Haploinsufficiency of the 22q11.2-microdeletion gene Mrpl40 disrupts short-term synaptic plasticity and working memory through dysregulation of mitochondrial calcium

Abstract

Hemizygous deletion of a 1.5- to 3-megabase region on chromosome 22 causes 22q11.2 deletion syndrome (22q11DS), which constitutes one of the strongest genetic risks for schizophrenia. Mouse models of 22q11DS have abnormal short-term synaptic plasticity (STP) that contributes to working memory deficiencies similar to those in schizophrenia. We screened mutant mice carrying hemizygous deletions of 22q11DS genes and identified haploinsufficiency of Mrpl40 (mitochondrial large ribosomal subunit protein 40) as a contributor to abnormal STP. Two-photon imaging of the genetically encoded fluorescent calcium indicator GCaMP6, expressed in presynaptic cytosol or mitochondria, showed that Mrpl40 haploinsufficiency deregulates STP via impaired calcium extrusion from the mitochondrial matrix through the mitochondrial permeability transition pore. This led to abnormally high cytosolic calcium transients in presynaptic terminals and deficient working memory but did not affect long-term spatial memory. Thus, we propose that mitochondrial calcium deregulation is a novel pathogenic mechanism of cognitive deficiencies in schizophrenia.

Introduction

Schizophrenia (SCZ) is a catastrophic disease that affects approximately 1% of the world’s population and is characterized by multiple symptoms that include cognitive abnormalities such as deficits in working memory, executive function, and learning¹. Mechanisms of cognitive symptoms of SCZ are poorly understood, partly because only weak associations have been identified between any single gene and the disease, and valid animal models have been lacking². Mouse models of 22q11 deletion syndrome (22q11DS) are among the few animal models that replicate abnormalities associated with SCZ. The 22q11DS is the most common multi-gene syndrome in humans and is considered a genetic risk factor for SCZ. The 22q11DS is caused by the hemizygous deletion of a 1.5- to 3-megabase region on the q arm of chromosome 22, resulting in the haploinsufficiency of multiple genes³. Approximately 30% of children with 22q11DS experience SCZ during late adolescence or early adulthood^{4, 5}. Symptoms of 22q11DS-related SCZ are indistinguishable from those of the idiopathic disease⁵, suggesting that the biological mechanisms involved in SCZ arising from the 22q11 deletion are similar to those involved in non–deletion-related SCZ.

The diagnosis of SCZ usually includes positive symptoms (i.e., disorderly thinking, hallucinations, and delusional ideas), negative symptoms (i.e., low levels of emotional arousal or social activity), and cognitive symptoms (i.e., deficits in attention, working memory, executive function, and learning and memory). Recognition of cognitive deficits as a core feature of SCZ and 22q11DS is increasing, as these deficits better predict disease progression than do the other symptoms^{6, 7}. Many cognitive symptoms of SCZ are thought to originate in the hippocampus^{8, 9}, a key brain region involved in learning and memory. Spatial working-memory deficits occur in patients with 22q11DS^{10, 11} and are also seen in 22q11DS mouse models. Mouse models of 22q11DS exhibit abnormal hippocampal short- and long-term synaptic plasticity^{12, 13}, which is consistent with the notion that synaptic plasticity is a cellular mechanism of learning and memory¹⁶. Short-term synaptic plasticity (STP) acting on the millisecond-to-minute time scale is believed to underlie reliable information transfer between hippocampal excitatory synapses in an activity-dependent manner^14–17, working memory¹⁸, and decision making¹⁹. STP predominantly occurs in presynaptic neurons²⁰. Several studies have shown that presynaptic abnormalities can be attributed to dysregulation of presynaptic calcium (Ca²⁺). For example, altered STP resulting from deregulated presynaptic Ca²⁺ are seen in models of neuropsychiatric diseases such as FMRP-related autism, Alzheimer disease, and 22q11DS^{12, 21, 22}. Because 22q11DS is a multi-gene deletion syndrome, more than one gene may affect STP. Initially, STP dysregulation in Df(16)1^+/− models of 22q11DS was linked to haploinsufficiency of microRNA-processing gene Dgcr8²³, which is mapped to a proximal part of the microdeletion. Depletion of microRNAs miR-185 and miR-25 leads to presynaptic Ca²⁺ dysregulation and abnormal STP through the abnormal elevation of (sacro)endoplasmic reticulum ATPase type 2 (Serca2), the Ca²⁺ pump that extrudes Ca²⁺ from the cytoplasm into the endoplasmic reticulum²³. SERCA2 is also elevated in the hippocampus of schizophrenic patients²³, and the most comprehensive genome-wide association study to date linked the ATP2A2 gene, which encodes SERCA2, with SCZ²⁴. Other genes that affect STP remain unknown.

Here we report results of our STP screening of the distal region of the 22q11DS microdeletion, which encompasses six genes: Cldn5, Cdc45l, Ufd1l, 2510002D24Rik, Mrpl40, and Hira. Using mutant mice carrying hemizygous deletions of individual genes, we discovered that haploinsufficiency of Mrpl40 (mitochondrial large ribosomal subunit protein 40, also known as Nlvcf) causes abnormal STP and short-term memory deficits via Serca2-independent deregulation of presynaptic Ca²⁺. Using two-photon Ca²⁺ imaging of the genetically encoded Ca²⁺ indicator GCaMP6²⁵, expressed either in the presynaptic cytosol or mitochondria, we showed that Mrpl40 haploinsufficiency hindered the extrusion of Ca²⁺ from the mitochondrial matrix through impaired mitochondrial permeability-transition pore (mPTP). This leads to abnormally high levels of Ca²⁺ in the presynaptic cytosol and elevated STP. Our data implicate Mrpl40 as a 22q11DS gene, the haploinsufficiency of which contributes to cognitive deficits in microdeletion-related SCZ.

Materials and Methods

Animals

Mature (16–20 weeks) mice of both sexes were used. Production and genotyping of Df(16)5^+/−, Dgcr8^+/−, Cldn5^+/−, and Hira^+/− mice were previously described^{23, 26–28}. To generate Cdc45/^+/− and Ufd1l^+/− mice, we obtained SIGTR embryonic stem (ES) cell clones containing gene-trap disruptions of the pGT01Lxr vector for Cdc45l (cell line AJ0425) and a Ufd1l allele (cell line AW0532) from the Mutant Mouse Regional Resource Center (University of California, Davis). The AJ0425 ES cell line carries a Genetrap insertion in exon 3 of the Cdc45l gene. Offspring were genotyped using the following primers: Cdc45lF: GCTGGGTACCTGAGTGTCATTG, Cdc45lR: CGAGACTGGTATGTGTGTGTGTG, and the betageo primer 2: ATTCAGGCTGCGCAACTGTTGGG, producing a 353-bp wild-type (WT) amplicon and a 309-bp mutant amplicon. The AW0532 ES cell line disrupts the Ufd1l gene in the first intron. This line was genotyped with the following primers: Ufd1lF: GTTGACGCTAACGTCCAGTCAC, Ufd1lR: GAAGCAGCGGTACTGCGTGGAG, and the Betageo primer: ATTCAGGCTGCGCAACTGTTGG, producing a 612-bp WT amplicon and a 304-bp mutant amplicon.

To generate the Mrpl40^+/− mice, we obtained sperm from the C57BL/6 strain carrying the Mrpl40^{tm1(KOMP)vlcg} allele of the Mrpl40 gene from the trans-NIH Knock-Out Mouse Project (www.komp.org). Sperm was used to produce Mrpl40^+/− offspring via in vitro fertilization. Mrpl40^+/− mice were genotyped using the following primers: Mrpl40F: CAGGCACACGTCAGACACA, Mrpl40R: GAGATCCCAGAAGGCCAGTAAG, and LacZR: CCCACCAGAGAGCTTC, producing a 102-bp WT amplicon and a 387-bp mutant amplicon. To expand the colony, we bred 2510002D24Rik^+/−, Mrpl40^+/−, Ufd1l^+/−, and Cdc45l^+/− mice harboring the disrupted alleles to C57BL/6J mice in the St. Jude Animal Resource Center.

All mouse strains in this study were back-crossed onto the C57BL/6J genetic background for at least 5 generations. Individual experiments were conducted in age-matched and sex-matched animals. The care and use of animals were reviewed and approved by the Institutional Animal Care and Use Committee at St. Jude Children’s Research Hospital.

Brain slice preparation

Mouse brains were quickly removed and placed in cold (4°C) dissecting artificial cerebrospinal fluid (ACSF) containing 125 mM choline-Cl, 2.5 mM KCl, 0.4 mM CaCl₂, 6 mM MgCl₂, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, and 20 mM glucose (295–300 mOsm), under 95% O₂/5% CO₂, and acute transverse hippocampal slices (400-µm thick) were prepared. After dissection, slices were incubated for 1 h in ACSF containing 124 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, and 10 mM glucose (285–295 mOsm), under 95% O₂/5% CO₂ at 32°C to 34°C and then transferred into the submerged recording chamber and superfused (1.5–2 mL/min) with ACSF. Long-term potentiation (LTP) experiments were performed as previously described¹². Because the bath temperature is crucial for short-term potentiation (STP) experiments²⁹, all electrophysiology and imaging experiments were conducted in a perfusion chamber maintained at near physiological temperature (33°C-34°C, measured at the center of the bath) by using both inline and bath heaters.

Whole-cell electrophysiology

Electrophysiological data were acquired using a Multiclamp 700B amplifier, filtered at 2 kHz and digitized at 10/20 kHz using a Digidata 1440 digitizer controlled by Clampex acquisition software. All offline analyses of electrophysiological data were done in Clampfit. To evoke excitatory postsynaptic currents (EPSCs) in CA1 neurons, we stimulated Schaffer collaterals with a concentric bipolar stimulating electrode (125-µm outer diameter; 12.5-µm inner diameter) connected to an Iso-Flex stimulus isolator with 100-µs pulses. Stimulating and recording electrodes were separated by at least 350 to 400 µm. For STP experiments, a cut was placed between the CA3 and CA1 regions to avoid recurrent stimulation. Recording electrodes were borosilicate glass capillaries pulled in a Sutter P-1000 puller and had resistances approximately 3.5 to 5 mΩ. CA1 neurons were whole cell voltage-clamped using electrodes filled with 130 mM K gluconate, 10 mM KCl, 0.5 mM EGTA, 2 mM MgCl2, 5 mM NaCl, 2 mM ATP-Na₂, 0.4 mM GTP-Na, 10 mM HEPES, and 10 to 25 µM Alexa 594 (pH 7.35, ~ 290 mOsm). EPSCs were recorded by holding the cells at –74 mV (accounting for a liquid junction potential of ~14 mV). Access resistance was monitored using a –5 mV step before each recording (range, 25–50 mΩ), and cells that displayed unstable access resistance (variations >20%) were excluded from the analyses. All EPSC recordings for STP experiments were done in the presence of the GABA_AR antagonist picrotoxin (100 µM) to prevent inhibitory responses and the NMDA receptor antagonist D-2-amino-5-phosphonovalerate (D-AP5, 50 µM) to avoid induction of long-term synaptic plasticity.

STP components were separated using an adapted procedure²⁹. A baseline of stable EPSCs was established to evoke 150- to 200-pA EPSCs. Stimulation intensities ranged from 30 to 60 µA and were held constant for each cell throughout the experiment. Stimulation intensity did not differ significantly between slices from WT and mutant littermates (P > 0.05). The stimulation protocol consisted of the following: 5 pre-train pulses (0.2 Hz), a high-frequency train (100 pulses, 80-Hz), 24 pulses at 5 Hz (to measure recovery from depression), and 24 pulses at 0.2 Hz (to measure augmentation). The 80-Hz (100 pulses) train was chosen because it induced both strong depression and augmentation on the basis of STP induced by trains of different frequencies (data not shown). EPSCs during and after the train were normalized to the average of the 5 pre-train EPSCs (shown as a single baseline point), and the amplitudes are reported as normalized EPSC peak amplitudes. For each cell, we averaged the data from 3 to 4 trials by using this stimulation protocol. To measure the EPSC amplitudes during the train, we resorted to a binning procedure, instead of the template waveform-based subtraction protocol, as previously described²⁹. The measured EPSC train of 1,250 ms (80 Hz × 100 pulses) was divided into 100 equal bins of 12.5 ms, and the peak amplitude within each bin was graphed as the normalized EPSC amplitude during the train. This procedure was sufficient to separate EPSCs in the train.

Augmentation, the slowest component of STP, has negligible contamination from depression and facilitation and is reported as measured without corrections. Statistical comparisons of augmentation were made using the measured peak augmentation at 5 s after the 80-Hz train, instead of the extrapolated augmentation at 0 s. Recovery from depression (200 to 4,800 ms) after the 80-Hz train was contaminated by augmentation and some residual facilitation. Because the EPSCs measured during the recovery phase had an interstimulus interval of 200 ms, the contamination by facilitation was minimal and corrected only for overlying augmentation. To that end, the measured augmentation decay curve (5–120 s after the 80-Hz train) was extrapolated from 200 to 4,800 ms after the train by using a standard exponential fit in Clampfit. The normalized EPSCs during the recovery phase were then corrected for the overlying extrapolated augmentation for each time point. Facilitation was measured in separate experiments by using paired pulses at interstimulus intervals of 20 to 1,000 ms.

To measure excitability, we held CA3 neurons in current clamp mode and added D-AP5 (50 µM) and picrotoxin (100 µM) to the external ACSF bath to avoid possible long-term effects. Action potential (AP) parameters were estimated by clamping the cell at –65 mV using automatic slow-current injection. Input resistance, AP threshold, rheobase (intensity of current reached at threshold), number of APs (evoked by holding at the minimum current needed to reach threshold) were measured by injecting 1-s steps of 25-pA current (12 steps from –25 pA to 250 pA). To measure AP widths, we evoked APs by injecting 5 short current steps (1,500–2,000 pA, 1-ms duration, 0.2 Hz). AP durations were calculated as the time interval between the up-stroke and down-stroke of the AP waveform at –10 mV. AP duration during the 80-Hz train was normalized to the AP duration at baseline, as previously described²¹. Electrophysiology experiments were done without blinding.

Generation of plasmids and viruses

To generate adeno-associated viruses (AAVs) expressing cytoplasmic GCaMP6f (GCaMP6) and mitochondrial-targeted GCaMP6f (mitoGCaMP6), we used PCR to amplify the human synapsin promoter (hSyn) from pAAV-6P-SEWB³⁰. The pAAV-GFP (Addgene plasmid 32395) was cut with SnaB1 and Sac1 to replace CMV with hsyn (pAAV-hsyn-GFP). EcoRI and BamHI were used to replace GFP with mCherry to generate pAAV-hsyn-mCherry. To generate pAAV-hSyn-mCherry-2A-GCaMP6, oligonucleotides containing the coding sequence for the 2A peptide were used for PCR amplification of GCaMP6 from pGP-CMV-GCaMP6 (Addgene plasmid 40755). To generate pAAV-hSyn-mCherry-2A-mitoGCaMP6, oligonucleotides containing the coding sequence for the 2A peptide and the mitochondrial-targeting sequence (MSVLTPLLLRGLTGSARRLPVPRAKIHSL) were used for PCR amplification of GCaMP6 from pGP-CMV-GCaMP6 (Addgene plasmid 40755). To generate AAV-Slc25a4 OE, we used PCR to amplify the open reading frame of Slc25a4. Drd2 open reading frame from AAV-CamKII-Drd2 OE³¹ was replaced with the Slc25a4 open reading frame using HindIII. DNA sequencing was used to verify the absence of PCR-induced mutations. Lentivirus vector siRNA plasmids (control shRNA, 5′-TACGTCCAAGGTCGGGCAGGAAGA-3′; Slc25a4 shRNA1, 5′- GCAAGGGATCTTCCCAGCGAGAATTCAAT-3′; Slc25a4 shRNA2, 5′-CGTTTGACACTGTTCGTCGTAGGATGATG-3′; Slc25a4 shRNA3, 5′- GCACATTATCGTGAGCTGGATGATTGCCC-3′) were generated by Applied Biological Materials (Richmond, BC, Canada). Viruses (1.8 × 10⁸ to 1 × 10⁹ particles/ml) were produced by either the St. Jude or University of Tennessee Health Sciences Center Viral Vector Cores.

Two-photon imaging of presynaptic calcium and in vivo injections

Two-photon laser-scanning microscopy was performed using an Ultima imaging system (Prairie Technologies, Middleton, WI), a Ti:sapphire Chameleon Ultra femtosecond-pulsed laser (Coherent Inc., Santa Clara, CA), and 60× (0.9 NA) water-immersion IR objectives (Olympus, Center Valley, PA). Calcium (Ca²⁺) transients in presynaptic terminals were recorded using GCaMP6 expressed in the CA3 hippocampal neurons.

To express GCaMP6, mice were anaesthetized using isoflurane (2% for induction and 1.5% for maintenance) in 100% oxygen, and their heads were restrained on a stereotaxic apparatus. An approximately 1-cm midline incision was made centered about 0.25 cm behind bregma. Viruses were injected into 3 locations within the CA3 region, in 1 or both hemispheres. The stereotaxic coordinates for the 3 injections were as follows, in relation to the bregma: (1) –1.5-mm anteroposterior, 1.8-mm lateral, and 1.7-mm deep; (2) 2.2-mm anteroposterior, 2.3-mm lateral, and 1.8-mm deep; (3) 2.5-mm anteroposterior, 2.8-mm lateral, and 2.2-mm deep. Craniotomy holes were drilled at these locations, and 200 nL of AAVs was slowly (20 nL/min) injected via a 33G cannula. After each injection, the cannula was left in place for 2 to 3 min before being retracted. Following injections, the skin was sutured, and the mice were allowed to recover before returning to the holding cages. Imaging experiments were performed 4 to 7 weeks after AAV injections. During each experiment, care was taken to limit the differences in post-injection durations to a maximum of 2 to 3 days across experimental groups to avoid substantial differences in the levels of AAV expression.

To visualize GCaMP6 or mitoGCaMP6, we used brain slices prepared from AAV-injected mice. Schaffer collaterals were stimulated via field stimulation using bipolar stimulating electrodes, as in STP experiments. In addition to D-AP5 and picrotoxin, the AMPA receptor antagonist NBQX (3 µM) was added to the bath ACSF to prevent excitation of postsynaptic neurons. GCaMP6 was visualized at 940 nm, and mCherry was visualized at 1,040 nm by two Ti:Si lasers. Presynaptic boutons were identified in a 34 µm × 34 µm region of interest by activity-dependent increase in GCaMP6 fluorescence during time-series scans. Line-scans through identified boutons were then used for experiments. Boutons from 4 to 8 regions of interests in the stratum radiatum of the CA1 area were imaged for each mouse. In some experiments, we used whole-cell recordings from CA3 neurons and filled the cells with Alexa Fluor 594 (30 µM) and Fluo 5F (300 µM) at 820 nm to visualize presynaptic Ca²⁺. Axons emanating from the cell bodies were identified based on their morphology and lack of spines. Presynaptic terminals were identified as boutons in secondary and tertiary axonal branches. Those axons could not be tracked beyond 200 µm from the CA3 cell bodies due to the limitation of the approach. APs were evoked by holding the cells at –70 mV (current clamp mode) and injecting depolarizing current (3.5–4.0 nA, 500 µs). We recorded Ca²⁺ transients in line-scan mode from 3 to 7 boutons per cell. For GCaMP6 or mitoGCaMP6 experiments, the baseline fluorescence (F₀) was used to calculate the change in signal (ΔF/F₀). For Fluo5F experiments, fluorescence changes (ΔG/R) were quantified as an increase in Fluo 5F fluorescence (ΔG) normalized to the respective Alexa 594 fluorescence (R). Imaging experiments were done without blinding.

Two-photon glutamate uncaging

For two-photon glutamate uncaging (TGU), MNI-glutamate (2.5 mM) was added to the recording ACSF. The timing and intensity of glutamate uncaging were controlled by TriggerSync (Prairie Technologies). In typical experiments, 0.2-ms pulses from a second Ti:sapphire Chameleon Ultra laser (720 nm) were delivered to the vicinity of a targeted dendritic spine, and TGU-evoked EPSCs (uEPSCs) were recorded. The duration and intensity of illumination of the uncaging laser were then adjusted to induce responses that mimicked spontaneous miniature EPSCs, which were recorded in CA1 neurons and averaged 10 to 15 pA. After the uncaging parameters (i.e., site, laser duration, and laser intensity) were adjusted for a single spine, the parameters remained constant for the STP experiments on the particular spine. An 80-Hz train of 100 TGU pulses was delivered to a single dendritic spine to measure TGU-induced STP.

Electron microscopy

Mice were anesthetized with ethyl carbamate (1.5 g/kg, 25% solution, intraperitoneal) and perfused transcardially with phosphate-buffered saline (PBS) for 1 to 2 min and then a fixative (2.5% gluteraldehyde and 2% paraformaldehyde in 0.2 M sodium cacodylate). Brains were isolated, stored at 4°C overnight in the same fixative, and sagittal sections (100-µm thick) were prepared on a Leica vibratome. Smaller regions (~ 500 µm × 500 µm) containing the stratum radiatum of the CA1 hippocampal region were processed for 3-dimensional (3D) scanning electron microscopy (SEM). The samples were stained with a modified heavy-metal–staining method, processed through a graded series of alcohol and propylene oxide, and then embedded in Epon hard resin³². Sections (0.5-µm thick) were cut to determine the correct area and then coated with iridium in a Denton Desk II sputter coater. The 3D EM images were collected on a Helios Nanolab 660 Dualbeam system. From the 3D stacks of electron micrographs (10 × 10 x 10 nm voxel size, 250 to 260 sections of 10 nm thickness and approximately 30 × 20 µm area, synapses were identified based on the presence of postsynaptic densities and presynaptic vesicles. The mitochondria in presynaptic terminals were identified and counted manually.

For transmission electron microscopy (TEM), 100-µm-thick vibratome sections containing the CA1 stratum radiatum region were prepared as described above and fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). Sections were post fixed in 2% osmium tetroxide in 0.1 M sodium cacodylate buffer with 0.3% potassium ferrocyanide for 1.5 h. After rinsing in the same buffer, the sections were dehydrated through a series of graded ethanol and propylene oxide solutions, infiltrated and embedded in epoxy resin, and polymerized at 70°C overnight. Thin sections (0.5 µm) were stained with toluidine blue for examination by light microscopy. Ultrathin sections (80 nm) were cut and imaged using an FEI Tecnai F 20 FEG Electron Microscope with A AT XR41 Camera.

Mitochondrial DNA quantification

RNA was isolated from the hippocampus of 4-month-old mice by using the mirVana RNA isolation kit (Ambion, Life Technologies) The SuperScript III reverse transcriptase kit (Invitrogen, Life Technologies) was used to synthesize cDNA from 1 µg RNA. The following primers were used in the qPCR experiments: COI (5'-GCCCCAGATATAGCATTCCC-3'and 5'-GTTCATCCTGTTCCTGCTCC-3'); ND2 (5'-CCCATTCCACTTCTGATTACC-3' and 5'-ATGATAGTAGAGTTGAGTAGCG-3'); 18S (5'-TAGAGGGACAAGTGGCGTTC-3' and 5'-CGCTGAGCCAGTCAGTGT-3'). The ratio of mitochondrial (CO1 and ND2) transcripts to the nuclear 18S transcript was used to quantify relative mitochondrial DNA.

Mitochondrial respiration

Mice were decapitated, and their brains were dissected immediately in cold (4°C) dissecting ACSF. Hippocampi were removed and washed in a mitochondrial isolation buffer (MIB) containing 250 mM sucrose, 1 mg/mL bovine serum albumin, 5 mM EDTA, and 10 mM Tris-HCl at pH 7.4. A crude mitochondrial fraction was then isolated by differential centrifugation. Briefly, the hippocampal tissue was finely minced on a pre-chilled glass dish, washed several times, resuspended in 1 mL MIB and homogenized in a 2-mL Dounce homogenizer (glass/glass, 5–7 runs in an ice bath). The homogenized tissue was centrifuged at 700 ×g for 5 min; the supernatant was then transferred to a new 1.5-mL tube and centrifuged at 7,000 ×g for 10 min. The resulting pellet was washed in 1 mL MIB and resuspended in 30 µL MIB. The protein concentration in the pellet was typically 25 to 35 mg protein/mL. The mitochondrial integrity was tested by measuring the glutamate/malate-dependent respiratory control ratio (i.e., State III/State IV of respiration). The resulting mitochondrial samples were used immediately to measure respiration (mitochondrial oxygen consumption) and Ca²⁺-retention capacity.

Mitochondrial oxygen consumption was measured using a Clark-type electrode (Hansatech Instruments, Ltd, Norfolk, U.K.) in a thermostatically controlled chamber equipped with a magnetic stirring device and a gas-tight stopper fitted with a narrow port for additions via a Hamilton microsyringe. Isolated mitochondria were placed in the respiration chamber at 37°C in 0.4 mL respiration buffer (250 mM sucrose, 1 g/L bovine serum albumin, 10 mM KH₂PO₄, 2.7 mM KCl, 3 mM MgCl₂, 40 mM Hepes, 0.5 mM EGTA, pH 7.1) to yield a final concentration of 0.5 mg/mL. State-III respiration was stimulated by the addition of 2 mM ADP. Respiratory control ratios were obtained by dividing the rate of oxygen consumption in the presence of ADP (state III) by that in the absence of ADP (state IV). The measurement protocol involved sequential addition (with 5-min intervals) of 5 mM glutamate, 2.5 mM malate (activating the CI-CIII-CIV span), 1 µM rotenone, 10 mM succinate (activating the CII-CIII-CIV span), 1 µM antymicin A, 10 mM ascorbate, 0.4 mM TMPD (activating the CIV span), and 5 mM KCN. The rates of oxygen consumption were calculated online, as first derivatives of the dioxygen-content changes by manufacturer-provided software.

Quantitative real-time PCR

The cDNA was generated from 1 µg hippocampal RNA by using Superscript III (Life Technologies, Waltham, MA). Primers for qPCR were as follows:

• Hira: TCCGCCATCCATCAATTC and CTATCCTTCACCAGCCTAG,

• Cldn5: CGCAGACGACTTGGAAGG and GCCAGCACAGATTCATACAC,

• Mrpl40: CTGGTAGTTAGAGATAGGTGGTG and GAGGAGCTGAAACTTGAATCTG,

• Ufd1l: TCAAGCATGTATTCATTCTGC and TTTATTTACAGTGACTCAGAAGG,

• 2510002D24Rik: GTGTTCCAGGTCAAGTAA and AGAAGGACAAGTGATAAGC,

• Cdc45l: GATTTCCGCAAGGAGTTCTACG and TACTGGACGTGGTCACACTGA.

• Co1: GCCCCAGATATAGCATTCCC and GTTCATCCTGTTCCTGCTCC,

• Nd2: CCCATTCCACTTCTGATTACC and ATGATAGTAGAGTTGAGTAGCG,

• 18s: TAGAGGGACAAGTGGCGTTC and CGCTGAGCCAGTCAGTGT.

We performed qPCR using SYBR green in an Applied Biosystems 7900HT Fast Real-time PCR system and the standard protocol. A serial dilution of cDNA was used to generate a standard curve for each primer set, and this curve was used to calculate gene concentrations for each sample. All samples were run in triplicate.

Mouse behavior

Mature animals (16- to 20-weeks) were used for all behavior experiments.

Morris water maze

One hour prior to testing, animals were brought into the testing room and allowed to habituate. Testing was performed during the animal’s inactive phase under dim-light conditions. Mice were allowed to navigate in the maze, and swim patterns were recorded with a video camera tracking system (HVS Image, Co., Buckingham, UK) mounted above the pool. Animals learned to find a hidden, clear platform by using the standard spatial version of the Morris water maze task for 4 successive days. Each day, animals were given four 1-min trials from each starting position with an inter-trial latency of 60 s. The order of the starting locations was counterbalanced each day by using a Latin-square design. A spatial learning (probe) trial was administered 1 h after the completion of spatial training. A spatial memory (probe) trial was administered 48 h after completion of the spatial learning. During both probe trials, the platform was removed, and the mice received a single 1-min trial in which the animal tried to find the escape platform. These trials originated from the starting location that was the farthest from the platform’s location throughout training. Mice also completed a nonspatial learning task at least 7 days after completion of the spatial protocol. In that task, mice were trained to find a black visible platform for 2 successive days. During Day 1, the escape platform was located in the same position used during spatial training. The next day, the escape platform was moved to a new quadrant. Each day, the mouse was given four 1-min trials in the same manner that occurred during spatial training. To avoid hypothermia, immediately after each round of training and testing trials, animals were dried with paper towels and placed in warmed holding cages.

Delayed non–matched-to-position task

To motivate mice to complete the delayed non–matched-to-position task, they were subjected to water restriction for 2 days prior to testing. Specifically, mice were allowed 2 h of free access to water per day. Mice were weighed daily to ensure that weight decrease during deprivation did not surpass the recommended 20% loss. One hour prior to testing, animals were brought into the testing room and allowed to habituate. Testing was performed during the animal’s inactive phase under well-lit lighting conditions. The testing apparatus consisted of a Y-maze (Cleversys Inc., Reston, VA) with a start arm (20 cm × 16 cm × 7 cm; l × w × h) leading to 2 goal arms (20 cm × 16 cm × 7 cm). Mice were allowed to habituate to the maze and were given a positive reinforcer (i.e., Chocolate Yoohoo) before behavioral testing. To achieve this, mice were allowed to investigate the Y-maze baited with Yoohoo for 15 min. Maze habituation was performed for 2 consecutive days.

After maze habituation was complete and the animal had consumed the food rewards, we conducted a test of spatial working memory. First, the mouse was constrained in the start arm with a guillotine door. Next, a sample arm was determined at random and the choice arm was closed off with a guillotine door (there was a limit of 2 consecutive same-side sample arms in a 10-trial test). Both arms were then baited with 20 µL Yoohoo. The mouse was released from the start chamber and allowed to run to the sample arm and consume the reward. The mouse was then returned to the start arm and the guillotine door to the nonsample arm was removed. All efforts were made to keep the intratrial interval at 5 s. The mouse was again released from the start chamber and allowed to run to either the previously entered sample arm or new choice arm to consume the reward. A return to the sample arm was counted as an incorrect response. Incorrect responses resulted in no reward and return to the start arm. A total of 10 trials were given.

The Y-maze spatial recognition

The maze was shaped like a “Y”, with 3 equally spaced arms (20 cm × 7 cm × 16 cm) radiating from a triangular center section. The Y-maze was constructed from blue opaque plastic to aid in video detection. The maze was located in a lit room with abundant extra-maze cues. The procedure consisted of an acquisition and a recognition session. During the acquisition session, the mouse was placed facing the distal end of an arm (start arm; determined semirandomly) and allowed to freely explore the maze for 15 min. During acquisition, the mouse was allowed to freely explore 2 of the 3 arms (determined semirandomly). After completing the acquisition session, the mouse was returned to its home cage for 1 h. Next, the mouse was given a 2-min recognition session, where all 3 arms were available for the mouse to freely explore. Time spent in each arm and the triangular center section and the total distance traveled during each trial were automatically recorded using TopScan software (CleverSys Inc.). Mice prefer novelty; therefore, if a mouse recognized and remembered which arms were familiar during the recognition session, the mouse would spend more time in the novel arm (expressed as a percent of total arm time) than would be expected by chance.

Acoustic startle and prepulse inhibition

Each day before testing, the mice were allowed a 1-h habituation in the testing room after being transported from the animal housing room. Before experiments were initiated, the mice were allowed to acclimate to the Plexiglas restraint chamber (6 cm × 6 cm × 4.8 cm) for 20 min. Acoustic-startle and prepulse inhibition (PPI) tests were performed in ventilated, sound-attenuated chambers (Med Associates, St. Albans City, VT). For acoustic-startle experiments, the mice had a 5-min acclimation period to a 65-dB background white noise, which played throughout the session. Three startle pulses (8 kHz, 120 dB, 40 ms) were then delivered at 15-s intertrial intervals.

For PPI experiments (conducted on different days than acoustic-startle experiments), mice had a 5-min acclimation period to a 65-dB background white noise, which played throughout the session. Three acoustic startles (broadband white noise click, 120 dB, 40 ms) were then delivered separated by a 15-s intertrial interval. The testing session consisted of 39 trials of 5 trial types: pulse alone, in which the startle pulse was presented; the combination of a 40-ms prepulse of 74 dB, 82 dB, or 90 dB preceding the startle pulse by 100 ms; and no stimuli. Trials were separated by 15 s and presented in a pseudo-random order. PPI was calculated as follows: 100 × (pulse-alone response – prepulse + pulse response)/pulse-alone response. All mouse behavior experiments were performed in a blind manner in respect to mouse genotypes.

Western blotting

AAV5-hSyn-mCherry-2A-mitoGCaMP6f or AAV5-hSyn-mCherry-2A-GCaMP6f viruses were injected in vivo into the mouse hippocampus as described above. Four weeks after injections, mice were euthanized and dorsal hippocampi were isolated at 4°C for fractionation into nuclear/cell debris, cytoplasm, and crude mitochondrial fractions. Freshly extracted hippocampi from each mouse were homogenized in isolation buffer (250 mM sucrose, 75 mM mannitol, 1 mM EGTA, and 5 mM Hepes, at pH 7.4) by using a glass-teflon tissue homogenizer on ice. Homogenized hippocampi were centrifuged twice at 1,400 ×g for 3 min at 4°C to separate nuclei and cell debris. The supernatant was then centrifuged at 17,200 ×g for 10 min at 4°C to separate cytosol and crude mitochondria. Ice-cold RIPA buffer (Santa Cruz Biotechnology, Dallas TX) with protease inhibitors (Roche, Basel, Switzerland) was added to the mitochondria but not the cytoplasmic fraction, and both were briefly sonicated. Samples were centrifuged at 17,200 ×g for 20 min at 4°C, and the pellet was kept at –80°C for protein quantification. Protein quantification was performed using a Pierce BCA Protein Assay Kit (Thermo Scientific). For loading-sample preparation, we used NuPAGE LDS sample buffer (Life Technologies).

The Western-blot experiments were performed similar to a previously described method¹². Briefly, NuPAGE 10% Bis-Tris gels and MES-SDS running buffer (Life Technologies) were used to load mitochondria and cytoplasmic fractions (10 µg total protein/well) and run the gels. The primary antibody used to detect GCaMP6 was rabbit–anti-GFP (1:1000, Abcam, 6556). We also used rabbit–anti-Prohibitin 1 (1:1000, Thermo Scientific, PA5-12274) and mouse–anti–β-actin (1:10,000, Sigma, 5316) primary antibodies. Secondary antibodies were Odyssey goat–anti-rabbit IRDye-680LT (1:40,000, LI-COR Biosciences, Lincoln, NE, 926-68021) or Odyssey donkey–anti-mouse IRDye-800CW (1:15,000, LI-COR Biosciences, 926-32212). Membranes were imaged using an Odyssey Infrared Imager (LI-COR Biosciences). Images were analyzed using Odyssey V3.0 software.

Subcellular localization of mitoGCaMP6 and GCaMP6

Neuro-2a (N2a) mouse neuroblastoma cells (ATCC, CCL-131) were plated in 4-well chamber slides (Thermo-Scientific Lab-Tek 177399) at 1.25 × 10⁵ cells/well and maintained in culture using Eagle’s Medium Essential Media (EMEM) (ATCC) plus 10% fetal bovine serum (heat inactivated, Life Technologies) and 1× PenStrep (Life Technologies) in an incubator at 37°C and 95% O₂/5% CO₂. Following a 24-h incubation, cells were transfected with hSyn-mCherry-2A-mitoGCaMP6f or hSyn-mCherry-2A-GCaMP6f plasmids (0.5 µg DNA/well) using Fugene HD transfection reagent (Promega) (Ration Fugene HD/DNA 1:4) in serum-free EMEM. Twenty-four hours later, cells were supplemented with additional fresh complete media, and 48 h later, they were incubated in complete media with Mitotracker DeepRed (1:2,000, Life Technologies) for 15 min at 37°C. Next, Mitotracker was washed out and replaced with fresh complete media, and cells were placed back to the incubator for 15 to 20 min. Cells were then washed with sterile PBS and fixed in 4% paraformaldehyde for 15 min at room temperature. Following fixation, cells were washed 3 times for 5 min with PBS and blocked with 10% normal goat serum and 0.1% TritonX-100 in PBS for 1 h at room temperature. We used the following antibodies: chicken–anti-GFP to detect GCaMP6 or mitoGCaMP6 (1:1000, Abcam, 13970) and goat–anti-chicken–Alexa-488 (1:1000, Life Technologies, A11039). Cell imaging was performed using an LSM-780 confocal microscope (Zeiss).

Other drugs and chemicals

Bongkrekic acid solution was purchased from Sigma-Aldrich. D-AP5, picrotoxin, tetrodotoxin, NBQX, CGP 37157, MNI glutamate were from Tocris. Stock solutions of these drugs were prepared in manufacturer recommended solvents and stored at −20°C. Ru360 was from Calbiochem.

Statistical analyses

Data are presented as the mean ± SEM. Statistical analyses were performed using SigmaPlot software. Parametric or nonparametric tests were chosen based on the normality and variance of data distribution. Independent or paired two-tailed t-tests (t value), Mann-Whitney Rank Sum test (U value), one-way analysis of variance (ANOVA) / Kruskal Wallis one-way analysis of variance on ranks followed by a multiple-comparison procedure (Dunn’s method), two-way ANOVA / two way repeated measures ANOVA with one factor repetition followed by Holm-Sidak multiple comparison procedure were the statistical tests used. F values were reported for ANOVA and Q values from multiple comparison procedure were reported for ANOVA on ranks. Differences with p < 0.05 were considered significant.

Results

Abnormal presynaptic augmentation underlies aberrant STP in Df(16)5^+/− mice

To determine whether haploinsufficiency of distal genes in the 22q11DS region contributes to hippocampal pathophysiology, we used Df(16)5^+/− mice³³ carrying a hemizygous deletion of six genes: Cldn5, Cdc45l, Ufd1l, 2510002D24Rik, Mrpl40, and Hira (Fig. 1a). Similar to the late onset of SCZ symptoms, synaptic plasticity abnormalities in mouse models of 22q11DS do not appear until later in life²³; thus, we used 4- to 5-month-old mice for these experiments. Df(16)5^+/− mice developed normally and had no visible gross morphologic abnormalities (data not shown).

Figure 1. Abnormal synaptic plasticity in Df(16)5^+/− mice is caused by Mrpl40 haploinsufficiency.

(a) Diagram depicting genes in the 22.q11.2 genomic region of the human chromosome 22 and the syntenic region of mouse chromosome 16. Red horizontal bar represents genomic regions hemizygously deleted in Df(16)5^+/− mice and grey horizontal bar represents genomic regions hemizygously deleted in Df(16)1^+/− mice. Note that 2510002D24Rik, Mrpl40, and Hira genes are mapped outside the Df(16)1 microdeletion. (b) Input–output relations in Df(16)5^+/− and WT littermates. (c) STP (comprising facilitation, depression, and augmentation) induced by the high-frequency (80-Hz) train. The first time point represents an average of 5 baseline EPSCs delivered at low frequency. The top inset shows the protocol for measuring STP, recovery from depression, and augmentation in the same experiment. (d-f) Average facilitation tested by paired-pulse ratio in separate experiments (d), recovery from depression tested 5 s after the 80-Hz train (e), and augmentation tested 5 to 120 s after the 80-Hz train in Df(16)5^+/− and WT mice (f). Insets show representative EPSC traces. (g,h) Mean STP of EPSCs induced by the 80-Hz train of synaptic stimulation of Schaffer collaterals (g) and augmentation (h) in Mrpl40^+/− mice and their WT littermates. Numbers of neurons are shown in parentheses or inside columns. Data are represented as mean ± SEM. *p <0.05.

Using the whole-cell voltage-clamp technique we recorded excitatory postsynaptic currents (EPSCs) at glutamatergic synapses between CA3 and CA1 pyramidal neurons in the hippocampus (CA3–CA1 synapses) by electrically stimulating Schaffer collaterals. Basal synaptic transmission measured as the input–output relation between stimulation intensity and EPSCs in Df(16)5^+/− mice was comparable to wild-type (WT) littermates [F_(1,12) = 1.435, p = 0.231] (Fig. 1b). EPSC kinetics, such as rise time [t₍₁₉₎ = 0.442, p = 0.663], half-width (U = 51, p = 0.805), and decay time [t₍₁₉₎ = 0.0377, p = 0.970] were also normal between mutants and controls (Supplementary Fig. 1). However, an 80-Hz train (100 stimuli) applied to Schaffer collaterals evoked a substantially larger STP [t₍₁₆₎ = 2.196; p = 0.010] in Df(16)5^+/− mutants than in WT mice (Fig. 1c). Because STP is not a unitary process but rather consists of several temporally and mechanistically distinct components (e.g., facilitation, augmentation, and depression of presynaptic transmission²⁰), we measured STP components individually.

The STP increase in Df(16)5^+/− mutants could arise from the increased facilitation/augmentation, decreased short-term depression, or a combination thereof²¹. To differentiate among these possibilities, we used an established protocol to separate STP components²⁹ and assess their individual contributions to the STP increase in Df(16)5^+/− mice. First, using the paired-pulse ratios of two consecutive EPSCs, we found no significant difference [F_(1,5) = 2.142, p = 0.146] in facilitation between Df(16)5^+/− and WT mice (Fig. 1d). We then examined whether the increased STP in Df(16)5^+/− mice resulted from reduced short-term depression. We examined the recovery from depression by using a 5-Hz stimulus train applied during the first 5 s after the 80-Hz train of Schaffer collateral stimuli. Because recovery from depression overlaps with the decay of augmentation during those 5 s, the actual EPSC peak amplitude reflects the net effect of the two processes²¹. To isolate the depression component, we corrected the synaptic responses for contribution from augmentation^{21, 29} (see Online Methods). This analysis revealed no significant difference [F_(1,23) = 1.947; p = 0.182] in recovery from short-term depression between Df(16)5^+/− and WT littermates (Fig. 1e).

Next, we assessed the role of augmentation, which is the longest-lasting component of STP and operates on a time scale of tens of seconds²⁹. Using a previously reported approach²⁹, we isolated augmentation by applying a single stimulus to Schaffer collaterals every 5 s for 2 min, starting 5 s after the 80-Hz train (Fig. 1f). Augmentation was significantly increased [t₍₁₆₎ = 4.758, p = 0.0002] in Df(16)5^+/− mice compared to WT controls (Fig. 1f). This increased augmentation in Df(16)5^+/− mice was maximal at the onset and decayed to normal values after 20 s. Elevated augmentation in Df(16)5^+/− mutants was not sensitive to the Serca inhibitor thapsigargin (4 µM) [without thapsigargin: t₍₁₆₎ = 4.404, p < 0.001; with thapsigargin: t₍₁₀₎ = 1.954, p = 0.0396] (Supplementary Fig. 2a). Further, Serca2 protein levels were normal (U = 107, p = 0.836) in Df(16)5^+/− mice (Supplementary Fig. 2b), suggesting that haploinsufficiency of genes within the Df(16)5 genomic region resulted in abnormal STP through Serca2-independent mechanisms. To ensure that the STP increase in Df(16)5^+/− mice originated from the presynaptic CA3 neurons, we examined the role of the postsynaptic component by performing two-photon glutamate uncaging (TGU) (100 TGU pulses, 80 Hz) to activate individual dendritic spines on CA1 neurons, the postsynaptic sites of CA3 inputs. Because TGU focally releases exogenous glutamate from inactive (caged) glutamate (MNI-glutamate), thereby bypassing the release of endogenous neurotransmitter from CA3 terminals, this method tests only postsynaptic mechanisms of synaptic transmission and plasticity. TGU-induced STP did not differ [t₍₂₅₎ = −0.405, p = 0.689] between Df(16)5^+/− and WT littermates (Supplementary Fig. 3a), suggesting that the causative locus of the abnormal STP increase in Df(16)5 mutants resides in the presynaptic CA3 neurons. CA3 pyramidal neuron recordings revealed that presynaptic neurons had normal resting membrane potential [t₍₁₈₎ = −1.621, p = 0.122], input resistance [t₍₁₈₎ = 0.978, p = 0.341], and depolarization-induced action potentials (APs) [rheobase: t₍₁₈₎ = 0.451, p = 0.658 ; AP width: t₍₁₈₎ = 0.384, p = 0.706) (Supplementary Fig. 3b). An 80-Hz train of depolarization pulses delivered to the CA3 soma progressively increased the AP width, but to the same extent [100th AP width: t₍₁₂₎ = 0.845, p = 0.414] in both Df(16)5^+/− and WT littermates (Supplementary Fig. 3c, d), thereby ruling out CA3 excitability as the culprit mechanism and implicating abnormal presynaptic glutamate release as a possible underlying mechanism of elevated STP in Df(16)5^+/− mice.

Mrpl40 haploinsufficiency causes abnormal STP in Df(16)5^+/− mice

To identify the culprit gene(s) whose haploinsufficiency causes abnormal augmentation and STP in Df(16)5^+/− mice, we tested STP parameters in Cldn5^+/− and Hira^+/− mice^{27, 28} and the following new mutants that we generated: Cdc45l^+/−, Ufd1l^+/−, 2510002D24Rik^+/−, and Mrpl40^+/− mice (Supplementary Fig. 4). All transcripts were reduced by approximately 50% in the Df(16)5^+/− mice [Cldn5^+/−: p < 0.001; Cdc45l^+/−: p = 0.03; Ufd1l^+/−: p = 0.004; 2510002D24Rik^+/−: p = 0.005; Mrpl40^+/− : p = 0.010; Hira^+/−: p = 0.003] and the respective individual mutants [Cldn5^+/−: p < 0.001; Cdc45l^+/: p < 0.001; Ufd1l^+/−: p < 0.001; 2510002D24Rik^+/−: p = 0.004; Mrpl40^+/− : p < 0.001; Hira^+/−: p = 0.03] (Supplementary Fig. 4).

Testing STP in all six mouse mutants revealed that Mrpl40^+/− mice had elevated STP and augmentation compared to their WT littermates [STP: t₍₂₉₎ = 2.368, p = 0.025; augmentation: t₍₂₉₎ = 3.150, p = 0.0037], whereas Cldn5^+/−, Cdc45l^+/−, Ufd1l^+/−, 2510002D24Rik^+/−, and Hira^+/− mutants had normal STP and augmentation [STP and augmentation, respectively: Cldn5^+/−: t₍₁₂₎ = 0.349, p = 0.733 and t₍₁₂₎ = −1.205, p = 0.250; Cdc45l^+/−: t₍₁₈₎ = 0.566, p = 0.579 and t₍₁₈₎ = −0.637, p = 0.532; Ufd1l^+/−: U = 43, p = 0.903 and U = 35, p = 0.438; 2510002D24Rik^+/−: t₍₃₅₎ = 0.789, p = 0.436 and t₍₃₅₎ =1.773, p = 0.085; Hira^+/−: t₍₁₁₎ = −0.644, p = 0.532 and t₍₁₁₎ = −0.687, p = 0.507] (Fig. 1g, h, Supplementary Fig. 5). Mrpl40 haploinsufficiency did not affect mRNA expression of other genes in the Df(16)5 microdeletion [Gapdh: U = 28, p = 0.871; Mrpl40: t₍₁₄₎ = 4.225, p < 0.001; Cldn5: t₍₁₄₎ = −0.0174, p = 0.986; Ufd1l: t₍₁₄₎ = −0.368, p = 0.719) (Supplementary Fig. 6). The full complement of Mrpl40 gene appeared to be essential for prenatal development as its homozygous deletion was embryonically lethal. The enhanced STP and augmentation in Mrpl40^+/− mice were comparable to that in Df(16)5^+/− mice, suggesting that hemizygous deletion of Mrpl40 underlies the abnormal synaptic plasticity in Df(16)5^+/− mice.

Df(16)5 deletion does not disrupt mitochondrial structure or oxidative phosphorylation

Because Mrpl40 is thought to be one of the proteins of the mitoribosome complex^{34, 35}, we investigated whether mitochondrial numbers, structure, or function were affected in Df(16)5^+/− mice. We found no significant difference in mitochondrial ultrastructure imaged with TEM in the CA1 area of the hippocampus (Fig. 2a), in total mitochondrial DNA (Co1: U = 89, p = 0.147; Nd2: U = 101, p = 0.318) or in oxidative phosphorylation [F_(1,8) = 0.108, p = 0.745] in isolated mitochondria from the hippocampus between Df(16)5^+/− and WT littermates, suggesting normal energy production in mice with a Df(16)5-hemizygous deletion (Supplementary Fig. 7). Furthermore, 3D scanning electron microscopy (SEM) imaging of the hippocampal CA1 area revealed a normal distribution of mitochondria in presynaptic terminals of Df(16)5^+/− mice compared to WT littermates (U = 34.5, p = 0.093) (Supplementary Fig. 8), suggesting normal trafficking of mitochondria to presynaptic terminals in Df(16)5^+/− mice.

Figure 2. Normal mitochondrial structure but abnormal presynaptic cytosolic and mitochondrial calcium regulation in Df(16)5^+/− and Mrpl40^+/− mice.

(a) Three representative TEM images of mitochondrial ultrastructure in the CA1 area of the hippocampus of WT and Df(16)5^+/− mice. (b) Representative fluorescent image of mCherry after infection of the CA3 area with AAVs encoding either GCaMP6 or mitoGCaMP6. (c) Line scan of mCherry and GCaMP6 fluorescence in a CA3 presynaptic terminal before and after the 80-Hz stimulation of Schaffer collaterals (arrow). (d-g) Mean normalized cytosolic GCaMP6 (d, f) and mitoGCaMP6 (e, g) fluorescence in CA3 presynaptic terminals imaged in the CA1 area of the hippocampus, before and after 80-Hz stimulation in Df(16)5^+/− and WT littermates (d, e) and Mrpl40^+/− and WT littermates (f, g). (h, i) Normalized mean peak amplitudes of GCaMP6 (h) or mitoGCaMP6 (i) in Df(16)5^+/− and WT littermates and Mrpl40^+/− and WT littermates. Numbers of fluorescent puncta are shown inside columns. *P <0.05.

Dysregulation of activity-dependent presynaptic and mitochondrial Ca²⁺ in Mrpl40^+/− mice

Because mitochondria also regulate presynaptic Ca²⁺ levels, we measured activity-dependent Ca²⁺ changes in response to the 80-Hz stimulation of Schaffer collaterals in presynaptic CA3 terminals in the hippocampal stratum radiatum (CA1 area). To this end, we took advantage of the highly sensitive genetically encoded Ca²⁺ indicator GCaMP6f (GCaMP6)²⁵. After infecting the CA3 area of the hippocampus in vivo with recombinant adeno-associated viruses (AAVs) encoding mCherry and GCaMP6, we observed high expression of fluorescent proteins in neuronal cell bodies in the CA3 area but not in the CA1 area (Fig. 2b). In the stratum radiatum we observed fluorescent boutons, which responded to 80-Hz electrical stimulation (10 pulses) of Schaffer collaterals with GCaMP6 fluorescence transients with fast rise and decay (rise time_20%-80%, 100.46 ± 4.03 ms; decay time (τ), 222.71 ± 9.40 ms) (Fig. 2c). These activity-dependent kinetics of GCaMP6 fluorescence were similar to those observed in CA3 presynaptic terminals in which the inorganic Ca²⁺ indicator Fluo 5F was used (data not shown), suggesting that GCaMP6 reliably measures cytosolic Ca²⁺ in presynaptic terminals. The activity-dependent increase in GCaMP6 fluorescence during the 80-Hz train of stimulations was substantially higher (U = 537, p < 0.001) in Df(16)5^+/− mice than in their WT littermates (Fig. 2d), which is consistent with the notion that higher presynaptic Ca²⁺ levels lead to elevated augmentation and STP²⁰.

To directly measure Ca²⁺ in the CA1 stratum radiatum mitochondria localized to the presynaptic terminals originating from CA3, we expressed GCaMP6 with a mitochondrial-localization signal (mitoGCaMP6) in the CA3 area using recombinant AAVs (Fig. 2b). We verified the specific localization of mitoGCaMP6 to mitochondria using subcellular fractionation followed by Western blotting and co-immunolocalization with mitochondrial markers (Supplementary Fig. 9). We also verified that the activity-dependent increase in mitoGCaMP6 fluorescence was sensitive to an inhibitor of the mitochondrial Ca²⁺ uniporter Ru360 (10 µM) (U = 15, p < 0.001) (Supplementary Fig. 10). An 80-Hz train (10 pulses) applied to the Schaffer collaterals induced an activity-dependent increase in mitoGCaMP6 but with substantially slower kinetics compared to cytosolic GCaMP6 (mitoGCaMP6: rise time_20%-80%, 210.21 ± 27.03 ms, decay time > 1 s; GCaMP6: rise time_20%-80%, 100.46 ± 4.03 ms; decay time (τ), 222.71 ± 9.40 ms; U = 540, p < 0.001) (Fig. 2e). Similar to cytosolic Ca²⁺, mitochondrial Ca²⁺ was elevated (U = 889, p = 0.020) in Df(16)5^+/− mice in response to the 80-Hz synaptic stimulation (Fig. 2e). The increases in cytosolic and mitochondrial Ca²⁺ were also observed in Mrpl40^+/− mice to a similar degree as in Df(16)5^+/− mice (GCaMP6: U = 482, p < 0.001; mitoGCaMP6: U = 618, p < 0.001) (Fig. 2f–i). These data suggest that Mrpl40 is the gene in the Df(16)5 genomic region that is responsible for the STP phenotype by deregulating activity-dependent mitochondrial and cytoplasmic presynaptic Ca²⁺ dynamics.

Mrpl40^+/− mice are deficient in short-term but not long-term spatial memory or long-term synaptic plasticity

To test if the hemizygous Mrpl40 deletion affects cognitive function, we compared the performance of Mrpl40^+/− and WT mice in several behavioral tests. Mrpl40^+/− mice behaved normally in the acoustic startle [F_(1,7) = 0.0711, p = 0.79] and pre-pulse inhibition [F_(1,2) = 1.314, p = 0.274] of acoustic startle tests (Fig. 3a, b), a measure of sensorimotor gating that is believed to be associated with positive symptoms of SCZ³¹. To test spatial working memory, we used a delayed, non-matched-to-position task (DNMPT), in which timing between runs ranged from 0 to 5 s. In this test Mrpl40^+/− mice performed significantly worse [t₍₂₈₎ = 3.3, p = 0.003] than WT littermates (Fig. 3c). However, Mrpl40^+/− mice performed normally in the tasks that assessed long-term memory (e.g., Morris Water Maze task). In this task, mutant mice learned to find the invisible escape platform (Fig. 3d) and retained this spatial memory for 48 h similar to WT controls [F_(1,3) = 0.0425, p = 0.837, Fig. 3e]. Mrpl40^+/− mice also performed comparably to WT controls [day 1: t₍₃₃₎ = −0.457, p = 0.651; day 2: U = 149.5, p = 0.921] when the escape platform was visible (Fig. 3f). Mrpl40^+/− mutants also showed no memory deficits in the Y-maze [t₍₁₄₎ = 0.160, p = 0.873], where we measured the amount of time a mouse spends in a novel arm 1 h after exploring the other two arms of the maze (Fig. 3g).

Figure 3. Deficient working memory and normal long-term spatial memory, long-term synaptic plasticity, and acoustic startle in Mrpl40^+/− mice.

(a, b) Mean maximal startle responses as a function of sound intensity (a) and prepulse inhibition (PPI) (b) in Mrpl40^+/− and WT littermates. (c) The mean numbers of correct choices made in the delayed non–matched-to-position task by Mrpl40^+/− and WT littermates. (d-f) Morris water maze tasks. Average time to find a submerged platform during the learning phase (d), time spent in quadrants during the probe test performed 24 h after the last learning session (e), and time to travel to a visible platform (f) in Mrpl40^+/− and WT littermates. Abbreviations: AdL, adjacent left quadrant; AdR, adjacent right quadrant; Opp, opposite quadrant; Tg, target quadrant. (g) Average percentage of time spent in the novel arm in the novel-recognition version of the Y-maze task by Mrpl40^+/− and WT mice. (h) Long-term potentiation at CA3–CA1 synapses measured as mean field EPSP (fEPSP) as a function of time before and after tetanization of Schaffer collaterals with 200-Hz trains (arrows) in Mrpl40^+/− and WT mice. Dashed lines in c and g indicate the level of performance expected by chance. Numbers of mice or slices are shown in parentheses or inside columns. *P <0.05.

These data suggest that Mrpl40 haploinsufficiency affects short-term (working) memory but not long-term memory. Consistent with this notion, STP was abnormal in Mrpl40^+/− mice (Fig. 1g, h), but long-term potentiation (LTP) of excitatory synaptic transmission, a major form of long-term synaptic plasticity at CA3–CA1 synapses, did not differ (t₍₅₃₎ = 0.161, p = 0.873) between Mrpl40^+/− mice and WT littermates (Fig. 3h).

Mitochondrial Ca²⁺–extrusion deficit underlies STP and Ca²⁺ phenotypes in Mrpl40^+/− mice

The abnormally high increase in cytoplasmic Ca²⁺ induced by the 80-Hz synaptic stimulation in Mrpl40^+/− mice coincided with the enhanced mitochondrial Ca²⁺ increase. A role for slow mitochondrial Ca²⁺ extrusion in STP has been implicated in crayfish neuromuscular junction³⁶ and led us to hypothesize that our results in the hippocampus could also be explained by impaired Ca²⁺ extrusion from mitochondria. Two major mechanisms extrude Ca²⁺ from the mitochondrial matrix to the cytoplasm: mitochondrial Ca²⁺ exchangers and the mPTP^{37, 38}. The selective antagonist of the mitochondrial Na⁺–Ca²⁺ exchanger, CGP 37157 (5 µM), had no effect on STP or augmentation in WT mice [STP: t₍₁₈₎ = 0.999, p = 0.333; augmentation: t(₁₈₎ = 1.007, p = 0.330], suggesting that this Ca²⁺ extrusion mechanism is not required for Ca²⁺ handling by mitochondria during 80-Hz–induced synaptic plasticity (Supplementary Fig. 11). However, bongkrekic acid (BKA, 2 µM), a non-selective inhibitor of the adenine nucleotide (ADP/ATP) translocases (ANTs)³⁹, which are required for sensitivity of mPTP to calcium⁴⁰, significantly increased STP and augmentation [STP: U = 14, p = 0.005; augmentation: t₍₁₉₎ = −3.169, p = 0.005] in WT mice (Fig. 4a, Supplementary Fig. 12a,c). This increase mimicked the STP and augmentation enhancement in Df(16)5^+/− and Mrpl40^+/− mice. Interestingly, BKA did not increase STP or augmentation further [STP: t₍₁₉₎ = −0.107, p = 0.916; augmentation: t₍₁₉₎ = −0.928, p = 0.365] in Df(16)5^+/− mice (Fig. 4a, Supplementary Fig. 12b,d). BKA also did not affect basal synaptic transmission [F_(3,12) = 1.179, p = 0.318], paired-pulse facilitation [F_(3,5) = 0.402, p = 0.752], or recovery from depression [F_(3,23) = 1.114, p = 0.358] in either WT or Df(16)5^+/− mice (Supplementary Fig. 13), indicating that the BKA effect is specific for the augmentation component of STP. Furthermore, BKA significantly enhanced the magnitudes of cytosolic and mitochondrial Ca²⁺ transients evoked by the 80-Hz train in WT mice (GCaMP6: Q = 2.821, p < 0.05; mitoGCaMP6: Q = 4.908, p < 0.05) (Fig. 4b,c, Supplementary Fig. 12e,g) but failed to increase Ca²⁺ transients in Df(16)5^+/− mice (GCaMP6: Q = 0.246, p > 0.05; mitoGCaMP6: Q = 1.767, p > 0.05) (Fig. 4b,c, Supplementary Fig. 12f,h), suggesting that the STP enhancement in Df(16)5^+/− mutants acts through the same mechanisms as BKA.

Figure 4. The mPTP inhibitor BKA mimics the STP and calcium transient phenotypes of Df(16)5^+/− and Mrpl40^+/− mice.

(a-f) BKA effect on augmentation (a,d), peak GCaMP6 (b,e), and peak mitoGCaMP6 fluorescence intensities (c,f) in WT and Df(16)5^+/− littermates (a-c) or WT and Mrpl40^+/− littermates (normalized to WT levels) (d-f). Numbers of neurons or fluorescent puncta are shown inside columns. *P <0.05.

Like Df(16)5^+/− mice, Mrpl40^+/− mice showed no effect of BKA on STP, augmentation, or Ca²⁺ transients [STP: U = 52, p = 0.371; augmentation: t₍₂₃₎ = −1.091, p = 0.287; GCaMP6: Q = 0.418, p > 0.05; mitoGCaMP6: Q = 0.052, p > 0.05] in presynaptic terminals or mitochondria, whereas BKA enhanced these parameters in WT littermates [STP: t₍₁₃₎ = −3.685, p = 0.0028; augmentation: t₍₁₃₎ = −0.2670, p = 0.02; GCaMP6: Q = 5.038, p < 0.05; mitoGCaMP6: Q = 2.645, p < 0.05] (Fig. 4d–f, Supplementary Fig. 14). These results suggest that haploinsufficiency of the Df(16)5 gene Mrpl40 impairs Ca²⁺ handling by mPTP. Therefore, we sought to rescue this deficit by enhancing mPTP function through overexpression of ANTs. To that end, we designed the AAV-Slc25a4 OE (Fig. 5), which overexpresses Ant1 (also known as Slc25a4), a gene encoding a regulator of mPTP activity^{40, 41}. Only three isoforms of Ant (Ant1, 2, and 4) have been found in mice^{42, 43}, and we identified Ant1 as the highest expressed isoform in the mouse hippocampus (data not shown). When expressed in the hippocampus, AAV-Slc25a4 OE increased the level of Slc25a4 protein [t₍₇₎ = 7.868, p < 0.001] (Supplementary Fig. 15a, b). We verified that AAV-Slc25a4 OE did not change the localization of Slc25a4 protein to mitochondria. Thus, Slc25a4 was co-localized with the mitochondrial fluorescent marker mitotracker (Supplementary Fig. 15c), and AAV-Slc25a4 OE did not change this pattern (Supplemental Fig. 15d). AAV-Slc25a4 OE expressed in the CA3 area of the hippocampus (Fig. 5a) rescued STP and augmentation in Mrpl40^+/− mice [STP: U = 15, p = 0.005; augmentation: t₍₂₂₎ = 2.543, p = 0.019] but did not affect WT littermates [STP: t₍₁₆₎ = 0.566, p = 0.578 ; augmentation: t₍₁₆₎ = −0.654, p = 0.522] (Fig 5b, Supplementary Fig. 16).

Figure 5. The mPTP gain- or loss-of-function molecular manipulations rescue or mimic, respectively, the STP phenotypes of Mrpl40^+/− mice.

(a) Overexpression of Slc25a4 and GFP in the CA3 area of the hippocampus. (b) Mean augmentation measured in sham- or Slc25a4-OE–injected WT and Mrpl40^+/− mice. (c) Mean augmentation measured in control or Slc25a4 shRNA–injected WT and Mrpl40^+/− mice. The data are shown for three different Slc25a4 shRNAs (shRNA1, shRNA2, shRNA3). Normalized to respective WT control levels. Numbers of neurons are shown inside columns. *P <0.05. (d) Model of mPTP-dependent mechanisms of STP dysfunction in 22q11DS. MRPL40 haploinsufficiency in 22q11DS reduces mitochondrial Ca²⁺ extrusion through impaired mPTP. This leads to a Ca²⁺ build-up in the mitochondrial matrix and enhanced Ca²⁺ transients in the mitochondrial matrix and cytosol during high-frequency activity, which in turn leads to enhanced synaptic vesicle release in presynaptic terminals. VGCC, voltage-gated calcium channels, NCX, Na⁺/Ca²⁺ exchanger, MCU, mitochondrial Ca²⁺ uniporter, VDAC1, voltage-dependent anion channel 1. Upper traces represent cytosolic Ca²⁺ transients (GCaMP6) and lower traces represent mitochondrial Ca²⁺ transients (mitoGCaMP6).

To confirm that the BKA-induced reduction in mPTP function mimics the Df(16)5^+/− phenotype, we downregulated Slc25a4 by using the shRNA approach. To that end, we injected lentiviruses encoding three different shRNAs against Slc25a4 into the CA3 area of the hippocampus. STP and augmentation of synaptic transmission at CA3–CA1 synapses increased in WT mice infected with all three Slc25a4 shRNAs compared to the control shRNA [shRNA1 STP: t₍₁₃₎ = −1.882, p = 0.041; augmentation: U = 9, p = 0.029; shRNA2 STP: t₍₁₅₎ = −1.910, p = 0.04; augmentation: t₍₁₇₎ = −2.757, p = 0.014; shRNA3 STP:U = 9, p = 0.02; augmentation: t₍₁₃₎ = −2.475, p = 0.028] (Fig. 5c, Supplementary Fig. 17). However, Slc25a4 shRNAs did not further affect STP or augmentation in Mrpl40^+/− mice [shRNA1 STP: U = 23, p = 0.382; augmentation: U = 28, p = 0.721; shRNA2 STP: U = 23, p = 0.091; augmentation: t₍₁₉₎ = 1.038, p = 0.312; shRNA3 STP U = 27, p = 0.267; augmentation: U = 24, p = 0.107] (Fig. 5c, Supplementary Fig. 17), suggesting that Mrpl40 haploinsufficiency altered STP by affecting mPTP function.

Discussion

Approximately 30% of patients with 22q11DS meet the diagnostic criteria for SCZ⁴. Cognitive symptoms of SCZ include deficits in working memory, attention, executive function, and learning and memory; these symptoms have a more prognostic value than do the positive or negative symptoms of the disease, and they contribute more to the patients’ functional disability⁴⁴. The mechanisms underlying the cognitive deficits in SCZ and 22q11DS are still debated. Here we presented evidence that 22q11DS affects working memory and STP through a novel pathogenic mechanism––abnormal Ca²⁺ handling by mitochondria. We also elucidated several features of this pathogenic mechanism: (1) By screening mice carrying hemizygous deletions of individual genes mapped within the distal part of the 22q11 microdeletion, we identified Mrpl40 as a culprit gene that causes abnormal STP. (2) We identified that the augmentation component of STP is specifically affected. (3) Mrpl40 haploinsufficiency led to abnormal STP through mitochondria-mediated deregulation of presynaptic Ca²⁺ levels. (4) We pinpointed a functional abnormality in mitochondrial Ca²⁺ extrusion through the mPTP as a pathogenic mechanism caused by Mrpl40 haploinsufficiency. (5) Mrpl40 haploinsufficiency led to deficits in working memory but not in long-term plasticity or long-term memory.

The hippocampus, prefrontal cortex, and interactions between these brain regions all contribute to working memory^45–48. Deficits in working memory in patients with SCZ and/or 22q11DS are well documented^{49, 50}. STP, an associative, short-lived synaptic strengthening, is considered a cellular correlate of short-term memory, a term often used synonymously with working memory^{20, 51}. Consistent with this notion, deficits in STP are well established in animal models of SCZ and 22q11DS^{12, 48, 52–54}.

Because symptoms of SCZ typically appear during late adolescence or early adulthood, age appears to be an important variable in synaptic plasticity phenotypes associated with 22q11DS mice. In our experiments, we used mice that were older than 4 months. At this age, STP and LTP at CA3–CA1 synapses are substantially increased in Df(16)1^+/− mouse models of 22q11DS, whereas at younger ages, both forms of synaptic plasticity are normal²³. These experiments led to the identification of the microRNA-processing gene Dgcr8 as the culprit gene residing in the proximal part of the microdeletion affecting both short- and long-term synaptic plasticity in the hippocampus. Deletion of one allele of Dgcr8 causes depletion of miR-185 and miR-25 and posttranscriptional upregulation of Serca2 in the forebrain of older but not younger mice²³.

Because 22q11DS is a multigene syndrome, it is extremely likely that more than one gene is involved in STP abnormalities. Our present work showed that deletion of one allele of Mrpl40 residing in the distal part of the microdeletion, which is outside of the Df(16)1 genomic region caused independent deficits in STP and working memory in 22q11DS. First, we identified the STP increase in Df(16)5^+/− mice, which were hemizygous for genes that mapped distally to Dgcr8. The STP screen revealed that only Mrpl40^+/− mice replicated the Df(16)5^+/− phenotype. Furthermore, the Mrpl40-related STP increase was mechanistically distinct from the STP increase identified in Dgcr8^+/− mice. Dgcr8 haploinsufficiency elevates Serca2 protein¹², but in the Df(16)5^+/− mice, Serca2 levels were normal, and the Serca inhibitor thapsigragin did not rescue the STP defect.

Mrpl40 was originally identified as a nuclear gene involved in mitochondrial function⁵⁵. Mitochondrial dysfunction has been strongly implicated in SCZ pathophysiology⁵⁶, though the exact connection between SCZ pathogenesis and mitochondria has not been established. For instance, the gene encoding DISC1 (disrupted-in-schizophrenia 1) has been localized to mitochondria and is involved in maintaining mitochondrial morphology and regulating mitochondrial transport⁵⁷. Several mitochondrial genes, including Mrpl40, are mapped within the 22q11.2 locus and expressed in the brain throughout development, thus implicating mitochondria in the pathogenesis of 22q11DS^{58, 59}. Mrpl40 haploinsufficiency does not affect the mitochondrial ultrastructure and does not reduce mitochondrial numbers in presynaptic terminals or total mitochondrial DNA, which is consistent with the view that SCZ is not a neurodegenerative disease. This is also consistent with our observations that Df(16)5^+/− or Mrpl40^+/− mice develop normally and have normal gross brain morphology.

Mitochondria provide energy through oxidative phosphorylation and regulate Ca²⁺ dynamics during synaptic transmission in neurons. Because of the mitochondrion’s high capacity for Ca²⁺ uptake, it acts as a rapid buffering system during periods of intense synaptic activity, then slowly releases Ca²⁺ when activity subsides. This rapid uptake of Ca²⁺ into mitochondria occurs through channels (i.e. voltage-dependent anion channel VDAC1) in the outer membrane and the mitochondrial calcium uniporter in the inner membrane. Ca²⁺ is slowly released from the mitochondria through the calcium exchangers and via the mPTP. mPTP has been shown to open during high frequency stimulation in mammalian neurons, and that transient activation of the mPTP may play a role in the normal contribution of mitochondria to STP⁶⁰.

Oxidative phosphorylation appeared to be normal, whereas Ca²⁺ dynamics in presynaptic terminals was substantially altered in Df(16)5^+/− and Mrpl40^+/− mice. Our two-photon Ca²⁺ imaging in presynaptic terminals revealed that synaptic high frequency stimulation that induces STP evokes enhanced Ca²⁺ transients in both the presynaptic mitochondrial matrix and presynaptic cytosol of mutant mice. This data argues for a problem with mitochondrial Ca²⁺ extrusion rather than with mitochondrial Ca²⁺ uptake. Indeed, if Mrpl40 haploinsufficiency reduced the Ca²⁺ uptake into mitochondria, we would expect to see increased amplitudes of cytoplasmic Ca²⁺ transients but decreased Ca²⁺ transients within mitochondria. Our data is more consistent with a model for a reduced Ca²⁺ extrusion from mitochondria. Impaired Ca²⁺ extrusion from mitochondria will result in Ca²⁺ accumulation in the mitochondrial matrix and this will contribute to the fast rise in mitochondrial Ca²⁺ that we observed with mitoGCaMP6. This will then reduce the effective mitochondrial buffering capacity and lead to enhanced cytoplasmic Ca²⁺ transients that we observed with the cytoplasmic GCaMP6 (Fig. 5d). Moreover, slow Ca²⁺ extrusion from mitochondria (evident from extremely slow decay times of mitoGCaMP6) is an unlikely contributor to fast cytosol Ca²⁺ transients that occur during high frequency synaptic activity.

Because inhibition of mPTP but not the mitochondrial Na⁺–Ca²⁺ exchanger increased presynaptic Ca²⁺ transients in WT mice, we concluded that the main route of Ca²⁺ extrusion from mitochondria during STP is mPTP. In WT mice, the pharmacological or molecular inhibition of mPTP with BKA or Slc25a4 shRNA, respectively, mimicked the STP and Ca²⁺-transient phenotypes we observed in Df(16)5^+/− and Mrpl40^+/− mice. Moreover, AAV-Slc25a4 OE rescued the STP abnormality in Mrpl40^+/− mice, indicating that deletion of one allele of Mrpl40 causes mPTP deficiency and STP increase.

Although our experiments demonstrated that Mrpl40 haploinsufficiency leads to abnormal Ca²⁺ handling by mitochondria during STP, the exact connection between Mrpl40 and mPTP function remains unclear. Some pharmacological agents that affect mPTP function⁶¹ can be beneficial adjuvants to antipsychotics to alleviate SCZ symptoms⁵⁶. However, because mPTP⁸ is involved not only in Ca²⁺ extrusion from mitochondria but also in ATP transport, we cannot rule out the possibility that impaired mPTP regulation leads to the abnormal production of ATP in presynaptic terminals, which in turn could result in abnormal presynaptic function. This theory could be addressed using a novel ATP probe that visualizes ATP dynamics in the presynaptic terminals of neurons in culture, but those experiments have not yet been tailored for acute brain slices⁶².

In summary, we report that haploinsufficiency of Mrpl40, a gene mapped within the 22q11.2 microdeletion associated with SCZ, causes deficient working memory. This behavioral abnormality, which typically manifests in patients with SCZ, is associated with abnormal STP changes in synaptic transmission in the hippocampus and caused by elevated presynaptic Ca²⁺ and impaired Ca²⁺ extrusion from mitochondria.