Computational modeling predicts ephemeral acidic microdomains in the glutamatergic synaptic cleft

Abstract

At chemical synapses, synaptic vesicles release their acidic contents into the cleft, leading to the expectation that the cleft should acidify. However, fluorescent pH probes targeted to the cleft of conventional glutamatergic synapses in both fruit flies and mice reveal cleft alkalinization rather than acidification. Here, using a reaction-diffusion scheme, we modeled pH dynamics at the Drosophila neuromuscular junction as glutamate, ATP, and protons (H⁺) were released into the cleft. The model incorporates bicarbonate and phosphate buffering systems as well as plasma membrane calcium-ATPase activity and predicts substantial cleft acidification but only for fractions of a millisecond after neurotransmitter release. Thereafter, the cleft rapidly alkalinizes and remains alkaline for over 100 ms because the plasma membrane calcium-ATPase removes H⁺ from the cleft in exchange for calcium ions from adjacent pre- and postsynaptic compartments, thus recapitulating the empirical data. The extent of synaptic vesicle loading and time course of exocytosis have little influence on the magnitude of acidification. Phosphate but not bicarbonate buffering is effective at suppressing the magnitude and time course of the acid spike, whereas both buffering systems are effective at suppressing cleft alkalinization. The small volume of the cleft levies a powerful influence on the magnitude of alkalinization and its time course. Structural features that open the cleft to adjacent spaces appear to be essential for alleviating the extent of pH transients accompanying neurotransmission.

Figure 1

The cleft of the Drosophila NMJ alkalinizes during activity. (A) The pH-sensitive pHusion-Ex probe can be targeted to the extracellular spaces of the NMJ with a transmembrane domain as an anchor. (B) When expressed in the muscle the probe is seen in the SSR that surrounds motor nerve terminal boutons (revealed here through forward-filling the live axon with Alexa Fluor 647 dextran 10,000 MW). (C) Cleft alkalinization can be seen in response to a single AP, providing signal averaging is used. Both traces represent the average of 100 APs delivered at 1 Hz. Orange trace, glutamate present at 7 mM. (D) Cleft alkalinization can be seen in response to a train of APs (without signal averaging) delivered at the neuron’s native firing frequency (21 Hz). In (C) and (D), NaHCO₃ was added to 15 mM and continuously bubbled with carbogen; stabilizing at pH 7.2. No phosphate or zwitterionic buffers are present. (E) Our interpretation of this phenomenon is summarized in a temporal series: in response to an AP, Ca²⁺ entry to the presynaptic terminal triggers the exocytosis of glutamate, co-loaded ATP, and H⁺ titrated by both. Activation of postsynaptic GluRs allows Ca²⁺ entry to the presynaptic muscle. The PMCA extrudes Ca²⁺ from both pre- and postsynaptic compartments using H⁺ as a counter-ion.

Figure 2

The Drosophila NMJ synaptic cleft is very narrow but continuous with an adjacent void. (A) Fluorescent images of two proteins at the NMJ revealed by immunohistochemistry: presynaptic Bruchpilot, defining the active zones (AZs), and PMCA found in both pre- and postsynaptic compartments. (B) An electron micrograph of a 50-nm section through the center of a presynaptic bouton and the surrounding muscle SSR. AZs, which are often marked by a “T-bar” extending into the cytosol from the presynaptic membrane (green arrow), can be found at a number of sites around the bouton’s periphery. (C) An enlarged view of the AZ indicated by the green arrow in (B). (D) Interpretation of the image in (C), emphasizing the location of the T-bar, the cleft, and the common observation that the cleft usually empties out into voids of the SSR ∼400 nm distant from the center of each AZ (see asterisk at far left). (E) Plot of the width of the cleft (between outer leaflets of opposing plasma membranes) at NMJs on muscle fibers #7/6, 13, and 12. Each point represents measurements at 12 AZs at each of five boutons on each of the muscle fibers, mean ± standard error. (F) Schematic of a cross section of a bouton embedded within the muscle SSR, showing release into the cleft from two out of five AZs. (G) For computational purposes, the cleft can be treated as a continuous layer, sandwiched between the planes of two opposing membranes; represented here in a cross section annotated with critical distances. (H) An expanded view of release and subsequent PMCA activity, emphasizing the location of PMCA activity and the location at which the cleft becomes continuous with the voids within the SSR.

Figure 3

The synaptic cleft shows substantial but brief acidification. (A) Output plots of the computational model. When 8000 glutamate molecules are released into the cleft containing 19 mM bicarbonate and a medium level of CA activity (enzymatic acceleration (En_acc of 600 (nominal units)), the pH at the mouth of fusion pore (asterisk) drops to 5.5 (i.e., equilibrates with SV contents) within microseconds. The cleft then alkalinizes, reaching a peak of ∼7.45 within ∼50 ms, before decaying to the prestimulus baseline at pH 7.2. (B and C) Bicarbonate buffering has little impact on cleft acidification at either the site of release or midway between the fusion pore and access to the SSR; phosphate buffering, however, is very effective (1 mM total). (D) If SV loading is reduced to a load of 1500 mM glutamate molecules (250 mM) and 610 ATP molecules (100 mM) and isosmotic with the cytosol (350 mM), the acid transient is as deep as when 8000 glutamate is released. (E and F) The time course of fusion pore opening also has little impact on the depth of acidification over a range of four orders of magnitude (τ = 0.1–100 μs; 8000 glutamate). Bicarbonate buffering is 19 mM in all scenarios except diffusion only. Fusion pore opening (τ) is 1 μs, except when indicated otherwise (E and F).

Figure 4

The synaptic cleft shows moderate but long lived alkalinization. (A) Identical to Fig. 3A, the model output plots in (A) provide context for the detail of (B and F). (B) and (C) show the minimal impact of bicarbonate buffering on cleft acidification caused by neurotransmitter release. The extended timescale encompasses the transition point from acidification to alkalinization (arrowhead) at ∼1 ms (when the impact of postsynaptic PMCA activity becomes apparent). (D) Left panel shows the first 80 ms after release, illustrating the shape of alkaline transient which does not decay until after postsynaptic Ca²⁺ entry ceases at ∼40 ms. Right panel shows detail of the boxed section on the left. #, cleft alkalinization as a result of presynaptic PMCA activity; †, cleft acidification as a result of exocytosis; @, realkalinization as a result of diffusion, buffering, and presynaptic PMCA activity; and ψ, alkalinization as a result of pre- and postsynaptic PMCA activity. (E and F) Bicarbonate, and especially phosphate buffering (1 mM), is effective at suppressing cleft alkalinization. Note, the 400-fold change in time base between (B and C) and (E and F). Shown are 8000 glutamate molecules released and 19 mM bicarbonate buffering in all scenarios. Fusion pore opening (τ) is 1 μs. Medium and high CA activity are 600 and 2000 En_acc nominal units, respectively. κ indicates the acidic transients have been truncated for presentation purposes.

Figure 5

Synapse morphology has a powerful influence on cleft pH. (A and B) Output plots of the computational model showing the impact of preventing access from the synaptic cleft to the void of the SSR (note the location of asterisks in A vs. B). The dashed line represents the control conditions, i.e., bicarbonate buffering in the presence of medium CA activity, full access to the SSR and membrane permeability to CO₂. Low, medium, and high CA activity are 100, 600, and 2000 En_acc nominal units, respectively. Note, because of the ability of CO₂ to cross membranes, high CA activity is better able to buffer pH changes than medium CA activity in combination with 1 mM total phosphate. When CA activity (En_acc) falls below ∼500 nominal units, cleft buffering fails. pH 8.8 represents the pH at which the PMCA is expected to stall (32). (C and D) Plots showing the impact of both preventing cleft access to the SSR and preventing CO₂ movement across membranes. Cleft pH buffering failed in all cases (note the location of asterisks in C vs. D). Shown are 8000 glutamate molecules released and 19 mM bicarbonate buffering in all scenarios. Fusion pore opening (τ) is 1 μs. κ indicates truncated acidic transients. Detail of acidic transients and first 80 ms of the alkaline transients are given in Fig. S4.

Figure 6

MNs must fire beyond usual rates before alkaline transients will summate at individual AZs. (A–C) A scheme that illustrates why pH reporter fluorescence summation from a bouton over multiple nerve impulses may not indicate pH transient summation at individual AZs. (A) Stylized representation of resting fluorescence of a cytosolic Ca²⁺ reporter (both sides of the synapse) and an extracellular pH reporter (muscle PM) at a single bouton of the NMJ. (B) Representation of the Ca²⁺ and pH responses to release from individual AZs over 10 APs; Ca²⁺ response are postsynaptic and pH responses are in the cleft. The Ca²⁺ reporter responses are shown to delineate the location of each releasing AZ. The average probability of release is 1/9, and nine AZs have been illustrated responding sequentially around the periphery of a bouton. The images capture the maximal response; the time of image capture is represented with an arrow on the pH plots for each AZ. The plots of pH change at each of the AZs are generated by the model (LHS ordinate). The images show the location of an AZ-delimited region of interest that would be expected to yield pH sensor transients (ΔF/F_rest) of the magnitude shown (right hand side ordinate). (C) Numerical summation of all AZ pH plots in (B) yields a pH change for the entire bouton that is well below that of an AZ, but unlike pH at an individual AZ, it summates over multiple APs. The right hand side ordinate shows the expected fluorophore transients (ΔF/F_rest) if the region of interest encompassed the entire bouton containing the nine AZs, and this is compatible with what is observed empirically (Fig. 1D). (D) Model generated plots of pH an individual AZ in which one release event follows another after a defined interval. (E) A plot of the potentiation of the second alkaline transient relative to the first, across different axon firing frequencies and showing the AZ release frequencies in register. The vertical dotted line represents the average firing frequency for this motor neuron.