Sequential Treatment of SH-SY5Y Cells with Retinoic Acid and Brain-Derived Neurotrophic Factor Gives Rise to Fully Differentiated, Neurotrophic Factor-Dependent, Human Neuron-Like Cells


  • Mario Encinas,

  • Montse Iglesias,

  • Yuhui Liu,

  • Hongyin Wang,

  • Ashraf Muhaisen,

  • Valentin Ceña,

  • Carme Gallego,

  • Joan X. Comella

  • Lippincott Williams & Wilkins, Inc., Philadelphia

  • Abbreviations used: BNDF, brain-derived neurotrophic factor; BrdU, 5-bromo-2′-deoxyuridine; cdk, cyclin-dependent kinase; DMEM, Dulbecco's modified Eagle's medium; GAP-43, growth-associated protein-43; GFAP, glial fibrillary acidic protein; MAP2, microtubule-associated protein 2; NGF, nerve growth factor; NF-L, -M, and -H, low-, medium-, and high-molecular-weight neurofilaments, respectively; NSE, neuron-specific enolase; PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline, PCD, programmed cell death; pRB, retinoblastoma susceptibility gene product; RA, retinoic acid; SDS, sodium dodecyl sulfate; Trk, tyrosine receptor kinase; TPA, 12-O-tetradecanoylphorbol 13-acetate; TUNEL, terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling.

Address correspondence and reprint requests to Dr. J. X. Comella at Grup de Neurobiologia Molecular, Departament de Ciències Mèdiques Bàsiques, Universitat de Lleida, Avda. Rovira Roure, 44, 25198 Lleida, Spain. E-mail:


Abstract: A rapid and simple procedure is presented to obtain nearly pure populations of human neuron-like cells from the SH-SY5Y neuroblastoma cell line. Sequential exposure of SH-SY5Y cells to retinoic acid and brain-derived neurotrophic factor in serum-free medium yields homogeneous populations of cells with neuronal morphology, avoiding the presence of other neural crest derivatives that would normally arise from those cells. Cells are withdrawn from the cell cycle, as shown by 5-bromo-2′-deoxyuridine uptake and retinoblastoma hypophosphorylation. Cell survival is dependent on the continuous presence of brain-derived neurotrophic factor, and removal of this neurotrophin causes apoptotic cell death accompanied by an attempt to reenter the cell cycle. Differentiated cells express neuronal markers, including neurofilaments, neuron-specific enolase, and growth-associated protein-43 as well as neuronal polarity markers such as tau and microtubule-associated protein 2. Moreover, differentiated cultures do not contain glial cells, as could be evidenced after the negative staining for glial fibrillary acidic protein. In conclusion, the protocol presented herein yields homogeneous populations of human neuronal differentiated cells that present many of the characteristics of primary cultures of neurons. This model may be useful to perform large-scale biochemical and molecular studies due to its susceptibility to genetic manipulation and the availability of an unlimited amount of cells.

In metazoan organisms, maintenance of homeostasis requires the proper relationships among cell proliferation, differentiation, and death. Somatic cells proliferate and divide by executing tightly regulated processes during the cell cycle, whereas apoptotic cell death allows an organism to eliminate unwanted cells through a safe, orderly process. In the case of the nervous system, roughly half of the neurons initially generated die by apoptosis in a well-defined interval, coincident with the establishment of synaptic connections. This cell death is known as naturally occurring or programmed cell death (PCD), and it is believed to match the number of innervating neurons to the size of the target cell population, as well as to eliminate aberrant synaptic contacts (Oppenheim, 1991; Henderson, 1996). Apoptotic neuronal cell death that occurs outside this developmental window may contribute to certain pathological conditions such as Alzheimer's disease (Anderson et al., 1996).

PCD seems to result from the failure of individual neurons to gain access to neurotrophic factors that are released in limited amounts by target tissues, glial cells, or afferent inputs (Burek and Oppenheim, 1996). The family of neurotrophins [composed mainly by nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin 3, and neurotrophin 4/5] is the best-characterized group of such neurotrophic factors and has been shown to promote survival of several types of neurons both in vivo and in vitro. Targeted disruption of the genes encoding for neurotrophins or their specific tyrosine kinase receptors (collectively known as Trks) causes cell loss of specific neuronal populations (reviewed by Klein, 1994; Snider, 1994). When cultured in vitro, different neuronal populations show a selective survival response to particular neurotrophins, and their removal triggers an apoptotic cell death that mimics the phenomenon of PCD (reviewed by Davies, 1994; Lewin and Barde, 1996).

The molecular mechanisms underlying the regulation of the apoptotic process are objects of intense study. Recent lines of evidence argue for a strong interrelationship between cell cycle control and apoptosis. These findings include morphological changes occurring during apoptosis that are reminiscent of mitosis, as well as involvement of cell cycle-regulatory molecules such as p53, retinoblastoma susceptibility gene product (pRB), E2F, and cyclin-dependent kinases (cdks) in the apoptotic process (reviewed by Kasten and Giordano, 1998; King and Cidlowski, 1998). In the nervous system, several cell cycle-regulatory molecules are expressed, including cyclins and cdk4 and 5, although neurons are postmitotic cells (Freeman et al., 1994). More important is that, in sympathetic neurons, NGF withdrawal induces a selective increase in the levels of cyclin D1 (Freeman et al., 1994), and overexpression of this molecule in N1E-115 cells leads to apoptotic cell death (Kranenburg et al., 1996). In staggerer and lurcher mutant mice, which show massive death of cerebellar granule cells due to defects in the development of their synaptic targets, an elevation of cyclin D and proliferating cell nuclear antigen levels and 5-bromo-2′-deoxyuridine (BrdU) uptake is observed during the onset of cell death (Herrup and Busser, 1995). Moreover, chemical inhibitors of cdks or dominant-negative forms of cdk4 and cdk6 promote survival of NGF-deprived sympathetic neurons (Park et al., 1997) and KCl-deprived cerebellar granule cells (Padmanabhan et al., 1999). These and other observations had led to the hypothesis that apoptosis occurs after an attempt to enter the S-phase of the cell cycle in an inadequate cellular context, where contradictory or conflicting growth signals may converge. Finally, alterations in the expression of cell cycle-related proteins have been described in neurodegenerative disorders such as Alzheimer's disease (Vincent et al., 1997; Nagy et al., 1997; Busser et al., 1998; Giovanni et al., 1999) and other pathological situations (Nagy and Esiri, 1998; Timsit et al., 1999).

Much of the above information has been obtained by the use of primary cultures of neurons. However, the reduced amount of cells obtained and the limited susceptibility of genetic manipulation of these systems make biochemical and molecular approaches difficult to perform. Another important handicap is the ethical problem that arises when using human embryonic neurons for pathophysiological studies of human neurodegenerative diseases. Several protocols that yield neuronal differentiated cells have been applied to human cell lines (see, for example, Pleasure et al., 1992; Hill and Robertson, 1997; Prince and Oreland, 1997). However, most of them are time-consuming and do not yield a population with known trophic dependencies, a requirement that must be accomplished to mimic the phenomenon of PCD that occurs in vivo and in primary cultures of neurons. In this work, we present a protocol of differentiation based on the sequential treatment of the SH-SY5Y human neuroblastoma cell line with retinoic acid (RA) and BDNF. This protocol yields homogeneous populations of neuronal differentiated cells that are strictly dependent on BDNF for their survival. When BDNF is removed from the culture medium, cells enter an apoptotic cell death accompanied by an attempt to reenter the cell cycle. This model seems to mimic naturally occurring cell death and could be a useful tool to perform large-scale biochemical and genetic studies.


Cell culture

The SH-SY5Y neuroblastoma cell line was kindly provided by Dr. D. Martin-Zanca (Salamanca, Spain). Cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 2 mM L-glutamine, penicillin (20 units/ml), streptomycin (20 mg/ml), and 15% (vol/vol) heat-inactivated fetal calf serum (GIBCO, Gaithersburg, MD, U.S.A.). Cells were maintained at 37°C in a saturated humidity atmosphere containing 95% air and 5% CO2. Cells were seeded at an initial density of 104 cells/cm2 in culture dishes (Corning, Corning, NY, U.S.A.) previously coated with 0.05 mg/ml collagen (Collaborative Biomedical Products, Bedford, MA, U.S.A.). all-trans-RA (Tocris Cookson, Bristol, U.K.) was added the day after plating at a final concentration of 10 μM in DMEM with 15% fetal calf serum. After 5 days in the presence of RA, cells were washed three times with DMEM and incubated with 50 ng/ml BDNF (Alomone Laboratories, Jerusalem, Israel) in DMEM (without serum) for different intervals.

Immunoprecipitation and western blot

Immunoprecipitation of Trk receptors was performed as previously described (Encinas et al., 1999). For western blot experiments, cells were washed with phosphate-buffered saline (PBS), lysed with 2% sodium dodecyl sulfate (SDS) and 125 mM Tris (pH 6.8), sonicated, and boiled for 5 min. Protein was quantified by means of the Bio-Rad DC protein assay. Protein (20-50 μg) was resolved in standard SDS-polyacrylamide gel electrophoresis (PAGE) minigels (Miniprotean; Bio-Rad) and were transferred to an Immobilon-P membrane (Millipore, Bedford) using a Pharmacia semidry Transblot. The SDS-PAGE conditions for pRB were slightly modified. Thus, pRB hyperphosphorylated forms were resolved in 6% acrylamide gels (see Fig. 4C, left panel), whereas the cleaved pRB product was identified in long (20-cm) 7.5% acrylamide gels (see Fig. 4C, right panel). Membranes were blocked in Tris-buffered saline with Tween 20 [20 mM Tris (pH 7.4), 150 mM NaCl, and 0.05% Tween 20] containing 5% nonfat dry milk for 30 min at room temperature. Primary antibodies were incubated for 1 h at room temperature and subsequently incubated with peroxidase-conjugated antibodies. Blots were finally developed with the enhanced chemiluminescence western blotting detection system (Amersham, Little Chalfont, Bucks, U.K.). The following antibodies were used at the dilutions recommended by the manufacturers: α-neuron-specific enolase (NSE; Cambridge Research Biochemicals, Cheshire, U.K.), α-growth-associated protein-43 (GAP-43; Transduction Laboratories, Lexington, KY, U.S.A.), α-glial fibrillary acidic protein (GFAP), α-low-molecular-weight neurofilament (NF-L; 68K), and α-medium-molecular-weight neurofilament (NF-M; 160K) (Sigma, Madrid, Spain), and α-pRB (PharMingen, San Diego, CA, U.S.A.).

Figure 4.

RA-BDNF-treated cells are arrested in G1. Withdrawal of BDNF induces S-phase entry and apoptotic cell death. A: Flow cytometric analysis of cell cycle distribution [assessed by propidium iodide (PI) staining] during RA-BDNF differentiation. Insets: BrdU uptake of pulse-labeled cells under the same conditions. cyc, naive (cycling) cells; RA, 5 days of RA; +BDNF, 5 days of RA followed by 3 days of BDNF; and -BDNF, the same as +BDNF followed by 2 days in the absence of neurotrophin. Bottom panels show floating (nonadherent) and attached (adherent) cells from -BDNF condition. B: The proportion of cells in each cell cycle phase was quantified for the analyzed experimental conditions. C, left panel: Analysis of the state of phosphorylation of pRB, as deduced by gel mobility shift. The treatments are similar as in A, but the time of incubation with BDNF was 7 days. Equal loading of the lanes was verified by Coomassie Blue staining of the membranes, and the most abundant band is shown in the lower panel (CB). Right panel: Removal of BDNF also caused the appearance of a lower-molecular-weight pRB species (arrowhead), which is likely to be a caspase-cleaved fragment of this protein. Note that the gels in the left and right panels differ in the running gel conditions (see Materials and Methods).

FIG. 4.

Terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling (TUNEL) assay

Cells were fixed in freshly prepared 2% paraformaldehyde for 30 min at room temperature and permeabilized with 0.1% Triton X-100 and 0.1% sodium citrate for 10 min at 4°C. Subsequently, cells were incubated with 50 μ1 of a reaction mixture containing 0.3 nmol of fluorescein-12-UTP (Boehringer Mannheim, Barcelona, Spain), 3 nmol of dATP, and 20 units of terminal deoxynucleotidyl transferase (GIBCO) for 1 h at 37°C in a humidified chamber. Cells were then washed with PBS and incubated with 0.05 μg/ml Hoechst 33258 (Sigma) for 30 min at room temperature. Finally, cells were mounted with Fluoprep (BioMérieux, Marcy l'Etoile, France) and observed under UV illumination in an epifluorescence microscope (Nikon, ECLIPSE model 600; Izasa, Barcelona) coupled to a digital CCD camera (AstroCam; Life Sciences Resources, Cambridge, U.K.). Image analysis was performed with Esprit Rego software (Life Sciences Resources). Quantitative data were obtained by scoring at least 500 cells under each experimental condition in three replicas. Experiments were repeated at least two times, and counts were made in a blinded manner.

Caspase activity

Cells were rinsed with ice-cold PBS and lysed in a buffer containing 100 mM HEPES (pH 7.4), 5 mM dithiothreitol, 5 mM EGTA, 0.04% NP-40, and 20% glycerol. Extracts were then centrifuged at 5,000 g for 10 min, and protein concentrations were determined by the assay of Bradford (1976). Cell extracts (10-20 μg) were diluted in 50 μl of reaction buffer [100 mM HEPES (pH 7.4), 5 mM dithiothreitol, 5 mM EGTA, 0.04% NP-40, and 20% glycerol] and incubated with 100 μM fluorescent substrate N-benzyloxycarbonyl-DEVD-7-amino-4-trifluoromethylcoumarin (Enzyme System Products, Livermore, CA, U.S.A.) at 37°C for 1 h. The fluorescent signals were determined with a spectrofluorometer (Bio-Tek Instruments, Winooski, VT, U.S.A.) at an excitation wavelength of 360 nm and an emission wavelength of 530 nm. Protease activity was expressed as the amount of cleaved substrate (7-amino-4-trifluoromethylcoumarin) per microgram of protein.

BrdU incorporation and cell cycle analysis

Cells were incubated with 20 μM BrdU (Sigma) for 4 h at 37°C in a saturated humidity atmosphere containing 95% air and 5% CO2. Subsequently, cells were fixed in 2% paraformaldehyde for 30 min and subjected to immunodetection using a fluorescein-conjugated anti-BrdU antibody from the In Situ Cell Proliferation Kit FLUOS (Boehringer Mannheim) following the manufacturer's instructions. To analyze the cell cycle distribution, cells were counterstained with 50 μg/ml propidium iodide and 20 μg/ml RNase A on ice for 15 min. FACS analysis was performed with an EPICS XL flow cytometer (Coulter).

Immunocytofluorescence of heavy-molecular-weight neurofilament (NF-H), microtubule-associated protein 2 (MAP2), and tau

Cells were fixed in 4% paraformaldehyde in PBS at room temperature for 20 min, washed three times with PBS containing 0.1% Triton X-100, and blocked for 40 min with PBS containing 1% bovine serum albumin and 0.1% Triton X-100. They were then incubated overnight with monoclonal α-NF-H (1:40; 200K; Sigma), α-MAP2 (1:50; Sigma), or α-Tau-1 (1:50; Boehringer Mannheim) antibodies diluted in PBS containing 1% bovine serum albumin and 0.1% Triton X-100, washed three times in PBS with 0.1% Triton X-100, and incubated at room temperature for 40 min with fluorescein isothiocyanate-conjugated donkey anti-mouse monoclonal antibody (1:250 in PBS with 0.1% Triton X-100; Jackson Immunoresearch, West Grove, PA, U.S.A.). Finally, they were washed three times with PBS and mounted with Mowiol (Calbiochem-Novabiochem, La Jolla, CA, U.S.A.) plus 0.1% 1,4-diazabicyclo[2.2.2]octane (Sigma) as an antifading agent. The observation was carried out in an epifluorescence microscope (see above).

Measurement of neurotransmitter content and noradrenaline release

The neurotransmitter content of the untreated cells and cells treated with RA for 5 days or RA for 5 days plus BDNF for 7 days was assayed by quantitative HPLC according to the procedure of Calvo et al. (1995).

For noradrenaline release assays, cells were seeded in 24-well dishes and either left untreated or treated with RA for 5 days or with RA for 5 days followed by BDNF for 7 days. Cells were then rinsed twice with HEPES-buffered saline (135 mM NaCl, 5 mM KCl, 0.6 mM MgSO4, 2.5 mM CaCl2, 10 mM HEPES, and 6 mM glucose) containing 20 μg/ml pargyline and 20 μg/ml ascorbic acid (pH 7.4) and loaded with 50 nM [3H]noradrenaline for 1 h at 37°C. After washing off the excess of noradrenaline, cells were treated with 10 nM 12-O-tetradecanoylphorbol 13-acetate (TPA) for 8 min immediately before evoking the release with 1 mM carbachol. Unreleased noradrenaline was extracted with 0.4 M perchloric acid, and the release was calculated as a percentage of the total amount of radioactivity present before evoking the release. Data are mean ± SEM values of three independent experiments (n = 4 for each condition), for which basal release has been subtracted.


Effects of long-term RA treatment in SH-SY5Y cells

The SH-SY5Y cell line is a third successive subclone of the SK-N-SH line, originally established from a bone marrow biopsy of a neuroblastoma patient (Biedler et al., 1973). The SK-N-SH parental line comprises at least two morphologically and biochemically distinct phenotypes: neuroblastic (N-type) and substrate adherent (S-type), which can undergo transdifferentiation (Ross et al., 1983). Although derived from a neuroblastic subclone, the SH-SY5Y line retains a low proportion of S-type cells (Fig. 1A, arrowheads). In agreement with previous reports (Pahlman et al., 1984), when cultures were treated with 10 μM RA for 5 days in complete medium (DMEM plus 15% fetal calf serum), a considerable proportion of neuroblastic (N-type) cells differentiated to a more neuronal phenotype by extending neuritic processes (Fig. 1B), whereas S-type cells did not undergo apparent morphological changes. Moreover, after ∼ 10 days of culture in the presence of RA, the percentage of S-type cells progressively increased (Fig. 1C), overgrowing the cultures in the subsequent days. To quantify the effects of RA in the cell proliferation rate, a growth curve was generated in the presence or absence of RA (Fig. 1D). In agreement with previous reports (Pahlman et al., 1984), RA inhibited the growth rate of SH-SY5Y cells during the first 8-10 days of treatment. However, longer periods of incubation with RA progressively increased the total number of cells owing to the accumulation of S-type cells. Thus, although short RA treatments were able to inhibit cell proliferation and induced a modest degree of neuronal differentiation, long-term treatments with this drug did not yield a homogeneously neuronal differentiated population. Actually, long-term RA treatments seemed to unbalance the proportion between the N-type and S-type phenotypes toward the S-type one.

Figure 1.

Effects of long-term RA treatment in SH-SY5Y cells shown in phase-contrast micrographs of SH-SY5Y cells growing in complete medium (DMEM with 15% fetal calf serum) without additives (A) and in complete medium containing 10 μM RA for (B) 5 or (C) 10 days. Note that untreated cells are mainly neuroblastic (N-type, arrows), but occasionally some substrate-adherent cells appear (S-type, arrowheads). Treatment with RA resulted initially in neurite outgrowth, but as time in culture increased, S-type cells progressively overgrew the cultures. Bar = 50 μm. D: Growth rate of SH-SY5Y cells in the presence or absence of RA. The cell number was calculated by detaching cells at the indicated time points and counting them in a hemocytometer. Data are mean ± SEM (bars) values of three independent experiments.

FIG. 1.

Effects of BDNF and serum in RA-pretreated cells

The side effects described above make long-term RA treatment of SH-SY5Y unsuitable for the acquisition of homogeneous populations of differentiated cells with neuronal characteristics. However, it has been reported that RA induces the expression of TrkB in SH-SY5Y cells, making them responsive to BDNF (Kaplan et al., 1993). Moreover, it has also been shown that BDNF enhances the differentiating effects of RA (Arcangeli et al., 1999). We first tested whether in our experimental conditions RA was able to induce the expression of TrkB. As shown in Fig. 2A, after 3 days of RA exposure, the levels of functional TrkB present on the cell surface, i.e., the amount of Trk capable of undergoing BDNF-induced autophosphorylation, were significantly increased with respect to untreated cells. The levels of TrkB were maximal at day 5 of treatment and slightly decreased in the subsequent days. Accordingly, we decided to remove RA from the culture medium at day 5 of treatment and then switch the cultures to BDNF in complete medium.

Figure 2.

Effects of serum and BDNF in RA-pretreated SH-SY5Y cells. A: Expression of TrkB after exposure of SH-SY5Y cells to RA. Cells were treated with 10 μM RA for the indicated times, stimulated with BDNF for 5 min, and immunoprecipitated (IP) with an anti-pan-Trk antiserum (203). The levels of active TrkB were detected by western blot (WB) with anti-phosphotyrosine antibody (top panel). The membrane was reincubated with antiserum 203 to visualize the amount of total Trk (bottom panel). Note that the highest levels of TrkB were reached at day 5 of RA. The PC12 lane corresponds to PC12 cells stimulated with NGF. B: After 5 days of treatment with RA, this agent was washed out from the medium, and BDNF (50 ng/ml) was added to analyze the biological effects elicited by TrkB. After 24 h of exposure to BDNF, cells showed a somewhat higher level of differentiation with more abundant and branched neurites. C: However, after 3 additional days under these conditions, S-type cells (arrowheads) overgrew the culture, and N-type cells (arrows) tend to aggregate. D: This could be circumvented by removing the serum from the culture medium at the time of addition of BDNF. In these conditions, virtually all the cells displayed a fully mature neuronal morphology, with round, phase-bright bodies and a robust network of neuritic processes connecting them. E: To assess the stability of the cultures treated with this protocol of differentiation, cells were pretreated 5 days with 10 μM RA and then incubated for 7 additional days in the presence of BDNF in serum-free medium. At this time point, there was no evidence of cellular degeneration, and S-type cells were not detectable. Moreover, cells remained in a differentiated state and showed a tendency to aggregate. F: In contrast, when in parallel cultures BDNF was not added to the medium after the removal of RA and serum, most cells retracted their neurites and died. Bar = 50 μm.

FIG. 2.

After the first day of incubation with BDNF, cells scattered over the culture plate and displayed neuritic processes, with most of them displaying growth cones typical of primary cultures (Fig. 2B). However, with increasing culture time, S-type cells gradually appeared, forming a continuous layer below the N-type cells, which remained as small aggregates of rounded, refringent cells (Fig. 2C). In contrast, in parallel cultures where serum was removed at the time of addition of BDNF, a homogeneous population of cells with neuronal morphology was obtained, with a very low amount of S-type cells (Fig. 2D). In these cultures, cells showed rounded, phase-bright bodies and a profuse neuritic arborization forming extensive networks over the culture dish surface. Under these conditions, cultures were stable for at least 3 weeks, showing neither signs of cellular degeneration nor reversion of the neuronal phenotype. Under these conditions S-type cells were almost undetectable.

Dependence on BDNF for survival: neurotrophic deprivation induces apoptotic cell death

In a previous work, we reported that BDNF promoted short-term survival of RA-pretreated SH-SY5Y cells (Encinas et al., 1999). We wanted to characterize further whether this dependence on BDNF for survival persisted in long-term BDNF-treated cultures. Figure 2E shows a phase-contrast micrograph of cells treated for 5 days with RA and 7 additional days in the presence of BDNF in serum-free medium. Cells appeared healthy and showed a high degree of differentiation. In contrast, when parallel cultures were left in serum-free medium alone for the same interval, only a few cells with retracted neurites persisted (Fig. 2F). When cultures were maintained for 7 days in the presence of BDNF and then deprived for 7 additional days, the vast majority of cells degenerated and died (data not shown). Thus, at least for the intervals examined, BDNF seemed to be necessary for the survival of these cells.

Because in primary cultures removal of trophic agents leads to apoptotic cell death, we were interested in analyzing the characteristics of the cell death after BDNF deprivation. DNA fragmentation, a typical feature of apoptotic cell death, was assessed by means of the TUNEL reaction. Cultures treated for 5 days with RA were switched to either BDNF-containing or serum-free medium for 24 h. When BDNF was present, positive cells only occasionally appeared (Fig. 3C and D), whereas in cultures deprived of BDNF for the same interval, a significant number of labeled cells was found (Fig. 3A and B). It is noteworthy that chromatin condensation, another characteristic of apoptosis, was only found in cells that had been deprived of BDNF (Fig. 3B, arrowheads). The time course of cell death was quantified by scoring positive nuclei with respect to total nuclei stained with Hoechst 33258, and a representative graph is shown in Fig. 3E. Thus, morphological criteria indicated that the cell death followed by BDNF removal was apoptotic.

Figure 3.

BDNF-deprived SH-SY5Y cells are TUNEL-positive and have increased DEVD-directed caspase activity. SH-SY5Y cells treated for 5 days with RA and then left in serum-free medium alone (A) or supplemented with BDNF (C) for an additional 24 h were examined by TUNEL staining. B and D: Corresponding micrographs show the same microscopic fields with nuclei stained with the DNA dye Hoechst 33258. Note that in some TUNEL-positive cells (arrowheads) the chromatin is condensed in a typical apoptotic pattern. E: Time course of the cell death triggered by BDNF withdrawal. Cells were treated for 5 days with RA and then left for the indicated times in the absence (-) or presence (+) of BDNF. The percentage of TUNEL-positive cells was scored by counting at least 500 cells in each condition. F: DEVD-specific caspases are activated after BDNF removal. Cells were treated as in E, and the amount of 7-amino-4-trifluoromethylcoumarin (AFC) produced was measured in a fluorometric assay. Data are mean ± SEM (bars) values of two independent experiments. *p < 0.05, **p < 0.01, by unpaired, two-tailed Student's t test with respect to BDNF treatment. Bar = 50 μm.

FIG. 3.

We also used biochemical approaches such as analysis of caspase activity to define this cell death as apoptotic. Caspases are cysteine proteases that cleave specific proteins in aspartic residues. Activation of these proteases accompanies the apoptotic process (for review, see Cryns and Yuan, 1998). Using a fluorogenic assay we observed a progressive increase in the caspase activity when cells treated for 5 days with RA were switched to serum-free medium. After 6 h of starvation, a detectable level of caspase activity above the background, i.e., the reaction mixture incubated without cell lysates, was measured, which increased until 24 h. When parallel cultures treated for 24 h in the presence of BDNF were subjected to such analysis, the amount of caspase activity was found to be about half of the one found in deprived cells (Fig. 3F). Taken together, these data strongly suggest that the cell death followed by BDNF withdrawal is apoptotic.

RA-BDNF-treated cells are arrested in G1

To analyze the cell cycle distribution during RA-BDNF treatment, we performed BrdU labeling experiments at different time points of the differentiation protocol. Cells were pulse-labeled for 4 h with BrdU before collecting and processing them for analysis with anti-BrdU fluorescein isothiocyanate-conjugated antibodies by means of flow cytometry. Propidium iodide staining was used to examine cell cycle distribution. Untreated cells displayed a typical profile of an asynchronously growing population, with ∼30% of the cells traversing S-phase. Accordingly, elevated numbers of cells were BrdU-positive (Fig. 4A and B). After 5 days of RA treatment, there was a significant depletion of cells in the S and G2/M phases of the cell cycle, concomitant with a decrease in the number of BrdU-positive cells (Fig. 4A and B). This agrees with the observed stabilization in total cell numbers after RA treatment depicted in Fig. 1D. However, after this period, a small, although significant, population of cells was still capable of incorporating BrdU. When cells were examined after 3 days of exposure to BDNF, the percentage of BrdU-positive cells was <5%, and ∼90% of cells were arrested in G1 (Fig. 4A and B). In conclusion, the RA-BDNF protocol yielded, as early as after 3 days in the presence of BDNF, a nearly pure population of cells arrested in the G1 phase of the cell cycle.

Withdrawal of BDNF induces S-phase entry and apoptotic cell death

When BDNF was withdrawn from the cultures, a subG1 peak typical of apoptotic cells was observed (Fig. 4A and B). A more detailed analysis showed that virtually all of these cells were present in the nonadherent fraction of cells, i.e., the cells that have died and detached from the dishes. It is interesting that BDNF removal prompted an attempt to reenter S-phase, as judged by the increase in the number of BrdU-positive cells.

We finally wanted to analyze the state of phosphorylation of pRB, by means of a gel mobility shift (Fig. 4C). In untreated cells, the most abundant pRB species was hyperphosphorylated, as expected from active cycling cells. After treatments with RA and RA-BDNF, pRB was found to be progressively hypophosphorylated, in agreement with the observed withdrawal of these cells from the cell cycle. Consistent with the attempt of cells to reenter S-phase after BDNF removal, pRB showed a tendency to hyperphosphorylate, accompanied by the appearance of a novel, lower-molecular-weight pRB species, which is likely to be a caspase-cleaved fragment of this protein (Fig. 4C).

RA-BDNF-treated cells express several neuronal markers

The morphology of RA-BDNF-treated cells examined by phase-contrast microscopy showed that these differentiated cells resembled primary neurons. To analyze whether this phenomenon was accompanied by the expression of neuron-specific markers, we checked for the presence of several proteins by either immunocytofluorescence or western blot throughout the period of differentiation. Neurofilament triplet expression changed during the acquisition of fully neuronal phenotype. NF-L was predominantly expressed in untreated or RA-treated cells, whereas its expression decreased in cells treated with BDNF for 7 days as assessed by western blot (Fig. 5). NF-M was expressed in naive cells, and its expression considerably increased after 5 days of RA. After 7 days with BDNF, however, we could not detect this protein by western blot (Fig. 5).

Figure 5.

RA-BDNF-differentiated SH-SY5Y express neuronal markers as shown by western blot analysis of the levels of NF-L, NF-M, NSE, GFAP, and GAP-43 in RA-BDNF-differentiated cells. Top panel: Cells were left untreated (lanes 1) or were treated with RA for 5 days (lanes 2) or with RA for 5 days followed by BDNF for 7 days (lanes 3). Note that cells were positive for NSE and negative for GFAP in all experimental conditions. Bottom panel: Cells were treated for the indicated days with RA (lanes designated R) or for 5 days with RA followed by the indicated days with BDNF (lanes designated B). Lane Br, mouse brain homogenates used as positive control.

FIG. 5.

Immunocytofluorescence revealed that NF-H was present in naive cells, with its localization being restricted to nuclear or perinuclear zones (Fig. 6). The same pattern of expression plus some labeling at the neurite extensions was observed in RA-treated cells, whereas cells treated for 7 additional days with BDNF showed a strong labeling at the neuritic processes, and besides perinuclear zones, cell bodies were also stained (Fig. 6). In BDNF-treated cells, MAP2 immunoreactivity was localized in the cytoplasm, with strong nuclear exclusion, and in short processes emanating from the cell body (Fig. 6). In naive and RA-treated cells, the pattern of distribution was similar except for the neuritic processes, with the majority of them being negative for this protein. In contrast, in RA-BDNF-treated cells, tau immunoreactivity was restricted to long, branched processes and was nearly undetectable in cell soma, whereas it was undetectable in naive cells. RA-treated cells showed a weak cytoplasmatic labeling that did not include neuritic processes. The pattern of labeling found in RA-BDNF-treated cells resembled the one found in cultures of primary neurons, in which MAP2 preferentially labels the somatodendritic domain whereas tau labels the axonal domain of neurons (Goedert et al., 1991). This observation suggests that RA-BDNF-differentiated cells are polarized.

Figure 6.

RA-BDNF-differentiated SH-SY5Y express neuronal markers as shown by immunocytofluorescence for NF-H, MAP2, and tau of SH-SY5Y cells under different culture conditions: naive, 5 days of RA, or 5 days of RA followed by 7 days of BDNF. The expression of all markers increased along with the RA and BDNF treatment. NF-H arose from perinuclear areas (naive) and progressively extended to the rest of the cell body and neurites (5 days of RA and 7 days of BDNF), whereas MAP2 and tau were excluded from the nuclei, and at 7 days of BDNF they showed a complementary pattern of staining at the neurites with MAP2 staining the proximal portions of neurites (respect to the cell body) and tau the distal ones. Bar = 50 μm.

FIG. 6.

The expression of NSE and GFAP was analyzed by western blot to confirm the neuronal lineage of these cells. As expected, NSE was present in all experimental conditions, whereas GFAP was never detectable (Fig. 5). Finally, analysis of the expression of GAP-43 revealed that this protein was transiently accumulated at day 5 of RA treatment and especially after the first day of BDNF exposure. These high levels returned to control values during the subsequent days of BDNF treatment (Fig. 5). This temporal pattern of expression coincides with the bulk of neurite outgrowth, in agreement with the proposed role of this protein in neurite elongation.

RA-BDNF-treated cells exhibit carbachol-evoked noradrenaline release

TPA-differentiated and naive SH-SY5Y cells have been reported to exhibit carbachol-evoked noradrenaline release (Scott et al., 1986; Murphy et al., 1991). To begin to characterize the neurotransmitter phenotype of RA-BDNF-treated cells, we performed noradrenaline release assays in untreated, RA-treated, or RA-BDNF-treated cells. A robust noradrenaline release was observed in each of the three conditions after carbachol stimulation, being 20.8 ± 0.88% in naive cells (values indicate the mean ± SEM percentage of noradrenaline released with respect to the total noradrenaline loaded), 17.2 ± 1.16% in RA-treated cells, and 8.9 ± 0.7% in RA-BDNF-differentiated cells. On the other hand, RA treatment has been shown to induce a slightly higher choline acetyltransferase activity in these cells, blocking TPA-induced noradrenaline production (reviewed by Pahlman et al., 1995). However, we failed to detect a significant choline acetyltransferase activity in any of the three experimental treatments (data not shown).

Finally, we measured the total content of several amino acid transmitters through the differentiation protocol by means of quantitative HPLC (Calvo et al., 1995). We were unable to detect significant changes in the total content of GABA, glycine, or taurine. However, glutamate levels decreased after 5 days of RA treatment (168.65 ± 24.6 ng in naive cells vs. 63.6 ± 16.3 ng in RA-treated cells) and were restored after RA-BDNF treatment (117.9 ± 13.9 ng). It should be noted that glutamate was by far the most abundant amino acid neurotransmitter present in any of the experimental conditions.


In this report we present a protocol of differentiation based on the sequential exposure of SH-SY5Y cells to RA and BDNF. This protocol yields homogeneous populations of fully neuronal differentiated cells. These cells are withdrawn from the cell cycle, express many of the typical neuronal markers, exhibit carbachol-evoked noradrenaline release, and are dependent on neurotrophic support for survival and differentiation. These characteristics make the cells obtained with this protocol very similar to primary neurons.

The effects of RA in SH-SY5Y are well documented. These include an attenuation of proliferation rate, extension of neuritic processes, and development of a slightly enhanced choline acetyltransferase activity (for review, see Pahlman et al., 1995). However, these studies generally do not consist of long-term analysis of the phenotype induced by RA treatment. In our hands, extended intervals in the presence of this agent favored the appearance of S-type cells, making further studies difficult to perform. This observation has been reported also by other authors in SH-SY5Y (Jensen, 1987; Hill and Robertson, 1997; Arcangeli et al., 1999) and other (Matsushima and Bogenmann, 1992) neuroblastoma cell lines. The phenomenon of transdifferentiation between N- and S-types seems to be common to most neuroblastoma cell lines, including those considered neuroblastic, like SH-SY5Y (Ross et al., 1983). This phenomenon may reflect the multipotential nature of neural crest cells from which neuroblastoma are presumed to arise. Therefore, it is considered that the prevalence of a given phenotype in a continuous neuroblastoma cell line is a consequence of slower rates of interconversion rather than loss of the potential to generate the other phenotype (Sadee et al., 1987). Moreover, in the case of neuroblastic subclones, during the processes of splitting and harvesting, S-type cells are selected against, owing to their tendency to remain attached to the culture plate (Jensen, 1987). Thus, the emergence of S-type cells after long-term treatments with RA could be primarily attributed to the spontaneous tendency of these cells to transdifferentiate when they are cultured for extended times without splitting them. The fact that S-type cells seem to be resistant to the growth inhibitory effects of RA would then contribute to the rapid expansion of these cells observed when SH-SY5Y are cultured in the presence of this agent. The possibility that RA may directly exert a dual differentiating effect over the two observed phenotypes cannot be excluded because this agent has been shown to promote differentiation of both adrenergic cells and melanocytes in neural crest cultures (Dupin and Ledouarin, 1995). However, the fact that treatment of SH-SY5Y cells with agents that promote growth inhibition of N-type cells such as NGF results in the appearance of S-type cells (Jensen, 1987) indicates that at least part of the transdifferentiation phenomenon is not triggered directly by RA.

Removal of RA after 5 days of treatment, followed by addition of BDNF in serum-containing medium, induced a more neuronal phenotype but did not circumvent the appearance of S-type cells as time in culture increased. In contrast, if serum was removed at this step, only residual amounts of S-type cells were present. A possible explanation for this behavior is that S-type cells, which do not express neuron-specific markers (Sadee et al., 1987; Ciccarone et al., 1989), fail to induce TrkB in response to RA and therefore die in the absence of trophic support. In this regard, it has been described that SH-EP, an S-type subclone of SK-N-SH, cannot survive in the absence of serum (see Leventhal et al., 1995).

The effects of BDNF on survival and differentiation indicate that TrkB receptors induced by RA are biologically active. In a previous report, we have shown that some of the relevant kinases involved in the signal transduction pathways mediated by BDNF become activated after addition of this neurotrophin to RA-pretreated SH-SY5Y cells. Moreover, the selective blockade of these pathways reverses the neurite-promoting and survival effects triggered by BDNF (Encinas et al., 1999). Previous studies have shown that RA promotes NGF survival responsiveness in cultured chicken sympathetic neuroblasts (Rodriguez-Tebar and Rohrer, 1991) by up-regulating the levels of TrkA (VonHolst et al., 1995, 1997). In contrast, RA inhibited the developmental increase in level of TrkA mRNA and the decrease in level of TrkC mRNA in mouse sympathetic neuroblasts (Wyatt et al., 1999). A recent work indicates that hippocampus-derived stem cell clones up-regulate expression of neurotrophin receptors when exposed to RA. Moreover, the sequential addition of RA and BDNF or NT-3 to these cells led to a significant increase in the number of mature neurons generated (Takahashi et al., 1999). Finally, RA and neurotrophins have been found to promote the survival and differentiation of pluripotent neural crest cells to a neuronal lineage (Sieber-Blum, 1991; Henion and Weston, 1994). These and other observations suggest that RA may be involved in the maturation of neural crest cells along a neuronal fate and that collaborative effects between RA and neurotrophins may be needed for the acquisition of a fully developed neuronal phenotype.

For the times analyzed, the viability of cells seemed to be conditioned by the continuous presence of BDNF in the culture medium. The removal of the neurotrophin triggered a cell death that has many of the features of apoptosis, including DNA fragmentation (detected by TUNEL and flow cytometry), chromatin condensation, and activation of caspases. This antiapoptotic role of BDNF has been demonstrated in several types of primary neurons (see, for example, Lewin and Barde, 1996). It should be noted that the rate of cell death followed by BDNF withdrawal was strongly dependent on the cellular density, being slower in high-density cultures (data not shown). This could be explained in terms of an autocrine survival loop of insulin-like growth factor-II operating in SH-SY5Y cells switched to serum-free medium (Martin and Feldman, 1993). We have found that the optimal densities that yielded healthy cultures in the presence of BDNF and a relatively rapid cell death after starvation were ∼104 cells/cm2 (data not shown), which are sixfold lower than the threshold density required for autocrine insulin-like growth factor-II-mediated growth (Martin and Feldman, 1993).

Another consistent feature of cells generated by sequential exposure to RA and BDNF is its progressive withdrawal from the cell cycle. It is generally accepted that terminal differentiation of neuroblasts occurs once they have been arrested in G0 (reviewed by Ross, 1996). It is interesting that removal of BDNF prompted an attempt to reenter S-phase, as judged by BrdU uptake and hyperphosphorylation of pRB. Unscheduled S-phase entry has been associated with apoptotic cell death in several systems and is thought to be a consequence of deregulated E2F-1 release from pRB/E2F complexes (reviewed by Kasten and Giordano, 1998; Macleod, 1999). In RB-null mice, aberrant S-phase entry and concomitant apoptotic cell death are observed in differentiating cells, which should be exiting the cell cycle to become postmitotic (Lee et al., 1994; Morgenbesser et al., 1994). Moreover, ectopic expression of viral genes that interferes with E2F-1 binding to pRB also results in apoptotic cell death (Howes et al., 1994; Pan and Griep, 1994, 1995; Symonds et al., 1994). It is interesting that the massive apoptotic cell death observed in RB—/— mice is abrogated by targeted disruption of E2F-1 (Tsai et al., 1998). Conversely, overexpression of E2F-1 causes apoptosis in mouse fibroblast cell lines after the cells enter S-phase (Qin et al., 1994; Shan and Lee, 1994; Wu and Levine, 1994).

In conclusion, the model presented herein accomplishes many of the characteristics presented by primary cultures of neurons. In particular, its trophic dependence toward BDNF makes this model system a suitable tool to approach the phenomenon of PCD and, more specifically, the relationship between this phenomenon and the cell cycle. Moreover, this model offers a convenient system to explore the therapeutic potential of neurotrophins in neurodegenerative diseases. Some differentiation procedures exist that yield fully mature human neuron-like cells. However, their trophic dependencies are not well established as most of them require serum factors for survival. Furthermore, many of the above models are time-consuming and require the use of antimitotic drugs that have been shown to be toxic for some primary cultures of neurons (Dessi et al., 1995; Sanz-Rodriguez et al., 1997; Anderson et al., 1999). In our model system, the use of antimitotics can be avoided, although SH-SY5Y cells have the ability to give rise to many neural crest derivatives. That the sequential use of RA and BDNF commits the vast majority of these cells to a neuronal phenotype suggests that this treatment could be mimicking some of the developmental signals that commit pluripotential neural crest cells to a neuronal fate.