| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,1
,
Departments of
* Genetics and Pathology, Rudbeck Laboratory,
Women and Child Health, University Hospital,
Animal Development and Genetics, Evolutionary Biology Centre, and
Medical Cell Biology, Biomedical Centre, University of Uppsala, SE 751 85 Uppsala, Sweden; ||Renal Section, Hammersmith Campus, Imperial College, London, Great Britain;
# Department of Pediatrics, University of Cologne, D-50924 Köln, Germany;
¶ Department of Pediatric Hematology and Oncology, Olgahospital, D-70176 Stuttgart, Germany;
** Department of Pediatric Pathology, Christian-Albrecht’s University in Kiel, D-24042 Kiel, Germany;
Childhood Cancer Research Unit, Karolinska Institute, SE 17176 Stockholm, Sweden; and
Biacore AB, SE-754 50 Uppsala, Sweden
1Correspondence: Department of Genetics and Pathology, Rudbeck Laboratory, SE-751 85 Uppsala, Sweden. E-mail: fredrik.hedborg{at}genpat.uu.se
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: angiogenesis • differentiation • sympathetic nervous system • insulin-like growth factor • vascular-endothelial growth factor
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
and is derived from immature cells of the sympathetic nervous system (2)
. A candidate progenitor cell type for neuroblastoma is the rosette-forming sympathetic neuroblast. These are abundant during fetal development but difficult to identify in histological specimens of the sympathetic nervous system after the age of three (3)
. The incidence of neuroblastoma parallels the decline in these sympathoblasts. It is one of the most common forms of congenital cancer (4
, 5)
; it peaks during the first year of life, then decreases gradually making neuroblastoma rare after age seven (6)
. Clinically it is a multifaceted disease with wide variation in malignant potential, symptoms, and endocrine activity (7)
. One peculiarity of neuroblastoma is that some cases exhibit spontaneous regression and maturation, which can result in benign ganglioneuroma (8)
. Unfortunately, modern chemotherapy regimens have resulted in only marginal improvement in the survival of patients with aggressive tumors (9
10
11)
, motivating further efforts to obtain a better understanding of its biology.
Neuroblastomas can be separated histologically according to their degree of cellular maturation and abundance of stromal cells of a glial, Schwann cell-like phenotype (12
, 13)
. Generally, unfavorable patient outcome is associated with less cellular maturation and fewer Schwann cell-like changes. Conversely, benign ganglioneuromas are rich in Schwann cell-like stroma and exhibit the highest degrees of cell maturity. There are many exceptions to this general rule, mainly because tumors in infants commonly exhibit primitive cell morphology and are stroma-poor despite having a good prognosis (14)
.
Aspects of cellular differentiation within these complex tumors are likely to affect their susceptibility to specific treatments. The mainstream paradigm for maturation/differentiation of neuroblastoma cells is based on the differentiation pathway for the sympathetic ganglion cell, but there is recent evidence indicating the existence of a neuroendocrine/chromaffin pathway of maturation in clinical neuroblastomas (15
16
17
18
19)
.
In the present study we sought to further elucidate the phenotype of these chromaffin-featured tumor cells, to examine their prognostic significance, and investigate the mechanisms underlying the phenomenon. In the first part of the study, we examined gene expression and histological features in a series of 116 clinical neuroblastomas. We show that chromaffin-differentiated regions express high transcript levels of the insulin-like growth factor II gene (IGF2) and low levels of the growth-associated protein-43 gene (GAP-43), consistent with their respective expression in chromaffin and neuronal cells in the fetal nervous system. The chromaffin phenotype was tightly associated with tumor regions that were predicted to be hypoxic. In the second part of the study, we used defined neuroblastoma cell lines. Mouse xenografts derived of these cell lines were shown to express chromaffin-associated markers in regions of hypoxia. Finally, we show that in monolayer culture, low oxygen tension induces aspects of the chromaffin phenotype, which is reversible upon reoxygenation.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Clinical tumor material
Tumor specimens from patients enrolled in the German Collaborative Neuroblastoma Studies during 1980-1995 were chosen from the comprehensive central collection of formalin-fixed, paraffin-embedded specimens at the Department of Pediatric Pathology of the Christian Albrechts University of Kiel. Evans classification (20)
; after 1990, the International Neuroblastoma Staging Systems (21)
were used for tumor staging. Of the 116 tumors, 89 had been sampled prior to any chemotherapy (39 were tumor stages 1, 2, and 4S and 50 were stages 3 and 4). Specimens of the other 27 tumors had been subjected to preoperative chemotherapy (7 were tumor stages 1, 2, and 4S and 20 were stages 3 and 4). Twelve of the 116 tumors were a consecutive series of infant tumors diagnosed via a population-wide screening program in the region of Stuttgart based on measurements of catecholamine metabolites in urine samples collected at 7 months of age. Minimal follow up time for the entire material was 25 months. Half (58) of the patients were diagnosed during infancy. Staging, tumor site, and outcome depending on age at diagnosis (more or < 1 year) are summarized in Table 1
. Written informed consent was given by the parents for documentation and analysis of medical data.
|
Experimental tumors
Seventy-one immunodeficient nude mice (NMRI nu/nu; Animal Department, BMC, Uppsala University) were injected subcutaneously with 20–50 million SH-SY5Y cells; 65 animals developed tumors that were allowed to reach an approximate diameter of 2 cm. Perfusion fixation under terminal anesthesia was carried out with 4% paraformaldehyde in Millonig’s phosphate buffer, pH 7.4, at a pressure of 100-140 mmHg via a 16-gauge cannula inserted into the thoracic aorta via the left ventricle. Two to four h before death, some of the animals were injected intraperitoneally with 7 mg 2-nitro-amidazole-theophyllin (NITP) as an extrinsic marker of hypoxia (22
, 23)
(kindly provided by Professor Juliana Denekamp, Department of Oncology, Umeå University, Sweden) dissolved in 0.45 mL peanut oil and 0.05 mL dimethylsulfoxide. Tumors were also established in two animals each from the neuroblastoma cell lines SK-N-AS, SK-N-DZ (kindly provided by Kristine Björnland, National Hospital, Oslo, Norway) and SK-N-SH and IMR 5. Four identically produced tumors from the IMR 32 neuroblastoma cell line were kindly provided by Lars-Gunnar Larsson, University of Agricultural Science, Uppsala, Sweden. After perfusion fixation, experimental tumors were excised, sectioned in half, and then immersed in fixative for 2-10 days before embedding in paraffin.
Cell culture experiments
Neuroblastoma cells were grown on 9 cm diameter plastic dishes (Labora, Stockholm, Sweden) at 37°C/5% CO2 in Eagle’s minimum essential or RPMI medium supplemented with 10% fetal calf serum, glutamine, penicillin, and streptomycin. Regardless of oxygen tension, cells were maintained in 5% CO2. The lowest oxygen tension was achieved with a commercial system for anaerobic bacterial culture (GasPack Plus SystemTM, BD Biosciences, San Jose, CA, USA). Cells were kept at 37°C and hypoxia was monitored using an indicator defining an oxygen level of no more than 0.1%. Less severe hypoxia was accomplished by flushing a 95:5 mix of N2/CO2 through an air-tight box with valves (Eurometric Instruments, Stockholm, Sweden) and continuously monitoring with a pO2 meter (PAC III Standard, Dräger Svenska AB, Svenljunga, Sweden). Cell density at initiation of hypoxia and of control cells was 
30% confluent, corresponding to 1-2 million cells per dish. Eagle’s minimum essential medium with no glucose added was used for glucose-free cell culture and RPMI medium supplemented with 30 nM selenium, 10 nM hydrocortisone, 10 nM ß-estradiol, and 30 µg partially Iron saturated transferrin (SHTE medium) was used for serum-free culture.
Tumor processing and histochemical procedures
Serial sections (3-6 µm) were collected on coated slides (Superfrost* Plus, Menzel-Gläser, Germany). Antigens were detected as follows: tyrosine hydroxylase (monoclonal antibody, diluted 1:40, Boehringer Mannheim, Mannheim, Germany) and chromogranin A (monoclonal antibody, BioGenex Laboratories, San Ramon, CA, USA, diluted 1:1000, and polyclonal antibody from Milab, Malmö, Sweden, diluted 1/3000). An insulin-specific antibody was used as a negative control (monoclonal antibody, MUO 29-UC, BioGenex Laboratories). NITP was identified immunohistochemically with a polyclonal antiserum (#T-2524, Sigma, St. Louis, MO, USA). Sections were generally microwave treated for two 5 min periods at full power (800W) in 0.1 M citrate buffer (pH 6.0) before incubating overnight with antibodies at 4°C. Detection was with the avidin-biotin-peroxidase complex method and diaminobenzidine (ABC kit, Dako, Glostrup, Denmark). The Grimelius method (24)
was used to detect neurosecretory granules.
RNA probes
[35S]- and [32P]-labeled riboprobes were made using a 680 bp cDNA fragment spanning the coding region in exon 9 of IGF2 (a kind gift from Professor Rolf Ohlsson, Uppsala) and a 740 bp EcoR1/BamH1 VEGFA165 cDNA fragment spanning the entire coding region (kindly provided by Arne Östman, Uppsala). A control 28S rRNA riboprobe was made from pTRI RNA 28S (Ambion Inc., Austin, TX, USA). Other radiolabeled riboprobes were made using an 800 bp HindIII/EcoRI cDNA fragment of GAP-43 (kind gift from Professor Sven Påhlman, Malmö), a 1.9 kb BamH1 cDNA fragment of human ß-actin (kind gift from Professor Monica Nistér, Uppsala), a 93 bp cDNA fragment of human LDH-A (nucleotides 275-367 of sequence accession no. X02152), and a 136 bp fragment of human GLUT-1 (nucleotides 1063-1198 of sequence accession no. K03195). Riboprobes were made using standard procedures, with supercoiled plasmid templates for [35S] probes for in situ hybridization and linearized templates for the [32P] probes used in RNase protection assays. The approximate specific activity of the [35S] probes was 250 Ci/mol.
In situ hybridizations
Riboprobes were hybridized to sections at 56°C overnight and washed under stringent conditions before RNase treatment (25)
. After 12 h to 8 days of X-ray film autoradiography, NTBII photographic emulsion (Eastman Kodak Co., Rochester, NY), diluted 1:1 in 2% glycerol, was applied, followed by exposure for 3 days to 4 wk. The slides were then developed and counterstained for <1 min with Mayer’s hematoxylin diluted in tap water.
Preparation of RNA and RNase protection analysis
Total RNA was prepared using a Nucleospin RNA II kit (Clontech Laboratories, Palo Alto, CA, USA). [32P]-Labeled antisense riboprobes were annealed to 5-10 µg of total RNA at 42°C for 15 h and analyzed with a RPA III kit (Ambion). After removal of excess probe by RNaseA/T1 degradation, denatured samples were analyzed on a 6% acrylamide-urea sequencing gel. Signals were quantified with a PhosphorImager using Image Gauge V3.45 software. For standardization of expression levels, values were related to their respective 28s rRNA signal.
Imaging and image analysis
Microphotographs were digitally processed from a Leica DMRX. Distances from vascular structures was analyzed using Leica QWin software.
Statistics
Statistical correlations were examined using 
2 cross tabulation (Statistica 4.1 software, StatSoft Inc., Tulsa, OK, USA).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
) and small intensely fluorescent (SIF) cells (Fig. 1D
) and were expressed at low to moderate levels in chromaffin cells of the adrenal medulla (data not shown). By contrast GAP-43, a gene involved in axonal growth (26)
, was expressed mainly by neuronal cells of the central and peripheral nervous systems (Fig. 1B
, data not shown) with only minor expression, just above the detection limit, in paraganglia (Fig. 1B
) and SIF cells (data not shown). GAP-43 expression thus exhibited cell specificity complementary to that of IGF2 within the early nervous system. We also tested fetal sympathetic cell types for expression of conventional chromaffin markers, which revealed good (but not absolute) specificity, with high-level signals in paraganglia and SIF cells for tyrosine hydroxylase (Fig. 1C, E
) and chromogranin A immunohistochemistry (Fig. 1F
and data not shown) and for Grimelius’ staining (data not shown).
|
Clinical tumors
We next analyzed the expression of IGF2 and GAP-43 in consecutive sections of 116 cases of clinical neuroblastoma. A total of 56 tumors expressed IGF2. In these tumors we observed strikingly complementary patterns of expression of IGF2 and GAP43 that depended on cell position in relation to tumor stroma. Generally, IGF2-expressing tumor cells were distant from the fibrovascular stroma, most frequently centrally in tumor nodules (Fig. 2
A–C), and were also seen in perinecrotic regions. GAP-43 was mainly expressed by blood vessel-apposed, IGF2 nonexpressing tumor cells (Fig. 2D-F
). Between these extremes was an intermediate zone of coexpression, with positive and negative gradients of IGF2 and GAP-43, respectively, as distance from the supporting tumor stroma increased (Fig. 2)
. Intra- (small vs. large boxed regions in Fig. 2A, D
) and inter-tumoral (Fig. 2
and Fig. 3
O/Q) variations in the proportion of GAP-43/IGF2 coexpressing cells were seen.
|
|
An effect of distance to the nearest blood vessel on tumor cell expression of IGF2 was evident in 50 of 56 tumors. This vascular dependence was seen in tumor nodules from tumors that were stroma-poor (Fig. 2B/C
and Fig. 3A/B, E
) and stroma-rich (Fig. 3J, O
), some of which also contained mature cells with a ganglion cell-like morphology (marked with asterisks in Fig. 3K-M
). The distance between capillary endothelium and the first cell layer exhibiting IGF2 expression varied between tumors with a mean of 110.5 µm (median 107 µm, SEM 4.8 µm, range 66–172 µm). In the remaining six IGF2-expressing tumors, one stroma-poor tumor showed a homogeneous IGF2 expression and five other stroma-poor tumors exhibited cellular heterogeneity in expression of the gene but with no clearly discernible effect of distance to the microvasculature.
Morphological correlates of IGF2 expression
Tumor cell expression of IGF2 correlated with striking specificity to an altered cell morphology (Fig. 3)
. Most consistently, IGF2-expressing cells displayed a distinct nuclear morphology. These nuclei were larger than nuclei of adjacent neuroblastic cells (Fig. 3C/D, F, P/R
) but smaller than those of adjacent ganglion cell-like tumor cells (Fig. 3I-M
), had one or two relatively small nucleoli, and displayed a punctate basophilia in a characteristically pale background with hematoxylin staining (Fig. 3D, F, K/L/M, P
). Within the developing sympathetic nervous system these nuclear features are characteristic of chromaffin cells (3)
and are specifically consistent with those of SIF cells, but differ clearly from the nuclear morphology of mature ganglion cells (Fig. 1E, F
). Other morphological features that correlated with IGF2 expression were a moderate increase in cell size and increased eosinophilia of the cytoplasm compared with adjacent neuroblastic cells (Fig. 3C/D, F
). Immunohistochemical staining of IGF2-expressing neuroblastoma cells (described below) revealed a distinct sperm-like cell shape with an ovoid body projecting a relatively thick and beaded neurite from one pole (Fig. 3H, I, M
), resembling the size and shape of normal SIF cells (Fig. 1E, F
). IGF2-expressing regions in the center of tumor nodules were furthermore associated with abundant formation of neuropil (Fig. 3C/D
). Not infrequently such neuropil formation was associated with a characteristic cell arrangement with IGF2-expressing cells lining up in parallel rows along the direction of these fibers (not shown).
Conventional histochemical evidence of a chromaffin phenotype in neuroblastoma
Conventional markers for a neuroendocrine cell phenotype are cytoplasmic immunoreactivity for the neurosecretory granule protein chromogranin A and a positive Grimelius’ silver stain. When testing these features on consecutive sections we found them to be specific for IGF2-expressing neuroblastoma cells (Fig. 3G/H, I/M, N
). Fifty-three of 56 specimens containing IGF2-expressing tumor cells were also chromogranin A immunoreactive. In general, IGF2 and chromogranin A were strictly coexpressed at the cellular level, as judged from the results of analyses on consecutive thin sections (Fig. 3E/G, I/J
). Tyrosine hydroxylase, the rate-limiting enzyme in catecholamine synthesis that is highly expressed by chromaffin cells (Fig. 1C, E
), was detected in neuroblastoma with a high specificity for IGF2-expressing tumor cells, which were either uniquely immunoreactive or showed a more intense staining than adjacent IGF2 nonexpressing tumor cells (Fig. 3L
and data not shown). These correlations between conventional chromaffin markers and expression of IGF2 were seen independent of whether tumors had been subjected to chemotherapy or not.
Taken together, the expression of IGF2, morphological characteristics, and histochemical features characterize a chromaffin-like tumor cell population in 50% of neuroblastomas that are strikingly similar to SIF cells and clearly different from ganglion-like tumor cells.
Histological subgrouping of clinical tumors
We next investigated the relationship between the presence of chromaffin-like features and the previously recognized parameters of ganglionic differentiation and presence of Schwannian-rich stroma. Cellular coexpression of IGF2 and chromogranin A was chosen as an operational definition of a chromaffin phenotype. In the complete series of tumors chromaffin differentiation was seen in 46% (53/116 tumors).
The great majority of the series of tumors (104/116) had a stroma-poor histology. Of these 104 tumors, 48 contained chromaffin-like cells. Ganglion-like cells were not found in these 104 tumors. The other 12 tumors in the series had a Schwannian stroma-rich histology of which 8 contained ganglion cell-like cells (large, rounded cells with a large, homogeneously staining nucleus with a prominent nucleolus; marked with asterisks in Fig. 3I-M
). Among these eight Schwannian-rich tumors with ganglionic differentiation, additional mature cells that fulfilled the chromaffin criteria were seen in three tumors (Fig. 3I-N
). In two of the stroma-rich tumors (2/12) that did not show evidence of ganglionic differentiation, cells of the chromaffin phenotype were present (Fig. 3O-R
). Consequently, tumors could be classified into four phenotypic categories in terms of maturation/differentiation of nonstromal tumor cells: exclusively chromaffin-featured tumors (43%; 50/116 tumors), exclusively ganglion cell-featured tumors (4%; 5/116 tumors), tumors containing a mixture of cells with these two respective phenotypes (3%; 3/116 tumors), and tumors with none of these signs, thus assigned as poorly differentiated (50%; 58/116 tumors). Among all 58 tumors with morphological or marker-based evidence of maturation/differentiation of nonstromal tumor cells, therefore, the chromaffin-like phenotype was the only 1 in 86% (50/58 tumors). With respect to the presence of a Schwannian-like stromal histology, chromaffin-like maturation was found in 42% (5/12 tumors) of stroma-rich tumors and in 46% (48/104 tumors) of stroma-poor tumors. Ganglion-like tumor cells were seen exclusively in tumors with a Schwannian stroma-rich histology and were present in a majority of this tumor subcategory (8/12 tumors).
Clinical characterization
Statistical evaluation (2 test, P value cutoff limit <0.05) suggested that the presence of chromaffin-like cells neither correlated to survival (P=0.17) nor to age at diagnosis (more or less than 1 year, P=0.46). A chromaffin tumor phenotype was significantly more frequently seen in stage 1-2B tumors than in stage 4 tumors (22/38 vs. 15/42; P=0.047) and in tumors detected via infant screening (9/12 vs. 44/104; P=0.03). Neither treatment prior to operation (treatment vs. no treatment, P=0.28) nor site of the tumor (adrenal vs. extra-adrenal, P=0.88) was statistically correlated with the presence of chromaffin-like cells.
Coexpression of VEGFA and chromaffin markers
The relationship between chromaffin-like features and distance from the vasculature suggested that tissue oxygenation might be important in this phenotype. We therefore examined expression of VEGFA, which is a hypoxia-inducible gene. Generally, VEGFA was remarkably coexpressed with IGF2 (Fig. 4
). In some tumors VEGFA was expressed by a smaller proportion of the IGF2-expressing tumor cells, albeit with a similar relationship to distance from the vasculature (data not shown). Overall, the colocalization of VEGFA expression with the chromaffin phenotype supported a possible role for hypoxia in this phenomenon.
|
Xenograft experiments
Clinical tumors are heterogeneous, and it is difficult to determine the precise role of genotype and microenvironment in generating phenotypic effects. Furthermore, in experimental studies it is possible to delineate tissue hypoxia with bioreductive markers. Therefore, to more directly examine the effect of tumor hypoxia, xenogenic tumors were generated from six human neuroblastoma cell lines; SK-N-DZ, SK-N-SH, SH-SY5Y, LAN 1, IMR 32, and SK-N-AS. The latter cell line was included because of its constitutive expression of IGF2 (27)
. These xenografts were tested at the cellular level for expression of the same set of markers as used for the clinical tumors. All tumors expressed VEGFA, with positive cells positioned at the periphery of blood vessel-embracing tumor cords, frequently bordering onto regions of necrosis/apoptosis (Fig. 5
A and data not shown). SK-N-AS-, SK-N-DZ-, and SH-SY5Y-derived tumors were positive for IGF2 in situ hybridization. The latter two cell lines displayed regions of strict coexpression of VEGFA, IGF2, and tyrosine hydroxylase (Fig. 5A-C
and data not shown). Furthermore, these regions were clearly distant from the vascular supply. Morphologically, tyrosine hydroxylase immunohistochemistry revealed the characteristic SIF cell-like shape of these cells with a prominent monopolar neurite (not shown). Analogous to clinical tumors, cells positioned closer to the blood vessels expressed the neuronal marker GAP-43 in SH-SY5Y xenografts (Fig. 5D
), with a narrow transitional zone of IGF2 coexpression (Fig. 5E
). VEGFA and tyrosine hydroxylase were strictly coexpressed also in SK-N-AS-derived tumors, whereas IGF2 was expressed by more cells, as expected. Tumor cell expression of this gene was neither ubiquitous nor constant in intensity, yielding a mosaic pattern of expression with no evident vascular dependence. Apart from this general and atypical mosaic pattern of IGF2 expression the gene was, however, discernibly coexpressed with VEGFA and tyrosine hydroxylase in some tumor regions, as in the other tumors (data not shown).
|
In vivo exposure of SH-SY5Y xenografts to NITP resulted in labeling of the peripheral-most part of the IGF2, VEGFA, and tyrosine hydroxylase coexpressing zones, thus verifying that these cells were perfused but severely hypoxic at the time of labeling (Fig 5F-H
). Therefore, these xenograft experiments provide clear evidence for a link between the chromaffin-like phenotype and hypoxia.
Effects of hypoxia in monolayer culture
The role of hypoxia in provoking changes in expression of IGF2 and GAP-43 in xenografted tumors was addressed further by subjecting the respective cell lines to various levels of oxygenation in monolayer culture. Before examining a range of oxygen tensions, we first examined the effects of very low pO2 levels over long periods on cell survival. Near-complete oxygen depletion was accomplished with the BDD GasPac Plus SystemTM. Using this near-anoxic system, survival times varied between 24 h for SK-N-DZ cells, 4 days for SH-SY5Y cells and 12-14 days for SK-N-AS cells after continuous exposure with a surplus of medium. We therefore exposed cells to low oxygen for 48 h, allowing experiments on all the cell lines at all tested pO2 levels except for near-anoxia in SK-N-DZ cells.
Under near-anoxic conditions RNase protection analysis of SH-SY5Y cells revealed a 110-fold increase of IGF2 transcripts and a decrease of GAP-43 expression to 8% of control values (normoxia), whereas expression of VEGFA increased 13-fold (Fig. 6
A). We analyzed expression of these genes at other levels of oxygenation and analyzed the expression of LDH-A and GLUT-1 (Fig. 6A
), two hypoxia-regulated genes coding for proteins involved in cell metabolism (28)
. These data indicated pO2 threshold levels for increased expression ranging from 2% (LDH-A, GLUT-1, VEGFA) to 1% for IGF2 and a pO2 threshold for decreased expression of GAP-43 at 0.3%. The data support a role for hypoxia in the transition of tumor phenotype, with reciprocal changes in neurone-specific GAP-43 expression and chromaffin-specific IGF2 expression. These data indicate these changes occur in the same or slightly lower pO2 range as those responsible for physiological angiogenesis and enhanced glycolysis. SH-SY5Y cells were also tested for the effects of 48 h culture with glucose- or serum-free media, representing other forms of cellular stress. Neither IGF2 nor VEGFA were induced under these conditions (data not shown).
|
SK-N-AS and SK-N-DZ cells were also analyzed for pO2 dependence of expression of these genes, with broadly similar results (Fig. 6B, C
). A maximal 15-fold increase of VEGFA transcript levels was seen in anoxic cells of both cell lines. GAP-43 expression was maximally decreased by 94 and 79%, respectively, whereas the effects of near-anoxia on IGF2 transcript levels were less dramatic, presumably due to high normoxic levels in SK-N-AS cells (4-fold increase by near-anoxia; Fig. 6B
) and to lack of detectable expression of IGF2 in SK-N-DZ cells at pO2 levels > 0.1% (4-fold increase between 0.1%/48 h and near-anoxia/24 h; Fig. 6C
). The pO2 threshold levels for changes in expression of the respective genes were generally lower for SK-N-DZ cells (0.3-0.1%) than for SK-N-AS (5-1%) and SH-SY5Y cells (2-0.3%) (Fig. 6A-C
).
We also addressed whether these hypoxia-induced changes in gene expression were reversible in SH-SY5Y cells. IGF2 and GAP-43 expression returned to normoxic values within 24 h after reoxygenation with kinetics that were slightly slower than the reversion of VEGFA expression (Fig. 6D
).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 30)
. Second, the severity of tumor hypoxia correlates with poor prognosis even when patients are treated with surgery (31
, 32)
. Third, hypoxia increases angiogenic growth factor secretion and enhances glycolysis, adapting tumor cells to the hypoxic environment (33
34
35)
. Here we describe for the first time to our knowledge that hypoxia can induce major phenotypic changes, which effectively constitute differentiation.
That hypoxia underlies this phenotypic switch was first suggested because the chromaffin phenotype, strongly reminiscent of SIF type I cells, was observed in regions predicted to be hypoxic on the basis of distance from the nearest blood vessel. Aspects of this phenotypic change showed the same relationship to the vasculature in xenografts of two neuroblastoma cell lines, providing a valuable model system to investigate this further. In xenografts we exploited an increasingly used (23
, 36)
bioreductive marker (NITP) to confirm that the chromaffin-featured regions were indeed hypoxic. Both in the clinical material and in xenografts, many other microenvironmental variables will alter with distance from blood vessels besides pO2, including glucose and pH. In this setting it would be conceivable that the association was not causative—for example, a developmental program that led to chromaffin differentiation might induce migration away from blood vessels or reduce angiogenic signaling. We were able to address these issues by examining the behavior of neuroblastoma cell lines in short-term monolayer culture. These experiments showed that changes in gene expression linked to the chromaffin phenotype were strikingly induced by hypoxia. Despite the insights from the model systems, they appear not to capture all aspects of the phenotypic changes we observed in clinical material. For example, although we observed some morphological features of the chromaffin phenotype in xenografts, we did not see the nuclear changes or chromogranin A reactivity that was found in natural tumors. The monolayer experiments represent a simplified, short-term system in which we could isolate the effect of hypoxia. That the changes in gene expression were reversed upon return to normal oxygenation should therefore be interpreted cautiously, since in the in vivo setting we have no direct evidence that increased oxygenation can reverse the phenotype. Nevertheless, it seems appropriate in light of this to term the phenomenon metaplasia rather than differentiation, since it may be reversible (37)
.
An interesting question is whether the lack of chromaffin features in 50% of tumors is because they lack regions that are sufficiently hypoxic or whether there are hypoxic regions without chromaffin-like differentiation. We favor the latter possibility for two reasons. First, the vascular arrangement was broadly similar in tumors with and without chromaffin features. Second, of the six neuroblastoma cell lines we examined in xenograft experiments, three did not express IGF2 in regions of VEGF expression.
We initially examined IGF2 and GAP-43 gene expression because of their specificity in the fetal nervous system. In that setting, sympathetic neuroendocrine cell types are unique in their expression of IGF2, with a particularly high expression exhibited by paraganglia and SIF cells, whereas expression of GAP-43 is specific for neuronal cells (reviewed in refs 2
, 38
). The remarkable oxygen-responsive expression patterns of these genes in neuroblastoma cell lines was therefore of considerable interest. During the course of our work, another group independently found that expression of IGF2 is regulated by hypoxia in hepatocellular carcinoma (39
, 40)
. There is also evidence for hypoxia-driven IGF2 induction in normal murine cells of embryonal origin (41)
. Although GAP-43 protein levels reportedly may show a slight increase under mild hypoxia in chromaffin glomus caroticum cells (42)
, the robust down-regulation of GAP-43 in response to more marked hypoxia found here has not been recognized previously.
Full understanding of the potentially complex relationships between hypoxia, differentiation, and expression of these genes in neuroblastoma will require further work. Two alternative hypotheses could be useful in this regard. In the first, reduced oxygenation directly regulates IGF2 and GAP-43 expression in an unconditional fashion in neuroblastoma. In the second, reduced oxygenation induces a change in cell phenotype and the observed changes in gene expression are conditional on this phenotypic switch. In either case, the mechanisms underlying the robust, reversible suppression of GAP-43 expression in response to low oxygen tension we observed in several well-characterized cell lines should be readily addressable. This is important because to date there has been a much better understanding of activation of gene expression in response to hypoxia (43
, 44)
rather than repression, partly because the effects described previously for other genes have been small in quantitative terms and have mainly been defined in primary hepatocytes (45)
.
The present data raise the important possibility that hypoxia could act as a differentiation cue in extra-adrenal chromaffin organogenesis. In view of the physiologically low oxygen tension of embryonal/fetal tissues, this might contribute to the abundance of this cell compartment during intra-uterine development. In situations of restricted oxygen supply, a further decrease of oxygen tension could also constitute a physiological feedback mechanism by which the fetus could be protected by a more potent chromaffin stress response via an increase in extra-adrenal chromaffin cell number. There is precedence for such an augmenting effect by low systemic oxygen tension on sympathetic chromaffin development in the postnatal rat (46)
. Low oxygen is also shown to act as a cue for differentiation at early stages of sympathetic development, by promoting the transition of neural crest cells into a sympathoadrenal phenotype (47)
. Studies of the transcription factor HIF-2
suggest this specifically hypoxia-responsive transcription factor has a critical role in extra-adrenal chromaffin function. Sympathetic expression of the Epas-1/Hif2 gene during murine development was found to be specific for paraganglia, and mouse fetuses with a homozygously defect Epas-1/Hif2 suffered bradycardia and died at midgestation, although they could be rescued when supplemented with a catecholamine analog (48)
. Our own preliminary data show that the same specificity of sympathetic EPAS-1/HIF2 expression applies to human development. It will therefore be of interest to determine the role of this transcription factor in the phenomena we observed in neuroblastoma.
Clinical considerations
One important clinical conclusion of this investigation is that the chromaffin tumor phenotype is a very common feature across different categories of neuroblastic tumors of the sympathetic nervous system (ganglioneuroma included). Unlike the less frequent ganglion cell-like tumor phenotype, this SIF cell-like phenotype was seen independent of a Schwannian stroma-rich histology. Together with the experimental findings, this suggests independent mechanisms behind chromaffin metaplasia and ganglionic differentiation. It is also noteworthy that the present results differ from those of previous studies, which were based on substantially smaller number of tumors and linked a chromaffin tumor phenotype with stroma-poor tumors of an extra-adrenal origin, mainly representative of infant disease (17
, 19)
.
The recognition of a phenotypic heterogeneity within the tumor tissue of neuroblastoma, which correlates to tumor cell oxygenation, may have therapeutic implications. For instance, we have tested the cellular uptake of mIBG in SH-SY5Y xenografts, which revealed a remarkable specificity for the hypoxic/chromaffin metaplastic cell compartment (unpublished results). If this specificity also applies in clinical tumors, mIBG could be used to specifically target regions of chronic tumor hypoxia, potentially offering a useful adjunct in preventing tumor relapse and decreasing angiogenic signaling.
|
ACKNOWLEDGMENTS |
|---|
Received for publication April 24, 2002. Accepted for publication December 22, 2002.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  L. K. Martens, K. M. Kirschner, C. Warnecke, and H. Scholz Hypoxia-inducible Factor-1 (HIF-1) Is a Transcriptional Activator of the TrkB Neurotrophin Receptor Gene J. Biol. Chem., May 11, 2007; 282(19): 14379 - 14388. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Beppu, K. Nakamura, W. M. Linehan, A. Rapisarda, and C. J. Thiele Topotecan Blocks Hypoxia-Inducible Factor-1{alpha} and Vascular Endothelial Growth Factor Expression Induced by Insulin-Like Growth Factor-I in Neuroblastoma Cells Cancer Res., June 1, 2005; 65(11): 4775 - 4781. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
