Neuronal hyperactivity causes Na⁺/H⁺ exchanger-induced extracellular acidification at active synapses

In biological systems, the H⁺ concentration gradient between the intracellular and the extracellular environments plays a crucial role in several physiological processes (Chesler, 2003). In the central nervous system (CNS), the control of pH homeostasis is essential for the regulation of neurotransmission and neuronal excitability. While pH affects neural activity, neural activity causes pH alterations that vary in location and time (Chen and Chesler, 1992; Dulla et al., 2005; Makani and Chesler, 2007). Local pH changes occur in the brain during normal function (Chesler and Kaila, 1992; Magnotta et al., 2012), and neuronal firing causes both intracellular and extracellular pH shifts (Chesler, 2003; Chesler and Kaila, 1992; Kim and Trussell, 2009; Trapp et al., 1996). Rapid changes in intracellular pH have been observed at Drosophila nerve terminals during high-frequency in vivo activity (Caldwell et al., 2013; Rossano et al., 2013), while in the mammalian CNS, intense and synchronous activity of neural networks is accompanied by an initial alkalinization followed by a slower wave of acidification of the extracellular environment (DeVries, 2001; Du et al., 2014). A variety of ion channels, such as voltage-gated Ca²⁺ channels (Iijima et al., 1986), AMPA- and NMDA-type glutamate receptors (Lei et al., 2001; Tang et al., 1990; Traynelis and Cull-Candy, 1990) and GABA_A receptors (Dietrich and Morad, 2010; Kaila, 1994) display sharp pH-sensitive conductance, indicating that pH itself can strongly influence neuronal excitability, neurotransmitter release and postsynaptic responses (Chen and Chesler, 1992; Dietrich and Morad, 2010; Makani and Chesler, 2007).

pH shifts are also associated with a number of pathological processes. Interstitial acidosis is present in several neuropathologies such as epilepsy (Gottfried and Chesler, 1996; Inamura et al., 1989; Krishtal et al., 1987; Somjen, 1984) and brain ischemia (Makani and Chesler, 2010; McCormick and Contreras, 2001; Nedergaard et al., 1991; Siesjo et al., 1996; Xiong and Stringer, 2000). The correlation between pH shifts and neuronal hyperexcitability is well established. However, because of the lack of accurate systems to measure the extracellular H⁺ concentration, addressing the detailed events underlying pH-induced modulation of neural excitability is still a major challenge.

pH-sensitive microelectrodes generated some of the earliest measurements of both extracellular and intracellular pH, although their size and effects on tissue integrity strongly limit their widespread application. The use of fluorescent pH indicators opened new avenues to the study of pH fluctuations. The use of the extracellular ratiometric pH sensor SNARF allowed measurement of the local acidification in PC12 cells, which were stimulated by exposure to high K⁺ concentrations (Shuba et al., 2008) and the stimulus-evoked non-quantal release of H⁺ from vestibular hair cells at the cell–calyx afferent synapse (Highstein et al., 2014). More recently, genetically encoded fluorescent pH reporters that can be targeted to specific cellular compartments have been employed. One such indicator is a variant of the green fluorescent protein (GFP) obtained via specific mutagenesis, called ‘ratiometric GFP’, which is characterized by multiple excitation and emission maxima at distinct pH values (Bizzarri et al., 2006; Kuner and Augustine, 2000; Llopis et al., 1998). The absorption spectrum of GFP is widely used to measure the H⁺ concentration in living organisms, and optimized versions of GFP for pH measurements have been generated (Kneen et al., 1998). Among various forms of GFP, ecliptic and super ecliptic pHluorins display no fluorescence when protonated, allowing sensitive detection of pH increases that are associated with biological processes (De Angelis et al., 1998; Miesenböck et al., 1998). GFP-based pH indicators have been extensively employed to study alterations of intracellular pH that are related to network activity (Raimondo et al., 2012) and that can be used as markers of synaptic release (Miesenböck et al., 1998). However, only a few studies have employed them to monitor extracellular H⁺ fluxes associated with neural excitability or under pathological conditions. Notably, very recently the super-ecliptic pHluorin has been targeted to synaptic compartments to detect release of H⁺ that is associated with synaptic activity, suggesting a potential role of H⁺ as an extracellular messenger (Du et al., 2014; Wang et al., 2014). However, superecliptic pHluorin, whose fluorescence is rapidly quenched in the acidic environment, is more sensitive to increases, rather than decreases, of pH.

The aim of the present study is to describe the spatio-temporal dynamics of extracellular pH fluctuations at synaptic sites during neuronal hyperexcitability. We employed E²GFP to create a membrane-targeted extracellular ratiometric pH indicator. In contrast to other fluorescent pH sensors that are quenched at acidic pH, E²GFP represents a very effective and exquisitely sensitive ratiometric pH indicator for acidic shifts. By monitoring variations of E²GFP fluorescence in real time in primary hippocampal neurons, we were able to quantify pH fluctuations during network hyperactivity induced by convulsant drugs or high-frequency stimulation. Sustained neural hyperactivity caused a pH decrease that was specifically located at active synapses and was not attributable to the outflow of intravesicular H⁺ into the cleft during intense exocytosis, but rather to the activation of the Na⁺/H⁺ exchanger. Our data could contribute to the understanding of the physio-pathological mechanisms associated with hyperexcitability in the epileptic brain.

To engineer a genetically encoded indicator specifically sensitive to acidic pH shifts and targeted to the extracellular side of the plasma membrane, we inserted the E²GFP sequence (Bizzarri et al., 2006) into the mammalian expression vector pDisplay™, which directs the encoded proteins to the cell surface. The resulting construct was named ‘ex.E²GFP’ (Fig. 1A). The construct was subsequently cloned into a lentiviral vector under the control of the ubiquitous phosphoglycerate kinase (PGK) promoter to allow effective transduction of primary neurons. The subcellular localization of the probe was studied in the HEK293 cell line and primary cortical neurons. Membrane localization was observed in both cell types (Fig. 1B). To quantify the extent of membrane targeting of the probe in primary neurons, we labeled membranes with the lipophilic dye FM4-64fx and performed three-dimensional reconstruction using confocal microscopy. This analysis revealed that the vast majority of the probe was indeed localized to the plasma membrane, while the small pool of intracellular ex.E²GFP is likely to correspond to the fraction of the protein that was being synthesized and translocated to the cell surface (Fig. 1C,D). To further demonstrate the plasma-membrane location of the probe, we performed subcellular fractionation of 17 days in vitro (DIV) neurons that had been infected with ex.E²GFP, lysed and analyzed by western blotting using anti-GFP antibodies (Fig. 1E). ex.E²GFP immunoreactivity was predominantly associated with the plasma-membrane-containing fraction P2, compared to the cytosolic/microsomal fraction S2.

E²GFP is ideally suited as a ratiometric biosensor as its emission can increase, decrease or be independent of pH based on the excitation wavelength, and its pK_a in the cytosolic form is about 6.8 (Bizzarri et al., 2006), thus very effective to detect pH changes just below the physiological range. We asked whether the optical properties of E²GFP were affected by exposure to the extracellular environment. Primary cortical neurons were infected with ex.E²GFP and fluorescence emission was evaluated at two extracellular pH values (5.8 and 7.4) using 405 and 488 nm as excitation wavelengths (λ_ex405 and λ_ex488, respectively) (Bizzarri et al., 2006, 2009) (Fig. 2A). This analysis revealed a much higher fluorescence intensity at acidic pH compared to physiological pH when ex.E²GFP was excited at 405 nm, whereas an opposite behavior (albeit within a smaller dynamic range) was observed with 488 nm excitation (Fig. 2B). Thus, the extracellular targeting of E²GFP does not alter its optical properties or its pH sensitivity.

Next, a calibration curve was produced by analyzing ex.E²GFP fluorescence generated by switching the excitation wavelength between 405 and 488 nm and by systematically changing the extracellular pH from 5.7 to 7.8. To render the signal independent of the expression level of the protein, the ex.E²GFP fluorescence ratio (λ_ex405/λ_ex488) was calculated and found to be strongly dependent on extracellular pH (Fig. 2C). Data were fitted according to a logistic sigmoidal function, with a pK_a value of 6.9±0.1 (mean±s.e.m.) that was very similar to that for soluble E²GFP. This calibration curve was used to convert fluorescence ratios into absolute pH values throughout.

Several studies have addressed the question of how neuronal activity elicits changes in intracellular and extracellular H⁺ concentration, sometimes yielding contrasting results (Caldwell et al., 2013; Chen and Chesler, 2015; Cichy et al., 2015; de Curtis et al., 1998; DeVries, 2001; Lu et al., 2012; Palmer et al., 2003; Xiong et al., 2000; Xiong and Stringer, 2000; Zhang et al., 2010; Zhao et al., 2011). We studied the effects of sustained heightened neural activity on the extracellular H⁺ concentration by using in vitro models of epileptic-like network states characterized by synchronized neuronal discharges. Changes in fluorescence were monitored by live imaging (0.4 Hz frame rate) in primary cortical neurons that had been infected with lentiviruses encoding ex.E²GFP, and pH values were subsequently extrapolated from ratio values using the calibration curve shown in Fig. 2C. A scheme of the time course of the experiments is shown in Fig. 3A – pH was first measured under baseline conditions (2 min), then network hyperactivity was induced by blocking GABA_A receptors with bicuculline (30 µM), and cultures were imaged for 10 min. Thereafter, a selective blocker of voltage-gated Na⁺ channels, tetrodotoxin (TTX; 300 nM), was added to suppress firing, and live imaging was continued for a further 10 min. During the time frame of the experiments, ex.E²GFP fluorescence showed marked activity-dependent fluctuations, with no significant photobleaching (Fig. S1).

We found that in vitro seizure-like activity was associated with marked acidic pH transients in neurons. Interestingly, such fluorescence variations were observed both at the cell body level and along neurites of infected cells, where they mostly localized at synapse-like bouton structures (Fig. 3B; Movie 1). We thus monitored extracellular pH variations around these boutons and neuronal cell bodies. At both locations, acidic pH transients typically began at the onset of the hyperactivity triggered by the convulsant. After an initial delay, extracellular pH slowly but progressively decreased, reached a minimum acidic plateau 4-5 min after the beginning of the ‘seizure-like’ period and recovered towards baseline levels 4 min after termination of the seizure induced by TTX (Fig. 3C,D).

We subsequently asked how the time scale of increased neuronal firing compared to that of the acidic shift after addition of bicuculline. We cultured cortical neurons on 12-well multielectrode array (MEA) plates. Neurons were infected at DIV 10 and analyzed at DIV 15. The spiking rate was measured for 2 min under basal conditions, then bicuculline was added and the recording was continued for 10 min (Fig. 3E). In parallel, fluorescence was monitored in the same culture preparation. The bicuculline-induced increase in spiking rate was already observed 10 s after drug addition, while the pH decrease took about 4 min to reach a plateau. Thus, the increase in spiking activity happens first, followed by the pH decrease that appears to be a much slower process, which is likely to be due to the progressive extracellular accumulation of H⁺ in the extracellular/synaptic space.

To determine the variations of extracellular pH in response to a physiological electrical stimulation, cortical neurons expressing ex.E²GFP were challenged with a stimulation of 1200 action potentials (APs) at 10 Hz (Fig. 4A; Movie 2). Electrical stimulation also triggered a fluorescence increase that was localized to discrete synapse-like boutons along the neurites. The acidic pH transient developed slowly during the stimulation and continued to reach an acidic plateau of pH 6.9±0.1 (mean±s.e.m.) about 50 s after the end of the stimulation (Fig. 4B,C). pH levels remained acidic until the end of the recording, showing only a slow and modest recovery. At the cell body level, the effects of electrical stimulation were much smaller and non statistically significant (Fig. 4B,C).

To assess whether synaptic acidification is also present under basal conditions, we analyzed the live-imaging data corresponding to the ‘basal’ conditions of the various experiments. We identified single synapses and followed the behavior of ex.E²GFP fluorescence over time in the corresponding regions of interest (ROIs). Interestingly, we found that 27% of the analyzed synapses did not show appreciable pH shifts (>0.1 pH units from basal), while 73% of synapses displayed spontaneous pH oscillations that reached, in a small percentage of synapses, values between 6.6 and 6.9 (Fig. 4D-F). Thus, the majority of synapses under basal conditions display spontaneous activity levels of sufficient intensity to generate synaptic pH shifts.

The punctate pattern of fluorescence observed in the neuropil upon induction of network hyperactivity was reminiscent of synaptic contacts. To investigate the subcellular distribution of ex.E²GFP and unambiguously identify the sites of pH change, we quantified the relative abundance of the probe in cell bodies and in synaptic boutons labeled by the lipophilic dye FM4-64fx upon bicuculline stimulation. Interestingly, the vast majority of ex.E²GFP co-distributed with FM4-64fx-labeled puncta along neurites (Fig. 5A,B). However, it was unclear whether this was a specific activity-dependent synaptic acidification or the mere consequence of a non-specific accumulation of the sensor to synaptic boutons. To determine whether this was the case, primary cortical neurons were co-infected with lentiviruses expressing mCherry–synapsin-I (syn I), used as marker of presynaptic terminals (Cesca et al., 2010), and ex.E²GFP. Comparable ex.E²GFP intensity levels were observed under basal conditions at synaptic (i.e. synapsin positive) and extra-synaptic (i.e. synapsin negative) areas, thus confirming the ubiquitous targeting of the probe and the synaptic localization of the pH shifts (Fig. 5C-E). Because the expression of ex.E²GFP is driven by the constitutive PGK promoter, the probe was expressed in both excitatory and inhibitory neurons. In fact, when primary cultures at various stages of maturation (between DIV 10 and 18) were double stained for GFP and markers of excitatory neurons (VGLUT1; also known as SLC17A7), inhibitory neurons (VGAT; also known as SLC32A1) and astrocytes (GFAP), ex.E²GFP was found to be expressed in all three cell populations, although neuronal expression was predominant due to the paucity of glial cells under our culture conditions (Fig. S2).

Next, we investigated the specificity of the pH shifts for synaptic sites upon hyperactivity. A digital mask corresponding to presynaptic sites was built using the mCherry fluorescence, which was constant under resting and stimulated conditions (Fig. S3A), and the percent overlap of the ex.E²GFP fluorescence with the masked area was calculated under basal conditions and upon bicuculline-induced hyperactivity. Stimulation caused an increase in ex.E²GFP colocalization with mCherry–syn-I-positive puncta up to about 25-30% (Fig. 5F,G), demonstrating a massive recruitment of synaptic sites in the acidic response to the convulsant. By contrast, when the mask was built using the ex.E²GFP fluorescence, which substantially increases in intensity and area upon stimulation (Fig. S3A), the ex.E²GFP overlap on syn-I-labeled synaptic sites did not change (≈60%) (Fig. S3B), confirming the synaptic localization of ex.E²GFP before and after stimulation, and the synaptic localization of the pH shifts.

Not all synapses labeled by synaptic markers such as syn I are functionally active (silent synapses). To relate ex.E²GFP-positive puncta to active synapses, we co-infected hippocampal neurons with lentiviruses expressing ex.E²GFP and synaptobrevin–pHluorin-mOrange2 (Syb2O), a marker of synaptic vesicle (SV) exo–endocytosis at active synaptic sites that has an emission spectrum that does not overlap with that of ex.E²GFP (Ramirez et al., 2012). Upon electrical stimulation with 1200 APs at 10 Hz, ex.E²GFP fluorescence and Syb2O fluorescence overlapped by approximately 60%, confirming the localization of the probe to active synapses (Fig. S3C,D).

We then asked whether the observed activity-dependent synaptic pH shifts were present in the whole synapse population, or whether they were preferentially localized to excitatory or inhibitory synapses. Primary cortical neurons were infected with ex.E²GFP and live-labeled with VGAT–Oyster650, a specific marker of active inhibitory synapses (Fig. 5H). A digital mask corresponding to the active GABAergic boutons was built using the loaded VGAT–Oyster650 fluorescence, which remained constant under resting and stimulated conditions (Fig. S3E), and the percent overlap with the mask of the ex.E²GFP fluorescence was calculated under basal conditions and upon bicuculline-induced hyperactivity. Stimulation caused an increase in ex.E²GFP colocalization up to 35% with the active GABAergic synapses (Fig. 5I). By contrast, when the mask was built using the ex.E²GFP fluorescence, which increases in intensity and area upon stimulation, the ex.E²GFP overlap on VGAT-labeled synaptic sites was roughly constant (≈30%) (Fig. S3E-G).

Neuronal activity triggers the mobilization of SVs, which are characterized by an acidic luminal pH due to the activity of the active H⁺ transporter V-ATPase (Südhof, 2013). ex.E²GFP, initially exposed on the plasma membrane at synapses, could get internalized into the acidic environment of endocytosed SVs. If this were the case, then the increase in ex.E²GFP fluorescence would reflect the re-acidification of the endocytosed SVs, rather than a drop in the extracellular pH. To distinguish these two possibilities, we treated primary neurons with bafilomycin, a specific inhibitor of V-ATPase, which prevents re-acidification of SVs after exocytosis (Forgac, 1989). Pretreatment of neurons with bafilomycin before the addition of bicuculline did not affect either the bicuculline-induced acidification or the temporal profile of the pH decrease (Fig. 6A,B). These data demonstrate that acidification reflects a variation in the extracellular environment and that the release of H⁺ from exocytosed SVs and/or the incorporation of the V-ATPase into the plasma membrane during SV exocytosis are not primarily involved.

Since the pH shifts at synaptic sites are not attributable to release of H⁺ from SVs into the synaptic cleft, we addressed the possibility that the secretion of H⁺ during hyperactivity occurs via the activation of the Na⁺/H⁺ exchanger. We pretreated primary neurons with the Na⁺/H⁺ exchanger inhibitor amiloride before the addition of bicuculline (Fig. 7A,B). The treatment with amiloride alone induced a slight, but significant, extracellular alkalinization, confirming the constitutive activity of the Na⁺/H⁺ exchanger under basal conditions. Interestingly, when amiloride was combined with bicuculline stimulation, the extent of hyperactivity-induced extracellular acidification was significantly decreased, demonstrating the involvement of Na⁺/H⁺ exchangers in mediating the activity-induced synaptic pH shifts (Fig. 7C,D). As amiloride also inhibits Na⁺/Ca²⁺ exchangers, we performed bicuculline treatment in the presence of 1 µM cariporide, a specific inhibitor of Na⁺/H⁺ exchanger isoform 1 (also known as SLC9A1), one of the most expressed Na⁺/H⁺ exchanger neuronal isoforms (Harguindey et al., 2013; Luo et al., 2005). We found that synaptic acidification was strongly inhibited under this experimental condition, indicating a major role of Na⁺/H⁺ exchanger 1 in the observed hyperactivity-induced synaptic acidification (Fig. 7E-H).

We also tested whether the blockade of bicuculline-induced extracellular acidification by amiloride was the result of an inhibition of network-firing activity by the drug. To this end, primary neurons were cultured on multi-well microelectrode arrays, and the network activity was analyzed in terms of firing rate, burst frequency and burst duration in the presence of the various treatments by following the same experimental protocol as that described in Fig. 7A (Fig. 8A). Notably, amiloride alone, bicuculline alone and amiloride+bicuculline all increased both the firing and the bursting frequencies (Fig. 8B-E), although acidification occurred only with bicuculline alone. The excitatory effect of amiloride on network activity confirms previous observations of a reduction of the sensitivity of GABA_A receptors toward GABA by amiloride (Fisher, 2002). Altogether, our results demonstrate the key role of the Na⁺/H⁺ exchanger in coupling firing activity with a shift in extracellular pH.

Extracellular pH has important effects on neuronal activity, and in turn neuronal activity is an important determinant of the extracellular H⁺ concentration. H⁺ released by the SVs during sustained activity acts as a pseudo-transmitter, modifying several regulators of excitatory and inhibitory synaptic transmission (Cho and von Gersdorff, 2014; Du et al., 2014; Wang et al., 2014). Extracellular acidification occurs in the brain with elevated neural activity and neuronal injury (Wemmie et al., 2013). In turn, several studies have demonstrated the pH sensitivity of synaptic components like voltage-gated Ca²⁺ channels (VGCCs) (Iijima et al., 1986), ligand-gated AMPA and NMDA receptors (Traynelis and Cull-Candy, 1990), and GABA receptors (GABARs) (Kaila, 1994). In addition, acid-sensing ion channels (ASICs) are activated by extracellular low pH and are associated with neuronal dysfunction or brain injury (Friese et al., 2007; Yin et al., 2013). However, most of the studies have been performed on brain slices by applying high-speed pH microelectrodes (Fedirko et al., 2006) or dyes (Shuba et al., 2008), which lack accurate spatial resolution to localize pH changes to restricted neuronal domains.

In this paper, we analyzed the dynamics of extracellular pH changes accompanying epileptic-like activity in primary neuronal networks. We engineered a genetically encoded ratiometric GFP-based pH sensor that provides high spatial and temporal resolution to quantify variations in extracellular pH. Superecliptic GFP (pHluorin) has been employed previously as an extracellular pH sensor when targeted to the synaptic cleft (Du et al., 2014; Wang et al., 2014). However, pHluorin is an excellent reporter of alkalinization but quenches very rapidly upon acidic pH shifts. We therefore focused our attention on E²GFP (Bizzarri et al., 2006), a ratiometric GFP derivative that successfully allows the live monitoring of the changes of intracellular pH towards acidosis (Bizzarri et al., 2006; Raimondo et al., 2012). We fused E²GFP to the N-terminus of the PDGF receptor transmembrane domain that exposes it on the extracellular side of the plasma membrane (ex.E²GFP). We demonstrated that the exposure to the extracellular environment does not affect E²GFP optical properties and pH sensitivity, and that the targeted pH sensor shows high photostability with negligible photobleaching. Such a chimera has several advantages, namely: (i) it is ubiquitously targeted to the plasma membrane, thus being potentially able to map acidification in any region of the neuronal network; (ii) it is a ratiometric sensor with opposing pH sensitivities depending on the excitation wavelength, thus allowing reporting of pH changes independently of the expression level; and (iii) when excited at 405 nm, its fluorescence increases as the pH drops, making it more precise and sensitive in evaluating acidification than the previously used pHluorins.

We characterized hyperactivity-induced acidification by performing chemical stimulation with bicuculline and by applying high-frequency electrical stimulation. Using chemical stimulation, we observed a relatively slow build-up of extracellular acidification that occurred at both cell bodies and synaptic sites, and that slowly returned to neutral pH upon silencing of electrical activity with TTX, with no apparent pre- or post-alkalinization phases. By contrast, the more physiological electrical stimulation showed a slower and milder synaptic acidification specifically localized to synaptic boutons. Interestingly, our probe was sensitive enough to detect also smaller acidic shifts associated with spontaneous activity. While ex.E²GFP, which is exquisitely sensitive to the acidic shifts, may not be ideal to monitor alkaline shifts, the novelty of our results lays in the following points: (i) the increased network firing causes acidification at cell bodies and synapses, in both excitatory and inhibitory neurons; (ii) acidification does not result from the endocytotic capture of the membrane-exposed probe by SVs; (iii) acidification is mostly contributed by the Na⁺/H⁺ exchanger.

The acidic shifts in synaptic pH occurring during intense neuronal activity imply quantal or non-quantal release of H⁺ from neurons. In principle, this can occur through (i) exocytosis of SVs; (ii) exposure of V-ATPase on the plasma membrane; (iii) increased production of CO₂ and metabolic acids, such as lactate, that are externalized by monocarboxylate transporters; (iv) activation of Na⁺/H⁺ exchanger (mostly the Na⁺/H⁺ exchanger 1) that transports Na⁺ into the cells and extrudes H⁺; and (v) activation of Na⁺-dependent anion (Cl⁻/HCO₃⁻) exchangers (such as SLC4A8 and SLC4A10) that transport Na⁺ and HCO³⁻ into the cell and extrude H⁺ and Cl⁻ (Sinning and Hubner, 2013). Moreover, a glial source of H⁺ cannot be excluded (Grichtchenko and Chesler, 1994), although this is unlikely to be the case in our cultures, which are virtually devoid of glial cells.

The demonstration that reduction of extracellular pH occurs in synaptic regions is consistent with recent studies (Cho and von Gersdorff, 2014; Du et al., 2014; Highstein et al., 2014; Wang et al., 2014). The synaptic distribution of acidic shifts emphasizes that an alteration of synaptic transmission is the first step for the initiation and maintenance of ictal activity (Lazarevic et al., 2013). Release of SV protons into the synaptic cleft upon intense exocytosis has been recently proposed as a mechanism of protonergic transmission (Du et al., 2014; Wang et al., 2014). However, this was not the case under our experimental conditions as the long delay between the onset of the hyperactivity state and the acidic shift makes it unlikely that the un-buffering of SV-released H⁺ upon SV fusion (DeVries, 2001) significantly contributes to the acidic shift. Moreover, blockade of SV acidification does not affect the kinetics or the extent of acidification, ruling out H⁺ release by SVs or sustained exposure of SV V-ATPase on the plasma membrane. This discrepancy is likely to be due to the different experimental models used in the various studies – i.e. in vitro cultures of mouse cortical neurons (present study), zebrafish retinal neurons (Wang et al., 2014) and mouse amygdala neurons in acute brain slices (Du et al., 2014). In addition, we have specifically studied the pH shifts caused by epileptic-like activity, which are likely to rely on different mechanisms with respect to physiological network activity.

Amongst the other potential mechanisms of extracellular acidification, we focused on the Na⁺/H⁺ exchangers as several members of this family are known to contribute to the onset and maintenance of epilepsy and other developmental brain disorders (recently reviewed in Zhao et al., 2016). Our data indicate that acidification during ictal activity is mostly due to the Na⁺/H⁺ exchanger as the addition of the Na⁺/H⁺ exchanger blocker amiloride strongly reduced the bicuculline-induced synaptic acidification. Na⁺/H⁺ exchanger 1 is one of the main Na⁺/H⁺ exchanger isoforms expressed in the CNS, and mutations in it are known to cause epilepsy in mice (Bell et al., 1999; Cox et al., 1997). Na⁺/H⁺ exchanger 1 is activated in response to intracellular acidification, and in neurons, the consequent extrusion of H⁺ into the confined space of the synaptic cleft could generate the acidic pH shifts that we observed in our cultures. The other main neuronal Na⁺/H⁺ exchanger isoform is Na⁺/H⁺ exchanger 5 (Diering et al., 2011; Diering and Numata, 2014); however, treatment with the specific Na⁺/H⁺ exchanger 1 inhibitor cariporide (Harguindey et al., 2013; Luo et al., 2005) virtually abolished the bicuculline-induced pH shifts, thus suggesting that Na⁺/H⁺ exchanger 1 is the main isoform involved in the observed synaptic acidification.

In summary, we describe a novel sensor for extracellular pH changes and use it to study the wave of extracellular acidification during hyperactivity of the neuronal network, which is independent of H⁺ released by SVs and is mostly generated by the activity of Na⁺/H⁺ exchangers. These results underline the strict links between synaptic activity and synaptic pH. The extracellular sensor engineered in this study could be employed to better understanding of ionic dynamics under physiological and pathological conditions with high spatio-temporal precision.

All biochemical reagents and drugs were purchased from Sigma-Aldrich unless otherwise specified. Tissue culture reagents and media were from Gibco – Thermo Fisher Scientific or Sigma-Aldrich. The following primary antibodies were used: anti-GFP (Thermo Fisher Scientific, #A-11122, dilution 1:500; Synaptic Systems, #132004, dilution 1:500 for triple staining with anti-Homer and anti-Bassoon antibodies), anti-GFAP (Sigma-Aldrich, #G3893, dilution 1:500), anti-VGLUT1 (Synaptic Systems, Goettingen, Germany, #1350304, dilution 1:500), anti-mCherry (Clontech, #632543, dilution 1:500), anti-Homer (Synaptic Systems, #16011, dilution 1:500), anti-Bassoon (Synaptic Systems, #141002, dilution 1:500), anti-PSD95 (Synaptic Systems, #124011, dilution 1:1000) and anti-actin (Sigma-Aldrich, #A4700, dilution 1:1000) antibodies. Fluorophore-conjugated secondary antibodies were Alexa Fluor® 488-conjugated anti-rabbit (#A1108; dilution 1:1000), Alexa Fluor® 488-conjugated anti-guinea pig Alexa Fluor® 568-conjugated anti-mouse (#A-11004; dilution 1:1000), Alexa Fluor® 546-conjugated anti-rabbit (#A-11035; dilution 1:1000), Alexa Fluor® 647-conjugated anti-mouse (#A-21236; dilution 1:1000) and Alexa Fluor® 647-conjugated anti-guinea pig (#A-2145; dilution 1:1000) (all Thermo Fisher Scientific) antibodies.

The ex.E²GFP sequence was amplified from the pcDNA3.1+E²GFP plasmid (kindly provided by Dr Ranieri Bizzarri, Consiglio Nazionale delle Ricerche, Pisa, Italy) (Bizzarri et al., 2006) and inserted into the pDisplay™ vector (Thermo Fisher Scientific) between the BglII and PstI sites to obtain pDisplay-E²GFP (ex.E²GFP). The ex.E²GFP sequence was subsequently cloned into the pLenti-hPGK viral vector (kind gift of Drs Mario Amendola and Luigi Naldini, Tiget, Milan, Italy) between the AgeI and XhoI sites. The sequences of all constructs were verified by direct sequencing. Vectors were transiently transfected into cultured cells using Lipofectamine 2000 (Thermo Fisher Scientific).

All experiments were performed in accordance with the guidelines established by the European Community Council (Directive 2010/63/EU of 22 September 2010) and approved by the Italian Ministry of Health. Human embryonic kidney cells (HEK293, purchased from American Type Culture Collection) and primary cortical cultures from wild-type C57Bl/6 mice (Charles River, Calco, Italy) were prepared and maintained following standard procedures (Baldelli et al., 2007; Cesca et al., 2015).

The production of VSV-pseudotyped third-generation lentiviruses was performed as previously described (De Palma and Naldini, 2002). Primary neurons were infected with ex.E²GFP alone or co-infected with mCherry–syn-I (Verstegen et al., 2014) at 10 DIV at 10 multiplicity of infection (MOI), and the experiments were performed 7-8 days post-infection (17-18 DIV). For multi-electrode array experiments, neurons were infected at 10 DIV, and recordings performed at 17-18 DIV. For pHluorin experiments, neurons were co-infected with synaptobrevin–pHluorin-m-Orange2 (Syb2O) (Ramirez et al., 2012) and ex.E²GFP lentiviruses at 13 DIV, and optical recordings were performed at 17-18 DIV. Subcellular fractionation of transduced 17 DIV neurons was performed as previously described (Huttner et al., 1983).

Immunohistochemistry on fixed cells was performed following standard procedures (Cesca et al., 2015). Coverslips were imaged by confocal microscopy (SP8, Leica Microsystems GmbH, Wetzlar, Germany) using a 63× (1.4 NA) magnification lens. Confocal images were analyzed with the Leica LAS AF software (Leica Application Suite Advance Fluorescence, version 3.3, Leica Microsystems). z-stack images for three-dimensional reconstruction were acquired in 300-nm steps, and offline analysis for colocalization quantification and determination of intensity profiles was performed using the JACoP plugin of the ImageJ software (Bolte and Cordelières, 2006).

For live-cell imaging experiments, an SP8 confocal laser-scanning microscope was used. The sequential excitation of ex.E²GFP at 405 nm and 488 nm was achieved with a multiline argon laser. Emitted fluorescence was collected between 510 and 560 nm using a single photomultiplier tube (PMT) at a constant voltage. To verify fluctuations in laser intensity and bleaching of the probe, fluorescence was recorded in the absence of stimulation for the whole duration of a typical experiment. Neurons and cells were grown and maintained in 25-mm cover glasses and placed in a chamber with Tyrode solution (140 mM NaCl, 4 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 10 mM glucose and 10 mM HEPES at pH 7.4). To build the calibration curve, neurons were incubated in solutions at controlled pH. For the acidic buffer (pH 5.8-6.2), HEPES was replaced with MES. Solutions were applied with a valve control perfusion system (Warner Instruments, Hamden, CT) after 2 min of baseline acquisition in physiological solution; 30 µM of bicuculline was added, followed by 300 µM of TTX. For amiloride experiments, after 2 min of baseline acquisition, 20 µM amiloride was added, followed by 30 µM bicuculline. For cariporide (Tocris, Avonmouth, Bristol, UK) experiments, after 2 min of baseline acquisition, 1µM cariporide was added, followed by 30 µM bicuculline. The acquisition rate was 0.4 Hz for each excitation wavelength. For field stimulation, time-lapse images were acquired every 2.58 s for 360 s. APs were evoked by passing 1-ms current pulses, yielding fields of 10 V/cm, through platinum-iridium electrodes using an AM 2100 stimulator (AM Systems, Carlsborg, WA). After 30 s of baseline acquisition, neurons were stimulated with a train of 1200 APs at 10 Hz.

Images were acquired with constant gain and exposure times across all experiments.

Dissociated cortical neurons were plated onto 12-well planar MEAs (768-GL1-30Au200 from Axion BioSystems, Atlanta, GA), comprising 768 electrodes in 12 wells (64 electrodes/well). Extracellular spontaneous activity was recorded throughout the experiment and analyzed as previously described (Cesca et al., 2015; Chiappalone et al., 2009). Data were analyzed using the Neural metric tool (Axion Biosystems, Atlanta, GA).

Neurons co-expressing mCherry–syn-I and ex.E²GFP were kept in a stimulation chamber (Warner Instruments, Hamden, CT) and imaged in Tyrode's solution. Bicuculline (30 µM) was added to the medium, and images were acquired after 10 min incubation using an SP8 fluorescence confocal laser-scanning microscope as described above. Five fields were analyzed using the JaCoP plug-in of ImageJ (Bolte and Cordelières, 2006).

For pHluorin experiments, neurons co-expressing Syb2O and ex.E²GFP were kept in a stimulation chamber and imaged in physiological solution. The chamber was positioned on the stage of an IX-81 motorized inverted epifluorescence microscope (Olympus, Tokyo, Japan). An MT20 Hg-Xe lamp (Olympus, Tokyo, Japan) was used as a light source with 480±20 nm excitation, 495 nm dichroic and 525±50 nm emission filters to detect ex.E²GFP signal. 560±40 nm excitation, 585 nm dichroic and 630±75 nm emission filters were instead used to detect the Syb2O signal. An electrical stimulation of 400 APs at 20 Hz was delivered to visualize active synapses. Images were acquired with an Hamamatsu Orca-ER CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan) using an UplanSapo 60×1.35 NA oil immersion objective (Olympus, Tokio, Japan) and the software Excellence RT (Olympus, Tokyo, Japan). Fields expressing both signals were analyzed to evaluate the colocalization under basal and stimulated conditions; 11 fields were analyzed using the JaCoP plugin of ImageJ (Bolte and Cordelières, 2006).

Labeling with VGAT–Oyster650 (Synaptic Systems, Goettingen, Germany; #131103; dilution 1:200) was performed by incubating ex.E²GFP-infected neurons for 20 min at 37°C under 5% CO₂ to load the dye into inhibitory presynaptic terminals through spontaneous activity. For basal conditions, images were acquired in Tyrode’s solution. For the hyperexcited state, 30 µM bicuculline was added to the medium, and images were acquired after 10 min of incubation. Experiments were performed on an SP8 fluorescence confocal laser-scanning microscope, as described above. For all experiments, colocalization was analyzed by generating Manders' colocalization coefficients (Manders et al., 1992).

Data are expressed as means±s.e.m. for the number of cells (n) and mouse preparations detailed in the figure legends. Normal distribution of data was assessed using D'Agostino-Pearson's normality test (for n>8 values/experimental group) or the Kolmogorov–Smirnov test (n<8 values/experimental group). To compare two normally distributed sample groups, a paired or unpaired Student's t-test was used. To compare more than two normally distributed sample groups, one- or two-way ANOVA, followed by the Bonferroni's multiple comparison test, was used. α levels for all tests were 0.05% (95% confidence intervals). Statistical analysis was performed using Origin Pro 9.1 (OriginLab Corp., Northampton, MA) and GraphPad Prism 5 (GraphPad Software) software.

We thank Dr Kavalali (Department of Neuroscience, The University of Texas Southwestern Medical Center, Dallas, TX) for the synaptobrevin–Orange2 constructs, Dr Naldini (Tiget, Milan, Italy) for lentiviral constructs and protocols, and Drs Nanni and Mehili (Istituto Italiano di Tecnologia, Genova, Italy) for assistance in the preparation of primary cultures. We are also grateful to Ms Donato for helping in the initial phases of the project and to Drs Moretti and Pecoraro for helping with the MEA experiments.