CNS Synaptogenesis Promoted by Glia-Derived Cholesterol

See allHide authors and affiliations

Science  09 Nov 2001:
Vol. 294, Issue 5545, pp. 1354-1357
DOI: 10.1126/science.294.5545.1354


The molecular mechanisms controlling synaptogenesis in the central nervous system (CNS) are poorly understood. Previous reports showed that a glia-derived factor strongly promotes synapse development in cultures of purified CNS neurons. Here, we identify this factor as cholesterol complexed to apolipoprotein E–containing lipoproteins. CNS neurons produce enough cholesterol to survive and grow, but the formation of numerous mature synapses demands additional amounts that must be provided by glia. Thus, the availability of cholesterol appears to limit synapse development. This may explain the delayed onset of CNS synaptogenesis after glia differentiation and neurobehavioral manifestations of defects in cholesterol or lipoprotein homeostasis.

The formation of synaptic contacts is a critical phase during brain development and plays a crucial role in long-term synaptic plasticity in the adult CNS, but the cell biological mechanisms that mediate the assembly of the synaptic machinery are still poorly understood. A possible role of glial cells in CNS synaptogenesis was indicated by a series of studies (1–3) on rat retinal ganglion cells (RGCs), CNS neurons that can be highly purified (4) and cultured under defined glia-free conditions (5). The previous reports (1–3) showed that neurons formed few and inefficient synapses in the absence of glia and that glial cells induced the formation of numerous and highly efficient synapses without affecting neuronal survival, excitability, or neurite outgrowth. These effects were mediated by a soluble factor secreted by macroglial cells that has remained elusive so far.

To identify the glial factor, we pursued two experimental approaches. First, we fractionated glia-conditioned medium (GCM) (6) chromatographically, treated glia-free cultures of purified RGCs with elution fractions, and assayed their activity by whole-cell recordings of spontaneous excitatory postsynaptic currents (EPSCs) (7). The GCM-induced increase in synaptic activity was mediated by a heparin-binding factor that could be further purified by cation exchange chromatography (8, 9). Subsequent gel filtration (10) indicated that the factor migrated with large components of GCM (Fig. 1A). Second, we studied GCM-induced changes in the protein composition of membranes prepared from RGCs by two-dimensional gel electrophoresis (2-DE) (11–13). GCM treatment reliably caused the appearance of a new silver-stained spot (n = 3 preparations) (Fig. 1B). Microsequencing by nanospray mass spectrometry (14, 15) showed that this spot contained apolipoprotein E (apoE) (Fig. 1C). In the CNS, apoE is produced by macroglial cells (16–18). Therefore, it appeared likely that the apoE occurring in GCM-treated cells was of glial origin. Accordingly, immunoblots showed a lack of apoE in RGCs grown under glia-free conditions (19).

Figure 1

Glia-derived cholesterol complexed to apoE-containing lipoproteins mediates the GCM-induced increase in synaptic activity. (A) Mean frequency of spontaneous EPSCs in RGC cultures treated with elution fractions obtained by gel filtration chromatography of prefractioned GCM. Inset: Calibration indicated a size range between 158 kD (9.6 nm) and 669 kD (17 nm) for the glial factor (gray shading, elution range of glial factor; arrows, molecular weight markers aldolase and thyroglobulin). (B) Silver-stained two-dimensional gels of membrane proteins from RGCs cultured without (left) and with (right) GCM (approximate range, molecular weight 28 to 45 kD, pI 4.5 to 5.5). Arrowhead, GCM-induced protein spot. (C) Identification of the protein marked in (B) as glia-derived apoE. The upper panel shows the nanoelectrospray mass spectrum of the peptide mixture from the trypsinized spot. Arrows indicate peptides selected for fragmentation. The lower panel shows the fragmentation spectrum of the doubly charged ion (mass-to-charge ratio, 620.34). Circles mark matches between peaks in the fragmentation spectrum and theoretical fragment masses for the identified apoE peptide (47). (D) Relative frequency of spontaneous EPSCs in individual RGC cultures treated with apoE (1 and 10 μg/ml), cholesterol (10 μg/ml), or GCM, all compared to untreated control cultures. (E) Cholesterol (upper panel: thick line, left axis) and protein concentration (thin line, right axis) and apoE content (lower panel) in gel filtration fractions. In fraction 11, which had the strongest effect on synaptic activity as shown in (A), the combination of apoE and cholesterol content was maximal. (F) Lowering the cholesterol concentration in GCM (LC-GCM, n = 5 cultures) reduced the GCM-induced increase in synaptic activity, and addition of cholesterol (10 μg/ml) to LC-GCM (LC-GCM+) fully restored the effect (n = 2 cultures). Inhibition of lipoprotein uptake by RAP (50 μg/ml) (n = 4 cultures) reduced the heparin eluate (HepElu)–induced increase in synaptic activity (31). In RAP-treated cultures, 6% of the RGCs tested (2 of 32) still showed strong synaptic activity. Open and solid columns represent mean values with and without these cells, respectively. Bars indicate SEM.

Initially, glia-derived apoE appeared to be a good candidate for the glial factor. It binds to heparin (20) and is part of glia-derived lipoprotein particles, whose size (21) falls in the same range that we had determined for the glial factor (Fig. 1A, inset). However, in RGC cultures treated with apoE (1 or 10 μg/ml) (22), the increase in EPSC frequency as compared to untreated controls averaged 0.8 ± 0.3–fold (mean ± SEM,n = 4 cultures; Fig. 1D), showing that apoE was not the glial factor. Because apoE-containing lipoproteins serve as cholesterol carriers (23), we then tested whether cholesterol mimicked the effects of GCM. Treatment of RGC cultures with cholesterol (10 μg/ml) (22) increased the frequency of spontaneous EPSCs on average by 12 ± 4–fold as compared to glia-free cultures and thus to a similar extent as treatment with GCM (15 ± 4–fold; 86 ± 20% of GCM control; n = 13 cultures) (Fig. 1D). We also tested whether this effect could be induced by other lipids contained in glia-derived lipoprotein particles (21,24). However, treatment of neurons with phosphatidylcholine (0.4 μM; 2 ± 1–fold; n = 4 cultures) or sphingomyelin (0.4 μM; 0.3 ±0.2–fold;n = 4) (22) did not significantly increase the frequency of spontaneous EPSCs as compared to untreated controls. At a higher concentration (2 μM), these lipids were toxic to purified RGCs. These results suggested that cholesterol complexed to large apoE-containing lipoproteins was the active component in GCM.

To test this hypothesis, we performed three experiments. First, we determined whether those gel filtration fractions with a strong effect on synaptic activity contained elevated levels of cholesterol and apoE (25). Our measurements showed that there was a close match between the synaptogenic activity of elution fractions and their combined content of cholesterol and apoE (Fig. 1E). Second, we tested whether lowering the cholesterol concentration in GCM would diminish its effect on synaptic activity. To produce low-cholesterol GCM (LC-GCM), we added mevastatin, an inhibitor of cholesterol synthesis (26), to glial cultures (27). Mevastatin (10 μM) lowered the cholesterol concentration from 5.0 ± 0.7 μg/ml in GCM to an undetectable level in LC-GCM (n = 4 cultures). In RGCs treated with LC-GCM, the spontaneous activity was strongly reduced as compared to the activity in cultures treated with normal GCM (Fig. 1F). This effect could be due to the drug-induced loss of a synaptogenic component other than cholesterol. However, supplementing LC-GCM with cholesterol (10 μg/ml) (LC-GCM+) fully rescued the strong increase in spontaneous synaptic activity (n = 2 cultures) (Fig. 1F), confirming that cholesterol was the active component. Third, we tested whether the GCM-induced increase in synaptic activity required apoE-binding receptors that mediate the uptake of lipoproteins (28). To accomplish this, we blocked members of the low-density lipoprotein (LDL) receptor family (29) with the receptor-associated protein (RAP) (30, 31). In RGCs treated with RAP and heparin-binding components of GCM, the spontaneous synaptic activity was reduced, as compared to the synaptic activity of controls treated with heparin eluate alone (Fig. 1F). In summary, five lines of evidence indicated that cholesterol produced by glial cells and secreted in apoE-containing lipoproteins was the glial factor. First, cholesterol alone strongly increased the frequency of spontaneous synaptic events, whereas other components of glia-derived lipoprotein particles did not affect synaptic activity. Second, there was a close match between the reported size of glial lipoproteins and the size of the synapse-promoting factor. Third, the level of activity in gel filtration fractions scaled with the combined content of cholesterol and apoE. Fourth, a drastic reduction of the cholesterol content in GCM by mevastatin eliminated its effects on synaptic activity. Fifth, inhibition of lipoprotein uptake reduced the effect of the heparin-binding fraction on synapses.

The identification of cholesterol as the glial factor prompted us to determine the content and distribution of this important membrane component (32) in RGCs grown in the presence and absence of GCM. To accomplish this, we stained cultures with filipin, a fluorescent antibiotic that binds specifically to cholesterol (33, 34). In the absence of GCM, somata and neurites of RGCs showed detectable filipin staining with the background-corrected fluorescence intensity in neuronal somata averaging 48 ± 1 analog-to-digital units (adu's) (n = 89 cells) (Fig. 2). In RGCs treated with GCM, however, filipin staining was much stronger, averaging 115 ± 4 adu's in neuronal somata (n = 60 cells) (Fig. 2). Thus, RGCs produced measurable levels of cholesterol, when cultured under defined conditions, and incorporated large amounts of glia-derived cholesterol when treated with GCM.

Figure 2

Cholesterol content of cultured RGCs. (A) Fluorescence micrographs of purified RGCs cultured for 18 days in the absence of GCM (left) or for the last 7 out of 18 days in the presence of GCM (right) and stained with the cholesterol-binding antibiotic filipin. Both images were acquired under the same conditions. Scale bar, 30 μm. (B) Cumulative relative frequency distribution of background-corrected fluorescence intensities in somata of RGCs cultured without (thin line, n = 89 cells) or with (thick line, n = 60 cells) GCM pooled from three cultures (P < 0.001, Student's t test).

How does glia-derived cholesterol increase the frequency of spontaneous EPSCs? Previous studies show that GCM increases spontaneous synaptic activity by inducing the formation of new synapses and by enhancing the quantal size and efficacy of transmitter release (2, 3). To determine whether these effects were mediated by cholesterol, we studied microcultures of RGCs (35), in which synapse development can be quantified in individual neurons (2). The results showed that cholesterol essentially mimicked the previously reported effects of GCM on autapse development (2). Cholesterol raised the number of immunostained autapses by 8-fold and the quantal content per neuron by 10-fold, and thus acted similarly to GCM (Fig. 3, A and B). Analysis of neurite growth showed that these increases were caused by a direct synaptogenic effect rather than by a corresponding increase in neurite length (36). Furthermore, cholesterol increased the efficacy of transmitter release, as indicated by the low failure rate (Fig. 3C) and the nonlinear increase in asynchronous release (Fig. 3C), as compared to spontaneous release (Fig. 3D). Cholesterol also increased the charge transfer of spontaneous excitatory autaptic currents (EACs), but to a smaller extent than GCM (Fig. 3D), suggesting that additional glial signals control the quantal size.

Figure 3

Cholesterol increases the number and release efficacy of synapses in single RGCs. (A) Fluorescence micrographs showing RGCs cultured in the absence (left) or presence (right) of cholesterol and labeled by a synapsin I–specific antibody. Scale bar, 5 μm. (B) Number of synapsin- and GRIP-positive puncta per neuron (black columns, left axis) and quantal content of evoked EACs (white columns, right axis) in glia-free microcultures [nfor RGC = 10 cells (left axis)/4 cells (right axis)], in the presence of cholesterol (10 μg/ml, +Chol 25/16) or of GCM (+GCM 12/36). (C) Failure rate of evoked EACs (black columns, left axis) and frequency of asynchronous events (white columns, right axis) (n for RGC = 12, +Chol 20, +GCM 45). (D) Frequency (black columns, left axis) and charge transfer (white columns, right axis) of spontaneous EACs (n for RGC = 12/31, +Chol 22/408, +GCM 45/1402). Error bars indicate SEM. All cholesterol- and GCM-induced changes were statistically significant (P < 0.05, Student's ttest).

How does cholesterol promote synapse formation? Cholesterol increased the number of synapsin I–positive puncta contained in neurites of RGCs (Fig. 3A) and thus mimicked a previously reported effect of GCM (2). We noticed, however, that cholesterol induced a stronger effect than GCM. In cholesterol-treated microcultures, the number of synapsin I–positive puncta per RGC was 69 ± 23% (n = 4 cultures) larger than in GCM-treated controls. This effect was not restricted to synapsin I; cholesterol also induced 57 ± 3% more synaptophysin-positive puncta than GCM (n = 2 cultures). Double immunostaining of cholesterol-treated cultures (37) revealed that the large majority of synapsin-positive puncta (61 ± 7%,n = 9 neurons) also contained synaptophysin, which indicates that these puncta represented synaptic vesicles or precursor material contained in presynaptic terminals or transport packets (38), respectively. These results suggested that cholesterol enhances the production of presynaptic components including synaptic vesicles (39) and release sites (40). Our previous finding (2) that GCM does not affect autapses immediately, but within 24 to 48 hours, suggests that the synaptogenic effect is mediated by such a slow multistep process.

Our study indicates that the ability of CNS neurons to form synapses is limited by the availability of cholesterol. RGCs cultured under serum- and glia-free conditions produced enough cholesterol to survive, to differentiate axons and dendrites (5), and to form a few immature synapses. Massive synaptogenesis requires large amounts of cholesterol and thus depends on cholesterol production by glial cells and its delivery via apoE-containing lipoproteins. Because cholesterol in the CNS is synthesized in situ rather than imported from blood (41–43), a link between synaptogenesis and glia-derived cholesterol could explain why most synapses in the developing brain are formed after the differentiation of macroglial cells (3, 44). Our results imply that genetic or age-related defects in the synthesis, transport, or uptake of cholesterol in the CNS (30, 45) may directly impair the development and plasticity of the synaptic circuitry.

  • * To whom correspondence should be addressed. E-mail: fw-pfrieger{at}


View Abstract

Navigate This Article