Vesicular glutamate transporters 1 and 2 (VGLUT1, VGLUT2) show largely complementary distribution in the mature rodent brain and tend to segregate to synapses with different physiological properties. In the hippocampus, VGLUT1 is the dominate subtype in adult animals, whereas VGLUT2 is transiently expressed during early postnatal development. We generated and characterized VGLUT1 knockout mice in order to examine the functional contribution of this transporter to hippocampal synaptic plasticity and hippocampus-dependent spatial learning. Because complete deletion of VGLUT1 resulted in postnatal lethality, we used heterozygous animals for analysis. Here, we report that deletion of VGLUT1 resulted in impaired hippocampal long-term potentiation (LTP) in the CA1 region in vitro. In contrast, heterozygous VGLUT2 mice that were investigated for comparison did not show any changes in LTP. The reduced ability of VGLUT1-deficient mice to express LTP was accompanied by a specific deficit in spatial reversal learning in the water maze. Our data suggest a functional role of VGLUT1 in forms of hippocampal synaptic plasticity that are required to adapt and modify acquired spatial maps to external stimuli and changes.
Glutamate is the most abundant excitatory neurotransmitter in mammalian central nervous system (CNS). Only recently it has become clear that it is packed by vesicular glutamate transporter (VGLUT) into vesicles for subsequent release during fast synaptic transmission. To date, 3 different subtypes (VGLUT1, VGLUT2, and VGLUT3) have been molecularly identified and functionally characterized, which show 74–90% sequence homology and almost identical transport properties (Aihara et al. 2000; Bellocchio et al. 2000; Fremeau et al. 2001, 2002; Takamori et al. 2001, 2002; Gras et al. 2002; Herzog et al. 2004). VGLUT1 and VGLUT2 are the predominant isoforms, accounting for most of the excitatory glutamatergic terminals in the CNS. In contrast to their comparable functional features, VGLUT1 and VGLUT2 have largely complementary distributions in the mature rodent brain with a partial overlap in only a few regions (Fremeau et al. 2001; Fujiyama et al. 2001; Herzog et al. 2001; Sakata-Haga et al. 2001; Hisano et al. 2002; Gras et al. 2005; Nakamura et al. 2005; Zhou et al. 2007). Notably, both transporters are expressed in hippocampus, cortex, and cerebellum, but VGLUT2 expression switches to VGLUT1 during early postnatal stages (Miyazaki et al. 2003; Fremeau et al. 2004; Gras et al. 2005). Unlike the 2 other isoforms, VGLUT3 is diffusely distributed in the brain, defining a discrete subpopulation of nonglutamatergic neurons (Gras et al. 2002, 2008; Schafer et al. 2002; Boulland et al. 2004; Herzog et al. 2004).
The (patho)physiological functions of VGLUTs are slowly being elucidated. VGLUT1 and VGLUT2 appear to be involved in brain disorders such as depression, schizophrenia, Alzheimer's disease, and Parkinson's disease (Harrison et al. 2003; Moutsimilli et al. 2005, 2008; Tordera et al. 2005; Kashani et al. 2006; Kirvell et al. 2006). It has been proposed that VGLUT levels are critical for the balance between excitation and inhibition (Cline 2005; Erickson et al. 2006; Seal et al. 2008). Accordingly, heterologous expression of VGLUT1 or VGLUT2 converted inhibitory neurons to excitatory ones (Takamori et al. 2000, 2001). Variations in VGLUT1 level critically affect the efficacy of glutamatergic synaptic transmission (Fremeau et al. 2004; Wojcik et al. 2004; Wilson et al. 2005), which is drastically reduced in cerebellar and hippocampal neurons from VGLUT1-deficient mice (Fremeau et al. 2004; Wojcik et al. 2004). Overexpression of VGLUT1 rescues these latter phenotypes or even results in enhanced excitatory neurotransmission (Wojcik et al. 2004; Wilson et al. 2005). In a recent study, heterozygous VGLUT1-deficient mice of another genetic origin were reported to display normal short-term but impaired long-term memory in the novel object recognition test, whereas spatial memory in the Morris water maze was unchanged (Tordera et al. 2007).
Some authors speculated that neurons expressing VGLUT1 and VGLUT2, respectively, could be differentially involved in types of synaptic plasticity such as long-term potentiation (LTP) (Varoqui et al. 2002; Nakamura et al. 2005). Here, we report a substantial reduction in the magnitude of LTP in the hippocampal CA1-region of heterozygous VGLUT1-deficient mice where the expression of VGLUT1 is reduced to about 60%. Heterozygous VGLUT2-deficient mice, in contrast, turned out to be virtually unchanged in their propensity to generate LTP. In accordance with their LTP deficit, VGLUT1+/− mice showed a reduced flexibility in relearning a new platform position in the water maze.
Materials and Methods
Generation of VGLUT1 Knockout Mice
VGLUT1 knockout mice were developed in collaboration with Lexicon Genetics Inc. Genomic DNA clones were isolated by screening of the 129SvEvBrd derived lambda pKOS genomic library (Wattler et al. 1999). A 7.4-kb genomic clone spanning the first 2 coding exons was used to generate the targeting vector. An IRESLacZ/MC1neo selection cassette was inserted as an SfiI fragment to replace a 72-bp VGLUT1 genomic fragment starting with the ATG startcodon in exon 1 after yeast-mediated homologous recombination. The NotI-linearized vector was electroporated into 129 Sv/Evbrd(LEX1) embryonic stem (ES) cells and neomycine-resistant ES cell clones were isolated and analyzed for homologous recombination by Southern blot analysis. Targeted ES cell clones were injected into C57BL/6(albino) blastocysts, and the resulting chimeras were mated with C57BL/6(albino) females to generate animals heterozygous for the mutation. These were subsequently crossed to generate all 3 genotypes employed in these studies. Polymerase chain reaction (PCR) was used to screen genotypes by using DNA isolated from mouse-tail biopsy samples. Primers 5′CAGGAGGAGTTTCGGAAGC3′ and 5GAGAGACAGATCAAAGTGGG3′ amplified a 508-bp band from the wild-type allele, whereas primers 5′GCAGCGCATCGCCTTCTATC3′ and 5′GAGAGACAGATCAAAGTGGG3′ amplified a 606-bp band from the knockout allele.
VGLUT2 Knockout Mice
VGLUT2 knockout mice were generated in collaboration with Lexicon Genetics Inc by homologous recombination resulting in deletion of exon 2 and disruption of the open reading frame as described elsewhere (Moechars et al. 2006). Heterozygous animals and their respective litter controls were used in all experiments. Mixed genotype groups were kept in standard animal cages under conventional laboratory conditions (12 h/12 h light–dark cycle, 22 °C), with ad libitum access to food and water (unless stated otherwise). Male animals aged 8–10 weeks were used for behavioral experiments, which were all conducted during the light phase of their activity cycle. All protocols have been reviewed and approved by the animal experiments committee of the Katholieke Universiteit Leuven, Belgium, and were carried out in accordance with the European Community Council Directive (86/609/EEC).
Quantitative RT-PCR Analysis
Quantitative RT-PCR (QRT-PCR) analysis was used to show expression of the VGLUT1 and VGLUT2 transcript. Total RNA was isolated from different tissues, using Trizol (Invitrogen; Carlsbad, CA) and first-strand cDNA synthesis was performed on 0.5 μg total RNA using random hexamer primers and SuperscriptII RT (Invitrogen). QRT-PCR was performed on an ABIPrism 7700 cycler (Applied Biosystems; Foster City, CA) using a Taqman PCR kit. Serial dilutions of cDNA were used to generate standard curves of threshold cycles versus the logarithms of β-actin, VGLUT1, VGLUT2 and VGLUT3 concentrations. A linear regression line calculated from the standard curves allowed the determination of transcript levels in RNA samples from 18-day-old mice. Expression levels were normalized to the expression levels of the β-actin gene (housekeeping gene). Following normalization relative expression levels of the VGLUT1, VGLUT2, and VGLUT3 genes were determined in +/+ (wild type), +/− (heterozygous) and −/− (homozygous) VLGUT1 mice. The following primer-probe combinations were used in these experiments:
probe 5′GGCGGCCTGGAACCACCCA3′ [5′]FAM [3′]TAMRA
probe 5′CCTTTTTCTCCCAGCCGTTAGGCCA3′ [5′]FAM [3′]TAMRA
probe 5′GGCGGCCTGGAACCACCCA3′ [5′]FAM [3′]TAMRA
probe 5′TGCTTGCTGATCCACATCTGCTGGA3′ [5′]FAM [3′]TAMRA
Mice of an age of 18 days were sacrificed by cervical dislocation and decapitation. Brains were carefully removed and fixed in acidic formalin or an alcohol-based fixative by immersion overnight at room temperature. All tissues were embedded in paraffin and sectioned at 6 μm. The detection for VGLUT1 and VGLUT2 was performed by immunofluorescence histochemistry. VGLUT1 immunofluorescence staining requires heat-mediated microwave antigen retrieval using ethylenediaminetetraacetic acid or citrate after deparaffination and rehydration, followed by additional enzymatic pretreatment (trypsin, 10 min at 37 °C). Prior to incubation with the primary rabbit polyclonal anti-VGLUT1 (AB5905, Chemicon,) and anti-VGLUT2 (AB5907, Chemicon) antibody (1/100 in phosphate buffered saline (PBS)/bovine serum albumin, overnight at 4 °C) sections were blocked with 10% normal goat serum (DakoCytomation, Denmark) for 20 min at room temperature. Subsequently, sections were rinsed and incubated with Cy3 conjugated goat antirabbit IgG (Jackson Immunoresearch Laboratories, Inc., PA) (1/500 in PBS pH7.4) for 2 h at room temperature and counterstained with Hoechst 33342 (Molecular Probes, Inc., OR) (1/500 in AD). To control for cross-reactivity caused by aspecific binding, negative controls were included by performing the immunostaining without adding primary antibodies to the sections. All reactions were negative confirming the noninterference of the secondary antibody. All morphological analysis and data acquisition were performed on an Axioplan-2 system (Carl Zeiss, Germany) with the integrated package image analysis software KS 400 or Axiovision Release 4.2.
Electrophysiological Long-Term Recordings
Electrophysiological recordings were performed as described previously (Balschun, Manahan, et al. 1999; Balschun et al. 2003). Transverse hippocampal slices (400 μm) of 5- to 8-month-old mice were kept in a submerged-type slice chamber and permanently perfused with 32 °C artificial cerebrospinal fluid (ACSF) (flow rate ∼2.8 mL/min). After preparation, slices were allowed to recover for at least 1 h. A glass electrode (filled with ACSF, 1–4 MΩ) was positioned in the apical dendritic layer to record field excitatory postsynaptic potentials (fEPSPs). For stimulation, a lacquer-coated stainless steel stimulating electrode was placed into the CA1 stratum radiatum about 200 μm apart. To assess basic properties of synaptic responses, input–output curves were constructed. The stimulation strength was adjusted to evoke a fEPSP slope of 35% of the maximum and kept constant throughout the experiment. Paired-pulse facilitation (PPF) was examined by applying 2 pulses in rapid succession (interpulse intervals, IPIs of 10, 20, 40, 100, 200, and 500 ms, respectively) at 120-s intervals. During baseline recording 3 single stimuli (0.1-ms pulse width; 10-s interval) were measured every 5 min and averaged. Once a stable baseline had been established, LTP was induced by 3 trains of theta-burst stimulation (TBS, 10 burst of 4 stimuli at 100 Hz, separated by 200 ms, 0.2-ms pulse width) (Larson and Lynch 1986; Larson et al. 1986) applied every 10 min. Immediately after every TBS train, recordings were taken at 1, 4, 7, and 10 min. Thereafter, the recording interval was 5 min. In all experiments, the recording of slices from mutant mice was interleaved by experiments with wild-type controls. Curves of the decay kinetics of potentiation after TBS stimulation were obtained by a nonlinear regression with the equation y = yo + a*e−t/tau (one-phase exponential decay with the asymptote yo, the span a and the decay-time constant tau) using GraphPad Prism 4.0 software (GraphPad Software, Inc., San Diego, CA). Statistical analysis of between-group differences was performed with the Mann–Whitney U-test. Within-group differences were evaluated with the Wilcoxon matched pairs signed rank test. To test for group differences between LTP time series and between the decay curves of fEPSPs during tetanization, analysis of variance with repeated measures (RM-ANOVA) was used (SPSS software; SPSS Inc. Chicago, IL).
Morris Water Maze
Spatial memory abilities were examined in the standard hidden-platform acquisition and retention (i.e., long-term memory) and working memory versions of the Morris maze (Stewart and Morris 1993). A 150-cm circular pool was filled with water, opacified with nontoxic white paint, and kept at 25 ± 1 °C as previously described (D'Hooge et al. 2005). A 15-cm round platform was hidden 1 cm beneath the surface of the water at a fixed position. Each daily trial block consisted of 4 swimming trials (10-min intertrial interval) starting randomly from each of 4 starting positions. Mice that failed to find the platform within 2 min were guided to the platform, where they remained for 15 s before being returned to their cages. Interspersed probe trials were conducted after 5 and 10 acquisition trial blocks. During these probe trials, the platform was removed from the pool, and the search pattern of the mice was recorded for 100 s. After establishing a stable acquisition curve and a clear preference for the target quadrant during probe trials, a reversal test was administered. In this test, platform was placed opposite the trained position and average distance to the platform was measured. Five swimming trials (10-min intertrial interval) starting semirandomly from each of 4 starting positions (swim 1, 2, and 5 from positions furthest away from the platform, whereas 3 and 4 from points close to the platform) were administered. Mice that failed to find the hidden platform within 2 min were guided to the platform, where they remained for 15 s before being returned to their cages. Swimming paths of the animals were recorded using “EthoVision” video tracking equipment and software (Noldus bv, Wageningen, The Netherlands). Three populations of mice were tested at different time points (group A, N = 15/7, group B, 13/13, and group C, 12/10 for VGLUT1+/+ and +/−, respectively). Only groups B and C were subjected to reversal tests (N = 26 and 23 for VGLUT1+/+ and +/−, respectively). Animals that were floating, defined as swim velocity less than 5 cm/s for either longer than 20 s or 40% of total swim time, were excluded from the analysis (2 VGLUT1+/−, 1 VGLUT1+/+).
In order to detect and visualize differences in spatial search strategy of VGLUT1+/− and WT mice, we transformed the raw tracking data obtained by EthoVision into average heat plots of each genotype using custom-made software written in MATLAB. Thus, all the images for each genotype were superimposed separately to generate an average heat plot per genotype. The heat plots were normalized to the number of animals and amount of samples (path length), to obtain comparable color scales. In a further step, the average heat plot of VGLUT1+/− was subtracted from VGLUT1+/+ yielding a differential heat plot with positive values (VGLUT1+/+ > VGLUT1+/−) in red, and negative values (VGLUT1+/− > VGLUT1+/+) in blue. Green zones indicate that there was no difference between groups.
Generation of the VGLUT1 Knockout Mice
In mouse, the VGLUT1 gene contains 12 exons. Homologous recombination resulted in deletion of the coding region of exon 1 and disruption of the open reading frame (Fig. 1A,B). Correct targeting in ES cells was confirmed by Southern blot analysis and loss of the wild-type VGLUT1 allele was confirmed by PCR analysis (results not shown). Expression of the VGLUT1 transcript was absent in VGLUT1−/− 18-day-old mice, whereas intermediate levels were obtained in the VGLUT1+/− littermates (Fig. 1C). The absence of the VGLUT1 protein was confirmed by western blotting (Fig. 1F,G). To look for potential compensatory mechanisms, VGLUT2 and VGLUT3 mRNA levels were analyzed in brain, spinal cord, and dorsal root ganglia (DRG) of 18-day-old mice (Fig. 1D,E). No changes in VGLUT2 nor VGLUT3 expression at RNA level could be detected between the genotypes. VGLUT1−/− animals were born at the expected Mendelian frequency and were indistinguishable from their VGLUT1+/+ and VGLUT1+/− littermates at birth. However, coincident with the developmental switch from VGLUT2 to VGLUT1 in the hippocampus (Fremeau et al. 2004; Wojcik et al. 2004) VGLUT1−/− mice were retarded in their growth (Fig. 1H) and failed to thrive from the third week onward. None of the VGLUT1−/− mice survived the end of the fourth week. Therefore, only heterozygous animals were used for the experiments described herein. No growth defect in body size or difference in body weight was detectable in VGLUT1+/− mice compared with their VGLUT1+/+ littermates during the life span of 3–8 months.
Immunohistological analysis of expression confirmed the differential expression pattern of VGLUT1 and VGLUT2 in murine brain that has been reported before (Fujiyama et al. 2001; Herzog et al. 2001; Gras et al. 2002; Kaneko et al. 2002; Fremeau et al. 2004; Liguz-Lecznar and Skangiel-Kramska 2007; Callaerts-Vegh et al., forthcoming). The expression pattern of VGLUT1 and VGLUT2 appeared in a dense granular way (data of VGLUT2 not shown). As depicted in Figure 2A, VGLUT1 is widely expressed in forebrain and cerebellum of VGLUT1+/+ mice. Brain regions showing abundant VGLUT1 in neuropil are cortex, hippocampus, caudate putamen, thalamus, nucleus accumbens, and both granular and molecular layers of cerebellum. Discrete VGLUT1-immunoreactivity has been seen in hypothalamus. Inspecting the hippocampus in more detail, VGLUT1 immunofluorescence showed strong immunoreactivity in both polymorph and molecular layers of dentate gyrus, in stratum oriens, lucidum and radiatum of the different CA-regions, in stratum lacunosum-moleculare and subiculum (Fig. 2C). Only few immunoreactive spots are observed in the granule cell layers of dentate gyrus and in the pyramidal cell layer of CA1 region (Fig. 2C). Lacking VGLUT1 immunoreactivity in VGLUT1−/− mice confirms the successful targeted disruption of the HVGLUT1 gene (Fig. 2B,D). VGLUT2 expression is comparable in both VGLUT1+/+ and VGLUT1−/− mouse brains (data not shown). The differential expression pattern of VGLUT1 was also confirmed by QRT-PCR (Fig. 2E).
It has been proposed that VGLUT1 is present at synapses known to exhibit long-term synaptic plasticity, whereas VGLUT2 is associated with synapses in sensory and autonomic pathways that display high-fidelity neurotransmission (Varoqui et al. 2002; Nakamura et al. 2005, but see Zhou et al. 2007 for contradictory findings in the cochlear nucleus). Therefore, we investigated whether the 2 VGLUT subtypes are differentially involved in long-term synaptic plasticity. Because LTP in the CA1 region of the hippocampal slice preparation is an established model for the investigation of this topic (Malinow et al. 2000; Dineley et al. 2001), we examined both mouse lines in a type of LTP that is induced by a strong TBS. TBS mimics natural patterns of discharge of CA1 pyramidal cells as occurring in animals during the sampling and analysis of learning relevant information (Otto et al. 1991), whereas the frequently used high-frequency stimulation (e.g., 100 Hz for 1 s) can be considered as a rather artificial stimulation protocol. To enable long-term recordings, we monitored fEPSPs in the apical dendritic layer of area CA1 elicited by stimulating the Schaffer collateral-commissural fibers. The initial slope of the fEPSPs was used as a measure of synaptic responses. In the first series of experiments, we tested whether a genetic reduction of VGLUT1 to about 60% in VGLUT1+/− mice has detrimental consequences on basic synaptic transmission. As shown in Figure 3A, recordings from slices of these mice tended to grow faster in response to an increasing stimulation strength, but this was not significant (Mann–Whitney U-test). They attained a maximum at about 2.25 V, whereas wild-type recordings further rose resulting in a significant difference at 2.75 V (P = 0.04 Mann–Whitney U-test).
Next, we studied whether PPF was affected in VGLUT1+/− mice. This presynaptically governed form of short-term plasticity is observed when 2 identical stimuli are delivered to homosynaptic afferent fibers in rapid succession (Anderson 1960; Curtis and Eccles 1960). Over the whole range of IPIs (10–500 ms), no significant difference was observed in PPF between +/− and their littermate controls, although VGLUT1+/− mice tended to achieve smaller PPF values at IPIs of 10–40 ms (Fig. 3B). Although there were no significant changes in short-term plasticity (PPF) and only a minor difference in basic excitability between the 2 genotypes, analysis of LTP revealed a marked difference. As depicted in Figure 3E, wild-type animals displayed an initial potentiation of 263.8 ± 11.2% (n = 11) after applying the first TBS train, which further increased to 270.8 ± 11.2% after completion of the third TBS. In contrast, VGLUT1+/− mice attained only a potentiation that was significantly smaller (first TBS: 223.9 ± 6.0%; second 230.6 ± 5.6%; third 243.6 ± 4.7%, n = 9, P = 0.018 for the induction phase LTP1–30 min, RM-ANOVA). Also of note, in wild types no further increase in the magnitude of potentiation could be observed between the second and third TBS trains indicating a saturation of potentiation. VGLUT1+/− mice, however, did not show any signs of saturation and displayed higher potentiation with every further TBS train. Although there was a clear difference between VGLUT1+/+ and +/− mice in LTP induction, the maintenance was not affected by the reduction in VGLUT1. Thus, the absolute difference between the potentiation of both groups did only slightly decline (third TBS [21 min]: 27.2%; 3 h: 18.7%) and both groups displayed almost the same decay kinetics as evaluated by a curve fit with a one-phase exponential decay function (y = yo + a*e−t/tau) yielding very similar decay-time constants tau: +/+ 67.07 ± 16.83 min; +/− 63.01 ± 16.67 min; computed values and 95% confidence interval). After 3 h, the potentiation of VGLUT1+/+ and VGLUT1+/− mice was 157.9 ± 8.8% and 139.2 ± 8.8%, respectively. The calculated plateau values (yo) of the curve fit were also clearly different with nonoverlapping 95% confidence intervals (+/+ 145.5 ± 8.3%; +/− 127.9 ± 7.3%). Overall, there was a significant difference between LTP1–180 min in +/− and +/+ mice (P = 0.047 RM-ANOVA). Thus, VGLUT1+/− mice displayed both lower initial potentiation and lower values during LTP maintenance.
To check whether the lower initial potentiation of VGLUT1+/− mice was due to a faster fatigue of potentials during tetanization, we analyzed the recordings during tetanization in more detail using the area under the first potential of every TBS burst (normalized to the area of the first potential of burst one) as a measure. As compared with other measures such as amplitude and slope, the area is more closely related to the attained depolarization at a particular time point of tetanization. Figure 3C illustrates that TBS stimulation resulted in a marked decay of the values from burst 1 to burst 10. Furthermore, the decrease is becoming faster with every additional TBS application (first Tet → third Tet) indicating an increasing presynaptic fatigue. Importantly, during all 3 TBS applications, potentials of VGLUT1 deficient mice decayed faster during the initial phase of TBS than the values of WT littermates, which resulted in a significant difference between the 2 genotypes (P < 0.001 RM-ANOVA).
For comparison, we also examined the effects of a reduction of VGLUT2 expression. In agreement with the low expression of this transporter in the adult hippocampus (Herzog et al. 2001, 2006; Nakamura et al. 2005) we did not observe any difference between the 2 genotypes. Thus, neither basal excitability as evaluated by input–output properties, nor short-term plasticity (data not shown) or LTP displayed any significant differences (LTP: first TBS: +/+ 222.5 ± 8.3%, n = 12, +/− 224.9 ± 5.8%, n = 13; second TBS: +/+ 232.2 ± 8.8%, +/− 239.7 ± 7.1%; and third TBS: +/+ 236.2 ± 9.2%, +/− 244.6 ± 7.5%. The similar LTP-induction phase of both genotypes was followed by an almost identical LTP maintenance (3 h: +/+ 131.9 ± 5.3%, +/− 128.8 ± 4.4%).
Morris Water Maze
The impairment of hippocampal synaptic plasticity in VGLUT1-deficient mice suggests a role of this transporter in hippocampus-dependent types of learning. Therefore, we examined spatial learning of VGLUT1+/− mice and their wild-type littermates in a Morris water maze. In this test, animals have to learn to use distant cues on the walls to find a platform that is hidden just underneath the surface of opacified water in a circular pool. As compiled in Figure 4A, all mice learned to find the hidden platform equally fast over 10 trial blocks. Multivariate RM-ANOVA analysis of the latency to find the hidden platform revealed no significant effect for factor “group” (F1,669=0.459, P > 0.1), a significant effect for factor “trial block” (F9,669=174.52, P < 0.001), with no significant interaction for group × trial block (F9,669 = 1.15). Other measures such as path length or thigmotaxis were not different between the 2 tested groups (data not shown).
After 5 and 10 trial blocks, interspersed probe trials were performed with the platform removed (P1 and P2, Fig. 4A). As evidenced by the significant increase in time spent in target quadrant at P1 (42 ± 1.9 5 and 43 ± 2.7% for VGLUT1+/+ and VGLUT1+/−, respectively) all groups learned the platform location after 5 acquisition trial blocks. RM-ANOVA of time spent in pool quadrants indicated a significant effect for “quadrant” (F3,267 = 57.28, P < 0.001) but no significant difference between the groups. Similarly, in the second probe trial, animals spent significantly more time (50%; RM-ANOVA F3,267 = 172.6, P < 0.001) in the target quadrant but no significant difference between the groups could be detected.
Because reversal trials, where the platform is moved to the opposite quadrant, were found to be very susceptible to even subtle functional changes in genetically modified mice (e.g., Balschun, Wolfer, et al. 1999) we tested VGLUT1+/+ and VGLUT1+/− after platform reversal. Analysis of reversal trial performance showed that both genotypes learned the novel platform position, as indicated by a decreasing average distance to the novel platform position with repeated reversal trials (Fig. 4B; RM-ANOVA for factor trial: F2,146 = 45.30, P < 0.001). Interestingly, when we compared the trials that started from the position furthest away from the platform (swim 1, 2, and 5), the average distance to the platform during the reversal trials was significantly increased in VGLUT1+/−. RM-ANOVA revealed a significant effect for genotype (F1,146 = 6.061, P < 0.05). However, the interaction of factors trial × genotype was not significantly different (F2,146 = 0.62, P = 0.5), indicating that the average distance from the platform for VGLUT1+/− was increased on all 3 reversal trials compared with VGLUT1+/+. Pairwise multiple comparison procedure (Holm–Sidak method) revealed that in the first reversal trial, the distance to the platform was significantly increased in VGLUT1+/− (t = 2.6, P = 0.01). Interestingly, when we analyzed the time spent close to the wall in former target (T) or current platform (O) quadrant (see insert in Fig. 4C), we found a significant effect for trial (F2,146 = 18.1, P < 0.001) and a significant group × trial interaction (F2,146 = 5.82, P < 0.005), but the effect for genotype was marginal (F1,146 = 3.14, P = 0.083) (Fig. 4C). All pairwise multiple comparison indicated that VGLUT1+/− mice spent significantly more time in T zone only in the first reversal trial than VGLUT1+/+ (t = 3.57, P < 0.005), but caught up by reversal trials 2 and 3 (P > 0.7). Because this could be due to a different spatial search strategy of VGLUT1+/− mice, we constructed average heat plots for the 2 genotypes from the normalized raw tracking data. To visualize differences between the 2 groups, a differential plot was generated by subtracting VGLUT1+/− from VGLUT1+/+. Figure 4D illustrates that VGLUT1+/− swam more along the wall of the former target quadrant, whereas VGLUT1+/+ showed a swim pattern closer to the novel platform position.
Because biochemical investigations failed to reveal major functional differences between the 3 yet identified VGLUTs, it has been tempting to address these questions by silencing each particular VGLUT subtype genetically. Apart from the one described here, 2 other VGLUT1-deficient strains have been raised independently that display growth retardation, behavioral impairments and a progressive neurological phenotype that includes blindness, hunched posture, and incoordination (Fremeau et al. 2004; Wojcik et al. 2004). In accordance with our data, none of these mutant lines showed a compensatory upregulation or changes in the distribution of VGLUT2 and VGLUT3, respectively (Fremeau et al. 2004; Wojcik et al. 2004). Furthermore, the neurological and behavioral phenotype is not accompanied by gross anatomical variations as evidenced by morphological examinations with different stainings and immunohistochemical markers (Fremeau et al. 2004; Wojcik et al. 2004). However, lethality is “constitutive” in 2 VGLUT1-deficient strains at an age of 2–3 weeks (Fremeau et al. 2004; Wojcik et al. 2004; present paper), but can be circumvented in a third strain (Fremeau et al. 2004; Wojcik et al. 2004). This difference could be attributed to 1) differences in genetic background as either JM1 (Fremeau et al. 2004; Wojcik et al. 2004), 129/ola (Fremeau et al. 2004; Wojcik et al. 2004), or 129 Sv/Evbrd(LEX1) (present paper) ES cells were used, 2) different targeting strategies that might result in different sets of flanking alleles, and/or 3) differences in subsequent breeding schedule. The VGLUT2-deficient mouse line, used in the present study for the comparison with VGLUT1+/−, could only be used as heterozygotes, because VGLUT2−/− mice die immediately after birth (Moechars et al. 2006).
In addition to their largely complementary anatomical distribution, VGLUT1 and VGLUT2 were found to segregate to synapses with different physiological properties, which is probably due to differences in the cytoplasmic C terminus of both subtypes that result in contrasting protein–protein interactions during synaptic targeting (Fremeau et al. 2004). Although the presence of either VGLUT1 or VGLUT2 does not seem to be decisive for the particular release probability of a presynaptic terminal (Wojcik et al. 2004), VGLUT1 is predominantly found at synapses known to show low probability of transmitter release and long-term synaptic plasticity. VGLUT2, in contrast, is mainly present at synapses with higher release probability in sensory and autonomic pathways that display high-fidelity neurotransmission (Fremeau et al. 2001; Varoqui et al. 2002; Nakamura et al. 2005). Opposite findings have been reported, for example, for the cochlear nucleus where VGLUT2 appears to support synaptic plasticity in afferent projections from the spinal trigeminal nucleus (Zhou et al. 2007).
Although Fremeau et al. (2004) reported near absence of fEPSPs in adult homozygous VGLUT1−/− mice, the reduction of VGLUT1 protein to about 60% in our heterozygotes did not lead to significant changes in the responses to a broad range of stimulation intensities. Because the reduced expression of VGLUT1 in our mice did not result in any noticeable compensatory upregulation of either VGLUT2 or VGLUT3, these findings indicate that reduction of VGLUT1 protein by about 40% does not affect basal transmission in adult mice. This is supported by data of another VGLUT1-deficient mouse line where a 50% reduction of transporter protein in 3-week-old animals was not accompanied by any changes in input–output properties of hippocampal neurons (Fremeau et al. 2004). In the present study, a significantly lower fEPSP slope at the highest stimulation strength of the I/O-curve and a markedly reduced initial magnitude of LTP in VGLUT1+/− mice suggest deteriorated response ability to either high-strength activation or bursts of stimulation. This was confirmed by a detailed analysis of the decay of responses during repeated tetanic stimulation. Potentials of VGLUT1+/− mice showed a significantly faster decay during tetanic stimulation, which is indicative of an increased synaptic fatigue. Thus, in contrast to the paired-pulse experiments, where the application of just 2 stimuli within a short-time interval does not represent a major challenge for the impaired vesicle recycling in VGLUT1+/− mice, during long trains of stimulation the functional impairment becomes discernable. Glutamatergic synapses are only equipped with a small pool of recycling vesicles, which requires a critical number of VGLUT molecules to ensure the rapid refilling necessary (Dobrunz and Stevens 1997; Harata et al. 2001). Thus, the reduced number of transporters in heterozygotes does not seem to replenish glutamate properly into the respective pool of vesicles. Tordera et al. (2007) reported unchanged glutamate and reduced γ-aminobutyric acid levels in the hippocampus of VGLUT1+/− mice, but they used a method that does not distinguish between synaptic and extrasynaptic transmitter levels.
It should be noted that studies that examined the reduced excitatory synaptic potentials after VGLUT1 silencing have not been consistent. In one VGLUT1-deficient mouse line, deletion of this transporter reduced both miniature excitatory postsynaptic current size and frequency as well as quantal size in hippocampal cultures (Wojcik et al. 2004). In acute hippocampal slices from another VGLUT1-deficient line (Fremeau et al. 2004), quantal size and release probability were unchanged and only the frequency of spontaneous release was significantly reduced. The latter was taken as an indication that loss of VGLUT1 affects a subpopulation of release sites, which reduces the frequency of release (Fremeau et al. 2004). Further electron-microscopic analysis suggested a specific loss of reserve pool vesicles in VGLUT1-deficient mice (Tordera et al. 2007). Therefore, the diminished LTP amplitude in VGLUT1+/− mice is most likely a consequence of reduced glutamate release during high-frequency stimulation caused by either a lower filling level of individual vesicles or a deteriorated vesicular recycling.
In our study, the electrophysiological examinations of hippocampal slices of VGLUT2+/− mice did not reveal any functional aberration. This is in accordance with data of autaptic hippocampal cultures of newborn VGLUT2−/− and VGLUT2+/− mice where the differential deficiency of VGLUT2 did not result in any discernible electrophysiological phenotype (Moechars et al. 2006). Furthermore, mature VGLUT2+/− mice did not show any change in hippocampus-dependent types of learning (Moechars et al. 2006). These findings indicate that the transient hippocampal expression of this VGLUT subtype during early development does not influence adult hippocampal physiology.
Because of the impairment of hippocampal synaptic plasticity in VGLUT1-deficient mice, it was tempting to test whether these deficits are accompanied by changes in hippocampus-dependent spatial learning. Such a functional link between hippocampal LTP and hippocampus-dependent learning has been questioned recently. Despite the prominent role of the glutamatergic system, some genetic manipulations of crucial components of this system were found to affect only hippocampal LTP but not spatial learning (and vice versa). For example, gene-targeted mice lacking the alpha symbol-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit GluR1 (GluR-A), a major constituent of AMPA receptors in CA1 pyramidal cells, showed a severe impairment of CA1-LTP but normal spatial reference memory in the Morris water maze (Zamanillo et al. 1999). Further investigation of these mice, however, revealed a specific deficit in spatial working memory (Reisel et al. 2002) and the presence of a late-onset CA1-LTP that is inducible by paired pre-and postsynaptic stimulation at the theta frequency (Hoffman et al. 2002). These results partially restored the link between hippocampal LTP and spatial learning.
In our study, VGLUT1+/− mice were rather inconspicuous in the acquisition phase and probe trial of the Morris water maze task, showing spatial learning abilities comparable with their wild-type littermates. Although VGLUT1+/− mice were able to acquire and use a spatial map in the standard version of the task, they showed problems to modify and reuse this spatial map once the platform was moved to the opposite quadrant. Our findings are corroborated by a study of (Tordera et al. 2007) who found no changes in water maze learning of another strain of VGLUT1 deficient mice (Wojcik et al. 2004) using a protocol without platform reversal. Hence, water maze protocols that include a platform reversal appear to be more sensitive to changes in hippocampal function than the “standard” task. For example, as shown by D'Adamo et al. (2004), mice lacking Rab3a, a protein that is involved in presynaptic function and transmitter release, display normal water maze acquisition and probe trial performance but a significant impairment in learning new platform positions. Likewise, deterioration of presynaptic function in VGLUT1+/− mice resulted in a more rigid behavior as compared with littermate controls implying a decreased flexibility to adapt an acquired spatial map to a new platform position. A decrease in synaptically released glutamate and the expression of VGLUT1 was implicated recently in the age-dependent decline of water maze performance in transgenic APdE9 mice carrying mutated human amyloid precursor protein and presenilin genes (Minkeviciene et al. 2008).
In conclusion, our data support a dominant role for VGLUT1 but not VGLUT2 in hippocampal synaptic plasticity. Evaluated together with the specific deficit in spatial reversal learning our data suggest a functional role of VGLUT1 in forms of hippocampal synaptic plasticity that are involved in the adaptation and modification of acquired spatial maps to external stimuli and changes in the environment.
Katholieke University Leuven (Impulse Fund IMP/04/006; OT/06/23 to D.B.), Fonds voor Wetenschappelijk Onderzoek (FWO grants; G.0271.06 and G.0496.07 to D.B. and R.D.); IWT/J&J-PRD research and development grant (to Z.C.V.).
We thank Kathrin Böhm, Ilse Goris, and Guy DaneeIs for excellent technical assistance; Patrick Callaerts for helpful suggestions; and Klaus Reymann, FAN gGmbH Magdeburg, Germany, for support. Conflict of Interest: None declared.