Mitochondria and plasma membrane Ca²⁺-ATPase control presynaptic Ca²⁺ clearance in capsaicin-sensitive rat sensory neurons

Abstract

The central processes of primary nociceptors form synaptic connections with the second-order nociceptive neurons located in the dorsal horn of the spinal cord. These synapses gate the flow of nociceptive information from the periphery to the CNS, and plasticity at these synapses contributes to centrally mediated hyperalgesia and allodynia. Although exocytosis and synaptic plasticity are controlled by Ca²⁺ at the release sites, the mechanisms underlying presynaptic Ca²⁺ signalling at the nociceptive synapses are not well characterized. We examined the presynaptic mechanisms regulating Ca²⁺ clearance following electrical stimulation in capsaicin-sensitive nociceptors using a dorsal root ganglion (DRG)/spinal cord neuron co-culture system. Cytosolic Ca²⁺ concentration ([Ca²⁺]_i) recovery following electrical stimulation was well approximated by a monoexponential function with a τ∼2 s. Inhibition of sarco-endoplasmic reticulum Ca²⁺-ATPase did not affect presynaptic [Ca²⁺]_i recovery, and blocking plasmalemmal Na⁺/Ca²⁺ exchange produced only a small reduction in the rate of [Ca²⁺]_i recovery (∼12%) that was independent of intracellular K⁺. However, [Ca²⁺]_i recovery in presynaptic boutons strongly depended on the plasma membrane Ca²⁺-ATPase (PMCA) and mitochondria that accounted for ∼47 and 40%, respectively, of presynaptic Ca²⁺ clearance. Measurements using a mitochondria-targeted Ca²⁺ indicator, mtPericam, demonstrated that presynaptic mitochondria accumulated Ca²⁺ in response to electrical stimulation. Quantitative analysis revealed that the mitochondrial Ca²⁺ uptake is highly sensitive to presynaptic [Ca²⁺]_i elevations, and occurs at [Ca²⁺]_i levels as low as ∼200–300 nm. Using RT-PCR, we detected expression of several putative mitochondrial Ca²⁺ transporters in DRG, such as MCU, Letm1 and NCLX. Collectively, this work identifies PMCA and mitochondria as the major regulators of presynaptic Ca²⁺ signalling at the first sensory synapse, and underlines the high sensitivity of the mitochondrial Ca²⁺ uniporter in neurons to cytosolic Ca²⁺.

Key points

• The first sensory synapse formed between the central processes of primary afferents and dorsal horn neurons plays an important role in controlling the flow of nociceptive information from the periphery to the CNS, and plasticity at this synapse contributes to centrally mediated pain hypersensitivity.

• Although exocytosis and synaptic plasticity are regulated by presynaptic Ca²⁺, the mechanisms underlying presynaptic Ca²⁺ signalling at the first sensory synapse are not well understood.

• In this study we show that the plasma membrane Ca²⁺-ATPase and mitochondria are the major regulators of presynaptic Ca²⁺ signalling in capsaicin-sensitive dorsal root ganglion neurons accounting for ∼47 and ∼40% of presynaptic Ca²⁺ clearance, respectively.

• Quantitative analysis of changes in cytosolic and mitochondrial Ca²⁺ concentrations demonstrates that the mitochondrial Ca²⁺ uniporter is highly sensitive to cytosolic Ca²⁺ at this synapse.

• These results help us understand presynaptic mechanisms at the first sensory synapse.

Introduction

Primary nociceptive neurons send their central processes to the dorsal horn of the spinal cord where they form glutamatergic synaptic connections with the second-order nociceptive neurons (Woolf & Salter, 2006; Kuner, 2010). This synaptic connection, often referred to as the first sensory synapse, plays a critical role in controlling the flow of nociceptive information from the periphery into the CNS. Inhibiting transmission at this synapse with blockers of voltage-gated Ca²⁺ channels or opioids produces analgesic effects (Cao, 2006; McGivern, 2006; Heinke et al. 2011), whereas facilitation of sensory transmission through various forms of synaptic plasticity (including wind-up, central sensitization and long-term potentiation) contributes to the pain hypersensitivity associated with inflammation or nerve injury (Ji et al. 2003; Woolf & Salter, 2006; Kuner, 2010; Ruscheweyh et al. 2011). The mechanisms responsible for various forms of synaptic plasticity at the first sensory synapse are complex and not fully understood.

Presynaptic Ca²⁺ is the principal regulator of neurotransmitter release and synaptic plasticity that acts via multiple Ca²⁺-sensing proteins to control almost every aspect of the synaptic vesicle life cycle. For example, Ca²⁺ triggers synchronous transmitter release via the low-affinity Ca²⁺ sensors synaptotagmins 1, 2 and 9 (Sugita et al. 2002; Xu et al. 2007), whereas asynchronous and spontaneous transmitter release are mediated by the high-affinity Ca²⁺ sensor Doc2 (Groffen et al. 2010; Yao et al. 2011). Endocytotic retrieval of synaptic vesicles is stimulated by Ca²⁺/calcineurin-dependent dephosphorylation of dynamin-1 and other endocytotic proteins (Cousin & Robinson, 2001; Clayton & Cousin, 2009). Residual presynaptic [Ca²⁺]_i accumulated during repeated electrical stimulation contributes to short-term synaptic plasticity by controlling the size of the release-ready pool of synaptic vesicles (Zucker & Regehr, 2002; Neher & Sakaba, 2008) through Ca²⁺-dependent activation of Munc13 and protein kinases A and C (PKA and PKC; Turner et al. 1999; Junge et al. 2004; Sørensen, 2004). To carry out these versatile synaptic functions, presynaptic Ca²⁺ signals must be precisely organized in time and space, which is achieved through the coordinated actions of Ca²⁺ channels, buffers and transporters. Therefore, identifying the specific components of presynaptic Ca²⁺ machinery is essential for understanding mechanisms of synaptic transmission and plasticity at any given synapse.

In spite of the key role of the first sensory synapse in transmitting sensory information and processing pain, the mechanisms that control presynaptic cytosolic Ca²⁺ concentration ([Ca²⁺]_i) at this synapse are not known. Here we used cytosolic and mitochondrial Ca²⁺ imaging to examine mechanisms of presynaptic [Ca²⁺]_i clearance at the synapses formed between dorsal root ganglion (DRG) and spinal cord (SC) neurons in culture. We focused our analysis on a subset of nociceptive DRG neurons that express TRPV1 (transient receptor potential vanilloid 1) receptors because of their well-established role in acute and chronic pain (Caterina & Julius, 2001; Szallasi et al. 2007). Our data demonstrate that plasma membrane Ca²⁺-ATPase (PMCA) and mitochondria are the major regulators of presynaptic [Ca²⁺]_i at this synapse.

Methods

Cell culture

DRG/SC co-cultures were prepared as previously described (Medvedeva et al. 2008, 2009). In brief, newborn (postnatal days 0–2) Sprague–Dawley rats were killed by decapitation with sharp scissors, according to a protocol approved by the University of Iowa Institutional Animal Care and Use Committee and in accordance with the guidelines of the National Institutes of Health. Every effort was made to minimize the number of animals used. DRG were dissected from the thoracic and lumbar segments and incubated in pronase E dissolved in DMEM (1 mg ml⁻¹) containing 20 mm Hepes (pH 7.4) and penicillin–streptomycin (100 U ml⁻¹ and 100 μg ml⁻¹, respectively) for 7 min at 37°C in a 10% CO₂ incubator. Ganglia were washed twice in cold DMEM containing Hepes (20 mm; pH 7.4) and then mechanically dissociated by trituration with flame-constricted Pasteur pipettes of decreasing diameter. The spinal cord was dissected into small segments (∼0.5 mm) and then digested in DMEM containing 0.025% trypsin, Hepes (20 mm, pH 7.4) and penicillin–streptomycin (100 U ml⁻¹ and 100 μg ml⁻¹, respectively) for 8 min at 37°C in a 10% CO₂ incubator. SC segments were then washed with cold complete DMEM medium supplemented with 5% heat-inactivated horse serum, 5% fetal bovine serum and penicillin–streptomycin (100 U ml⁻¹ and 100 μg ml⁻¹, respectively) and dissociated using the procedure described above for DRG dissociation. Suspensions of SC and DRG cells were plated onto 25 mm glass coverslips coated with poly-l-ornithine and laminin. Cells were grown in DMEM supplemented with 5% heat-inactivated horse serum, 5% fetal bovine serum, insulin (6 μg ml⁻¹) and penicillin–streptomycin (100 U ml⁻¹ and 100 μg ml⁻¹, respectively) in a 10% CO₂ incubator at 37°C. After 48 h, 5 μm cytosine β-d-arabinofuranoside (AraC) was added to cultures for 24 h. Two days later, cells were again treated with 5 μm AraC for 24 h. Cultures were grown for 10–16 days before use; 50% of the culture medium was replaced every 5–6 days.

Electrophysiological recordings

The methods for monitoring whole-cell Ca²⁺ currents and excitatory postsynaptic currents (EPSCs) were similar to those previously described by our group (Medvedeva et al. 2008, 2009). Whole-cell patch-clamp recordings were obtained using a Axopatch 200B patch-clamp amplifier and a Digidata 1322A analog-to-digital converter (Molecular Devices, Union City, CA, USA). Data were collected (filtered at 2 kHz and sampled at 5 kHz) and analysed using the pClamp 9 software (Molecular Devices). Patch pipettes were pulled from borosilicate glass (Narishige, New York, USA; 3–5 mΩ) on a Sutter Instruments (Novato, CA, USA) P-87 micropipette puller. For EPSC recordings, spinal cord neurons were voltage-clamped at –60 mV using patch pipettes that were filled with the following solution (mm): 125 potassium gluconate, 10 KCl, 3 Mg-ATP, 1 MgCl₂, 5 EGTA and 10 Hepes (pH 7.25 adjusted with KOH; 290 mosmol kg⁻¹ with sucrose). The standard extracellular recording solution contained (mm): 140 NaCl, 5 KCl, 1.3 CaCl₂, 0.4 MgSO₄, 0.5 MgCl₂, 0.4 KH₂PO₄, 0.6 Na₂HPO₄, 3 NaHCO₃, 10 glucose and 10 Hepes (pH 7.4 with NaOH; 310 mosmol kg⁻¹ with sucrose). The extracellular solution also contained 10 μm bicuculline and 2 μm strychnine to block inhibitory postsynaptic currents mediated by GABA_A and glycine receptors, respectively. Evoked EPSCs in SC neurons were elicited every 4 s using a glass extracellular stimulating electrode (0.2–0.4 ms pulse) positioned near the cell body of a nearby DRG neuron. For Ca²⁺ current recordings, extracellular solution contained (mm): 115 choline chloride, 30 TEACl, 1.3 CaCl₂, 1 MgCl₂, 10 glucose, 10 Hepes and 1 μm tetrodotoxin (pH 7.4 with TEAOH; 310 mosmol kg⁻¹ with sucrose); and the patch-pipette solution had the following composition (mm): 135 caesium gluconate, 3 Mg-ATP, 1 MgCl₂, 10 EGTA and 10 Hepes (pH 7.25 adjusted with CsOH; 290 mosmol kg⁻¹ with sucrose). Voltage-gated Ca²⁺ currents were evoked by step depolarizations from –60 to +10 mV (100 ms, every 20 s).

Measurements of [Ca²⁺]_i in axonal boutons

The protocols for monitoring presynaptic [Ca²⁺]_i were similar to those we have previously described (Medvedeva et al. 2008, 2009). Cells were placed in a flow-through chamber mounted on the stage of an inverted IX-71 microscope (Olympus, Tokyo, Japan). For [Ca²⁺]_i measurements, DRG neurons were loaded with the low affinity (K_d∼ 5.5 μm) Ca²⁺ indicator Fura-FF (200 μm) or, in some experiments involving small-amplitude [Ca²⁺]_i measurements (see Figs 4D, E and 8C, D) the high-affinity (K_d = 225 nm) Ca²⁺ indicator Fura-2 (200 μm), via patch pipettes by holding neurons in the whole-cell configuration for 3–5 min; recordings began within 15–20 min of the pipette withdrawing. Only neurons that had membrane potential more negative than –50 mV immediately after patch membrane rupture were used for further experimentation. Fluorescence of Fura-FF or Fura-2 was alternately excited at 340 nm (12 nm bandpass) and 380 nm (12 nm) using a Polychrome IV monochromator (TILL Photonics, Gräfelfing, Germany) via a 40× oil-immersion objective (NA = 1.35, Olympus). Emitted fluorescence was collected at 510 (80) nm using an IMAGO CCD camera (TILL Photonics). Pairs of 340/380 nm images were sampled at 10 Hz. The fluorescence ratio (R = F₃₄₀/F₃₈₀) was converted to [Ca²⁺]_i according to the formula: [Ca²⁺]_i = K_dβ(R–R_min)/(R_max–R) (Grynkiewicz et al. 1985). The dissociation constants (K_d) used for Fura-2 and Fura-FF were 225 and 5500 nm, respectively (Molecular Probes Handbook). R_min, R_max and β were determined by applying 10 μm ionomycin in Ca²⁺-free buffer (1 mm EGTA) and saturating Ca²⁺ (1.3 mm Ca²⁺). For Fura-2, calibration constants were: R_min = 0.23, R_max = 2.70 and β = 4.75. For Fura-FF, calibration constants were: R_min = 0.21, R_max = 2.25 and β = 5.30. Fluorescence was corrected for background, as determined prior to loading of cells with the indicators.

Figure 4. The rate of presynaptic Ca²⁺ clearance is not affected by inhibition of SERCA.

A, [Ca²⁺]_i transients in axonal bouton (white box) were evoked by trains of action potentials (10 Hz for 2 s) in the absence or presence (5 μm) of the SERCA inhibitor cyclopiazonic acid (CPA). The CPA treatment started 3 min prior to the third stimulation. B, [Ca²⁺]_i transients generated in the absence (grey dots) or presence of 5 μm CPA (red dots) are superimposed. The [Ca²⁺]_i traces were offset along the y-axis to facilitate comparison. Traces obtained by monoexponential fitting of the corresponding [Ca²⁺]_i data points are shown by continuous lines. C, plot summarizing time constants for [Ca²⁺]_i recovery in controls (contr) and in the presence of 5 μm CPA (paired Student's t test, n = 18 boutons/4 cells). D, [Ca²⁺]_i changes recorded from two axonal boutons (b1 and b2, squares) and the cell body of a Fura-2-loaded DRG neuron in response to electrical stimulation (6 Hz for 5 s; arrow) or the ryanodine receptor activator caffeine (10 mm). E, comparison of the amplitudes of [Ca²⁺]_i responses to 10 mm caffeine in axonal boutons (n = 27 boutons/5 cells) and somata (n = 5) of DRG neurons. **P < 0.01, unpaired Student's t test.

Figure 8. Quantification of mitochondrial Ca²⁺ uptake as a function of [Ca²⁺]_i elevation.

A, DRG neurons were transfected with mtPericam (green) and subsequently loaded with Rhod10K to identify presynaptic boutons. B, [Ca²⁺]_mt changes in axonal boutons (white box) were evoked by depolarization of incremental magnitude using solutions containing 15 mm KCl, 20 mm KCl or 30 mm KCl. C, axonal distribution of Fura-2 in a resting DRG neuron. D, [Ca²⁺]_i changes in axonal boutons (white box) recorded in Fura-2-loaded DRG neurons using the same stimulation protocol as in B. E, the initial rate of the [Ca²⁺]_mt increase is plotted as a function of peak [Ca²⁺]_i elevation in response to K⁺-induced depolarization or TRPV1 stimulation. [Ca²⁺]_i and [Ca²⁺]_mt changes were evoked by applying 10, 15, 20, 30, 50 or 90 mm KCl (30 s, grey; n = 8–24 boutons from 4–8 cells) or 1 μm capsaicin (30 s; black; n = 16 boutons/5 cells for [Ca²⁺]_i and n = 15 boutons/4 cells for [Ca²⁺]_mt), as described in B and D. [Ca²⁺]_i was measured using Fura-2 (10, 15, 20, 30 and 50 mm KCl) or Fura-FF (90 mm KCl and 1 μm capsaicin); [Ca²⁺]_mt was measured using mtPericam. The initial rate of rise in [Ca²⁺]_mt was calculated by dividing half of the [Ca²⁺]_mt amplitude by the time required to reach half of the [Ca²⁺]_mt peak. The data were fitted using the Hill equation (Origin 7 software), which produced a K_0.5 = 606 nm and a Hill coefficient = 2.3. F, RT-PCR analysis of neuron-enriched DRG culture shows that several putative molecular components of mitochondrial Ca²⁺ transport, Letm1, MCU, MICU1 and NCLX, are expressed in DRG neurons.

After dye loading and patch pipette withdrawal, intact DRG neurons were stimulated with trains of action potentials using extracellular field stimulation as previously described (Usachev & Thayer, 1999). In brief, field potentials were generated by passing current between two platinum electrodes via a Grass SS stimulator and a stimulus isolation unit (Quincy, MA, USA). Trains of 1 ms pulses were delivered at 10 Hz. The stimulus voltage threshold sufficient to elicit a detectable increase in [Ca²⁺]_i from neuronal processes and axonal boutons was determined before beginning an experiment, and the stimulus voltage for further experimentation was set at 20 V higher than the threshold voltage. An increase of the stimulus voltage above the threshold did not lead to a change in the amplitude of the resulting [Ca²⁺]_i transients. For the experiments involving substitution of intracellular K⁺ with Cs⁺ to examine the role of K⁺-dependent Na⁺/Ca²⁺ exchange (see Fig. 2E), DRG neurons were stimulated under the whole-cell voltage-clamp mode by applying 20 5-ms voltage pulses from a holding potential of –60 mV to +10 mV at a rate of 10 Hz.

Figure 2. The rate of presynaptic [Ca²⁺]_i recovery is slightly reduced by inhibition of the plasma membrane Na⁺/Ca²⁺ exchanger.

A, [Ca²⁺]_i transients in an axonal bouton (white box) were evoked by trains of action potentials (20 action potentials at 10 Hz for each train; arrows) using extracellular field stimulation under control conditions and after equimolar substitution of extracellular Na⁺ with Li⁺ (Na⁺-free). B, superimposition of [Ca²⁺]_i transients obtained under control (grey dots) and Na⁺-free (red dots) conditions, for the axonal bouton in A. Continuous lines are the result of the monoexponential fitting procedure. C, summary of time constants (τ) describing presynaptic [Ca²⁺]_i recovery in control conditions and after substitution of extracellular Na⁺ with Li⁺ (Na⁺-free). Experiments were performed as in A. The value of τ was calculated for each [Ca²⁺]_i transient. Time constants for the control condition were obtained by averaging τ for the first and second [Ca²⁺]_i transients, and those for the Na⁺-free condition were obtained by averaging τ for the third and fourth [Ca²⁺]_i transients. *P < 0.05; paired Student's t test; n = 15 boutons/5 cells D, RT-PCR analysis of NCX1–3 and NCKX1–5 expression in neuron-enriched rat DRG culture. E, summary of time constants describing presynaptic [Ca²⁺]_i recovery obtained under control and Na⁺-free (equimolar substitution with Li⁺) conditions in the absence of intracellular K⁺ (equimolar substitution with Cs⁺ in the patch pipette). The experimental protocol was similar to that described in A, except that the cells were stimulated under the whole-cell mode of patch-clamp (voltage-clamp), by 20 voltage steps from –60 to +10 mV, for 5 ms each and applied at 10 Hz. The control time constants were obtained by averaging τ for the first and second [Ca²⁺]_i transients, and the Na⁺-free time constants were obtained by averaging τ for the third and fourth [Ca²⁺]_i transients. *P < 0.05, paired Student's t test.

Analysis of the rate of [Ca²⁺]_i clearance

The recovery phase of each [Ca²⁺]_i transient was fitted with a monoexponential function defined by the formula [Ca²⁺]_i = [Ca²⁺]_i0+Aexp(–(t–t₀)/τ) using a non-linear, least-squares curve fitting algorithm (Origin 7 software), where [Ca²⁺]_i0 is the basal [Ca²⁺]_i level, A is the amplitude of the [Ca²⁺]_i response, t₀ is the time at the peak [Ca²⁺]_i and τ is the time constant describing [Ca²⁺]_i recovery. For a monoexponential [Ca²⁺]_i recovery process the rate of [Ca²⁺]_i clearance (–d[Ca²⁺]_i/dt) is a linear function of [Ca²⁺]_i that can be determined by the formula: –d[Ca²⁺]_i/dt = k[Ca²⁺]_i, where the rate constant k = 1/τ. To calculate the contribution of each specific Ca²⁺ transporting system, τ and k were calculated for each presynaptic [Ca²⁺]_i transient under control conditions (τ_control, k_control) and after inhibition of the Ca²⁺ transporter (τ_inhibit, k_inhibit) of interest. The rate constants were averaged, and the partial contribution of a specific Ca²⁺ transport was calculated as:

All the data are presented as mean ± SEM.

Measurements of mitochondrial Ca²⁺ concentrations ([Ca²⁺]_mt) in axonal boutons

[Ca²⁺]_mt was measured using the mitochondria-targeted Ca²⁺ indicator mtPericam (Nagai et al. 2001) transferred to DRG neurons using lentivirus as previously described (Medvedeva et al. 2008). To visualize axonal boutons, DRG neurons overexpressing mtPericam were loaded with tetramethylrhodamine 10K dextran (Rhod10K; Invitrogen, Carlsbad, CA, USA) via patch pipettes (200 μm). mtPericam fluorescence was excited at 410 (12) nm via a 40× oil-immersion objective (NA = 1.35, Olympus) and measured at 530 (44) nm (sampling rate = 5 Hz). At 410 nm, a decrease in mtPericam fluorescence corresponds to the [Ca²⁺]_mt increase. [Ca²⁺]_mt changes were presented as

where F is the current fluorescence intensity and F₀ is the fluorescence intensity in the resting cell. The initial rate of [Ca²⁺]_mt increase was determined by dividing half of the [Ca²⁺]_mt amplitude by the time that was required to reach one half of [Ca²⁺]_mt peak.

Reagents

Fura-2, Fura-FF and Rhod10K were obtained from Invitrogen; capsaicin was from Tocris (Ellisville, MO, USA); ionomycin was from Calbiochem (San Diego, CA, USA); agatoxin IVA was from Bachem (Torrance, CA, USA); and ω-conotoxin GVIA was from Alomone Labs (Jerusalem, Israel). All other reagents were purchased from Sigma (St Louis, MO, USA).

RT-PCR analysis

Enriched neuronal DRG cultures were prepared as previously described (Usachev et al. 2002). In brief, DRG cultures were grown in the presence of AraC (5 μm) for 5–6 days to eliminate non-neuronal cells. To provide trophic support for this glial cell-deficient culture, fresh culture medium was mixed with an equal volume of medium conditioned by DRG cultures grown in the absence of AraC for 4–6 days. These growth conditions yield cultures containing ∼75% of neurons and do not alter [Ca²⁺]_i homeostasis in DRG neurons (Usachev et al. 2002). Total RNA was isolated using the RNase easy mini kit (Qiagen, Valencia, CA, USA) and reverse transcribed and amplified using the one-step RT-PCR kit (Qiagen) according to the manufacturer's protocols. RT-PCR reactions were performed (Hybaid PCR Sprint Thermal Cycler, MA, USA) under the following conditions: 50°C for 30 min for reverse transcription, followed by 95°C for 15 min for initial PCR activation, then 30 cycles of 95°C for 1 min, 55°C for 1 min and 72°C for 1 min, and 72°C for 10 min. Amplified transcripts were separated by electrophoresis on 1% agarose gels and detected using ethidium bromide. The sets of PCR primers are listed in Table 1.

Table 1.

Primer sequences

Target gene (accession no.)	Primer sequence
NCX1	F: 5′-CTTCGTCCCACCTACAGAAT-3′
(NM_019268.2)	R: 5′-TGGTAGATGGCAGCAATGGA-3′
NCX2	F: 5′-TGCCATCCTGCTTTGACTAC-3′
(NM_078619.1)	R: 5′-GTCAACAGTGTGACCGAGAA-3′
NCX3	F: 5′-GAAACATGCAGCAGAGCAAG-3′
(NM_078620.1)	R: 5′-GACATTGCTCAGTCTCACGA-3′
NCKX1	F: 5′-CACCTTCCTGGGATCCATCATC-3′
(NM_020090.1)	R: 5′-CGATCTTCTAACATCACACTGATC-3′
NCKX2	F: 5′-TTATCATGTGGTGGGAAAGC-3′
(NM_031743.2)	R: 5′-GCTTTTTCTCTGAACCTCCC-3′
NCKX3	F: 5′-GACGTTTGCTTCCTCTACACTGT-3′
(NM_053505.1)	R: 5′-AACTCCGTCATGATGGAGAAA-3′
NCKX4	F: 5′-GGTGTGGCTGGTGACTATTATTG-3′
(NM_001108051.1)	R: 5′-CGTTGCTTCCGATGGTGTTAGAG-3′
NCKX5	F: 5′-CTTTCTCTAAGTTTACATGGTCTTAGT-C-3′
(NM_001107769.1)	R: 5′-CTTTCGGTCTAGTTTCCAGCC-G-3′
Letm1	F: 5′-AATAAGCTGCCCTCCACATTTG-3′
(NM_001005884.1)	R: 5′-TGTTTCAGCTGGCCCTTCAG-3′
NCLX	F: 5′-AGTCACCGCAGCCAAGTTCTT-C-3′
(NM_001017488.1)	R: 5′-AGTGCAGATGATGACCGTGACC-3′
MICU1	F: 5′-ACCCAATGAGAAGCAGCCAG-3′
(NM_199412.1)	R: 5′-ACGTCATGCTGCAGTTTACGC-3′
MCU	F: 5′-AGTACGGTTGTGCCCTCTGATG-3′
(NM_001106398.1)	R: 5′-AGTGGTCCTCTTCTCCGCTTTC-3′

Results

DRG neurons were co-cultured with spinal cord neurons for 10–16 days as previously described, to enable formation of sensory synapses in vitro (Medvedeva et al. 2008, 2009). Synapses formed by DRG and SC neurons grown in co-culture possess many important characteristics of the first sensory synapse, such as: (1) EPSCs are inhibited by the antagonists of 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid (AMPA)-type glutamate receptors (Gruner & Silva, 1994; Gu & Macdermott, 1997; Bao et al. 1998; Medvedeva et al. 2008; Heinke et al. 2011); (2) N-type voltage-gated Ca²⁺ channels (Ca_V2.2) play the predominant role in controlling evoked glutamate release (Fig. 1A and B; Gruner & Silva, 1994; Bao et al. 1998; Heinke et al. 2004; Cao, 2006); (3) a large proportion of these synapses are sensitive to TRPV1 agonists such as capsaicin and NADA that produce dramatic increases in the frequency of spontaneous EPSCs (Fig. 1C–E; Yang et al. 1998; Huang et al. 2002; Sikand & Premkumar, 2007; Medvedeva et al. 2008); and (4) these synapses demonstrate wind-up and long-term potentiation (LTP)-like forms of synaptic plasticity (Vikman et al. 2001; Ji et al. 2003; Woolf & Salter, 2006). Consequently, the DRG/SC co-culture system has been widely used as a reliable tool for studying the first sensory synapse under highly defined recording conditions (Gu & Macdermott, 1997; Vikman et al. 2001; Tsuzuki et al. 2004; Sikand & Premkumar, 2007; Medvedeva et al. 2008, 2009).

Figure 1. Characterization of synaptic transmission and action potential-evoked presynaptic [Ca²⁺]_i transients in DRG/SC co-culture.

A, evoked EPSCs were recorded in DRG/SC co-culture as previously described (Medvedeva et al. 2008, 2009). The N-type voltage-gated Ca²⁺ channel (VGCC) inhibitor ω-conotoxin GVIA (1 μm; ω-CnTx) nearly completely abolished the EPSCs (n = 7), whereas the P/Q-type VGCC inhibitor agatoxin IVA (200 nm; AgaTx) produced a much smaller effect (n = 5). The EPSCs were not affected by the L-type VGCC inhibitor nimodipine (5 μm; n = 7). The superimposed EPSC traces represent recordings obtained before (control; black) and during the application of VGCC inhibitors (grey) for individual synaptic pairs. B, plot summarizing the effects of VGCC inhibitors (grey bars) or vehicle control (white bar) on EPSC amplitudes. To quantify the effects of VGCC inhibitors, 10 EPSC traces obtained in the presence of a given VGCC inhibitor (3 min of treatment) were averaged and normalized to the mean amplitude of the first 10 EPSC for each synaptic pair. *P < 0.05, ***P < 0.001, one-way ANOVA with Dunnett's post test (relative to control). C, capsaicin (1 μm) strongly increases the frequency of spontaneous EPSCs. This effect is reversibly inhibited by the AMPA receptor antagonist CNQX (10 μm). Enhanced glutamate release continues long after the capsaicin washout as a result of prolonged mitochondria-dependent presynaptic [Ca²⁺]_i plateau (Medvedeva et al. 2008). D and E, two short segments of EPSC recordings from the experiment described in are shown using an expanded time scale during the treatments with capsaicin alone (D) or capsaicin combined with CNQX (E). The times corresponding to these segments are shown by arrowheads in C. F, presynaptic [Ca²⁺]_i transients evoked by repeated trains of action potentials in capsaicin-sensitive DRG neurons. Presynaptic [Ca²⁺]_i transients were evoked by trains of action potentials (2 s at 10 Hz for each train; arrows) applied at the time points t = 0, 2, 5 and 7 min. Image on the left shows the distribution of Fura-FF fluorescence (λ_ex = 380 nm) in an unstimulated cell. [Ca²⁺]_i recording was made from the axonal bouton indicated by the white box. G, the second [Ca²⁺]_i transient from the experiment in F is shown on an expanded time scale (grey dots), and the monoexponential function obtained by fitting the recovery phase of this [Ca²⁺]_i transient is shown by a continuous line. H, [Ca²⁺]_i response to the TRPV1 agonist capsaicin (1 μm, 30 s), recorded from the axonal bouton in F. [Ca²⁺]_i rapidly recovered to a new steady-state level ([Ca²⁺]_i plateau), and remained elevated for an additional 5–15 min, as previously described (Medvedeva et al. 2008). I, summary of time constants (τ; open circles) and the rate constants (k; grey triangles) calculated for each presynaptic [Ca²⁺]_i transient, using the same experimental protocol as in F. Data are for 27 presynaptic boutons/6 cells. Time and rate constants did not change significantly during the time course of the experiments; repeated one-way ANOVA with Bonferroni post test.

[Ca²⁺]_i transients in presynaptic boutons of DRG neurons were measured using the low-affinity fluorescent Ca²⁺ indicator Fura-FF (K_d∼ 5.5 μm). This dye is ideally suited for monitoring [Ca²⁺]_i changes predicted to be in the micromolar range, while the low Ca²⁺ affinity of the indicator minimizes its influence on Ca²⁺ buffering processes (Neher & Augustine, 1992). Fura-FF was loaded into small to medium-sized DRG neurons (18–32 μm) via a patch pipette, which enabled visualization of multiple axonal boutons along the processes (Figs 1–6). We have previously shown that these structures are associated with the presynaptic proteins synaptophysin and bassoon as well as the postsynaptic marker PSD95 (Medvedeva et al. 2008, 2009). [Ca²⁺]_i transients were evoked by a train of 20 action potentials applied at 10 Hz, consistent with firing patterns of primary sensory neurons in vivo (Meyer & Campbell, 1981; Slugg et al. 2000). Such stimulation produced [Ca²⁺]_i elevation from the basal level of 0.12 ± 0.05 μm to a peak of 2.14 ± 0.12 μm (n = 79 boutons/24 cells; Fig. 1F). After reaching its maximum, [Ca²⁺]_i rapidly recovered to the prestimulus level. For the majority of axonal boutons (79 of 84 tested), [Ca²⁺]_i recovery was well approximated using a monoexponential function with a time constant τ = 2.02 ± 0.11 s (n = 79 boutons/24 cells; Fig. 1G). Each cell was stimulated four times as described in Fig. 1F, and τ and the rate constant (k = 1/τ) for each [Ca²⁺]_i transient were calculated. The first two stimuli were used as controls, and treatments directed at inhibiting various Ca²⁺ transporters were applied during the third and fourth stimuli. The contribution of a specific Ca²⁺ clearance mechanism was calculated as (k_control–k_inhibit)/k_control100%, where k_control is the average rate constant in the control, and k_inhibit is the average rate constant under conditions when a specific Ca²⁺ transporter is inhibited. In the absence of treatment, τ and k did not change significantly throughout the course of the experiment (Fig. 1I). Functional expression of TRPV1 was determined at the end of each experiment by applying 1 μm capsaicin (30 s; Fig. 1H), and only capsaicin-responding cells (mean presynaptic [Ca²⁺]_i elevation ≥ 100% above baseline) were used for further analysis. We used the described approach to determine the roles of plasma membrane Na⁺/Ca²⁺ exchangers, PMCA, sarco-endoplasmic reticulum Ca²⁺-ATPase (SERCA) and mitochondria in clearing Ca²⁺ from the presynaptic boutons of DRG neurons.

Figure 6. The rate of presynaptic [Ca²⁺]_i recovery is significantly reduced by inhibition of the mitochondrial uniporter.

A and B, action potential-evoked (10 Hz for 2 s; arrows) [Ca²⁺]_i transients were recorded in control cells (A) and cells loaded with the MCU inhibitor ruthenium red (RuR; B). In the latter case, 10 μm RuR was included in the patch pipette solution, along with 200 μm Fura-FF. White boxes indicate axonal boutons from which recordings were made. C, control [Ca²⁺]_i trace (grey dots) is superimposed with a [Ca²⁺]_i trace recorded in the presence of 10 μm RuR (red dots). [Ca²⁺]_i transients were normalized to facilitate comparison of the recovery kinetics. Continuous lines indicate the plots obtained by monoexponential fitting of the corresponding [Ca²⁺]_i traces. D, [Ca²⁺]_i transients were studied in cells containing 10 μm RuR under normal (pH = 7.4) or basic (pH = 8.8) extracellular pH. [Ca²⁺]_i responses were induced by trains of action potentials (10 Hz for 2 s, arrows) generated by extracellular field stimulation. E, superimposition of [Ca²⁺]_i transients obtained under normal (pH = 7.4; grey dots) or elevated (pH = 8.8) extracellular pH (red dots), for the same cell as shown in D. Continuous lines represent mononexponential fitting of the corresponding [Ca²⁺]_i data points. F, summary of the effects of RuR and high pH on the time constant of [Ca²⁺]_i recovery. The data were obtained from experiments such as those shown in A, B and D. ***P < 0.001, one-way ANOVA with Bonferroni post test; n = 79 boutons/24 cells for control; n = 33 boutons/7 cells for RuR only; n = 16 boutons/5 cells for RuR + pH 8.8.

Presynaptic Ca²⁺ clearance in DRG neurons depends to a small extent on plasma membrane Na⁺/Ca²⁺ exchange

Plasma membrane Na⁺/Ca²⁺ exchangers are low-affinity Ca²⁺ transporters that are expressed in excitable and non-excitable cells (Blaustein & Lederer, 1999; Lytton, 2007). Two large families of the exchangers have been identified: the K⁺-independent Na⁺/Ca²⁺ exchangers (NCX proteins), and the K⁺ -dependent Na⁺/Ca²⁺ exchangers (NCKX proteins). The NCX transporters utilize the transmembrane Na⁺ gradient to extrude 1 Ca²⁺ ion from the cell in exchange for 3 or 4 Na⁺ ions moved into the cell, whereas the NCKX transporters extrude 1 Ca²⁺ and 1 K⁺ ion in exchange for 4 Na⁺ ions (Blaustein & Lederer, 1999; Lytton, 2007). We blocked plasma membrane Na⁺/Ca²⁺ exchange by replacing extracellular Na⁺ with Li⁺. This treatment produced a small but significant slowing of presynaptic [Ca²⁺]_i recovery (Fig. 2). In these experiments, the time constant of [Ca²⁺]_i recovery increased from 1.99 ± 0.24 s in control to 2.43 ± 0.31 s in Na⁺-free external solution (n = 15 boutons/5 cells; P < 0.05, paired Student's t test). The corresponding rate constants were 0.59 ± 0.06 and 0.52 ± 0.07 s⁻¹, respectively (n = 15 boutons/5 cells; P < 0.05, paired Student's t test). The overall contribution of plasma membrane Na⁺/Ca²⁺ exchange to presynaptic Ca²⁺ clearance in DRG neurons was ∼12%. Using the high-affinity Ca²⁺ indicator Fura-2 we found that the Na⁺/Ca²⁺ exchanger also controls presynaptic [Ca²⁺]_i at rest, as replacement of extracellular Na⁺ with Li⁺ resulted in a small but significant [Ca²⁺]_i elevation from 107 ± 4 to 134 ± 8 nm (n = 20 boutons/6 cells; P < 0.001, paired Student's t test; data not shown).

The NCX protein family comprises three isoforms (NCX1–3) and the NCKX family five (NCKX1–5). With the exception of NCKX1 (rod-specific isoform), all the NCX and NCKX isoforms are found in the brain (Annunziato et al. 2004; Lytton, 2007). DRG neurons express NCX2 and NCX3, whereas NCX1 is found primarily in satellite cells in the DRG (Persson et al. 2010). Which NCKX isoforms are expressed in DRG neurons is unknown. Using RT-PCR, we detected transcripts of NCX1, NCX2, NCX3 and NCKX2 in neuron-enriched DRG cultures (Fig. 2D). A very weak band corresponding to NCKX3 was also detected. Given the presence of some NCKX isoforms in DRG neurons, we examined whether presynaptic Ca²⁺ extrusion mediated by Na⁺/Ca²⁺ exchange depends on K⁺. Substitution of intracellular K⁺ with Cs⁺ inhibits NCKX2 and NCKX3 by 80 and 95%, respectively (Lee et al. 2007b). Therefore, we replaced K⁺ with an equimolar amount of Cs⁺ in the patch pipette solution (Fig. 2E). Since action potentials could not be evoked under these conditions, we stimulated the neurons by applying 20 short (5 ms) depolarization pulses (from –60 to +10 mV) at 10 Hz using the patch clamp technique. In the absence of intracellular K⁺, the slowing of [Ca²⁺]_i recovery induced by Na⁺ removal was similar to that observed in the presence of intracellular K⁺ (compare graphs in Fig. 2C and E), suggesting that NCKX exchangers do not significantly contribute to Ca²⁺ clearance.

Presynaptic Ca²⁺ clearance in DRG neurons is highly dependent on PMCA

PMCAs are key components of Ca²⁺ extrusion in central and peripheral neurons (Benham et al. 1992; Usachev et al. 2002; Wanaverbecq et al. 2003; Empson et al. 2007; Gover et al. 2007b). Four major PMCA isoforms (1–4) have been identified (Carafoli, 1991; Strehler & Zacharias, 2001), and we previously showed that PMCA2 and PMCA4 are the major isoforms expressed in DRG neurons (Usachev et al. 2002). In addition, strong PMCA4 immunoreactivity was detected in the superficial layers of the dorsal horn (Tachibana et al. 2004), the area where the majority of TRPV1-positive central processes terminate (Guo et al. 1999; Cavanaugh et al. 2011). Here we examined the functional significance of PMCAs in presynaptic boutons of DRG neurons. Ca²⁺ extrusion by PMCA requires the countertransport of protons, and thus can be blocked by raising extracellular pH (Carafoli, 1991; Schwiening et al. 1993; Usachev et al. 2002; Wanaverbecq et al. 2003). Using Fura-2, we found that blocking PMCA by raising extracellular pH to 8.8 led to baseline [Ca²⁺]_i elevation from 98 ± 5 to 178 ± 12 nm (n = 29 boutons/6 cells; P < 0.001, Student's t test; data not shown). Fura-FF-based measurements determined that PMCA inhibition markedly slowed [Ca²⁺]_i recovery in the axonal boutons of DRG neurons (Fig. 3). The time constant increased from 2.01 ± 0.19 s in control to 4.29 ± 0.53 s at pH 8.8 (n = 27 boutons/7 cells; P < 0.001, paired Student's t test). The corresponding rate constants were 0.60 ± 0.06 and 0.32 ± 0.03 s⁻¹, respectively (n = 27 boutons/7 cells; P < 0.001, paired Student's t test). The contribution of PMCA to presynaptic [Ca²⁺]_i clearance was ∼47%.

Figure 3. The rate of presynaptic [Ca²⁺]_i recovery is markedly slowed by inhibition of PMCA.

A, [Ca²⁺]_i transients were evoked by trains of 20 action potentials at 10 Hz (arrows) using extracellular field stimulation under control conditions and after raising extracellular pH to 8.8 to block PMCA. The white box indicates an axonal bouton from which [Ca²⁺]_i traces were recorded. B, superimposition of control [Ca²⁺]_i (grey dots) with that obtained under conditions of elevated extracellular pH (red dots). Continuous lines show the results of monoexponential fitting procedure for the corresponding traces. C, plot summarizing τ obtained from seven independent experiments similar to that shown in A. Control time constants were obtained by averaging τ for the first and second [Ca²⁺]_i responses, and the time constants for pH = 8.8 were obtained by averaging τ for the third and fourth [Ca²⁺]_i responses. ***P < 0.001, paired Student's t test, n = 27 boutons/7 cells. D, raising extracellular pH to 8.8 does not affect the amplitude of voltage-activated Ca²⁺ currents (I_Ca). Plot (left) summarizing I_Ca recordings (100 ms depolarizations from –60 to +10 mV) in DRG neurons (n = 6) in control and following treatments with either pH 8.8 or a blocker of voltage-gated Ca²⁺ channels, Cd²⁺ (500 μm). The I_Ca amplitudes were normalized to the amplitude of the first I_Ca response recorded for each cell. The right panel shows a superimposition of representative I_Ca traces in control conditions (black) and following extracellular pH increase to 8.8 (red).

We also found that PMCA blockade resulted in a mild but significant increase of the amplitude of electrically evoked presynaptic [Ca²⁺]_i responses from 1.83 ± 0.14 μm in control to 2.25 ± 0.19 μm at pH 8.8 (n = 27 boutons/7 cells; P < 0.01, paired Student's t test). Raising extracellular pH to 8.8 did not have any significant effect on voltage-gated Ca²⁺ currents in DRG neurons, as determined by whole-cell patch-clamp recordings (Fig. 3D). Thus, PMCA strongly contributes to multiple aspects of presynaptic [Ca²⁺]_i signalling in DRG neurons, including maintenance of low [Ca²⁺]_i at rest, regulation of the [Ca²⁺]_i clearance rate and setting the amplitude of [Ca²⁺]_i elevation in response to action potentials.

Presynaptic Ca²⁺ clearance does not depend on SERCA

SERCA is a high-affinity Ca²⁺ pump that serves to lower cytosolic Ca²⁺ and replenish intracellular Ca²⁺ stores for subsequent Ca²⁺ release via the ryanodine or inositol 1,4,5-trisphosphate (IP₃) receptors (Henzi & MacDermott, 1992; Pozzan et al. 1994; Thomas & Hanley, 1994). Three SERCA isoforms (1–3) have been identified, two of which (SERCA2 and SERCA3) are present in the brain (Mata & Sepulveda, 2005; Vangheluwe et al. 2009). SERCA plays an important role in regulating [Ca²⁺]_i in the somata of DRG and trigeminal neurons, and its contribution to [Ca²⁺]_i clearance can be 50% and higher, depending on the stimulation protocol and sensory neuron population (Shmigol et al. 1994; Usachev & Thayer, 1999; Lu et al. 2006; Gover et al. 2007a). We tested the effects of a selective SERCA inhibitor, cyclopiazonic acid (CPA, 5 μm), on [Ca²⁺]_i in the presynaptic boutons of DRG neurons. Application of 5 μm CPA did not affect the recovery kinetics of presynaptic [Ca²⁺]_i transients (Fig. 4A–C), and the time constants were 1.97 ± 0.27 and 2.21 ± 0.31 s in control and in the presence of CPA, respectively (n = 18 boutons/5 cells; P = 0.28, paired Student's t test). The corresponding rate constants were 0.66 ± 0.08 and 0.62 ± 0.09 s⁻¹, respectively (P = 0.39, paired Student's t test). Similar results were obtained using another SERCA inhibitor, thapsigargin (Tg) (Thomas & Hanley, 1994; Thayer et al. 2002). Treatment with 1 μm Tg did not significantly change the time constant, which was 2.18 ± 0.17 s in the absence and 2.13 ± 0.19 s in the presence of 1 μm Tg (n = 16 boutons/4 cells; P = 0.72, paired Student's t test; data not shown). Thus, Ca²⁺ clearance in the presynaptic boutons of DRG neurons is independent of SERCA.

To test whether functional endoplasmic reticulum (ER) Ca²⁺ stores are present in the axonal boutons of DRG neurons, we applied an activator of Ca²⁺ release from ryanodine-sensitive stores, caffeine, in a Ca²⁺-free buffer. To improve the detection of potentially small [Ca²⁺]_i changes in response to caffeine, we used a high-affinity Ca²⁺ indicator, Fura-2 (K_d = 225 nM), in these experiments. Application of 10 mm caffeine produced a small but detectable Ca²⁺ release in presynaptic boutons with an amplitude of 76 ± 11 nm (n = 27 boutons/5 cells; Fig. 4D and E). This value was significantly smaller than that for the caffeine-induced [Ca²⁺]_i responses in the somata of DRG neurons (159 ± 39 nm; n = 5; P < 0.01, non-paired Student's t test). In contrast, [Ca²⁺]_i transients induced by electrical stimulation were larger in the axonal boutons than those in the somata (Fig. 4D; arrow). Notably, after caffeine-induced depletion of the ER Ca²⁺ stores, switching from a Ca²⁺-free to normal solution containing 1.3 mm Ca²⁺ led to a pronounced [Ca²⁺]_i overshoot in both presynaptic boutons and somata. This transient [Ca²⁺]_i elevation is probably the result of activation of store-operated Ca²⁺ channels, as previously described for DRG neurons (Usachev & Thayer, 1999; Lu et al. 2006; Gemes et al. 2011). Moreover, depleting Ca²⁺ stores with 5 μm CPA or 1 μm Tg induced small but significant presynaptic [Ca²⁺]_i elevations from 106 ± 3 to 152 ± 5 nm for CPA (n = 38 boutons/7 cells; P < 0.001 paired Student's t test) and from 116 ± 8 to 134 ± 7 nm for Tg (n = 14 boutons/4 cells; P < 0.001 paired Student's t test), consistent with the functional presence of store-operated Ca²⁺ channels in presynaptic boutons of DRG neurons.

Mitochondria regulate the amplitude and duration of presynaptic [Ca²⁺]_i transients in DRG neurons

Presynaptic terminals and boutons are densely packed with mitochondria that support high energy demand at the synapses (Hollenbeck, 2005; Ly & Verstreken, 2006). Mitochondria also contribute to presynaptic Ca²⁺ signalling and synaptic plasticity at several central synapses, as well as at the neuromuscular junction (Billups & Forsythe, 2002; David & Barrett, 2003; Jonas, 2004; Garcia-Chacon et al. 2006; Lee et al. 2007a). We recently reported that mitochondria control TRPV1-mediated prolonged elevations of presynaptic [Ca²⁺]_i and glutamate release at the first sensory synapse (Medvedeva et al. 2008). Here we examined whether mitochondria are also involved in shaping presynaptic [Ca²⁺]_i transients in response to brief trains of action potentials. Ca²⁺ uptake by mitochondria is mediated by the mitochondrial Ca²⁺ uniporter (MCU), and the mitochondrial membrane potential ΔΨ_mt (∼–150 mV) provides the driving force for the Ca²⁺ uptake (Bernardi, 1999; Nicholls & Budd, 2000; Thayer et al. 2002). Mitochondrial Ca²⁺ uptake can be blocked by depolarizing mitochondria with a protonophore, carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). We examined the effects of 1 μm FCCP on presynaptic [Ca²⁺]_i transients induced by electrical stimulation (10 Hz for 2 s). FCCP treatment was combined with application of the F₁,F₀ ATP-synthase inhibitor oligomycin (2 μm) to prevent ATP depletion via the reverse mode of the synthase (Nicholls & Ward, 2000; Toescu & Verkhratsky, 2003). Using Fura-2, we found that FCCP+oligomycin increased baseline presynaptic [Ca²⁺]_i from 126 ± 6 to 163 ± 7 nm (n = 17 boutons/4 cells; P < 0.001 Student's t test). Fura-FF-based measurements revealed that FCCP produced a marked slowing of [Ca²⁺]_i recovery in the axonal boutons of DRG neurons (Fig. 5). The time constants were 2.11 ± 0.21 s in control and 4.60 ± 0.39 s after the FCCP treatment (n = 17 boutons/5 cells; P < 0.001, paired Student's t test). The corresponding rate constants were 0.54 ± 0.05 and 0.25 ± 0.02 s⁻¹, respectively (n = 17 boutons/5 cells; P < 0.001, paired Student's t test). We also found that the amplitudes of the [Ca²⁺]_i responses increased from 2.17 ± 0.24 μm in control to 3.45 ± 0.42 μm in the presence of FCCP (n = 17 boutons/5 cells; P < 0.01, paired Student's t test). Whole-cell patch-clamp recordings showed that FCCP+oligomycin treatment reduced voltage-gated Ca²⁺ currents by ∼30% in DRG neurons (Fig. 5D), probably as a result of reduced Ca²⁺ buffering and, consequently, increased Ca²⁺-dependent inhibition of voltage-gated Ca²⁺ channels (Hernandez-Guijo et al. 2001). Overall, our data suggest that mitochondria play an important role in buffering presynaptic Ca²⁺ influx and clearing [Ca²⁺]_i elevations after periods of neuronal activity.

Figure 5. Inhibition of mitochondrial uptake by FCCP strongly reduces the rate of presynaptic Ca²⁺ clearance.

A, electrically evoked [Ca²⁺]_i transients (10 Hz, 2 s; arrows) were recorded from axonal boutons of capsaicin-sensitive DRG neurons (image on the left) in control and during treatment with the mitochondrial protonophore FCCP (1 μm) and F1,Fo ATP-synthase inhibitor oligomycin B (2 μm). B, superimposition of [Ca²⁺]_i traces in control (grey dots; second trace in A) conditions and in the presence of 1 μm FCCP + 2 μm oligomycin (red dots; third trace in A). Plots obtained by monoexponential fitting of [Ca²⁺]_i recovery are shown by continuous lines. C, plot summarizing the time constant data for all the experiments as those shown in A. ***P < 0.001, paired Student's t test. D, treatment with 1 μm FCCP + 2 μm oligomycin decreased the amplitude of voltage-activated Ca²⁺ currents (I_Ca) in DRG neurons. Amplitudes of I_Ca (depolarization from –60 to +10 mV for 100 ms) were normalized to the amplitude of the first evoked I_Ca for each cell (n = 5). The effect of FCCP + oligomycin was significant starting from t = 5 min (P < 0.001; repeated one-way ANOVA with Dunnett's post test relative to control I_Ca taken at t = 3 min). I_Ca was completely blocked by 500 μm Cd²⁺. Plot on the right shows superimposition of representative I_Ca traces in control conditions (black) and after 2 min of treatment using FCCP + oligomycin (red).

MCU-mediated Ca²⁺ transport can be blocked directly by ruthenium red (RuR; Bernardi, 1999; Thayer et al. 2002). An important advantage of this approach is that it does not disrupt either the mitochondrial electrochemical gradient or the ΔΨ_mt. RuR is membrane impermeable and thus was delivered to the cell through a patch pipette (10 μm) along with Fura-FF (see Methods). The amplitudes of action potential-evoked [Ca²⁺]_i transients were significantly larger in RuR-loaded boutons (4.63 ± 0.45 μm; 33 boutons/7 cells) than in control cells (2.02 ± 0.12 μm; 79 boutons/24 cells; P < 0.001, unpaired Student's t test). As shown in Fig. 6, treatment with RuR also led to a marked slowing of presynaptic [Ca²⁺]_i recovery. The time constants were 2.06 ± 0.10 s in control cells (n = 79 boutons/24 cells) and 3.35 ± 0.16 s in RuR-treated cells (n = 33 boutons/7 cells; P < 0.001, unpaired Student's t test). The corresponding rate constants were 0.58 ± 0.03 s⁻¹ in control cells (n = 79 boutons/24 cells) and 0.35 ± 0.02 s⁻¹ in RuR-loaded cells (n = 33 boutons/7 cells; P < 0.001, unpaired Student's t test). Thus, mitochondria blunt presynaptic [Ca²⁺]_i elevations and play important role in presynaptic Ca²⁺ clearance in DRG neurons. The mitochondrial contribution to [Ca²⁺]_i recovery process ranges between 40% (based on RuR experiments) and 54% (based on FCCP experiments).

It has been proposed that the Ca²⁺ accumulated in mitochondria can stimulate several Ca²⁺-dependent dehydrogenases and thereby enhance the production of ATP (Hajnoczky et al. 1995; Denton, 2009). Thus, the inhibition of mitochondrial Ca²⁺ uptake could potentially slow ATP-dependent transport, such as PMCA-mediated Ca²⁺ efflux. We therefore tested the role of PMCA in RuR-treated neurons. Inhibiting PMCA by raising extracellular pH to 8.8 produced a more than two-fold increase in the time constant, from 3.44 ± 0.20 s in RuR only-treated cells to 7.80 ± 0.53 s after a combined RuR + pH 8.8 treatment (n = 16 boutons/5 cells; P < 0.001, paired Student's t test; Fig. 6D-F). The magnitude of the pH 8.8 effect in RuR-treated cells was similar to the pH 8.8 effect in control cells (Fig. 3), indicating that PMCA-mediated Ca²⁺ transport was independent of the RuR-sensitive mitochondrial Ca²⁺ uptake under the described experimental conditions. We were unable to perform similar experiments in FCCP-treated cells because the combined application of FCCP and pH 8.8 destabilized presynaptic [Ca²⁺]_i in DRG neurons (n = 14 boutons/3 cells; data not shown).

Properties of Ca²⁺ uptake by presynaptic mitochondria in DRG neurons

Given that mitochondria significantly contribute to the amplitude and time course of changes in presynaptic [Ca²⁺]_i, we next examined the properties of Ca²⁺ transport by presynaptic mitochondria in capsaicin-sensitive DRG neurons. In these experiments, we directly monitored [Ca²⁺]_mt using a mitochondria-targeted Ca²⁺ indicator, mtPericam (Nagai et al. 2001), as previously described (Medvedeva et al. 2008). DRG neurons transfected with mtPericam were subsequently loaded with Rhod10K through a patch pipette to enable visualization of axonal varicosities (Fig. 7A). Stimulating DRG neurons with trains of action potentials produced a rapid [Ca²⁺]_mt elevation. Only 3–5 action potentials were required to produce detectable Ca²⁺ uptake by mitochondria (Fig. 7B). Comparison of the time courses of [Ca²⁺]_i and [Ca²⁺]_mt showed that mitochondrial Ca²⁺ uptake developed with a 1–2 s delay relative to the presynaptic [Ca²⁺]_i rise. For example, in response to a stimulus with 20 action potentials (10 Hz, 2 s), [Ca²⁺]_i and [Ca²⁺]_mt reached their peak values within 1.9 ± 0.1 s (79 boutons/24 cells) and 3.5 ± 0.2 s (n = 22 boutons/5 cells; P < 0.001, unpaired Student's t test compared to the [Ca²⁺]_i rising time) after beginning of the stimulation, respectively. In addition, Ca²⁺ removal from mitochondria occurred at a markedly slower rate (half recovery time = 9.2 ± 0.8 s; n = 22 boutons/5 cells) than Ca²⁺ clearance from the presynaptic cytosol (Fig. 7D). The fact that [Ca²⁺]_mt remained elevated long after [Ca²⁺]_i has fully recovered suggests that mitochondria serve as an integrator of consecutive presynaptic [Ca²⁺]_i spikes produced by repeated neuronal stimulation.

Figure 7. Ca²⁺ changes in presynaptic mitochondria ([Ca²⁺]_mt) monitored using mitochondria-targeted Ca²⁺ indicator mtPericam.

A, DRG neurons were transfected with mtPericam (green) and subsequently loaded with rhodamine 10K dextran (Rhod10K; 200 μm; red) via a patch pipette to facilitate the visualization of axonal boutons. B, mitochondrial Ca²⁺ changes were evoked by 3, 5 or 20 action potentials (10 Hz) and recorded from two axonal boutons (b1 and b2) indicated by white boxes in A. C, [Ca²⁺]_mt responses to 1 μm capsaicin (30 s) for the same axonal boutons as in B. D, comparison of typical [Ca²⁺]_i (black) and [Ca²⁺]_mt (green) changes induced by 20 action potentials (10 Hz for 2 s). The [Ca²⁺]_mt trace was obtained by averaging traces from five axonal boutons of the cell depicted in A. The [Ca²⁺]_i trace was generated by averaging traces from four axonal boutons of a different cell loaded with Fura-FF. The traces were normalized along the vertical axis.

Next we addressed the important and widely debated question of how mitochondrial Ca²⁺ uptake depends on [Ca²⁺]_i. The reported values for the Ca²⁺ affinity of MCU cover a wide range (0.5–20 μm), depending on the preparation, cell type and method of monitoring mitochondrial Ca²⁺ uptake (Gunter & Pfeiffer, 1990; Bernardi, 1999; Spat et al. 2008). In neuronal preparations, high Ca²⁺ sensitivity of MCU (K_0.5∼0.4–0.5 μm) was reported for presynaptic mitochondria at nerve–muscular junctions of the fruit-fly and lizard (David, 1999; Chouhan et al. 2010). In contrast, mitochondrial Ca²⁺ uptake in the cell body of bullfrog sympathetic neurons was estimated to have a much lower Ca²⁺ affinity (K_0.5∼ 10 μm; Colegrove et al. 2000). Comparable quantifications in mammalian neurons are lacking. To determine the dependence of mitochondrial Ca²⁺ uptake on [Ca²⁺]_i in axonal boutons of DRG neurons, we compared [Ca²⁺]_i and [Ca²⁺]_mt elevations induced by depolarizing pulses of incremental amplitude (extracellular buffers containing 10, 15, 20, 30, 50 or 90 mm KCl; Fig. 8), as well as by treating cells with 1 μm capsaicin, which is known to produce large presynaptic [Ca²⁺]_i and [Ca²⁺]_mt elevations in DRG neurons (Medvedeva et al. 2008). To facilitate detection of small presynaptic [Ca²⁺]_i changes in response to mild to moderate depolarization (10–50 mm KCl), we used the high-affinity Ca²⁺ indicator Fura-2, while the low-affinity Ca²⁺ indicator Fura-FF was employed for monitoring [Ca²⁺]_i produced by the stronger stimuli such as 90 mm KCl and 1 μm capsaicin. Since the excitation spectra of Fura-2 (Fura-FF) and mtPericam overlap significantly, the measurements of [Ca²⁺]_i and [Ca²⁺]_mt were carried out in parallel experiments. Figure 8 shows representative presynaptic [Ca²⁺]_i and [Ca²⁺]_mt traces in response to incremental depolarization. [Ca²⁺]_i changes in axonal boutons were detectable in response to 10 mm KCl, whereas triggering [Ca²⁺]_mt response required 15 mm KCl to produce a stronger depolarization. Analysis of the initial rate of the [Ca²⁺]_mt rise as a function of presynaptic [Ca²⁺]_i elevations showed that Ca²⁺ uptake by mitochondria was induced by a [Ca²⁺]_i elevation as little as 200–300 nm (K_0.5∼ 600 nm); the rate of [Ca²⁺]_mt rise (V_MCU) progressively increased with the [Ca²⁺]_i increase, until V_MCU reached its maximum once [Ca²⁺]_i began to exceed ∼2 μm (Fig. 8E). Thus, the mitochondrial Ca²⁺ uptake mechanism in axonal boutons of DRG neurons is highly sensitive to cytosolic Ca²⁺. Moreover, the Hill coefficient was 2.3, suggesting that MCU activation by Ca²⁺ is cooperative (Fig. 8E).

Finally, we examined the expression of several recently identified components of mitochondrial Ca²⁺ transport. MCU has been proposed to mediate mitochondrial Ca²⁺ uptake (Baughman et al. 2011; De Stefani et al. 2011), whereas the Na⁺/Ca²⁺/Li⁺ exchanger (NCLX) is probably responsible for mitochondrial Na⁺/Ca²⁺ exchange (Palty et al. 2010). Leucine zipper EF hand-containing transmembrane protein (Letm1) was identified as a mitochondrial Ca²⁺/H⁺ antiporter that can contribute to both mitochondrial Ca²⁺ uptake and release, depending on the mitochondrial Ca²⁺ concentration (Jiang et al. 2009). In addition, the mitochondrial calcium uptake 1 (MICU1) has been proposed to function as an essential regulator of mitochondrial Ca²⁺ uptake (Perocchi et al. 2010; Mallilankaraman et al. 2012). RT-PCR analysis of RNA extracted from neuron-enriched DRG cultures demonstrated the presence of all four transcripts (Fig. 8F).

Discussion

The present study identifies mechanisms of presynaptic [Ca²⁺]_i regulation in an in vitro model of the first sensory synapse. This synapse controls nociceptive transmission from the periphery to the CNS, and plasticity at this synapse contributes to bidirectional modulation of pain processing (Ji et al. 2003; Woolf & Salter, 2006; Kuner, 2010; Ruscheweyh et al. 2011). We found that presynaptic [Ca²⁺]_i in DRG is at ∼100 nm at rest, and that this low baseline [Ca²⁺]_i is maintained by Ca²⁺ extrusion via PMCA and NCX, as well by Ca²⁺ buffering into intracellular Ca²⁺ organelles. Presynaptic [Ca²⁺]_i transients induced by brief electrical stimulation rapidly recovered to the baseline with a time constant of ∼2 s. The ER Ca²⁺ pumps did not significantly contribute to presynaptic [Ca²⁺]_i clearance at this synapse under the described stimulation conditions. On the other hand, PMCA-mediated Ca²⁺ extrusion and Ca²⁺ uptake by presynaptic mitochondria had major impacts on the rate of [Ca²⁺]_i clearance, accounting for ∼47 and 40%, respectively, of the [Ca²⁺]_i recovery. Ca²⁺ imaging in presynaptic mitochondria revealed a high sensitivity of MCU to [Ca²⁺]_i elevations, with 3–5 action potentials being sufficient to produce detectable Ca²⁺ uptake by presynaptic mitochondria. Quantification of the rate of mitochondrial Ca²⁺ uptake as a function of [Ca²⁺]_i yielded a K_0.5 of ∼600 nm, and a Hill coefficient of 2.3. These values are consistent with the high Ca²⁺ affinity of presynaptic MCU (K_0.5∼ 0.4–0.5 μm) reported for non-mammalian synapses (David, 1999; Chouhan et al. 2010). In addition, the Hill coefficient of 2.3 is similar to the previously reported value of ∼2 for isolated mitochondria and intact cells (Gunter & Pfeiffer, 1990; Colegrove et al. 2000), and suggests that Ca²⁺ cooperatively binds to and activates MCU in the presynaptic mitochondria of DRG neurons.

Our finding that presynaptic [Ca²⁺]_i recovers in DRG neurons with τ∼ 2 s is in good agreement with the [Ca²⁺]_i recovery rates reported for presynaptic boutons in the hippocampus (1–3 s) (Scotti et al. 1999; Scott & Rusakov, 2006) and neurohypophysial axonal terminals (1–2 s) (Lee et al. 2002), although faster [Ca²⁺]_i clearance kinetics have been observed in presynaptic terminals of Purkinje and cerebellar granule cells (τ∼ 200–300 ms) (Mintz et al. 1995; Regehr & Atluri, 1995). Our data are also in good agreement with the observations in human nociceptive C-fibres, where electrically evoked [Ca²⁺]_i transients recovered to baseline with a half-recovery time of ∼2–3 s (Mayer et al. 1999). However, a 5–10-fold slower rate of [Ca²⁺]_i clearance was reported for sensory terminals of the rat cornea (Gover et al. 2007b), suggesting that Ca²⁺ clearance mechanisms at these sensory terminals originating from the trigeminal ganglia appear to be quite different from those in the peripheral and central processes of DRG neurons. This may not be surprising given that mitochondria do not participate in [Ca²⁺]_i clearance from corneal terminals (Gover et al. 2007b), but significantly impact presynaptic [Ca²⁺]_i recovery in DRG neurons (Figs 5 and 6). It is also possible that use of the high-affinity Ca²⁺ indicator Oregon Green BAPTA-1 dextran (K_d∼170 nm) to measure [Ca²⁺]_i in the corneal terminals (Gover et al. 2007b) might have also contributed to a slowed rate of [Ca²⁺]_i recovery in that work (Neher & Augustine, 1992; Kirischuk et al. 1999).

Our finding that SERCA does not significantly contribute to presynaptic Ca²⁺ clearance in DRG neurons (Fig. 4) is somewhat surprising given that SERCA plays major roles in setting the duration and amplitude of [Ca²⁺]_i responses in the somata of DRG and trigeminal sensory neurons (Usachev & Thayer, 1999; Lu et al. 2006; Gover et al. 2007a; Rigaud et al. 2009). Just the same, our data are consistent with a negligible role of SERCA in peripheral sensory terminals of the cornea (Gover et al. 2007b), suggesting that the impact of SERCA on [Ca²⁺]_i regulation differs substantially between the cell body and axons in sensory neurons. However, we cannot rule out the possibility that SERCA blockade led to enhanced function of PMCA or other Ca²⁺ transporters, thereby masking the effects of SERCA inhibitors. It is likely that the presynaptic boutons of primary sensory neurons contain functional intracellular Ca²⁺ stores associated with the ER. First, the SERCA inhibitor Tg was shown to modulate glutamate release at the first sensory synapse (Tsuzuki et al. 2004). Second, we found that an activator of ryanodine receptors, caffeine, induced Ca²⁺ release from the intracellular stores in presynaptic boutons of DRG neurons (Fig. 4D). Although SERCA pumps do not contribute significantly to the rate of presynaptic [Ca²⁺]_i recovery, Ca²⁺ stores are involved in the maintenance of low [Ca²⁺]_i at rest, as the SERCA inhibitors CPA and Tg produced 20–50 nm elevation of the [Ca²⁺]_i baseline. The latter could be explained either by activation of Ca²⁺ influx through presynaptic store-operated Ca²⁺ channels (Usachev & Thayer, 1999; Gemes et al. 2011) or the role of SERCA pumps in controlling presynaptic [Ca²⁺]_i at rest.

Na⁺/Ca²⁺ exchange of the plasma membrane is another mechanism that can contribute to Ca²⁺ clearance. Based on our RT-PCR data (Fig. 2D), DRG neurons express three Na⁺/Ca²⁺ exchanger isoforms that are K⁺-independent (NCX1, NCX2 and NCX3) and one that is K⁺-dependent (NCKX2). The functional analysis showed that plasma membrane Na⁺/Ca²⁺ exchange regulated the resting [Ca²⁺]_i and contributed approximately 12% to presynaptic Ca²⁺ clearance in DRG neurons (Fig. 2), and that this transport was probably mediated by the NCX, but not NCKX, isoforms (Fig. 2E). Only limited information is available about the specific roles of NCX and NCKX at various synapses. NCKX was shown to be the major mechanism mediating Ca²⁺ extrusion from axonal terminals of the rat neurohypophysis (Lee et al. 2002), whereas both NCX and NCKX contribute to presynaptic [Ca²⁺]_i clearance at the calyx of Held synapse (Kim et al. 2005). Notably, deletion of NCX2 results in increased paired-pulsed facilitation of synaptic transmission and enhanced LTP in the CA1 region of the hippocampus (Jeon et al. 2003), whereas deletion of NCKX2 leads to a profound loss of the hippocampal LTP (Li et al. 2006). Thus, it is likely that the NCX and NCKX isoforms play distinct roles at various synapses.

Identification of PMCA as a key regulator of presynaptic [Ca²⁺]_i at the sensory synapses is in agreement with the major role of PMCA in controlling [Ca²⁺]_i recovery in the somata of DRG and trigeminal sensory neurons (Benham et al. 1992; Usachev et al. 2002; Gover et al. 2007a), as well as in peripheral sensory terminals innervating the cornea (Gover et al. 2007b). PMCA is a high-affinity (0.1–0.2 μm) ATP-dependent Ca²⁺ transporter that extrudes Ca²⁺ from the cell (Carafoli, 1991; Elwess et al. 1997). The PMCA family of proteins consist of four major isoforms, PMCA1–4, and alternative RNA splicing of these gives rise to nearly 30 PMCA proteins that differ in their tissue distribution, function and regulatory properties (Strehler & Zacharias, 2001; Thayer et al. 2002). PMCA2 (splice variants 2a and 2b) and PMCA4 (splice variant 4b) are the major PMCA isoforms expressed in the DRG (Usachev et al. 2002). PMCA4 is highly expressed in the superficial lamina of the dorsal horn, where the majority of TRPV1-positive sensory fibres terminate; PMCA2, on the other hand, is found in the deeper lamina of the dorsal horn (Tachibana et al. 2004). PMCA2 has very high affinity for Ca²⁺ (∼0.09 μm for PMCA2a and ∼0.06 μm for PMCA2b), suggesting that it plays an important role in maintaining low resting [Ca²⁺]_i (Elwess et al. 1997). PMCA4b exhibits a somewhat lower apparent affinity for Ca²⁺ (∼0.16 μm) (Elwess et al. 1997), but its activity can be markedly increased by PKC-mediated phosphorylation (Enyedi et al. 1996). Notably, this mechanism is responsible for a marked PKC-mediated acceleration of Ca²⁺ efflux from the somata and axonal boutons of DRG neurons (Usachev et al. 2002).

PMCA2 and PMCA4 are also highly sensitive to stimulation by calmodulin (CaM). Ca²⁺-dependent binding of CaM to the PMCA autoinhibitory domain, located in its carboxy-terminus, disinhibits the pump, which results in a marked increase of both Ca²⁺ affinity and maximal velocity (V_max) of PMCA (Carafoli, 1991; Strehler & Zacharias, 2001). The two major PMCA isoforms in DRG neurons, PMCA2b and PMCA4b (Usachev et al. 2002), are characterized by unusually slow dissociation of CaM (>20 min) (Caride et al. 1999, 2001). This implies that even a brief elevation in [Ca²⁺]_i will produce a long-lasting CaM/Ca²⁺-dependent stimulation of PMCA. Indeed, a priming [Ca²⁺]_i elevation has been reported to result in long-term enhancement of Ca²⁺ extrusion (∼1 h) in DRG neurons (Pottorf & Thayer, 2002). Thus, PMCA-mediated Ca²⁺ extrusion in sensory neurons is subject to modulation by both neuronal activity and PKC. In the context of spinal pain processing, the described mechanisms are predicted to accelerate presynaptic [Ca²⁺]_i clearance during central sensitization, which is known to be associated with heightened neuronal activity in the spinal cord and enhanced release of neuromodulators coupled to PKC signalling (e.g. bradykinin and substance P). However, direct testing of this idea is hampered by a lack of isoform-specific inhibitors of PMCA and technical difficulties associated with presynaptic Ca²⁺ imaging in spinal cord slices or in vivo.

Ca²⁺ uptake by presynaptic mitochondria is another major mechanism responsible for presynaptic [Ca²⁺]_i clearance in DRG neurons. Electron microscopy has demonstrated numerous presynaptic mitochondria in the central processes of primary sensory neurons, including TRPV1-expressing boutons (Maxwell & Rethelyi, 1987; Rethelyi et al. 1989; Hwang et al. 2003). We also found that axonal boutons are intensely labelled by the mitochondria-targeted Ca²⁺ indicator mtPericam (Medvedeva et al. 2008; Figs 7 and 8). Presynaptic clustering of mitochondria is consistent with the high energy demand at the synapses. Mitochondria also play important roles in shaping presynaptic [Ca²⁺]_i signals at various synapses (Billups & Forsythe, 2002; David & Barrett, 2003; Jonas, 2004; Garcia-Chacon et al. 2006; Lee et al. 2007a), including the first sensory synapse (Medvedeva et al. 2008; Figs 5 and 6). Mitochondria buffer excessive Ca²⁺ via the Ca²⁺ uniporter mechanism, and subsequently release accumulated Ca²⁺ back into the cytosol via Na⁺/Ca²⁺ exchange or other mechanisms (Bernardi, 1999). Ca²⁺ buffering by mitochondria limits the amplitude of the cytosolic [Ca²⁺]_i elevation, and accelerates the initial recovery of presynaptic [Ca²⁺]_i (Medvedeva et al. 2008; Figs 5 and 6). Our data suggest that the impact of mitochondria on the amplitude of presynaptic [Ca²⁺]_i responses is much greater than that of PMCA (FCCP or RuR induced ∼60–100% increase in the amplitude of [Ca²⁺]_i response compared to ∼20% increase produced by pH 8.8). One possible explanation of this difference is that Ca²⁺ transport by the mitochondrial uniporter is much faster than that by PMCA (Miller, 1991; Elwess et al. 1997; Gunter & Sheu, 2009).

Our comparison of the time courses of [Ca²⁺]_i and [Ca²⁺]_mt changes induced by electrical stimulation demonstrated that [Ca²⁺]_mt reaches its peak by 1–2 s later than presynaptic [Ca²⁺]_i (Fig. 7D), indicating that mitochondrial Ca²⁺ uptake continues during the initial phase of [Ca²⁺]_i recovery. This observation is consistent with the role of mitochondrial Ca²⁺ uptake in [Ca²⁺]_i recovery in the axonal boutons of DRG neurons. Based on the effects of the mitochondrial Ca²⁺ uniporter inhibitor RuR, we estimate that mitochondria account for at least 40% of the [Ca²⁺]_i recovery in the axonal boutons of DRG neurons. Although our analysis using the protonophore FCCP suggested that MCU has a somewhat stronger effect on the rate of presynaptic Ca²⁺ clearance (∼54%), this difference could be due to a partial inhibition of ATP synthesis by FCCP.

Ca²⁺ initially accumulated by mitochondria can be subsequently released back into the cytosol. Following intense (e.g. tetanic) stimulation, strong Ca²⁺ efflux from mitochondria can markedly slow recovery of presynaptic [Ca²⁺]_i leading to a [Ca²⁺]_i plateau lasting many minutes after stimulation is terminated (Tang & Zucker, 1997; Garcia-Chacon et al. 2006; Lee et al. 2007a). We previously reported that TRPV1 activation or intense electrical stimulation (20 Hz for 30 s) results in a long-lasting presynaptic [Ca²⁺]_i plateau associated with enhanced glutamate release in DRG neurons (Medvedeva et al. 2008). In contrast, mild electrical stimulation such as that used in the present study (10 Hz for 2 s) was insufficient to generate a [Ca²⁺]_i plateau in the same axonal boutons, in spite of the contribution of mitochondria to Ca²⁺ clearance. Thus, a specific role of presynaptic mitochondria at the sensory synapse depends on the type and intensity of the stimulus.

In addition to regulating presynaptic [Ca²⁺]_i, mitochondrial Ca²⁺ uptake can stimulate the production of ATP via Ca²⁺-dependent dehydrogenases (Hajnoczky et al. 1995; Denton, 2009; Chouhan et al. 2012) and the generation of reactive oxygen species (ROS) (Nicholls & Budd, 2000; Hongpaisan et al. 2004). Both effects require strong neuronal activation (Hongpaisan et al. 2004; Chouhan et al. 2012), and this probably explains why mitochondrial Ca²⁺ uptake following the mild stimulation used in our study did not detectably impact ATP-dependent Ca²⁺ transport (Fig. 6D–F). In pathological pain states, neuronal activity in the spinal cord is markedly intensified, and the resulting enhancement of ROS production is thought to contribute to central sensitization (Schwartz et al. 2009). Notably, recent data suggest that mitochondrial Ca²⁺ uptake is essential for ROS-dependent plasticity in the spinal cord, and for the development of persistent pain (Kim et al. 2011).

In summary, this study provides the first detailed characterization of Ca²⁺ clearance mechanisms in presynaptic boutons of DRG neurons, and identifies PMCA and mitochondria as major regulators of Ca²⁺ signalling at the first sensory synapse. Development of new pharmacological and genetic tools targeting specific PMCA isoforms and mitochondrial Ca²⁺ transporters will be required to determine their precise contribution to the various forms of synaptic plasticity involved in spinal pain processing, such as wind-up, LTP and central sensitization.

Glossary

AraC

cytosine β-d-arabinofuranoside

[Ca²⁺]_i

cytosolic Ca²⁺ concentration

[Ca²⁺]_mt

mitochondrial Ca²⁺ concentration

CaM

calmodulin

CPA

cyclopiazonic acid

DRG

dorsal root ganglia

ΔΨ_mt

mitochondrial membrane potential

EPSC

excitatory postsynaptic current

ER

endoplasmic reticulum

FCCP

carbonyl cyanide p-trifluoromethoxyphenylhydrazone

MCU

mitochondrial Ca²⁺ uniporter

MICU1

mitochondrial calcium uptake 1

Letm1

leucine zipper EF hand-containing transmembrane protein

LTP

long-term potentiation

NCX

Na⁺/Ca²⁺ exchanger

NCKX

K⁺-dependent Na⁺/Ca²⁺ exchanger

NCLX

Na⁺/Ca²⁺/Li⁺ exchanger

PKA

protein kinase A

PKC

protein kinase C

PMCA

plasma membrane Ca²⁺-ATPase

Rhod10K

tetramethylrhodamine 10K dextran

RuR

ruthenium red

SC

spinal cord

SERCA

sarco-endoplasmic reticulum Ca²⁺-ATPase

Tg

thapsigargin

TRPV1

transient receptor potential vanilloid 1

Author contributions

Y.M.U. and L.P.S. designed the work. L.P.S., Y.M.U., M.S.K., P.R.H. and Y.V.M. performed the experiments. Y.M.U. wrote the paper. All authors provided critical intellectual input to the text and figures and approved the final version of the article.