Introduction

The neo- or isocortex is a morphologically homogeneous brain region that covers highly diverse functions within its specialized areas, ranging from early sensory processing and motor control up to higher order associations, cognition and consciousness (Douglas and Martin, 2004; Harris and Shepherd, 2015). Notably, all of these diverse functions are executed by the same archetypes of neurons and synapses. However, depending on the area, they are engaged to different degrees in lower or higher order processing. Whether area-specific functional differences in the mature neocortex are associated with or even arise from differences in the functional presynaptic nanoarchitecture of the same principal types of synapses is currently unclear.

At presynaptic active zones, the coupling distance between voltage-gated Ca²⁺ channels (VGCCs) and transmitter-filled synaptic vesicles (SVs) is a major determinant of key synaptic properties, including speed, efficacy and reliability of synaptic transmission (Rozov et al., 2001; Fedchyshyn and Wang, 2005; Bucurenciu et al., 2008; Baur et al., 2015; Bornschein et al., 2019a; Chen et al., 2024), and also short-term plasticity, although the latter can be obscured by vesicle replenishment (Miki et al., 2016; Doussau et al., 2017; Bornschein et al., 2019b; Lin et al., 2022). Studies at different synapses in different parts of the brain suggest that excitatory synapses engaged in reliable information transfer switch from loose microdomain to tight nanodomain coupling during development (Fedchyshyn and Wang, 2005; Baur et al., 2015; Nakamura et al., 2015; Bornschein et al., 2019a). This includes synapses between layer 5 pyramidal neurons (L5PNs) in the primary somatosensory cortex (S1), an area of early sensory processing (Bornschein et al., 2019a). On the other hand, loose microdomain coupling was found in the mature brain to date only at a highly plastic hippocampal synapse and has been suggested to provide a molecular framework for presynaptic plasticity (Vyleta and Jonas, 2014). However, if microdomain coupling also represents a synaptic correlate of higher order neocortical function or if it restricted to specific and highly plastic synapses in the hippocampus is currently unclear. Therefore, here we tested the hypothesis that loose microdomain coupling is the synaptic correlate of higher order functions of the mature neocortex, whereas tight nanodomain coupling is the synaptic correlate of earlier processing stages. Our data show that in the mature cortex the same types of pyramidal neuron synapses operate with loose microdomain coupling if they are located in the prefrontal cortex (PFC), whereas in S1 they operate with tight nanodomain coupling. Thus, our results corroborate the hypothesis and suggest microdomain coupling as a synaptic correlate of higher order function in the neocortex.

Results

Synaptic efficacy, reliability and short-term plasticity differ between PFC and S1

We focused on the lateral prefrontal cortex (PFC) and medial prefrontal cortex (mPFC) as areas of higher order processing and on primary somatosensory cortex (S1) as an area of lower order processing. Within these areas we investigated excitatory synapses between pyramidal neurons (PNs), which are the principal building blocks of the neocortex. We first analyzed the properties of synapses connecting neighboring layer 5 PNs (L5PNs) in mature PFC (P21-26) by performing paired whole-cell patch-clamp recordings and compared their properties to those we previously reported for the same synapses and age in S1 (Bornschein et al., 2019b; Bornschein et al., 2019a) (Figure 1A). We found that the efficacy of transmission, i.e. the size of a unitary EPSC evoked by a single action potential (Rozov et al., 2001), was significantly lower in PFC (8 pA, 5-15 pA) than in S1 (29 pA, 18-53 pA), whereas the synaptic failure rates were significantly higher in PFC (0.16, 0.08-0.28) than in S1 (0, 0-0.03; Figure 1B-D). In addition, we found that the synaptic delays (2.46 ms, 2.20-2.94 ms) as well as their standard deviation (SD_Delay: 0.60 ms, 0.49-0.81 ms) were significantly larger in PFC than in S1 (delay: 1.99 ms, 1.67-2.25 ms; SD_Delay: 0.24 ms, 1.14-0.37 ms; Figure 1E, F), indicating that transmitter release is less tightly coupled to the time of the action potential in PFC than in S1 (Bullmann et al., 2024). Together these data show that synaptic speed, efficacy and reliability during single action potentials are lower in PFC than at the same synapses in S1.

Figure 1.

When pairs of action potentials were applied at high frequency (Figure 1G-I), differences in the synaptic success rates were associated with a significant difference in the paired pulse ratios (PPR) between PFC and S1. In PFC paired pulse facilitation (PPF, 1.18, 0.98-1.52) prevailed associated with increased success rates in the 2^nd pulse (1_x: 0.81, 0.73-0.91; x_1: 0.91, 0.82-0.97). In contrast, in S1 paired pulse depression (PPD, 0.78, 0.69-0.90) was found with high initial success rates (1_x: 1.00, 0.97-1.00; x_1: 0.99, 0.95-1.00). We observed the difference in short-term plasticity over a range of interstimulus intervals (ISIs) of 5 to 50 ms (SI Appendix, Figure S1A, B). Thus, the differences in synaptic efficacy and reliability between S1 and PFC during the first action potential correlate with differential short-term plasticity, with the emphasize in S1 being on the first transmission process while in PFC facilitation emphasized the subsequent release.

Next, we investigated transmission at synapses between L2/3 and L5PNs, using minimal extracellular stimulation near somata of L2/3PNs (Figure 1A). We found that also L2/3-L5PN synapses in PFC showed PPF (1.36, 1.14-1.88) as opposed to PPD in S1 (0.87, 0.72-0.99; Figure 1J, K). We observed this over a range of ISIs up to 100 ms (SI Appendix, Figure S1C, D). In addition, at PFC synapses the SD of the synaptic delays (0.46, 0.29-0.66) was again significantly larger than in S1 (0.25, 0.15-0.37; Figure 1L). Furthermore, we found larger EPSC decay time constants (τ_decay) in PFC compared to S1, possibly reflecting a contribution of different postsynaptic glutamate receptors (SI Appendix, Figure S2A, B). Indeed, application of the N-methyl-D-aspartate receptor (NMDAR) antagonist APV removed the difference in τ_decay between PFC and S1 (SI Appendix, Figure S2C-F). We extended the paired pulse experiments at L2/3-L5PN synapses in PFC to even older mice (P90-100) and to the mPFC (P21-26). We found that PPF persisted in the older age window (1.25, 1.14-1.41) and that also the synapses in mPFC showed PPF (1.14, 1.03-1.42; Figure 1K).

Together these results from glutamatergic synapses onto L5PNs show significant differences in the synaptic delays, efficacy, reliability and short-term plasticity between PFC and S1. The results would be consistent with area-specific differences in the functional coupling distances between VGCCs and SVs at the presynaptic active zones.

EGTA sensitivity of release is higher in PFC than in S1

To directly test for differences in the coupling distances, we investigated the sensitivity of release to synthetic Ca²⁺ chelators (Figure 2). The sensitivity of release to low millimolar concentrations of the kinetically slow Ca²⁺ chelator EGTA is a standard indicator of loose coupling (Eggermann et al., 2012). First, we performed recordings at pairs of L5PNs in mature PFC (Figure 2A, B). After 10 min of stable baseline recordings, the presynaptic neuron was patched a second time and equilibrated for 30 min with 10 mM EGTA-containing pipette solution. In these paired recordings we found a significant effect of EGTA onto release in PFC (0.67, 0.63-0.73; control: 1.02, 0.75-1.20). In identical experiments in S1, we had previously found release to be insensitive to the same concentration of EGTA (0.94, 0.71-1.02) (Bornschein et al., 2019a). These data provide evidence that the synapses between L5PNs in mature PFC do indeed operate with loose coupling, in contrast to their counterparts in S1.

Figure 2.

We proceeded by comparing the EGTA sensitivity of release at L2/3-L5PN synapses between S1 and PFC, using extracellular stimulation and incubation with Ca²⁺ chelator-AM compounds (Figure 2C-F) (Baur et al., 2015; Kusch et al., 2018). Following 10 min of stable baseline recordings, slices were perfused with ACSF that either contained the DMSO/Pluronic containing solvent alone (control) or 10 µM of dissolved EGTA-AM or BAPTA-AM. The fast BAPTA interferes with release irrespective of the coupling topography, thereby, providing a positive control for the proper functioning of the AM-method (Eggermann et al., 2012). EPSC amplitudes were quantified during a subsequent 10 min period with perfusion with normal ACSF. During this test period, BAPTA significantly reduced EPSC amplitudes both in PFC (0.28, 0.12-0.32; control: 0.99, 0.84-1.02) and in S1 (0.08, 0-0.18; control: 0.85, 0.83-0.92). On the other hand, EGTA significantly reduced EPSC amplitudes only in PFC (0.57, 0.50-0.70; 58% of control) but not in S1 (0.85, 0.81-0.90; 100% of control).

In S1 we previously found that coupling switches from loose to tight during postnatal development between ∼P10 and ∼P20 (Bornschein et al., 2019a). To probe for a developmental retardation of this process in PFC, we extended the EGTA-AM experiments at L2/3-L5PN synapses to the older age-window of P90-100. Also, in these experiments we found a significant sensitivity of release to EGTA (0.76, 0.65-0.78; Figure 2D; SI Appendix, Figure S3A), indicating that loose coupling is a functionally persistent feature of synapses in PFC. Finally, we also extended the EGTA-AM experiments to synapses in mPFC and again found significant effects of EGTA on the EPSC amplitudes (0.65, 0.49-0.75; Figure 2D; SI Appendix, Figure S3B).

Together the results obtained with the slow Ca²⁺ chelator EGTA suggest that major glutamatergic synapses in mature PFC operate with loose coupling in contrast to tight coupling in the mature S1. The data further suggest that loose coupling is a developmentally persistent property of synapses in PFC. Finally, they indicate that the above functional differences between synapses in S1 and PFC originate from the differences in Ca²⁺ influx-release coupling.

Release probabilities are similar in PFC and S1

To test whether the release probabilities (p_N) of the differentially coupled release sites in PFC and S1 differ, we performed multiple probability fluctuation analysis (MPFA) (Clements and Silver, 2000; Brachtendorf et al., 2025) at different extracellular Ca²⁺ concentrations ([Ca²⁺]_e; Figure 3) (Bornschein et al., 2019a). The parabolic MPFA-fits to variance-mean plots from recordings of pairs of L5PNs in PFC yielded a median p_N of 0.50 (0.37-0.55; Figure 3A, D). This value was slightly but not significantly smaller than p_N of 0.66 (0.55-0.70), which we quantified previously for this connection in S1 (Bornschein et al., 2019a).

Figure 3.

We also quantified p_N at L2/3-L5PN synapses, both in PFC and in S1, using MPFA with extracellular stimulation (Baur et al., 2015). We found that p_N was as high as at the L5PN-L5PN synapses in both areas with values (PFC: 0.62, 0.47-0.67; S1: 0.48, 0.39-0.58) being not significantly different between areas (Figure 3B, C, F). Interestingly, in our previous study (Bornschein et al., 2019a) we found that the developmental switch from loose to tight coupling at L5PN synapses in S1 also had no significant impact on p_N.

From MPFA in paired recordings it is possible to estimate the number of release sites (N) and the quantal size (q), additionally. N was significantly smaller in PFC than in S1 being only 2.1 (1.9-2.3) rather than 8 (3-19). q tended to be smaller in PFC (6 pA, 5-7 pA) than in S1 (9 pA; 8-11 pA) and the difference became significant when evoked quantal EPSCs recorded in low [Ca²⁺]_e were compared (PFC: 4 pA, 4-5 pA; S1: 9 pA, 7-11 pA). We consider it unlikely that postsynaptic receptor saturation or desensitization influenced the determination of quantal parameters since experiments with the competitive low-affinity glutamate receptor antagonist γ-DGG (γ-D-glutamylglycine) (Chanda and Xu-Friedman, 2010) performed at the S1 synapses had revealed no signs of this (p_N=0.58, 0.52-0.60; N=8, 8-12; Figure 3D). Due to the smaller size of the EPSCs, testing on the PFC synapses was not possible. Estimating EPSC amplitudes (EPSC = N p_N q; PFC, 6 pA; S1, 48 pA) and failure rates (F = (1-p_N)^N; PFC, 0.25, S1, 0.0001) from the quantal parameters yielded values that were well suited to explain the area-specific differences in synaptic efficacy during individual action potentials (Figure 1).

Presynaptic Ca²⁺ signals are similar in PFC and S1

For a thorough interpretation of the above results knowledge about the presynaptic Ca²⁺ dynamics in the different synapses is required. We performed dual-dye two-photon Ca²⁺ imaging (Sabatini et al., 2002) at presumed presynaptic boutons located on axon collaterals of L5PNs in PFC and in S1. We quantified single action potential-mediated elevations in green over red fluorescence signals (ΔG/R) that were converted to increases in intracellular calcium (Δ[Ca²⁺]_i) (Figure 4) (Bornschein et al., 2019a). We found that neither the Ca²⁺ transients nor the decay time constants were significantly different between the cortical areas (Figure 4A-F; PFC: G/R=0.25, 0.23-0.31; Δ[Ca²⁺]_i=279 nM, 244-453 nM; basal [Ca²⁺]_i=30 nM, 21-34 nM; τ_decay,1=9 ms, 6-19 ms; τ_deay,2=91 ms, 75-120 ms; S1: G/R=0.29, 0.25-0.39; Δ[Ca²⁺]_i=238 nM, 208-303 nM; basal [Ca²⁺]_i=22 nM, 13-38 nM; τ_decay,1=9 ms, 3-15 ms; τ_decay,2=86 ms, 63-212 ms). The similarity in the presynaptic Ca²⁺ transients of boutons in PFC and S1 suggests that differences in the presynaptic Ca²⁺ influx are unlikely to be the cause for the differential Ca²⁺ chelator effects. Rather, they support the view that these differences result from loose coupling in PFC and from tight coupling in S1.

Figure 4.

Estimate of the coupling distance in PFC verify loose microdomain coupling in PFC

To derive a quantitative picture of the coupling topography in L5PN synapses in PFC we used numerical computer simulations that were based on and constrained by the experimental data. First, we simulated the measured Ca²⁺ transient with the indicator dye present in the simulation (Figure 5A). Subsequently, the indicator dye was removed and a release sensor model for Synaptotagmin-1-triggered release from L5PN boutons was included in the simulations (Figure 5B, C; SI Appendix, Figure S4) (Bornschein et al., 2025). VGCCs were assumed to form a ring like structure around a vesicle (Figure 5D), similar to the release site topography in young S1 (Bornschein et al., 2019a). Release rates as reported by the sensor model were integrated over time to yield the p_N values. The coupling distance between the sensor and the ring of VGCCs as well as the number of VGCCs forming the ring were iteratively varied until the simulation correctly predicted the experimental p_N values from MPFA obtained under control conditions and in the presence of EGTA. By this procedure we quantified an average coupling distance of 48-51 nm between a microdomain of several VGCCs and the release sensor (Figure 5B; SI Appendix, Figure S4), similar to the estimate for these synapses in immature S1 (Bornschein et al., 2019a). Thus, a microdomain model similar to immature S1 is suitable to predict the data from mature PFC, whereas the same synapses in mature S1 operated with Ca²⁺ signaling nanodomains in which only 1-3 VGCCs at distances of 11-16 nm trigger fusion (Bornschein et al., 2019a; Bornschein et al., 2025).

Figure 5.

To experimentally test whether overlapping Ca²⁺ domains from several VGCCs trigger release in PFC (microdomain), we analyzed the effects of the unspecific VGCC blocker Cd²⁺ onto the PPR. If fusion is triggered by such VGCC microdomains, application of a subsaturating concentration of Cd²⁺ will increase the PPR, while it will leave PPR unaffected if only a single or few VGCCs trigger release (Hefft et al., 2002; Bucurenciu et al., 2008; Scimemi and Diamond, 2012; Baur et al., 2015; Bornschein et al., 2019a). In paired pulse experiments, at L2/3-L5PN synapses in PFC we indeed found a significant increase in the PPR following the application of 5 µM Cd²⁺ (rPPR=1.20±0.06), whereas the average PPR remained unaffected in S1 (rPPR=1.01±0.10; Figure 5E, F). Thus, our data and simulations suggest that the nanotopography of PN synapses in mature PFC is reminiscent of the same synapses in young S1.

Discussion

Our results provide evidence that the same archetypes of glutamatergic synapses show area-specific differences in their functional release site nanoarchitectures in the mature neocortex. They suggest loose microdomain coupling as a synaptic correlate of higher order neocortical functions.

The differences in coupling gave rise to larger synaptic delays in PFC compared to S1 and were associated with differences in synaptic reliability and efficacy, whereas the p_N values were not significantly different between areas. A major determinant of p_N is the size of the Ca²⁺ signal at the release sensor. Accordingly, tightening of coupling can increase p_N (Eggermann et al., 2012; Baur et al., 2015; Bornschein and Schmidt, 2019). However, large Ca²⁺ elevations at the release sensor can also be obtained with Ca²⁺ microdomains. At L5PN-L5PN synapses in S1 the developmental switch from microdomain to nanodomain coupling had no effect on p_N (Bornschein et al., 2019a). At the calyx of Held (Iwasaki and Takahashi, 2001; Taschenberger et al., 2002; Fedchyshyn and Wang, 2005; Koike-Tani et al., 2008; Nakamura et al., 2015) and at cerebellar basket cell to Purkinje cell synapses (Chen et al., 2024) p_N even decreased during development, whereas coupling got tightened. The decrease in p_N was compensated by an increase in N at these synapses, which increased their reliability and efficacy during development. We found N to be larger in S1 than in PFC. Thus, it appears that differences in coupling are typically associated with other characteristic changes in the release parameters. Overall, this suggests that reliable synapses switch from an initial state with high plasticity to a matured state with high reliability. On the other hand, PN synapses in the PFC appear to remain in a more juvenile state that favors plasticity over reliability.

We observed differences in short-term plasticity that suggest that S1 emphasized the first transmission process, whereas in PFC facilitation emphasized the subsequent release. The mechanisms of short-term plasticity are complex and include the regulation of p_N, the speed of vesicle replenishment and the recruitment of release sites (Neher and Brose, 2018; Schmidt, 2019; Neher, 2023; Brachtendorf et al., 2025). We consider the differential emphasize on the available presynaptic efficacy between PFC and S1 as a signature of higher plasticity and flexibility in the processing of information in PFC compared to S1.

Our results have implications for long-term plasticity at neocortical PN synapses. Both, long-term potentiation (LTP) and depression (LTD) have a presynaptic locus at these synapses and likely involve changes in p_N, but the final steps that alter release are not well understood (Feldman, 2009; Castillo, 2012; Feldman, 2012). A recent study on L5PN synapses in S1 showed an increase in the number of primed vesicles as a major mechanism of LTP (Weichard et al., 2023). PN synapses in S1 are already tightly coupled and therefore a further decrease in the coupling distance is an unlikely mechanism of LTP. However, this does not exclude that in S1 LTD induction breaks tight coupling and vice versa that LTP in PFC involves tightening of coupling. In addition, in PFC we found an increased contribution of NMDAR on the postsynaptic site. Thus, these synapses have increased potential for plasticity on the pre- and the postsynaptic site. During induction of long-term plasticity, the postsynaptic site signals back to the presynaptic site via retrograde messengers (Regehr et al., 2009) and loose coupling will provide the substrate for enhanced regulatory capacity (Vyleta and Jonas, 2014).

In general, loose coupling provides more flexibility to regulate [Ca²⁺]_i at the release sensor (Vyleta and Jonas, 2014). Prefrontal networks carry out higher-order computations that transform sensory information and memory to perform flexible output such as delayed response, inference and planning (Narayanan et al., 2025). Our findings suggest that loose coupling provides a synaptic foundation for such flexible neocortical network functions.

Methods

Slice preparation

C57BL/6J mice at P21-26 of either sex were used in this study. All experiments were performed in accordance with the guidelines for the welfare of experimental animals issued by the European Communities Council Directive (2010/63/EU) and with the German Protection of Animals Act (Tierschutzgesetz). Mice were bred in the animal facility of the Medical Faculty of Leipzig University and were housed in individually ventilated cages in a specific pathogen free environment and in a 12h/12h light dark cycle with access to food and water ad libitum. Experiments were approved by the animal welfare office of the University Medical Center, Leipzig, as well as the local governmental authorities (Landesdirektion Leipzig, registration numbers T10/20, T05/21-MEZ).

C57BL/6J mice at P21-26 of either sex were decapitated under deep Isoflurane (Curamed) inhalation anesthesia. The brain was excised rapidly and placed in cooled (0-4°C) artificial cerebrospinal fluid (ACSF) containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 26 NaHCO₃, 1 MgCl₂, 2 CaCl₂, and 20 glucose, equilibrated with 95% O₂ and 5% CO₂ (pH 7.3-7.4). Coronar neocortical slices (150-250 μm thick) were cut from the PFC or S1 region (Figure 1A) with a vibratome (HM 650 V, Microm). For some paired recordings between L5PNs and for calcium imaging experiments parasagittal slices (200 µm) were prepared from PFC region. Slices were incubated for 30 min at 35°C and subsequently stored at room temperature. For experiments, slices were transferred to a recording chamber and continuously perfused with ACSF (2-3 ml per min, supplemented with 10 μM (-)-bicuculline methiodide (Tocris) at 31-33°C. Unless stated otherwise, chemicals were from Sigma-Aldrich.

Electrophysiological recordings

Patch-clamp recordings from L5PNs were established as described in detail previously (Bornschein et al., 2019a; Bornschein et al., 2025). Patch pipettes were prepared from borosilicate glass (Hilgenberg) with a PC-10 puller (Narishige) and had final resistances of 6-8 MΩ when filled with the following standard pipette solution (in mM): 150 K-gluconate, 4 NaCl, 3 MgCl₂, 3 Na₂ATP, 0.3 NaGTP, 0.05 EGTA, 10 KHEPES, dissolved in purified water. The pH was adjusted to 7.3 with KOH. Recordings were performed under optical control (BX51WI, Olympus), using an EPC10/2 amplifier and Patchmaster software (version v2x90.2, HEKA). EPSCs were recorded in the whole-cell configuration at a holding potential (V_hold) of -90 mV (online corrected for a liquid junction potential of 16 mV), filtered at 5 kHz and sampled at 10 kHz. Series resistance (R_s) was continuously compensated to a fixed value between 10 and 15 MΩ. The average uncompensated R_s was 18±1 MΩ in L5PN-L5PN paired recordings (n=41 cells, mean ± SEM) and 23±1 MΩ in L2/3-L5PN recordings (n=133 cells). Holding current (I_hold) was also monitored continuously and was <180 pA (-134±16 pA in L5PN-L5PNs; -168±11 in L2/3-L5PNs). Presynaptic L5PNs were stimulated in on-cell configuration (200-500 mV, 1-2 ms).

For chelator wash-in experiments, presynaptic neurons were stimulated in the whole-cell configuration after re-patching with a pipette solution supplemented with 10 mM EGTA (K-gluconate concentration was reduced to 135 mM to adjust osmolarity). EPSCs were evoked every 10 s for at least 30 min (Figure 2). Averaged EPSC amplitudes were calculated before (≥10 min baseline) and 20-30 min after repatching (test period). For recordings from L2/3-L5PN connections ACSF-filled patch pipettes were placed close to the somata of PNs in L2/3 and extracellular stimulation was performed using an ISO-Stim 01 DPI (NPI electronics). Minimal stimulation (0.5 to 5 V) was used and yielded EPSC amplitudes of ∼40 pA (47±3 pA, n=131 pairs), making it likely that we investigated transmission mainly between single or few L2/3PNs and L5PNs in these experiments. Following 10 min of stable recordings in normal ACSF (baseline), the bath solution was exchanged for ACSF containing either 0.1% DMSO and 0.01% Pluronic (control solution), 10 µM EGTA-AM or 10 µM BAPTA-AM in DMSO/Pluronic for 30 min (incubation period) and thereafter replaced by normal ACSF again for another 10 min (test period). Averaged EPSC amplitudes were calculated during baseline and test period. The time courses of chelator effects on EPSC amplitudes were analyzed by binning EPSC amplitudes within 2 min intervals of recording time to average amplitudes. Chelator effects were expressed as averaged EPSC amplitudes during test period normalized to the corresponding baseline values.

VGCCs were inhibited by bath-application of ACSF containing a subsaturating concentration of 5 µM CdCl₂ (Figure 4J). NMDA receptors were blocked by the selective antagonist 2-Amino-5-phosphonovaleriansäure (APV, 50 μM; SI Appendix, Figure S2). For assessing the effects of the individual blockers, EPSCs were recorded in ACSF for at least 10 min (10 s intervals), subsequently the blocker was perfused and the blocker effects were quantified after a stable block had been established (at least 5 min test period).

Quantification of quantal synaptic parameters

Quantal synaptic parameters were estimated from parabolic fits to mean-variance (MV) plots of EPSC amplitudes recorded at different [Ca²⁺]_e (1, 2, and 5 mM, ≥30 repetitions per concentration; [Mg²⁺]_e was correspondingly adjusted to 2, 1, 0 mM, respectively), assuming binominal release statistics (Clements and Silver, 2000; Scheuss et al., 2002; Silver, 2003). The recordings always started in 2 mM [Ca²⁺]_e. The variance of EPSCs (σ²) was calculated according to ref. (Scheuss et al., 2002), plotted against the averaged amplitudes (I), and fitted by a parabola of the form:

where q is the quantal size, N a binominal parameter, and CV_I and CV_II the coefficients of intrasite and intersite quantal variability, assumed to be 0.3 (Clements and Silver, 2000). The fits were constrained to pass through zero. The variance of the variance was calculated according to ref. (Meyer et al., 2001). Stationarity of EPSC amplitudes was established when, after full exchange of the bathing solutions (≥5 min), EPSC amplitudes no longer showed a tendency to increase or decline. For MV analysis in L2/3- L5PN connections, the bath solution was supplemented with the competitive AMPA receptor antagonist kynurenic acid (Kyn, 0.25 mM), which relieves their desensitization and saturation (Neher and Sakaba, 2001) and with 50 µM APV to prevent NMDAR activation (Figure 3). Due to the small size of EPSCs in L5PN-L5PN connections no blocker was used in these recordings. The size of qEPSCs was quantified from evoked miniature EPSCs recorded in low [Ca²⁺]_e (1 mM) and high [Mg²⁺]_e (2 mM).

Ca²⁺ imaging

Action potential-evoked (current injection of 1-4 nA for 1-2 ms) fluorescence changes were recorded at boutons located on axon collaterals of L5PNs as described previously (Bornschein et al., 2019a; Bornschein et al., 2025). L5PNs were filled with EGTA-free, Fluo-5F (200 µM, Invitrogen) and Alexa-594 (50 µM, Molecular Probes) containing pipette solution via somatic whole-cell patch-pipettes. Cells were dialyzed for at least 20 min to yield sufficient equilibration with the dyes. Volume averaged fluorescence signals were recorded in point mode at 500 kHz temporal resolution and subsequently binned to 500 Hz, using a custom-build two-photon microscope based on a Fluoview-300 scanner (Olympus), a 60x/0.9

N.A. objective, a mode-locked Ti:sapphire laser (Tsunami, Newport-Spectra Physics, set to a center wavelength of 810 nm), and a Pockels cell (350-80 KD^*P, Conoptics). The fluorescence background was determined in a subsequent point mode recording performed near the boutons under investigation. Typically, at least 5 boutons per L5PN were recorded and averages per cell were calculated. The fluorescence signals were filtered (HC647/75, Semrock HC525/50, 720-SP, AHF), detected by two external PMT modules (H7422-40, Hamamatsu; PMT-02M/PMM-03, NPI electronics) monitoring red and green epifluorescence, respectively at fixed PMT voltages, and digitized with the Fluoview system. The Ca²⁺-dependent green fluorescence was normalized to the Ca²⁺-insensitive red fluorescence and expressed as background-corrected ΔG/R (Sabatini et al., 2002). ΔG/R signals were converted to changes in [Ca²⁺]_i based on an in vitro quantification of the K_D of Fluo-5F (439 nM) in our pipette solution, adjusted with CaEGTA and K₂EGTA (100 mM stock solutions, 10 mM HEPES added) to contain different free Ca²⁺ concentrations that were calculated with the MaxChelator (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/CaMgATPEGTA-NIST.htm). R_min and R_max values were measured after each successful recording in sealed patch pipettes that were placed in the slices near the recorded cells. The pipettes contained the EGTA-free, Fluo-5F- and Alexa-594-containing pipette solution with either 0 mM CaCl₂ and 10 mM K₂EGTA (zero Ca²⁺ solution) or 20 mM CaCl₂ (high Ca²⁺ solution).

Modeling

The kinetic gating scheme of the VGCCs and the reaction schemes for Ca²⁺ binding to buffers and transmitter release were converted to ordinary differential equations (ODEs). The system of ODEs was numerically solved using “NDSolve” or “NDSolveValue” of Mathematica 14 (Wolfram) as described previously (Schmidt et al., 2013; Bornschein et al., 2019a; Bornschein et al., 2025). Spatial resolution was achieved by placing the ODEs in concentric hemi-shells (1 or 2 nm thickness) that were coupled via diffusion.

The presynaptic action potential was modeled with a half-duration of 0.2 ms, τ_rise of 0.1 and τ_decay of 0.2 ms (Borst et al., 1995). Voltage dependent activation of VGCCs was simulated using the gating scheme for P/Q-type channels, a single channel conductance of 2.5 pS, and a Ca²⁺ equilibrium potential of 130 mV (Li et al., 2007). The model included ATP as mobile buffer and immobile endogenous buffers with a buffer capacity (κ_E) of 25 (Tran and Stricker, 2018). Model parameters are given in Table S1 (SI Appendix).

The model for each channel consisted of five closed (C) and one open (O) state. The transition between the first five steps was voltage dependent and the last step voltage independent:

a and b are the rate constants for transitions between C₄ and O, and α_i(V) and β_i(V) are the voltage dependent forward and backward transition rates that are given by:

where α_i,0 and β_i,0 are the forward and backward rate constants at 0 mV and k_i a slope factor.

Ca²⁺ binding to all buffers (B) was simulated by second order kinetics with forward and backward binding rate constants k_on and k_off, respectively:

Diffusion of all species (X) was simulated according to

where D_X is the diffusion coefficient, A the surface, V the volume and r the radius of a shell (n). Ca²⁺ was cleared by a linear, surface-based extrusion mechanism driven by the difference between [Ca²⁺]_i and the resting [Ca²⁺]_i (30 nM according to the imaging data).

For fitting the model to the average experimental Ca²⁺ transient, the number of VGCCs was adjusted, yielding 33 VGCCs (Figure 4F). The other free parameter of the simulation was the extrusion rate, which however affected only the later phase of the decay of the transient. During fitting, the model included the Ca²⁺ indicator dye. Fitting was first performed in a single compartment model (Helmchen and Tank, 2005). Subsequently, fitting was performed in the spatially resolved model with the numbers of VGCCs remaining unaltered and [Ca²⁺]_i as reported by the dye was calculated from the concentration of the Ca²⁺- dye complex assuming equilibrium conditions (Schmidt et al., 2003). Since L5PNs are characterized by an immobile and small κ_E with unknown binding kinetics, a slow (EB_s) and a fast immobile endogenous buffer (EB_f) were assumed (Helmchen et al., 1996; Ohana and Sakmann, 1998; Tran and Stricker, 2018). The fractions of EB_s and EB_f were adjusted to optimize the fit to the measured Ca²⁺ transient and to sum up to κ_E of 25. Following fitting, the indicator dye was removed from the model and the recently developed model for release from L5PN terminals was included (Bornschein et al., 2025). Either the “Syt1 priming model” (Equation 6; Figure 4) or the “Syt1 model” (Equation 7; SI Appendix, Figure S4) were placed at varying distances from the Ca²⁺ sources.

In the priming model (Equation 6), a single Ca²⁺-dependent priming step with Michaelis-Menten kinetics (K_M) and a priming rate k_prim is assumed. V is the vesicular release sensor of a primed vesicle (Equation 6, 7) that binds a maximum of 5 Ca²⁺ ions. k_on and k_off are the forward and backward Ca²⁺ binding rate constants, βⁱ (i=0-4) is a cooperativity factor, and γ the fusion rate. p_N was obtained by integrating the release rates over time.

A ring like arrangement of the VGCCs around a vesicle was assumed (Figure 4I; ring 1 model in ref. (Bornschein et al., 2019a)). The number of VGCCs driving release was continuously reduced from the maximal value of 33 (Figure 4F) and the coupling distances at which the simulated p_N values matched the experimentally determined p_N were obtained. Subsequently, EGTA was included and the consistency between the experimentally quantified EGTA effect and the simulated EGTA effect at a given coupling distance was determined (Figure 4G). The best match between experiments and simulation was obtained for 9 VGCCs at a coupling distance of 51 nm (Figure 4H).

Quantification and statistical analysis

Data from L5PN-L5PN pairs in PFC were compared to previously published data from these pairs in S1 (Bornschein et al., 2019b; Bornschein et al., 2019a). Electrophysiological data and Ca²⁺ imaging data were analyzed using custom written routines in Igor Pro (version 6.32 and 8.03, Wavemetrics). Data are shown as median and interquartile range (IQR) or as mean ± standard error (SEM).

Summarized data are either shown in boxplots as median ± IQR and mean values included as dashed lines in addition or in bar graphs as mean ± SEM. Normality was tested using the Shapiro-Wilk Test. Normally distributed data were compared with the t-test (two groups) or an ANOVA (more than two groups). Non-normally distributed or small samples of data were compared with the Mann-Whitney-U rank sum test (MWU; two groups) or an ANOVA on ranks (more than two groups). To compare pre- and post-treatment data the paired t-test or the Wilcoxon signed rank test (WSR) was used, depending on the distribution of the data. All statistical tests were two-tailed. P values are indicated as ^*P<0.05, ^**P< 0.01 and ^***P<0.001. The number of experiments n represents the number of cell pairs or cells and was chosen sufficiently high to permit reliable statistical analysis. All of the statistical details of experiments can be found in the figures and corresponding figure legends. Statistics were performed with Sigma Plot 11.0 (Dundas Software) and Mathematica 14.

Acknowledgements

We thank Gudrun Bethge for technical assistance.

This work was supported by a grant of the German research foundation (DFG SCHM1838/6-1 to HS).

Data availability

All data are available in the main text or in the supplemental information and will be shared by the corresponding author upon request. All original code has been deposited at Github and is publicly available as of the date of publication.

Additional information

Author contributions

Conceptualization: H.S., G.B.; Funding acquisition: H.S.; Methodology: all authors; Investigation: M.S., G.B., A.B., A.A., S.B., H.S.; Data Analysis: M.S., G.B., A.B., A.A., S.B., H.S.; Visualization: M.S., G.B., A.B.; Resources: H.S.; Modeling: H.S.; Supervision: H.S.; Writing – original draft: H.S., G.B.; Writing – review & editing: all authors.

Funding

Deutsche Forschungsgemeinschaft (DFG) (SCHM1838/6-1)

• Hartmut Schmidt

Additional files

Supplemental information. Figures S1–S4, Table S1 and Supplemental references.