The 22q11 deletion syndrome (22q11DS) is characterized by multiple physical and psychiatric abnormalities and is caused by the hemizygous deletion of a 1.5–3 Mb region of chromosome 22. It constitutes one of the strongest known genetic risks for schizophrenia; schizophrenia arises in as many as 30% of patients with 22q11DS during adolescence or early adulthood. A mouse model of 22q11DS displays an age-dependent increase in hippocampal long-term potentiation (LTP), a form of synaptic plasticity underlying learning and memory. The sarco(endo)plasmic reticulum Ca2+ ATPase (SERCA2), which is responsible for loading Ca2+ into the endoplasmic reticulum (ER), is elevated in this mouse model. The resulting increase in ER Ca2+ load leads to enhanced neurotransmitter release and increased LTP. However, the mechanism by which the 22q11 microdeletion leads to SERCA2 overexpression and LTP increase has not been determined. Screening of multiple mutant mouse lines revealed that haploinsufficiency of Dgcr8, a microRNA (miRNA) biogenesis gene in the 22q11DS disease-critical region, causes age-dependent, synaptic SERCA2 overexpression and increased LTP. We found that miR-25 and miR-185, regulators of SERCA2, are depleted in mouse models of 22q11DS. Restoration of these miRNAs to presynaptic neurons rescues LTP in Dgcr8+/− mice. Finally, we show that SERCA2 is elevated in the brains of patients with schizophrenia, providing a link between mouse model findings and the human disease. We conclude that miRNA-dependent SERCA2 dysregulation is a pathogenic event in 22q11DS and schizophrenia.
Schizophrenia affects ∼1% of the world's population and is characterized by symptoms that include hallucinations and delusions (positive symptoms), antisocial behavior and blunted emotions (negative symptoms), and deficits in working memory, executive function, and learning and memory (cognitive symptoms). Mechanisms of schizophrenia are poorly understood, in part because no single gene mutation is associated with development of the disease. Consequently, research in genetic models of schizophrenia has been difficult to interpret (Nestler and Hyman, 2010).
One well known genetic predictor of schizophrenia is the 22q11 deletion syndrome (22q11DS). This syndrome is caused by the hemizygous deletion of a 1.5–3-megabase region of the q arm of chromosome 22, resulting in the haploinsufficiency of 30–40 genes (Scambler et al., 1992; Burn et al., 1993; Ryan et al., 1997; Scambler, 2000; Oskarsdóttir et al., 2004). Schizophrenia develops in ∼30% of patients with 22q11DS during adolescence or early adulthood (Pulver et al., 1994; Bassett et al., 2005; Chow et al., 2006). Symptoms of 22q11DS-related schizophrenia are indistinguishable from those of the idiopathic disease (Pulver et al., 1994; Murphy et al., 1999; Chow et al., 2006), suggesting that lessons learned from deletion-related forms of schizophrenia may also shed light on the mechanisms of schizophrenia in general.
Cognitive deficits are central to schizophrenia and are among its least treatable symptoms (Green, 1996; Green et al., 2000; Gold, 2004). Cognitive deficits have been linked to the hippocampus (Heckers et al., 1998; Weinberger, 1999; Tamminga et al., 2010), a brain region central to learning and memory. Synaptic plasticity at excitatory synapses is a mechanism of hippocampus-related learning and memory (Milner et al., 1998; Martin et al., 2000) that provides an excellent means to probe cellular events related to cognition in animal models of schizophrenia.
The 22q11DS-critical region is largely conserved on mouse chromosome 16, allowing for the generation of 22q11DS mouse models. Df(16)1/+ mice carry a hemizygous deletion of 23 genes in the syntenic region of chromosome 16 (Lindsay et al., 1999) and develop a spatial memory deficit and enhanced synaptic plasticity in the form of long-term potentiation (LTP) by 16 weeks (Earls et al., 2010). This age-dependent alteration is caused by an aberrant increase in the level of sarco(endo)plasmic reticulum ATPase (SERCA2), which maintains Ca2+ levels in the endoplasmic reticulum (ER). SERCA2 upregulation increases LTP by enhancing Ca2+ entry into the presynaptic cytoplasm and releasing excessive neurotransmitter during synaptic plasticity induction (Earls et al., 2010). Age-dependent synaptic abnormalities in Df(16)1/+ mice may affect the cognitive decline observed at the onset of schizophrenia. Identification of the culprit genes within the 22q11DS-critical region that cause these abnormalities may provide new insight into the disease's pathophysiology.
Here we used a panel of mutant mice carrying hemizygous deletions of genes within the 22q11DS-critical region to screen for genes involved in the age-dependent increase in LTP. This screen identified Dgcr8 deficiency as a contributor to synaptic abnormalities. The hemizygous loss of Dgcr8 causes an age-dependent increase in LTP that depends on the synaptic upregulation of SERCA2. We also provide evidence that SERCA2 is upregulated in postmortem brain samples from patients with schizophrenia. We propose that the molecular events described in these mouse models of 22q11DS may be relevant to the human disease.
Materials and Methods
Young (8–10 weeks) and mature (16–20 weeks) mice of both sexes were used for all experiments except the microarray experiments, for which only male mice were used. Dgcr8+/− mice were generated from the XH157 ES cell line (Bay Genomics) as previously described (Stark et al., 2008; Schofield et al., 2011). Zdhhc8+/− mice were generated from the IST14452C2 ES cell line (TIGM), which contains a gene-trap insertion downstream of the first exon. To expand the colony, Dgcr8+/− and Zdhhc8+/− mice harboring the disrupted alleles were bred to C57BL/6J mice in our animal-housing facility. Production and genotyping of Df(16)1/+ (Lindsay et al., 1999), Df(16)2/+ (Lindsay et al., 2001), Znf74l-Ctp/+(Kimber et al., 1999), Prodh+/−(Gogos et al., 1999), Comt+/−(Gogos et al., 1998), and Rtn4r+/−(Kim et al., 2004) mice have been previously described. All mouse strains in this study were back-crossed onto the C57BL/6J genetic background for at least six generations. The care and use of animals was reviewed and approved by the St. Jude Children's Research Hospital Institutional Animal Care and Use Committee.
Acute transverse hippocampal slices (400 μm) were prepared as previously described (Earls et al., 2010). Briefly, mouse brains were quickly removed and placed in cold (4°C) dissecting artificial CSF (ACSF) containing the following (in mm): 125 choline-Cl, 2.5 KCl, 0.4 CaCl2, 6 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, and 20 glucose (285–295 mOsm), under 95% O2/5% CO2. After dissection, slices were incubated for 1 h in ACSF containing the following: (in mm): 125 NaCl, 2.5 KCl, 2 CaCl2, 2 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose (285–295 mOsm), under 95% O2/5% CO2 at room temperature, and then transferred into submerged recording chambers and superfused (2–3 ml/min) with warm (30−32°C) ACSF. Field recordings were performed using a setup with eight submerged recording chambers (Campden Instruments). Recordings in each chamber were performed independently. Field EPSPs (fEPSPs) from the CA1 stratum radiatum were recorded by using an extracellular glass pipette (3–5 MΩ) filled with ACSF. Schaffer collateral fibers in the stratum radiatum were stimulated with a bipolar tungsten electrode placed 200–300 μm away from the recording pipette. Stimulation intensities were chosen to produce an fEPSP with a 0.5 V/s slope. LTP was induced by three periods of 200 Hz tetanization delivered every 5 min. Each period of tetanization consisted of 10 trains of 200 Hz stimulation delivered at the same intensity for 200 ms (40 stimulations) every 5 s.
Western blots were performed as previously described (Earls et al., 2010). For human brain tissue, dissected samples arrived from brain banks frozen on dry ice. Mouse tissue was dissected at 4°C and prepared either as whole-tissue lysates or as crude P2 synaptosomal fractions. Synaptosomes were prepared as described previously (Gray and Whittaker, 1962). In brief, tissue was homogenized in 10 mm HEPES (pH 7.4)/0.32 m sucrose using a motorized glass-Teflon homogenizer. To separate the P2 synaptosomal fraction, the homogenate was spun for 5 min at 800 × g; the supernatant from three successive spins was then centrifuged for 20 min at 12,000 × g. Tissue or synaptosomal pellets were lysed by freezing and thawing, subsequent syringe passage in ice-cold radioimmunoprecipitation assay buffer (50 mm Tris-HCl [pH 7.4], 1% NP-40, 0.25% sodium deoxycholate, 150 mm NaCl, 1 mm EDTA, and protease inhibitor cocktail tablets (Roche), and finally brief sonication. Concentrations of protein lysates were determined by the BCA assay (Thermo Scientific). A 25 μg sample of each protein extract was electrophoresed on a 10% SDS-PAGE gel, and protein was transferred onto polyvinylidene difluoride membranes (Invitrogen). The primary antibodies used were goat anti-SERCA2 (1:250, sc-8095; Santa Cruz Biotechnology) and mouse anti-β-actin (1:10,000, A5316; Sigma-Aldrich). SERCA and β-actin Western blots were probed with anti-mouse (1:5000) and anti-goat (1:5000) secondary antibodies conjugated to IR dye 680 or 800 (LI-COR Biosciences). These blots were imaged and quantified using the Odyssey infrared imaging system (LI-COR Biosciences).
Total RNA was isolated from 4-month-old male wild-type (WT) and Df(16)1/+ hippocampi using the miRVana RNA isolation kit (Ambion). A mouse miRNA microarray (Agilent-029298; Agilent Technologies) consisting of probes for 690 mouse miRNAs from the Sanger miRBase (release 14.0) was designed and manufactured by Agilent Technologies. Array hybridization was performed according to the manufacturer's recommended protocols. In brief, total RNA was labeled using the Agilent miRNA labeling kit. Hybridization was performed in an Agilent oven at 55°C for 20 h at 20 rpm, followed by standard wash procedures. The microarray was then scanned in an Agilent scanner at 5 μm resolution, and the array data were extracted using the default miRNA settings of Agilent Feature Extraction Software (v10.5.1.1) with the miR_105_Jan09 protocol. Signal intensity was normalized at the 80th percentile, scaled across the sample set, and log2 transformed. The minimum detecting miRNA expression signal was set at a threshold greater than the 99th percentile of those from the negative-control probes. A p value was calculated using a t test of samples from two different experimental conditions.
Quantitative real-time PCR.
Total RNA (1 μg) was polyadenylated and reverse transcribed using an oligo-dT primer with an attached universal sequence tag according to the miRNA First-Strand cDNA Synthesis kit (Stratagene/Agilent). The qPCR was then performed using SYBR green (Applied Biosystems), a forward primer specific to the miRNA of interest: mmu-miR185 (tggagagaaaggcagttcct), mmu-miR-299* (tggtttaccgtcccac), mmu-miR-337–3p (ttcagctcctatatgatg), mmu-miR-411 (tagtagaccgtatagcgta), mmu-miR-411* (tatgtaacacggtccacta), mmu-miR-874 (ctggcccgagggacc), mmu-miR-374 (atataatacaacctgctaag), mmu-miR-379 (tggtagactatggaacgta), mmu-miR-337–5p (gaacggcgtcatgcaggag), mmu-miR-329 (aacacacccagctaaccttt), mmu-miR-674* (cacagctcccatctcagaac), mmu-miR-323–3p (tacagttgttcaaccagtta), mmu-miR-582–5p (tacagttgttcaaccagtta), mmu-miR-98 (tgaggtagtaagttgtattg), mmu-miR-672 (tgaggttggtgtactgtgtgt), mmu-miR-421 (atcaacagacattaattgggc), mmu-miR-409–5p (aggttacccgagcaactttgc), mmu-miR-872 (aaggttacttgttagttca), mmu-miR-532–3p (cctcccacacccaaggcttg), mmu-miR-425 (aatgacacgatcactcccgtt), mmu-miR-25 (cattgcacttgtctcggtct), U6snRNA_Forward (cgcttcggcagcacatatac), U6snRNA_Reverse(ttcacgaatttgcgtgtcat), and a universal reverse primer specific to the sequence tag (miRNA First-Strand cDNA Synthesis kit, Stratagene/Agilent). qPCR was performed in an Applied Biosystems 7900HT Fast Real-Time PCR System using the following cycling parameters: 55°C (2 min), 95°C (10 min), 40 cycles of 95°C (15 s), 60°C (30 s), and 72°C (20 s). miRNA concentrations were calculated using cycle threshold values and a standard curve made from serial dilutions of WT cDNA samples (dilutions: 1:3, 1:9, 1:27, 1:81, 1:243, 0). The control, U6 snRNA [primers used from (Thomson et al., 2006)], showed no difference between WT and Df(16)1/+ samples. Therefore, all miRNA concentrations were normalized to the U6 level for a given animal. All PCR experiments were conducted in duplicate. PCR products were run on a 5% polyacrylamide gel to confirm the presence of a single band of the expected size, and PCR products were cloned and sequenced to confirm the identity of each miRNA. Mature forms of miRNAs produced a single band ∼80 bp in size in agreement with estimations (22 bp miRNA + polyA tail of varying size + 32 bp universal tag sequence) (data not shown).
In vivo viral injections.
Adeno-associated viruses (AAVs) were made by cloning chimeric hairpins of the miRNAs of interest with hsa-miR-30a into the pAAV-6P-SEWB vector, as previously described (Christensen et al., 2010). The following primers were used: ChimericmiR185up1 (gtacagctgttgacagtgagcgactggagagaaaggcagttcctgatgtgaa), ChimericmiR185up2 (gccacagatggtcaggaactctttctctccagctgcctactgcctcggaa), ChimericmiR185low1 (ccatctgtggcttcacatcaggaactgcctttctctccagtcgctcactgtcaacagct), Chimer icmiR185low2 (agctttccgaggcagtaggcagctggagagaaagagttcctga), ChimericmiR25up1 (gtacagctgttgacagtgagcgacaggcggagacttgggcaattgctgtgaa), ChimericmiR25up2 (gccacagatgggcaattgccagtctccgcctgctgcctactgcctcggaa), ChimericmiR25low1(ccatctgtggcttcacagcaattgcccaagtctccgcctgtcgctcactgtcaacagct), and ChimericmiR25low2 (agctttccgaggcagtaggcagcaggcggagactggcaattgc).
Viruses were prepared by cotransfecting each plasmid into HEK293 cells with both pDp1 and pDp2 helper plasmids. AAVs were then harvested and purified as previously described (Zolotukhin et al., 1999). Titers were in the range of 1011 TU/ml. Anesthesia in mice (10 weeks of age) was induced with 2–2.5% isoflurane (in 100% O2). Anesthesia was maintained during surgery using 1.5% isoflurane. A 1/32-G cannula was inserted into the brain to deliver 2 μl AAV-miRNA or AAV-empty in three locations centered on the CA3 region of the hippocampus (AP: –1.8 mm; L: ±2.4 mm; V: –2.0 mm). AAV-miRNA was injected into one hemisphere, and green fluorescent protein (GFP)-only control AAV was injected into the contralateral hemisphere. Following AAV injections, incisions were sutured, and mice were allowed to recover and were then returned to holding cages. Electrophysiological and imaging experiments in hippocampal slices from these mice were performed 6–8 weeks after AAV injections.
Spatial memory testing
We tested spatial memory in the Morris water maze as previously described (Earls et al., 2010). Briefly, animals learned to find a platform hidden under water clouded with nontoxic, water-based paint. The platform was located in the same “training quadrant” of the pool, and mice learned to locate its position using the standard version of the Morris water maze spatial learning task for 10 successive days. A spatial memory probe trial was administered on the day after the completion of spatial learning. With the platform removed, animals received a single 1 min trial in which they tried to find the escape platform in the training quadrant. This trial started from the point that was the farthest from the platform's location on the previous training day. The overall path length was measured for each mouse, and the relative path length for each quadrant was calculated.
Human brain tissue.
Postmortem samples of human prefrontal cortex were obtained from the Alzheimer Disease and Schizophrenia Brain Bank (Mount Sinai School of Medicine, New York, NY). Samples from 17 patients with schizophrenia and 22 unaffected controls were used. Postmortem samples of human hippocampus were obtained from the Harvard Brain Tissue Resource Center (McLean Hospital, Belmont, MA). Samples from five patients with schizophrenia and six age-matched unaffected controls (all males) were used.
Drugs were purchased from Sigma-Aldrich, except thapsigargin, which was purchased from Tocris Bioscience.
All data are represented as mean ± SEM. Statistical analyses for all experiments were performed using nonparametric Mann–Whitney rank sum, Wilcoxon signed rank, or t tests measured in Sigma Stat (Systat Software). Kruskal–Wallis one-way ANOVA by ranks was used to compare >2 independent groups.
Age-dependent LTP increase in Dgcr8+/− mice
In the Df(16)1/+ mouse model of 22q11DS, LTP of synaptic transmission measured at excitatory hippocampal synapses between CA3 and CA1 pyramidal neurons (CA3–CA1 synapses) increases in mature (16–20 weeks) but not young (8–10 weeks) animals (Earls et al., 2010). LTP was >200% higher in mature Df(16)1/+ mice than WT littermates, whereas basal synaptic transmission at CA3–CA1 synapses in Df(16)1/+ mice was normal.
To narrow the location of genes involved in the LTP phenotype, we took advantage of mice that carry smaller hemizygous subdeletions within the Df(16)1 region (Kimber et al., 1999; Lindsay et al., 2001). We tested basal synaptic transmission and LTP of fEPSPs in acute hippocampal slices from Df(16)2/+ mice, which carry a hemizygous deletion of genes at the proximal end of the Df(16)1 region (Es2el-Trxr2) (Fig. 1A). As in Df(16)1/+ mice, LTP was elevated in Df(16)2/+ animals in an age-dependent manner (Fig. 1B,C). In young animals, the increase in fEPSP slope (measure of LTP) quantified 6 h after tetanization of Schaffer collaterals was not significantly different between Df(16)2/+ and WT littermates (Df(16)2/+: 20 slices, 4 mice; WT: 23 slices, 5 mice; p = 0.238). However, in mature Df(16)2/+ mice, LTP was higher than in WT littermates (Df(16)2/+: 22 slices, 6 mice; WT: 27 slices, 7 mice; p = 0.006) (Fig. 1C). On average, LTP in Df(16)2/+ mice was ∼100% higher than that in WT littermates.
As in Df(16)1/+ mice, basal synaptic transmission was normal in both young and mature Df(16)2/+ mice (Fig. 1D,E). Thus, the input–output relationship between fEPSPs and stimulation intensity was comparable in young and mature WT and Df(16)2/+ mice (young Df(16)2/+: 20 slices, 4 mice; young WT: 23 slices, 5 mice, p > 0.05; mature Df(16)2/+:22 slices, 6 mice; mature WT: 27 slices, 7 mice, p > 0.05).
Changes in LTP are often associated with changes in learning and memory. The substantial increase in LTP at CA3–CA1 synapses in mature Df(16)1/+ mice is accompanied by a mild deficit in hippocampus-dependent spatial memory, as measured in the Morris water maze task (Earls et al., 2010). In mature Df(16)2/+ mice, however, spatial memory tested in the Morris water maze task was normal (data not shown, 27 Df(16)2/+ and 26 WT mice, p = 0.875), indicating that additional genetic factors are required for the full effect on learning and memory (see Discussion).
To identify the culprit gene within the Df(16)2 region, we further narrowed the LTP-critical region by measuring LTP in Znf74l-Ctp/+mice (Kimber et al., 1999), which carry a hemizygous, 150 kb subdeletion that includes the three most proximal Df(16)2 genes Es2el, Gscl, and Ctp (Fig. 2A). LTP was normal in mature Znf74l-Ctp/+mice compared with WT littermates (Fig. 2D), indicating that Es2el, Gscl, and Ctp do not contribute to the phenotype observed in Df(16)2/+ mice. Among the remaining genes in the Df(16)2 region, five have been previously implicated in the pathogenesis of 22q11DS and schizophrenia. To test whether these genes affect age-dependent LTP, we used mice deficient for these genes. These included Prodh+/−(Gogos et al., 1999), Rtn4r+/−(Kim et al., 2004), and Comt+/−(Gogos et al., 1998) mice and Zdhhc8+/− and Dgcr8+/−mice (Fig. 2A), which were generated during the course of this study. Zdhhc8+/− and Dgcr8+/− mice had significant reductions of the targeted genes, verifying haploinsufficiency of these genes in the heterozygous mutants (Figures 2B,C). A comparison of LTP between mutants and WT littermates from each mutant line at 16 weeks of age revealed an LTP increase only in Dgcr8+/− mice (Figures 2D, 3). Similar to Df(16)2/+ mice, the LTP increase in Dgcr8+/− mutants was age dependent; there was no increase in LTP in WT littermates in 8- to 10-week-old animals (Dgcr8+/–: 31 slices, 7 mice; WT: 39 slices, 7 mice, p = 0.850) (Figure 3A). However, mature Dgcr8+/− mice had increased LTP, which was equivalent to that of Df(16)2/+ mutants (Dgcr8+/−: 29 slices, 6 mice; WT: 32 slices, 7 mice, p < 0.001) (Figs. 2D, 3B). Input–output coupling did not differ between young (data not shown) and mature Dgcr8+/− and WT littermates, indicating that the hemizygous deletion of Dgcr8 does not affect basal synaptic transmission (Fig. 3C). These results implicate Dgcr8 as the gene from the Df(16)2 region that is responsible for the observed abnormalities in synaptic plasticity.
Age-dependent upregulation of SERCA2 in Dgcr8+/− hippocampus
We previously showed that age-dependent overexpression of SERCA2 is crucial for the LTP increase observed in Df(16)1/+ mice (Earls et al., 2010). SERCA2 is increased in the hippocampus of mature but not young Df(16)1/+ mice, and SERCA inhibitors rescue the LTP increase in mature Df(16)1/+ mice (Earls et al., 2010). Therefore, we tested SERCA2 protein levels in the hippocampus of young and mature Dgcr8+/− mice. Although we found no difference in SERCA2 levels between Dgcr8+/− and WT whole-hippocampal extracts (data not shown), SERCA2 protein levels in synaptosomal preparations from the hippocampus of mature Dgcr8+/− mice were significantly elevated compared with that in WT littermates (p < 0.001) (Fig. 4A). In contrast, this synaptic increase in SERCA2 levels was not present in younger mice, indicating a correlation between SERCA2 elevation and the LTP increase in Dgcr8+/− mice (Fig. 4A).
To test whether this increase in SERCA2 is necessary for enhanced LTP in Dgcr8+/− mice, we measured LTP in the presence of the SERCA inhibitor thapsigargin (4 μm). SERCA inhibition rescued the LTP increase in Dgcr8+/− slices to WT levels (Fig. 4B). In the absence of thapsigargin, the fEPSP slope increase measured 6 h after induction was ∼120% stronger than in the presence of thapsigargin in slices from Dgcr8+/− mice (8 mice: 21 slices vehicle, 21 slices thapsigargin; p = 0.014). Furthermore, in the presence of thapsigargin, LTP in Dgcr8+/−mice did not significantly differ from that in WT animals, indicating full rescue (p = 0.854). Similar to previously shown results (Earls et al., 2010), thapsigargin did not affect LTP measured in slices from WT littermates (8 mice: 24 slices vehicle, 19 slices thapsigargin; p = 0.974). These results suggest that SERCA2 is necessary for the observed LTP increase in the mature Dgcr8+/− hippocampus.
Identification of miRNAs responsible for enhanced LTP in mouse models of 22q11DS
Dgcr8 is an miRNA biogenesis gene, and miRNAs typically act as negative regulators of protein translation. DGCR8 binds to primary miRNA transcript hairpins and recruits the nuclease DROSHA, which cleaves the hairpins. Further processing produces mature miRNAs that bind to complementary seed sites in the 3′ untranslated regions (3′UTRs) of target mRNA transcripts and negatively regulate protein translation through recruitment of the RNA-induced silencing complex (Bartel, 2009). We therefore hypothesized that SERCA2 is upregulated in 22q11DS brains due to the age-dependent loss of its regulatory miRNAs. To identify potentially responsible miRNAs, we performed a microarray comparison of hippocampal miRNAs between Df(16)1/+ and WT littermates (seven mice per genotype) at 16 weeks, the age of onset of the LTP and SERCA2 increase. No miRNAs were elevated, but 50 miRNAs were significantly reduced at this stage in the Df(16)1/+ mutants (Fig. 5A; Table 1). We verified the depletion of mature forms of 20 of these miRNAs in Df(16)1/+ hippocampus by qPCR (Fig. 5B). For a subset of these miRNAs, we also verified their depletion in the Dgcr8+/−hippocampus at 16 weeks, and found that the same miRNAs were depleted in both animal models (Fig. 5C).
miRNAs affect their target mRNAs by binding to complementary seed sites within the 3′UTR and recruiting the RNA-induced silencing complex to the transcript to prevent translation (Ambros, 2004). We therefore used miRNA target prediction algorithms to narrow down the list of candidate miRNAs found from the microarray to only those predicted to target the Serca2 transcript. Using the algorithms miRBase (Enright et al., 2003), TargetScan (Lewis et al., 2005), DIANA-microT v3.0 (Kiriakidou et al., 2004), and miRDB (Wang and El Naqa, 2008), we identified potential seed sites for miRNAs within the 3′UTR of the murine Serca2 transcript (Table 2). Of the miRNAs depleted in 22q11DS mice, three were predicted to target the SERCA2 3′UTR: miR-25, −98, and −185. Depletion of mature forms of miR-25 and miR-185 in Df(16)1/+ mice was verified by qPCR (Fig. 5B). However, depletion of mature miR-98 in Df(16)1/+ mice could not be verified due to the AT-rich nature of this miRNA. We nonetheless hypothesize that the depletion of any of these miRNAs contributes to SERCA2 upregulation and abnormal LTP in 22q11DS mouse models. A diagram of the murine SERCA2 3′UTR with the locations of predicted seed sites for these three miRNAs is shown in Figure 6A.
Rescue of the LTP increase in Dgcr8+/− mice by presynaptic restoration of miR-25 or miR-185
To test whether depletion of the miRNAs of interest is required for the observed LTP increase in 22q11DS models, we attempted to restore miR-25 or miR-185 to hippocampal neurons of Dgcr8+/− mice to rescue the LTP phenotype. To do this, we generated recombinant AAVs that encode GFP just upstream of either miR-25 or miR-185 under the control of the neuron-specific Synapsin promoter to infect adult neurons in vivo according to a previously described method (Christensen et al., 2010). Because the increase in LTP measured at CA3–CA1 synapses in mature Df(16)1/+ hippocampus is caused by presynaptic abnormalities (Earls et al., 2010), we injected virus into the CA3 region of the hippocampus in vivo. Injections were performed at 10 weeks of age, before the onset of LTP abnormalities. We injected AAV expressing a given miRNA into one hemisphere and an empty AAV (expressing only GFP) into the contralateral hemisphere as a control. We then measured LTP in both miRNA-injected and control-injected hippocampi at 16 weeks. AAV-driven GFP expression was robust in presynaptic CA3 neurons, but GFP was absent from postsynaptic CA1 neurons (Fig. 6B), suggesting successful targeting of presynaptic CA3 neurons. Using double staining with the nuclear fluorescent dye DAPI (a measure of overall number of cells) and GFP (a measure of infected neurons) we estimated the infection rate of neurons by AAVs in the pyramidal layer of the CA3 area to be ∼88% (10 slices) (Fig. 6B). Using qPCR, we verified that injection of AAV-miRNAs at 10 weeks of age resulted in elevation of respective miRNAs at 16 weeks (Fig. 6C). AAV-driven overexpression of these miRNAs appeared modest, mainly because we measured miRNA levels throughout the hippocampus after injecting our rescue viruses only in the CA3 region. Therefore, the levels of miRNAs measured in these experiments are an underestimation.
Further, to verify that miR-25 and miR-185 target the Serca2 transcript and thus affect SERCA2 protein levels, we injected these viruses into the CA3 region of 4 WT animals at 10 weeks of age. We injected GFP-only vector on the contralateral side as a control. At 16 weeks of age, we harvested hippocampi and used quantitative Western blotting to compare SERCA2 levels. As shown in Figure 6D, SERCA2 decreased with overexpression of either miR-25 or miR-185. As with the qPCR results, this is an underestimation of the effect of these miRNAs on SERCA2, because the injection was specific to CA3 while Western blotting was performed in the entire hippocampus.
Importantly, presynaptic expression of either miR-25 or miR-185 was sufficient to rescue the increased LTP in Dgcr8+/− hippocampus (Fig. 6E). These results indicate that presynaptic depletion of SERCA2-targeting miRNAs causes the LTP increase in 22q11DS mouse models, and restoration of any of these miRNAs is sufficient to rescue this abnormality in synaptic plasticity.
Elevation of SERCA2 in postmortem brain tissue from patients with schizophrenia
In this work and previously, we showed that SERCA2 is elevated in the hippocampus of the Df(16)1/+ mouse model of 22q11DS (Earls et al., 2010). However, the hippocampus is clearly not the only brain region involved in psychiatric disease associated with the deletion. To determine the specificity of this upregulation, we tested SERCA2 levels in various tissues of Df(16)1/+mice. SERCA2 was elevated in all tested brain regions of Df(16)1/+ mice, including the cortex (125.2 ± 5.6% of the WT level, p = 0.012; 4 mice per genotype) and cerebellum (124.6 ± 7.5% of the WT level, p = 0.048; 4 mice per genotype), but not in non-neural tissues such as liver (96.9 ± 4.7% of the WT level, p = 0.689; 3–4 mice per genotype). These findings indicate that changes in SERCA2 expression are brain specific and found throughout the brain.
Because the elevation of SERCA2 has serious consequences for neural function, we questioned whether the molecular findings in these 22q11DS mouse models might translate to human disease. We compared SERCA2 levels in postmortem tissue samples from the hippocampus and prefrontal cortex of patients with schizophrenia and unaffected controls. This comparison revealed a significant increase in SERCA2 levels in both brain regions of patients with schizophrenia (Fig. 7A). Given the severe phenotype associated with SERCA2 elevation in mouse neurons, this finding suggests that elevation in SERCA2 protein may contribute to neural deficits in schizophrenia. The seed sites for hsa-miR-25, hsa-miR-98, and hsa-miR-185 are conserved in the 3′UTR of human Serca2B (Fig. 7B), suggesting that modulation by these miRNAs is a potential mechanism for the observed SERCA2 protein overexpression in schizophrenia.
Schizophrenia is a devastating disease that severely affects cognitive function and currently has no cure. Here we report three new findings that expand our understanding of 22q11DS, which poses a major risk for schizophrenia. First, we showed that 22q11DS and schizophrenia are associated with elevated SERCA2 levels in the brain. SERCA2 was elevated in the brains of mouse models and patients with schizophrenia. Given previous work showing that elevated SERCA2 substantially affects synaptic function (Earls et al., 2010), altered SERCA2 levels may be involved in schizophrenia symptoms. Second, we determined that Dgcr8, a miRNA-processing gene, is the 22q11 gene responsible for SERCA2 elevation in the mouse. Third, we showed that the LTP increase in mice is caused by the loss of miRNAs that target SERCA2 (miR-25 and miR-185). This finding was verified by the fact that restoration of depleted miRNAs rescues LTP abnormalities in Dgcr8+/−mice. Furthermore, since presynaptic injection of miRNAs was sufficient to rescue the LTP deficit, this validates previous findings that functional alterations at this synapse are presynaptic in nature. Our model of the pathway for miRNA modulation of SERCA2 and synaptic plasticity (Fig. 8) outlines potential novel targets for therapeutic approaches to treat cognitive deficits in 22q11DS and schizophrenia.
The hemizygous loss of a group of genes causes 22q11DS, which is characterized by physical and cognitive abnormalities. Identifying the genes responsible for each category of 22q11DS symptoms is crucial for understanding the pathogenesis of the disease. Past work has shown that haploinsufficiency of the 22q11.2 gene Tbx1 causes some physical abnormalities in 22q11DS (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001). However, the genes involved in the cognitive deficits have remained elusive, perhaps because behavioral assays reveal only mild deficits in 22q11DS mouse models and therefore are not amenable to nonbiased screening. In contrast, the substantial age-dependent increase in LTP observed in Df(16)1/+ mice provides a robust assay for identifying culprit genes that contribute to cognitive symptoms in 22q11DS (Earls et al., 2010).
Using Df(16)2/+ and Znf74l-Ctp/+ subdeletion mice, we narrowed the causal region of age-dependent LTP increase to the Trxr2-Dgcr6 genomic region. Further screening of multiple strains deficient in individual genes within this region revealed that hemizygous deletion of Dgcr8 recapitulates the Df(16)2/+ phenotype. The magnitude of the LTP increase was comparable between mature Df(16)2/+ and Dgcr8+/− mice, suggesting that Dgcr8 is the sole contributor to the Df(16)2/+phenotype.
Because DGCR8 controls miRNA biosynthesis, we hypothesized that neural deficits observed in our 22q11DS mouse models are caused by the age-dependent depletion of miRNAs. The involvement of miRNAs in age-related alterations in plasticity is an attractive model for several reasons. First, miRNAs are developmentally expressed (Miska et al., 2004), and the loss of miRNAs during a critical period may explain the age dependence of this phenotype. Second, we previously showed that SERCA2 is upregulated at the protein level but not at the transcript level in Df(16)1/+ mice (Earls et al., 2010). miRNAs typically affect protein levels of their targets independently of transcript levels, which is consistent with the modulation of SERCA2 observed in our mouse model. Finally, although the changes seen in SERCA2 levels were significant, they were modest. This is also consistent with modulation by miRNAs, which fine-tune protein expression levels rather than result in robust expression changes (Baek et al., 2008).
Using microarray analysis, we identified miRNAs that were depleted in Df(16)1/+ mice at the age of the LTP phenotype onset. Only seven (14%) were common with those identified in a previous microarray analysis of young (8- to 10-week-old) 22q11DS mouse models (Stark et al., 2008). This speaks to the dynamic nature of miRNA expression during development. Individual primary miRNA transcripts may be sensitive to Dgcr8 loss at different developmental stages. We identified three miRNAs (miR-25, −98, and −185) that are predicted to target the 3′UTR of Serca2. We confirmed depletion of miR-25 and miR-185 in mouse models of 22q11DS by qPCR. We were unable to validate depletion of miR-98 due to the AT-rich nature of this miRNA. However, all miRNAs tested by qPCR produced similar results to the microarray, suggesting that the microarray list, in general, is valid and that miR-98 is a potential contributor to the phenotypes discussed here. Restoring either miR-25 or miR-185 levels in presynaptic neurons rescued the LTP phenotype in Dgcr8+/− mice, indicating that reintroducing a single SERCA2-targeting miRNA in presynaptic neurons is sufficient to return LTP to its normal level. Previous work has shown that miR-25, miR-98, and miR-185 are present or enriched in synapses (Lugli et al., 2008; Smalheiser, 2008; Siegel et al., 2009). This may explain why the increase in SERCA2 was detected in synapses in the mature Dgcr8+/−hippocampus.
miRNA-185 may be particularly important for the pathology of 22q11DS, because the genomic region encoding miR-185 is contained within an intron of the 22q11.2 gene T10. Here, we show that miR-185 is also modulated by Dgcr8 loss, as miR-185 is significantly reduced in Dgcr8+/− mice. Thus, in 22q11DS, miR-185 is downregulated at both the genomic and transcript levels. Additionally, the 3′UTR of the human SERCA2 transcript contains four predicted seed sites for miR-185 not found in the mouse transcript. Together, these findings suggest that miR-185 regulates cognitive and psychiatric symptoms of patients with 22q11DS.
Because 22q11DS is responsible for only a subset of schizophrenia cases (International Schizophrenia Consortium, 2008; Stefansson et al., 2008; Xu et al., 2008), we sought to determine how the above pathway is affected in schizophrenia that is not associated with the deletion. Surprisingly, we found that the increase in SERCA2 levels shown in mouse models of 22q11DS is also present in patients with schizophrenia. Because miRNA loss leads to SERCA2 elevation in mouse models of 22q11DS, we hypothesized that miRNA regulation of SERCA2 translation could also mediate the observed elevation of SERCA2 protein in patients with schizophrenia. Several groups have attempted to quantify miRNA levels in postmortem schizophrenia brains, but the results have shown little overlap (Perkins et al., 2007; Beveridge et al., 2010; Kim et al., 2010; Moreau et al., 2011). These differences are most likely due to different sampling of miRNAs in various studies, but variations in postmortem intervals and tissue-handling methods may also contribute to these differences. As we were unable to obtain quality RNA from the postmortem samples in this study, we could not reliably measure miRNA levels. Future studies will seek to determine the correlation between the levels of miRNAs and SERCA2 in human brain samples collected after shorter postmortem intervals.
SERCA2 upregulation in patients with schizophrenia may also be caused by miRNA-independent mechanisms. The SERCA2 pump, as a part of the Ca2+ homeostasis machinery, is very tightly regulated (Vandecaetsbeek et al., 2011). This control is essential, especially in presynaptic terminals of central synapses, where the relationship between cytoplasmic Ca2+ concentration and neurotransmitter release is described by a fourth-power relationship (Mintz et al., 1995; Borst and Sakmann, 1996). Thus, slight changes in SERCA2 levels can have sizable consequences on presynaptic function, synaptic plasticity, and cognition. SERCA2 upregulation in the 22q11DS model mice causes an aberrant Ca2+ increase in the presynaptic terminals, which in turn leads to abnormally enhanced LTP (Earls et al., 2010). Similar changes in plasticity may occur in patients with schizophrenia who have increased SERCA2 levels. Although miRNA modulation may be one way to achieve this, SERCA2 is also regulated by other mechanisms (Vandecaetsbeek et al., 2011). Future studies will seek to determine which of these mechanisms contributes to SERCA2 misregulation in the brains of patients with schizophrenia. Regardless of how SERCA2 changes, our results indicate that SERCA2 upregulation at synapses may be a pathogenic mechanism of schizophrenia.
In this study, we identified the contribution of Dgcr8 to the age-dependent plasticity deficits observed in the 22q11DS mouse model. The magnitude of LTP increase measured in Dgcr8+/− and Df(16)2/+ mice was smaller than that observed in Df(16)1/+ mice (Earls et al., 2010). Furthermore, hippocampus-dependent spatial learning and memory measured in Morris water maze tasks was normal in mature Df(16)2/+ mice. These results suggest that combinatorial loss of multiple genes within the disease-critical region is necessary to achieve the effect of the full microdeletion. This complexity is not surprising, because no single-gene deletions have been found that recapitulate the symptoms of 22q11DS and because schizophrenia is widely believed to be a polygenic disorder. Future LTP screening for additional contributors outside the Df(16)2 region may reveal other genes that contribute to cognitive deficits in 22q11DS.
Since the original description of schizophrenia, the disease's causative molecular mechanisms have remained elusive. Schizophrenia has a complex etiology, affects multiple brain systems, and may arise from different molecular insults. Presynaptic perturbation of Ca2+ signaling due to SERCA2 upregulation contributes to the cognitive deficits that are central to this disease. The finding that SERCA2 is upregulated in the human brain with schizophrenia provides a mechanistic link between 22q11DS and schizophrenia. Here we identified miR-25 and miR-185 as regulators of SERCA2. These miRNAs are depleted in 22q11DS due to Dgcr8 deficiency. Restoration of these miRNAs in presynaptic neurons rescued synaptic plasticity deficits in Dgcr8-deficient mice. These findings suggest that progressive miRNA-dependent SERCA2 upregulation at central synapses is a pathogenic mechanism of 22q11DS.
This work was supported in part by National Institute of Mental Health Grants R01MH079079, R01MH095810 (S.S.Z.) and the American Lebanese Syrian Associated Charities. We thank Dr. Elizabeth Illingworth (Dulbecco Telethon Institute) for providing Df(16)2/+ mice, Dr. Anthony Wynshaw-Boris (University of California San Francisco) for providing Znf74l-Ctp/+ mice, Dr. Joseph Gogos (Columbia University) for providing Comt+/− and Prodh+/− mice, and Dr. Stephen Strittmatter (Yale University) for providing Rtn4r+/− mice for this study. We also thank the Harvard Brain Tissue Resource Center, which is supported in part by Public Health Service Grant R24MH068855, and Dr. Vahram Haroutunian (The Mount Sinai School of Medicine) for providing postmortem human brain samples. L.T.B. was supported in part by Grant 5R25CA023944 from the National Cancer Institute. We thank Dr. John Gray and the St. Jude Vector Core for producing AAVs, Granger Ridout and the St. Jude Hartwell Center for Bioinformatics for performing microarrays, and Vani Shanker for editing this manuscript. The funding sources had no role in study design, data collection and analysis, decision to publish, or preparation of this manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Stanislav S. Zakharenko, Department of Developmental Neurobiology, MS 323, St. Jude Children's Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105-3678.