Characterization of a mouse model overexpressing beta-site APP-cleaving enzyme 2 reveals a new role for BACE2


  • G. Azkona,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    3. Department of Neuroscience, University of Basque Country (UPV/EHU), E-48003, Bizkaia, Spain
    Search for more papers by this author
  • A. Amador-Arjona,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    Search for more papers by this author
  • C. Obradors-Tarragó,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    Search for more papers by this author
  • E. Varea,

    1. Departments of Histology and Anatomy, University Miguel Hernández, Elche, E-03550, Alicante, Spain
    Search for more papers by this author
  • G. Arqué,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    Search for more papers by this author
  • R. Pinacho,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    Search for more papers by this author
  • C. Fillat,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    Search for more papers by this author
  • S. De La Luna,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    3. Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain
    Search for more papers by this author
  • X. Estivill,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    Search for more papers by this author
  • M. Dierssen

    Corresponding author
    1. Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), and
    2. CIBER de Enfermedades Raras (CIBERER), E-08003 Barcelona, Catalonia, Spain
    Search for more papers by this author

* M. Dierssen, Neurobehavioural Phenotyping of Mouse Models of Disease, Genes and Disease Program, Center for Genomic Regulation (CRG), Barcelona Biomedical Research Park (PRBB), Dr. Aiguader, 88, 08003 Barcelona (Spain). E-mail:


BACE2 is homologous to BACE1, a β-secretase that is involved in the amyloidogenic pathway of amyloid precursor protein (APP), and maps to the Down syndrome critical region of chromosome 21. Alzheimer disease neuropathology is common in Down syndrome patients at relatively early ages, and it has thus been speculated that BACE2 co-overexpression with APP would promote the early neurodegenerative phenotype. However, the in vivo function of BACE2 has not yet been elucidated. The aim of the present work has been to analyse the impact of in vivo BACE2 overexpression using a transgenic mouse model. Our results suggest that BACE2 is not involved in the amyloidogenic pathway, cognitive dysfunction or cholinergic degeneration. However, TgBACE2 animals showed increased anxiety-like behaviour along with increased numbers of noradrenergic neurones in locus coeruleus, thus suggesting an unexpected role of BACE2 overexpression.

Progressive formation and extracellular aggregation of amyloid-β (Aβ) peptide has been considered one of the causal factors for the pathogenesis of Alzheimer's disease (AD) (Yan et al. 2001). This neuropathological hallmark is also common in Down syndrome (DS) patients at relatively early ages and has been attributed to the overexpression of amyloid precursor protein (APP) (Murphy et al. 1990), a large type I transmembrane glycoprotein precursor that maps to human chromosome 21 (HSA21). In the amyloidogenic pathway, a β-secretase cleaves APP to generate APPsβ, a soluble N-terminal fragment and a C-terminal fragment (C99). A γ-secretase cleaves C99 to form the mature Aβ peptide comprising 39–42 amino acids (Mattson 2004). Increased accumulation of the C99 fragment has been observed in the brain of DS individuals (Busciglio et al. 2002; Sun et al. 2006), suggesting that abnormal processing at the APP β-site might be involved in the common degenerative phenotype of DS and AD patients.

One of the principal players in the amyloidogenic pathway is the β site APP cleaving enzyme 1 (BACE1), encoded by a gene located on HSA11 (Sambamurti et al. 2004). BACE1 is expressed in the hippocampus and the cerebral cortex, colocalizes with APP in the Golgi compartment, and its overexpression results in an increase of Aβ peptide (Hussain et al. 1999; Vassar et al. 1999). However, there is still no consistent evidence of its role in the early appearance of DS neurodegenerative phenotypes (Cheon et al. 2008). BACE1 has a paralogous gene in vertebrates, BACE2, which in humans, maps to HSA21 at 21q22.3 in the DS critical region (Acquati et al. 2000; Solans et al. 2000). It has thus been speculated that BACE2 and APP co-overexpression would promote the early appearance of amyloid plaques in DS patients. BACE2 is a transmembrane glycoprotein with aspartic protease activity (Bennett et al. 2000), which overexpression has been reported in DS fetal brain samples (Barbiero et al. 2003), although it is controversial (Cheon et al. 2008). Of interest to AD phenotype, immunoreactivity for BACE2 has been observed in neurofibrillary tangle-bearing neurones from elderly DS brains with AD-type neuropathology (Motonaga et al. 2002).

There has been an intense debate about BACE2 function. In vitro studies showed that BACE2 cleaves APP at the β-secretase site (Farzan et al. 2000; Hussain et al. 2000), although cleavage at the α-secretase site has been also reported (Basi et al. 2003; Farzan et al. 2000; Yan et al. 2001). However, a recent study showed that BACE2 cleaves APP at a novel site named θ, generating APP C80 fragments, an effect that would reduce Aβ production (Sun et al. 2006). It is thus necessary to elucidate the function of BACE2 in vivo and its possible involvement in DS and/or AD phenotypes. Among those, even though the most important is the learning and memory deficit possibly related to cholinergic degeneration, a noradrenergic phenotype has also been reported in AD patients and in AD mouse models, where loss of locus coeruleus (LC) noradrenergic neurones (Bondareff et al. 1987; Kalinin et al. 2007; Mann 1983; O’Neil et al. 2007) may depend on the increase of APP C99.

For the first time, here we show that overexpression of BACE2 in a transgenic mouse model gives rise to significant alterations in the noradrenergic system accompanied by increased anxiety-like behaviour, thus suggesting a novel role for this enzyme.

Materials and methods

General procedures

Animals were housed in standard macrolon cages (four to five animals per cage, 40 × 25 × 20 cm) with freely available food and water in standard environmental conditions (constant humidity and temperature of 22 ± 1°C) and a 12 h light/dark cycle (lights on at 7:00 a.m.). All animal procedures were approved by the local ethical committee (CEEA-IMIM and CEEA-PRBB), and met the guidelines of the local (law 32/2007) and European regulations (EU directive n° 86/609, EU decree 2001-486) and the Standards for Use of Laboratory Animals n° A5388-01 (NIH). The CRG is authorized to work with genetically modified organisms (A/ES/05/I-13 and A/ES/05/14) and the experimenters hold the official accreditation (law 32/2007).

Construction of Thy-1/BACE2HA transgene

The open reading frame corresponding to the longest isoform of human BACE2 (Solans et al. 2000) was C-terminally fused in-frame to the hemaglutinin (HA) tag. A blunted DNA fragment containing these sequences was inserted into the blunted XhoI site of the murine Thy1 cassette (Moechars et al. 1996), which was kindly provided by D. Moechars (Janssen Pharmaceutica NV, Belgium). The orientation of the complementary DNA (cDNA) within the cassette was verified by sequencing with specific primers, and the complete expression cassette was designated Thy1-BACE2.

Generation of transgenic mice and genotyping

Transgenic mice were generated by standard pronucleus microinjection of the 8.7 kb fragment of the Thy1-BACE2 construct, without the plasmid vector sequences. Two transgenic lines were obtained on different genetic backgrounds, to help discern to what extent the genetic background may modify the phenotypic impact of BACE2 overexpression. We used a hybrid C57BL/6J × CBA/J F1 (B6/CBAF1/J; retinal degeneration mutation free) and on a C57BL/6J × SJL F1 (B6/SJLF1/J) genetic background. The presence of the transgene was assessed in DNA from tail biopsies by digestion with BamHI and Southern blot analysis by using a fragment of BACE2 cDNA as a probe. Two transgenic lines were obtained and maintained by backcrossing to a hybrid B6/CBAF1/J or B6/SJLF1/J background in heterozygosity. Genotyping was performed routinely by polymerase chain reaction (PCR) analysis using the primer pairs: BACE2f, 5′-ATCCACAAATGCGCTGGT-3′ and BACE2r, 5′-GCGGCCGTTACTAGTGGA-3′. Hybrid founders were backcrossed extensively in order to attenuate littermate's genetic differences. All experiments were performed in mice from the F15 generations. In all cases, transgenic mice were directly compared with non-transgenic littermates.

RNA expression analysis

For expression analysis of BACE2 messenger RNA (mRNA), total RNA from brain samples of both transgenic lines generated (L2 and L11) and their wild-type animals were isolated with the TriPure kit (Boehringer, Mannheim, Germany). reverse transcription polymerase chain reaction (RT-PCR) was carried out by reverse-transcribing total RNAs (1 μg) using Superscript reverse transcriptase (Gibco BRL, San Francisco, CA, USA), and followed by PCR amplification with primers BACE2f and BACE2r. Absence of genomic DNA contamination was determined by the amplification of a 125 bp PCR fragment from cDNA samples with primers for GdX transcript (GdXf, 5′-GGCAGCTGATCTCCAAAGTCCTGG-3′ and GdXr, 5′-AAC GTTCGATGTCATCCAGTGTTA-3′).

Western blot

For each experiment four wild-type and four TgBACE2 were used (see Figure legends for details of each experiment). Animals were sacrificed, brains rapidly removed and dissected on ice. Tissues were homogenized in lysis buffer (10 mm HEPES pH 7.5, 150 mm NaCl, 1 mm ethylenediaminetetraacetic acid (EDTA), 0.1 mm MgCl2, phosphate-buffered saline (PBS) 0.2% Triton X-100 and a protease inhibitor cocktail (Roche, Mannheim, Germany). After clearance of the lysates by centrifugation (1400 × g, 20 min at 4°C), protein quantification was performed following the bicinchoninic acid (BCA) Protein Assay Reagent (Pierce, Rockford, IL, USA) protocol. Western blot analysis was performed using 50 μg of protein resolved on a 10% sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) and electro-blotting onto nitrocellulose membranes (Hybond-C, Amersham Pharmacia Biotech, Freiburg, Germany). Membranes were blocked with 5% non-fat dry milk in Tris-buffered saline including 0.1% Tween-20 (TBS-T) and incubated with the primary antibodies in 5% non-fat dry milk in TBS-T overnight at 4°C. The following antibodies were used as primary antibodies: goat anti-Bace2 antibody (1:500; D-20; Santa Cruz, Heidelberg, Germany), rabbit anti-HA antibody (1:1000; Eurogenic, Emeryville, CA, USA), mouse anti-tyrosine hydroxylase (TH) antibody (1:15000, Sigma, St Louis, MO, USA) and anti-actin antibody (1:2000; Sigma, St Louis, MO, USA). Incubation with horseradish peroxidase (HRP)-conjugated anti-goat, anti-rabbit or anti-mouse IgG (Pierce, Rockford, IL, USA), followed by enhanced chemiluminescence (ECL, Pierce, Rockord, IL, USA) assay allowed detection. Quantification was made by densitometric analysis of non-saturated films (Quantity One image software).


Four wild types and four TgBACE2 (L11) for each age were perfused transcardially with 0.1 M PBS, pH 7.4 and then with 4% paraformaldehyde (Sigma, St Louis, MO, USA). The brains were removed from the skull and left in the same fixative for 24 h, then cryoprotected in 30% sucrose and kept frozen at–80°C. Coronal sections were obtained using a cryostat (50 μm). Serial tissue sections were processed free-floating using the streptavidin-biotin-peroxidase complex immunohistochemical method (DAKO, LSAB system, Ely, UK). Briefly, after peroxidase blocking, sections were incubated with 10% fetal bovine serum (FBS) and 0.25% gelatine. Then incubation with the primary antibody was performed overnight at 4°C at a dilution of 1:1000 for anti-ChAT antibody (choline acetyltransferase, Chemicon, Temeluca, CA, USA), 1:2000 for anti-p75NGFR antibody (Chemicon, Temeluca, CA, USA), 1:8000 for anti-TH antibody (Sigma, St Louis, MO, USA) or 1:1000 for anti-HA antibody (Eurogenic, Emeryville, CA, USA) in PBS containing 0.2% Triton X-100 and 1% FBS. The sections were then incubated with the biotinylated link and the streptavidin-HRP as indicated in the manufacturer's instructions. Peroxidase activity was visualized with 0.05% diaminobenzidine and 0.01% hydrogen peroxide.

Quantification of Aβ1−40and A β1−42

1−40 and Aβ1−42 peptide levels were determined using an enzyme-liked immunoadsorbent assay (ELISA) (Covance, Dedham, MA, USA) using soluble extracts from cerebral cortex and hippocampus of four mice per genotype and age (L11), and following the manufacturer's instructions.

Stereological analysis: quantification of ChAT, p75NGFRand TH-positive cells

Two basal forebrain (BF) cholinergic nuclei, the medial septal nucleus (MSN) and the vertical diagonal band (VDB), and the LC were analysed in all immunohistochemical experiments. Stereological estimations of the density of ChAT and p75NGFR positive neurones in BF and of TH-positive neurones in LC were obtained. For the stereological analysis the CAST-GRID software package (Olympus, Ballerup, Denmark) adapted to an OLYMPUS BX51 microscope was used. The studied areas were MSN and VDB (Bregma 1.54 to 0.26 mm) and LC (Bregma–5.34 to–5.8 mm) according to the stereological coordinates adopted from the Mouse Brain Atlas (Keith & George Paxinos 1997). Estimation of the volume of the selected regions was performed using the Cavalieri method and the optical dissector method was used as previously described by Dierssen et al. (2006).

Behavioural analysis

To check behavioural genotype-associated changes at two different ages, we studied adult (6 months) wild-type (n = 24) and TgBACE2 (n = 17) and old (22 months) wild-type (n = 14) and transgenic (n = 12) male mice. The behavioural characterization consisted in a neurological test battery, analysis of the locomotor activity, anxiety-like behaviour and cognitive profile (see below). The experiments were performed with an increasing gradient of stress to avoid interference in the results.

Neurological assessment (SHIRPA protocol)

SHIRPA primary screen is a comprehensive semiquantitative routine testing protocol to identify and characterize phenotype impairments during which 40 separate measurements are recorded for each animal, including somatometry (Rogers et al. 1997). Assessment of each animal began with observation of undisturbed behaviour in a cylindrical clear Perspex viewing jar (15 cm height, 11 cm diameter) for wild running or stereotypes. Mice were then transferred to an arena (56 × 34 cm) for observation of motor behaviour and sensorial function. Animals underwent screening exams for visual acuity, vibrissae, corneal and pinna responses to an approaching cotton swab, auditory (Preyer reflex) and vestibular function (contact righting reflex and negative geotaxis), and grip strength and body tone. In the last part of the test battery, changes in excitability, aggression, general fear, vocalization and salivation, and piloerection (for analysis of autonomic function) were recorded.

Locomotor activity

Locomotor activity was measured by using actimetry boxes (45 × 45 cm; Panlab SL, Barcelona, Spain) contained in a soundproof rack mount cabinet. Back and forward movements were monitored by means of an infrared beams grid and used as an index of locomotor activity (counts). Counts were integrated every hour and added to obtain total locomotor activity for a 24-h period maintaining the 12:12 h light–dark schedule, during four consecutive days. The measured parameters in the present study were total distance travelled by the animals (cm) and mean velocity (cm/sec).

Open field test

The open field was a white melamine box (70 × 70 × 50 cm high) divided into 25 equal squares and under high intensity light levels (300 Lux). Mice tend to avoid brightly illuminated, novel, open spaces, so the open field environment acts as an anxiogenic stimulus and allows for measurement of anxiety-induced locomotor activity and exploratory behaviours. Thus, two zones, centre (1764 cm2) and periphery (3136 cm2) were delineated, being the centre more anxiogenic. At the beginning of the test session, mice were left in the periphery of the apparatus and during 5 min we measured and analysed the latency to cross from the periphery to the centre, total distance travelled, average speed and time spent in various parts of the field (e.g. the border areas vs. the open, central area).

Light and dark box

The light and dark box test is based on the innate tendency of mice to seek refuge in a dark box. We used a box consisting of a small (15 × 20 × 25 cm) compartment with black walls and black floor dimly illuminated (25 Lux), connected by a 4-cm long tunnel to a large compartment (30 × 20 × 25 cm) with white walls and a white floor, intensely lit (500 Lux). Mice were individually placed in the dark compartment facing the tunnel at the beginning of the 5-min observation session. Number entries to light and dark zones, and in the tunnel connecting both zones, and time spent in each were recorded, as well as the latency to the first visit to the light zone.

Elevated plus maze

The elevated plus maze consisted of a black Plexiglas apparatus with four arms (29 cm long × 5 cm wide) set in cross from a central square (5 cm × 5 cm). Two opposite arms were delimited by vertical walls (closed arms), and the other two had unprotected edges (open arms). The maze was elevated 40 cm above the ground under dim light (100 Lux). At the beginning of the 5-min observation session, each mouse was placed in the central zone, facing one of the open arms. The total numbers of visits to the closed and open arms, and the time spent in open and closed arms were recorded. An arm visit was recorded when the mouse moved all four paws into the arm.

Zero maze

The zero maze consisted of a circular path (runway width 5.5 cm and 46 cm diameter) with two open and two closed segments (walls 8 cm high) and was elevated 50 cm above ground. Animals were placed into the closed segment and their movements were recorded for 5 min. The latency to enter to the open segment, the number of entries and the total time spent into both segments were measured.

Morris water maze

The swimming pool, 120 cm diameter and 0.5 m height, was filled with water (24 ± 1°C) made opaque with non-toxic white paint and several fixed room cues were constantly visible from the pool. In the first day (training session), the escape platform (15 cm diameter, 24 cm height) was visible and placed in the centre of the pool to train the animal to escape from water. Four training trials were performed, entering the mice for four starting positions (north, south, east or west). During the following 5 days (Days 2–6) animals were tested for place learning acquisition with the escape platform located in the middle of the northwest quadrant, 1 cm below water surface. Four trials per day were performed (30 min inter-trial interval), mice entering randomly from each one of the starting positions and allowed to swim until they located the platform. Mice failing to find the platform within 60 seconds were placed on it and left there for 20 seconds, as the successful animals. On the day 7, we removed the platform from the pool, and four probe trials (60 seconds) were performed, in which the time spent and distance travelled in the trained and non-trained quadrants were recorded. Finally, a session with a cued visible platform, situated in the centre of the pool and 1 cm above the water surface, was carried out. All the trials were recorded and traced with an Image tracking system (SMART, Panlab SL, Barcelona, Spain) connected to a video camera placed above the pool.

Passive avoidance

We used a step-down passive avoidance test, which consisted of a transparent Plexiglas circular cage (40 cm in height, 30 cm in diameter) with a grid floor and a circular platform (4 cm diameter) in the centre. During the training session, animals were placed on the platform and their latency to step down with all four paws was measured. Immediately after stepping down on the grid, animals received an electric shock (0.6 mA, 2 seconds). Retention test sessions were carried out 24 h (short-term) and 7 days after training (long-term). Step-down latency was used as a measure of memory retention. A cut-off time of 300 seconds was set.

Data analysis

Statistical analysis was performed using the software SPSS 12.0. To assess possible differences between both transgenic lines (L2 and L11), for each experiment we analysed both main effects and interactions of line. Then, we checked that the genotype effects persisted if line was added as a factor to the analysis of variance (ANOVA) model. As we did not observe any significant differences between transgenic lines or wild-type lines, we decided to combine the results of both lines. Data were summarized as mean ± standard error of mean (SEM) when normality might be assumed. Between-group comparisons were analysed using one-way ANOVA or a two-way ANOVA with genotype and age as factors and with acquisitions trials as a repeated measure and significant effects were analysed post hoc using Bonferroni test. The Passive avoidance test was analysed using Mann–Withney U non-parametric test. For the analysis of the Western blot and ELISA results, Student's t-test analysis was used. In all tests, a difference was considered to be significant if the obtained probability value was P < 0.05.


Generation and general characterization of TgBACE2 mice

To generate transgenic mice overexpressing BACE2, we cloned the human BACE2 open reading frame (518 amino acids isoform) fused at the C-terminus to the HA-tag under the control of the mouse neurone-specific Thy-1.2 promoter (Fig. 1a). We used the Thy-1.2 promoter because it has been shown to drive efficient and specific expression of transgenes in the brain (Moechars et al. 1996). Two transgenic lines, designated as line 2 (B6/CBAF1/J background) and 11 (B6/SJLF1/J background), carrying 5 and 10 copies of the transgene, respectively, were established (data not shown). TgBACE2 were maintained by crossing TgBACE2 males with wild-type females, and transgenic pups were born at the expected frequency (L2: wild type = 45.71%, TgBACE2 = 54.28% and L11: wild type = 52.38%, TgBACE2 = 47.62%). Transgene expression at the protein level was confirmed by Western blot using an anti-HA antibody (Fig. 1b) and increased BACE2 protein levels were detected as shown by Western blot analysis with an anti-BACE2 antibody in adult brain from (L 2: 21 ± 3.6%; n = 4; L11: 27 ± 4.3%; n = 4; Fig. 1c). Immunohistochemistry studies showed that overexpressed BACE2-HA was present in the cytoplasm similar to what has been described for the endogenous BACE2 subcellular localization (Fig. 1d).

Figure 1.

Construction of Thy-1/BACE2HA transgene. Genotyping and expression analysis of TgBACE2 mice. (a) Schematic representation of the Thy-1/BACE2HA chimeric gene. Dark gray boxes represent exons E1-E4 from mouse Thy1 gene. (b) Representative Western blots of adult (6 months) brain from wild type (WT) and TgBACE2 (Tg, Line 2) with anti-HA antibody. (c) Representative Western blots of adult (6 months) brain from wild-type (WT) and transgenic (L11) mice using anti-Bace2 antibody; two animals per genotype are shown. Actin was used as a control of equal loading. (d) Representative photomicrographs of anti-HA immunohistochemistry of different brain areas of adult (6 months) TgBACE2 (L11). Panels show: a, cerebral cortex (Bregma–0.94 mm); b,d, cerebellar cortex (Purkinje cells) (Bregma–6.64 mm); and c,e, medulla-pons (Bregma–6.72 mm). Upper panels scale bar: 50 μm. Lower panels scale bar: 10 μm.

Physical characteristics such as body weight and the presence of bald patches and appearance of behavioural anomalies in the home cages were registered systematically with no differences between genotypes. Neurological assessment using modified Primary SHIRPA protocol showed that spontaneous activity or sensory, motor and autonomic functions were not affected by BACE2 overexpression (Table S1). No genotype-dependent differences were found in total locomotor activity rcorded during the 4 days, and the results showed that there was a habituation process in all mice along days (Fig. 2). However, aged mice of both genotypes showed the characteristic reduction in general activity (Fig. 2).

Figure 2.

Locomotor activity. Total distance travelled in the actimetry box during four consecutive days. No differences were observed comparing wild-type vs. TgBACE2 mice of each age and each day. Note the progressive reduction of activity along days in both genotypes and ages. Older mice of both genotypes were less active than their young adult counterparts. Data are expressed as mean ± SEM.

BACE2 overexpression does not lead to alterations in learning and memory, increase of Aβ peptides or cholinergic degeneration

In the Morris water maze (Fig. 3a), age-related but no genotype-related (two-way ANOVA; age effect: F(1,66) = 22.1, P = 0.0001; genotype and genotype × age effects: F(1,66) < 1.2) differences were observed in procedural learning during training session. Along the acquisition session, also age-related but no genotype-related differences were observed (two-way ANOVA repeated measure; age effect: F(1,66) = 11.3, P = 0.001; genotype and genotype × age effect: F(1,66) < 1.0), observing that all groups of animals learned the task during the five consecutive days (two-way ANOVA repeated measure; acquisition effect: F(4,66) = 34.2, P = 0.0001; age, genotype or age × genotype effects: F(4,66) < 0.34). In the removal session (Fig. 3b), a significant increase in the percentage of time spent in the trained quadrant (northwest) comparing to non-trained quadrants was observed (two-way ANOVA, F(1,66) = 58.6, P = 0.001). However, no genotype or age significant differences were observed (two-way ANOVA, genotype × age effect: F(4,66) = 0.32, P = 0.79). Finally, in the cued session, no significant motor or motivational problems were detected (two-way ANOVA, age, genotype and genotype × age effects: F(4,66) < 2.0; Fig. 3a).

Figure 3.

Learning and memory tests. (a,b) Morris water maze. (a) Escape latencies along the different sessions of the test. No differences between genotypes were observed at either studied ages and the age-dependent memory loss was similar in both genotypes (T = training; A = acquisition). (b) In the removal session no differences were observed between genotypes in the preference for the trained quadrant (NW = northwest, trained quadrant; NE = northeast; SW = southwest; SE = southeast). Data are expressed as mean ± SEM. (c) Step-down passive avoidance. No differences were observed between genotypes at both studied ages, in the latency to exit the platform. Data are expressed as median and interquartile ranges.

The data obtained in the passive avoidance show that overexpression of BACE2 affects neither short-term nor long-term memories as shown by the similar step-down latencies in both genotypes at 24 h (Mann–Withney U-test; 6 months, P = 0.80; 22 months: P = 0.89) and 7 days (Mann–Withney U-test; 6 months: P = 0.50; 22 months: P = 0.85) (Fig. 3c).

To determine whether the overexpression of BACE2 led to an increase of Aβ peptides in the cerebral cortex and the hippocampus, the more affected areas in DS and AD, we measured the concentration of Aβ1−40 and Aβ1−42 by ELISA at two different ages, 6 and 22 months. The results showed no differences in the amount of any of the peptides in transgenic mice of either age when compared to control littermates (Table 1).

Table 1.  ELISA quantification of Aβ1−40 and Aβ1−42 levels in cerebral cortex and hippocampus wild-type and TgBACE2 animals
6 months22 months6 months22 months6 months22 months6 months22 months
  1. No differences were observed between genotypes and ages. TgBACE2 (L11, n = 4 per age) vs. wild type (n = 4 per age). Data (pg/ml) are expressed as means ± SEM.

Wild type1.09 ± 0.373.04 ± 1.54.02 ± 0.354.47 ± 1.371.68 ± 0.723.64 ± 0.342.52 ± 0.732.92 ± 0.95
TgBACE21.35 ± 0.454.05 ± 0.84.22 ± 0.824.38 ± 0.281.16 ± 0.693.38 ± 0.712.50 ± 0.812.97 ± 0.89

Despite the lack of Aβ peptides accumulation, it was relevant to determine whether BACE2 overexpression had an impact in cholinergic neurodegeneration similar to the observed in DS mouse model and DS and AD patients (Fodale et al. 2006; Granholm et al. 2000). To this end, unbiased stereological methods were used to compare the density of cholinergic neurones of BF in MSN and VDB. Cholinergic neurones were stained with specific markers, namely ChAT, the rate-limiting enzyme in the acetylcholine synthesis, and p75NGFR, the low affinity nerve growth factor receptor. No genotype-dependent differences were detected in the total volume of the nuclei studied (data not shown) or in ChAT or p75NGFR cell density at the ages analysed (Table 2). These results suggest that BACE2 overexpression does not produce early degeneration of the cholinergic system.

Table 2.  Stereological analysis of densities of cholinergic neurones in the basal forebrain
 ChAT density p75NGFR density
6 months22 months6 months22 months
  1. No differences between genotypes and ages were observed in ChAT and p75NGFR cell density of MSN and VDB. Wild type (n = 4 per age) vs. TgBACE2 (L11, n = 4 per age). Data (no. of cells/mm3) are expressed as means ± SEM.

Wild type5282.012 ± 972.685603.85 ± 450.026622.73 ± 225.566250.01 ± 974.78
TgBACE24199.37 ± 1185.614870.68 ± 247.276519.74 ± 877.964401.88 ± 317.66

Behavioural and molecular characterization showed increased anxiety-related phenotype in TgBACE2

In the open field there was a significant increase in the time spent in the periphery (ANOVA, 6 months: F(1,40) = 7.31, P = 0.03; 22 months: F(1,25) = 5.30, P = 0.04; Fig. 4b). We also observed an increase in the latency to cross from center to periphery in both age groups of transgenic animals (ANOVA, 6 months: F(1,40) = 4.51, P = 0.04; 22 months: F(1,25) = 6.45, P = 0.03; Fig. 4a). Importantly, both distance travelled and mean speed were unaffected (Fig. 4c,d), indicating that the anxiety-like behaviour was not dependent on alterations of locomotor activity. In the light and dark boxes, a similar phenotype was observed in young TgBACE2 mice, characterized by an increase in the time spent in the dark box, an accepted measure of anxiety-like behaviour (ANOVA, F(1,40) = 4.98, P = 0.04; Fig. 5b), and in the latency to cross from dark to light compartment (ANOVA, F(1,40) = 6.14, P = 0.03; Fig. 5a). Both young and old transgenic animals presented reduced percentage of entries to the light compartment (ANOVA, 6 months: F(1,40) = 8.56, P = 0.03; 22 months: F(1,25) = 3.92, P = 0.04; Fig. 5c).

Figure 4.

Open field test. (a) Transgenic animals spent more time in the periphery of the maze. (b) Latency to cross from periphery to center expressed in seconds. Transgenic animals show an increased latency that was more marked in older mice. TgBACE2 displayed normal activity levels (c) and velocity (d). Data are expressed as mean ± SEM *P < 0.05.

Figure 5.

Light and dark box. (a) Percentage of time spent in the dark zone. Transgenic young mice spent more time in this zone as compared to their control group. (b) Latency to cross from dark to light compartment expressed in seconds. Young transgenic animals show a significantly increased latency. (c) Percentage of entries to the light side. Both transgenic groups entered less often to the lit side than their controls. Data are expressed as mean ± SEM *P < 0.05.

For a further characterization of the anxiety-like behaviour, we tested the animals in the elevated plus maze. There were no differences in the time spent in the closed arms (Fig. 6a), but young transgenic mice presented reduced entries into the open arms as compared to their control group (ANOVA, F(1,40) = 8.01, P = 0.02; Fig. 6b) along with unaffected total distance travelled (Fig. 6c). The latency to cross from closed to open segment in the zero maze was higher in TgBACE2 old mice as compared to their control group (ANOVA, F(1,25) = 5.11, P = 0.04; Fig. 7a). Transgenic young mice showed an increase in the time spent in the closed segments (ANOVA, F(1,40) = 3.12, P = 0.07), being significant in old transgenic animals (ANOVA, F(1,25) = 4.89, P = 0.04; Fig. 7b). Finally, TgBACE2 young mice showed reduced entries to the open segments (ANOVA, F(1,40) = 7.31, P = 0.03; Fig. 7c), with no changes in the total distance travelled (Fig. 7d). Taken together results indicate an increased anxiety-like behaviour in TgBACE2 mice.

Figure 6.

Elevated plus maze. (a) Young transgenic animals visited less the unprotected (open) arms. (b) Percentage of time spent in the protected (closed) arms, indicated no differences between genotypes. (c) No differences in the total distance travelled were observed. Data are expressed as mean ± SEM *P < 0.05.

Figure 7.

Zero maze. (a) Latency to enter in the unprotected (open) segment in seconds. Transgenic animals showed a higher latency to enter to the open segment. (b) Transgenic animals spent more percentage of time in protected (closed) segment. (c) Number of entries to the unprotected segment. A reduction was observed in young transgenic animals. (d) No differences were observed in the total distance travelled. Data are expressed as mean ± SEM *P < 0.05.

We hypothesized that the noradrenergic system could be involved the increased anxiety phenotype observed in TgBACE2 mice. To explore this possibility, we determined the expression levels of TH, the rate-limiting enzyme of the synthesis of catecholamines, in the medulla-pons region by Western blotting analysis. Our results showed a significant increase in TH levels in young transgenic mice vs. wild types (Student's t-test, P = 0.003) and a tendency to an increase in old TgBACE2 animals (Student's t-test, P = 0.088) (Fig. 8a).

Figure 8.

Molecular characterization of the noradrenergic system. (a) Western blot analysis to evaluate TH expression levels in the medulla-pons region of TgBACE2 (Tg, L11; n = 4 per age) and wild-type mice (WT; n = 4 per age) at two different ages, as indicated. Actin expression was used for normalization. (b) Increased density of TH-positive cells in LC from TgBACE2 (L11; n = 4 per age) vs. wild-type (WT, n = 4 per age) mice. Data are expressed as mean ± SEM *P < 0.05; **P < 0.01.

Since these results may indicate an increase in LC noradrenergic neurone density, we quantified the density of TH-positive cells in the LC. The density of TH-positive of cells was higher in transgenic animals of both ages as compared to their respective controls (ANOVA, 6 months, F(1,7) = 4.6; P = 0.04; 22 months, F(1,7) = 4.38; P = 0.03; Fig. 8b), without differences in the total volume of the LC (data not shown).


Because of its high homology to BACE1, during many years β-secretase activity was attributed to BACE2 (Acquati et al. 2000; Yan et al. 1999); if confirmed, it could contribute to the Alzheimer's like neuropathology observed in DS patients by the concomitant overexpression with APP (Acquati et al. 2000; Murphy et al. 1990). The present study shows that overexpression of BACE2 in vivo is not involved in the age-dependent motor or cognitive impairment, and aged TgBACE2 mice did not show increased Aβ production or cholinergic degeneration signs.

In fact, previous in vitro studies had already questioned the proamyloidogenic role of BACE2, showing controversial results. It has been shown that overexpression of BACE2 in APP-expressing cells profoundly reduced Aβ formation (Bennett et al. 2000; Farzan et al. 2000; Hussain et al. 2000; Sun et al. 2006; Yan et al. 2001), and that selective downregulation of BACE2 by RNA interference increased Aβ secretion (Basi et al. 2003). All these data pointed to a protective role for BACE2 more than a proamyloidogenic factor. Our results using ELISA, show that in vivo neuronal BACE2 overexpression does not affect Aβ1−40 or Aβ1−42production in specific areas involved in learning and memory, such as the hippocampus or the cerebral cortex. Whereas some authors found that BACE2 can cleave wild-type or Swedish mutant APP at β-secretase (Farzan et al. 2000; Hussain et al. 2000), more recently, it has been show that BACE2 generates a C80 fragment, as a result of a novel cleavage site designated as APP θ-secretase cleavage site (Sun et al. 2006).

On the other hand, a role of β-amyloid and/or inflammation has been suggested in the degeneration of cholinergic synaptic structures, and premature loss of BF cholinergic neurones, one of the neuropathological hallmarks of DS and AD (Casanova et al. 1985; Mann 1988; Mann et al. 1985). This alteration, also observed in Ts65Dn mice (Cooper et al. 2001; Granholm et al. 2000), is associated with learning and memory deficits in cognitive tasks (Fodale et al. 2006; Granholm et al. 2000; Hyde et al. 2001). In our experiments, TgBACE2 mice did not show any impairment in visuo-spatial learning and memory, or in recent memory in the passive avoidance test, which has been shown to depend on the integrity of the cholinergic system (Dierssen et al. 1992). Along with the lack of cognitive impairment, TgBACE2 mice did not present signs of cholinergic degeneration, thus indicating that single overexpression of BACE2 is not able to develop the cholinergic and cognitive phenotype observed in DS, AD patients and DS mouse models.

Interestingly, behavioural characterization showed an increased anxiety-like behaviour in TgBACE2. This anxiety-like behaviour is similar to that described for Bace1 knock-out mice (Dominguez et al. 2005), but opposite to that shown by BACE1 transgenic mice which exhibited a reduced anxiety-like behaviour (Harrison et al. 2003), suggesting antagonistic roles for BACE2 and BACE1 in anxiety pathways. Several mechanisms could be involved in the anxiety-like behaviour that include an elevated noradrenergic activity or inappropriate activation of the LC, which has a physiological role in emotionality (Dierssen et al. 2006; Priolo et al. 1991). Consistent with this, we have observed an increase of TH-positive neurones density in LC of both transgenic groups and a significant increase of TH levels in the medulla-pons region of TgBACE2 young mice, and a tendency in old transgenic animals. Unfortunately, there are no studies analysing noradrenergic phenotype in the other available BACE1 mouse models and thus we cannot compare the effects of the two β-secretases in this particular aspect.

However, a noradrenergic phenotype has been reported in AD patients and in AD mouse models, where loss of LC noradrenergic neurones has been observed (Bondareff et al. 1987; Kalinin et al. 2007; Mann 1983; O’Neil et al. 2007), so the increase of APP C99 may contribute to this phenotype. Moreover, it has been suggested that noradrenergic innervations from LC are needed to maintain adequate Aβ clearance (Kalinin et al. 2007).

On the other hand, DS patients and the Ts65Dn mice, a DS trisomic model, show alterations in noradrenergic synaptic transmission in the cerebral cortex and the hippocampus (Dierssen et al. 1996; Dierssen et al. 1997; Lumbreras et al. 2006), although no gross anatomical differences were observed in the cholinergic, serotonergic and catecholaminergic systems in trisomic mice (Megias et al. 1997) or in DS fetuses (Lubec et al. 2001). Phenotypically, reduced behavioural inhibition and/or reduced attention to relevant stimuli in the open field and elevated plus maze (Coussons-Read & Crnic 1996; Escorihuela et al. 1995; Escorihuela et al. 1998) and anxiety and panic behaviour responses to a predator in the Mouse Defense Test Battery (Martinez-Cue et al. 2006) have been reported in Ts65Dn mice, thus suggesting that functional or structural alterations of the noradrenergic system may be at least partially contributed by BACE2 overexpression.

In conclusion, our results indicate that BACE2 overexpression in vivo is not involved in the cholinergic-dependent cognitive dysfunction observed in DS patients. However, we show that BACE2 overexpression induces an anxiety-related behaviour and gives raise to structural changes in the noradrenergic system.


The laboratories of MD, SL and CF are supported by the DURSI (2009SGR1313) and the Department de Salut of the Catalan Autonomous Government (‘Generalitat de Catalunya’). This work was supported by grants from the Spanish Ministry of Education and Science (FCT-08-0782, SAF2007-60827, SAF 2007-31093-E), FIS (PI 082038), Marató TV3 (062230), Jerome Lejeune (JMLM/AC/08-044), Reina Sofía and Areces Foundations and EU (LSHG-CT-2006-037627). GA received additional support from the Basque Government (BFI05.48). The CIBER of Enfermedades Raras is an initiative of the ISCIII.