Hippocampal proteoglycans brevican and versican are linked to spatial memory of Sprague–Dawley rats in the morris water maze



Proteoglycans (PGs) are major constituents of the extracellular matrix and have recently been proposed to contribute to synaptic plasticity. Hippocampal PGs have not yet been studied or linked to memory. The aim of the study, therefore, was to isolate and characterize rat hippocampal PGs and determine their possible role in spatial memory. PGs were extracted from rat hippocampi by anion-exchange chromatography and analyzed by nano LC-MS/MS. Twenty male Sprague–Dawley rats were tested in the morris water maze. PGs agrin, amyloid beta A4 protein, brevican, glypican-1, neurocan, phosphacan, syndecan-4, and versican were identified in the hippocampi. Brevican and versican levels in the membrane fraction were higher in the trained group, correlating with the time spent in the target quadrant. α-amino-3-hydroxy-5-methylisoxazole-4-propionate receptor GluR1 was co-precipitated with brevican and versican. Levels for a receptor complex containing GluR1 was higher in trained while GluR2 and GluR3-containing complex levels were higher in yoked rats. The findings provide information about the PGs present in the rat hippocampus, demonstrating that versican and brevican are linked to memory retrieval in the morris water maze and that PGs interact with α-amino-3-hydroxy-5-methylisoxazole-4-propionate receptor GluR1, which is linked to memory retrieval.


Proteoglycans (PGs) are major constituents of the extracellular matrix of the brain and were proposed to contribute to synaptic plasticity. This report addressed PGs in rat hippocampus and suggests that PGs brevican and versican are linked to spatial memory, and form a complex with the GluR1 subunit of the AMPA receptor, a key signaling molecule in memory mechanisms.

Abbreviations used



enhanced chemiluminescence


long-term potentiation




perineuronal net

The brain's structural plasticity is largely modified during its development and maturation, and also at the time of remodeling processes in adulthood. In adult hippocampus, synaptic plasticity in association with memory formation is considered a prerequisite for the normal function of this brain area. This process involves cellular signaling from extracellular environment to the cell nucleus. Brain synapses are in a wrapped form, a dense meshwork of extracellular matrix comprising lectican-bound hyaluronic acid and tenascin, known as perineuronal net (PNN) (Celio et al. 1998), maintaining excitatory/inhibitory homeostasis in the adult brain (Hensch 2005). PNNs are diminished in patients with schizophrenia (Pantazopoulos et al. 2010) but preserved in patients with Alzheimer′s disease (Morawski et al. 2012). Loss of PNNs is associated with aberrant hippocampal activity and abnormal dopaminergic function (Shah and Lodge 2013).

More than 15 proteoglycans are reported in the central nervous system (Bandtlow and Zimmermann 2000), however, no systematic studies have been carried out to identify proteoglycans from hippocampus. Although the pivotal role of proteoglycans in synaptic plasticity has been proposed (Daniels 2012; Orlando et al. 2012; Senkov et al. 2012), their role in cognitive functions including memory formation and storage remain elusive. Knockout studies of the proteoglycan brevican revealed that it was not involved in memory formation (Brakebusch et al. 2002). Nevertheless, PNNs were shown to prevent fear memory from extinction (Gogolla et al. 2009) and affected lateral diffusion of α-amino-3-hydroxy-5-methylisoxazole-4-propionate (AMPA) receptors (Frischknecht et al. 2009), key elements of signaling in memory processes (Krugers and Hoogenraad 2009; Conboy and Sandi 2010; Rust et al. 2010; Yuen et al. 2011). These few studies suggest that PNN-associated proteoglycans may play a role in learning and memory. AMPARs execute their function through assembly to complexes rather than subunits and two major populations of AMPA receptor complexes are reported from rat hippocampus: Glutamate receptor subunits (GluR)1-GluR2 and GluR2-GluR3 and a limited population of GluR1-GluR3 complexes (Wenthold et al. 1996). The GluR1-GluR2 complex is trafficked to the membrane upon synaptic activity, while GluR2-GluR3 complexes undergo constitutive cycling (Emond et al. 2010). Therefore, it is challenging to study the link between expression of memory-linked proteoglycans and membrane AMPA receptor complexes for better understanding of the molecular mechanisms involved in memory mechanisms.

In this study, proteoglycans were isolated by anion-exchange chromatography from rat hippocampus and identified by mass spectrometry. Selected proteoglycans were quantified from hippocampi of rats that were tested in the spatial paradigm morris water maze (MWM). Modification of AMPAR complexes were analyzed and shown to interact with proteoglycans brevican and versican.

Materials and methods


Rats were bred and maintained in Makrolon cages, supplied with autoclaved woodchips, standard rodent diet (Altromin, Lage, Germany), and water in the Core Unit of Biomedical Research, Division of Laboratory Animal Science and Genetics, Medical University of Vienna. Room temperature was 22 ± 1°C and relative humidity was 50 ± 10%. The light/dark rhythm was 14 : 10 h. Ventilation with 100% fresh air resulted in an air change rate of 15 times per hour. The room was illuminated with artificial light at an intensity of about 200 lx in 2 m from 5 AM to 7 PM Behavioral tests were performed between 8 AM and 1 PM. All procedures were carried out according to the guidelines of the European Communities Council Directive of 24 November 1986 (86/609/EEC) and were approved by Federal Ministry of Education, Science and Culture, Austria.

Morris water maze

Ten male Sprague–Dawley rats, 12–14 weeks old were trained in four daily acquisition sessions (four trials per day with 20 min inter-trial interval) to find a hidden platform in a circular pool (150 cm diameter, wall depth 60 cm) filled with water (21 ± 1°C). Rats that failed to locate the platform within 120 s were manually placed on the platform and allowed to remain on it for 30 s. Memory retrieval was assessed on day 5 (60 s, platform removed) (Sase et al. 2013). Ten rats spending the same time in the MWM without training were used as controls. All experiments were recorded by computerized tracking/image analyzer system (video camcorder: 1/3' SSAM HR EX VIEW HAD coupled to the computational tracking system: TiBeSplit). The rats were bred and maintained at the Core Unit of Biomedical Research, Division of Laboratory Animal Science and Genetics, Medical University of Vienna.

Isolation and identification of proteoglycans

Rat hippocampi were homogenized in a buffer containing 50 mM Tris (pH 8.0), 0.15 M NaCl, 1% {3-[(3-Cholamidopropyl)dimethylammonio]-1-propanesulfonate}, and one complete protease inhibitor tablet [(Roche Molecular Biochemicals, Mannheim, Germany) per 50 mL]. Following incubation at 4°C for 1 h, samples were centrifuged at 20 000 g for 45 min. The pellet was discarded, while the supernatant was retained for further analysis. The efficiency of proteoglycan extraction is shown in Figure S2a. Proteoglycans (PGs) were isolated by using HiTrap DEAE Sepharose FF according to the manufacturer′s instructions (17-5055-01; GE Healthcare, Freiburg, Germany). Isolated PGs were treated with chondroitinase ABC (C3667-5UN; Sigma, St Louis, MO, USA) (Herndon and Lander 1990) in order to cleave glycosaminoglycans to allow electrophoretic separation. Resulting PG proteins were separated by 8% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and in-gel digested peptides were identified by Nano-LC-ESI-CID/ETD-MS/MS (Ghafari et al. 2012a,b; Wan et al. 2012).

Hippocampal membrane protein preparation

All procedures were carried out at 4°C. Hippocampi were homogenized in ice-cold homogenization buffer [10 mM HEPES, pH 7.5, 300 mM sucrose, one complete protease inhibitor tablet (Roche Molecular Biochemicals) per 50 mL] by Ultra-Turrax (IKA, Staufen, Germany). The homogenate was centrifuged for 10 min at 1000 g and the pellet was discarded. The supernatant was centrifuged at 50 000 g for 30 min in an ultracentrifuge (Optima- L-90K, Beckman Coulter Inc., Brea, CA, USA). The pellet was resuspended in washing buffer (homogenization buffer without sucrose), kept on ice for 30 min and centrifuged at 50 000 g for 30 min to obtain a membrane fraction. Extraction buffer (1.5 M 6-aminocaproic acid, 300 mM Bis–Tris, pH 7.0) with 1% n-Dodecyl β-d-maltoside was added to the membrane pellets, incubated for 1 h with vortexing every 10 min. Following solubilization, samples were cleared by centrifugation at 20 000 g for 1 h. The protein content was estimated using the bicinchoninic acid protein assay kit (Pierce, Rockford, IL, USA). Extracted proteins were then aliquoted and stored at −80°C till use (Sase et al. 2012).

Gel electrophoresis and western blotting

For the quantification of AMPA receptor complexes and PGs hippocampal membrane fractions were used. AMPA receptor complexes were separated on 5–13% of blue native polyacrylamide gel electrophoresis (BN-PAGE) as described previously (Sase et al. 2012) while PGs were separated on 8% sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Separated proteins were transferred to polyvinylidene difluoride membranes and blocked with 5% of non-fat dry milk in Tris-Buffered Saline and Tween 20 for 1 h. Primary antibodies were added and incubated overnight at 4°C. After membranes were washed for six times with Tris-Buffered Saline and Tween 20, corresponding secondary species-specific horseradish peroxidase-conjugated antibodies against IgG were added and incubated for 1 h at 21°C (Antibody details are given in Table S1). Membranes were developed with the enhanced chemiluminescence (ECL) solution (Clarity Western ECL Substrate, Bio-Rad) and hyperfilms (Amersham Hyperfifilm ECL, GETM Healthcare, GE Healthcare Europe GmbH, Vienna, Austria) were used. Eight to ten animals per group were used for quantification analysis. A rectangular area of immunoreactive bands was selected and arbitrary units of optical density were measured using the software Image J (http://imagej.nih.gov/ij/features.html or http://howtowesternblot.net/data-analysis-3/quantification/). Proteins from bands at the apparent molecular weights in the individual lanes from 35 kDa to 150 kDa for proteoglycans and from 720 kDa to 1236 kDa for AMPA receptors were used for normalization from coomassie blue R-350 stained membranes (Figure S1) (Welinder and Ekblad 2011; Lee et al. 2013). The intensity of individual immunoreactive band was divided by the corresponding lane intensity of coomassie blue R-350 stained membrane to get the normalized values. Selection of these particular molecular ranges would give low variations between samples and groups. This does not rule out, however, that learning and memory may change total protein levels in the gels although this fact has never been addressed in neurochemistry so far.

Immunoprecipitation and antibody-shift assay

For immunoprecipitation (IP) experiments, 3 μg of GluR1 (AB31232; Abcam, Cambridge, MA, USA), 5 μg of GluR2 (AB1768; Millipore Corporation, Bedford, MA, USA), 10 μg of GluR3 (3437; Cell Signaling Technology, Beverly, MA, USA), 20 μg of brevican (610895; BD Biosciences, San Jose, CA, USA), and versican (sc-25831; Santa Cruz Biotechnology, Santa Cruz, CA, USA) antibodies were used per 500–1000 μg of hippocampal membrane proteins. Anti-cAMP response element-binding protein antibody (Cell Signaling) was used for control IP (Figure S2b). IP procedures were carried out according to the manufacturer′s instructions (Thermo Scientific Pierce Direct IP Kit, Rockford, IL, USA). For antibody-shift assays, 0.5 μg of either GluR1 or brevican antibodies were added to 100 μg of membrane proteins, incubated for 1 h at 4°C and then separated by 5–13% of BN-PAGE. GluR1 and brevican antibody-treated samples were probed with brevican and GluR1 antibodies, respectively.

Statistical analysis

One-Way anova combined with post hoc multiple comparison test (Bonferroni) was used to analyze latency, path length, and speed. Time spent in target quadrant and western blot data were analyzed by an unpaired t-test. Pearson correlation coefficients were used for correlation studies. A probability level of p < 0.05 was considered statistically significant. Calculations were performed using Graphpad Prism 6.


Identification of rat hippocampal proteoglycans by nano LC-MS/MS

From 12 mg of total hippocampal protein ~ 15 μg of PGs were isolated. In order to get sufficient amounts of PGs for mass spectrometric analysis, five hippocampi were pooled for PGs isolation as described above. Agrin, amyloid beta A4 protein, brevican, glypican-1, neurocan, phosphacan, syndecan-4, and versican were identified by mass spectrometry. Tenascin-R, a major component of PNN glycoproteins, was also identified (Table 1). Information on peptides for unambiguous identification of proteoglycans is given in Table S2.

Table 1. Mass spectrometric-based identification of proteoglycans and tenascin-R from rat hippocampus
Gene IDProtein NameSwiss Prot No.
AppAmyloid beta A4 proteinP08592
Gpc1Glypican 1P35053

Changes of PG levels upon spatial memory retrieval

Spatial memory was accessed by the MWM. During the training period of 4 days, latencies to reach the platform (Fig. 1a) and path lengths (Fig. 1b) were significantly reduced as animals learned the task. Speed remained comparable between training days (Fig. 1c). On day 5, the time spent in the target quadrant was significantly increased in the trained animals as compared to remaining quadrants during the probe trial (Fig. 1d). Times spent in the target quadrant were comparable to the other quadrants in the yoked group (Fig. 1e).

Figure 1.

Morris water maze. Rats were trained in morris water maze (MWM). Time taken to reach the platform (latency; a), distance traveled to reach platform (pathlength; b) and the average speed (c) during training days were calculated with One-way ANOVA analysis. (d) Unpaired t-test was performed between time spent in the target quadrant and remaining quadrants of trained animals during probe trails. (e) Unpaired t-test was performed between time spent in the target quadrant and remaining quadrants of yoked group during probe trails. Values are means ± SD. ****p < 0.0001.

Membrane fraction, which contains both PNN-associated proteoglycans (Shah and Lodge 2013) and membrane-bound proteoglycans, were subjected to analyze memory-related proteoglycans as PNN had been shown to play a critical role in fear memory extinction (Gogolla et al. 2009) and lateral diffusion of AMPA receptor (Frischknecht et al. 2009). Agrin, brevican, glypican-1, phosphacan, tenascin-R, and versican were selected for western blotting-based quantitative analysis between yoked and trained animals. Western blot patterns are given in Fig. 2. Brevican bands at 145 kDa and 50 kDa and versican levels were significantly increased in hippocampi of trained animals as compared to yoked animals while the > 145 kDa band of brevican was not increased significantly (Table 2). Pearson correlation studies showed significant positive correlation between brevican and versican and the time spent in the target quadrant (Fig. 3a–c). So far PG changes were only demonstrated in the membrane fraction which are not necessarily because of changed synthesis or secretion but may simply be because of changed solubility or anchoring at membranes.

Table 2. Quantification of proteoglycans and tenascin-R in yoked and trained rat hippocampus
PGsYoked Mean ± SD, nTrained Mean ± SD, np value
Agrin 200 kDa0.2978 ± 0.0917, 90.3169 ± 0.1048, 80.6861
Agrin 90 kDa0.1934 ± 0.157, 90.2394 ± 0.088, 80.476
Brevican > 145 kDa0.1424 ± 0.064, 90.1951 ± 0.073, 100.118
Brevican 145 kDa0.1255 ± 0.045, 90.1928 ± 0.046, 100.0071
Brevican 50 kDa0.5702 ± 0.111, 90.7813 ± 0.2496, 100.0251
Glypican 1 1.4669 ± 0.7612, 81.4451 ± 0.4902, 80.9432
Phosphacan 0.0596 ± 0.0213, 90.0793 ± 0.036, 80.2058
Tenascin-R0.1233 ± 0.0315, 100.1174 ± 0.0179, 100.6204
Versican 0.0468 ± 0.0087, 100.1211 ± 0.0776, 100.0075
Figure 2.

Western blot imaging of proteoglycans and tenascin-R. Hippocampal membrane proteins were prepared from morris water maze (MWM) trained and yoked groups. Hippocampal membrane proteins of yoked and trained animals were separated on 8% of sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and probed with indicated proteoglycan and tenascin-R antibodies.

Figure 3.

Pearson correlations. Learning and memory altered proteoglycans were statistically correlated with the time spent by the rats in the target quadrant of MWM. Pearson's correlations between the 145 kDa band of brevican and the time spent in the target quadrant (a), the 50 kDa band of brevican and the time spent in the target quadrant (b), and versican levels and time spent in the target quadrant (c).

Brevican and versican form a complex with GluR1

Although PNNs were shown to affect AMPA receptor (AMPAR) mobility, to the best of our knowledge direct interaction between PNN-associated proteoglycans and AMPARs was not reported yet. IP was performed using brevican and versican antibodies as both PGs were significantly modified in spatial memory retrieval and subsequently probed with a GluR1 antibody. Results from IP showed that brevican (145 kDa) and versican were co-immunoprecipitated with GluR1 suggesting that these proteins are in a complex (Fig. 4a). However, 50 kDa band of brevican was not detected with GluR1 co-immunoprecipitation. In addition, antibody-shift assays were performed to confirm that GluR1 and brevican are in the same complex. Membrane protein fractions were pre-incubated with an antibody against either GluR1 or an antibody against brevican before separation by BN-PAGE. Samples pre-incubated with GluR1 antibody were probed using an anti-brevican antibody and extracts pre-incubated with brevican using an antibody against GluR1, respectively. The clear shifts in antibody-incubated lanes confirmed that GluR1 and brevican were in the same complex (Fig. 4b).

Figure 4.

Complex analysis. (a) Immunoprecipitation was performed from hippocampal membrane fraction using brevican and versican antibodies and subsequently probed with indicated antibodies. (b) Antibody-shift assay; in the first panel hippocampal membrane proteins were incubated with GluR1 antibody, separated on 5–13% blue native polyacrylamide gel electrophoresis (BN-PAGE) and probed with brevican antibodies. In the second panel membrane, proteins were incubated with brevican or GluR1 antibody, separated on 5–13% BN-PAGE and probed with GluR1 antibody. Membrane proteins without antibody were used as controls (first lanes). (c) Hippocampal membrane proteins of yoked and trained animals were separated on 5–13% of BN-PAGE and probed with antibodies against AMPA receptor subunits GluR1-3. Note: In the western blot image of GluR2 and GluR3, blank spaces indicate that no samples were loaded as wells were damaged.

Modification of AMPAR complex levels in spatial memory formation

AMPAR complex level changes between trained and yoked groups were analyzed. Western blot bands are shown in Fig. 4c. Levels of a GluR1-containing receptor complex were significantly increased in the trained group while levels of GluR2-and GluR3- containing receptor complexes were significantly increased in the yoked group (Table 3). Pearson correlation studies showed that levels of the GluR1-containing complex were positively correlating while levels of the GluR3-containing complex were negatively correlated with the time spent in the target quadrant (Fig. 5a and b) GluR1-containing receptor complex levels correlated significantly with levels of the 145 kDa form of brevican (Fig. 5c) while complex levels containing GluR3 correlated significantly with versican (Fig. 5d).

Table 3. AMPAR complexes in yoked and trained rat hippocampus
AMPAR complexesYoked Mean ± SD, nTrained Mean ± SD, np value
GluR1- complex0.1499 ± 0.0158, 100.1832 ± 0.012, 10< 0.0001
GluR2- complex 0.0168 ± 0.0059, 100.0111 ± 0.0054, 100.0388
GluR3- complex 0.1153 ± 0.0321, 100.0404 ± 0.009, 10< 0.0001
Figure 5.

Pearson correlations. Pearson's correlations were performed between time spent in the target quadrant and significantly altered AMPA receptor complexes. To link AMPA receptor complexes with proteoglycans, correlations were also performed between significantly altered proteoglycans and AMAP receptor complexes. Correlations between the GluR1-containing complex levels and the time spent in the target quadrant (a), GluR3-containing complex levels and the time spent in the target quadrant (b), GluR1-containing complex levels and levels of the 145 kDa band of brevican (c), and GluR3-containing complex levels and versican levels (d).


The most salient findings of this study show that a series of PGs are identified in rat hippocampus, and that brevican and versican are linked to spatial memory retrieval. A complex formation of brevican and versican with the AMPAR GluR1 has been observed, along with the fact that AMPARs GluR1- and GluR3-containing complexes correlate with the time spent in the target quadrant, the parameter used for evaluation of retrieval in spatial memory.

As shown in the results, agrin, amyloid beta A4 protein, brevican, glypican-1, neurocan, phosphacan, syndecan-4, tenascin-R, and versican were observed in rat hippocampus by mass spectrometry. Agrin was suggested to be associated with synapse maturation (Ferreira 1999) and amyloid beta A4 protein was proposed to be associated with Alzheimer disease and Down syndrome (Beyreuther et al. 1993). Hypoxic-ischemic brain injury altered brevican and versican expression in rat hippocampus (Aya-ay et al. 2005; Leonardo et al. 2008) and loss of synapses was associated with loss of brevican (Morawski et al. 2012). Neurocan knockout mice showed normal early-phase long-term potentiation (LTP) but late-phase LTP was impaired (Zhou et al. 2001). Proteolytic degradation of phosphacan was linked to neuronal degeneration in the hippocampus (Kurazono et al. 2001) and this PG was shown to contribute to mossy fiber outgrowth and regeneration (Butler et al. 2004). Electrophysiological studies have shown that Tenascin-R is involved in synaptic plasticity (Saghatelyan et al. 2000; Nikonenko et al. 2003). In this study, increased levels of brevican and versican in the trained group may point to involvement of these PGs in spatial memory in the rat. Brakebusch et al. (2002) have shown that brevican-deficient mice present with impaired LTP but learning and memory as evaluated in a MWM were not affected. Their finding of unchanged learning and memory does not contradict our results.

As shown in the Results section, brevican and versican co-precipitated with AMPAR GluR1 and an antibody-shift assay clearly indicated that brevican and GluR1 were in the same complex. This complex formation was not shown before, but glypican 4 and 6 led to clustering of GluR1 on the cell surface (Allen et al. 2012) and the PNN, which mainly consists of PGs, was shown to affect AMPAR trafficking (Frischknecht et al. 2009). It is, however, unclear whether those proteoglycans interacted with GluR1 through core protein or glycosaminoglycan side chains. Like brevican and versican, GluR1-containing complexes were correlating with the time spent in the target quadrant and indeed, GluR1-containing complexes were linked to spatial memory in the mouse (Ghafari et al. 2012a,b) and the link between GluR1 and memory per se is evident and well-documented. The correlation between GluR1-containing complex levels and brevican again would point to a role of this PG in spatial memory retrieval. Versican was found to negatively correlate with GluR3-containing complex levels, which were decreased in trained rats, and positively correlate with the time spent in the target quadrant, probably suggesting a role also for versican during memory retrieval. A previous study demonstrated that GluR3-containing complexes were decreased in spatial memory training (Falsafi et al. 2012), and here it was shown that GluR3-containing complex levels were decreased in memory retrieval. AMPARs exert function by assembling into homo-and hetero-oligomers, and therefore AMPARs complexes rather than subunits were investigated (Emond et al. 2010). The complexity of AMPAR complexes has been addressed by several groups (Wenthold et al. 1996; Reimers et al. 2011; Schwenk et al. 2012) and we, hereby, are adding two proteoglycans, brevican and versican, to the probable interactome of AMPARs.

Taken together, we have shown the presence of several PGs in the rat hippocampus, revealed the association of brevican and versican with AMPARs-containing GluR1 or GluR3 and the time spent in the target quadrant representing memory retrieval performance in the MWM. Moreover, it has been demonstrated that brevican and versican are probable complex components of GluR1- or GluR3-containing complexes. Therefore, we propose a tentative role of brevican and versican in spatial memory either directly or by interaction with AMPARs in the MWM.

Acknowledgments and conflict of interest disclosure

The study was funded by the Medical University of Vienna, Austria.

All experiments were conducted in compliance with the ARRIVE guidelines. There are no conflicts of interest to declare.