|Home | About | Journals | Submit | Contact Us | Français|
Binding of the neurotrophin brain-derived neurotrophic factor (BDNF) to the TrkB receptor is a major survival mechanism during embryonic development. In the aged brain, however, BDNF levels are low, suggesting that if TrkB is to play a role in survival at this stage additional mechanisms must have developed. We here show that TrkB activity is most robust in the hippocampus of 21-d-old BDNF-knockout mice as well as in old, wild-type, and BDNF heterozygous animals. Moreover, robust TrkB activity is evident in old but not young hippocampal neurons differentiating in vitro in the absence of any exogenous neurotrophin and also in neurons from BDNF −/− embryos. Age-associated increase in TrkB activity correlated with a mild yet progressive loss of cholesterol. This, in turn, correlated with increased expression of the cholesterol catabolic enzyme cholesterol 24-hydroxylase. Direct cause–effect, cholesterol loss–high TrkB activity was demonstrated by pharmacological means and by manipulating the levels of cholesterol 24-hydroxylase. Because reduced levels of cholesterol and increased expression of choleseterol-24-hydroxylase were also observed in the hippocampus of aged mice, changes in cellular cholesterol content may be used to modulate receptor activity strength in vivo, autonomously or as a way to complement the natural decay of neurotrophin production.
During development, neurotrophins are mandatory for the survival, differentiation, and growth of different neuronal populations (Reichardt, 2006 ). In the mature nervous system, neurotrophins are important for the modulation of neuronal connectivity and activity-dependent plasticity (Conover and Yancopoulos, 1997 ; Blum and Konnerth, 2005 ). Neurotrophins bind and activate receptor tyrosine kinases (RTKs), in turn leading to multiple intracellular signaling pathways, most notoriously those involving mitogen-activated protein kinases and phosphatidylinositol 3-kinase (PI3K) (Kaplan and Miller, 2000 ; Reichardt, 2006 ). In the hippocampus, a region of the brain critically involved in certain types of learning and memory, the most prominently expressed neurotrophin receptor is TrkB (Tokuyama et al., 1998 ), whose cognate ligand is BDNF (brain-derived neurotrophic factor). In agreement with a role in memory-associated processes, loss-of-function studies of both TrkB and BDNF result in changes in affective and cognitive states in mammals and humans (Minichiello et al., 1999 ; Pozzo-Miller et al., 1999 ; Xu et al., 2000 ; Egan et al., 2003 ; Yeo et al., 2004 ). Although there is no doubt that BDNF is the main modulator of TrkB activity, a number of evidences indicate that certain roles mediated by the activation of TrkB may occur independently from BDNF. For instance, TrkB conditional knockout mice present clear defects in pre- and postsynaptic morphogenesis in the hippocampus (Luikart et al., 2005 ), yet this is not the case in BDNF conditional knockout mice (Gorski et al., 2003 ; Hill et al., 2005 ). The last observations are in turn consistent with the lack of an overt effect on the development of the hippocampus in BDNF knockout animals (Ernfors et al., 1994 ; Jones et al., 1994 ). Furthermore, the absolute need for BDNF for all and every TrkB-mediated activity in the adult hippocampus is challenged by reports showing that the levels and activity of BDNF in the hippocampus start to decay in early adulthood in rats, reaching lowest levels at the end of the first year of life (Gooney et al., 2004 ; Shetty et al., 2004 ; Hattiangady et al., 2005 ; see however Katho-Semba et al., 1998 ). One simple conclusion from these results is that mechanisms alternative or complementary to BDNF binding have evolved during cellular/neuronal differentiation, to guarantee TrkB activity in the adult brain. The best characterized of such mechanisms is the promiscuity of the lower affinity ligands NT3/NT4 to bind to TrkB (Yan et al., 1993 ; Davies et al., 1995 ). Paradigmatically, in vivo studies demonstrated that the lack of BDNF is not always compensated by NT4 and vice versa (Conover et al., 1995 ; Stenqvist et al., 2005 ). Robust TrkB signaling in the mature brain can in principle also occur through the increased expression of TrkB subunits, and/or of its p75 coreceptor (Zaccaro et al., 2001 ; Hartmann et al., 2004 ; Marshak et al., 2007 ). Contradicting this possibility, recent studies have shown the down-regulation of full-length TrkB in the hippocampus, starting 2 mo after birth (Silhol et al., 2005 ). It also appears possible that TrkB signaling in the adult brain could occur via the up-regulation of ligands like adenosine or gangliosides, which have been shown to act as potent inducers of TrkB activity in cells in vitro (Lee and Chao, 2001 ; Duchemin et al., 2002 ).
Irrespective of the possibilities of TrkB binding to a combination of classical or noncanonical ligands, one can also envision that cells have evolved non-ligand–mediated mechanisms to guarantee the most efficient transduction of this important signaling receptor. One such mechanism may be the modulation of the content/ratios of different membrane lipids, most notoriously cholesterol. Cholesterol plays a crucial role in the generation of ordered domains in the plasma membrane that laterally segregate certain proteins, thus reducing their rate of lateral diffusion and, by virtue of this, increasing clustering and consequently signaling strength (Edidin, 2003 ; Hancock and Parton, 2005 ; and Hancock, 2006 ). In fact, the number of studies in this regard is steadily increasing, also for RTKs (Paratcha and Ibanez, 2002 ; Suzuki et al., 2004 ; Nicolau et al., 2006 ; Hanzal-Bayer and Hancock, 2007 ; Jacobson et al., 2007 ; Pereira and Chao, 2007 ). These observations prompted us to address the hypothesis that the activity of TrkB in the differentiated hippocampus may be controlled via a physiological modulation of cholesterol levels in the neuronal plasma membrane.
Primary cultures of embryonic rat hippocampal neurons were prepared as described (Kaech and Banker, 2006 ). Hippocampal neurons from BDNF knockout mice were obtained from crossing BDNF +/− mice (Ernfors et al., 1994 ). The two hippocampi from each 14–15-d embryo were put in separate tubes; a piece of cortex from each embryo, for later determination of genotype with BDNF specific primers, was also put in separate tubes. The hippocampi were then trypsinized and dissociated as for the rat hippocampal cultures, followed by plating in individual dishes. Each dish was later classified as BDNF +/+, +/−, or −/−, according to the PCR analysis of the corresponding cortical specimens. For biochemical analysis, 105 cells were plated into 3-cm plastic dishes coated with poly-l-lysine (0.1 mg/ml) and containing minimal essential medium with N2 supplements (MEM-N2). Neurons were kept under 5% CO2 at 37°C. Where indicated, cells were incubated with particular inhibitors for 2 h to assess target inhibition or for 36 h to measure cell death. The monoclonal anti-Flotillin 1 antibody, clone 18, was from Transduction Laboratories (Franklin Lakes, NJ); the polyclonal chicken anti-BACE 1 antibody, raised against Fc-Asp 2-fusion protein, and the monoclonal anti-transferrin receptor antibody were from Zymed Laboratories (San Francisco, CA). Antibodies used to detect pTrk were polyclonal rabbit either from Santa Cruz Biotechnology (Santa Cruz, CA) or from Cell Signaling (Beverly, MA). Antibodies used to detect TrkB were polyclonal rabbit either from Santa Cruz Biotechnology or from BD Transduction Laboratories (Lexington, KY). Mouse monoclonal anti-tubulin were all from Cell Signaling, and the rabbit antiserum anti-p85 was from Upstate Biotechnology (Lake Placid, NY). CYP46A1 antibody was provided by Dr. Russell (University of Texas Southwestern Medical Center, Dallas, TX) and described previously (Lund et al., 1999 ).
Predifferentiated 10 days in vitro (10 DIV) hippocampal neurons were cultured in a serum-free medium and stimulated in independent experiments with 100 ng/ml BDNF for 30 min, 50 ng/ml nerve growth factor (NGF) for 10 min, or with a conditioned medium derived from 26 DIV neurons for 120 min. Neurotrophin activities were neutralized in 10 DIV cell culture medium by adding agonist antibodies raised against BDNF, NGF, NT3, NT4/5 (anti-neurotrophin antibody sample pack 1; Chemicon, Temecula, CA) at a concentration of 5 mg/ml for 120 h.
Mature rat hippocampal neurons were lysed by osmotic swelling in 25 mM MES Buffer, pH 7.0, 5 mM dithiothreitol, 2 mM EDTA and protease inhibitor cocktail (Roche, Indianapolis, IN), in ice for 15 min. Extracts were homogenized with a syringe (22 gauge) and centrifuged for 5 min at 700 × g and at 4°C. Supernatants were brought to a final concentration of 1.6 M sucrose, put on the bottom of an SW50 rotor centrifuge tube (Beckman Instruments, Fullerton, CA), overlaid with a continuous sucrose gradient from 1.6 to 0.4 M, and subsequently centrifuged at 12,0000 × g for at least 12 h at 4°C. Eight fractions were collected from the top to the bottom of the tube. The fractions that contain plasma membrane were subjected to centrifugation at 10,0000 × g for 18 h at 4°C. The fractions were finally analyzed by Western immunoblot using antibodies against specific markers of membrane compartments.
Detergent-resistant membranes (DRMs) were prepared from hippocampal membranes (see above) by Triton X-100 extraction, and soluble and insoluble (DRMs) fractions were separated as described previously (Tansey et al., 2000 ). For some experiments, total plasma membrane extracts were brought to 60% sucrose in MES Buffer Saline (MBS), and a sucrose step gradient was overlaid (35 and 5% sucrose). After centrifugation at 100,000 × g for 18 h at 4°C, fractions were collected from the top of each tube. Fractions 4 and 5 were identified as the DRM fractions by the presence of the DRM marker Flotillin 1.
Rat or mouse hippocampal tissues were homogenized in PBS containing 9% sucrose, protease inhibitors (CLAP: pepstatin, antipain, and chymostatin, each at a final concentration of 25 mM) and 1 mM sodium orthovanadate using a dounce homogenizer and 10 passages through a 22-gauge syringe. Samples were centrifuged for 10 min at 2500 × g, and supernatants were considered as total extracts. A further centrifugation step was performed at 10,000 × g for 1 h at 4°C to pellet the membrane fraction.
Total and membrane pellets of hippocampal neurons were extracted with two different detergents depending on the aim of the experiment: Nonidet P-40 (1% Nonidet P-40, 10% glycerol, 100 mM NaCl, 2 mM EDTA, 10 mM Tris/HCl, 500 mM sodium orthovanadate, and protease inhibitors) or Triton X-100 buffer (100 mM MES, pH 7, 150 mM NaCl, 1% Triton X-100) in the presence of 1 mM sodium orthovanadate and protease inhibitors. Extracts were clarified by centrifugation, and the protein concentrations were quantified by the BCA method (Bio-Rad Laboratories, Hercules, CA). Proteins were then transferred onto nitrocellulose membranes and probed with primary antibodies for 16 h. Species-specific peroxidase-conjugated secondary antibodies were subsequently used to perform enhanced chemiluminescence detection (Amersham, Little Chalfont, United Kingdom).
For immunoprecipitations, hippocampal neurons were extracted as described above. The soluble and insoluble fractions were diluted with the appropriate buffer to equalize detergent amount (1% Triton and 0.1% SDS). For TrkB and p85 immunoprecipitations, 1 μg of polyclonal antibody/sample was added to protein A–Sepharose beads, and samples were rotated at 4°C overnight. Beads were washed with cold lysis buffer. Immunoprecipitated complexes were separated by 10% PAGE-SDS electrophoresis and subjected to Western blot analysis. Quantification was done by densitometry of the autoradiograms using the NIH Image J software package (http://rsb.info.nih.gov/ij/).
Lipids were extracted from plasma membranes obtained from rat hippocampal tissues or cultured neurons according to Bligh and Dyer (1959) . Extracted lipids (cholesterol, ceramide, and sphingomyelin) were subsequently analyzed by thin-layer chromatography (TLC) on silica gel 60 HPTLC plates using a double-step system (hydrophilic running solvent: chloroform/acetone/acetic acid/methanol/water 50:20:10:10:5) and hydrophobic solvent: hexane/ethyl acetate (5:2) for cholesterol and ceramide. The plate was then dried and developed by spraying with 7% sulfuric acid and heating at 150°C in an oven.
Total cholesterol was measured in membrane extracts from hippocampal neurons in vitro or whole hippocampus as previously described (Abad-Rodriguez et al., 2004 ). For the induction of membrane cholesterol reduction, 0.4 μM mevilonin and 0.5 mM β-methyl-cyclo-dextrin were added daily to 5-d-old neurons during 4 d (96 h). At the end cholesterol was measured, to ascertain that the treatment did not result in more than 25% reduction. Only validated cases membranes were utilized for further experimentation. Cells were scraped in MES Buffer-CLAP at 4°C, and centrifuged at 1000 × g. Postnuclear supernatants were further centrifuged at 10,0000 × g for 1 h at 4°C to get the membrane pellet. Protein and cholesterol concentrations were measured after resuspension in MES Buffer-CLAP with 0.1% Triton X-100.
Cholesterol-MCD inclusion complexes were prepared as described (Klein et al., 1995 ). These complexes, containing 0.3 mM cholesterol, were added to the medium at a final concentration of 0.3 mM together with 2 mg/ml free cholesterol. The treatment was performed for 15 min, and then cells were washed and scraped with MES Buffer-CLAP 0.1% Triton X-100 at 4°C
Total RNA samples were prepared from 5 × 105 hippocampal neurons at 10, 15, 20, and 25 DIV using the ABI Prism 6100 Nucleic Acid PrepStation (Applied Biosystems, Foster City, CA). One hundred nanograms of total RNA were first incubated with poly-T oligonucleotides and subjected to reverse transcription using the Superscript First-Strand Synthesis System (Invitrogen, Carlsbad, CA), following the manufacturer's instructions. cDNA was used as a template for the real-time PCR analysis based on the SYBR Green PCR Master Mix assay with the ABI PRISM 7000 Sequence Detection System (Applied Biosystems, Foster City, CA). Real-time PCR primers were designed using primer-Express software (Applied Biosystems). Expression of genes encoding TrkA, TrkB, and BDNF and of genes involved in cholesterol synthesis and metabolism was analyzed, and transcriptional expression was normalized using the housekeeping gene encoding for the β Actin (actb) as reference in order to avoid differences due to possible RNA degradation or different reverse transcription efficacy. Relative expression levels of the targets were calculated with the comparative cycle threshold (CT) method according to the following formula: sample ΔCT = (gene CT − actbCT). To follow expression during ageing, results were normalized to a calibration standard sample of RNA (10 DIV sample). This control was assigned the normalization ratio = 1 and the normalization ratio of each sample was calculated according to the formula: 2−ΔΔCT = 2−(sample ΔCT − calibrator ΔCT).
For the light microscopy detection of Cyp46, neurons on glass coverslips were washed in PBS (0.01M), fixed with 4% paraformaldehyde, and processed for immunofluorescence by using anti-CYP46. For the analysis of phospho-TrkB on the surface of young and old cells, cells were first incubated with the corresponding antibody (see above) at 15°C for 15 min to minimize loss by internalization, and excess antibody washed away and then fixed and processed for immunofluorescence microscopy. The surface distribution of the ganglioside GT1b was detected with anti-GT1b mAb (Seikagaku, Tokyo, Japan). Samples were analyzed on an Olympus IX81 fluorescence microscope (Melville, NY). Quantification was performed using the NIH ImageJ software, and signal intensities were normalized by area.
The cyp46A1 targeting sequence was designed by using the RNA interference design algorithm at www.dharmacon.com/DesignCenter/DesignCenterPage.aspx, and 46-sense and 46-antisense oligos encoding cyp46A1-short hairpin RNA (shRNA) were designed as reported by Rubinson et al. (2007) . To generate pSi46 plasmid, these oligonucleotides were annealed and cloned into the HpaI and XhoI sites from pLentiLox 3.7 (Rubinson et al., 2007 ): 46-sense oligo: 5′-TAGATGTACCGTGCGATTCATTCAAGAGATGAATCGCACGGTACATCTTTTTTTC-3′; 46-antisense oligo: 5′-TCGAGAAAAAAAGATGTACCGTGCGATTCATCTCTTGAATGAATCGCACGGTACATCTA-3′.
The full ORF expression clone p46+ including the full cyp46A1 mouse open reading frame was purchased from RZPD (German Resource Center for Genome Research).
Primary dissociated hippocampal cells isolated from rat embryos (on embryonic day 18) were transfected using the Rat Neuron Nucleofector Kit from Amaxa Biosystems (Köln, Germany).
Consistent with previous work showing a decay in BDNF levels in the mature and aged brain (Hattiangady et al., 2005 ; Silhol et al., 2005 ), the mRNA levels were lower in 10- and 21-mo-old mouse hippocampi as compared with those in 1-mo-old mice (Supplementary Figure 1A). To test whether decreased BDNF levels during aging were paralleled by a decreased activity of TrkB (the main Trk receptor in the hippocampus; Merlio et al., 1992 ; Cellerino, 1996 ), phosphorylated TrkB levels were measured in hippocampal membranes of mice as they grew old. Quite paradoxically, Western blotting revealed weak yet noticeable levels of phosphorylated TrkB in the hippocampi of 21-d-old mice and much higher in 10- and 20-month-old animals (Figure 1A). In support of the levels of receptor phosphorylation leading to activation of downstream survival effectors, a similar curve was observed for the downstream target Akt (Figure 1B). To further test that hippocampal cells can potently activate the TrkB pathway in low-BDNF situations, we performed Western blot analysis of hippocampal membranes from BDNF heterozygous (+/−) and knockout (−/−) mice, which have, respectively, half or no BDNF compared with wild-type (see Ernfors et al., 1994 ; Agerman et al., 2003 ). Supporting the above prediction, the membranes of 21-d-old BDNF −/− mice and the membranes from 10-mo-old BDNF +/− and +/+ mice all presented a highly phosphorylated TrkB signal (Figure 1, C and D).
Numerous studies have shown that embryonic hippocampal neurons from wild-type rodents, mouse or rat, acquire in vitro many of the complex aspects of neuronal architecture and function of their in vivo counterparts. Importantly, these features are acquired without the need of supplementation by BDNF, or any other neurotrophins, to the culture medium (Kaech and Banker, 2006 ). Furthermore, survival and acquisition of mature molecular and functional phenotype does not seem to require endogenous ligand. In agreement, we observed that the levels of BDNF in the culture medium of these neurons, as determined by ELISA, ranged between 0.07 and 4.38 pM. Because the highest of such values is 30-fold lower than the concentrations required to elicit TrkB phosphorylation by BDNF in these same cells (see Suzuki et al., 2004 ), we have considered these cells as a most suitable system to study the possible mechanisms that may trigger TrkB activity independently from high levels of BDNF, as just was demonstrated to be the case in situ. Furthermore, this system is also most suitable as TrkB is the main Trk isoform in this culture (Supplementary Figure 1B).
To determine if hippocampal aging in vitro is accompanied by increased TrkB phosphorylation, as in the in vivo situation (see Figure 1A), the levels of phosphorylated TrkB were measured in membranes from hippocampal neurons maintained in culture for 10, 15, and 26 DIV, representative of immediate terminally differentiated, fully mature, and aging neurons, respectively. This study revealed weak levels of TrkB phosphorylation in 10 DIV neurons, 10% higher in 15 DIV neurons, and a further and a most significant increase in aged 26 DIV neurons (Figure 2A). In agreement with the biochemical data, immunofluorescence microscopy analysis of phospho-TrkB distribution revealed more and more peripherally distributed structures in aged (20 DIV) than in 10 DIV neurons (Figure 2B). Because 20 DIV neurons also presented fully active TrkB at the biochemical level, identical to 26 DIV neurons, 20 or 26 DIV were indistinctly taken as representative of the in vitro aged situation.
In support that higher levels of TrkB phosphorylation during aging were transduced downstream, Akt presented an activity curve similar to that of TrkB: low levels at 10 DIV, evident at 15 DIV, and highest by 26 DIV (Figure 2C), and the incubation of 15 DIV neurons with the kinase inhibitor K252a (Lee et al., 2002 ) blocked the occurrence of Akt phosphorylation (Figure 2D).
To further strengthen the notion that hippocampal neurons in vitro is a good system to study the possible mechanisms behind robust TrkB activity in low neurotrophin conditions, the activity of the receptor was next assessed in hippocampal neurons in vitro from BDNF knockout embryos (obtained by crossing BDNF +/− mice; see Materials and Methods), grown in the absence of any exogenous neurotrophin. In these cells, rate of hippocampal neuron survival and the degree of morphological differentiation was, as observed in vivo (Ernfors et al., 1994 ), undistinguishable from wild-type neurons (Figure 3A). More relevant for the purpose of this work, 15 DIV neurons from BDNF −/− embryos presented a phosphorylated TrkB signal, not different from that of neurons from wild-type littermates at the same age in vitro (Figure 3B). This last result conclusively demonstrates that hippocampal neurons have mechanisms to elicit TrkB activity independently from the presence of the receptor natural ligand. To test whether or not such activity could be attributed to excess production/release of low-affinity ligands, we shifted back to the rat embryo hippocampal neurons from wild-type animals, which present robust TrkB activity in the absence of exogenous neurotrophins (see Figure 2A). For this purpose, the culture medium of 26 DIV neurons, in which low-affinity ligands would be highest if responsible for the presence of high receptor phosphorylation at this stage, was added to 10 DIV cells, that present only weak TrkB activity (see Figure 2 A). This treatment did not induce TrkB activity in 10 DIV cells (Figure 3C), suggesting that activation in the old neurons was not due to increased production of NT3/NT4 or adenosine (see Introduction). In further agreement, incubation of 15 DIV neurons with a cocktail of neurotrophin function-blocking antibodies neither prevented nor reduced the spontaneous appearance of TrkB activity in these neurons (Figure 3D). On the other hand, the antibody cocktail was efficient to inhibit the activity induced by supraphysiological amounts of BDNF (Figure 3E), ruling out that their inefficacy to block spontaneous activation in old neurons (Figure 3D) was due to lack of inhibitory activity or the capacity to reach the sites of receptor location on the neuronal surface (synaptic cleft). Surely, these data do not exclude that activation of TrkB during aging may be due to binding to as yet unidentified or nonsoluble ligands, which might have become up-regulated with aging. On the other hand, these data, together with the previous in situ results, clearly demonstrate that differentiated hippocampal neurons posses mechanisms that guarantee TrkB activity even under conditions of minimal availability of canonical neurotrophins.
Plasma membrane cholesterol levels play a critical role in receptor signaling strength (see Introduction). Hence, we found it most natural to test the possibility that high TrkB activity in aging neurons could be due to the occurrence of changes in the content of this lipid. Therefore, we measured cholesterol levels in the membranes of 10, 15, and 26 DIV hippocampal neurons. Colorimetric measurement revealed that 10 DIV neurons have 10 and 25% more cholesterol than 15 and 26 DIV neurons, respectively (Figure 4A). Together with the data of Figure 2, these last results indicate the existence of an inverse cholesterol level–TrkB activity correlation (see Figure 2).
The biochemical isolation of DRMs is a useful tool to confirm the existence of changes in the organization of cellular membranes (Brown and London, 1997 ). Therefore we speculated that the decrease in cholesterol content with ageing in vitro would be reflected in the content of this lipid in DRMs prepared from hippocampal neurons cultured for 10, 15, and 26 d (see Materials and Methods). TLC of equal protein loads confirmed the clear reduction in cholesterol content in15 and 26 DIV neurons, compared with 10 DIV neurons (Figure 4B). Because changes in cholesterol levels affect the degree of activation of numerous proteins, including Trks, and this is reflected in changes in the way these proteins associate with DRMs (Paratcha and Ibanez, 2002 ; Hancock, 2006 ; Jacobson et al., 2007 ), we next determined the degree of TrkB phosphorylation in DRMs throughout aging. This revealed TrkB activity confined to DRMs at 26 DIV. In contrast, TrkB activity was undetectable in the DRMs of 10 DIV neurons (Figure 4C, p-Trk). To test if the fractions containing active receptors corresponded to DRMs (despite their lower cholesterol levels), the same membranes were blotted with an antibody against the glycosylphosphatidylinositol-anchored PrPc. That this is the case is shown in Figure 4C (PrPc). In further support that TrkB is present in DRMs of old neurons in vitro and that this activity may be relevant to neuronal survival, the p85 subunit of PI3K, a downstream target of TrkB activity, was precipitated exclusively from the DRMs of differentiated neurons only in complex with the active receptor (Figure 4D). On the other hand, the transmembrane protein BACE1, which was shown is associated with DRMs of neurons in a cholesterol-dependent manner (Ledesma et al., 2003 ; Abad-Rodriguez et al., 2004 ) became more concentrated in detergent-soluble membranes of the aging neurons (Figure 5A). In good agreement with the presence of TrkB in DRMs of cells with low cholesterol, the TrkB DRM-localizing enzyme Fyn (Pereira and Chao, 2007 ) was also found considerably enriched in DRMs of aging neurons, compared with younger cells (Figure 5B).
Increased activity due to cholesterol loss was observed for Fyn (Ko et al., 2005 ), epidermal growth factor receptor (Oh et al., 2007 ), and TGF-β (Chen et al., 2007 ). To test if this is also the case for the increased activity of TrkB in aging neurons, the levels of cholesterol were lowered in 10 DIV neurons, until reaching those of aging neurons (25% reduction, see Figure 4A). This was accomplished by the addition of a cholesterol synthesis inhibitor (mevilonin) and a cholesterol removal drug (β-methyl-cyclodextrin; Ledesma et al., 2003 ; Abad-Rodriguez et al., 2004 ; see Materials and Methods). Consistent with the existence of a direct causal relationship, treated 10 DIV neurons presented strong TrkB phosphorylation levels, far above those of nontreated neurons of similar age (Figure 6A). This effect occurred in the DRMs of these neurons, similarly to what happens in untreated aging neurons (see Figure 4D). In agreement with the effect on TrkB activity being due to the true reduction of cholesterol levels, the association of Flotilin1 with DRMs was decreased in the cholesterol-deprived 10 DIV neurons (Figure 6B). To further control that the effects on TrkB were due to the loss of cholesterol and not to a β-methyl-cyclodextrin–mediated inhibition of the lateral mobility of the receptor, as shown for other proteins (Goodwin et al., 2005 ), cholesterol was added back to the mevilonin/cyclodextrin-treated 10 DIV neurons. This resulted in the suppression of TrkB signal in these cells (Figure 6A). In further agreement, cholesterol replenishment resulted in the reappearance of Flotilin 1 in the DRMs of these cells (Figure 6B). Treatment of young neurons with mevilonin did not result in receptor activation (Supplementary Figure 1C), consistent with the inefficacy of this treatment alone to reduce the levels of cholesterol in these cells (Simons et al., 1998 ).
To test the involvement of cholesterol loss in TrkB activation by means different from pharmacological reduction/replenishment, which may lead to a number of side effects (Kwik et al., 2003 ; Hancock, 2006 ), we searched for the possible molecular event responsible for cholesterol loss in aging cells. To this aim, we performed a real-time PCR-based analysis of the major cholesterol anabolic and catabolic enzymes, utilizing as template mRNA from hippocampal neurons at different stages of development. This study revealed a twofold increase in the expression of the brain-specific catabolic enzyme cholesterol-24-hydroxylase, also known as CYP46 (Lund et al., 1999 ; Figure 7A), without significant changes in the levels of the metabolic enzymes. In agreement, also the levels of the protein were higher in 20 DIV neurons (Figure 7B). Having established this link, we next modified the neurons' cholesterol levels by altering the expression of cholesterol-24-hydroxylase through canonical gain (cDNA expression) and loss (shRNA knockdown) of function approaches. To test if cholesterol loss can produce increased TrkB activity, cyp46A1 was overexpressed in 10 DIV neurons, which, as shown earlier, present weak to undetectable TrkB activity. Consistent with the importance of low cholesterol levels, overexpressing cells showed higher levels of TrkB phosphorylation (Figure 7C). In further agreement, knocking cyp46A1 down in aged cells, led to reduced levels of TrkB activity detected either by Western blot (Figure 7C) or by immuno-fluorescence (IF) microscopy (Figure 7D).
Numerous evidences have accumulated over the years that unequivocally prove that the BDNF-TrkB pathway is a crucial player in the survival of certain neuronal populations during development and in synapse fine-tuning in the mature nervous system (Egan et al., 2003 ; Blum and Konnerth, 2005 ; Reichardt 2006 ). We here showed that BDNF can be dispensable for TrkB activation toward survival in the mature hippocampus, both in vitro and in situ. Because the adult hippocampus and hippocampal cells do, however, present strong TrkB survival activity, a most logical conclusion from the BDNF-independent data are that in the adult brain TrkB activity may be guaranteed by more than one mechanism, acting in parallel or complementarily to BDNF. One such mechanisms is via the modulation of cholesterol content. This conclusion is supported by a temporal correlation, i.e., cholesterol is reduced during ageing in vitro and in situ and also by a direct cause–effect evidence; i.e., 20–30% reduction of cholesterol in young neurons elicited receptor activation. It is difficult to argue that such effect might have derived from inhibition of the lateral diffusion of TrkB (induction of clustering) triggered by the cholesterol-removing drug-utilized, β-methyl-cyclodextrin (Goodwin et al., 2005 ), because TrkB activity in the young cells was reverted to control levels (i.e., undetectable) upon restoring cholesterol to normal (for this age) levels. Whether or not a loss of cholesterol may cause TrkB activity in vivo is a matter that needs further attention. However, in support that this may be the case, reduced cholesterol levels were observed in the hippocampus of aged mice, together with a reduction of Flotilin1 from DRMs and increased expression of cholesterol-24-hydroxylase (Figure 8, A–D), clearly resembling the in vitro data.
The demonstration that cholesterol reduction can lead to increased TrkB activity constitute, at first sight, a most paradigmatic example of cholesterol control of cellular signaling. In fact, for the most cases, cholesterol loss leads to reduced signaling (Simons and Toomre, 2000 ). On the other hand, recent articles have shown that the opposite effects (i.e., reduction leading to increased signaling) is also possible (Kalvodova et al., 2005 ; Chen et al., 2007 ; Oh et al., 2007 ), indicating that the outcome of cholesterol reduction will depend on the type of receptor, cell type, and stage of differentiation. The increased activity for TrkB under low cholesterol conditions may reflect autoactivation by dimerization due to slower diffusion on the plane of the membrane. Alternatively, cholesterol reduction may have triggered receptor activation in the early secretory pathway or in endosomes, because of altered transport or recycling. In any event, the way that cholesterol modulates TrkB activity in differentiated neurons is certainly different from that in developing neurons, in which receptor activation is reduced by cholesterol-loss treatments (Suzuki et al., 2004 ). The dissection of the mechanisms of TrkB cholesterol dependence and independence in adult and young neurons is of major biological relevance and will need extensive work before we can understand it. Unfortunately, the overexpression of TrkB variants may not be very useful for this purpose, because it leads to receptor activation independently from ligand and cholesterol concentrations.
Cholesterol loss in differentiated neurons, both in vivo and in vitro, seems to occur via up-regulation of the catabolic enzyme cholesterol-24-hydroxylase. This brain specific enzyme is responsible for the conversion of cholesterol into 24S-hydroxycholesterol, the excreted form of cholesterol (Lund et al., 1999 , 2003 ), an event that progressively increases with age (Lutjohann, 2006 ). Although this does not rule out that other events contribute to brain cholesterol reduction during ageing, such as a decay in the activity of synthesizing enzymes or a reduction in cholesterol transport from astrocytes, overexpression of cholesterol24-hydroxylase induced TrkB activation in young neurons and its knockdown prevented its occurrence in the old neurons, proving that modulation of this enzyme can act as a sufficient and necessary stimulus for receptor activation. Although the finding that the activity of the enzyme is also high in the differentiated hippocampus in situ makes us feel confident that this pathway might operate in vivo, direct validation is needed. Quite encouraging that this may be the case, the only (up to now) event demonstrated to increase the expression of cholesterol-24-hydroxylase activity is oxidative stress (Ohyama et al., 2006 ). Unfortunately, the use of cholesterol-24-hydroxylase knockout mice cannot be envisioned for this purpose, as these mice compensate decreased cholesterol excretion by a reduction in de novo synthesis (Lund et al., 2003 ).
The observation that up-regulation of cholesterol-24-hydroxylase can activate TrkB makes unavoidable to discuss its relevance for Alzheimer's disease (AD). In fact, although still controversial, several groups have found a significant association between intronic mutations in this gene and increased risk of AD (Kolsch et al., 2002 ; Papassotiropoulos et al., 2003 ; Borroni et al., 2004 ; see however, Desai et al., 2002 ; Wang and Jia, 2007 ). Although we here report that cholesterol loss, triggered by the up-regulation of cholesterol-24-hydroxylase, leads to TrkB activation, we have previously demonstrated that cholesterol loss of more than 30% can lead to apoptosis (Ledesma et al., 2003 ; Abad-Rodriguez et al., 2004 ). In this context, and considering that TrkB activity is important for survival of differentiated cortical neurons under stress (Gates et al., 2000 ), it is quite conceivable that intronic mutations could predispose to AD through a “gain-of-function” process, i.e., premature and constant loss of cholesterol leading to cell death upon the occurrence of further loss (beyond 30%) triggered when age-associated (i.e., cumulative) stress passes the physiological threshold. But it is also possible that the intronic mutations found in certain cases of AD may predispose to disease by a “loss-of function” mechanism, whereby this enzyme would not respond to physiological stress and therefore prevent the activation of an important antiapoptotic pathway like that of TrkB. To distinguish between these possibilities is another interesting derivation from this work.
We thank Outi Nikkila for preparing the BDNF +/− crosses; Kristel Vennekens (University of Leuven) for the hippocampal cultures; Francesca Gilli and Fabiana Marnetto for help with the real-time PCR experiments; and Laura Doglio, Jose Abad-Rodriguez, and Maria D. Ledesma for discussions and technical advice. The following institutions provided financial support: Swedish Medical Research Council and the Alzheimer's Association (T.H.), Sigrid Juselius Foundation and Academy of Finland (E.C.), and Fund for Scientific Research Flanders (FWO), Federal Office for Scientific Affairs (IUAP P6/43/), and Flemish Government (Methusalem grant; C.G.D.).
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E07-09-0897) on February 20, 2008.