Ma et al., 2014) instead of hippocampal neurons, potentially indicating that distinct kinase cascades may contribute to CREB activation across neuronal cell types. Yet in contrast, other studies in cortical neurons found that ERK-dependent IEG activation following bath KCl depolarization is controlled through rapid recruitment of RNA Pol2 to promoters that are primed for transcriptional initiation by bound transcription factors, including CREB (Tyssowski et al., 2018). Thus, given our results showing that ERK-mediated activation of CREB is only effectively engaged by somatic Ca2+ influx, the rapid IEG induction that occurs following neuronal stimulation in both cortical and hippocampal neurons might largely reflect somatic LTCC→ERK signaling. However, we did not investigate whether somatic LTCC activation of cAMP→PKA signaling to CREB, either directly or indirectly via ERK-MSK1 regulation (Sindreu et al., 2007; Vierra et al., 2021, 2023), could also contribute to CREB E–T coupling as measured in our experiments. Nonetheless, our present findings support the dominant role of ERK signaling in CREB regulation and help reconcile previous inconsistencies in the literature regarding the involvement of CaMK versus ERK pathways, including by highlighting problems with the use of KN93 to study LTCC→CaMKII signaling in many previous studies.
In summary, our findings identify a previously unappreciated rapid dendrite-to-soma LTCC→ERK signaling pathway that drives CREB synapse-to-nucleus E–T coupling. Furthermore, we conclude that LTCC dendrite-to-soma Ca2+ signal propagation is the common mediator of E–T coupling in hippocampal neurons for both CREB and NFAT (Wild et al., 2019), two of the most prominent phosphorylation-regulated transcription factors. Importantly, we previously determined that this mode of NMDAR→LTCC signaling from synapse to nucleus produces CaV and NaV spiking that closely resembles dendritic plateau potentials/complex spikes observed ex vivo and in vivo in CA1 neurons (Golding et al., 1999, 2002; Remy and Spruston, 2007; Grienberger et al., 2014). Importantly, such complex spiking promotes multisynapse LTP in dendrites (Golding et al., 2002; Lisman and Spruston, 2005; Remy and Spruston, 2007; Takahashi and Magee, 2009), produces neuron-wide Ca2+ responses similar to those we observed here with 1 Hz GU dendritic stimulation (Takahashi and Magee, 2009; Xu et al., 2012; Grienberger et al., 2014), and is both required and sufficient for behavioral timescale synaptic plasticity that establishes hippocampal place cells during spatial learning and memory in vivo (Bittner et al., 2015, 2017). Thus, the use of our CRE-GFP reporter paired with Ca2+ imaging and behavioral monitoring of CA1 neuron activity could be used for future investigations of how CREB E–T coupling is regulated during spatial learning and memory in vivo. Notably, genome-wide association studies have linked polymorphisms in the human CACNA1C gene encoding the predominant neuronal CaV1.2 LTCC pore-forming subunit to multiple neuropsychiatric disorders including autism spectrum disorders (ASD), schizophrenia, major depressive disorder, and bipolar disorder that are all characterized by altered cognition (Bhat et al., 2012; Smoller et al., 2013; Nie et al., 2015). In addition, gain-of-function missense mutations in both CaV1.2 and CaV1.3 LTCCs are associated with syndromic and nonsyndromic ASD, respectively (Splawski et al., 2004; Bader et al., 2011; Pinggera et al., 2015). This strong association of LTCCs with multiple nervous system disorders is consistent with the central role these channels play in learning and memory through E–T coupling control of neuronal gene expression, thus further underscoring the importance of understanding how these channels mediate synapse-to-nucleus information relay as elucidated here.
Footnotes
We thank Dr. Angela Wild for the initial subcloning work used in construction of the CRE-2xsGFP reporter plasmid. This work was supported by National Institutes of Health R01MH123700 (M.L.D.A. and K.H.Z.) and T32GM007635 (K.H.Z.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Mark L. Dell’Acqua at mark.dellacqua{at}cuanschutz.edu.