Ganglioside Synthesis during the Development of Neuronal Polarity MAJOR CHANGES OCCUR DURING AXONOGENESIS AND AXON ELONGATION, BUT NOT DURING DENDRITE GROWTH OR SYNAPTOGENESIS*

Changes in the levels and types of gangliosides occur during neuronal differentiation and development, but no studies have correlated these changes with defined events in neuronal morphogenesis. Here, we have analyzed the relationship between ganglioside synthesis and the development of axons and dendrites in polarized neurons, using hippocampal neurons cultured in such a way that axons and dendrites are generated by a defined sequence of events and in which there is virtu- ally no contamination by glial cells. Neurons were labeled with [4,5- 3 H]dihydrosphingosine, which was rapidly incorporated into cells and metabolized to 3 H-labeled glycosphingolipids. The rate of 3 H-labeled glycosphingolipid synthesis was directly proportional to the initial rate of [4,5- 3 H]dihydrosphingosine uptake and was linear versus time for up to 9 h of incubation. The major changes in 3 H-labeled ganglioside synthesis occurred during the period of axonogenesis and rapid axon growth. During axonogenesis, there was a significant increase in the synthesis of complex gangliosides ( i.e. G M1 , G D1a , G D1b , and G T1b ) with a corresponding reduction in the synthesis of glucosylceramide and ganglioside G D3 . During the stage of rapid axon growth, the ratio of a- to b-series gangliosides rhodamine Contax

The sialic acid-containing glycosphingolipids (GSLs), 1 the gangliosides, are found at high levels in the plasma membrane (1) of neuronal tissues and may play important roles in neuronal function (2,3). GSL synthesis begins at the cytosolic surface of the endoplasmic reticulum, with formation of the sphingoid long chain base (4) and of dihydroceramide (4,5), and is com-pleted in the Golgi apparatus, where most of the glycosylation reactions occur (6,7). The first ganglioside synthesized is G M3 , 2 formed by sialylation of lactosylceramide (Fig. 1). G M3 can be sialylated to G D3 , the first ganglioside of the b-series, or modified by addition of N-acetylglucosamine to G M2 , the first ganglioside of the a-series. Gangliosides G M3 and G D3 are referred to as "simple" gangliosides, and all subsequent gangliosides in the metabolic pathway as "complex" gangliosides ( Fig. 1) (6). Ganglioside synthesis can be studied using various precursors, including radioactive sphingoid bases such as [4, H]dihydrosphingosine (5,8,9).
Many studies have demonstrated that ganglioside levels and types change during neuronal differentiation and development (see, for example, Refs. 10 -18). However, the ability to correlate ganglioside synthesis with a specific developmental event has been limited by the use of neuronal cell lines, in which the main developmental event is transformation of a non-neuronal to a neuronal phenotype (i.e. formation of a neurite), or of primary cultures, in which neuronal development is not well characterized and in which the contribution of glia to ganglioside synthesis is difficult to distinguish from that of neurons. Ganglioside synthesis has also been analyzed in developing brain (13), but similar problems of characterization and glial contamination exist.
Here, we have analyzed ganglioside synthesis in hippocampal neurons cultured in such a way that axons and dendrites develop by a known sequence of events and can be distinguished both morphologically and biochemically (19). Hippocampal neurons have been extensively characterized when cultured at low densities (19 -25). Cells are cultured from the hippocampus of embryonic day 18 rats (20); at this time, the hippocampus is inhabited mainly by pyramidal neurons that are at the transition stage of withdrawal from proliferation and at the beginning of differentiation (26). In the initial stages of growth (stages 1 and 2; see Ref. 19), each neuron develops a number of short processes, and after some hours, one of the processes starts to grow rapidly. The rapidly growing process develops axonal characteristics (stage 3) and can be distinguished from the minor processes by the presence of proteins such as GAP-43 (22) and synaptophysin (23). During the next stage of growth (stage 4), minor processes begin to elongate, * This work was supported by the German-Israel Foundation for Scientific Research and Development. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Incumbent of the Recanati Career Development Chair in Cancer Research. To whom correspondence should be addressed. Tel.: 972-8-9342704; Fax: 972-8-9344112; E-mail: bmfuter@weizmann.weizmann. ac.il. 1 The abbreviations used are: GSLs, glycosphingolipids; MAP-2, microtubule-associated protein-2; GlcCer, glucosylceramide; LacCer, lactosylceramide; PSL, phospholuminescence. albeit at much slower rates than the axon, and eventually acquire the typical characteristics of dendrites; dendrites can be distinguished from axons by the presence of microtubuleassociated protein-2 (MAP-2) (21). During this period, axons continue to grow, although the rates of axon growth compared with earlier stages have not been determined. The final stages of development (stage 5) result in fully polarized neurons with functional synapses.
We have examined the relationship between the synthesis of particular gangliosides, or of classes of gangliosides, and these well defined stages of neuronal development. Surprisingly, the major changes in synthesis occur during the first 5 days of culture, the period of axonogenesis and rapid axon elongation, but few changes occur at later stages of development, including the period of dendrite growth and synaptogenesis.

EXPERIMENTAL PROCEDURES
Hippocampal Cultures-Hippocampal neurons were cultured at high density based on methods used for low density hippocampal cultures (20). The dissected hippocampi of embryonic day 18 rats (Wistar) were dissociated by trypsinization (0.25% (w/v) for 15 min at 37°C). The tissue was washed in Mg 2ϩ /Ca 2ϩ -free Hanks' balanced salt solution (Life Technologies, Inc.) and dissociated by repeated passage through a constricted Pasteur pipette. Cells were plated in minimal essential medium with 10% horse serum at a plating density of 240,000 cells/ 24-mm glass coverslip (precoated with poly-L-lysine (1 mg/ml)). After 3-4 h, coverslips were transferred into 100-mm Petri dishes (Nunc) containing a monolayer of astroglia. Coverslips were placed with the neurons facing downwards and separated from the glia by paraffin "feet." Cultures were maintained in serum-free medium (minimal essential medium) that included N 2 supplements (20), ovalbumin (0.1%, w/v), and pyruvate (0.1 mM). Glial proliferation was prevented by ad-dition of cytosine ␤-D-arabinoside (5 M) to the medium on day 2 of culture.
Characterization of Neuronal Development-Neuronal development was assessed by immunofluorescence using antibodies against axonspecific (GAP-43 and synaptophysin) and dendrite-specific (MAP-2) proteins (20). Glia were identified using an anti-glial fibrillary-associated protein antibody. Neurons were fixed in 4% paraformaldehyde in phosphate-buffered saline containing 4% sucrose for 20 min at 37°C, permeabilized with 0.25% Triton X-100 for 5 min at 37°C, and then incubated for 1 h at 37°C with primary antibodies (Biomakor, Kiryat Weizmann, Rehovot, Israel); anti-glial fibrillary-associated protein, anti-MAP-2, and anti-synaptophysin antibodies were diluted 1:200 in phosphate-buffered saline, and the anti-GAP-43 antibody was diluted 1:1000. A rhodamine-conjugated goat anti-mouse second antibody (Jackson Im-munoResearch Laboratories, Inc.) was used for detection. Cells were observed using Plan Apochromat 40ϫ/1.3 NA or 63ϫ/1.4 NA oil objectives of a Zeiss Axiovert 35 microscope with an appropriate filter for rhodamine fluorescence. Cells were photographed using a Contax 167MT camera and Eastman Kodak 400 film.
Synthesis of [4, H]Dihydrosphingosine- [4, H]Dihydrosphingosine was synthesized by reduction of D-erythro-sphingosine (Sigma) with NaB 3 H 4 (10 Ci/mmol) (5,27). Sphingosine (10 mg) was dissolved in tetrahydrofuran/H 2 O (1:1, v/v), frozen in liquid N 2 , and overlaid with Pd(Ac) 2 (5 mol), glacial acetic acid (20 l), and NaB 3 H 4 (100 mCi; dissolved in 1 M NaOH). After each overlay, the reaction mixture was refrozen. Tetrahydrofuran was passed over an activated basic alumina column prior to use to remove peroxides. The frozen vial was flushed with argon, sealed, and allowed to thaw to 25°C, and the reaction mixture was stirred for 24 h. The reaction was terminated by filling the vial with water, and the reaction mixture was passed over an RP-18 reverse-phase chromatography column (Merck  [4, H]dihydrosphingosine (10 Ci/mmol) was added to the Petri dishes so that the final concentration of ethanol in the medium did not exceed 0.5%. After various times of incubation, coverslips were washed with Hanks' balanced salt solution, cells were removed by scraping with a rubber policeman, and coverslips were washed four times with water. The suspended cells were lyophilized, and the dry material was extracted using CHCl 3 /CH 2 OH/H 2 O/pyridine (60:30:6:1, v/v/v/v) for 48 h at 48°C (28). Phospholipids were degraded by mild alkaline hydrolysis with methanolic NaOH (100 mM) for 2 h at 37°C; ϳ93% of the 3 H radioactivity was recovered after alkaline hydrolysis. Lipid extracts were desalted by reverse-phase chromatography using an RP-18 column (29); 96% of the 3 H radioactivity was recovered after the RP-18 column chromatography step. 3 H-GSLs were separated by TLC using CHCl 3 /CH 2 OH/9.8 mM CaCl 2 (60:35:8, v/v/v) as the developing solvent; 150,000 cpm of 3 H-GSLs were loaded on each lane of the TLC plate. Upon metabolism of [4, H]dihydroceramide to [ 3 H]ceramide, we assume that 50% of the 3 H radioactivity is lost due to dehydrogenation of the 4,5-double bond; this was taken into account when quantifying 3 H-GSL synthesis. In some cases, 3 H-GSLs were visualized by autoradiography using EN 3 HANCE TM spray (Du Pont NEN). 3 H-GSLs were identified using authentic ceramide, glucosylceramide (GlcCer), and lactosylceramide (LacCer) (all from Sigma) and gangliosides G M3 , G M2 , G M1 , G D3 , G D1a , G D1b , G T1b (Matreya, Inc., Pleasant Gap, PA). Confirmation of the identity of 3 H-GSLs was obtained by comparison with [ 14 C]galactose-and [ 14 C]serine-labeled gangliosides extracted from mouse cerebellar neurons (30).
In most experiments, 3 H radioactivity was quantified using a Fuji BAS 1000 phosphoimager. TLC plates were dried and exposed to a tritium-sensitive imaging plate (BAS-TR2040S, Fuji Photo Film Co., Ltd., Tokyo, Japan). After overnight exposure, phospholuminescence (PSL) was quantified for each 3 H-labeled lipid spot using MacBAS Version 2.0 software and a Macintosh Quadra 840 computer. To convert PSL units to cpm, a calibration curve was generated by applying increasing amounts of [4, H]dihydrosphingosine to a TLC plate. Analysis of the ratio between PSL units and cpm was linear for all values in the range 10 2 to 6 ϫ 10 6 cpm. Data were fitted to a straight line, the slope of which is cpm ϭ (18.36/(Ϫ0.0062 ϩ (0.0012 ϫ min))) ϫ (PSL Ϫ blank), where min is the time of exposure to the imaging plate, and [4, H]dihydroceramide by dihydroceramide synthase. Simple GSLs (GlcCer, LacCer, G M3 , and G D3 ) are shown above the dashed line, and complex GSLs below. a-and b-series gangliosides are also defined as shown, based upon van Echten and Sandhoff (6). The enzymes of GSL synthesis are indicated as follows: SAT-I, sialyltransferase I; SAT-II, sialyltransferase II; GalNAcT, Gal-NAc-transferase; GalT-II, galactosyltransferase II; SAT-IV, sialyltransferase IV; SAT-V, sialyltransferase V. Note that GalNAc-transferase, galactosyltransferase II, and sialyltransferase IV are able to glycosylate substrates from both the a-and b-series gangliosides (39).

FIG. 1. Proposed pathway of utilization of [4,5-3 H]dihydrosphingosine by cultured hippocampal neurons. [4,5-3 H]Dihydrosphingosine is metabolized to
blank is a background PSL value obtained from an area of identical size on an unrelated region of the TLC plate.

Characterization of Neuronal
Cultures-Hippocampal neurons cultured at low density have been well characterized (19,20). However, due to the paucity of cellular material, low density cultures are not suitable for biochemical analysis. We have now characterized the development of hippocampal neurons cultured by a similar method but at sufficiently high densities to provide enough material for biochemical analysis of 3 H-GSL synthesis.
Within a few hours after plating, neurons extend lamellipodia, followed by the formation of minor processes ( Fig. 2A). Within the next 24 -48 h, one of the processes starts to grow rapidly (Fig. 2B), and within 2-3 days, a dense network of neuronal processes has formed, rendering impossible the identification of the processes as either axons or dendrites by light microscopic criteria (Fig. 2, C and D). However, by 5 (not shown) and 8 days in culture (Fig. 2, E and F), dendrites can be distinguished using an anti-MAP-2 antibody; most of the processes are not labeled using this antibody, indicating that they are axons. Likewise, after 15 days, MAP-2-positive dendrites can be distinguished from the dense axonal network (Fig. 2, G and H), and by 15 days, synaptic connections have been formed, as indicated by the localization of synaptophysin (23) to presynaptic boutons along the axon (Fig. 2, I and J). Axons were also identified using an anti-GAP-43 antibody (22).
The number of cells remaining on the coverslips at various days of culture was ascertained by counting six separate fields on two individual coverslips/day; analyses were combined from two separate neuronal cultures. On day 0, there were ϳ110,000 cells/coverslip; day 1, 92,000 cells; day 5, 80,000 cells; day 8, 55,000 cells; and day 15, 45,000 cells.
The number of glial cells was estimated by counting the number of glial fibrillary-associated protein-positive cells. The percent of glial cells, compared with neurons, was 3.6 Ϯ 2.5, 8.2 Ϯ 1.0, and 7.0 Ϯ 1.0% (mean Ϯ S.E.; 40 microscopic fields were analyzed per coverslip/day for two individual cultures) on days 6, 8, and 15, respectively. The extremely low levels of glial contamination, and the sequence of developmental events identical to that observed in low density cultures, render high density cultures of hippocampal neurons an attractive and unique system for analyzing GSL synthesis during the development of neuronal polarity. However, the amount of protein/ coverslip could not be reliably estimated (analyzed as described in Ref. 31). On days 0 -5, protein levels were usually below 1 g/coverslip, whereas on days 8 and 15, ϳ3-5 g of protein was detected per coverslip.
During the first hour, the rate of uptake of [4, H]dihydrosphingosine was 1452 and 3210 cpm/h/10 4 cells for 2-and 8-day-old neurons, respectively (Fig. 3A). The rapid initial rate of uptake of [4,5-3 H]dihydrosphingosine into neurons is similar to those observed in rat microsomal membranes (5) and results from the rapid transfer and partitioning of [4, H]dihydrosphingosine into membranes. The 2.21-fold difference in the amount of uptake between 2-and 8-day-old neurons is presumably due to the greater membrane surface area/cell in older neurons (see Fig. 2, C and E), which results from rapid axon and dendrite growth.
A lag period of between 30 min and 1 h was observed before significant amounts of 3 H radioactivity were recovered in 3 H-GSLs. However, for incubations of between 1 and 9 h, the rate of synthesis of total 3 H-GSLs was linear versus time (Fig. 3B) for both 2-and 8-day-old neurons, with rates of synthesis of 182 and 452 fmol/h/10 4 cells, respectively. The similarity between the ratios of the initial rates of [4,5-3 H]dihydrosphingosine uptake (2.21) and the ratio of the rates of 3 H-GSL synthesis (2.48) on days 2 and 8 suggests that the rate of 3 H-GSL synthesis is proportional to the initial rate of [4,5-3 H]dihy-drosphingosine uptake. This is supported by observations that the rate of 3 H-GSL synthesis remains linear even after longer periods of incubation (Fig. 3B), despite a significant decrease in the rate of accumulation of 3 H radioactivity in neurons after 1 h (Fig. 3A).
Comparison of initial rates of synthesis (0 -1 h) of individual 3 H-GSLs with later rates of synthesis (1-6 h) demonstrated that there is a significant lag period in the synthesis of complex 3 H-GSLs, but not of simple 3 H-GSLs (Table I). For instance, the initial rate of synthesis of ganglioside [ 3 H]G T1b was 20.6 fmol/ h/10 4 cells in day 2 neurons, but increased to 64.9 fmol/h/10 4 cells during the next 5 h. In contrast, the rate of synthesis of both [ 3 H]GlcCer and [ 3 H]G M3 decreased as the time of incubation increased (Table I). Moreover, the ratio of the rates of initial versus late synthesis of individual 3 H-GSLs increased in a fashion consistent with the pathway of ganglioside synthesis shown in Fig. 1 Table I). Consistent with data obtained by analysis of the rates of total 3 H-GSL synthesis on days 2 and 8 of culture (Fig. 3B), the synthesis of 3 H-GSLs increased by 2.58-fold between days 2 and 8 during a 6-h incubation (Table II). Between days 0 and 15, 3 H-GSL synthesis increased by 12.5-fold. A particularly noticeable increase in total 3 H-GSL synthesis occurred between days 0 and 1 of culture. During this period, the surface area of most neurons increased significantly due to the rapid growth of one neuronal process and of a number of short processes (Fig. 2, A and B).
The synthesis of all individual 3 H-GSLs increased versus days of culture (Table II), but the amount of increase varied for individual 3 H-GSLs (Table III) Fig. 3), lipids were extracted, and the rates of synthesis of individual 3 H-GSLs were calculated by linear regression analysis. "Initial rates" were calculated by analyzing the levels of synthesis during the first hour of incubation with [4, H]dihydrosphingosine, and "late rates" during 1-6 h of incubation. The ratio of the rates of initial and late synthesis was calculated; a value Ͼ1 indicates that the rate of synthesis increases between 1 and 6 h compared with 0 and 1 h, demonstrating a lag period in the synthesis of the particular 3 H-GSL, and a value Ͻ 1 and indicates that the rate of synthesis of a particular 3 H-GSL decreases as the time of incubation increases.  synthesis increased by 30-fold (Table II). Analysis of the distribution of 3 H radioactivity between individual 3 H-GSLs ( Fig.  4 and Table III) (Tables II and III); together with [ 3 H]Lac-Cer and [ 3 H]G M3 , these simple GSLs compose 60% of the total 3 H-GSLs synthesized on day 0 (Fig. 5). However, on day 1 of culture, by which time one of the minor processes has begun to grow rapidly (stage 3; see Fig. 2B), the simple GSLs compose Each lane was loaded with 150,000 cpm of the extracted 3 H-GSLs, which were visualized using EN 3 HANCE spray. Note that lipids extracted from 0 -8-day-old neurons were separated on a different TLC plate than 15-day-old neurons, but the same amount of 3 H radioactivity was loaded onto each plate, and the same exposure times to x-ray film were used.  only 33% of the total 3 H-GSLs synthesized (Fig. 5), with significantly higher levels of synthesis of complex gangliosides of both the a-and b-series (Table III).
(ii) The ratio of simple to complex gangliosides does not change significantly after day 1 of culture, but the ratio of a-to b-series gangliosides increases rapidly during the first 4 days. After formation of the process that is destined to become the axon (stage 3), a period of rapid axon growth continues for the next 3-4 days (Fig. 2, C and D), during which time the minor processes grow only very slowly (19). During this period, the ratio of simple to complex gangliosides does not change (Fig. 5). However, there is significant change in the ratio of complex ato b-series gangliosides (Fig. 6). On day 1 of culture, a-series gangliosides compose 21% of the total 3 H-GSLs synthesized, but by day 5, they compose 29% (Table III). Since there is a corresponding decrease in the synthesis of b-series gangliosides (Table III), a significant increase in the ratio of a-to b-series gangliosides is obtained (Fig. 6).
(iii) There is a small increase in the ratio of a-to b-series gangliosides from days 5 to 14, by which time the major gangliosides synthesized are G D1a , G T1b , G D1b , and G M1 . Between days 5 and 15, minor processes acquire the characteristics of dendrites (stage 4; see Fig. 2, E-H), and synaptogenesis occurs (stage 5; see Fig. 2, I and J). Despite these major morphological and functional changes, there is little alteration in the pattern of ganglioside synthesis during this period. The ratio of a-to b-series gangliosides continues to increase, but at a much slower rate than that observed between days 1 and 5 (Fig. 6). On day 8, gangliosides G M1 , G D1a , G D1b , and G T1b compose 59% of the total 3 H-GSLs synthesized, but by day 15, these four gangliosides compose 71% (Table III), and simple gangliosides compose only 23% of the total 3 H-GSLs synthesized (Fig. 5). 3 H-GSL Synthesis in Glia-To determine whether the small number of glial cells found on neuronal coverslips (between 3 and 8%; see above) contributed significantly to 3 H-GSL synthesis, glia were grown to near confluence and incubated with [4,

DISCUSSION
In this study, we have analyzed ganglioside synthesis during the development of neuronal polarity. Ganglioside synthesis was studied after relatively short times of incubation (up to 6 h), unlike many previous studies, which either analyzed total GSL content or analyzed GSL synthesis after long times of incubation (24 -48 h), in which the rate of synthesis could not be accurately determined due to turnover and degradation of labeled GSLs.
Ganglioside Synthesis during Axonogenesis and Rapid Axon Growth-The most dramatic changes in ganglioside synthesis are the increase in complex ganglioside synthesis on day 1 (stages 2 and 3; see Ref. 19) compared with day 0 (stage 1) and the increase in the ratio of a-to b-series gangliosides between days 1 and 5 (stage 3). A decrease in G D3 content and an increase in the ratio of a-to b-series gangliosides were observed in extracts of rat brain (13) between embryonic days 18 and 20, but none of these changes could be ascribed definitively to neurons or to specific stages of neuronal development since whole brain tissue (which contains many types of neurons, all at different developmental stages) was analyzed. In hippocampal neurons, changes in ganglioside synthesis occur during axonogenesis and during the subsequent period of rapid axon growth. However, ganglioside synthesis does not appear to be a prerequisite for these events since inhibition of synthesis by either fumonisin B 1 (34) or D-threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol (35) has no effect on the formation of the initial axon between days 0 and 2 of culture (see Fig. 1 in Ref. 9). This suggests that the increase in complex ganglioside synthesis is a result of (rather than the cause of) axonogenesis. In contrast, inhibition of GSL synthesis does affect the formation of collateral axonal branches between days 2 and 3 of culture (9), during which time there is a shift toward the synthesis of a-series gangliosides.
Ganglioside Synthesis during Dendrite Growth-Surprisingly, between days 3 and 8 of culture, there is little alteration in the pattern of ganglioside synthesis, even though morphologically distinct dendrites form during this time. Although a minor ganglioside in hippocampal neurons, the synthesis of G M2 appears somewhat elevated during this period (Table II), but the low levels of synthesis make reliable and accurate quantification difficult (see Fig. 4). If G M2 is indeed elevated,  Table III. Simple (f) and complex (Ⅺ) GSLs are defined as in Fig. 1.   FIG. 6. Ratio of a-to b- this would lend support to the idea, based on the accumulation of G M2 during ectopic dendrite formation in lysosomal storage diseases (36), that G M2 plays an important role in dendritogenesis. In cerebellar Purkinje cells, evidence has been provided, based on immunolocalization, that a minor ganglioside, G D1␣ , is specifically localized to proximal dendrites and cell bodies (37). Since we did not analyze the synthesis of this ganglioside in hippocampal neurons, we could not ascertain whether the synthesis of this or any other ganglioside that may be localized to dendrites is elevated during dendrite growth.
Ganglioside Synthesis during Synaptogenesis-Our recent observation that long-term inhibition of GSL synthesis does not effect the polarized segregation of axonal and dendritic markers in mature neurons (see Fig. 8 in Ref. 9) suggests that there is no direct correlation between the development of neuronal polarity and the synthesis of any major ganglioside. There is nevertheless a significant difference between the profile of gangliosides synthesized in immature neurons (i.e. days 0 -3) and that observed in mature neurons with functional synapses (day 15). Whether gangliosides G M1 , G D1a , G D1b , and G T1b play a particular role in synaptic function remains to be established.
In summary, we have shown that major changes in ganglioside synthesis occur during axonogenesis and axon elongation, but not during dendritogenesis and dendrite growth or synaptogenesis. Although our study does not address the function of gangliosides in these events, it does demonstrate that the rapid growth of axons results in an increase in complex and a-series gangliosides, whereas dendritogenesis and dendrite growth have little, if any, effect on ganglioside synthesis.