Alterations in neuroblastoma ganglioside synthesis by induction of GD1b synthase by retinoic acid

Recent findings link increased expression of the structurally complex ‘b’ pathway gangliosides GD1b, GT1b, GQ1b (CbG) to a favourable clinical and biological behaviour in human neuroblastoma (NB). Seeking a model to probe these observations, we evaluated four human NB cell lines. Very low CbG content (4–10%) in three of the four cell lines (LAN-5, LAN-1, SMS-KCNR) reflected the ganglioside pattern observed in the most aggressive NB tumours. Pharmacological alterations of complex ganglioside synthesis in vitro by a 5–7 day exposure to 5–10 μM retinoic acid, which is employed in maintenance therapy of disseminated NB, included markedly increased (i) relative expression of CbG (6.6±2.0-fold increase, P=0.037), (ii) relative expression of the analogous ‘a’ pathway gangliosides, termed CaG (6.4±1.4-fold increase in GM1a and GD1a; P=0.010), and (iii) total cellular ganglioside content (2.0–6.3-fold), which in turn amplified the accumulation of structurally complex gangliosides. Substantial increases (2.7–2.9-fold) in the activity of GD1b/GM1a synthase (β-1,3-galactosyltransferase), which initiates the synthesis of CbG and CaG, accompanied the all-trans retinoic acid (ATRA)-induced ganglioside changes. Thus, increased CbG synthesis in NB cell lines is attributable to a specific effect of ATRA, namely induction of GD1b/GM1a synthase activity. Since the shift towards higher expression of CbG and CaG during retinoic acid-induced cellular differentiation reflects a ganglioside pattern found in clinically less-aggressive tumours, our studies suggest that complex gangliosides may play a role in the biological and clinical behaviour of NB.

Increasing evidence has implicated gangliosides, a specific class of cell surface glycosphingolipids, in the biological and clinical behaviour of many types of tumours including human neuroblastoma (NB). Gangliosides are overexpressed and actively shed by tumour cells (Wu et al, 1986;Valentino et al, 1990;Li and Ladisch, 1991) and have a number of biological properties that could conceivably alter tumour-host interactions to influence the survival of the malignant cells that carry these molecules (Hakomori, 1996). Our previous studies have linked specific ganglioside changes in human NB tumours to differences in the clinical and biological behaviour of this tumour (Kaucic et al, 2001;Hettmer et al, 2003), and raise the possibility that pharmacologically induced modulation of NB ganglioside content could have important consequences for outcome.
Gangliosides consist of a sialic acid-containing carbohydrate portion and a lipid portion (ceramide) embedded in the outer leaflet of the cell membrane. They are synthesised via two prominent pathways, designated 'a' (GM2, GM1a, GD1a) and 'b' (GD3, GD2, GD1b GT1b, GQ1b), from a common precursor (GM3) derived from lactosylceramide ( Figure 1). Each ganglioside is structurally more complex than its precursor molecule, and the stepwise addition of monosaccharide or sialic acid residues by specific membrane-bound glycosyltransferases in the Golgi apparatus is catalysed by the same glycosyltransferases in both pathways (van Echten and Sandhoff, 1993).
Our previous studies have characterised ganglioside expression in human NB, a paediatric malignancy of neural crest origin (Wu et al, 1986;Valentino et al, 1990;Kaucic et al, 2001;Hettmer et al, 2003). We have linked alterations in the expression of complex 'b' pathway gangliosides (CbG) downstream of GD1b/GM1a synthase (GD1b, GT1b and GQ1b) to differences in the biological phenotype and clinical behaviour that can be used to predict patient outcome in NB (Hettmer et al, 2003). Further, recent experimental evidence suggests that GD1b, GT1b and GQ1b can modulate a number of biological processes that are believed to counteract malignant transformation and progression, including tumour cell proliferation, host immune function and signal transduction mechanisms (Hynds et al, 1995;Fukumoto et al, 2000;. We hypothesise that high CbG content may contribute to reduced tumour aggressiveness, and that alteration of CbG expression might be a potential therapeutic target in NB. One pharmacological agent that is well known to alter cellular ganglioside metabolism and successfully used in the oral maintenance therapy of disseminated NB is retinoic acid (Matthay et al, 1999). Previous in vitro studies have shown that morphological signs of neuronal differentiation in response to retinoic acid are accompanied by an increase in total ganglioside content and a relative increase in the expression of certain complex gangliosides of both the 'a' and 'b' pathway in NB cells (Li and Ladisch, 1992;Rebhan et al, 1994), as well as in normal and transformed embryonal cells (Levine and Flynn, 1986;Rizzo et al, 1995;Liour et al, 2000). On the basis of these findings, we have undertaken a comprehensive analysis of retinoic acid-induced changes in ganglioside expression and metabolism in human NB cell lines.
Here, we demonstrate that low CbG levels in some NB cell lines are analogous to the ganglioside pattern observed in clinically and biologically unfavourable NB tumours, providing an in vitro model to study pharmacological agents that might alter the metabolism of complex ganglioside subspecies. We further demonstrate that treatment with retinoic acid markedly enhances the activity of GD1b/GM1a synthase, resulting in increased expression of the complex gangliosides downstream of this enzyme, namely GD1b and GT1b (CbG), and GM1a and GD1a (CaG). Our results thus identify a specific effect of retinoic acid on the expression and biosynthesis of complex gangliosides in NB cells.  Figure 1 Schematic representation of the major pathways of ganglioside biosynthesis. (A) The monosialoganglioside GM3, derived from lactosylceramide, is the common precursor for both 'a' and 'b' pathway gangliosides. Each ganglioside species consists of a ceramide backbone (CER), and a carbohydrate chain (glc ¼ glucose, gal ¼ galactose, GalNAc ¼ N-acetylgalactosamine) containing one or more sialic acid (SA) residues. 'a' and 'b' pathway gangliosides downstream of GD1b/GM1a synthase were designated complex 'a' (CaG) and complex 'b' (CbG) gangliosides, respectively. (B) Parallel steps in both pathways are catalysed by the same glycosyltransferases of the Golgi apparatus (Li and Ladisch, 1997): (1), GD3 synthase (a-2,8-sialyltransferase): (2), GM2/ GD2 synthase (b-1,4-N-acetylgalactosaminyltransferase); (3), GD1b/GM1a synthase (b-1,3-galactosyltransferase); (4), GT1b/GD1a synthase (a-2,3sialyltransferase); (5), GQ1b/GT1a synthase (a-2,8-sialyltransferase). For each enzyme, accession numbers according to the NCBI protein bank are included.

Cell culture
Human NB cell lines LAN-1, LAN-5, SMS-KCN and SMS-KCNR were derived from three children with disseminated NB and have been previously described. SMS-KCN and SMS-KCNR were established from the same patient, from the primary tumour and bone marrow at relapse, respectively (Kohl et al, 1983;Reynolds et al, 1986;Rettig et al, 1987).
Retinoic acid (all-trans or 13-cis; Sigma Chemical Co., St Louis, MO, USA) was dissolved in ethanol to a concentration of 10 À2 M and kept as stock solution for up to 4 weeks at À701C, protected from light. For each experiment, stock solutions were diluted with the growth medium to a final concentration of 5 or 10 mM retinoic acid. These concentrations are equivalent to the serum levels effective in the oral maintenance therapy of disseminated NB with 13-cis retinoic acid (13-cis RA) (7.275.3 mM) (Reynolds and Lemons, 2001). The final concentration of ethanol in the culture medium was p0.2% (v v À1 ).
Cells were seeded at 7 Â 10 4 cells cm À2 . After 24 h, the medium was replaced with a medium containing retinoic acid. Culture medium was replaced with fresh medium containing retinoic acid every other day. Untreated control cells were grown with and without ethanol. Cells were harvested after 5 -7 days of exposure to retinoic acid, followed by ganglioside purification or extraction of crude membrane extracts, as described below. Viable cells were counted using trypan blue dye exclusion and the cellular protein content was estimated by the Bio-Rad D C Protein Assay (Bio-Rad Laboratories, Hercules, CA, USA), using a bovine serum albumin standard.

Ganglioside purification and quantification
Methods for extraction and purification of gangliosides were previously described (Ladisch and Gillard, 1985). Briefly, total lipid extracts of cell pellets were obtained by extracting the lyophilised cell pellets twice with chloroform/methanol (1 : 1 v v À1 ) at 41C with stirring. Gangliosides were isolated by partitioning the dried total lipid extract in diisopropyl ether/1-butanol/17 mM aquaeous NaCl (6 : 4 : 5 v v À1 ). Gangliosides, in the lower aqueous phase, were further purified by Sephadex G-50 gel filtration.
High-performance thin layer chromatography (HPTLC) analysis of gangliosides (Ledeen and Yu, 1982) was performed using 10 Â 10-or 10 Â 20 cm precoated silica gel HPTLC plates (Merck, Darmstadt, Germany). The plates were developed in chloroform/ methanol/0.25% CaCl 2 Á H 2 O (60 : 40 : 9 by volume), and the gangliosides were stained with resorcinol. Absolute cellular ganglioside content was determined by HPTLC densitometry using known concentrations of human brain gangliosides (HBG) as the standard. Radioactive products were visualised by exposure of HPTLC plates to an X-ray film (BioMax, Kodak).
Individual gangliosides were identified using a human brain ganglioside standard or 14 C-labelled rat brain gangliosides. The area under the peak generated by each ganglioside band was integrated after the plates/films were scanned (Microtek Scan-Maker 5), and the relative percentage was calculated. To categorise individual ganglioside species, we used a previously established classification scheme based on biosynthetic pathways (Hettmer et al, 2003). Structurally complex molecules downstream of GD1b/ GM1a synthase (b-1,3-galactosyltransferase) were designated complex 'a' gangliosides (CaG: GM1a, GD1a and GT1a) and complex 'b' gangliosides (CbG: GD1b, GT1b and GQ1b), respectively.
Preparation of crude membrane extracts Crude membrane fractions, used as the source of enzymes, were prepared by suspending cell pellets in five volumes of cold 0.25 M sucrose containing 0.01 M 2-mercaptoethanol. Cells were disrupted using eight cycles of freezing in liquid nitrogen and thawing in a water bath at 371C. After centrifugation at 2000 g for 10 min to remove cellular debris, the supernatant fractions were centrifuged at 10 000 g for 10 min and then sedimented at 100 000 g for 1 h (Kemp and Stoolmiller, 1976). The sedimentable fractions were resuspended in unbuffered sucrose (0.25 M) and the amount of membrane protein was determined using the Bio-Rad D C Protein Assay (Bio-Rad Laboratories, Hercules, CA, USA).
Processing of glycosyltransferse assays Triton X-100 (50 mg, Sigma Chemical Co., St Louis, MO, USA), ganglioside precursors and radiolabelled sugar donors in chloroform/methanol (1 : 1 v v À1 ) were transferred to 1 ml conical glass tubes and the solvent was evaporated under a stream of nitrogen. The corresponding buffer mixture and unlabelled sugar donor were added, the mixture was shaken vigorously, and each sample was sonicated for 30 s. Reaction mixtures were completed by addition of 50 -150 mg membrane protein to a final reactive volume of 50 ml and shaken gently. After incubation at 371C for 1 h, reactions were terminated by addition of 1 ml chloroform/methanol (2 : 1 v v À1 ) (Kemp and Stoolmiller, 1976;Pohlentz et al, 2000).
For each glycosyltransferase assay, buffer mixtures containing MES (Calbiochem, San Diego, CA, USA), KCl, MnCl 2 and MgCl 2 (Sigma Chemical Co., St Louis, MO, USA) were freshly prepared. Radiolabelled and 'cold' sugar donors were obtained from Perkin -Elmer Life Sciences Inc. (Boston, MA, USA) and Calbiochem (San Diego, CA, USA), respectively. All enzyme assays were performed in duplicates with the appropriate blank.
Separation of lipid-soluble reaction products Lipid-soluble compounds in the reaction mixture were isolated by passing them through columns of Sephadex G25 Superfine (1 Â 3 -5 cm in 150 mm Pasteur pipettes, stuffed with glass wool). Sephadex G25 Superfine (Sigma Chemical Co., St Louis, MO, USA) was suspended in water (5 g per 60 ml), stored overnight at 41C and shaken vigorously. Before pouring the columns, the gel was brought to room temperature and washed twice with the original volume of methanol (X30 min each time). One-third of the supernatant was decanted, the mixture was shaken vigorously to obtain a homogenous solution and the Pasteur pipettes were filled to the top. After the solvent had drained, the gel was overlayed with 1.5 ml methanol and equilibrated with chloroform/methanol/ water (120 : 60 : 9 v v À1 ). The stopped samples were applied to the columns, the sample tubes washed with 1 ml chloroform/methanol/water (120 : 60 : 9 v v À1 ) and elution of the products was completed by adding 2 ml of the latter (Pohlentz et al, 2000).
Determination of glycosyltransferase activities and product identification To measure the rate of incorporation of radiolabelled sugars, the column effluents were transferred in part (40%) to glass scintillation vials, evaporated to dryness under a stream of nitrogen, and dissolved in 10 ml Betafluor (National Diagnostics, Mannville, NJ, USA). Radioactivity (measured as d.p.m.) was determined in a Beckman LS6500 liquid scintillation counter. Quench correction was accomplished using the AutoDPM program. Enzyme activity (pmol h À1 mg protein À1 ) was calculated from the ratio of the products and the radioactivity of the total sample. The remainder of the sample was used to confirm the identity and the purity of the product by TLC autoradiography.

Statistical analysis
The relative expression of individual gangliosides and ganglioside subsets (CbG and CaG) in each of the four cell lines studied were expressed as the mean7s.e.m. for three to five separate experiments. For the three cell lines that were responsive to retinoic acid, the relative ganglioside content was similarly reported as the mean7s.e.m. and, alternatively, the per cent of the expression of parallel control cells for two separate experiments per cell line. Data from six separate experiments involving three different cell lines were normalised by relating results in each experimental group to their respective controls by expressing the experimental (exp) levels as a per cent of the control (ctrl) levels (exp/ctrl Â 100). Control values of zero were set to 1 before the computation. To test the hypothesis that experimental values differed from controls, we employed a t-test to examine whether the experimental values differed from 100%. The same method was used to compare glycosyltransferase activities among retinoic acid-treated and control cells in retinoic acid-responsive cell lines (LAN-1, LAN-5, SMS-KCNR). Results were considered statistically significant if the P-value was less than 0.050.
To examine whether experimental values differed from controls in studies of 13-cis RA, mean treatment values were compared to control values (no treatment with 13-cis RA) using a t-test.

Ganglioside expression in human NB cell lines
The impetus for this study were our findings (Hettmer et al, 2003) and those of others (Schengrund et al, 1985;Sung et al, 1995;Yates et al, 1999), suggesting that differences in tumour tissue CbG (GD1b, GT1b and GQ1b) expression are linked to the clinical and biological behaviour of NB and other tumours. In order to obtain an in vitro model that permits a more dynamic view of NB ganglioside biosynthesis, we analysed ganglioside expression in four well-described NB cell lines, LAN-1, LAN-5, SMS-KCNR and SMS-KCN.
The prominent position of GD2 as a major ganglioside component in the NB cell lines studied (44% in LAN-5, 36% in LAN-1, 60% in SMS-KCNR and 26% in SMS-KCN) mirrors the  ubiquitously high expression of GD2 that characterizes NB tumours (Wu et al, 1986). Importantly, the low CbG expression observed in each of the cell lines is also analogous to the CbG levels found in clinically and biologically unfavourable NB tumours (Hettmer et al, 2003). We therefore used these four NB cell lines as an in vitro model to determine whether retinoic acid, as a pharmacological agent, might alter the expression and biosynthesis of CbG.

Retinoic-acid induced changes in CbG expression in NB cell lines
Retinoids are a class of vitamin A derivatives that differ in their isomeric conformations, and include all-trans retinoic acid (ATRA), 9-cis retinoic acid and 13-cis RA. Owing to its superior pharmacokinetics (higher and more sustained plasma levels), 13cis RA is the isomer that is primarily used clinically. However, 13cis RA isomerises in vivo to both ATRA and 9-cis retinoic acid, which are believed to be the active compounds in vivo in nonocular tissue (Lovat et al, 1994). Since retinoic acid treatment has been associated with changes in the ganglioside complement of various cell types (Li and Ladisch, 1992;Rebhan et al, 1994;Liour et al, 2000), and the oral maintenance therapy of disseminated NB with 13-cis RA results in improved event-free survival (Matthay et al, 1999), we sought to specifically explore the effect of retinoic acid on CbG expression. We, therefore, first treated NB cells in vitro with 10 mM ATRA, then evaluated ATRA-induced changes in the composition of individual gangliosides and in total cellular ganglioside content in LAN-1, LAN-5, SMS-KCNR and SMS-KCN cells (Table 2). Strikingly, ATRA induced a shift from synthesis of simpler gangliosides towards more complex species within both the 'a' and 'b' pathways in three of the four cell lines (LAN-1, LAN-5 and SMS-KCNR). In these three cell lines, the redistribution towards complex gangliosides included increased relative expression of GD1b and GT1b ('b' pathway) and GM1a and GD1a ('a' pathway), compounded by a substantial increase in total cellular ganglioside content, making the increase in the absolute content of CbG and CaG even more striking (Table 2). In LAN-5 cells for example, the ATRA-induced changes in cellular ganglioside expression included a 3.7-fold increase in CbG content (from 6 to 22%), a 2.2-fold increase in CaG content (from 4 to 9%) and a two-fold increase in absolute ganglioside content (from 57 to 112 nmol 10 8 cells À1 ). SMS-KCN, the cell line with the highest baseline level of CbG, was the only cell line of the four which was not responsive to ATRA treatment; specifically, the change in complex ganglioside content was minimal (23 vs 23% CbG and 24 vs 18% CaG in controls compared to ATRAtreated cells), as was the change in total cellular ganglioside  ). When normalised data from the three responsive cell lines were analysed together (Table 3), ATRA treatment clearly resulted in significant increases in the cellular content of CbG (6587202% of control, P ¼ 0.037) and CaG (6407137% of control, P ¼ 0.010), and, concomitantly, significant decreases in the cellular content of GD2 (69711% of control; P ¼ 0.040) and GD3 (35712% of control; P ¼ 0.002).
To illustrate these changes, Figure 3 summarises the ATRAinduced changes in cell morphology and ganglioside expression of the NB cell line SMS-KCNR. The morphological changes in response to ATRA, including rounding up of the cell body, extension of long processes with the appearance of neurites and cell aggregation into tight clusters ( Figure 3A), are consistent with previous studies evaluating the differentiating effects of ATRA on NB cells (Li and Ladisch, 1992;Melino et al, 1997). Qualitative changes in the composition of cell-surface gangliosides accompanied morphological signs of differentiation in response to ATRA. The most prominent ganglioside components in untreated SMS-KCNR cells were the 'b' pathway ganglioside GD2 and the 'a' pathway ganglioside GM2. Treatment with ATRA increased the proportions of the CbG (GD1b and GT1b), as well as the CaG (GD1a and GM1a), such that these two subgroups predominated relative to other ganglioside species ( Figure 3B). In view of the increase in total cellular ganglioside content (Table 2), this relative change resulted in significant absolute changes in the membrane expression of complex ganglioside molecules. Thus, the mean four-fold increase in relative CbG expression translated into a 25-fold increase in absolute cellular CbG expression in SMS-KCNR cells, indicating a striking accumulation of these biologically active molecules.
Since 13-cis RA is currently the only retinoid compound in clinical use to treat NB (Reynolds and Lemons, 2001), we also exposed one NB cell line, LAN-1, to 13-cis RA at two different concentrations (5 mM, 10 mM). We observed that 13-cis RA caused a concentration-related increase in the expression of structurally complex gangliosides in LAN-1 cells ( Figure 4). As with ATRAtreated cells, the increase in CbG expression was paralleled by a concurrent increase in CaG, suggesting a specific effect of retinoids on the activity of GD1b/GM1a synthase (b-1,3-galactosyltransferase), which initiates the synthesis of both CbG and CaG, from GD2 and GM2, respectively.

Effect of retinoic acid on the activity of ganglioside glycosyltransferases in human NB cells in vitro
To delineate the metabolic basis for the ATRA-induced shift towards expression of complex gangliosides, we determined the effect of ATRA treatment on the activity of three key enzymes in the synthesis of these molecules. Precursor molecules enter 'b' pathway ganglioside biosynthesis through conversion of GM3 into GD3 by GD3 synthase (a-2,8-sialyltransferase). GD2/GM2 synthase (b-1,4-N-acetylgalactosaminyltransferase) and GD1b/GM1a synthase (b-1,3-glactosyltransferase) catalyse the two subsequent enzymatic reactions resulting in the synthesis of the most complex gangliosides in each pathway (van Echten and Sandhoff, 1993) ( Figure 1). We measured the activities of GD3 synthase, GD2 synthase and GD1b synthase in crude membrane extracts from LAN-5, SMS-KCNR, SMS-KCN, using excess GM3, GD3 and GD2 as exogenously added ganglioside precursors.
When the retinoic acid-responsive LAN-5 or SMS-KCNR cells were exposed to 10 mM ATRA for 6 -7 days (Table 4), we observed a marked increase (2.7 -2.9-fold, Po0.050) in GD1b synthase activity. This was accompanied by a lesser (0 -1.8-fold, P ¼ 0.198) increase in GD2 synthase activity, and only small changes in GD3 synthase activity (1.2 -1.4, P ¼ 0.161). Thus, in both LAN-5 and SMS-KCNR cells, increased expression of complex gangliosides downstream of GD1b sythase was associated with a significant (over two-fold) increase in GD1b synthase activity. In contrast, in SMS-KCN cells, the cell line that was unresponsive to ATRA treatment, GD1b synthase activity was only slightly increased (1.5-fold) in ATRA-treated vs control cells. Figure 5 illustrates the autoradiographic visualisation of enzyme products resulting from ATRA-induced changes in ganglioside metabolism in SMS-KCNR cells, confirming the major increase in GD1b synthase activity ( Figure 5A), a comparatively minor increase in GD2 synthase activity ( Figure 5B) and almost no change in GD3 synthase activity ( Figure 5C). It is interesting to note that in the case of the ATRA-treated cell samples, the parallel blank (which does not contain exogenously added ganglioside precursors) also shows the synthesis of GD1b and GM1a from the trace amounts of endogenous gangliosides present in the cell pellet ( Figure 5A), further confirming the substantial increase in the activity of GD1b/GM1a synthase following treatment with ATRA. Taken together, our data demonstrate that the increase in GD1b synthase activity is specifically associated with increased expression of complex ganglioside species, induced by ATRA in NB cells in vitro, and provides strong evidence that elevated CbG and CaG expression can be accounted for by a specific effect of ATRA on this enzyme.

DISCUSSION
Treatment with retinoic acid is shown here to induce a dramatic shift from synthesis of simpler gangliosides towards predominant expression of structurally complex 'a' and 'b' pathway ganglioside molecules downstream of GD1b/GM1a synthase in some NB cell lines. The increase in total cellular ganglioside content combined with the relative changes in the proportion of complex gangliosides results in a significant increase in the absolute cellular content of these molecules. The findings are consistent with the previous observations of a dramatic increase in membrane expression and shedding of total cellular gangliosides and enhancement in GD1a and GT1b content in LAN-5 NB cells treated with retinoic acid (Li and Ladisch, 1992).
Among the four cell lines employed in this study, the NB cell line SMS-KCN appears to be biologically distinct. Three differences are evident. First, SMS-KCN cells consist of two morphologically distinct cell populations, one neuroblastic (N-type) and the second substrate-adherent (S-type) with epithelial or fibroblast-like morphology (Rettig et al, 1987), while LAN-1, LAN-5 and SMS-KCNR exhibit a uniform phenotype with small, round, neuroblastic cells that have short neuritic processes. Second, constitutive CbG content in SMS-KCN was higher (1.8-to 5.2-fold) than in the other cell lines studied. Third, treatment with retinoic acid did not induce the alterations in cellular ganglioside biosynthesis observed in the other cell lines. These differences in ganglioside metabolism do not have an obvious explanation, but it is possible that they are accounted for by constitutive differences between N-type and S-type cells. One possible explanation for the retinoic acid-induced changes in ganglioside composition in LAN-1, LAN-5 and SMS-KCNR cells is an increased rate of synthesis of certain gangliosides. A previous study in embryonal carcinoma cells demonstrated enhanced activities of various ganglioside glycosyltransferases at different stages of retinoic acid-induced neuronal differentiation, and a parallel increase in mRNA expression, suggesting that retinoic acid activated the transcription of these glycosyltransferases or stabilised their expression (Osanai et al, 1997;Liour et al, 2000). Our data support that the retinoic acid-induced increase in the expression of complex gangliosides both within the 'a' and the 'b' pathway is attributable to an induction of GD1b/GM1a synthase (b-1,3-glactosyltransferase) activity. The retinoic acid-induced alterations in the ganglioside complement of NB cells seem to reflect, therefore, a specific effect on ganglioside anabolism, although, theoretically, alterations in ganglioside catabolism or shedding are also possible contributing factors. Increased expression of complex gangliosides is a well-described feature of normal embryonic neurogenesis and accelerated axonogenesis (Svennerholm et al, 1989;Hirschberg et al, 1996). Similar ganglioside changes to those we describe here have been reported in retinoic acid-treated Xenopus embryo cells, murine embryonal carcinoma cells and SHSY-5Y NB cells during neuronal maturation (Rebhan et al, 1994;Rizzo et al, 1995;Liour et al, 2000). Thus, predominant expression of complex gangliosides can be considered a biochemical marker of increasing neuronal differentiation. Further, the retinoic acid-induced changes in the ganglioside, complement of NB cell lines, are analogous to reported differences in CbG content among prognostically distinct categories of NB tumours (Schengrund et al, 1985;Schengrund and Shochat, 1988;Hettmer et al, 2003). Specifically, we previously found that among 74 human NB tumours, high (X35%) CbG expression (a) characterised nonprogressive vs progressive tumours (median 41 vs 18% of total gangliosides), and (b) was strongly predictive of a favourable outcome, even when other prognostic factors were considered (Hettmer et al, 2003). Related studies, associating low CbG levels in astrocytoma and medulloblastoma with higher histological grades and lower survival (Sung et al, 1995;Yates et al, 1999), further support the value of CbG in discriminating between different categories of malignant disease. In this context, the retinoic acid-induced increase in CbG expression in NB cells represents a transition into a ganglioside pattern which is associated with (i) neuronal differentiation in normal cells and (ii) clinically less-aggressive NB tumours, suggesting that CbGs may play a role in the biological behaviour of NB.
Tumour gangliosides of aberrant structure are generally believed to favour tumour progression (Hakomori, 1996), and a substantial body of previous experimental evidence associates the shedding of tumour cell surface gangliosides with increased tumorigenicity (Ladisch et al, 1987). The relationship between the concentration of shed GD2 in the plasma of NB-bearing patients and reduced long-term survival (Valentino et al, 1990) is consistent with this concept. There is also experimental evidence that suggests that the biological effects of tumour-derived gangliosides differ greatly depending on their carbohydrate and ceramide structures (Floutsis et al, 1989;Ladisch et al, 1994). Here we hypothesise that the significantly higher CbG levels in clinically favourable NB tumours (Hettmer et al, 2003) may lead to reduced NB tumour aggressiveness. Substantial experimental evidence corroborates the concept that the prognostic value of CbG reflects their functional involvement in the cellular processes that define the malignant behaviour of NB. Malignant transformation of murine epidermal cells in response to tumour-promoting phorbol esters is accompanied by a decrease in GT1b synthesis (Srinivas and Colburn, 1982). Proliferation of and IL-8 production by human metastatic melanoma cells is inhibited by GD1b, GT1b and GQ1b . Platelet-derived growth factor (PDGF)-mediated cell growth and PDGF receptor activation in NB and glioma cells are inhibited by complex ganglioside species (Hynds et al, 1995). These gangliosides also modulate immunoglobulin synthesis in peripheral blood mononuclear cells Tamaki, 1998, 1999a, b) and cause a shift from Th-2 to Th-1 cytokine production in PHA-stimulated T cells . Finally, increased expression of GD1b and GT1b in rat pheochromocytoma cells transfected with GD3 synthase occurs in parallel with trk-A dimerisation (Fukumoto et al, 2000), which in turn is associated with improved prognosis in NB (Nakagawara et al, 1993). Taking these findings together, we hypothesise that CbG may impede tumour progression, either by acting directly on the tumour cell or by increasing the host resistance.
Similarly, increased tumour CbG expression might contribute to the success of retinoic acid in the maintenance therapy of disseminated NB. The beneficial role of retinoic acid is generally attributed to a reversal of the malignant state through differentiation of the tumour cells into more mature ganglion-like cells (Matthay et al, 1999;Reynolds and Lemons, 2001). Substantial experimental evidence suggests that complex gangliosides are biologically relevant molecules during cellular differentiation and development. First, outgrowth of cellular extensions, one aspect of neuronal differentiation, can be induced by treating NB cells in vitro with exogenous gangliosides, and the CbG species GT1b and GQ1b have been repeatedly described as the two most potent ganglioside molecules with differentiating properties (Tsuji et al, 1983;Leskawa and Hogan, 1985). Secondly, increased expression of GD1b and GT1b in murine NB cells transfected with GD3 synthase occurs in parallel with morphological signs of increased cellular differentiation (Kojima et al, 1994). Thirdly, absence of all gangliosides downstream of GM2/ GD2 synthase (CbG þ GD2 and CaG þ GM2) in genetically altered mice with a disrupted gene for that enzyme is associated with decreased myelination and axon degeneration in the central and peripheral nervous system (Sheikh et al, 1999), although brain development and gross behaviour is normal (Takamiya et al, 1996). Finally, although retinoic acid-induced neurite outgrowth is unaffected by abrogation of cellular ganglioside synthesis in LAN-5 cells grown in a serum-supplemented environment (Li and Ladisch, 1997), reduction of cellular ganglioside synthesis is associated with decreased NGF-induced outgrowth of neuritic processes in a NB cell line grown in serum-free medium, and is reversible by exogenous addition of gangliosides (Rosner, 1998).
Previous findings and our demonstration of induction of CbG (and CaG) synthesis by retinoic acid underscore the potential role of complex gangliosides in cellular differentiation and other cellular functions important to reversion of the malignant phenotype. Consequently, the specific role of complex gangliosides compared to structurally simpler molecules in influencing cellular differentiation and other cellular functions in NB and other neuroectodermal tumours will be important to delineate.