|Home | About | Journals | Submit | Contact Us | Français|
Defining growth factor requirements for progenitors facilitates their characterization and amplification. We characterize a peripheral nervous system embryonic dorsal root ganglion progenitor population using in vitro clonal sphere-formation assays. Cells differentiate into glial cells, smooth muscle/fibroblast (SM/Fb)-like cells, and neurons. Genetic and pharmacologic tools revealed that sphere formation requires signaling from the EGFR tyrosine kinase. Nf1 loss-of-function amplifies this progenitor pool, which becomes hypersensitive to growth factors, and confers tumorigenesis. DhhCre;Nf1fl/fl mouse neurofibromas contain a progenitor population with similar growth requirements, potential, and marker expression. In humans, NF1 mutation predisposes to benign neurofibromas, incurable peripheral nerve tumors. Prospective identification of human EGFR+;P75+ neurofibroma cells enriched EGF-dependent sphere forming cells. Neurofibroma-spheres contain glial-like progenitors that differentiate into neurons and SM/Fb-like cells in vitro and form benign neurofibroma-like lesions in nude mice. We suggest that expansion of an EGFR-expressing early glial progenitor contributes to neurofibroma formation.
Schwann cells are peripheral nerve glia derived from the neural crest, a multipotent cell population that delaminates from the dorsal neural tube at E9 in the mouse. In addition to Schwann cells, rodent trunk level crest cells in vivo give rise to neurons of the peripheral nervous system (PNS) including those of the dorsal root ganglion (DRG), melanocytes and endoneurial fibroblasts, in vitro and in vivo (Baroffio et al., 1988; Chen and Lechleider, 2004; Fernandes et al., 2004; Ito et al., 1993; Joseph et al., 2004; Morrison et al., 1999; Shah et al., 1996; Stemple and Anderson, 1992). Growth factor signaling is broadly implicated in the maintenance of progenitor populations. In neural crest, combinations of Wnt and BMP suppress neural crest stem cell differentiation and maintain multipotency (Kleber et al., 2005; Lee et al., 2004), and notch and BMP2 signals influence gliogenic and neurogenic fate determination (Lo et al., 2002; Morrison et al., 1999).
The epidermal growth factor receptor (EGFR) has been specifically implicated in central nervous system progenitor expansion. Ligand (EGF) exposure confers multipotency to CNS subventricular zone progenitors at the transit amplifying stage (Doetsch et al., 2002; Fernandes et al., 2004; Reynolds and Weiss, 1992). EGFR signaling influences fate choices and chemotaxis of neural progenitors (Aguirre et al., 2005; Burrows et al., 1997; Lillien and Wancio, 1998). EGF has been included in culture medium for neural progenitors not only in the CNS, but also PNS boundary cap cells, and dermal neural crest derived progenitors (Fernandes et al., 2004; Hjerling-Leffler et al., 2005; Maro et al., 2004; Sieber-Blum et al., 2004), yet the effects of EGFR and its ligands have not been analyzed in dorsal root ganglion or peripheral nerve progenitors.
Neural crest cells develop into Schwann cell precursors between E11 and E13 in mouse sciatic nerve, immature Schwann cells by E15, and Schwann cells by E18 (Dong et al., 1999; Jessen and Mirsky, 2005). Progenitors identified after the establishment of the dorsal root ganglia (DRG) and the peripheral nerve have more limited self-renewal and differentiation potential than neural crest stem cells (Bixby et al., 2002; Kleber et al., 2005; Nagoshi et al., 2008). Products of the neuregulin-1 (Nrg-1) gene, acting through the receptor tyrosine kinase heterodimer ErbB2/3, promote survival of Schwann cell precursors and their differentiation into S100β+ Schwann cells (Dong et al., 1999; Leimeroth et al., 2002; Shah et al., 1996). The roles of most growth factors have not been analyzed in the developing PNS, especially in the neural crest-Schwann cell precursor transition.
The most common PNS tumor is the neurofibroma, hallmark of the inherited disorder neurofibromatosis type 1 (NF1) (Rasmussen and Friedman, 2000; Carroll and Ratner, 2008). Biallelic NF1 mutations have been detected in neurofibroma Schwann cells, implying that the Schwann cell lineage drives neurofibroma formation (Serra et al., 2001). Multiple Cre recombinase driver lines have been used to conditionally inactivate Nf1. Those that cause in vivo GEM-neurofibroma formation share expression at the Schwann cell precursor or boundary cap stages of Schwann cell development, placing the neurofibroma cell of origin subsequent to the neural crest stage (Joseph et al., 2008; Wu et al., 2008; Zheng et al., 2008; Zhu et al., 2002). It is debated as to whether the neurofibroma initiating cell is a committed glial cell, a de-differentiated Schwann cell, or a post-crest progenitor cell.
Failure of cellular differentiation and subsequent augmentation of a progenitor population may provide a basis for tumorigenesis (Jordan et al., 2006). Here we tested the hypothesis that the expression of EGFR defines a normal peripheral nerve glial progenitor population susceptible to Nf1 mutation, because 1 – 2 % of cells in neurofibromas express the Schwann cell marker S100β and the epidermal growth factor receptor (EGFR), which is not expressed by Schwann cells (DeClue et al., 2000). Also, some cells in neurofibroma tissue sections express the stem cell marker CD34 (Khalifa et al., 2000). In addition, E12.5 Nf1−/− DRG cultures are abnormally abundant in EGFR+ cells expressing the Schwann cell precursor marker Blbp/Fabp7 (DeClue et al., 2000; Kim et al., 1997; Miller et al., 2003).
We show that Nf1 loss in mouse expands numbers of EGFR-dependent peripheral progenitor cells, and confers tumorigenic capacity. Additionally, neurofibromas in the DhhCre;Nf1fl/fl model contain multi-potent progenitors. Finally, EGFR expression defines and allows isolation of tumorigenic progenitors from primary human neurofibromas.
Studies of various neural progenitors have demonstrated the utility of the in vitro sphere assay to assess self-renewal capability of a cell population (Fernandes et al., 2004; Joseph et al., 2008; Reynolds and Weiss, 1992). We dissociated E12.5 DRG of wild type, Nf1+/− and Nf1−/− embryos and cultured the suspension under sphere conditions defined for crest-derived progenitors (Fernandes et al., 2004). Nf1−/− DRG cell preparations gave rise to spheres in the presence of EGF and bFGF at a rate of about 1:700 by 14 days, significantly higher than wild type and Nf1+/− DRG cell suspensions (approximately 1:3000; p < 0.05, Figure 1A). Postnatal day 2 (P2) and 120 (P120) sciatic nerves were dissociated and plated in sphere-forming conditions. Nf1+/− P2 nerve cells generated primary spheres, but wild type cells did not; at P120 cells from neither genotype yielded spheres (Figure 1B). E12.5 DRG-derived spheres were dissociated to single cell suspension and re-plated. Cells of all genotypes formed secondary spheres, and the frequency of formation was not statistically different among genotypes (Figure 1C). Cells from spheres derived from embryonic DRG from all genotypes could be passaged for at least eight generations. P2 Nf1+/− spheres, in contrast, failed to give rise to secondary spheres (Figure 1D). Early postnatal Nf1+/− nerves contain EGFR+;P75+ cells (Wu et al., 2006), and may be the primary sphere-forming cells at P2. The data imply that the number of progenitors is increased in Nf1−/− embryonic DRG, but that the self-renewal capacity of mutant cells is unaltered.
A widely accepted measure of cell self-renewal is testing capability of sustained clonal expansion. To test whether clonal growth accounts for sphere formation, purified tertiary spheres from wild type and Nf1−/− embryos were transduced with lentiviral eGFP, dissociated and plated at limiting densities with equal numbers of unlabeled cells. All spheres were clonal as they were either fluorescent (green) or non-fluorescent (white) when plated at densities of 250 cells/cm2 or lower (Figure 1E–H, O). Higher plating densities yielded colonies which contained both fluorescent and non-fluorescent cells (not shown). Nuclear counterstain with DAPI (false-colored red) confirmed that all nuclei in newly forming spheres at clonal density were exclusively eGFP+ or negative at low density (Figure 1I–N). Clonal spheres from both genotypes could be dissociated and replated, giving rise to secondary low-density spheres (not shown), indicating that spheres form by expansion of single cell clones, not by cell aggregation. In conjunction with our ability to passage spheres for at least 10 passages in vitro, the data indicate that embryonic spheres contain clonogenic self-renewing cells.
To correlate the phenotype of sphere cells with the stages of development relevant to the embryonic DRG, we examined expression of a panel of crest and glial lineage markers using RT-PCR. Wild type and Nf1−/− spheres express the neural crest markers Twist, Nestin, Wnt5a, P75, and Sox9 (Figure 2A). Spheres from both genotypes expressed the glial markers Sox10, Mpz, Blbp (Fabp7), Dhh, and Gfap (Figure 2B). This may reflect that spheres contain cells representative of neural crest and/or immature glia, or an intermediate cell.
Sphere formation medium contains EGF and bFGF, implying that receptors for one or both of these factors are present on cells. We plated single spheres derived from wild type, Nf1+/, and Nf1−/− DRG and assayed EGFR and P75 (a neural crest lineage marker) immunoreactivity. Spheres from all genotypes contained cells that coexpressed EGFR and P75 (Figure 2C–E). All spheres analyzed contained EGFR+;P75+ cells (n = 8 spheres from each of three embryos per genotype).
To further characterize the external signals required for progenitor expansion, we plated dissociated cells from purified spheres in various doses of EGF, bFGF, or both. Both wild-type and Nf1−/− sphere formation were significantly increased when cultured with both factors versus either factor alone (Figure 2F). Additionally, Nf1−/− sphere-forming cells expanded (7.7-fold increase) when cultured in EGF and bFGF at low concentration (0.2 ng/ml) as compared to wild type (2.3% sphere formation versus 0.3%), suggesting higher sensitivity to growth factor-induced progenitor expansion. Spheres were maintained at saturating concentrations of growth factors (20 ng/ml) for subsequent experiments. Sphere formation was impaired when cells from dissociated primary spheres were subjected to 3 μM erlotinib (Figure 2G), which blocks EGFR tyrosine kinase activity (Norman, 2001). We confirmed the action of erlotinib on EGFR activity by western blot (Figure 2H). We also used a genetic approach to assess the degree to which EGFR activity drives sphere formation in the embryonic DRG. We generated spheres from E12.5 DRG from mice heterozygous for the EGFR mutation Waved-2 (Wa2/+) and their wild type littermate controls. Cells from the DRG of these mutants showed a 3-fold reduction in sphere formation versus wild type which was maintained even when spheres were passaged, as secondary sphere formation is quantified in Figure 2I. Thus, EGFR function is necessary for normal sphere formation.
The elevated response to low concentrations of EGF in Nf1−/− cells suggested that they may release additional EGFR ligands, which could act in an autocrine fashion. We examined whether EGFR ligands were being synthesized by sphere cells by RT-PCR. Cells from both wild type and mutant spheres express mRNA transcripts for Amphiregulin, Betacellulin, Tgf-a, HB-Egf, and Epiregulin (Figure 2J). The relevance of these factors remains to be determined.
As a defining characteristic of stem cells and progenitor cells is their ability to differentiate along multiple lineages, we tested whether cells in spheres from wild type or Nf1−/− embryonic DRG showed multi-lineage differentiation. Cells from clonally-derived spheres were plated in serum in absence of supplemental EGF and bFGF. Lineage differentiation markers were not expressed in growing spheres or directly after plating (Supplemental Figure 1 A, C, E, left panels). In some experiments, we monitored spontaneous differentiation. In other experiments, we directed differentiation according to previous reports (see experimental procedures) by exposing cells from spheres to lineage specific growth factors, after plating in neurogenic, smooth muscle/fibroblastic, or gliogenic conditions. Both wild type and Nf1 mutant spheres formed Schwann cells with cytoplasmic S100β expression, bipolar morphology and elongated nuclei (Supplemental Figure 1A right panels, B). Cells from both genotypes differentiated into smooth muscle/fibroblast-like cells, as assayed by smooth muscle actin (SMA) immunoreactivity and large round nuclei with a flattened, fibroblastic morphology (Supplemental Figure 1 C right panels, D). When placed into neurogenic conditions, wild type cells very rarely expressed neuron-specific β-tubulin (β-Tub). Cells from Nf1−/−sphere cultures generated many β-Tub+ cells with large cell bodies and long neurites (Supplemental Figure 1E, right panels). Of spheres which contained positive cells, there was a 10-fold increase in the number of β-Tub+ cell in Nf1−/− spheres versus wild type spheres (2.9% vs 0.27% respectively, Supplemental Figure 1F) in the spontaneous format. The cued paradigm also showed a neurogenic bias for the Nf1−/− spheres (2.1% vs. 0.7% respectively, Supplemental Figure 1F). Neurons also expressed neurofilament (NF-M, Inset; Supplemental Figure 1E). We were unable to stimulate pigmentation by cells from either genotype, although neural crest cells from mouse E9.5 showed robust pigmentation under the same conditions (not shown). We conclude that wild type cells from embryonic DRG-derived spheres differentiate into Schwann cells and smooth muscle fibroblasts, and infrequently into neurons. In contrast, Nf1 mutant cells appear less restricted.
To provide additional evidence supporting the ability of sphere cells to differentiate into neurons and glia we used an avian xenograft assay of neural crest cell differentiation. Sphere cells, wild type and Nf1−/−, migrated to the DRG and peripheral nerve. After in ovo differentiation most cells express the Schwann cell marker myelin basic protein (MBP), with some cells expressing the neuronal marker NeuN within the DRG (Supplemental Figure 2).
To assess the tumorigenic capacity of cells within spheres, we injected dissociated sphere culture cells subcutaneously into the flanks of nude mice. No lesions formed after injection of wild type cells (n= 11, 1.0–5.0 × 105 cells/injection) even after 18 weeks. Nf1 mutant spheres were dissociated and injected (1 – 1.5 × 105 cells/injection) and formed neurofibroma-like lesions in 11/21 injections (Figure 3A–C). We observed regions reminiscent of neurofibroma following hematoxylin and eosin staining of tissue sections (Figure 3D), containing occasional mast cells and cells with elongated nuclei (presumptive Schwann cells or fibroblasts). No features of malignancy (hypercellularity, failure to differentiate), necrosis, or edema were observed. Tumor areas were close to nerve fibers, as evident by neurofilament (Figure 3E) and S100β (Figure 3F) staining. Some cells were S100β+, bipolar Schwann-like cells (Figure 3G).
The mouse data demonstrates that the loss of Nf1 leads to an increase in EGFR+ progenitors which can be isolated, expanded as spheres, and are tumorigenic in nude mice. To test if human neurofibromas contain similar cells, we plated primary neurofibroma cells in medium containing EGF (20ng/ml) and bFGF (40ng/ml) on low-binding plastic dishes and observed selective expansion of cells that grew as floating spheres (Figure 4A). When grown on tissue culture plastic dishes, spheres formed as attached colonies, allowing for accurate quantification of growth (Supplemental Table 1).
We stained cells from colonies re-plated on chamber slides. As shown in Figure 4B – E, we identified EGFR+;P75+ cells (n=3 experiments, less than 2% total cells were double-positive). To test for EGF-dependence of colony formation, we plated 4 × 104 primary neurofibroma cells per well in triplicate using a 24-well plate cells in human sphere medium in the presence or absence of 20 ng/ml EGF. In this case, cells were plated onto tissue culture plastic, so that cells attached to the uncoated plastic for quantification. Fourteen days later colony (attached sphere) formation was scored. In three experiments in triplicate wells from two different patient tumors, in the presence of EGF, an average of 43 colonies per well formed, eight-fold higher than in the absence of EGF (Figure 4F – I ). When cell were plated in EGF together with the EGFR function-blocking antibody Cetuximab, an average of only 4 colonies per well appeared (Figure 4H–I). Similarly, when plated in EGF with the EGFR antagonist erlotinib (Tarceva; OSI-774) an average of 4 colonies per well formed (data not shown). Thus, human neurofibroma cell colony (sphere) formation is responsive to EGF and dependent on EGFR activity.
We tested whether EGFR+ cells could be prospectively isolated from human neurofibroma. We analyzed the co-expression of EGFR and P75 surface markers by flow cytometry on cells from aliquots of primary neurofibroma cells dissociated and then frozen. The EGFR+;P75+ cells also expressed a stem cell marker, CD34 (Figure 4J–L), previously shown to be expressed in neurofibroma tissue sections (Khalifa et al., 2000). The average final percentage of live cells that express all three antigens is 1.3%; all EGFR+;P75+ cells also express CD34 (data from two tumors from separate patients, sorted twice each). We repeated this analysis on two additional independent patient samples, and sorted live double-positive cells for culture in the sphere-forming assay. The sorted double-positive population showed 4 – 5 fold enrichment in primary colony-forming cells as compared to bulk, unsorted cells and depletion of the EGFR+ lineage (P75+) cells from the same patients prevented sphere formation. While bulk cells were rapidly depleted of sphere-forming potential, the EGFR+; P75+ sorted cells retained this capacity through tertiary passage (Figure 4M). Thus, EGFR expression correlates with enrichment of human neurofibroma sphere forming capacity.
Expression of CD34 raised the possibility that cells were blood-derived circulating progenitors similar to hematopoietic stem cells found in muscle tissue. To exclude this possibility we performed cobblestone area forming cell assays (CAFC), a classic in vitro hematopoietic stem cell/progenitor assay system on bulk neurofibroma cells. We identified only two colonies after two weeks; these did not possess hematopoietic morphology; in the CAFC assay, very rarely CNS neural progenitors will form colonies with a non-hematopoietic morphology (unpublished observation, H. Geiger). We incubated sorted EGFR+;P75+;CD34+ cells from a second patient neurofibroma in growth factor containing methylcellulose medium for 10 days. In this definitive hematopoetic stem and progenitor assay system, no neurofibroma cells formed colonies (not shown). Thus EGFR+;P75+;CD34+ cells do not contain bone marrow stroma-dependent hematopoietic stem or progenitor cells.
We have as yet failed to grow sorted EGFR+;P75+;CD34+ cells under clonal conditions, and have thus been unable to test whether the triple-positive cells meet this criterion to be defined as a stem cell; indeed the inability to maintain spheres for more than three passages suggests that they might resemble progenitors rather than stem cells. We first examined bulk neurofibroma tissue for progenitors, using the Schwann cell precursor marker BLBP (FABP-7). Both P75+;BLBP+ and EGFR+;BLBP+ cells were observed (Figure 5A, C). We confirmed CD34 and P75 expression in EGFR+ cells (Figure 5B, D). To assess whether spheres contain neural crest cells, we examined spheres from neurofibromas for expression of various known crest markers. In addition to BLBP, RT-PCR of neurofibroma-derived EGFR-dependent spheres revealed expression of mRNAs encoding the neural crest markers P75, SOX9, and TWIST, and the stem cell marker CD34 (Figure 5E).
The expression of these markers is associated with neural crest and/or early glial lineage progenitor cells. Spheres do not contain SMA+ or β-tubulin+ non-glial lineage cells (Figure 5G, H), but can be differentiated in vitro to SMA+ smooth muscle/fibroblast-like cells (Figure 5K) and β-tubulin+ (Figure 5L) and neurofilament+ (Figure 5M) neurons. They also became elongated and expressed cytoplasmic S100β as do Schwann cells after exposure to gliogenic conditions (Figure 5J). However, this marker is expressed in sphere cells prior to differentiation (Figure 5F). Consistent with the expression of early crest markers, differentiation did not yield MBP immunoreactivity, a later marker of Schwann cell differentiation, even though cultured normal human Schwann cells were MBP+ (not shown). When injected into nude mice (1.0 – 2.0 × 105 cells/injection; 3/6 injections from 3 independent patients), human spheres grew slowly (masses less than 3 cubic mm, by 8–12 weeks) as benign lesions containing S100β+ cells, mast cells, and rare Mib1+ cells (Figure 5N–Q). Only at a tenfold increase in cell number (1.0 – 2.0 × 106 cells/injection; 6/8), did injection of normal human Schwann cells cause growths of similar size. Hematoxylin and eosin staining of sections from normal cell grafts showed homogeneous S100β+ cells with microcysts containing myxoid material. No mast cells were observed (data not shown). Thus EGFR-expressing multi-lineage progenitors are present within neurofibroma-derived spheres, can be enriched prospectively, and grow as a mixed cell population in athymic nu/nu mice.
To determine whether an EGFR-dependent progenitors are present in a mouse neurofibroma, we studied the DhhCre;Nf1fl/fl mouse (Wu et al., 2008). EGFR+;P75+ cells were visualized by immunofluorescence in GEM-neurofibroma sections (Figure 6A–C). To assess the presence of progenitors, we plated dissociated tumor cells into medium permissive for sphere formation. Spheres grew within 10–14 days (Figure 6D). Due to large amounts of collagenous debris (even after filtering), quantification of cell number at plating was not reliable, preventing determination of the frequency of primary spheres. However, clonal density for secondary sphere formation was determined via limiting dilutions to be between 1 in 300 and 1 in 400 cells/cm2 (data not shown). We next tested tumor-derived sphere growth in EGF and bFGF. A significant increase in sphere formation (0.49 ± 0.17% vs. 0.11 ± 0.01%; p < 0.01) was observed in medium containing EGF and bFGF versus bFGF alone or medium without growth factors (Figure 6E). Notably, the presence of EGF alone promoted a significant increase in sphere formation, demonstrating EGF dependence in this model, as in human and embryonic spheres.
We then used RT-PCR to determine the developmental status of sphere cells (Figure 6F–H). GEM-neurofibroma spheres express transcripts characteristic of early crest cells (Snail, Slug, Twist1, Nestin, Sox9, Sox10, Wnt5a; Figure 6F), Schwann cell precursors (Blbp, Dhh, Ap2α)and mature Schwann cells (Gfap, MPZ, MBP, and Oct6); Figure 6G. Spheres do not express neuronal progenitor transcripts Ascl1 or Ngn1, or mature peripheral neuron transcripts Phox2B and Brn3a (Figure 6H). To determine the ability of sphere cells to differentiate, we plated cells in serum and identified neurons (β-tubulin+; neurofilament+; Figure 6I–K) as well as S100β+ glia and SMA+ smooth muscle/fibroblast-like cells (Figure 6L–N).
We describe a self-renewing, multi-potent, EGFR-dependent progenitor in the E12.5 mouse PNS (Fig. 7A). Nf1 controls the size of the progenitor pool, and influences its potential. Loss of Nf1 correlates with growth of cells in immunocompromised mice (Fig. 7B). We identify EGFR+ cells within human neurofibroma which form colonies and spheres, undergo multi-lineage differentiation, and grow in nude mice, and EGFR+ cells in DhhCre;Nf1fl/fl GEM-neurofibromas that form EGFR-dependent, multi-potent spheres. We suggest that this EGFR-dependent progenitor is relevant to NF1 tumorigenesis.
Our data identify a normal EGF-dependent embryonic PNS progenitor. DRG-derived spheres self-renew in an EGF-dependent manner, and an inhibitor of EGFR activity blocks sphere formation. Although erlotinib might also block the related receptor ErbB2, we demonstrate the expression and activation of ErbB1 in sphere cells. Confirming relevance of EGFR activity, cells from the Wa2 hypomorphic EGFR mutant show impaired sphere formation. Wild type and Nf1−/− cells each form clonally derived, EGFR-dependent, self-renewing multi-potent spheres. As multi-potent sphere-forming cells have been identified in developing (Bixby et al., 2002; Hagedorn et al., 1999; Morrison et al., 1999) and adult PNS (Nagoshi et al., 2008), EGFR may help define them. The presence of an EGFR expressing cell at the Schwann cell precursor stage of Schwann cell development E12.5 does not exclude relevance of EGFR to earlier and/or later stages. Indeed, GEM-neurofibromas show disruption of non-myelin forming Schwann cells in Remak bundles (Ling et al., 2005; Zheng et al., 2008; Wu et al., 2008). Some evidence supports an additional role for EGFR signaling in the perinatal period, via mast cell recruitment, to this phenotype (Monk et al., 2008).
Nf1−/− progenitors maintained a neurogenic bias over their wild type counterparts, consistent with loss of Nf1 causing survival and neurotrophin-independent neuronal differentiation (Vogel et al., 1995). Nf1 mutation also expands the progenitor population. While we were not able to cause cells to form melanocytes, Nf1−/−; EGFR+;Blbp+ E12.5 DRG-derived embryonic cells did form melanocytes in vivo (Rizvi et al., 2002). Strongly supporting a role for an at least bi-potent progenitor in human NF1, identical biallelic NF1 mutations were identified in neurofibroma Schwann cells and melanocytes (Maertens et al., 2007).
Although we cannot exclude a contribution of increased expression of EGFR ligand(s) expressed by Nf1−/− cells to self-renewal/expansion of the population, hyper-response to mitogens (EGF, and to EGF plus bFGF) is likely to result from loss of Nf1’s RAS-GAP function (Kim et al., 1995). Thus Nf1−/− blood progenitors show Ras-dependent heightened response to granulocyte-macrophage colony stimulating factor (Birnbaum et al., 2000; Zhang et al., 1998).
We identified EGFR+;P75+ cells within human neurofibromas that form spheres in vitro; FACS-sorting for EGFR and P75 enriched cells capable of sphere formation. Cells with a propensity to form colonies in neurofibroma cultures had been noted, but not studied (Muir et al., 2001). Human sphere cells are similar to mouse EGFR+ sphere cells in their response to and dependence on EGFR signaling, multi-potency, and the ability to form subcutaneous benign lesions in nude mice. Mouse sphere cells grown as subcutaneous xenografts also showed features of neurofibromas, including S100β+ and S100β− cells, occasional mast cells, association with nerves and blood vessels, and no evidence of malignancy. The data are consistent with our enriching neurofibroma initiating cells. Nf1−/− mid-gestation neural crest stem cells from E12.5 Nf1−/− mouse DRG did not form neurofibromas when injected into the nerve, possibly because nerve injury associated with this procedure provided an unsuitable environment for tumorigenic progenitor survival or maintenance or because of different culture conditions (Joseph et al., 2008).
Unlike mouse progenitors, human sphere-forming cultures were not expandable at clonal density and depleted by passage four. A possible trivial explanation for the limited clonogenic capacity of the human cell reflects lack of relevant cytokines in vitro. Alternatively (and possibly relevant to the clinical observation of transient neurofibroma growth), the human cell may have limited self-renewal because it is isolated after neurofibroma formation (Riccardi, 1992). We favor the idea that the human EGFR+;P75+ cell is a more committed, but not exclusive, progenitor, rather than an earlier stem-like cell. This interpretation is consistent with the identification of cells expressing the precursor marker AP-2α+ in human neurofibroma (Harder et al., 2006), and our finding that BLBP is expressed by P75+;EGFR+ neurofibroma cells. Our finding that mouse DhhCre;Nf1fl/fl tumors contain EGFR+;P75+ cells, and that dissociated tumor cells are capable of sphere formation and multi-lineage differentiation, also supports relevance of a similar progenitor population. DhhCre;Nf1fl/fl GEM-neurofibroma spheres express transcripts of multiple stages of glial development including the Schwann cell precursor (Blbp and Ap2α). Dhhcre;Nf1fl/fl neurofibromas, spheres derived from these neurofibromas, both wild type and Nf1−/− spheres from E12.5 embryos, and spheres from human neurofibromas all express EGFR and BLBP. This data provides strong support for the idea that the tumor initiating cell for human and mouse neurofibroma is a progenitor characterized by expression of these markers. The data is supported by expression of desert hedgehog, an additional marker of BLBP progenitors, in mouse neurofibromas, spheres derived from these neurofibromas, and both wild type and Nf1−/− spheres from E12.5 embryos.
Other reports identify proliferating cells within GEM-neurofibroma and postnatal Nf1-deficient nerve as P75+ (neural crest, precursor, non-myelinating Schwann cell and Schwann cell dissociated from axon marker) or Gfap+ (Schwann cells dissociated from axons and non-myelinating Schwann cells) (Joseph et al., 2008; Zheng et al., 2008). S100β+ (myelinating Schwann cell and Schwann cells lacking axonal contact)/BrdU+ cells and Blbp+ (progenitor) cells also occur in GEM-neurofibroma and nerve (Wu et al., 2008). Thus a panel of glial lineage markers is expressed in the GEM-neurofibroma, possibly reflecting stages of Schwann cell differentiation within the neurofibroma, paralleling our finding that human neurofibromas contain P75+;BLBP+;EGFR+ cells. As progenitors exist and non-myelinated fibers are disrupted within neurofibromas, we favor the view that aberrant progenitors ultimately cause non-myelinating Schwann cell dysfunction.
A hierarchical model of tumor stem/progenitor cells has been developed stating that cellular events at an early point in a lineage allow for tumor formation as altered development proceeds. Much about stem cell-initiated tumors has been learnt from the hematopoietic system. For instance, chronic myeloproliferative disorders are stem cell clonal disorders, resulting from mutations that dysregulate key cellular proliferation pathways (e.g., BCR/ABL or JAK2 mutants) (reviewed in Tefferi A, 2008). These mutations induce “benign” lineage-specific cell proliferation, which themselves do not represent oncogenic transformation but instead favor development of second hits which are the base of malignant transformation into acute leukemias. Cancer stem/progenitor cells have also been identified in solid tumor models (Jordan et al., 2006). Our findings that progenitor cells can be obtained from GEM-neurofibroma and human neurofibroma support the hypothesis that these examples are relevant to benign neurofibroma formation. The identification of an EGFR+ neurofibroma-initiating progenitor population provides a novel avenue for study of the cellular biology of neurofibroma tumorigenesis enabling novel therapeutic targeting strategies.
We maintained mice per guidelines of the Animal Use and Care Committee at the Cincinnati Children’s Hospital Medical Center. We mated Nf1+/− C57Bl/6 females (Brannan et al., 1994) with Nf1+/− 129X1/SvJ Nf1+/− males (Jackson; Bar Harbor, ME). For timed dissections, the morning on which a vaginal plug was observed was E0.5. DhhCre;Nf1fl/fl mice were generated as described (Wu et al., 2008).
For primary cultures, we dissociated DRG from E12.5–13.5 embryos with 0.25% Trypsin (Mediatech; Herndon, VA) for 20 min. at 37° C and obtained single-cell suspensions with narrow-bore pipettes and a 40 μm strainer (BD-Falcon). For postnatal mouse sciatic nerve dissection, we chopped tissue into 1–3 mm3 pieces, plated in sphere medium for 3 days, centrifuged and resuspended in L-15 (Mediatech) plus 0.5 mg/ml collagenase type 1 (Worthington; Lakewood, NJ), and 2.5 mg/ml dispase protease type II (Cambrex; East Rutherford, NJ) at 37°C for 25 min. For all mouse cells, we used trypan blue exclusion (Stem Cell Technologies, Vancouver, BC) to plate 2 × 104 live cells per well in 24-well plates in sphere medium. Sphere medium (modified from Fernandes et al., 2004): DMEM:F-12 (3:1) + 20 ng/ml rhEGF (R&D Systems), 20 ng/ml rh bFGF (R&D Systems), 1% B-27 (Invitrogen), 2 μg/ml heparin (Sigma). We maintained cultures at 37° C and 5% CO2 and counted floating spheres after 14 days. Half of the volume was removed and replaced with fresh medium every 4 days, taking care to centrifuge the used medium and collect any cells or spheres that may have been floating to return to the plate. To passage, we centrifuged sphere cultures, treated with 0.05% Trypsin-EDTA for 3 min., dissociated and plated at 2 × 104 cells/ml in 50% conditioned and 50% fresh medium. We counted secondary spheres after 14 days. For EGFR inhibition experiments, erlotinib (Genentech; San Francisco, CA), was dissolved in DMSO, added at 3 μM, and replenished in fresh medium every 3 days. For each experiment, we show a representative of >3 independent experiments.
We obtained human plexiform neurofibromas incidental to therapeutic surgery and after IRB approval. We cut tumors into 1 mm3 pieces in L-15 supplemented with Pen/Strep (Mediatech), Gentomycin, Collagenase Type I, and Dispase protease II. We dissociated tumors at 37°C for 4 h. For sphere formation, we plated 4 × 104 cells per well in 24-well plates, and counted spheres after 21 days. Human sphere medium was DMEM/F-12 (1:1) + 1% N2 supplements, 1% B27, 20 ng/ml EGF, 20 ng/ml bFGF, 0.1 mM β-ME (Sigma), 2 μg/ml Heparin (Sigma). EGF and EGFR inhibitor application was as for mouse. The monoclonal antibody Cetuximab was also used at 5 μg/ml (ImClone). For culture details, see Supplemental Table 1. For immunocytochemistry, we seeded spheres onto poly-L-lysine (PLL) and laminin-coated Lab-tek II tissue culture plates (Nalge Nunc, Rochester, NY). Differentiation conditions were as for mouse.
We collected 6,000 triple-positive cells and plated them at 8 concentrations (200 cells per well through 1.6 cells per well; 15 wells per dilution) to identify hematopoietic progenitor and stem cells (Geiger et al., 2001). We scored for clones 3 times per week for 7 weeks.
DhhCre;Nf1fl/fl tumors were dissociated in the same manner as human tumors, filtered with 40 micron cell strainers, and plated in mouse sphere medium; differentiation and immunocytochemistry were conducted as above. Spheres were dissociated as above and split after 10 – 14 days; feeding was the same as for embryonic sphere cultures.
We produced lentivirus particles encoding GFP by transient co-transfection of 293T cells, as described (Arumugam et al., 2007). We dissociated spheres and plated 1.0 × 103 cells in 96-well plates in 200 μl sphere proliferation medium. After 24 hours we removed growth medium, added a minimal volume (~50–100 μl) of purified virus at an MOI of 10 and incubated at 37° C. We removed virus-containing medium the next day and applied fresh sphere medium.
We combined equal ratios of lentivirally labeled, eGFP+ cells and unlabeled cells from the same embryo and replated cells from dissociated mouse spheres at limiting dilutions on PLL coated chamber slides. We validated single-cell plating. After 10 days percentages of adherent single-color and mixed spheres were scored. We dissociated and replated spheres from wells yielding only single-color spheres. All samples gave rise to secondary spheres.
We incubated neurofibroma cell suspensions with mouse anti-human monoclonal antibodies against CD34 (8G12/HPCA-2, Becton-Dickinson; San Jose, CA) bound to allophycocyanin (APC), p75/NGFR (C40–1457, Becton-Dickinson) bound to phycoerythrin (PE) and EGFR (EGFR1, Research Diagnostics Inc; Flanders, NJ) bound to FITC and anti-CD16/CD32 antibody (Becton-Dickinson) to block Fc gamma receptor unspecific binding, on ice in a solution containing PBS/0.2%BSA/0.01% NaN3 for 30 minutes. After washing, we resuspended cells in PBS/0.2%BSA/0.01% NaN3/2 μg/mL 7-aminoactinomycin D (7-AAD, Invitrogen). We carried out isotipic controls with irrelevant mouse IgG1-APC, mouse-IgG1-PE and mouse IgG2b-FITC in parallel. We acquired cell suspensions in a dual-laser (Argon 488 and dye laser 630 or HeNe 633) FACSCalibur or FACSCanto (Becton-Dickinson) and analyzed on an “alive” gate based on light scatter parameters and 7-AAD staining negativity. We used the same incubation conditions to isolate cell populations (positive and negative for the antigens CD34, p75/NGFR and/or EGFR) by FACS. We acquired cells in a FACSVantage DiVa equipped with a 488 nm-Argon laser, a 633 nm-HeNe laser and a tunable Argon-UV laser, sorting a total of 6×106 cells at one time.
We plated DRG-derived spheres on PLL and laminin-coated chamber slides. Base medium was DMEM: F-12 (3:1), + 10% FBS. For neurons, + 40 ng/ml bFGF for 5 days only, then +10 ng/ml each NT-3 and BDNF (R&D Systems), and NGF (Harlan) for 7 days (Fernandes et al., 2004). For glia, +5 ng/ml recombinant humanβ-heregulin and 2 μM forskolin (both R&D Systems) for 10 days (Ratner et al., 2006). For smooth muscle fibroblasts, +40 ng/ml bFGF for 5 days only, then + 1 ng/ml TGF-β1 (Sigma-Aldrich) for 7 days (Chen et al., 2004). For all conditions, we changed medium every 3 days.
Following fixation, we permeablized cells with 0.2% TX 100 and blocked with 10% normal serum for one hour at room temperature. Primary antibodies were: goat anti-EGFR (Santa Cruz, SC-03-G, 1:10), rabbit anti-P75 (Chemicon 1:400), mouse anti-β-Tubulin (Chemicon 1:200), rabbit anti-S100β (DAKO 1:2000), mouse anti-SMA (Sigma 1:500), and rabbit anti-NF-M (Chemicon, 1:200). Secondary incubations were with host-appropriate FITC or TRITC (Jackson Immunoresearch; West Grove, PA). We labeled nuclei with DAPI (Sigma-Aldrich; St. Louis, MO). We acquired microscopic images with Openlab software suites on a Zeiss Axiovert 200.
We anaesthetized athymic nu/nu mice in isoflurane and subcutaneously injected 1–4.5 × 105 mouse or human sphere cells/injection into left flanks. Tumors were monitored weekly. We dissected visible tumors and fixed them in 4% paraformaldehyde overnight, then embedded in paraffin. We stained 6uM sections with hematoxylin and eosin, anti-S100β (1:2000, DAKO) and anti-neurofilament (1:2000, Chemicon) visualized with HRP-conjugated goat anti rabbit antibody with diaminobenzidine reaction.
We isolated total RNA from spheres (RNeasy mini-kit, Qiagen) and made cDNA (Superscript III RT, Invitrogen). We conducted RT-PCR as described (Miller et al., 2003). Mouse sequences and protocols for Twist1, Nestin, Wnt5a, P75, Sox9, and Mpz are published (Fernandes et al., 2004)as are mouse Egfr (Represa et al., 2001), mouse and human BLBP (Miller et al., 2003), and human P75 (Toma et al., 2005). Additional primer sequences are shown in Supplemental Table 2.
All experiments were analyzed with a paired t-test, and data are shown as mean β standard error of the mean.
We thank M. Nakafuku (Cincinnati Children’s) for numerous helpful suggestions, A. Stemmer-Rachamimov, (Massachusetts General Hospital) and M. Collins (Cincinnati Children’s) for assistance with pathology, and B. Ehmer for her assistance with confocal microscopy. This work was supported by NIH R01-NS 28840 to NR.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.