Voltage Gated Calcium Channel Activation by Backpropagating Action Potentials Downregulates NMDAR Function

The majority of excitatory synapses are located on dendritic spines of cortical glutamatergic neurons. In spines, compartmentalized Ca2+ signals transduce electrical activity into specific long-term biochemical and structural changes. Action potentials (APs) propagate back into the dendritic tree and activate voltage gated Ca2+ channels (VGCCs). For spines, this global mode of spine Ca2+ signaling is a direct biochemical feedback of suprathreshold neuronal activity. We previously demonstrated that backpropagating action potentials (bAPs) result in long-term enhancement of spine VGCCs. This activity-dependent VGCC plasticity results in a large interspine variability of VGCC Ca2+ influx. Here, we investigate how spine VGCCs affect glutamatergic synaptic transmission. We combined electrophysiology, two-photon Ca2+ imaging and two-photon glutamate uncaging in acute brain slices from rats. T- and R-type VGCCs were the dominant depolarization-associated Ca2+conductances in dendritic spines of excitatory layer 2 neurons and do not affect synaptic excitatory postsynaptic potentials (EPSPs) measured at the soma. Using two-photon glutamate uncaging, we compared the properties of glutamatergic synapses of single spines that express different levels of VGCCs. While VGCCs contributed to EPSP mediated Ca2+ influx, the amount of EPSP mediated Ca2+ influx is not determined by spine VGCC expression. On a longer timescale, the activation of VGCCs by bAP bursts results in downregulation of spine NMDAR function.


INTRODUCTION
The dendritic surface of most excitatory projection neurons is covered with thousands of spines which receive over 90% of glutamatergic synapses (Harris and Kater, 1994). Spines and the corresponding synapses are plastic, they undergo activity-dependent functional state changes. This includes spine-and synapse specific structural, morphological and functional changes ultimately affecting synapse strength and stability (Alvarez and Sabatini, 2007;Korte and Schmitz, 2016;Segal, 2017). An important signaling molecule related to different types of spine plasticity is calcium (Ca 2+ ). Ca 2+ serves as a transducer between fast and transient electrical signals at the membrane and biochemical and structural changes outlasting the initial electrical signal (Hille, 2001). Therefore, spine Ca 2+ signals evoked by neuronal activity substantially determine plasticity processes.
Voltage gated Ca 2+ channels (VGCCs) contribute to spine Ca 2+ signals in synaptically excited spines. Theory predicts that spine VGCCs could actively enhance synaptic excitatory postsynaptic potentials (EPSPs; Miller et al., 1985;Segev and Rall, 1988;Araya, 2014). In a number of experimental studies, the contribution of VGCCs to synaptic depolarization during EPSPs has been described as marginal (Bloodgood and Sabatini, 2007;Palmer and Stuart, 2009;Popovic et al., 2015). In CA1 pyramidal neurons, they are part of a negative feedback loop that dampens synaptic EPSPs (Bloodgood and Sabatini, 2007;Giessel and Sabatini, 2011;Wang et al., 2014Wang et al., , 2015. In addition to specific direct synaptic activation of spines, VGCCs are also activated globally in spines that are activated by electrotonic spread of depolarization and not by direct synaptic inputs (Higley and Sabatini, 2008). Whenever suprathreshold synaptic activation evokes an action potential (AP), the depolarization travels back into the dendrite in the form of backpropagating action potentials (bAPs). bAPs activate VGCCs in dendrites and spines. The resulting global Ca 2+ signals transmit neuronal activity levels to a large population of spines that do not receive direct synaptic activation (Waters et al., 2005). Recently, we observed that neuronal AP firing upregulates spine VGCCs. In these experiments, we also observed a larger range of single spine VGCC responses to bAPs in excitatory layer 2 neurons of the MEC when directly compared to CA1 pyramidal neurons (Johenning et al., 2015). The plasticity and interspine variability of VGCC expression has not been taken into account by previous studies examining the acute effect of spine VGCCs on synaptic transmission. In addition to acute effects resulting from direct electrical interactions in spines with a high density of VGCCs, VGCC expression levels may also interfere with the long-term regulation of synaptic strength. We therefore tested if synaptic properties of spines with large depolarization-mediated VGCC Ca 2+ transients differ from spines with small Ca 2+ transients. AMPAR mediated synaptic transmission appeared functionally uncoupled from VGCC-mediated Ca 2+ influx and VGCCs do not electrically amplify spine EPSPs. However, we found a reduction of the spine-specific NMDAR-response by bAP-Ca 2+ transients. This establishes a functional link between global VGCC activation by dendritic backpropagation of APs and synaptic transmission.

Slice Preparation and Electrophysiology
Acute brain slices were prepared from Wistar rats (postnatal day 17-25, see Supplementary Figure S5 for age distributions in different experiments) in accordance with the national and institutional guidelines as described in Beed et al. (2010). All procedures were approved by the local health authority and the local ethics committee (Landesamt für Gesundheit und Soziales, Berlin; animal license number T0073/04). Briefly, brains were placed in ice-cold artificial cerebrospinal fluid (ACSF; pH 7.4) containing (in mM): 87 NaCl, 26 NaHCO 3 , 25 Glucose, 2.5 KCl, 7 MgCl 2 , 1.25 NaH 2 PO 4 , 0.5 CaCl 2 and 75 Sucrose. Slices were cut at 300 µm thickness, and incubated at 35 • C for 30 min. The slices were then transferred to standard ACSF containing (in mM): 119 NaCl, 26 NaHCO 3 , 10 Glucose, 2.5 KCl, 2.5 CaCl 2 , 1.3 MgCl 2 , and 1 NaH 2 PO 4 . The slices were stored at room temperature in a submerged chamber for 0.5-5 h before being transferred to the recording chamber. One micromolar of Gabazine was added for all experiments involving synaptic stimulation. For application of Ni 2+ and the interleaved control experiments, NaH 2 PO 4 was omitted from the standard ACSF. Application of peptide toxins CtxGIVA,AgaIVA) and the interleaved control experiments were performed in a perfusion system where tubing was coated with 0.1 mg/ml cytochrome added to the standard ACSF. TTA-P2, Nimodipine, PD-173212, SKF-96365 and D-APV were added to the standard ACSF after baseline recordings. For voltage clamp experiments in Figure 4, 1 µM TTX and 10 µM D-Serine was added to the bath. In experiments shown in Figure 5, we wanted to isolate NMDAR EPSCs at hyperpolarized potentials. In these experiments, slices were recorded in low-Mg 2+ ACSF consisting of (in mM): 119 NaCl, 26 NaHCO 3 , 10 Glucose, 2.5 KCl, 2.5 CaCl 2 , 0.1 MgCl 2 , 1 NaH 2 PO 4 , 1 µM Gabazine and 20 µM NBQX.

Two-Photon Calcium Imaging and Uncaging
A Femto 2D two-photon laser scanning system (Femtonics Ltd., Budapest, Hungary) was equipped with two femtosecond pulsed Ti:Sapphire lasers (Cameleon, Coherent, Santa Clara, CA, USA). The imaging laser was tuned to λ = 805 nm for Alexa-594 (red fluorescence (R), morphometric dye) and Fluo-4FF or Fluo-5F (green fluorescence (G) low and high affinity Ca 2+ indicators, respectively). The uncaging laser was tuned to 720 nm. The laser lines were directly coupled into the microscope, precise overlay of the imaging and uncaging laser lines was checked directly before starting an experiment using fluorescent spheres. Imaging and uncaging were controlled by the Matlab-based MES software package (Femtonics Ltd., Budapest, Hungary). For delivery of uncaging pulses and detection of fluorescence we used a water immersion objective (LUMPLFL 60×/1.0 NA, Olympus, Hamburg, Germany). Transfluorescence and transmitted infra-red light were detected using an oil immersion condenser (Olympus). We filled the cells for at least 25 min with dye before multiple line-scans of dendritic spines and the adjacent dendritic segment were taken. Spines were within 150 µm of the soma to reduce space-clamp errors. Imaged spines were located up to 75 µm under the slice surface. The average scanning speed was 300 Hz and the intermediate sections were jumped over within 60 µs using a spline interpolated path (Lörincz et al., 2007). bAP-Ca 2+ transients of the doublet test stimulus were measured every 60 s. uEPSP Ca 2+ transients where measured at 300 Hz by interleaving the uncaging point and the imaging line in the multiple line scan mode applying the aforementioned spline interpolated path (Bywalez et al., 2015). 0.5-1 ms duration uncaging pulses were generated using an electro-optical modulator (Pockels cell, Conoptics). Energy deposition of the uncaging laser for each individual spine of interest was calibrated by placing a fluorescent calibration microsphere from the same batch (InSpeck Green, Thermo Fischer Scientifics) in a sealed patch pipette in the focal plane next to the spine. Laser power was then set to always reach the same absolute brightness value under identical detection settings (Supplementary Figure S3). Based on the minuscule intrabatch variability of the beads, we assumed identical interslice and inter-animal energy deposition of the uncaging laser. At a distance of about 0.5 µm from the spine head, we then searched for the maximal activation point of each spine. Before each uncaging sweep, the position of the spinehead and the uncaging point were realigned to compensate for movements. Care was taken to only image isolated spines >2 µm from a neighboring spine in order to minimize cross-activation of different spines.
Calibration of G max /R values (ratio of maximal green fluorescence change under saturating [Ca 2+ ] over red) was performed at the tip of a sealed pipette in the imaging plane of the slice using a calibration solution consisting of 50 µl recording solution and 50 µl 1 M CaCl 2 for each batch of recording solution. For statistical comparisons and display of averages from different experiments (Supplementary Figure S4), changes in [Ca 2+ ] are reported as G/G max obtained by dividing ∆G/R values by their corresponding G max /R values (Holderith et al., 2012). Typical traces from single experiments are averages of 2-6 sweeps and displayed as ∆G/R values.
For quantification of bAP-Ca 2+ amplitudes, G/G max was averaged over a 70 ms time interval starting 10 ms after the first AP. uEPSP Ca 2+ amplitudes were averaged over a 20 ms time interval starting 20 ms after the uncaging pulse. The depolarization pulse in voltage clamp was averaged over a 30 ms time interval. The analysis intervals are based on visual inspection of the data for maximizing the signal to noise ratio. ∆G/G max amplitudes are averages of 2-6 sweeps.
To eliminate out of focus line scan measurements, measurements were excluded when the Alexa-594 intensity was below 80% of the average baseline intensity. A further exclusion criterion was the rise in the background corrected baseline green over red (G 0b R) above 50% of the average baseline value in three consecutive sweeps. Increases in G 0b R indicate a rise in baseline Ca 2+ suggesting a deterioration in cell health (Yasuda et al., 2004).
In experiments where relative changes of the bAP-Ca 2+ transient were measured (Figure 1), spines were only included if at least two out of six sweeps in the baseline and analysis time window fulfilled the above mentioned criteria for focus and cell health. In addition, in order to form a meaningful post/pre percentage ratio, we defined an inclusion criterion for the signalto-noise ratio (S/N) of the pre-induction signal: in averaged traces of the pre-induction sweeps, the amplitude averaged in the 20 ms time interval 10 ms after the second AP in a doublet had to be 2.5 times larger than the standard deviation of a 40 ms pre-AP baseline stretch.
For morphological reconstructions, we performed post hoc high-resolution z-stacks of the recorded spines with a ∆z of 0.2 µm. Head size and spine length were estimated from maximum intensity projections of z-stacks of the spines and the adjacent dendritic segment. The apparent spine size was approximated by measuring the FWHM of the maximal spine diameter x. The diameter of spines is below the resolution limit of a two-photon microscope. We therefore implemented a correction factor k by dividing the maximal brightness of a spine by the maximal brightness of the adjacent dendritic segment. This correction is based on the assumption that the dendritic segment is larger than the resolution limit of the 2P microscope (Holtmaat et al., 2005). Spine length was determined from the origin of the spine at the dendrite to the middle of the spine head.

Statistics
For statistical comparisons, we used GraphPad Prism Software 1 or the MATLAB Statistics toolbox. Datasets were tested for normal distribution using a Shapiro-Wilk normality test. In case of non-normally distributed data, the non-parametric two-tailed Wilcoxon-Mann-Whitney U test was used unless otherwise noted. When data was distributed normally, two-tailed t-tests were used as indicated. Correlations were tested using the non-parametric Spearman's rank order test unless otherwise noted. Significance level for all statistical tests was at p < 0.05. Unless otherwise noted, data are reported as median ± interquartile range.

VGCC Content of MEC Layer 2 Spines Measured With bAPs
The bAP-Ca 2+ transient is mainly set by the VGCC channel conductances (Bloodgood and Sabatini, 2007). We can therefore 1 https://www.graphpad.com/scientific-software/prism/ define the bAP-Ca 2+ transient as a proxy for the spine VGCC content. In acute brain slices, we studied excitatory neurons in layer 2 of the medial entorhinal cortex (MEC) in the whole-cell patch clamp configuration. We performed two-photon Ca 2+ imaging of bAP-Ca 2+ transients in spines using Fluo-5F (500 µM) as a Ca 2+ indicator and Alexa 594 (30 µM) as a morphometric dye ( Figure 1A). We analyzed the contribution of different subtypes of VGCCs to doublet bAP-Ca 2+ transients using different antagonists. For these experiments, we chose a relatively high concentration of a medium-affinity Ca 2+ indicator in order to improve our signal to noise ratio and suppress VGCC plasticity (Johenning et al., 2015). Every 60 s, we imaged doublet bAP-Ca 2+ transients evoked by current injection. The effect of drug-wash-in was normalized to the 5 min pre-wash-in baseline. Drug effects were compared to the averaged control experiments (ctrl) interleaved with the drug application experiments under similar recording conditions (time-matched controls) and expressed as ∆ctrl-drug in Figure 1K.

Instantaneous Consequences of Specific Block of Spine VGCCs on Synaptic Compound EPSPs
Ligand and voltage-gated channels in the spine head can interact on different timescales. Instantaneous electrical interactions between VGCCs and iGLURs would have an effect on spine excitability and modulate synaptic potentials. At this timescale, VGCCs could contribute to both the depolarization underlying the EPSP and the EPSP Ca 2+ transient.
Theory suggests that VGCCs could electrically amplify the spine depolarization underlying EPSPs resulting in active local boosting of EPSPs (Koch and Poggio, 1985;Miller et al., 1985;Segev and Rall, 1988). In this context, recruitment of especially low voltage activated T-Type VGCCs by larger synaptic responses has also been discussed (Deisz et al., 1991;Gillessen and Alzheimer, 1997;Seong et al., 2014). Another acute effect of VGCCs on synaptic potentials was demonstrated in CA1 pyramidal cells, where R-Type Ca 2+ channels specifically interact with K + channels. Ca 2+ influx from R-type channels activates K + channels, which results in a dampening of the EPSP, with conflicting results regarding the identity of the target K + channel (Bloodgood and Sabatini, 2007;Giessel and Sabatini, 2011;Wang et al., 2014Wang et al., , 2015. Based on this hypothesis, we used a KMeSO3 intra to enhance the impact of Ca 2+ activated K + channels, which are reduced by K + -gluconate based intracellular solutions (Velumian et al., 1997).
Previously, our experiments demonstrated significantly larger VGCC mediated Ca 2+ transients in proximal spines of layer 2 cells in the MEC compared to proximal spines in CA1 pyramidal cells (Johenning et al., 2015). Theory predicts that active nonlinear interactions between VGCCs and AMPARs occur in spines with a high VGCC density (Segev and Rall, 1988). Consequently, MEC layer 2 cells are a good model to study the interaction of VGCCs and glutamatergic transmission in spines.

Combined Two-Photon Ca 2+ Imaging and Glutamate Uncaging for Direct Measurements of Single-Spine VGCCs and iGLURs
Previously, we demonstrated that there are large interspine differences with respect to the amplitude of bAP Ca 2+ transients in layer 2 cells of the MEC. In addition, the responses are much larger when directly compared to proximal spines of hippocampal CA1 cells (Johenning et al., 2015). This variability has not been taken into account by previous studies analyzing the contribution of postsynaptic VGCCs to synaptic transmission and spine electrogenesis, and may be averaged out when stimulating a large population of spines with extracellular stimulation. VGCCs could also affect synaptic plasticity processes on a longer time scale via mechanisms independent of electrical interactions.
We therefore wanted to characterize synaptic properties of single spines with different levels of VGCC mediated Ca 2+ influx. While it is possible to analyze the Ca 2+ transients evoked in single spines by extracellular synaptic stimulation, it is not possible to relate these to synaptic properties as extracellular stimulation inevitably recruits several synapses on spines which may not even be in the field of view (Johenning et al., 2009). Two-photon uncaging of glutamate is a tool that permits for selected activation of individual spines (Matsuzaki and Kasai, 2011). We combined two-photon glutamate uncaging, Ca 2+ imaging and whole-cell patch clamp recordings with a K + -gluconate based intracellular solution in which we switched between current clamp and voltage clamp mode. This way, we related the strength of iGLUR mediated uncaging excitatory postsynaptic currents (uEPSCs), interactions between iGLURs and voltage gated conductances during uncaging EPSPs (uEPSPs), uEPSP Ca 2+ transients and bAP-Ca 2+ transients at the single spine level (Figure 3A).
A prerequisite for our experiments was a protocol that permitted inter-slice and inter-animal comparisons of spine uEPSC and uEPSP amplitudes in two-photon uncaging experiments. As opposed to slice cultures, where spines are located in a thin layer of tissue, the focal depth of spines in acute slices differs significantly. For inter-experimental comparability, we therefore had to compensate for focal-depth dependent differences in energy deposition of the uncaging laser. The strategy we chose was to use InSpeck microspheres from the same batch, which are standardized fluorescent spheres with a very low inter-sphere variability in fluorophore brightness based on the manufacturer's specifications (see ''Materials and Methods'' section). Using identical detection settings between experiments, we positioned the InSpeck microspheres at the tip of a pipette at the same focal plane as our spine of interest (Supplementary Figure S3A). Assuming the InSpeck microsphere's brightness and the detection efficiency of our two-photon system were constant between different trials, we set the uncaging laser intensity so that we would always detect the same brightness of the bead. This way, we assured identical inter-slice and inter-animal energy deposition of the uncaging laser. We kept inter-experimental caged glutamate concentrations constant by always bath-applying the same concentration of caged glutamate (2.5 mM). For each spine, we located the uncaging spot generating a maximal response. Before each uncaging experiment, the relative position of this uncaging spot and the spine of interest was readjusted to compensate for drift.
Using this approach, we could reproduce the correlation between spine size and uEPSC amplitude demonstrated before in slice cultures (Matsuzaki et al., 2001;Zito et al., 2009) in a population of 50 spines from 30 cells from 21 different animals (p = 0.01 and r = 0.36, Spearman correlation, Supplementary  Figures S3B,D). Using our calibration approach, we generated uEPSCs with an amplitude of 26.03 ± 1.54 pA (mean ± SEM). This is in the same range as the unitary amplitudes of local inputs onto layer 2 cells in the MEC we measured previously in age-matched rats (Beed et al., 2010). These local inputs correspond to the proximal apical spines under investigation here (Ma et al., 2008). In the same spine population, we correlated neck length with the EPSC amplitude but could not find a significant correlation (p = 0.07 and r = −0.25, Spearman correlation, Supplementary Figures S3B,C).
We next wanted to find out whether we could observe a comparable APV effect on the synaptically evoked compound EPSP. Here, we again compared APV-wash-in to time-matched controls. However, as opposed to our uncaging experiments, APV did not have a significant effect on the synaptically evoked first or second compound EPSP amplitude (APV: n = 8/4 (cells/animals), ctrl: 10/6 (cells/animals), EPSP1: p = 0.18 (t-test), EPSP2: p = 0.24 (t-test), Supplementary Figure S4E). This means that all NMDAR-related measures described here when using two-photon uncaging do not reflect synaptic NMDARs exclusively but also depend on the extrasynaptic NMDARs on a spine. This has to be kept in mind when comparing uEPSPs and uEPSCs to synaptic activation by extracellular stimulation. The physiological relevance of this extrasynaptic pool of spine NMDARs is not clear. Extrasynaptic NMDARs on spines have been proposed to get recruited by synaptic glutamate spillover during repetitive synaptic inputs .

Instantaneous Effect of VGCCs on Local Ca 2+ Transients in Dendritic Spines
To test whether VGCC expression in spines is linked to their synaptic properties, we obtained a dataset with single spine measurements of bAP-Ca 2+ transients, EPSP related Ca 2+ transients, somatic EPSP amplitudes and somatic EPSC amplitudes measured in voltage clamp ( Figure 3A). We measured spine Ca 2+ transients with the low affinity Ca 2+ indicator Fluo-4FF. In contrast to the pharmacology experiments in Figure 1, we used a low-affinity Ca 2+ indicator in order to minimize dye nonlinearities that could occur when measuring synaptic Ca 2+ transients (Yasuda et al., 2004).
Previous work shows that VGCCs locally enhance the EPSP related Ca 2+ influx on the single spine level. This local influx was not mirrored by VGCC related changes of the somatic EPSP (Bloodgood et al., 2009;Seong et al., 2014;Bywalez et al., 2015).
Probing with bAPs, we can distinguish between spines with different levels of VGCC mediated Ca 2+ influx. The previously demonstrated absence of a correlation between bAP-Ca 2+ transient amplitude and spine morphology leaves VGCC activation levels as the most likely explanation for spine to spine differences in bAP-Ca 2+ transients (Johenning et al., 2015).
First, we tested whether the synaptic Ca 2+ transient depends on the amount of synaptically mediated depolarization. When plotting the uEPSP amplitude against the uEPSP Ca 2+ transient, we did not see a correlation (n = 22/11/9 (spines/cells/animals), r = 0.102, p = 0.65, Figure 3B). To find out what determines the size of the uEPSP Ca 2+ transient, we first wanted to test if the VGCC content assessed by bAP-Ca 2+ transients can explain differences in local glutamate mediated spine Ca 2+ influx. However, these two values were not correlated (n = 22/10/9 (spines/cells/animals), p = 0.92, r = −0.023, Spearman correlation, Figure 3C). Nevertheless, in APV we still observe Ca 2+ influx upon glutamate uncaging (Supplementary Figure S4C). From this we conclude that while VGCCs can contribute to uEPSP-related Ca 2+ influx, they do not determine its amplitude.
Theoretical studies predict that large local synaptic depolarizations which would significantly recruit additional voltage gated conductances are preferentially reached in spines with a large spine neck resistance (Segev and Rall, 1988). To test this hypothesis, we related the uEPSP Ca 2+ transient as a proxy for local depolarization to the spine neck length. As can be seen in Figure 3D, spine neck length does not correlate with the uEPSP Ca 2+ transient (n = 22/11/9 (spines/cells/animals), p = 0.49, r = 0.15, Spearman correlation, Figure 3D). Our experiments therefore do not permit the conclusion that there may be a strong VGCC mediated contribution to the uEPSP spine Ca 2+ transient in spines with a larger neck resistance. In sum, our data permits the conclusion that on the single spine level, spine VGCCs do not actively contribute to spine electrogenesis in the form of spine-specific Ca 2+ spikes.

Long-Term Interactions Between VGCCs and Ionotropic GLURs
bAPs may still contribute to AMPAR mediated transmission beyond the level of direct electrical interactions. This interaction would occur on the level of plasticity processes. Synaptic plasticity changes the synaptic weight on a longer time scale that outlasts the timeframe of individual stimuli (Korte and Schmitz, 2016). The amplitude of the bAP-Ca 2+ transient is inversely related to the induction of LTD induced by NMDAR activation (Hayama et al., 2013). In addition, bAP-burst mediated Ca 2+ transients and NMDARs interact during the induction of LTP (Kampa, 2006). A recent study has proposed a central role for bAP-mediated VGCC activation for the induction of synaptic plasticity (Tigaret et al., 2016).
We hypothesized that a spine's VGCC content reflected by the amplitude of bAP-Ca 2+ transients should correlate with synaptic strength if these parameters interact directly and significantly. In order to test whether there is a relationship between the AMPAR response in voltage clamp and the bAP-Ca 2+ transient, we measured the bAP-Ca 2+ transient and the uEPSC in individual spines ( Figure 3E). However, there is no significant correlation between the two parameters (n = 33/15/11 (spines/cells/animals), p = 0.95, r = 0.01, Spearman correlation, Figure 3E). So far, we used synaptic stimulation (Figure 2) and two-photon uncaging (Figure 3) to probe for AMPARs at the single spine level. Based on these sets of experiments, we conclude that fast AMPAR mediated synaptic transmission and VGCCs are functionally uncoupled.
In addition to synaptic AMPARs, NMDARs can also be bidirectionally modified by patterned synaptic stimulation (Rebola et al., 2010;Hunt et al., 2013). NMDARs and VGCCs are the main Ca 2+ sources during neuronal activity and together determine the amount of Ca 2+ entering the spine during synaptic activity. We therefore wanted to test whether the number of VGCCs on a spine could relate to a single spine's NMDAR pool.
As a measure of a spine's NMDAR pool, we switched to voltage clamp measurements using a Caesium-based intracellular solution and TTX to optimize voltage control. We measured AMPAR based uEPSCs (uAMPA) at −60 mV ( Figure 4A2) and NMDAR driven uEPSCs (uNMDA) at +40 mV ( Figure 4A3). The amplitude of the NMDAR uEPSC was measured 50-60 ms after stimulus onset to avoid contamination by AMPA currents. In order to accentuate the contribution of synaptic NMDARs to our uncaging evoked currents, the synaptic NMDARspecific coagonist Serine was added in these experiments (Papouin et al., 2012). Under these conditions optimized for voltage clamp recordings, bAPs cannot be evoked by current injection. To probe for single-spine VGCCs, we applied 30 ms voltage steps from −60 mV to 0 mV to measure the VGCC mediated Ca 2+ influx ( Figure 4A4). Similar to the results when using a K + -based intracellular solution, the AMPAR-mediated uEPSC was not related to the single-spine VGCC mediated Ca 2+ influx (n = 16/9/6 (spines/cells/animals), p = 0.09, r = −0.44, Spearman correlation, Figure 4B). When grouping the spines by the median of the VGCC mediated Ca 2+ influx, there was also no significant difference in AMPA-current between the small VGCC Ca 2+ response spines and the large VGCC Ca 2+ response spines (p = 0.51 (t-test)). However, there was a significant inverse correlation between the NMDAR mediated current at +40 mV and the VGCC mediated Ca 2+ influx (n = 16/9/6, p = 0.04, r = −0.52, Spearman correlation, Figure 4C). When grouping the spines by the median of the depolarization induced Ca 2+ transient, there was a significantly smaller NMDAR current in the group with the larger VGCC mediated Ca 2+ influx (p = 0.013; t-test).

Effect of Spine VGCC Enhancement on NMDARs
Spines with large bAP-Ca 2+ transients could therefore undergo a reduction in NMDARs. In our previous work, we demonstrated that suprathreshold activity results in long-lasting enhancement of bAP-Ca 2+ transients. We next wanted to test if bAP-Ca 2+ transients can be directly related to a reduction in NMDARs. In this experiment, we wanted to test if NMDARmediated currents evoked by extracellular synaptic stimulation can be modulated by period of bAP bursts previously demonstrated to enhance spine VGCCs (Johenning et al., 2015). A K + -gluconate based intracellular solution was used so that bAPs could be evoked. To relieve the Mg 2+ block of the NMDAR at hyperpolarized membrane potentials we used a low Mg 2+ extracellular solution. AMPARs were blocked with NBQX. We then stimulated synaptically using an extracellular stimulation electrode. Single presynaptic pulses do not saturate the postsynaptic spine NMDARs . This lack of saturation could be attributed to the recruitment of extrasynaptic NMDARs by repetitive stimuli (Pankratov and Krishtal, 2003;Harris and Pettit, 2008). For compatibility with our two-photon uncaging experiments, where we most likely also stimulated extrasynaptic NMDARs, we stimulated with double pulses and focussed our analysis on the second EPSC in a double pulse ( Figure 5A). We then compared the effect of a 5 min burst of bAPs to a control group that did not fire APs. The effect of the bAPs and the control interval are normalized to the pre-induction baseline.

DISCUSSION
Ca 2+ is a major player in long-term synaptic adaptations underlying plasticity. Depolarization of dendritic spines activates VGCCs, both by direct synaptic activation and by electrotonic spread of depolarization mediated by dendritic bAPs. Here, we demonstrate that the major VGCC subtypes contributing to bAP Ca 2+ transients in layer 2 cells of the MEC are Rand T-type channels. R-and T-type channel activation by synaptic activation of single spines contributes to the spine Ca 2+ response, but does not contribute to the depolarization underlying EPSPs.
In addition to direct interactions on a shorter time scale, VGCCs may also affect the strength of ionotropic glutamatergic signaling on a longer time scale by interfering with plasticity processes. While there is no relationship between spine VGCCs and AMPARs, VGCC-mediated transients are inversely correlated with the NMDAR current of individual spines. In addition, bAP-mediated Ca 2+ transients induce downscaling of NMDAR currents. These two findings demonstrate a dose-dependent global modification of the spine NMDAR content by VGCC mediated Ca 2+ influx on a longer timescale. Our study provides a direct link between global VGCC activation by dendritic backpropagation of APs and synaptic function. The activity dependent selective downscaling of NMDARs could serve to homeostatically stabilize glutamatergic synapses on active cells by increasing their plasticity threshold without scaling down AMPAR-mediated synaptic strength.

Effects of Depolarization Mediated VGCC Ca 2+ Influx on Synaptic Potentials
Previous studies have generated mixed results regarding the direct contribution of VGCCs to synaptic depolarization. Some reports using extracellular synaptic stimulation suggest that there is direct boosting of large compound EPSPs by VGCCs Gillessen and Alzheimer, 1997). Voltage imaging of EPSPs in dendritic spines suggests that there is no significant VGCC contribution to local synaptic depolarization (Palmer and Stuart, 2009;Popovic et al., 2015). Here, we utilized drugs that are highly selective for the two major postsynaptic VGCC subtypes in dendritic spines of layer 2 cells in the MEC (Figure 1). We could therefore directly test VGCC contribution to large compound EPSPs evoked by direct extracellular stimulation. This approach also takes cooperative effects between several synapses into account. Our results show that in MEC layer 2 cells, the amount of depolarizing currents added to the compound synaptic potential by spine VGCCs is too small to play a quantitatively significant role.
The absence of a T-and R-type VGCC block mediated effect on synaptic depolarization measured at the soma is also relevant with respect to another interaction between VGCCs and EPSPs: in CA1 pyramidal cells, R-type VGCCs initiate a negative feedback loop by activating Ca 2+ activated K + channels. This decreases the EPSP amplitude in single spines (Bloodgood and Sabatini, 2007;Giessel and Sabatini, 2011;Wang et al., 2014Wang et al., , 2015. In layer 5 prefrontal cortex neurons, compound EPSPs are dampened by this negative feedback loop, whereas single spine uEPSPs are not affected (Seong et al., 2014). In excitatory layer 2 neurons of the MEC, the VGCC-mediated negative feedback loop described in other principal neurons is absent even in compound EPSPs. It seems that the degree of coupling between VGCCs and Ca 2+ activated K + channels with an effect on the EPSP is cell-type specific and does not occur in all excitatory neurons in the forebrain.
So far, we focussed our discussion on instantaneous interactions between VGCCs and iGLURs when measuring EPSPs at the cell soma. However, this might have filtered out a VGCC-mediated boosting of the EPSP in spines (Harnett et al., 2012;Acker et al., 2016). Therefore, the limited electrogenic role of VGCCs when measuring synaptic activity at the soma does not rule out the local contribution of VGCCs to synaptic responses. This way, VGCCs could still shape and determine local EPSP Ca 2+ transients.
Our results using two-photon uncaging for glutamatergic stimulation are not compatible with local non-linear electrogenic effects of VGCCs in single spines of layer 2 MEC neurons. Our experimental results in APV clearly suggest that VGCCs contribute to EPSP Ca 2+ transients (Supplementary Figure S4C). This contribution needs to be considered from a quantitative perspective. We therefore tested whether the quantitative variability observed in spine VGCC content could explain the variability of the uncaging-evoked spine Ca 2+ response. From our experiments, we conclude that the amplitude of EPSP Ca 2+ transients is regulated independent of VGCC expression. This hints at a dominant role for the interaction of AMPARs and NMDARs in setting the magnitude of the EPSP Ca 2+ transient. These findings further argue against a significant local effect of VGCCs in synaptic depolarization and against local active processes like single-spine Ca 2+ spikes.

Long-Term Effects of Depolarization Mediated VGCC Ca 2+ Influx by bAPs
Alternatively, while contributing to EPSP Ca 2+ transients, the main functional relevance of VGCCs may only be mediated by bAP-Ca 2+ transients. This interaction would occur on a more prolonged timescale. Exclusive VGCC activation by electrotonic spread could permit for a distinction between local specific mechanisms in synaptically driven spines and global unspecific mechanisms in spines that are depolarized but not synaptically driven.
In synaptically driven spines, a recent study has identified a central role for bAP-mediated VGCC activation for the induction of synaptic plasticity (Tigaret et al., 2016). In this study, partial block of VGCCs with different pharmacological compounds could not block the induction of synaptic plasticity, only a cocktail of VGCC blockers incorporating R-, T-and L-type VGCCs had an effect. Our interpretation is that while coactivation of VGCCs per se is a prerequisite for plasticity induction, the range of VGCC-mediated [Ca 2+ ] for plasticity induction is rather large. This large safety margin for VGCC recruitment during plasticity induction is reflected by our finding that AMPAR mediated synaptic currents are independent of the physiological range of the VGCC mediated bAP-Ca 2+ transient and the depolarization mediated spine Ca 2+ influx measured in voltage clamp (Figures 3, 4).
What happens to synapses that are not synaptically activated but experience bAP-Ca 2+ transients when a neuron receives suprathreshold activation? Our results suggest that spine AMPARs and, by extension, AMPAR plasticity are independent of a spine's VGCC content and the related bAP-Ca 2+ transient. In addition to AMPARs, NMDARs are involved in the induction of both synaptic depression and potentiation (Malenka and Bear, 2004). In CA3 neurons, bidirectional hebbian plasticity of NMDARs has been demonstrated to increase or decrease the threshold for LTP induction (Rebola et al., 2011;Hunt et al., 2013). In the context of homeostatic plasticity, NMDARs are globally upregulated by long phases of reduced network activity and globally downregulated by long phases of enhanced network activity in neuronal cultures (Watt et al., 2000). Here, we contribute a new mechanism of global activity-dependent NMDAR regulation: the activitydependent enhancement of bAP-Ca 2+ transients by bAP mediated depolarization results in the downregulation of NMDAR function. We infer this from two observations: first, on the single spine level the NMDAR content is inversely correlated with the VGCC mediated depolarization associated Ca 2+ influx (Figure 4). Accordingly, the degree of NMDAR downscaling in a spine seems to depend on the amplitude of the depolarization associated Ca 2+ influx. Second, AP bursts result in downscaling of NMDARs when compared to controls that do not experience AP firing during the induction period ( Figure 5). Based on methodological constraints of two-photon uncaging, we cannot differentiate synaptic NMDARs from spine NMDARs located extrasynaptically (Supplementary Figure S4) when measuring uEPSPs and uEPSCs. We added 10 µM of the NMDAR-coagonist Serine in our voltage-clamp experiments when we quantified the relationship between step depolarization mediated Ca 2+ influx and NMDAR currents (Figure 4). It has previously been published that Serine predominantly activates synaptic NMDARs (Papouin et al., 2012). In addition, experiments in Figure 5 demonstrate that synaptic NMDAR-EPSCs evoked by synaptic stimulation are functionally reduced by AP firing. This implies a causal link between AP firing, dendritic backpropagation and downscaling of synaptic and extrasynaptic NMDARs on spines. So far and to the best of our knowledge, the bAP-Ca 2+ transient is the only biochemical transducer of bAP mediated spine depolarization. The most parsimonious explanation therefore is that bAP-Ca 2+ transients induce downregulation of spine NMDARs in a dose-dependent manner.
In this context, we would like to compare LTP induction, synaptic NMDAR content and bAP mediated Ca 2+ influx between layer 2 cells in the MEC and hippocampal CA1 neurons. Layer 2 cells in the MEC have larger bAP-Ca 2+ transients than CA1 pyramids (Johenning et al., 2015). Interestingly, this is accompanied by a smaller NMDAR/AMPAR ratio and difficulties in LTP induction in MEC layer 2 cells of rats in the same age range (Deng and Lei, 2007).
Future experiments will have to show whether the downregulation of NMDARs indeed has a metaplastic effect on synaptic plasticity by shifting the induction threshold and occurs homeostatically in response to reduced or elevated activity levels of individual neurons in a network. In this case, downregulation of NMDARs by bAPs may serve as a homeostatic mechanism specific for synapses that have not been involved in the suprathreshold activation of individual neurons. This heterosynaptic adaptation would result in potentiation of only a subset of repeatedly activated synapses, while less frequently activated synapses would have a higher plasticity threshold. This type of metaplasticity would mainly affect dendritic spines that have not been specifically activated. We therefore propose a stabilizing mechanism that does not interfere with information transfer. In this model, silent neurons would keep high levels of spine NMDARs, constituting a pool of plasticity competent neurons that can be integrated in the network by novel activation patterns.

AUTHOR CONTRIBUTIONS
AKT and FWJ performed experiments and analyzed the data. AKT, FWJ, DS, GK and BR conceptualized the research, designed experiments, interpreted the data, wrote the manuscript and prepared the figures. GK and BR contributed reagents.