Address correspondence and reprint requests to A. L. Epstein and D. Jerusalinsky. A. L. Epstein, CMCG, University C Bernard Lyon1, Villeurbanne, France. E-mail: email@example.com D. Jerusalinsky, Institute of Cellular Biology and Neuroscience ‘Professor E De Robertis', School of Medicine, University of Buenos Aires, Buenos Aires, Argentina. E-mail: firstname.lastname@example.org
Amplicon vectors derived from herpes simplex virus type 1 were built to modify NMDA receptors by expressing antisense RNA for the essential NR1 subunit. Their ability to modify endogenous levels of NR1 was tested in cultures of rat embryo neocortical neurons. We studied behaviour and tested for expression in adult rats injected with those vectors into the dorsal hippocampus to find out which cells and how many appear involved in memory formation. Rats injected with vectors expressing NR1 antisense performed significantly worse than control rats in an inhibitory avoidance task. Immunohistochemistry was performed in brain slices from the same animals. The transduced cells represented 6–7% of pyramidal neurons in CA1, showing that a single gene knockdown of NR1 in a small number of neurons significantly impaired memory formation. Perhaps neurons undergoing synaptic plasticity are more susceptible to NR1 knockdown, and hence NMDAR are particularly required in those neurons undergoing synaptic plasticity during learning, or perhaps, and more likely, there is not a high level of redundancy in the hippocampal circuits involved, leading to the idea that a certain level of NR1 expression/availability appears necessary for memory formation in most of CA1 pyramidal neurons.
The NMDA subtype of glutamate receptor (NMDAR) is involved in activity-dependent synaptic plasticity, such as associative long-term potentiation (Bliss and Collingridge 1993), and in related central functions, such as learning and memory (Morris et al. 1986; Collingridge and Singer 1990). The hippocampus is crucial in consolidating different kinds of memories (Scoville and Milner 1957). Changes in synaptic connection have been proposed as the basis for long-term information storage, and the NMDAR has been considered a molecular coincidence-detector requiring both presynaptic glutamate release and postsynaptic depolarization for the channel to open.
Molecular and cellular mechanisms underlying behaviour have been investigated using classical pharmacology. NMDAR antagonists impair memory (Morris et al. 1986; Jerusalinsky et al. 1992). However, a somatic genetic deletion, whenever it could be temporally and regionally confined, might be better than a pharmacological blockade regarding molecular and anatomical specificities. The NMDAR consists of NR1 and NR2, and sometimes NR3 subunits, NR1 being essential and always present (Moriyoshi et al. 1991; Nakanishi 1992; Hollmann and Heinemann 1994).
Tsien et al. (1996), using a NR1-knockout (KO) mouse restricted to the hippocampal CA1 region, showed that NMDAR-mediated synaptic plasticity was essential for spatial learning, but not for a non-spatial task. McHugh et al. (1996) showed that NMDAR was necessary for proper representation of space. As most, if not all of the supposed pyramidal cells in CA1 (judged by size and location) suffered the deletion, it is unclear which of those cells and how many are needed to support a memory trace. Also, KO from the middle of the third week is likely to reduce but does not preclude the possibility of developmental alterations. In addition, that technology is restricted to mice and depends on availability of specific transcription promoters.
We have shown previously that hippocampal injection in adult rats of viral vectors carrying the antisense transcription sequence for the NR1 subunit resulted in significant impairments of the ability of rats to recognize a new environment (Cheli et al. 2002).
The goals of the present work were to study behaviour in rats injected with viral vectors able to knockdown the NR1 subunit into the dorsal hippocampus by expressing the NR1 antisense RNA, to test for expression of the transgene in the same animals, and to determine the cell type and numbers of cells infected as a way to investigate which NMDAR-expressing cells and how many appear to be involved in recording the memory trace.
Behaviour in an open field was observed to evaluate possible sensorimotor disturbances. Then, the rats were trained in an inhibitory avoidance, a well-characterized task with long-lasting memory traces.
All the procedures (housing, handling, surgical and behavioural methods) involving animals were carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) and the School of Veterinary Science of Buenos Aires University Review Committee for use of Animal Subjects.
Materials and methods
Amplicon vectors (A)
Amplicon plasmids containing an HSV-1 (herpes simplex virus type 1) origin for DNA replication and the HSV-1 packaging sequence, with two promoter-transgene cassettes carrying either the NR1 sense (+) or antisense (–) sequence and that for GFP (green fluorescent protein) (Adrover et al. 2003) were used to construct the vectors according to Zaupa et al. (2003). The vectors were grown using the LaLΔJ helper system, leading to vector stocks practically devoid of helper virus (Zaupa et al. 2003). The vectors were named according to the transgenes: A-GFP-NR1(+) (titre: 1.4 × 108 amp/mL) and A-GFP-NR1(–) (titre: 6.7 × 107 amp/mL).
Expression of NR1 in primary culture of neurons
The neocortex of Wistar rats at embryonic day 18–19 was dissected. The tissue was dissociated in 0.05% trypsin and 0.1 mg/mL DNaseI in high glucose Dulbecco's modified Eagle's medium and incubated in Neurobasal/B27 (Brewer 1995) at 36.5°C with 5% CO2. After 14 days the cells were incubated for 48 h with either A-GFP-NR1(+) or A-GFP-NR1(–). Then immunocytochemistry and immunoblots were performed.
The cells were washed with 0.1 m phosphate-buffered saline pH 7.2–7.4 and harvested in 25 mm Tris-HCl, 0.32 m sucrose, 5 mm EDTA, 1 mm EGTA, 1 mm phenylmethylsulfonyl fluoride and 10 µg/mL bacitracine (pH 7.4). Aliquots of 25 µg protein were submitted to sodium dodecyl sulfate–polyacylamide gel electrophoresis. Both primary antibodies anti-NMDAR1 and anti-actin were rabbit polyclonal (Chemicon, Temecula, CA, USA); the secondary antibody was goat anti-rabbit IgG-HRP (horseradish peroxidase) conjugate (Vector, Burlingame, CA, USA). The bands were visualized by chemiluminescent detection (ECL, Amersham, Piscataway, NJ, USA) and quantified by ImageJ (NIH).
The cells were washed with phosphate-buffered saline at 37°C and fixed in 4% paraformaldehyde in phosphate-buffered saline for 45 min. Then the cells were incubated with the primary anti-NMDAR1 antibody (1:100) over night and washed-out, and then the secondary antibody goat anti-rabbit tetramethylrhodamine isothiocyanate (TRICT) 1:100 was added. The fluorescence was detected under a microscope (Olympus BX100, Olympus Optical Co. Ltd, Tokyo, Japan) with the appropriate filters (‘green’, 488 nm; ‘red’, 543 nm). The analysis was performed with Image-Pro Plus (Media Cybernetics, Silver Spring, MD, USA).
Adult male Wistar rats (220–270 g) were maintained from birth in inverted 12 h light:dark cycles, eight animals per cage with food and water ad libitum. The vectors were bilaterally injected in a final volume of 1 µL/side into the dorsal hippocampus (stereotaxic coordinates anteroposterior 3.6; lateral 3.0; dorsoventral 2.3; Paxinos and Watson 1996) of rats anesthetized with 0.2 mL/100 g of body weight of 3:1 ketamine (50% solution; Kensol®):xilacine (2% solution; Rompun®) i.m. There were three treated groups of 12 rats each, injected (i) with saline – sham operated, (ii) with A-GFP-NR1(+), and (iii) with A-GFP-NR1(–), and a fourth group of 12 naive animals. The behavioural assays were performed under low-intensity red light in a double-blind way. Five days after injection, the behaviour in an open field (1.2 m × 1.2 m × 0.4 m high, with 64 quadrants on the floor) was observed for 2 min in order to evaluate sensorimotor disturbances. Crossings from one quadrant to another, rearings – denoting exploratory behaviour – and groomings were quantified. The rats were then trained in an inhibitory avoidance to a foot-shock. Each animal was put on an isolated platform (30.0 cm long × 8.0 cm wide × 3.0 cm high) on one side of an acrylic box (60.0 cm long × 30.0 cm wide × 40.0 cm high), with an electrifiable grid-floor made of parallel bronze bars (Jerusalinsky et al. 1992). The training latency was the time to get down with its four paws onto the grid-floor and receiving a foot-shock (0.5 mA, 2 s). In the test session 24 h later, the test latency, limited to 120 s, was measured. Latency difference was an indication of consolidation/retention (see Izquierdo et al. 1992).
Transgene expression in the injected hippocampus
After the behavioural assays, the same rats were perfused under 0.2 mL/100 g of body weight of 3:1 ketamine (50% solution; Kensol®):xilacine (2% solution; Rompun®) anesthesia, with 400 mL of 4% paraformaldehyde in cool phosphate-buffered saline. Serial brain slices 30 µm thick, were obtained with a cryostat (Leica): some were immunostained using anti-GFP and GSA-B4 (Vector); alternate slices were stained with anti-NeuN (neuronal nuclei) (Molecular Probes, Eugene, OR, USA), anti-GFP and Hoechst 33342 (1 µg/mL, KPBS 0.12 m). Images were recorded every 3 µm by confocal microscopy. The CA1 region was selected and cells stained for GFP, NeuN and Hoechst were quantified using the Metamorph® software.
Results and discussion
In vitro studies
Primary cultures of cortical neurons were infected with A-GFP-NR1(+) and A-GFP-NR1(–) vectors at high multiplicity of infection (MOI = 5–7). About 50% of the cells expressed GFP, showing that both vectors transduced the cells (Figs 1b and d). By immunocytochemistry (Figs 1a and c), a high expression of NR1 was observed in cells infected with A-GFP-NR1(+), whereas there was an evident decrease in the immunostaining when the culture had been infected with A-GFP-NR1(–), compared to non-transduced cells. Accordingly, the density of bands in the immunoblot (Fig. 1e) of cells infected with A-GFP-NR1(+) showed a significant increase in NR1 (433.9% ± 17.7, p < 0.001, n = 4), whereas those infected with A-GFP-NR1(–) showed a significant decrease (26.1% ± 1.5, p < 0.01, n = 4) compared to non-infected cells (100% ± 8.5) (one-way anova and Tukey's Multiple Comparison Test).
We had previously shown that cells tranfected with A-GFP-NR1(–) from the same stock produced the NR1 antisense RNA (Adrover et al. 2003); here we show the decrease of NR1 in neurons infected with A-GFP-NR1(–). NR1 was overexpressed in neurons infected with A-GFP-NR1(+), whereas GFP was conspicuously expressed from both vectors. Therefore, the constructed amplicon vectors, practically devoid of helper virus, were efficient at expressing the transgenes and modifying endogenous levels of NR1 in neurons; hence, they were appropriate to be used for our in vivo assays.
In vivo studies
Rats injected with the vectors were tested in an open field 5 days later. The naive animals showed no differences compared to sham-operated rats in every assay for all the recorded parameters (not shown). The number of crossings and rearings showed a normal distribution (Wilk-Shapiro modified test, Mahibbur and Govindarajulu 1997). Media ± standard error of mean (SEM) of number of crossings and rearings were, for sham operated, 122.2 ± 5.4 and 22.6 ± 1.4, for A-GFP-NR1(+) injected, 113.3 ± 8.9 and 19.3 ± 1.8, and for A-GFP-NR1(–), 106.9 ± 6.4 and 19.3 ± 0.8 (n = 12 for each group). There were no significant differences between groups (one-way anovaF = 0.91, p > 0.05 for crossings, F = 2.67, p > 0.05 for rearings). These results strongly indicated that there were no alterations in locomotor activity or in exploratory behaviour of injected rats.
Then, the animals were trained in the inhibitory avoidance task. As test latencies were limited to 120 s and they were not normally distributed (Wilk–Shapiro modified test), the data were analysed by non-parametric statistics (Conover 1999). There were no differences between groups for training latencies (Kruskal–Wallis); median with interquartile ranges [25; 75] were, for sham-operated rats, 3 s [2.00; 3.49], for A-GFP-NR1(+), 3.32 s [2.90; 6.27], and for A-GFP-NR1(–), 2.72 s [1.61; 4.59] (n = 12 for each group). In the test session 24 h later, A-GFP-NR1(+) injected rats showed no differences compared to sham-operated animals; test latencies were significantly higher than training latencies (p < 0.05, Wilcoxon test). Median of test latencies for sham-operated animals was 16.57 s [10.54; 65.73] and for A-GFP-NR1(+) injected rats, 15.22 s [9.93; 57.27]. In other words, they learned and remembered to spend longer on the platform. Although there was an overexpression of NR1 in primary culture of neurons transduced with the A-GFP-NR1(+) vector, in our in vivo experiments neither significant improvements in performance nor overexpression of the subunit were detected. Genetic enhancement of NMDAR function in transgenic mice resulted in an enhancement of both synaptic coincidence detection and learning/memory (Tang et al. 1999; Wong et al. 2002). Barria and Malinow (2002) reported that exogenously expressed NR1 in culture neurons transfected with the corresponding plasmid was not assembled as NMDAR unless coexpressed with NR2. If NR1 protein expressed from the vectors was not properly assembled in vivo, an improvement in learning and memory would not be expected.
On the other hand, rats injected with A-GFP-NR1(–) did not show a significant increase in test latencies. Median for training latencies was 2.72 s [1.61; 4.59] and for test latencies it was 5.46 s [4.25; 9.43]. These results corroborated previous findings and strongly supported the involvement of hippocampal NMDAR in recording the trace for this task (Jerusalinsky et al. 1992; Adrover et al. 2003). Furthermore, we have demonstrated previously that rats injected with A-NR1(–) vectors did not habituate to the open field, and there was a slight decrease in the NR1 subunit in the hippocampus (Cheli et al. 2002). Those behavioural tests have been reproduced in the present work with similar results (not shown).
We used confocal microscopy to determine and quantify the cells transduced by the vectors in vivo. GFP expression was restricted to a small region in the dorsal hippocampus (Fig. 2a); most of the ‘green cells’ were pyramidal neurons. The volume for quantification (1000 µm thick × 400 µm wide × 70 µm high) was calculated from serial brain slices. Total cell number (Hoechst), neurons (NeuN) and transduced cells (GFP) were quantified (Mouton 2002) (Figs 2b and c). About 6–7% of pyramidal neurons in CA1 and only about 1–1.5% of granule cells just below those pyramidal neurons were transduced, as judged by GFP expression. This implies that reduction of NMDAR in only a small proportion of CA1 pyramidal neurons affects ability to learn and/or remember the task. When larger hippocampal regions were affected, i.e. with NMDAR antagonists (Jerusalinsky et al. 1992), the performance in similar tasks did not appear to be different from the results in this study.
Cain (1998) reported that rats receiving NMDAR antagonists displayed sensorimotor disturbances. Inada et al. (2003), delivering NR1-antisense to the rat hippocampus, caused a mild knockdown that did not affect spatial memory but it appeared to produced a deficit in sensorimotor gating. From the behavioural point of view, neither the conditional CA1 NR1-KO mice (Tsien et al. 1996), nor our infected rats appeared to have sensorimotor alterations or disturbances in exploratory behaviour. The learning deficits for CA1 NR1-KO mouse have been extended to non-spatial tasks (Rampon et al. 2000; Shimizu et al. 2000). Since most of CA1 pyramidal neurons in that mouse suffered the deletion, it was not clear which cells and how many were needed to support the trace.
The injection of a neurotropic amplicon vector expressing NR1 antisense RNA seems a rapid and efficient procedure to modify the receptor, with temporal and regional restrictions. Here we showed that the knockdown of the NR1 subunit in as few as 6–7% of CA1 pyramidal cells was sufficient to impair learning of an inhibitory avoidance task. Interestingly, Rumpel et al. (2005) have reported that memory of fear conditioning was reduced if AMPA receptor synaptic incorporation was blocked in about 10% of lateral amygdala neurons.
A-GFP-NR1(+) vector caused the same average rate of infection as A-GFP-NR1(–); this strongly suggests that expression of NR1 antisense RNA, rather than any other related variable, was responsible for learning impairment.
Although the hippocampus appears crucial for memory acquisition in rodents, taking into account that severe neuropathology or large surgical lesions do not seem to easily perturb memory formation, it is at least surprising that dysfunction of just a few neurons in CA1 could cause significant impairments in recognizing a new environment (Cheli et al. 2002) and in learning an inhibitory avoidance task. Assuming that our amplicon vectors, which mainly transduced pyramidal neurons, did not preferentially infect certain of them (they were administered 5 days before the behavioural assays, reaching a ‘plateau’ in expression on the third day), one must conclude that affecting a small fraction of pyramidal neurons in CA1 is sufficient to disturb memory formation. It might be speculated that neurons and their synapses underlying learning-related synaptic plasticity were more susceptible to the expression level of the transgene and, therefore, NMDAR are particularly required in those neurons undergoing synaptic plasticity during learning. But assuming that CA1 neurons were infected at random, the results suggest that there would not be a high level of redundancy in the circuits involved, leading to the idea that a certain level of NR1 subunit expression/availability is required during memory formation in most of CA1 pyramidal neurons bearing NMDAR.
Financial support: J S Guggenheim Memorial Foundation (USA), CNRS (France), CONICET (Argentina), ANPCYT-FONCYT and University of Buenos Aires (Argentina).
Plasmids with the NR1 ORF were a gift from Dr S Nakanishi. We thank to Dr M Conn and Dr S Ojeda (ONPRC-OHSU) for facilitating the use of the confocal microscope and providing technological assistance (Technological Assistance Fogarty Grant).