Departments of Psychiatry and Behavioral Sciences, Neurobiology, and Cell Biology, Mouse Behavioral and Neuroendocrine Analysis Core Facility, Duke University Medical Center, Durham, North Carolina, USA
Departments of Psychiatry and Behavioral Sciences, Neurobiology, and Cell Biology, Mouse Behavioral and Neuroendocrine Analysis Core Facility, Duke University Medical Center, Durham, North Carolina, USA
Copper (Cu), an essential trace element present throughout the mammalian nervous system, is crucial for normal synaptic function. Neuronal handling of Cu is poorly understood. We studied the localization and expression of Atp7a, the major intracellular Cu transporter in the brain, and its relation to peptidylglycine α-amidating monooxygenase (PAM), an essential cuproenzyme and regulator of Cu homeostasis in neuroendocrine cells. Based on biochemical fractionation and immunostaining of dissociated neurons, Atp7a was enriched in post-synaptic vesicular fractions. Cu followed a similar pattern, with ~ 20% of total Cu in synaptosomes. A mouse model heterozygous for the Pam gene (PAM+/−) was selectively Cu deficient in the amygdala. As in cortex and hippocampus, Atp7a and PAM expression overlap in the amygdala, with highest expression in interneurons. Messenger RNA levels of Atox-1 and Atp7a, which deliver Cu to the secretory pathway, were reduced in the amygdala but not in the hippocampus in PAM+/− mice, GABAB receptor mRNA levels were similarly affected. Consistent with Cu deficiency, dopamine β-monooxygenase function was impaired as evidenced by elevated dopamine metabolites in the amygdala, but not in the hippocampus, of PAM+/− mice. These alterations in Cu delivery to the secretory pathway in the PAM+/− amygdala may contribute to the physiological and behavioral deficits observed.
Atp7a, a Cu-transporting P-type ATPase, is localized to the trans-Golgi network and to vesicles distributed throughout the dendritic arbor. Tissue-specific alterations in Atp7a expression were found in mice heterozygous for peptidylglycine α-amidating monooxygenase (PAM), an essential neuropeptide-synthesizing cuproenzyme. Atp7a and PAM are highly expressed in amygdalar interneurons. Reduced amygdalar expression of Atox-1 and Atp7a in PAM heterozygous mice may lead to reduced synaptic Cu levels, contributing to the behavioral and neurochemical alterations seen in these mice.
Copper (Cu), an essential trace element with diverse roles in biology, is known for its role as an obligate cofactor in oxidation–reduction reactions catalyzed by a dozen enzymes. These cuproenzymes play key roles in aerobic respiration, free radical detoxification, and neurotransmitter biosynthesis (Andreini et al. 2008). A much larger number of proteins are dedicated to the handling and transport of Cu (Uriu-Adams and Keen 2005). Cu homeostasis from the cellular to the multi-organ level is not yet well understood, but it is clear that disruption leads to disease.
Mutations in the two intracellular Cu transporting ATPases, ATP7A, and ATP7B (Lutsenko et al. 2007), disrupt Cu homeostasis. Menkes disease (OMIM #309400), with mutations in ATP7A, causes Cu deficiency with gray kinky hair, frail skin, immune compromise, seizures, and death before 3 years of age without treatment (Kaler 2011). Wilson's disease (OMIM #277900), with mutations in ATP7B, causes Cu overload with hepatic cirrhosis, hemolytic anemia, Kayser-Fleischer rings in the cornea, neurodegeneration, and death if left untreated (de Bie et al. 2007). Other diseases, including Alzheimer's and prion-related neurodegenerative diseases, are associated with imbalances in Cu homeostasis (Zomosa-Signoret et al. 2008; Lin et al. 2010). The severity of these conditions highlights the critical nature of maintaining tight control of Cu levels, especially in neuronal tissue.
Both Atp7A and Atp7B receive cytosolic Cu from Atox-1 and transport Cu into the lumen of the secretory pathway. Cu is bound by lumenal cuproenzymes like peptidylglycine α-amidating monooxygenase (PAM) (El Meskini et al. 2003; Steveson et al. 2003) and dopamine β-monooxygenase (DβM) (Rush and Geffen 1980, 1980; Klinman 2006, 2006) or packaged into vesicles for secretion. Atp7a is the major Cu transporting ATPase in the brain and levels of the amidated peptides produced by PAM and norepinephrine, which is produced by DβM, are reduced in mouse models of Menkes Disease (Niciu et al. 2006; Donsante et al. 2013). As in other cell types, Atp7a trafficking in neurons is responsive to Cu (Schlief et al. 2005; La Fontaine and Mercer 2007). Although little is known about Cu homeostasis in neurons, it is clear that NMDA receptor stimulation results in the Atp7a-dependent post-synaptic secretion of Cu (Schlief et al. 2005, 2006).
We recently developed a mouse model in which to explore the effects of altered Cu handling on fear learning. Mice heterozygous for the Pam gene (PAM+/−) display impaired learned fear behaviors, deficits in amygdalar synaptic plasticity and decreased levels of Cu in the amygdala (Gaier et al. submitted). An increase in PAM+/− dietary Cu intake ameliorated the fear learning deficit and eliminated the deficits observed in vitro in PAM+/− amygdala slice recordings. Moreover, chelation of endogenous Cu in amygdala synaptic plasticity experiments abolished long-term potentiation (LTP) in wild-type mice. These studies provide compelling evidence of an essential role for Cu in amygdalar synaptic function.
In this study, we examined the subcellular localization of Atp7a and Cu in the hippocampus and cortex. Next, we explored biochemical evidence for disrupted Cu handling in the PAM+/− amygdala in order to explore mechanisms through which PAM haploinsufficiency might cause the behavioral and electrophysiological deficits observed in these animals. We hypothesize that alterations in Cu handling underlie the physiological and behavioral effects of Pam heterozygosity on amygdala function, and highlight the importance of Cu homeostasis in neuronal and synaptic function.
Male and female mice for these studies were generated from female wild-type x male PAM+/− matings in the University of Connecticut Health Center (UCHC) animal facility. Wild-type and PAM+/− littermates (> 20 generations bred into C57/BL6J background) were weaned between post-natal days 19 and 21 and group housed until experiments. Animals were maintained under a 12 h light/dark cycle (lights on at 7:00 A.M.) and were given ad libitum access to food and water. All experiments were conducted with approved protocols from the UCHC Institutional Animal Care and Use Committees, in accordance with National Institutes of Health guidelines for animal care and the ARRIVE guidelines.
Timed-pregnant wild-type C57/BL6J mice were sacrificed at gestational day 18, embryos were removed and cortices were dissected into ice-cold Hanks balanced salt solution (Gibco, Carlsbad, CA, USA). Tissue was rinsed, trypsinized, and triturated using polished glass Pasteur pipettes. Dissociated neurons were collected by centrifugation, suspended in Dulbecco's modified Eagle's medium: nutrient mixture F-12 with 10% FCS and penicillin/streptomycin and plated at low density (~ 1 × 105 cells per cm2) onto coverslips coated with poly-l-lysine (1 mg/mL in Borate Buffer: 50 mM Boric acid, 23 mM sodium tetraborate, pH 8.5). Medium was half-changed with Neuronal Growth Medium (Neurobasal medium containing 3% horse serum, 2% B27, 0.4 mM Gibco GlutaMAX, 0.5% penicillin/ streptomycin) after 24 h, and then with Neuronal Maintenance Medium (Neurobasal medium containing B27, GlutaMAX, pen/strep) every 2–3 days thereafter. For immunocytochemistry, cultures were fixed with 4% formaldehyde in phosphate-buffered saline (PBS) after 23 days in vitro. Cultures were permeabilized using 0.075% Triton X-100 in PBS, then incubated in blocking buffer (2 mg/mL bovine serum albumin in PBS) for 1 h. Immunocytochemistry was conducted with indicated combinations of the following antisera: Atp7a [CT77 (Niciu et al. 2006, 2007), rabbit], MAP2 (Sigma, St. Louis, MO, USA; mouse), post-synaptic densities (PSD)-95 (NeuroMab, Davis, CA, USA; mouse), and Vglut1 (Millipore, Bedford, MA, USA; guinea pig). Primary antibody binding was visualized using spectrally distinct, fluorescently tagged secondary antibodies: Cy3-donkey anti-rabbit, Alexa-633 donkey anti-mouse, FITC-donkey anti-guinea pig. Images were taken with a Zeiss Axiovert 200M microscope using 40X, 63X, and 100X oil objectives and AxioVision software (Carl Zeiss Microscopy, LLC, Thornwood, NY, USA). Optical sectioning was achieved with the Zeiss ApoTome module, and 3–5 sections of 0.2–0.4 μm each were compressed for each image.
Immunohistochemical staining of tissue sections from perfusion-fixed mice was described previously (Ma et al. 2001, 2002, 2008). Briefly, male wild-type and PAM+/− littermates were perfused transcardially with 4% paraformaldehyde/0.1 M sodium phosphate buffer (pH 7.4) under deep ketamine/xylazine anesthesia. After fixation, brains were post-fixed in 4% paraformaldehyde for 3 h. Coronal sections were cut (15 μm) through the amygdala using a cryostat and immunostained simultaneously with rabbit antiserum JH629 to the Exon A (exon 16) region of PAM1 (Maltese and Eipper 1992; Hansel et al. 2001) and mouse monoclonal IgG to Atp7a (Neuromab, University of California Davis, CA, USA) (Niciu et al. 2006, 2007). Alexa-488 donkey anti-rabbit (H+L) antibody (Invitrogen, Molecular Probes, Eugene, OR, USA) and Cy™3-conjugated AffiniPure F(ab')2-fragment donkey anti-mouse (H+L) antibody (Jackson ImmunoResearch Laboratories Inc., West Grove, PA) were then applied. To-Pro3 (Invitrogen-Molecular Probes) was added separately as described (Ma et al. 2001). Images are single focus planes captured using an automated Zeiss LSM 510 Meta confocal microscope at the Center for Cell Analysis and Modeling, UCHC. Low power images employed a 20X 0.75 NA air objective using a 92 μm pinhole; high power images employed a 63X 1.4 NA oil DIC objective using a 92 μm pinhole. Sections from PAM+/− and wild-type mice were imaged under identical conditions. Blocking and specificity controls for this Atp7a antibody were reported previously (Niciu et al. 2006, 2007).
PSD fractionation and western blot analysis
Subcellular fractionation was used to prepare samples enriched in endoplasmic reticulum/Golgi, cytosol, synaptosomal membranes, synaptic vesicles, and synaptosomal cytosol (Huttner et al. 1983; Ma et al. 2008). Purified PSDs were prepared using a modification of published procedures (Carlin et al. 1980; Ma et al. 2008). Samples removed from the interface of the 1.0/1.2 M sucrose layers of an equilibrium gradient were diluted, pelleted, and then solubilized by incubation for 30 min at 4°C with 0.5% Triton X-100 (TX-100), 10 mm HEPES, pH 7.4. PSDs were pelleted and the TX-100-soluble fraction was saved for analysis. Protein concentrations were determined using the bicinchoninic acid assay (Pierce Biotechnology, Rockford, IL, USA) with bovine serum albumin as the standard. Polyclonal rabbit antiserum to Atp7a (CT-77) was described previously (Niciu et al. 2006). Commercial mouse monoclonal antibodies to NR2B (clone N59/20; NeuroMab), PSD95 (clone K28/43; NeuroMab), Vglut1 (clone N28/9; NeuroMab), synaptophysin (clone svp-38; Sigma), GM130 (BD Biosciences, Franklin Lakes, NJ, USA), and βIII-tubulin (Covance, Princeton, NJ, USA) were used. Antigen–antibody complexes were visualized using horseradish peroxidase-conjugated secondary antibody and Super Signal West Pico chemiluminescence substrate (Pierce); images were quantified using Gene Tools (SynGene, Frederick, MD, USA).
Bilateral 2 mm diameter punches from individual mice containing the amygdalae or dorsal hippocampi were isolated from 1 mm coronal sections and immediately transferred from isotonic phosphate buffered solution, pH 7.4, into 500 μL TRIzol. RNA and cDNA were prepared essentially as per the manufacturer's instructions, with minor modifications (Mains et al. 2011). All primers (Table 1) were chosen to have identical melting temperatures and to produce products of 120 ± 5 nt. The maximal fold of amplification was 1.92 ± 0.14. All data were normalized with respect to GAPDH and data for five to six mice were averaged.
Table 1. Quantitative PCR primer sequences and melting temperatures
GAPDH - F
GAPDH - R
mPAM - F
mPAM - R
Atox-1 - F
Atox-1 - R
Atp7a - F
Atp7a - R
GABAA α1 - F
GABAA α1 - R
GABAB 2 - F
GABAB 2 - R
GAD65 - F
GAD65 - R
GAD67 - F
GAD67 - R
Cu level measurements
Inductively coupled plasma mass spectrometry (ICPMS) analysis was performed using an Agilent 7700x equipped with an ASX 250 autosampler. The system was operated at a radio frequency power of 1550 W, an argon plasma gas flow rate of 15 L/min, and Ar carrier gas flow rate of 1.04 L/min. Elements were measured in kinetic energy discrimination (KED) mode using He gas (4.3 mL/min). For the analysis, 50 μL homogenates were transferred into 1% HNO3 treated 15 mL conical centrifuge tubes and digested with 100 μL 50% HNO3 solution. Following overnight, 25 °C digestion of the samples, 850 μL of 1% HNO3 was added, the samples spun at 4700 g for 10 min to remove insoluble residue, and the supernatant transferred into clean, acid treated 15 mL conical tubes. Data were quantified using a 9-point calibration curve [0, 0.5, 1, 2, 5, 10, 50, 100, 1000 ppb (ng/g)] with external standards for Fe, Cu, and Zn (Common Elements Mix 2, Multi-Element Aqueous Standard; VHG Labs, Manchester, NH, USA) in 1% HNO3 (trace metal grade; Thermo Fisher, Waltham, MA, USA). For each sample, data were acquired in triplicate and averaged. An internal standard (Internal Standard Multi-Element Mix 3; VHG Labs) introduced with the sample was used to correct for plasma instabilities; frequent measurements of a 10 ppb Cu solution as well as a blank (containing 1% HNO3 only) were used as quality control and to determine the coefficient of variance. To access recovery rates of elements and probe background contamination from buffers and containers the following controls were treated, prepared, and analyzed by the same method as the samples: certified NIST standard reference material (serum; 1598a), homogenization buffers, acid control (containing 50% HNO3 only), and certified elemental standard for ICPMS (Common Elements Mix 2, Multi-Element Aqueous Standard; VHG Labs).
Catecholamine and metabolite measurements
Bilateral 2 mm diameter punches from individual mice (7 wild-type and 10 PAM+/− adult males) containing the amygdalae or dorsal hippocampi were taken from 1 mm coronal sections and immediately transferred from isotonic phosphate buffered solution, pH 7.4, into marked microfuge tubes and stored in −80°C until analyzed using a modified previously reported method (Anderson et al. 2008). Tissue samples were weighed and then sonicated in ice-cold 0.1 M HClO4 with 100 pg/μL isoproterenol, an internal standard. Homogenates were filtered with 0.22 μm filters (Millipore) and 10 μL of filtrate was injected onto the HPLC. The Coulochem III HPLC system with 5014B analytical cell and 5020 guard cell (ESA Inc., Chelmsford, MA, USA) was used for analysis. The guard cell potential was set at +0.380 mV and the first and second electrodes of the analytical cell were at −0.15 and +0.35 mV, respectively. Separation was carried out at 22 °C on MD-150 C18 150 × 3.2 mm column with 0.55 mL/min mobile phase containing 90 mM sodium dihydrogen phosphate, 50 mM citric acid, 1.7 mM 1-octane sulfonic acid sodium salt, 50 μM EDTA, 9.5% acetonitrile (pH 3). Peak areas were compared against internal standard using EZChrom Elite Version 3.2 (Agilent Technologies, Pleasanton, CA, USA) and concentrations were reported as ng/mg of wet tissues.
Quantitative PCR and catecholamine studies were analyzed using two-way anova by genotype and brain region unless otherwise specified. Interactions were decomposed by Bonferroni corrected pair-wise comparisons. Hypothesis-driven comparisons between genotypes within each brain region were performed using Student's t-tests (as indicated in figures). In all cases, p <0.05 was considered significant.
Atp7a and PAM co-localize in cortex and hippocampus
We first examined the co-localization of Atp7a and PAM protein in the cortex and hippocampus. Coronal slices of perfusion-fixed wild-type mouse brains were stained using antibodies against Atp7a and PAM (Fig.1). Immunostaining controls were published previously for the Atp7a (Niciu et al. 2006, 2007) and PAM (Hansel et al. 2001) antibodies. In somatosensory cortex, Atp7a staining was most intense in the large pyramidal neurons of layer 5 and in pyramidal neurons in layers 2/3 (Fig. 1a); much less staining was evident in layer 4. Atp7a staining was also present in the apical dendrites of layer 5 pyramidal neurons that could be appreciated extending through layers 2/3. PAM staining shared a similar pattern, including expression in the same neuronal somata and extending into apical dendrites. Atp7a and PAM staining co-localized in all neuronal somata, dendrites, and in the layer 5 and layers 2/3 neuropil of the cortex.
In the hippocampus, staining of both Atp7a and PAM was observed in all regions examined, including the dentate gyrus, CA3, and CA1 regions (Fig. 1b–d). In the dentate gyrus, Atp7a staining was present in the molecular layer, the granule cell layer and in the hilus (Fig. 1b). A perinuclear staining pattern was observed in granule cell somata. Atp7a staining was especially intense in putative GABAergic interneurons in the hilus. Atp7a staining intensity was no different in the polymorphic cell layer compared to the granule cell layer. In contrast, PAM staining was highest in neuronal somata in the granule cell layer, but was also present in the molecular layer and hilus. The same putative GABAergic interneurons that stained highly for Atp7a in the hilus also stained intensely for PAM. Neurons staining highly for PAM in the hilus were also GAD67 positive (data not shown), indicating they use GABA as a neurotransmitter. Staining patterns for Atp7a and PAM overlapped, indicating co-localization of these proteins in hilar interneurons.
Atp7a staining in the CA3 region of the hippocampus was most intense in the somata of neurons in the pyramidal layer, and this staining overlapped closely with PAM staining (Fig. 1c). Atp7a staining was also intense in putative GABAergic interneurons (basket cells) in the Stratum Oriens and overlapped closely with PAM staining. Staining for both Atp7a and PAM was present in the neuropil of the Stratum Oriens and Reticularis as puncta; however, PAM staining in these layers was more prominent than Atp7a staining. The CA1 region of the hippocampus showed similar patterns of Atp7a and PAM staining; staining for both overlapped in neurons in the pyramidal layer, but was most prominent in putative interneurons in the Strata Oriens and Radiatum (Fig. 1d). Punctate Atp7a and PAM staining was also observed in the neuropil of the Strata Oriens and Radiatum. Under higher power magnification of the CA1 region, perinuclear prominence and co-localization of Atp7a and PAM staining was observed in putative interneurons (inset). These data show that Atp7a and PAM co-localize in pyramidal neurons and interneurons of the hippocampus.
Subcellular localization of Atp7a
Atp7a staining in the neuropil of the cortex and hippocampus suggested that it was present in dendrites as well as in the cell soma. Atp7a has a C-terminal PDZ-binding motif (Linz and Lutsenko 2007; Kaler 2011), which could facilitate its localization to the PSD, the gigadalton complex of proteins localized at the tips of dendritic spines. To address this possibility, we employed a well-established subcellular fractionation technique to examine the localization of Atp7a in adult mouse cortex and hippocampus (Fig. 2a). Equal amounts of protein from each subcellular fraction were analyzed. As expected, calreticulin, a soluble endoplasmic reticulum protein, was enriched in the P3 fraction and PSD95, a major component of the PSD, was enriched in the LP1 (lysed synaptosomal membrane) fraction (Fig. 2b and c). Golgi membranes, identified using an antibody to GM130, a peripheral membrane protein (Nakamura et al. 1995), were recovered in the P3 and LP2 fractions; GM130 was also recovered from soluble fractions. Synaptic vesicles, identified using an antibody to synaptophysin, were enriched in the LP2 fraction.
Atp7a was recovered in the P3, LP1, and LP2 fractions (Fig. 2b and c). The LP1 fraction was next layered onto a sucrose density gradient; after centrifugation, the lighter myelin membranes and denser mitochondria are separated from synaptosomal membranes. As expected, most of the PSD95 was recovered in fraction D (at the 1.0/1.2 M sucrose interface) (Fig. 2a); synaptophysin was present in less dense fractions (fractions B+C) as well as in fraction D (Fig. 2b and c). For both cortex and hippocampus, Atp7a was recovered from the lighter fractions, along with fraction D, demonstrating its presence in synaptosomal membranes.
To determine whether Atp7a is concentrated at the PSD, where ionotropic glutamate receptors are located, we treated the synaptosomal membranes recovered in Fraction D with 0.5% Triton X-100, separating TX-soluble (TX) proteins from the insoluble PSD complex (Fig. 3a). Blotting for PSD95 demonstrated its presence in the pellet fraction from both cortex and hippocampus; very little PSD95 was solubilized by 0.5% TX-100. Two synaptic vesicle markers, Vglut1 and synaptophysin, were visualized for comparison; both integral membrane proteins were largely solubilized by TX-100. Most of the Atp7a recovered in Fraction D was also solubilized by TX-100.
The behavior of the marker proteins and Atp7a was quantified by determining their enrichment in the LP1, PSD, and TX fractions in comparison to the starting material (S1) (Fig. 3c). Similar patterns were observed in cortex and hippocampus. Like PSD95, Atp7a, and Vglut1 were slightly enriched in the LP1 fraction. Unlike PSD95, Atp7a, Vglut1, and synaptophysin were largely solubilized by TX-100 (Fig. 3c). To determine whether the Atp7a associated with synaptic membranes was located pre- or post-synaptically, we turned to a cell culture system.
Synaptic localization of Atp7a
Cortical neurons prepared from embryonic day 18 mouse pups were grown in culture for 23 days. Neurons were fixed and stained simultaneously for Atp7a, a dendritic marker (MAP2) and a presynaptic marker (Vglut1) (Fig. 4a and b). Vglut1 positive presynaptic endings decorate the dendritic arbor, as expected for a culture of this age. Atp7a staining was present in the cell soma (Fig. 4a) and throughout the dendritic arbor; very little staining for Atp7a was detected in the Vglut1 positive endings aligned along the dendrites.
Separate cultures of the same age were stained simultaneously for Atp7a, PSD95, and Vglut1 (Fig. 4c and d). Atp7a staining appeared as small puncta throughout the width of dendritic branches, and Vglut1 staining was again aligned along the periphery of each dendrite. PSD95 staining was punctate and juxtaposed to Vglut1 puncta, with limited but occasional overlap with Atp7a staining (Fig. 4c and d). Single channel images are shown in Figure S1. Taken together with our biochemical fractionation data, we conclude that a significant fraction of the Atp7a protein in the nervous system is localized to vesicular structures distributed throughout the dendritic arbor, a position from which it could participate in the regulation of Cu homeostasis at synapses.
Cu content of subcellular fractions
Next we turned our attention to Cu itself, using the same subcellular fractionation scheme to prepare samples from the cortices of individual wild-type mice. Cu concentrations were determined for each fraction; the data are presented as percentage of total Cu content in each fraction (Fig.5). The cytosolic fraction (S3) contained approximately 60% of the total Cu in the cortex; Cu bound to chaperone proteins and metallothioneins would be recovered in this fraction. The concentration of Cu in S3 was high (80.5 ± 11.6 fg/mg protein). The ER/Golgi-enriched P3 fraction contained less than 5% of the total Cu and the concentration of Cu in this fraction was low (15.5 ± 0.5 fg/mg protein). After hypotonic lysis of the crude synaptosomes (P2), approximately half of the Cu was recovered from the synaptosomal membrane fraction (LP1); the concentration of Cu in LP1 was 43.8 ± 5.7 fg/mg protein. Further fractionation of the supernatant (LS1), which contains both the soluble and vesicular components of synaptosomes, revealed a substantially higher amount of the total Cu in the soluble fraction (LS2, synaptosomal cytosol) than in the synaptic vesicle fraction (LP2). Over 25% of the total Cu in the cortex was recovered in the crude synaptosomal pellet. Together, these results indicate that while the majority of the Cu in the cortex is contained in the cytosol, Cu is also found in various synaptic compartments, consistent with a role for endogenous Cu in synaptic transmission.
PAM and Atp7a co-localize in lateral amygdalar interneurons
Since our behavioral and electrophysiological studies relating Pam heterozygosity to altered Cu homeostasis focused on the amygdala, we explored the co-localization of PAM and Atp7a in this brain region. We perfusion-fixed wild-type and PAM+/− mouse brains, made coronal slices through the amygdala, and simultaneously visualized Atp7a and PAM (Fig.6). Atp7a staining was indistinguishable in wild-type and PAM+/− amygdala (Fig. 6a and b); staining was most intense in the basolateral complex, especially in the basolateral nucleus. Interestingly, Atp7a staining was prominent in the same afferent fiber bundles targeted for stimulation in physiology experiments performed in PAM+/− mice (Fig. 6a and b; arrows) (Gaier et al. 2010) (Gaier et al. submitted). PAM staining was also abundant in the amygdala (including the lateral and basolateral nuclei) (Fig. 6a and b), confirming previous results (Gaier et al. 2010); based on western blot analysis, PAM expression in PAM+/− tissues is about half the level observed in the corresponding wild-type tissue (Bousquet-Moore et al. 2010b; Gaier et al. 2010). PAM staining overlapped with Atp7a staining in these nuclei. In addition, PAM staining was especially intense in the central nucleus, where Atp7a staining was relatively sparse. Thus, the gross expression patterns of Atp7a and PAM were similar in the wild-type and PAM+/− amygdala.
Electrophysiological experiments support the hypothesis that Cu is essential for synaptic plasticity in the lateral nucleus of the amygdala (Gaier et al. submitted). PAM haploinsufficiency is associated with behavioral and physiological effects similar to those caused by Cu deficiency in wild-type mice, and Cu levels were found to be lower in the amygdala of PAM+/− mice than in the amygdala of wild-type mice (Gaier et al. in preparation). To investigate the relationship between PAM and Cu handling, we examined PAM and Atp7a localization in the lateral nucleus of the amygdala under higher power magnification (Fig. 6c and d). In the wild-type lateral nucleus, Atp7a staining was present in all neurons, and especially high in a subset of neurons comprising approximately 15% of the neuronal population (Fig.6C; arrows). PAM showed a similar staining pattern, with intense staining in the same subset of neurons. We showed previously that the subset of neurons with especially high PAM levels was GABAergic (Gaier et al. 2010). Therefore, Atp7a expression is highest in the GABAergic interneurons in the lateral nucleus of the amygdala. Both PAM and Atp7a staining were most intense in the somata of GABAergic interneurons (Fig.6c) (Niciu et al. 2006). In contrast, in pyramidal neurons there was clearly more Atp7a than PAM staining in the somata (Fig. 6c; arrow heads). The Atp7a and PAM staining patterns in the PAM+/− lateral nucleus were similar to those observed in wild-type mice, with especially intense staining in GABAergic interneurons (Fig. 6d; arrows) (Gaier et al. 2010). Highest expression of both Atp7a and PAM was observed in GABAergic interneurons, independent of Pam genotype.
There was substantially more Atp7a and PAM staining in the basolateral nucleus than in the lateral nucleus. This difference was evident in staining in neuronal somata and the neuropil for both proteins in mice of both genotypes (Fig. 6e and f). Pyramidal neurons had higher intensities of Atp7a staining than PAM staining. The basolateral nucleus contained larger neurons with high intensity staining for Atp7a and PAM in both genotypes, consistent with higher levels of GABAergic innervation in this area (Gaier et al. 2010). Within neurons, staining for both proteins was concentrated in the somata (insets). There were no differences in Atp7a or PAM staining patterns between wild-type and PAM+/− mice in this nucleus. Together, these data demonstrate a close relationship between Atp7a and PAM in amygdalar nuclei in a pattern consistent with other brain regions.
Gene expression is altered in PAM+/− mouse amygdala
Regulated intramembrane proteolysis of PAM generates a cytosolic fragment that enters the nucleus and is thought to alter gene expression (Francone et al. 2010). Expression of the mRNA encoding Atox-1, the cytosolic chaperone that delivers Cu to Atp7a (Hamza et al. 2001), is increased when PAM-1 expression is increased in a corticotrope cell line (Francone et al. 2010); consistent with these results, Pam heterozygosity reduced Atox-1 transcript levels in the mouse pituitary (Bousquet-Moore et al. 2010b). To determine whether a similar change in Atox-1 mRNA levels occurred in the amygdala of PAM+/− mice, we used quantitative polymerase chain reaction (qPCR) (Fig.7) (Prigge et al. 2000; De et al. 2007; Bousquet-Moore et al. 2010b). RNA was prepared from the bilateral amygdalae and dorsal hippocampi of individual wild-type and PAM+/− mice. There was a significant interaction between genotype and brain region for PAM mRNA (p <0.05), suggesting PAM expression in each brain region was dependent on genotype. As with all other tissues collected from PAM+/− mice, the amygdala and hippocampus contained about half the level of PAM mRNA compared to wild-type (p values < 0.05; t-test) (Fig. 7a). PAM mRNA levels in the amygdala were higher than in the hippocampus (p values < 0.05; t-test). The PAM+/− amygdala contained significantly lower levels of Atox-1 and Atp7a mRNA than wild-type amygdala (p values < 0.05; t-test) (Fig.7a); similar changes were not observed in the hippocampus. As for PAM, Atox-1 mRNA levels in the amygdala exceeded levels in the hippocampus (p values < 0.05) (Fig. 7a). Comparable levels of Atp7a mRNA were found in the amygdala and hippocampus of both genotypes. These data indicate that there are region-specific differences in the levels of transcripts encoding the machinery required for Cu delivery to the secretory pathway and suggest that delivery of Cu to the secretory pathway may be decreased in PAM+/− amygdala neurons.
In agreement with the qPCR data, we quantified Atp7a protein levels in the samples of amygdala and hippocampus that were analyzed for Cu content (Gaier et al. submitted). Lysates from individual mice were analyzed, with levels of Atp7a normalized to levels of βIII-tubulin (Fig. 7b). The level of Atp7a in the PAM+/− amygdala was 80% of the level in wild-type mice (p <0.05; t-test). The level of Atp7a in the PAM+/− hippocampus was not significantly different from the level in wild-type mice. There was significantly more Atp7a protein in the hippocampus compared to the amygdala (p <0.05) (Fig. 7b). Measurement of both Atp7a mRNA and protein levels indicated that expression of Atp7a was diminished in the amygdala of PAM+/− mice.
Next, we evaluated changes in the amygdalar and hippocampal GABAergic systems in wild-type and PAM+/− mice. The synaptic plasticity deficit in PAM+/− mice is GABA-dependent, and PAM+/− mice display enhanced GABAergic inhibition in the basolateral amygdaloid complex (Gaier et al. 2010). To elucidate a molecular basis for these changes, we quantified mRNAs encoding the predominant GABAA receptor subunit (α1), the obligatory GABAB receptor subunit (2), and both the vesicular and cytosolic forms of glutamic acid decarboxylase (GAD65 and GAD67, respectively). No genotypic differences were observed for the GABAA receptor subunit in either brain region (Fig. 7c). There were significantly higher levels of the GABAA receptor α1 subunit in the amygdala than the hippocampus (p <0.05). There was a significant interaction between genotype and brain region for GABAB receptor subunit mRNA levels (p <0.05). GABAB receptor subunit mRNA levels were reduced in the PAM+/− amygdala compared to wild-type (p values < 0.05; Bonferroni pairwise comparisons, t-test); no genotypic difference was observed in the hippocampus (Fig. 7c). There were significantly higher GABAB mRNA levels in the amygdala compared to the hippocampus only in wild-type mice, and not in PAM+/− mice (p <0.05). No genotypic differences were observed for GAD65 or GAD67 in either the amygdala or hippocampus. As with the GABA receptors, GAD65 and GAD67 mRNAs were higher in the amygdala compared to the hippocampus (p values < 0.05). Thus, PAM+/− mice exhibit a region-specific reduction in GABAB receptor expression that may reflect alterations in GABAergic inhibition in this region.
Pam heterozygosity affects other cuproenzyme function
Alterations in Cu homeostasis induced by Pam heterozygosity could affect Cu delivery to other cuproenzymes in the secretory pathway. To address this question, we assessed catecholaminergic metabolism as an indicator of DβM activity. PAM+/− mice have reduced Cu concentrations in the amygdala, but not the hippocampus. Therefore, we compared catecholamine metabolite concentrations in these brain regions in wild-type and PAM+/− mice (Fig.8). The precursor for the reaction catalyzed by DβM, dopamine, was found at approximately 40-fold higher concentrations in the amygdala than the hippocampus (p <0.05) (Fig. 8a), consistent with previous reports (Chuang et al. 2011). There was no effect of genotype on dopamine concentration. In contrast, the product of the DβM reaction, norepinephrine, was found at significantly higher concentrations in the hippocampus than the amygdala (p <0.05) (Fig. 8b). As we observed for dopamine, there was no genotype effect on norepinephrine concentration.
Catecholamines have a short half-life and are quickly metabolized. Dopamine is metabolized into 3,4-dihydroxyphenylacetic acid (DOPAC) by monoamine oxidase, and further metabolized to homovanillic acid (HVA) by catechol-O-methyltransferase. As expected, these metabolites were much more prevalent in the amygdala than in the hippocampus (p values < 0.05). Concentrations of DOPAC and HVA were similar in amygdala, and there was approximately twofold more HVA than DOPAC in the hippocampus. There was a significant interaction between genotype and brain region with respect to DOPAC concentration (p <0.05), with a twofold increase in DOPAC concentration in the amygdala (p values < 0.05; Bonferroni pairwise comparisons, t-test), but no change in the hippocampus (Fig. 8c). In contrast, there was no genotypic difference in either brain region in HVA concentration data not shown. We summed these metabolites and normalized the total to dopamine levels [(DOPAC+HVA)/dopamine] as an indicator of dopamine turnover that could result from DβM impairment (Fig. 8d). We excluded the hippocampus from this analysis because of the very low levels of dopamine and its metabolites. There was a significantly higher metabolite:dopamine ratio in the PAM+/− amygdala compared to wild-type (p <0.05; t-test). These data represent an elevation in PAM+/− dopamine metabolites that could signify DβM impairment specifically in the amygdala of these animals, which is known to be Cu deficient.
In this study, we provide evidence for localization of Cu and its handling machinery at synapses in the mouse hippocampus, cortex and amygdala. These data are quite compelling in the context of work demonstrating the essential nature of Cu in synaptic plasticity (Gaier et al. submitted) (Fig.9). The synaptic localization of Atp7a provides strong evidence for its potential role as a regulator of synaptic transmission. The close overlap between Atp7a and PAM expression patterns builds on previous experiments demonstrating the bi-directional relationship between PAM and Cu and indicates that these proteins may be involved in a common regulatory pathway for Cu homeostasis (Bousquet-Moore et al. 2010a). In this way, disruption in PAM expression by haploinsufficiency could lead to altered Cu homeostasis in affected brain regions, namely the amygdala. Such disruptions have reversible physiologic and behavioral consequences (Gaier et al. 2010) (Gaier et al. submitted). Therefore, our data expand the current fund of knowledge concerning the regulation and potential role of Cu in neuronal and synaptic function.
Atp7a and Cu are present at synapses
The majority of Atp7a staining in the hippocampus and cortex was in glutamatergic and GABAergic neurons (Fig.1). The close association of Atp7a and PAM staining is consistent with an important role for Cu in the secretory pathway. Atp7a staining that extends into the neuropil and dendrites provides visually compelling evidence that Cu may be transported to synapses. Previous studies have examined the expression pattern of Atp7a in dendrites with subcellular resolution in these brain regions (Ke et al. 2006; Niciu et al. 2006). In these studies, Atp7a staining was primarily perinuclear, and co-localized with trans-Golgi network markers. Atp7a will redistribute to more peripheral neuronal domains with exposure to Cu (Petris et al. 1996; La Fontaine and Mercer 2007) or neuronal activation (Schlief et al. 2005). However, the signaling pathways that control these effects are not well understood.
Taking a biochemical approach, we demonstrated the presence of Atp7a in synaptic membranes isolated from mouse neocortex (Figs2 and 3). Like synaptophysin and Vglut1, Atp7a was readily solubilized by Triton; very little Atp7a fractionated with the PSD complex. Immunostaining of mature neuronal cultures revealed the presence of Atp7a in close proximity to synaptic terminals, though Atp7a puncta were not always precisely coincident with PSD95 staining. These results suggest novel functions for Atp7a in post-synaptic Cu handling, which could include Cu transport into vesicles for exocytosis and/or release into the synaptic cleft. Depolarization elicits Cu secretion from isolated synaptosomes; in the synaptic cleft, Cu is thought to reach concentrations of 100 and 250 μM (Kardos et al. 1989), well above the concentration at which Cu has effects on many synaptic receptors and proteins (Gaier et al. submitted). Future experiments could examine whether in vivo correlates of Cu exposure/deficiency and neuronal stimulation influence the distribution of Atp7a within neurons and synapses.
Pam heterozygosity as a model for Cu deficiency
It is postulated that Atp7a contributes to Cu efflux to prevent Cu overload within the cell (La Fontaine and Mercer 2007). In support of this, Atp7a expression and peripheral localization are increased with Cu exposure (Greenough et al. 2004), and over-expression of Atp7a in the rodent brain results in a reduction of brain Cu levels (Ke et al. 2006). We found a reduction in Atox-1 and Atp7a mRNA that was specific to the PAM+/− amygdala (Fig. 7). PAM+/− mice have reduced Cu levels in this brain region (Gaier et al. submitted). It is not apparent from our data what drives the parallel decreases in Atox-1 and Atp7a in the PAM+/− amygdala, but these changes could contribute to the observed decrease in amygdalar Cu levels and deficient LTP (Gaier et al. 2010).
Atox-1 expression is affected directly by alterations in PAM expression in a corticotrope cell line (Francone et al. 2010). While the mechanism is unclear, the ability of PAM, Atox-1 and other Cu-binding proteins to signal to the nucleus and affect gene transcription is quite compelling (Fig.9) (Itoh et al. 2008; Rajagopal et al. 2009; Lutsenko 2010). We found the majority of the Cu in cortical tissue to be cytosolic (Fig.5), where Atox-1-bound Cu resides. Therefore, changes in Atox-1 levels could contribute to reductions in overall Cu levels in this brain region. The network of genes involved in Cu homeostasis is much more complex than previously thought.
Pam heterozygosity and amygdalar dysfunction
The amygdala is unique in its sensitivity to deficiencies in PAM with respect to Cu homeostasis. PAM+/− mice have whole brain Cu levels comparable to wild-type mice (Bousquet-Moore et al. 2010b), yet there is reduced Cu specifically in this region (Gaier et al. submitted). Moreover, altered expression of Cu handling machinery (Fig.7), and increased DOPAC/DA ratio (Fig.8) provide evidence for reduced Cu within the secretory pathway. The DOPAC/DA ratio is elevated in the plasma of Menkes patients (Goldstein et al. 2009), while a high Cu diet produces a lowered DOPAC/DA ratio (De Vries et al. 1986) and Cu catalyzes the rapid breakdown of DOPAC in a test tube (Mefford et al. 1996). These alterations in Cu homeostasis correspond to deficits in amygdalar synaptic function and amygdala-related behavioral tasks, suggesting impaired Cu homeostasis may contribute to the PAM+/− fear phenotype. The essentiality of Cu in amygdalar synaptic plasticity is a compelling potential link between these biochemical and behavioral changes. How Cu influences synaptic plasticity and relating these changes to behavior will require further investigation.
Another way in which Pam heterozygosity may affect amygdalar function is through GABAergic transmission. Cu blunts the anti-convulsant effect of intra-amygdaloid injection of diazepam in rats, suggesting Cu may mediate synaptic effects at GABAA receptors (Kardos et al. 1984). There is enhanced GABAergic signaling in the PAM+/− amygdala, and the PAM+/− anxiety-like phenotype is hypersensitive to benzodiazepines (Gaier et al. 2010). Adding GABAB receptor blockade ameliorates the synaptic plasticity deficit in the PAM+/− amygdala. Correspondingly, the PAM+/− deficit in fear-potentiated startle is rescued at higher shock intensities. These observations may represent heightened GABAergic signaling obstructing the gating of synaptic plasticity and learning. In this study, we show that expression of the obligate GABAB receptor subunit is reduced specifically in the PAM+/− amygdala. In contrast, mRNA encoding the predominant GABAA receptor and GABA biosynthetic enzyme were unaffected. GABAB receptor expression in the amygdala is largely in GABAergic interneurons (McDonald et al. 2004). Moreover, the enhanced GABAergic transmission in the PAM+/− amygdala may result from disinhibition of local interneurons through a reduction in GABAB receptor signaling. How Pam heterozygosity affects GABAB expression is less clear. Alterations in Cu homeostasis could affect GABAB receptor signaling through adenosine monophosphate kinase activity [as discussed in Gaier et al. (2013)]. This hypothesis ties altered GABA signaling to disrupted Cu homeostasis in the PAM+/− model, potentially identifying GABA as a key player through which Cu regulates synaptic plasticity in the amygdala.
A role for Cu in synaptic function
The association of Atp7a and Cu with synaptic membranes puts Cu in a key position to affect synaptic transmission. Indeed, Cu affects the function of multiple synaptic receptors and ion channels (for review see Mathie et al. 2006; Gaier et al. 2013). Channel receptors such as GABAA, AMPA, and NMDA receptors are inhibited by physiological levels of Cu. Cu also affects other synaptic proteins, including amyloid precursor protein, prion protein, and matrix metalloproteinases. Cu inhibits hippocampal synaptic plasticity (Doreulee et al. 1997; Leiva et al. 2003), but is essential for amygdalar LTP (Gaier et al. submitted). Thus, the role of Cu in synaptic function is complex and region specific. Future experiments should focus on simple models under tightly controlled conditions. Such work will provide valuable insight into the role Cu plays in brain function and behavior, and potentially uncover Cu as a target for therapeutic intervention in emotional neuropsychiatric disease states.
We thank Darlene D'Amato and Yanping Wang for tireless support in the lab. This study was supported by grants from the National Institutes of Health DK32949 (BAE, REM) and NS41224. The authors declare no conflicts of interest.
ARRIVE Guidelines: We have received these guidelines and have followed them.