Pigment epithelium-derived factor (PEDF) in neuroblastoma: a multifunctional mediator of Schwann cell antitumor activity

Although neuroblastomas are the most common extracranial tumors found in childhood, therapeutic advances over the past two decades have not substantially altered the dismal 5-20% five-year survival rate for the nearly 40% of all neuroblastoma patients who present beyond one year of age with evidence of metastatic disease (Albano et al., 1997; Brodeur and Castleberry, 1997). Even predicting outcome in a child diagnosed with neuroblastoma is difficult due to the marked biological heterogeneity of the tumors and to clinical outcomes that can range from spontaneous regression to rapidly progressive metastatic disease (Brodeur et al., 1997; Van Noesel et al., 1997; Castleberry, 1997; Katzenstein et al., 1998; Linet et al., 1999; Haase et al., 1999).

In vivo neuroblastomas are composed of a highly variable admixture of cells including malignant neuroblasts derived from the embryonic neural crest destined for the adrenal gland (Israel, 1993), ganglion cells, vascular endothelial cells and Schwann cells. Better clinical outcome has been associated with an increased density of differentiated neuronal cells and Schwann cells (Brodeur, 1996) in the stromal compartment, whereas poor outcome correlates with N-myc amplification and with increased microvascular density, a hallmark of angiogenesis (Meitar et al., 1996; Eggert et al., 2000). Like most tumors, neuroblastomas are dependent on angiogenesis (Folkman, 1990a) and xenografts are sensitive to a variety of anti-angiogenic drugs (Wassberg, 1999; Rowe et al., 2000; Shusterman et al., 2000).

Although the Schwann cell component was initially thought to arise from neuroblastoma tumor cells (Sidell et al., 1986; Tsokos et al., 1987), subsequent morphological and cytogenetic studies suggest that they are more likely to be a normal, reactive cell population that infiltrates the tumor (Ambros and Ambros, 1995; Ambros et al., 1996; Katsetos et al., 1994). Like most normal cells, Schwann cells are not angiogenic (Sheela et al., 1990; Huang et al., 2000). Their secretions induce neuronal differentiation in several neuroblastoma tumor cell lines (Kwiatkowski et al., 1998). These Schwann-cell-derived soluble factors have been referred to as ‘anti-neuroblastoma’ agents (Brodeur, 1996), although the active ingredients have not yet been identified.

The cellular heterogeneity of neuroblastomas is enhanced by the ability of tumor cells themselves to switch their phenotype. Pure cultures of tumor cells can change from less differentiated, more malignant N-type neuroblasts to less malignant, more differentiated S-type cells either spontaneously or in response to a variety of stimuli. Numerous exogenous agents have been found to induce differentiation of neuroblastoma tumors including the thymidine analogue 5-bromo-2′-deoxyuridine (Tsunamoto et al., 1988) and agents known to increase intracellular calcium (Wu et al., 1998), but the identification of endogenous differentiating factors has been elusive.

Recently, pigment epithelium-derived factor (PEDF), a 50 kDa glycoprotein originally isolated from retinal pigment epithelial cells and recognized for its neurotrophic activity on cells derived from the neural crest (Tombran-Tink et al., 1991; Becerra et al., 1993; Steele et al., 1993; Becerra, 1997), was shown to be a potent inhibitor of angiogenesis (Dawson et al., 1999). It is a non-inhibitory serpin that can promote the survival of cerebellar granule cells (Taniwaki et al., 1995), induce differentiation in retinoblastoma tumor cells in vitro (Seigal et al., 1994) and protect retinal neurons from apoptotic death (Cao et al., 1999). Additionally, a PEDF-derived peptide promotes survival and neurite outgrowth in spinal motor neurons, even after axotomy in animal models (Houenou et al., 1999). In contrast to its anti-apoptotic effects on neural cells, PEDF acts in the opposite way on the endothelial cells that are forming new vessels, inducing their apoptosis and thereby inhibiting neovascularization (Stellmach et al., 2001).

Here PEDF is shown to be produced by the differentiated elements within neuroblastomas and to be the major inhibitor of angiogenesis in media conditioned by Schwann cells or by differentiated S-type neuroblastoma cells. PEDF also promoted the survival of Schwann cells and the differentiation of tumor cells. Taken together, data presented below demonstrate that PEDF has the potential to act in a multifunctional antitumor manner in neuroblastomas, targeting both the tumor and stromal components and setting up autocrine and paracrine loops with great potential for stabilizing or even regressing these aggressive tumors.

Histological sections of human neuroblastoma tumors (n=15) showing varying degrees of ganglionic differentiation were immunostained using an antibody raised against PEDF-derived peptide (Dawson et al., 1999), and sections of SK-N-BE(2) induced tumors in mice that were injected with PBS vehicle (n=4) or PEDF (n=6) were stained with an antibody that recognized neurofilament (DAKO). Sections were dried at 42°C for 24 hours and then deparaffinized, incubated at room temperature with the primary antibody, followed by incubation with an avidin-biotin peroxidase conjugated secondary antibody, and counterstained with hematoxylin. Those cells exhibiting a brown color were scored positive for PEDF or neurofilament, respectively.

Schwann cells were cultured from rat sciatic nerves as described (Brockes et al., 1979). Briefly, nerves were removed from 3-day-old rat pups, digested for 2 hours in 0.03% collagenase (Serva) at 37°C, and triturated thoroughly to achieve dissociation. Cells were maintained as monolayers in low glucose DMEM supplemented with 10% FBS, 100 units/ml penicillin and 100 μg/ml streptomycin at 37°C in a humid atmosphere of 10% CO₂/90% air. Contaminating fibroblasts were inhibited by 72 hours treatment with 10 μM cytosine arabinoside. Schwann cells were also cultured from human ganglioneuromas (Casella et al., 1996) and maintained in low glucose DMEM supplemented with 20% FBS and 2 μM forskolin. Purity of Schwann cell cultures were confirmed by staining with antibody raised to S-100 protein (DAKO), and only cultures that contained >95% S-100 positive cells were used in experiments. Two neuroblastoma cell lines, SK-N-BE(2) and SH-SY5Y, were obtained from the American Tissue Type and Culture, and N-type and S-type neuroblastoma cells were a generous gift of Susan Cohn (Northwestern University, Chicago, IL). Cell lines were maintained in DMEM supplemented with 10% FBS (Flow Laboratories, McLean, VA) in 37°C and 5% CO₂.

Cell cultures were grown to 80% confluence, media aspirated, cells rinsed twice with 5 ml of phosphate buffered saline (PBS), and incubated in serum-free DMEM for 4 hours. The rinse media were aspirated, fresh serum-free DMEM added, and cells incubated at 37°C/5% CO₂. After 48 hours the media were collected and subjected to centrifugation (1000 g, 5 minutes) to remove debris, concentrated, and dialyzed against PBS using Millipore Ultrafree centrifugal filters with a 10 kDa cutoff.

Protein samples were loaded into the wells of a 10% SDS-polyacrylamide gel, subjected to electrophoresis under reducing conditions, and blotted onto Hybond-C nitrocellulose membranes (Amersham). Blots were blocked with a solution of 5% nonfat dry milk/Tris-buffered saline (TBS) pH 8.0/1% Tween-20 for 1 hour at room temperature; rinsed twice for 5 minutes in TBS/0.1% Tween-20; exposed to a 1:2000 dilution primary anti-PEDF antibody (Dawson et al., 1999) or preimmune serum in 0.1% nonfat dry milk/TBS/0.1% Tween for 1 hour; rinsed three times as described above; incubated with a 1:5000 dilution of anti-rabbit IgG-HRP (Pierce); rinsed three times in TBS/0.1% Tween-20, 10 minutes per rinse; rinsed once in TBS without Tween-20 for 10 minutes. Signals were developed with Luminol reagent for western blotting (Santa Cruz).

Purified histidine-tagged recombinant human PEDF was prepared from media conditioned by human embryonal kidney cells forced to express and secrete the rPEDF as described (Dawson et al., 1999).

Bovine adrenal capillary endothelial cells (kindly provided by J. Folkman, Children’s Hospital, Boston, MA) were maintained in DMEM containing 10% donor calf serum (Flow Laboratories, McLean, VA) and 100 ug/ml endothelial cell mitogen (Biomedical Technologies, Stoughton, MA) and used between passage 13 and 15. Human microvascular endothelial cells (Clonetics) were maintained in endothelial growth media (Clonetics). Migration was assayed as described (Polverini et al., 1991). DMEM containing 0.1% BSA was used as the negative control, and 10 ng/ml of bFGF or 100 pg/ml VEGF as a positive control. Conditioned media were tested at 20 μg protein/ml, antibodies against PEDF or VEGF were used at a final concentration of 5 μg protein/ml or 10 μg protein/ml, respectively. At a minimum, each sample was tested in triplicate in a single experiment, and each experiment was repeated at least twice. Results were plotted as percentage maximum migration induced by the positive control (bFGF or VEGF) used in the experiment after the background migration of endothelial cells toward the 0.1% BSA in DMEM was subtracted.

Female Fischer 344 rats (Harlan Industries, Indianapolis, IN) weighing 120-140 g were used for the assay. Briefly, Hydron pellets (Interferon Sciences, New Brunswick, NJ) of <5 μl were prepared containing the test media at 200 μg/ml and/or bFGF at 100 ng/ml. Pellets were implanted into the avascular corneas of anesthetized rats 1.0-1.5 mm from the limbus. Neovascularization was assessed 7 days later and a permanent photographic record of the response was obtained by perfusion with colloidal carbon to highlight vessels. During the experiment, all animals maintained their weight and appeared healthy throughout the 7-day period. Animals were treated according to National Institutes of Health guidelines for animal care and use, and protocols were approved by the Animal Care and Use Committee Northwestern University.

Neuroblastoma cells were harvested, and 1 ml aliquots of cell suspensions containing 1.25×10⁴ cells/ml were used to seed each well of 24-well plates. Twenty four hours later, purified recombinant PEDF or Schwann cell conditioned media were added at the indicated concentrations to triplicate wells, and SH-SY5Y cells incubated for an additional 24 hours and SK-N-BE(2) for an additional 4 hours. The percentage of differentiated cells was determined by counting the total number of cells in three non-overlapping 1 mm² areas per well. A cell was scored as differentiated if neurite outgrowths were greater than 50 μm in length. Morphological counts were confirmed by checking for neuronal differentiation, and observing all cells possessing neurites showed immunohistochemical localization of neurofilament (DAKO, Carpinteria, CA).

Rat Schwann cells were suspended at 5×10⁴ cells/ml of DMEM, 10% FCS and 2 μM Forskolin; 1 ml was used to seed each well of a 24 well plate. After 24 hours, rPEDF was added at the indicated concentrations to duplicate wells. After 6 days, cells were harvested and counted.

Neuroblastomas were experimentally induced in athymic (nu/nu) mice by injecting 1×10⁶ SK-N-BE(2) cells subcutaneously at two sites on the hind flanks of each mouse. When tumors grew to approximately 8 mm in diameter, treatment was initiated. A total of 2 ug of purified histidine-tagged PEDF in 100 ul of PBS was injected into three sites/tumor each day for four consecutive days. On the fifth day, the mice were euthanized by an overdose of metaphane or halothane, the needle injection sites marked on the skin with a solvent resistant marker, and the tumors surgically removed. Tumor tissue was sliced, fixed in 10% buffered formalin for at least 24 hours, embedded in paraffin, and four μm sections were placed on poly-L-lysine-coated slides for histological examination and immunohistochemical studies. Animals were treated according to National Institutes of Health guidelines for animal care and use, and protocols were approved by the Animal Care and Use Committee Northwestern University.

To determine whether PEDF was expressed in neuroblastomas, sections from 15 different tumors (neuroblastomas with focal ganglionic differentiation, n=5; ganglioneuroblastomas, n=5; mature ganglioneuroma, n=5) were stained with an anti-PEDF antibody. The least differentiated tumors, which are composed predominantly of neuroblasts with a high nuclear to cytoplasmic ratio and numerous small caliber vessels, showed minimal staining for PEDF in both tumor cells and endothelial cells (Fig. 1A). Moderate staining was noted in the differentiating ganglion cells that could be recognized by their peripherally located multilobated nuclei and more abundant cytoplasm (Fig. 1B). The ganglioneuroblastomas and ganglioneuromas showed the most intense positivity for PEDF where it was present in both mature ganglion cells and in Schwann cells (Fig. 1C).

Immunoblot analysis confirmed that Schwann cells derived from human ganglioneuromas and rat sciatic nerves elaborated an approximately 50 kDa protein that was detected by antibodies raised against PEDF-derived peptides (Fig. 2). No bands were detected when blots were probed with preimmune sera. In in vitro angiogenesis assays, Schwann cell conditioned media (CM) were inhibitory as they suppressed the induction of endothelial cell migration by the well characterized inducer of angiogenesis, bFGF, when it was added to the media (Fig. 3). We tested whether PEDF was responsible for this anti-angiogenic activity as PEDF is a potent inhibitor of angiogenesis, capable of inhibiting the activity of multiple purified angiogenic inducers including bFGF and VEGF (Dawson et al., 1999), and may act by triggering apoptosis in endothelial cells activated by inducers of angiogenesis (Stellmach et al., 2001). When an antibody that neutralizes the anti-angiogenic activity of PEDF was added to the Schwann cell CM, the inhibitory activity was abolished and underlying inducing activity was revealed. This inducing activity was probably not due to VEGF as this cytokine was below the level of detection in Schwann cell conditioned media measured by ELISA (see below).

Similar findings were observed when media conditioned by spontaneously differentiated neuroblastoma S cells were tested (Fig. 3). The S-type cell media blocked approximately 80% of migration of endothelial cells elicited by the potent inducer of angiogenesis, VEGF. Addition of a neutralizing antibody to VEGF did not significantly alter endothelial cell migration from that observed when S-type media were tested alone. In vivo confirmation of these results was obtained using the neovascularization of the rat cornea assay. Since PEDF is an endogenous inhibitor of angiogenesis in the cornea, to avoid a neovascular response due to the incorporation of this antibody into test pellets, PEDF was removed from the CM by immunoprecipitation prior to the media being incorporated into the test pellets (Dawson et al., 1999). Removal of the PEDF from the S cell conditioned media confirmed that PEDF was a key anti-angiogenic component produced by these cells, and revealed inducing activity present in the media (Table 1).

In contrast to the differentiated S-type cells and normal stromal Schwann cells, the undifferentiated N-type neuroblastoma tumor cell conditioned medium was potently angiogenic in vitro and in vivo (Fig. 3; Table 1). This difference in angiogenic activity was due to high levels of VEGF produced by N-type cells. VEGF levels in the conditioned media were quantified by ELISA and N cells secreted 26.2 pg VEGF/ug CM protein, whereas VEGF was not detectable in the media conditioned by S type or Schwann cells (limit of detection was calculated to be 0.73 pg/ug CM protein). In addition, when neutralizing VEGF antibody was added to media conditioned by N-type cells or the SK-N-BE(2) neuroblastoma cells, their ability to induce migration of endothelial cells was relieved (Fig. 3), as was the ability of N-type CM to induce angiogenesis in the cornea (Table 1); no consistent difference could be detected in PEDF levels between N and S cells (data not shown), suggesting that the change in VEGF was driving their switch in angiogenic phenotype.

PEDF has been shown to have neurotrophic properties and protect cells of the nervous system from an apoptotic death (Becerra et al., 1993; Steele et al., 1993; Cao et al., 1999; Houenou et al., 1999; Bilak et al., 1999), and it had similar effects on Schwann cells. Using a method that triggers apoptosis in Schwann cells (Campana et al., 1999), cells were grown on chamber slides to 80% confluence with or without 1 nM PEDF and then serum starved over a 5 day period. Terminal deoxynucleotidyltransferase-mediated UTP end labeling (TUNEL) staining was used to identify apoptotic cells. The PEDF-treated cells had a significantly lower percentage of apoptotic cells when compared with the control cells (27% versus 43%; P<0.05; Fig. 4). Not only did PEDF protect Schwann cells from apoptosis, PEDF also stimulated Schwann cell growth. This effect of PEDF on Schwann cell growth were measured by adding purified recombinant PEDF (rPEDF) to Schwann cell growth media. The rPEDF stimulated Schwann cell growth in a dose-dependent fashion (Table 2), and at 1.0 nM was as effective as the well known potent Schwann cell mitogen, NDF-β (Baek and Kim, 1998; Raabe et al., 1996).

In agreement with previous reports (Kwiatkowski et al., 1998), Schwann cell conditioned media triggered differentiation of cultured neuroblastoma cells, SK-N-BE(2) (Fig. 5B,D) and SH-SY5Y (data not shown). To determine whether PEDF was involved, anti-PEDF antibody was added to the Schwann cell conditioned media and approximately 60% of the neurite outgrowth was blocked, indicating that PEDF was one of the major differentiation factors secreted by normal Schwann cells (Fig. 5C,D). Additionally, nanomolar concentrations of purified rPEDF were capable of inducing neurite outgrowths in a dose-dependent manner for SH-SY5Y (Fig. 5E) and SK-N-BE(2) (data not shown). The effective dose that triggered 50% of the maximum inducible differentiation, ED₅₀, for the SH-SY5Y and SK-N-BE(2) cell lines (0.9 nM and 0.75 nM, respectively) closely approximated that reported for the anti-angiogenic activity of PEDF (Dawson et al., 1999).

Subcutaneous tumors that resulted from injecting SK-N-BE(2) neuroblastoma tumor cells into the flanks of nude mice were allowed to grow to approximately 8 mm in diameter. Tumors were injected at three sites once a day for 4 days with either rPEDF (n=6) or vehicle PBS (n=4), and tumor injection sites were marked with ink. On the fifth day, the animals were euthanized and tumors excised. Gross evaluation of each of the six PEDF-treated tumors revealed clearly discernable pale tan, somewhat firmer regions underlying the injection sites. This was in striking contrast to the hemorrhagic, friable tissue located in the same tumors but more distant to the injection sites. All of the vehicle-treated tumors had a uniformly red, hemorrhagic appearance. Histological examination of the PEDF-treated tumor samples revealed that the grossly noted pale regions corresponded to confluent areas of spindle shaped cells characterized by bland nuclei and more abundant cytoplasm. These cells were quite distinct from the more primitive neuroblasts observed within the tumor distant from the injection site. The vehicle-treated tumors were composed of predominantly undifferentiated neuroblasts with only a few clusters of spindled cells. Immunohistochemical staining using anti-neurofilament antibody showed that the spindled cells along the injection sites in the PEDF-treated tumors were strongly positive for neurofilament (Fig. 6C,D), whereas only a rare single cell was positive in the vehicle-treated tumors (Fig. 6A,B). These data demonstrate that even an abbreviated treatment regimen with PEDF in vivo was able to promote discrete areas of tumor differentiation.

Previous work has described Schwann cell secretions as ‘antineuroblastoma’ agents (Brodeur, 1996; Ambros and Ambros 1995) and an increased density of Schwann cells in neuroblastomas is associated with more favorable prognosis and spontaneous maturation (Ambros et al., 1996; Shimada et al., 1985). Data presented above suggests that PEDF is a key antitumor component of these Schwann cell secretions and could be the major reason why Schwann cells are associated with reduced malignancy. PEDF was produced by Schwann cells in vitro and in vivo and was capable of three different antitumor activities: induction of tumor cell differentiation, inhibition of angiogenesis, and augmentation of the Schwann and ganglion cells that in turn produce more PEDF. The autocrine and paracrine interactions investigated here are summarized in Fig. 7. The paracrine interactions within these complex tumors are certainly not limited to those proposed here. For example, endothelial cells secrete a number of factors that may influence angiogenesis and could also impact on the phenotypes of a variety of cells within the tumor (Folkman, 1990b; Rak et al., 1996).

PEDF was responsible for most of the well documented ability of Schwann cell secretions to induce the differentiation of cultured neuroblastoma tumor cells (Brodeur, 1996; Kwiatkowski et al., 1998). The observations that low doses of recombinant PEDF induced tumor cell differentiation in vivo and that the amount of immunoreactive PEDF in tumor tissue correlated with tumor cell differentiation suggest that PEDF is an active differentiation factor in vivo. The fact that about 25% of differentiation inducing ability of Schwann cell conditioned media was insensitive to anti-PEDF antibodies indicates that additional Schwann cell-derived differentiating agents remain to be discovered.

Differentiated tumor cells would be expected to be less aggressive in vivo. In vitro they were also less angiogenic. As undifferentiated tumor cells produced high levels of VEGF and differentiated tumor cells expressed low levels of VEGF, the PEDF:VEGF ratio is altered so that their secretions changed from angiogenic, where stimulatory VEGF dominated, to anti-angiogenic, where PEDF was in control. VEGF was also the major angiogenic factor secreted by several tumor lines. Various inducers of angiogenesis have been shown to be produced by cells isolated from neuroblastoma tumors including VEGF (Eggert et al., 2000; Rossler et al., 1999; Meister et al., 1999), bFGF (Eggert et al., 2000), and IL-8 (Ferrer et al., 2000) using immunolocalization and/or RT/PCR. Our functional studies are in agreement with in vivo studies demonstrating that interference with VEGF restricts tumor growth in xenograft models (Rowe et al., 2000; Davidoff et al., 2000; Klement et al., 2000).

PEDF was elaborated by Schwann cells. A strong immunoreactive band was observed by immunoblot analysis of media conditioned by Schwann cells with an apparent molecular weight of 50 kDa, in agreement with the size previously reported (Pignolo et al., 1993; Ortego et al., 1996). A report examining media derived from Schwann cells failed to detect PEDF (Huang et al., 2000); however, several factors may account for this discrepancy including the use of a different antibody, sample processing and blotting conditions. Subsequent analysis of Schwann cell conditioned medium collected by this group was analyzed in our laboratory and a distinct protein of approximately 50 kDa was detected with our antibody (S.E.C. and S. L. Cohn, unpublished).

PEDF was also a potent anti-angiogenic agent elaborated by cultured Schwann cells. Although these cells are known to produce other molecules that are capable of blocking neovascularization (S.E.C. and X.H., unpublished) (Huang et al., 2000), our functional analysis using neutralizing PEDF antibodies indicated that PEDF is either the major anti-angiogenic factor made by Schwann cells, or it is an essential synergistic partner of other inhibitors present in their conditioned media.

The third antitumor activity of PEDF was its ability to support the expansion of the normal Schwann cell population by stimulating growth and by providing protection from apoptotic stimuli. PEDF has been shown by others to be capable of protecting a variety of neural cells, including mature ganglion cells similar to those found in neuroblastomas, from entering apoptosis but this is the first example of its doing so for Schwann cells. PEDF also enhanced Schwann cell growth. As cell numbers doubled in response to PEDF, this activity could not be attributed the simple enhancement of survival.

PEDF was secreted in abundance by Schwann cells in vitro and in situ within human tumors suggesting that it could be a major factor responsible for the expansion of the Schwann cell population within differentiating stroma-rich neuroblastomas. In vitro, Schwann cells respond to a number of other growth factors that may also be found in tumors including TGF-β, glial growth factor isoforms (Rutkowski et al., 1995; Ridley et al., 1989), and VEGF (Schratzberger et al., 2000; Sondell et al., 1999). It is not known how these growth factors affect the growth of Schwann cells within neuroblastomas or the expression of PEDF in vivo.

Perhaps the most remarkable aspect of PEDF’s role in neuroblastoma is its potential for participating in feedback loops that enhance its own synthesis (Fig. 7). Differentiated tumor cells, Schwann cells and ganglion cells, whose presence in ever increasing numbers is favored in the first place by PEDF, all in turn produce more PEDF. The multiplicity of PEDF’s antitumor activities coupled with this possibility of autocrine and paracrine loops to enhance its production indicate that PEDF is capable of providing a firm blockade to neuroblastoma tumor growth. They also provide an explanation for why neuroblastomas with high levels of PEDF-producing Schwann cells are more benign and raise the possibility that endogenous PEDF could be a major contributor to spontaneous tumor regression. In addition, they suggest that exogenous PEDF, used as a single agent or as an adjunct to traditional chemotherapeutic modalities, may prove to be unusually beneficial in stabilizing or suppressing the growth of these often highly aggressive childhood tumors.

We thank Mona Cornwell for her excellent technical assistance, Helen R. Salwen and Susan Cohn for providing N-type and S-type neuroblastoma cells, Rachid Idriss for Fig. 7 and the National Cancer Institute for support through grants CA52750 (to N.B.) and CA64239 (to S.E.C.).