Introduction

The childhood malignancy neuroblastoma is derived from cells of the developing sympathetic nervous system. The neuroblastoma cells are arrested at immature stages of sympathetic differentiation, although the stage of tumor cell maturation varies from one tumor to another and often also within a single tumor. This immaturity is frequently manifested in phenotypic characteristics resembling those of sympathetic progenitor cells such as expression of the embryonic and fetal transcription factors dHAND and human achaete-scute homolog-1 (HASH-1) (Gestblom et al, 1997, 1999. Such progenitor cells give rise to three major closely related, sympathetic cell lineages: the neuronal/ganglionic, the small intensely fluorescent, and the chromaffin cell lineages (Patterson, 1990). The hormones and growth factors determining rat sympathetic progenitor cell lineage selection are grossly understood and involve glucocorticoids, classical growth factors, and neurotrophins (Anderson, 1993, 1997). For example, development into and along the neuronal lineage requires sequential exposure to insulin-like growth factors, basic fibroblast growth factor, and the neurotrophins nerve growth factor (NGF) and neurotrophin(NT)-3 (DiCicco-Bloom and Black, 1988; Stemple et al, 1988; Verdi and Anderson, 1994; Zackenfels et al, 1995).

In mammals, the neurotrophin family of growth factors consists of NGF, brain-derived neurotrophic factor (BDNF), NT-3, and NT-4 (also known as NT-5) (Barbacid, 1995; Barde, 1994). Their high affinity receptors belong to the trk gene family, encoding tyrosine kinase receptors. Thus, NGF binds p140trkA, BDNF, and NT-4 bind p145trkB; and NT-3 recognizes p145trkC. In addition, there is some cross-talk between the various neurotrophins and their receptors (Barbacid, 1995). Both NT-3 and NGF are involved in the late phases of sympathetic neuronal differentiation, and in agreement with this, trkA and trkC are expressed in human sympathetic neuroblasts (Hoehner et al, 1995). Furthermore, mice lacking either NGF, trkA, or NT-3 have a reduced number of sympathetic ganglion cells; the NT-3 effect is apparently being mediated via p140trkA, showing explicitly the vital roles of NGF and NT-3 for development and survival of these neurons (Barbacid, 1995). In contrast, expression data indicate that trkB is not expressed in human sympathetic neurons or neuroblasts (Hoehner et al, 1995); in trkB or BDNF/NT-4–deficient mice, the number of sympathetic neurons appears to be unaffected (Barbacid, 1995). This strongly suggests that neither BDNF nor NT-4 have major roles during normal sympathetic neuronal differentiation. However, trkB is expressed in the chromaffin cells of human paraganglia at fetal stages, suggesting that p145trkB signaling is involved in the development of this sympathetic cell type (Hoehner et al, 1995).

Despite their sympathetic characteristics, neuroblastoma cells can express trkB. Truncated trkB mRNA is found in some differentiated tumors, while full-length trkB mRNA is almost exclusively expressed in MYCN-amplified, highly malignant neuroblastoma tumors (Nakagawara et al, 1994). Furthermore, all-trans-retinoic acid (RA) induces trkB expression in cultured human SH-SY5Y neuroblastoma cells (Kaplan et al, 1993; Matsumoto et al, 1995). The receptors are functional and mediate BDNF-induced morphologic differentiation. Under serum-free growth conditions, the RA-BDNF–treated cells appear to terminally differentiate (Encinas et al, 2000). These findings are difficult to explain in a context of normal sympathetic neuronal differentiation, considering the apparent lack of trkB expression in nonmalignant sympathetic neuroblasts and neurons. One putative explanation could be that the RA-BDNF–treated neuroblastoma cells shift toward a nonsympathetic neuronal or to a sympathetic neuroendocrine (chromaffin) phenotype. Previous studies have shown that RA alone induces morphologic differentiation of SH-SY5Y neuroblastoma cells and induces a small increase in choline actyltransferase activity (Adem et al, 1987; Påhlman et al, 1984), without promoting extensive norepinephrine or epinephrine production, the major transmitters of the sympathetic nervous system. However, the SH-SY5Y cells have the capacity to acquire a more developed noradrenergic phenotype, eg, 12-O-tetradecanoyl-phorbol-13-acetate (TPA) induces sympathetic neuronal differentiation in these cells, including a drastic increase in synthesized and stored norepinephrine (Påhlman et al, 1981, 1991). Thus, RA might induce a switch in the differentiation program of the SH-SY5Y cells, which would be in keeping with the well-established capacity of RA to shift cells into specific lineages during normal development.

In this report we have investigated if MYCN overexpression is causally linked to trkB expression in neuroblastoma cells and found that this is not the case. We have also induced trkB expression by treating SH-SY5Y cells with RA and analyzed the differentiated phenotype of RA-BDNF–treated SH-SY5Y cells. This treatment led to an increased expression of general neuronal marker genes and the cholinergic marker genes vesicular acetylcholine transporter (VAChT) and choline acetyl transferase (ChAT), suggesting an RA-BDNF–induced phenotypic shift from a sympathetic noradrenergic toward a cholinergic neuronal phenotype of these neuroblastoma cells. We further show that RA induces expression of platelet-derived growth factor receptor-α (PDGFR-α) and that RA-induced growth inhibition could be reversed by stimulation with cognate ligands.

Results

trkB Expression During Human Fetal Sympathetic Development and in MYCN-Amplified Neuroblastoma Tumors

The trkA and trkB expression in human fetal abdominal cross-sections (developmental weeks 3 to 24) containing sympathetic ganglia and paraganglia was analyzed immunohistochemically. p145trkB immunoreactivity was detected solely in sympathetic paraganglia (Fig. 1B), most prominently at the later gestational ages tested (21 to 24 weeks). For comparison, p140trkA immunoreactivity was analyzed in serial sections. Expression of p140trkA was detected in both sympathetic ganglia and paraganglia (Fig. 1A). The chromaffin cells of the human fetal adrenal medulla, detected by tyrosine hydroxylase expression (Fig. 1C), showed a weak p140trkA and a strong p145trkC staining, whereas only a trace amount of p145trkB immunoreactivity was detected, as exemplified in sections from a 24-week-old fetus (Fig. 1, D to F). In addition, adrenal sympathetic neuroblasts were p140trkA and p145trkC positive, and p145trkB negative (not shown, Hoehner et al, 1995). Thus, during the developmental stages investigated, p145trkB immunoreactivity was not detectable in sympathetic neuroblasts, while expression was strong in chromaffin cells of sympathetic paraganglia and weak in adrenal chromaffin cells.

Figure 1
figure 1

Immunohistochemical characterization of p140trkA (A and D), p145trkB (B and E), p145trkC (F), and tyrosine hydroxylase (TH, C) expression in human fetal sympathetic ganglia, paraganglia, and adrenal chromaffin cells; and MYCN amplification in two neuroblastoma tumors analyzed by fluorescence in situ hybridization (G to I). Abdominal cross-sections of a 16-week developmental age fetus were immunostained for p140trkA (A) and p145trkB (B), respectively, and counterstained with hematoxylin. The panels show sympathetic ganglia (arrows), immunoreactive for p140trkA but not p145trkB, juxtaposed to a p140trkA and p145trkB positive sympathetic paraganglion (asterisk). Scale bar, 200 μm. (C to F) Chromaffin cells (arrows) of the adrenal medulla of a 24-week-old human fetus stained for TH (C), p140trkA (D), p145trkB (E), and p145trkC (F). Scale bars, 200 μm (C) and 100 μm (D to F). As seen in D and F, the antibodies directed against p140trkA and p145trkC, respectively, gave a slight background staining. (G and H) Cells from MYCN-amplified cases ST130 and ST131, with high number of MYCN gene copies, shown as several red signals in the nucleus. (I) Control cells showing the normal two copies of MYCN (arrows).

Full size image

In light of the trkB expression pattern in developing sympathetic cells, the expression of trkB in highly malignant neuroblastoma cells seen both in vitro and in vivo (Kaplan et al, 1993; Nakagawara et al, 1994) is difficult to explain because these tumors are considered to be of sympathetic neuronal origin (Hoehner et al, 1996a). Expression of functional, full-length trkB in neuroblastoma tumor specimens was only detected in tumors with an amplified MYCN gene (Nakagawara et al, 1994), and presumably these tumor cells have a high MYCN expression. To verify that high MYCN expression due to MYCN amplification frequently is seen together with trkB expression and that trkB expression is not abundant in non–MYCN-amplified neuroblastomas, a set of cDNAs derived from tumor specimens was analyzed by semiquantitative PCR for the expression of MYCN and full-length trkB. As shown in Figure 2, A and B, none of the seven tumors lacking an amplified MYCN gene showed detectable trkB expression using RA-treated SH-SY5Y cells as a positive trkB control. In contrast, two out of five analyzed tumors with an amplified MYCN gene, determined by either Southern blot (Hedborg et al, 1992) or fluorescence in situ hybridization (FISH) analyses (Fig. 1, G and H), expressed trkB (Fig. 2C), thus confirming previously reported findings (Nakagawara et al, 1994). To test if overexpression of MYCN in non–MYCN-amplified neuroblastoma cells is sufficient to induce trkB expression, an MYCN-transfected SK-N-SH cell clone (SKMYC2) and MYCN-amplified neuroblastoma cell lines with high MYCN expression were analyzed for trkB expression. As shown in Figure 2D, overexpression of MYCN did not induce trkB expression in SK-N-SH cells, despite the fact that trkB expression can be induced by RA in the SK-N-SH derived sub clone, SH-SY5Y (Kaplan et al, 1993). In addition, RA did not induce MYCN expression in the SH-SY5Y cells (Fig. 2D). Furthermore, trkB was not expressed in any of the tested cell lines with MYCN amplification (Fig. 2D). The cell line data are in accordance with the findings of Nakagawara et al (Nakagawara et al, 1994), who reported only one out of six tested MYCN amplified cell lines as trkB-positive.

Figure 2
figure 2

Expression of MYCN and trkB in neuroblastoma tumors and cell lines. (A) RT-PCR analysis with MYCN-specific PCR primers of tumor samples from one MYCN-amplified (FH135, marked by an asterisk) and seven nonamplified tumors. After 24 PCR cycles, samples were analyzed by agarose gel electrophoresis. In parallel the same quantity of cDNA was amplified for 22 cycles with GAPD-specific primers as a reference for the amount of cDNA in each sample. cDNA from the MYCN-amplified LA-N-1 tumor cell line (asterisk) was included as positive control. (B) RT-PCR analysis with trkB-specific PCR primers of the same tumor samples analyzed in A with SH-SY5Y/wt cells treated for 6 days with 1 μm RA as a positive control for trkB expression. After 36 PCR cycles, samples were analyzed. As in A, amplification with GAPD-specific primers for 22 cycles was used as a reference for sample cDNA content. (C) RT-PCR analysis with MYCN- and trkB-specific PCR primers of tumor samples from five MYCN-amplified tumors (asterisks) with cDNA from SH-SY5Y/wt cells (± 1 μm RA) included as a negative and positive control, respectively, for trkB expression. After 26 (MYCN) or 34 (trkB) PCR cycles, samples were analyzed. In parallel the same quantity of cDNA was amplified for 23 cycles (for ST131, 24 cycles) with GAPD-specific primers as a reference for the amount of cDNA in each sample. (D) Northern blot analysis of expression of MYCN (endogenous MYCN 2.7 kb, exogenous MYCN 2.2 kb) and trkB (8.2 kb). SH-SY5Y/wt cells were grown for 6 days in serum-containing medium supplemented with 1 μm RA or 1 μm RA combined with 100 ng/ml brain-derived neurotrophic factor. Fifteen μg of total RNA was loaded in each lane and GAPD (1.5 kb) mRNA levels were used as a reference for the amount of loaded RNA. Asterisks denote MYCN-amplified cell lines. The bars in B through D indicate the trkB PCR product and mRNA, respectively.

Full size image

Phenotypic Characterization of RA- and RA-BDNF–treated SH-SY5Y Neuroblastoma Cells

Previously we reported that RA induces morphologic (neurite growth) and biochemical (neuron-specific enolase) differentiation of SH-SY5Y cells, but RA treatment inhibited the phorbol ester-induced and differentiation-related increase in norepinephrine levels (Påhlman et al, 1984). At a concentration of 1 μm, RA induced a distinct morphologic differentiation of SH-SY5Y cells characterized by growth cone-terminated processes bearing varicosities (Fig. 3C). As reported by Kaplan et al (Kaplan et al, 1993), BDNF alone (up to 100 ng/ml) did not induce any apparent morphologic changes compared with nontreated controls (Fig. 3, A and B), while RA treatment led to trkB expression and BDNF responsiveness. Compared with the morphology of cells treated with RA only, RA-BDNF–treated cells had a more distinct phenotype, manifested by longer neurites and a considerably higher number of varicosities (Fig. 3, C and D).

Figure 3
figure 3

Morphology of wild-type SH-SY5Y cells cultured in serum-containing medium for 6 days with the following additives: (A) nontreated controls, (B) 100 ng/ml brain-derived neurotrophic factor (BDNF), (C) 1 μm RA, and (D) 1 μm RA + 100 ng/ml BDNF. Bars represent 25 μm.

Full size image

As the phenotype of the RA-BDNF–treated cells grown in serum has not been further characterized and the induced trkB expression suggested that these cells did not develop into a noradrenergic, neuronal sympathetic phenotype, we extended the characterization by employing marker genes previously used to phenotypically describe the neuronal and neuroendocrine cells of the developing human sympathetic nervous system (Gestblom et al, 1997, 1999; Hedborg et al, 1995; Hoehner et al, 1996a). The axonal nature of the neurites of both RA- and RA-BDNF–treated cells was corroborated by the increased expression of GAP-43 (Fig. 4), a gene encoding an axonal protein involved in growth cone motility and synapse formation. RA induced less than a 2-fold increase in GAP-43 expression for up to 9 days of treatment. Consistent with previously published results from experiments performed under slightly different conditions (Encinas et al, 1999; Lesser et al, 1997), RA-BDNF treatment resulted in a 5- to 8-fold increase after 6 and 9 days of cell exposure. For comparison, we also included in our analysis SH-SY5Y/trkA cells treated for 4 days with NGF (Fig. 4), a treatment leading to increased GAP-43 expression and neuronal sympathetic differentiation (Edsjö et al, 2001; Lavenius et al, 1995).

Figure 4
figure 4

Northern blot analysis of trkB, GAP-43, and NPY expression in RA and RA–brain-derived neurotrophic factor (BDNF)–treated SH-SY5Y/wt cells and in nerve growth factor–treated SH-SY5Y/trkA cells. Cells were grown in serum-containing medium, with the addition of 1 μm RA and 1 μm RA + 100 ng/ml BDNF (SH-SY5Y/wt) or 100 ng/ml NGF (SH-SY5Y/trkA), for the indicated time periods. Total RNA (15 μg) was analyzed for the expression of trkB (8.2 kb), GAP-43 (1.4 kb), and NPY (0.8 kb), using GAPD (1.5 kb) mRNA levels as a measure of the amount of loaded mRNA. The band intensities were analyzed with phosporimaging, and the relative amounts of mRNA (normalized to GAPD levels and controls) are shown. The increase in trkB mRNA levels could not be quantified since no reliable signal was obtained from the control cell RNA.

Full size image

Neurotransmitter Synthesis in RA- and RA-BDNF–Treated SH-SY5Y Cells

Norepinephrine and NPY are two important neurotransmitters produced in catecholaminergic cells of the sympathetic nervous system. The concentration of the major neurotransmitter, norepinephrine, was analyzed in RA- and RA-BDNF–treated cells. No significant increase in norepinephrine levels compared with control cells was detected, but there was a tendency toward increased norepinephrine levels in the RA-BDNF–treated cells (Fig. 5A). However, if real, this effect was small compared with the 6-fold increase in the norepinephrine concentration induced by NGF treatment of SH-SY5Y/trkA cells (Fig. 5A). The rate-limiting enzyme of the catecholamine synthesis, tyrosine hydroxylase (TH), was analyzed in these cells at the protein level by Western blotting. Neither RA nor the combination of RA and BDNF induced an increased TH-expression during a time period (3 to 6 days) when the cells differentiated morphologically and biochemically (Fig. 5B). Instead a slight reduction in the TH levels was observed after 6 days of RA and RA-BDNF treatment.

Figure 5
figure 5

Phenotypical characterization of RA- and RA–brain-derived neurotrophic factor (BDNF)–treated SH-SY5Y cells. (A) HPLC analysis of norepinephrine levels in cell lysates from SH-SY5Y/wt cells treated for 6 days with 1 μm RA or 1 μm RA + 100 ng/ml BDNF (n = 5) and in SH-SY5Y/trkA-transfected cells treated for 4 days with 100 ng/ml NGF (n = 7). The mean norepinephrine levels are represented by bars; error bars indicate sem. (B) Western blot analysis of tyrosine hydroxylase protein levels in RA- and RA-BDNF–treated SH-SY5Y/wt cells. Cells were grown in serum-containing medium, with the addition of 1 μm RA and 1 μm RA + 100 ng/ml BDNF. After 3 or 6 days, cells were harvested, and 30 μg of whole cell lysate proteins was loaded in each lane. (C) Northern blot analysis of VAChT and HASH-1 expression in RA- and RA-BDNF–treated SH-SY5Y/wt and LA-N-5 cells. Cells were grown in serum-containing medium, with the addition of 1 μm RA or 1 μm RA + 100 ng/ml BDNF, where indicated. After 6 (SH-SY5Y/wt) and 7 days (LA-N-5), cells were harvested and lysed. Fifteen μg of total RNA from each sample was analyzed for the expression of VAChT (2.5 kb) and HASH-1 (2.7 kb), using GAPD (1.5 kb) mRNA levels as a measure of the amount of loaded mRNA. The band intensities were analyzed with densitometry, and the relative amounts of mRNA (normalized to GAPD levels and controls) are shown. (D) RT-PCR analysis of samples from SH-SY5Y/wt cells grown as in (C, RA, RA + BDNF), using ChAT-specific primers. After 32 PCR cycles, samples were analyzed. In parallel the same quantity of cDNA was amplified for 19 cycles using GAPD-specific primers and was used as a reference for sample cDNA content.

Full size image

The expression of NPY, which in the sympathetic nervous system characteristically is expressed in the neuronal but not in the chromaffin paraganglionic lineage (Gestblom et al, 1997), increased significantly (approximately 4-fold) after 9 days of RA-BDNF exposure (Fig. 4). The NPY mRNA levels were considerably lower after stimulation for shorter time periods, and treatment with RA alone resulted in decreased NPY mRNA levels at all analyzed time points (Fig. 4). Although the elevation of NPY expression in cells treated for 9 days with RA-BDNF was significant, it was lower than that seen in TPA-treated SH-SY5Y/wt cells and NGF-treated SH-SY5Y/trkA cells (Fig. 4) and in previous results (Lavenius et al, 1995; Påhlman et al, 1991).

During normal development, a subpopulation of catecholaminergic sympathetic neuroblasts transdifferentiate and become cholinergic. Neuroblastoma cells, as well as catecholaminergic sympathetic neuroblasts, show cholinergic traits with weak expression of cholinergic marker genes like ChAT (Biedler et al, 1978; Francis and Landis, 1999; Yokomori et al, 1983). We have previously reported that RA treatment of SH-SY5Y cells promotes a cholinergic phenotype measured as an increase in choline acetyltransferase activity (Adem et al, 1987). A switch to a cholinergic phenotype was recently reported to occur during RA treatment of human NB69 and LA-N-5 neuroblastoma cells (Hill and Robertson, 1997; Handler et al, 2000). We, therefore, analyzed the effect of BDNF on RA-treated SH-SY5Y cells on the expression of the cholinergic marker genes, the acetylcholine transporter VAChT and ChAT. As shown in Figure 5C, RA induced a minute increase in VAChT expression, which was enhanced (1.5-fold over control) by BDNF. The latter increase was in parity with that seen in RA-treated LA-N-5 cells (1.3-fold increase) used as a positive control of neuroblastoma cells acquiring a cholinergic phenotype (Hill and Robertson, 1997). As expected, ChAT expression was also induced by RA and RA-BDNF (Fig. 5D). The induced cholinergic features of the RA and RA-BDNF–treated SH-SY5Y cells were also accompanied by a complete down-regulation of HASH-1 (Fig. 5C), a transiently expressed transcription factor essential for the commitment and differentiation of migrating neural crest cells toward a noradrenergic sympathetic neuroblastic phenotype.

PDGF Receptor Expression in RA- and RA-BDNF–Treated SH-SY5Y Cells

The RA-treated SH-SY5Y cells expressing trkB are highly proliferating in the presence of BDNF and serum, and the BDNF stimulation also enhances a migratory response (Matsumoto et al, 1995). Thus, the BDNF-induced differentiation does not lead to growth inhibition, as opposed to for instance the TPA-induced differentiation of SH-SY5Y cells (Påhlman et al, 1991). The development of the phorbolester-driven phenotype requires serum or a growth factor such as PDGF. The receptor, PDGFR-α, is expressed in SH-SY5Y cells and transduces a mitogenic response upon stimulation with PDGF-AA (Påhlman et al, 1992). As this receptor is down-regulated in TPA-differentiated SH-SY5Y cells, presumably as part of an induced sympathetic differentiation program (Figure 6A; previous results [Påhlman et al, 1992]), we analyzed the expression pattern of this receptor in RA and RA-BDNF–treated cells. RA induced a several-fold increase in PDGFR-α expression, and the combination with BDNF slightly reduced that increase. In contrast, both NGF-treated SH-SY5Y/trkA and TPA-treated SH-SY5Y wild-type cells down-regulated PDGFR-α (Fig. 6A). Thus, with regard to PDGFR-α expression, there was a distinct difference between RA-treated and TPA-treated SH-SY5Y cells.

Figure 6
figure 6

Platelet-derived growth factor receptor (PDGFR) expression and PDGF effects on cell numbers. (A) Northern blot analysis of PDGFR-α expression in SH-SY5Y/wt cells treated with 1 μm RA, 1 μm RA + 100 ng/ml brain-derived neurotrophic factor and 16 nm TPA, and in SH-SY5Y/trkA cells treated with 100 ng/ml NGF. The cells were harvested after the indicated treatment periods, and 15 μg of total RNA was analyzed for the expression of PDGFR-α (6.5 kb), using GAPD (1.5 kb) mRNA levels as a measure of the amount of loaded mRNA. The band intensities were analyzed with phosporimaging, and the relative amounts of mRNA (normalized to GAPD levels and controls) are shown. (B) Number of viable cells in SH-SY5Y/wt cell cultures, grown for 3 days in 1% or 10% FCS, as measured by MTT assay (n = 8 to 12). Where indicated, 50 ng/ml PDGF-AA, 50 ng/ml PDGF-BB, and/or 1 μm RA was added to the medium. (C) Isolated effect of PDGFs ±RA. Data from 6B, expressed as percent increase in cell numbers in cultures treated with PDGF-AA or –BB ± RA, compared with the amount of viable cells in untreated control and RA-treated cells, respectively.

Full size image

RA induces differentiation accompanied by growth inhibition in many neuroblastoma cell lines, which is the rationale behind treatment of minimal residual disease in neuroblastoma. It was therefore of interest to test whether PDGF-AA (binds exclusively to the PDGFR-α) and PDGF-BB (binds both PDGFR-α and PDGFR-β) could mitogenically stimulate the RA-treated SH-SY5Y cells. In the absence of RA, SH-SY5Y cell growth was virtually unaffected by PDGF-AA treatment but increased slightly with PDGF-BB (Fig. 6, B and C). In contrast but in agreement with PDGFR-α expression data, PDGF-AA potently stimulated cell growth of RA-treated SH-SY5Y cells, both at 1% and 10% serum concentration. PDGF-BB, on the other hand, stimulated growth also in non–RA-treated cells, an effect markedly potentiated in RA-treated cells. Considering the significant effect of PDGF-AA on these cells, the augmented PDGF-BB response was conceivably due to PDGF-BB stimulation of the up-regulated PDGFR-α (Fig. 6, B and C).

Discussion

Data from characterization of neuroblastoma tumors employing a broad panel of neuronal and neuroendocrine marker genes are consistent with neuroblastomas arising from cells of the sympathetic nervous system (Hoehner et al, 1996a). There are, however, examples of genes expressed in neuroblastoma that usually are not transcribed in developing or mature sympathetic cells. Many of these genes are expressed in other neural crest derived lineages, and one hypothesis is that the expression of such genes reflects the immature or stem cell character of these neuroblastoma cells (Påhlman and Hedborg, 2000). The expression of trkB in some tumors and in RA-treated cultured neuroblastoma cells is one example in which the expression profile does not appear to easily fit into such a developmental model. Although trkB is expressed in developing chromaffin cells of the paraganglia, the differentiation markers of RA-BDNF–treated SH-SY5Y cells are not in keeping with a paraganglionic phenotype. The formation of long neurites, and the GAP-43 induction in RA-BDNF–treated cells, indicate that this treatment promotes a neuronal phenotype (Gestblom et al, 1997). Furthermore, the low norepinephrine levels and lack of TH induction also argue against these cells becoming chromaffin because high catecholamine production is a hallmark of a chromaffin cell. The phenomenon as such, ie, transdifferentiation of neuroblastoma cells displaying sympathetic neuronal features into a sympathetic extra-adrenal chromaffin lineage, can occur spontaneously, preferentially in low-stage neuroblastoma tumors (Gestblom et al, 1997). Instead, RA-BDNF induced the expression of cholinergic marker genes; the increase in VAChT expression appeared primarily to be a BDNF effect, while the increased ChAT expression was mainly RA-induced. These changes, together with the increase in NPY expression, might suggest that the RA-BDNF–treated SH-SY5Y cells develop into a sympathetic cholinergic neuronal lineage, although it is to our knowledge not known if human sympathetic cholinergic neuroblasts or neurons express trkB.

There are now several examples of cultured neuroblastoma cells that respond to RA treatment by expressing cholinergic features (Handler et al, 2000; Hill and Robertson, 1997), suggesting that this is a general effect of RA on neuroblastoma cells. Consistent with this conclusion, cultured sympathetic neuroblasts also switch from a catecholaminergic to a cholinergic phenotype when treated with RA (Matsushima and Bogenmann, 1993). As discussed below, the in vitro RA-induced tumor cell phenotype, including expression of functional growth factor receptors, might in vivo result in tumors with new growth properties potentially adding to an increased malignant behavior; however, we do not know if the induced trkB and PDGFR-α expression is linked to the cholinergic phenotype. To our knowledge, there is only one report in which the clinical behavior of catecholaminergic and cholinergic neuroblastomas has been compared (Yokomori et al, 1983). Based on pretreatment measurements, no clear difference in outcome between catecholaminergic and cholinergic tumors was found in this report. However, in relapsed tumors treated with chemotherapy or radiation, there was a tendency toward a change from a catecholaminergic to a cholinergic phenotype, which might suggest that the cholinergic phenotype is associated with aggressive tumor behavior (Yokomori et al, 1983).

Highly malignant neuroblastomas, including established cell lines, have many characteristics (eg, HASH-1 and MYCN expression) in common with the sympathoadrenal progenitor cell defined primarily in rodents. Neuroblastoma cells might thus be arrested at such an early differentiation stage that they can behave as progenitor cells. If this is indeed the case, it is reasonable to assume that conditions can be found at which cultured neuroblastoma cells can be forced into cell lineages other than those of the sympathetic nervous system. The neural crest-derived progenitor cells giving rise to the sympathetic nervous system are also progenitors for some of the neurons of the enteric nervous system (Blaugrund et al, 1996; Carnahan et al, 1991). Interestingly, enteric neurons do express trkB (Hoehner et al, 1996b) and NPY (Gaetano et al, 1991; Gordon-Weeks, 1988). Furthermore, although mature intrinsic enteric neurons appear not to express norepinephrine or epinephrine, immature enteric neurons do (Kapur, 1993). Thus, as many of the characteristics of the RA-BDNF–treated cells appear to be similar to those of immature or differentiating enteric neurons, differentiation into this lineage is, formally, an alternative explanation of our data.

Amplification of MYCN in neuroblastomas correlates strongly to poor prognosis. MYCN functions as a heterodimeric transcription factor, forming a complex with Max. It is generally believed that overexpression of MYCN in neuroblastoma results in altered gene expression, although few target genes have been identified (Lasorella et al, 2000; Schwab, 2000; Shohet et al, 2002). One question we wanted to answer was whether high MYCN expression was directly linked to the trkB expression observed in some neuroblastomas. We conclude that the trkB expression detected in neuroblastoma tumors is not directed by high MYCN expression. It is a fact, though, that expression of full-length trkB in neuroblastoma specimens primarily is found in tumors with MYCN amplification. Whether the trkB expression is regulated via co-amplified genes like DDX1 (George et al, 1997; Godbout and Squire, 1993) or merely is a reflection of a phenotypic heterogeneity among neuroblastoma tumors, possibly requiring high MYCN expression, needs to be further addressed.

The fact that RA induces trkB and PDGFR-α expression in SH-SY5Y cells and that these receptors transduce mitogenic signals to these cells are findings that might have clinical relevance. RA and RA homologues have for a long time been recognized as promising drugs for treatment of neuroblastoma, with the anticipation that they will induce tumor cell differentiation and growth arrest in vivo, similar to the effects seen in vitro. With one noticeable and promising recent exception (Matthay et al, 1999), RA and its homologues have not lived up to these expectations, and RA treatment has, so far, not dramatically changed the outcome of neuroblastoma patients. One molecular explanation for a weak in vivo effect of RA on neuroblastoma growth could be that this treatment induces growth factor receptor expression, similar to the in vitro results presented in this study. If that is the case, these growth factor receptors could transduce mitogenic signals, counteracting the growth inhibitory effect that would follow from the RA-induced differentiation. A future treatment strategy might therefore be to combine RA with one or several of the novel receptor tyrosine kinase inhibitors now becoming available.

Materials and Methods

Cell Cultures, Tumor Materials, and Assessment of Cell Numbers

SH-SY5Y wild-type (SH-SY5Y/wt) neuroblastoma cells, a neuroblastic subclone of the non–MYCN-amplified SK-N-SH cell line (Biedler et al, 1973), as well as stably trkA-transfected SH-SY5Y cells (SH-SY5Y/trkA, clone 6:2) (Lavenius et al, 1995), were cultured in Eagle’s minimum essential medium with the following additives: 10% FCS (FCS), 100 IU/ml penicillin V, and 100 μg/ml streptomycin. LA-N-1, LA-N-2, and LA-N-5 cells were grown in RPMI 1640 medium with 10% FCS and the same antibiotics added. To induce differentiation the following additives were used: RA (Sigma, St. Louis, Missouri), human recombinant BDNF (PeproTech EC Ltd., London, United Kingdom), mouse NGF 2.5S (Promega, Madison, Wisconsin), and PDGF-AA and -BB (PeproTech). Media, FCS, and antibiotics were from Life Technologies Inc. (Rockville, Maryland). The SKMYC2 clones of SK-N-SH cells transfected with MYCN has previously been described (Judware and Culp, 1995). The RNA analyzed was extracted from cells grown with 200 μg/ml of Hygromycin B (Calbiochem, San Diego, California). The amount of viable cells in cell cultures was analyzed measuring the conversion of the tetrazolium salt MTT to formazan (CellTiter 96, Promega). Cells were plated at a density of 7500 cells/well in 96-well plates with factors used added prior to the addition of cells.

The tumor material used has been described previously (Hedborg et al, 1992; Martinsson et al, 1995). MYCN amplification was determined by either Southern blot (Hedborg et al, 1992; Martinsson et al, 1995) or FISH analyses (Martinsson et al, 1995). FISH analysis was performed as previously described (Fröstad et al, 1999).

Northern Blot and Semiquantitative RT-PCR Analyses

The guanidine-isothiocyanate/phenol-chloroform extraction method (Chomczynski and Sacchi, 1987) was used for isolation of total cellular RNA. Fifteen μg of RNA was electrophoretically separated on a 1% agarose-formaldehyde gel and blotted onto a nylon membrane (Hybond-N, Amersham Pharmacia Biotech, Uppsala, Sweden). The following cDNA probes were labeled with [32P] dCTP using an oligonucleotide labeling kit (Amersham Pharmacia Biotech): Neuropeptide tyrosine (NPY) (Minth et al, 1986), MYCN (Kohl et al, 1983), glyceraldehyde-3-phosphate-dehydrogenase (GAPD) (Tso et al, 1985), growth-associated protein 43 (GAP-43) (Örtoft et al, 1993), trkA (Martin-Zanca et al, 1989), trkB (Klein et al, 1989), VAChT (a 1605 bp probe amplified from LA-N-2 cDNA, using the 30-mer 5′-CGC GGA TCC ATG GAA TCC GCG GAA CCT GCG-3′ as forward and the 29-mer 5′-CGC GAA TTC TAG CTG CGG GTG TAG TAG TA-3′ as reverse primer), and human PDGFR-α (Claesson-Welsh et al, 1989). Hybridizing probes were visualized by autoradiography, and RNA levels were quantified and related to the corresponding GAPD mRNA level by phosphor image analysis, using a Fuji BAS 2000 equipment (Fujifilm, Tokyo, Japan) or by densitometric analysis.

Total RNA from snap-frozen tumor specimens and cultured cells were extracted as described above and treated with RNase-free DNase I (Appligene Oncor, Illkirch, France) prior to cDNA synthesis. cDNA synthesis was performed using the Gibco BRL SuperScript First-Strand Synthesis System for RT-PCR (Life Technologies Inc.). Duplicates of each sample were made and the reverse transcriptase was added to one of them (RT+), while the other (RT−) was used as a negative control for contaminating genomic DNA in the samples. The cDNA was purified using the QIAquick PCR Purification kit (Quiagen, Venlo, the Netherlands).

In the subsequent PCR reactions, template cDNAs were mixed with 10 × PCR buffer (Sigma) and 20 nm of each primer in a total volume of 60 μl along with 2.5 U of TaqDNA polymerase (Sigma). Amplification was carried out for 19 to 36 cycles, with a denaturation temperature of 95° C, annealing temperature of 57° C, and an extension temperature of 72° C for 30 seconds each. RT− samples were run in parallel for at least as many cycles as the RT+ samples to check for genomic DNA. Ten μl of the collected aliquots were loaded on a 1% agarose gel stained with SYBR Green I Nucleic Acid Gel Stain (Roche, Basel, Switzerland). Band intensities were recorded and analyzed using LAS-1000 CCD imaging equipment with accompanying image analyzing software (Fujifilm). For each primer set, the exponential phase of amplification was determined by removing aliquots at four different cycle numbers. The genes analyzed were ChAT (forward primer: 5′-CTGCTGCTATGGTCCTGTGG-3′, reverse: 5′-TCAAG-GTTGGTGTCCCTGGC -3′); GAPD (forward primer: 5′-CCATCACCATCTTCCAGGAGC-3′, reverse: 5′-AGTGA-TGGCATGGACTGTGGTC-3′); MYCN (forward primer: 5′-TTCCTCCTCCAACACCAAGG-3′, reverse: 5′-CGCCT-CGCTCTTTATCTTCT-3′); and trkB (forward primer: 5′-GAG CAT CAT GTA CAG GAA AT-3′, reverse: 5′-CTT GAT GTT CTT CCT CAT GT-3′). The trkB primers will generate a 235-bp PCR product covering a part of the 3′ end-coding for the tyrosine kinase domain, thereby only amplifying trkB cDNA generated from full-length transcripts and not the cDNAs from the two known truncated trkB mRNAs (Barbacid, 1994; Klein et al, 1989, 1990; Middlemas et al, 1991).

Immunohistochemical and Western Blot Analyses

Human fetal tissue was obtained from either elective or spontaneous abortions from 9- to 25-weeks gestational age, kindly provided by Dr. B. Sandstedt, Stockholm. Ethical approval (Dnr 93–216) was obtained from the local ethical committee of Karolinska Hospital, Stockholm, Sweden. p140trkA, p145trkB, and p145trkC expression were analyzed immunocytochemically as described (Hoehner et al, 1995), in which the specificities of the antisera employed (Santa Cruz, CA) were characterized and reported. Sections were counter-stained with hematoxylin.

For Western blot analysis, cells were lysed in 10 mm Tris-HCl (pH 7.2), 160 mm NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1 mm EGTA, 1 mm EDTA in the presence of Complete Protease Inhibitor (Roche). For TH-protein–expression analysis, 30 μg of whole-cell lysate protein was separated on an SDS-PAGE gel and blotted onto a Hybond C extra membrane (Amersham Pharmacia Biotech). A monoclonal antityrosine hydroxylase antibody (Roche), diluted 1:500, was used along with a horseradish peroxidase coupled secondary antibody diluted 1:5000 (Amersham Pharmacia Biotech).

Norepinephrine Analysis

RA and neurotrophin-treated cells (106/10-cm plate) were grown for 4 days and then washed and harvested in ice-cold PBS. The pelleted cells were resuspended in 0.5 M trichloroacetic acid and then centrifuged at 21,000 × g for 5 minutes in 4° C. The supernatant was saved and pH adjusted to pH 4 with NaOH. The pellet was resuspended in 1 M NaOH for protein determination according to Bradford. After addition of an internal standard (3,4-dihydroxynorepinephrine) the supernatant samples were extracted and derivatized with 1,2 diphenylethylenediamine as previously described (van der Hoorn et al, 1989). The samples were analyzed by HPLC using a 25 × 2–mm precolumn (Perisorb RP 18, 30- to 40-μm particle diameter) and a 250 × 4.6–mm column (Hypersil ODS, 5-μm particle diameter) using 50 mm Na acetate (pH 7.0):acetonitrile:methanol (5:4:1) as mobile phase. The fluorescence was detected with either a Shimadzu RF-10A XL or a Jasco Fluorescence Detector 821, and the signal was integrated with a Shimadzu C-R3A integrator.