Peddinti, Zeine et al. Clinical Cancer Research 2007
Human Cancer Biology
Prominent Microvascular Proliferation in Clinically
Radhika Peddinti,1 Rana Zeine,2 Dragos Luca,5 Roopa Seshadri,1 Alexandre Chlenski,2 Kristina Cole,4
Bruce Pawel,4 Helen R. Salwen,2 John M. Maris,4 and Susan L. Cohn3
Abstract Purpose: Tumor vasculature is disorganized and glomeruloid microvascular proliferation (MVP)
has been identified as a poor prognosticator in some adult cancers. To determine the clinical sig-
nificance of MVP, including glomeruloid MVP in neuroblastoma, we initially examined vessel
architecture in tumor sections from 51 children diagnosed at Children’s Memorial Hospital
(CMH) and subsequently evaluated 154 neuroblastoma tumors on a tissue microarray con-
structed at Children’s Hospital of Philadelphia (CHOP).
Experimental Design: H&E sections were examined for the presence of structurally abnormal
vessels and further characterized by immunostaining for CD31and von Willebrand factor to high-
light endothelial cells and a-smooth muscle actin for pericytes. Tumors with thickened walls con-
taining a complete layer of hypertrophic endothelial cells plus additional layers of vascular mural
cells were classified as MVP positive. Associations between MVP and established clinicopatho-
logic features and outcome were assessed.
Results: In both series, MVP was significantly associated with Schwannian stroma-poor histol-
ogy (CMH, P = 0.008; CHOP, P < 0.001) and decreased survival probability (CMH, P = 0.017;
CHOP, P = 0.014). In the CHOP series, MVP was associated with high-risk group classification
(P < 0.001), although this association was not seen in the smaller CMH cohort.
Conclusions: The association between MVP and poor outcome provides further support for the
concept that angiogenesis plays an important role in determining the biological behavior of neu-
roblastoma tumors. Our results also indicate that angiogenesis is regulated differently in Schwan-
nian stroma-rich versus stroma-poor neuroblastoma tumors. Further studies investigating the
activity of angiogenic inhibitors in children with clinically aggressive stroma-poor neuroblastoma
Neuroblastoma, the most common extracranial solid tumor in cells, and the amount of Schwannian stroma strongly affects
children, is unique among pediatric cancers for its broad prognosis (2). Favorable outcome is associated with tumors
spectrum of clinical behavior (1). This tumor is composed of with abundant Schwannian stroma. In addition to tumor
two main cell types, neuroblastic/ganglion cells and Schwann histology, various genetic and biological factors have been
shown to correlate with outcome. However, the cellular
mechanisms that underlie the clinical variability observed in
neuroblastoma remain largely unknown. It is widely accepted
Authors’ Affiliations: 1Department of Pediatrics, Children’s Memorial Hospital; that malignant solid tumors must acquire a new blood supply
Robert H. Lurie Comprehensive Cancer Center, Northwestern University; 3Institute
for their progressive growth to a clinically relevant size and for
for Molecular Pediatric Sciences, University of Chicago, Chicago, Illinois; 4Department
of Pediatrics, Children’s Hospital of Philadelphia and the University of Pennsylvania metastasis (3, 4). In contrast to normal vessels, tumor
School of Medicine, Philadelphia, Pennsylvania; and 5Department of Pathology, vasculature tends to be disorganized with vessels that are
Children’s Hospitals and Clinics of Minnesota, Minneapolis, Minnesota dilated and tortuous and have uneven diameters and excessive
Received 2/2/07; revised 3/14/07; accepted 3/29/07. branching (5). Microvascular proliferation (MVP) ranges from
Grant support: Clinical Oncology Research Training grant T32 CA079447, NIH
grants R01 NS049814 and P01 CA97323, the Neuroblastoma Children’s Cancer
slight to extreme degrees. Focal proliferative buddings of
Society, Friends for Steven Pediatric Cancer Research Fund, the Elise Anderson endothelial cells resembling a renal glomerulus [glomeruloid
Neuroblastoma Research Fund, Neuroblastoma Kids, Alex’s Lemonade Stand MVPs (GMP)] have been described in some types of adult
Foundation, and The Robert H. Lurie Comprehensive Cancer Center, NIH, National cancers, and several studies have shown an association between
Cancer Institute Core grant 5P30CA60553.
the presence of GMP and shorter survival (6 – 8). Wesseling
The costs of publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked advertisement in accordance et al. (9) have shown by electron microscopy that the process of
with 18 U.S.C. Section 1734 solely to indicate this fact. MVP involves not only endothelial cell proliferation but also a
Requests for reprints: Susan L. Cohn, Clinical Sciences, Institute for Molecular significant contribution from pericytes. In the WHO grading of
Pediatric Sciences, University of Chicago, 5841 Maryland Avenue, MC 4060, astrocytic neoplasms, GMP is a criterion for increasing the
Room N114, Chicago, IL 60637. Phone: 773-702-2571; Fax: 773-834-1329;
E-mail: scohn@ peds.bsd.uchicago.edu.
tumor grade (10). However, the clinical significance of MVP
F 2007 American Association for Cancer Research. has not been evaluated previously in neuroblastoma or other
doi:10.1158/1078-0432.CCR-07-0237 pediatric cancers.
www.aacrjournals.org 3499 Clin Cancer Res 2007;13(12) June 15, 2007
Human Cancer Biology
There is significant evidence that angiogenesis contributes to Histologic evaluation. H&E-stained slides were examined histolog-
the aggressive behavior of neuroblastoma tumors. In retrospec- ically for Schwannian stroma and neuroblast differentiation. The entire
tive studies, high vascular density has been correlated with poor tissue section was evaluated from one block for each of 51 tumors from
the CMH series. Of these, 33 were neuroblastomas, 13 were
clinical outcome (11, 12). Furthermore, high levels of
ganglioneuroblastomas of which 9 were intermixed and 4 were
angiogenesis activators have been detected in clinically aggres-
nodular, and 5 were ganglioneuromas. One to four cores sampled on
sive neuroblastoma tumors (13). Conversely, increased levels of the TMA were assessed for 154 tumors adequate for judging vascular
endogenous inhibitors of angiogenesis are present in Schwan- architecture, of which 118 were neuroblastomas, 24 were ganglioneur-
nian stroma-rich tumors that are associated with favorable oblastomas of which 13 were intermixed and 11 were nodular, and 12
outcome (14, 15). Preclinical studies have also shown that were ganglioneuromas. Without knowledge of the patient’s stage,
neuroblastoma growth can be inhibited by agents that target MYCN status, or clinical course, the tumors were evaluated for
blood vessels (16 – 20). However, no correlation between morphology of blood vessels, hemorrhage, and necrosis. Blood vessels
vascular variables and survival was seen in a study reported were classified into two types according to the vessel wall structure by
by Canete et al. (21). The conflicting results most likely reflect two independent reviewers (R. Peddinti and R. Zeine). On H&E,
sections showing vessels with thickened walls containing a complete
differences in techniques used to measure vessel number, a
layer of hypertrophic endothelial cells plus additional layers of vascular
difficulty encountered in reconciling the results of studies of
mural cells were classified as MVP positive. The degree of MVP varied
other solid tumors, such as breast cancer (22). from slight to GMP, which was defined as florid proliferation of small
Because the architecture of tumor blood vessels is distinct vessels with the formation of complex glomeruloid structures (25).
from normal vasculature, we hypothesized that structurally Tumor sections that contained only thin walled vessels with no more
abnormal vessels would be identified in neuroblastoma than one layer of flat, spindle-shaped endothelial cells were classified as
tumors that are highly angiogenic and clinically aggressive. MVP negative. Special care was taken to avoid confusing tangentially cut
To test this hypothesis, we initially evaluated vessel structure vessels. MVP noted within lymph nodes or in intense inflammatory
in neuroblastoma tumors from 51 children diagnosed at infiltrates was excluded from analysis. Arteries and veins were also
Children’s Memorial Hospital (CMH) in Chicago. In this excluded from analysis.
cohort, MVP was significantly associated with stroma-poor Immunohistochemistry. The structure of the tumor vasculature was
further evaluated by examination of adjacent sections from all CMH
histology and decreased survival. These findings were con-
cases stained by immunohistochemistry for endothelial cell marker,
firmed using a neuroblastoma tissue microarray (TMA) cons- CD31, and for pericytes with a-smooth muscle actin (a-SMA). The TMA
tructed at the Children’s Hospital of Philadelphia (CHOP) was stained for von Willebrand factor to highlight endothelial cells.
that contained 154 tumor samples. Our results indicate that Sections (4 Am) were deparaffinized and heat-induced antigen retrieval
angiogenesis plays a critical role in neuroblastoma pathogen- was carried out in a steamer for 20 min in citrate buffer (Target Retrieval
esis and suggest that the process is regulated differentially in Solution (pH 6), DakoCytomation) for von Willebrand factor and
Schwannian stroma-rich versus stroma-poor tumors. CD31 and for 20 min in Target Retrieval Solution (pH 9; DakoCyto-
mation) for a-SMA. Subsequently, slides were immersed in peroxidase
block solution (DakoCytomation) and then incubated for 1 h at room
Materials and Methods temperature with the following primary antibodies: monoclonal mouse
anti-human CD31 (clone JC70A, DakoCytomation) at 1:40 dilution,
Patients and tumor specimens. The CMH patients were selected polyclonal rabbit anti-human von Willebrand factor (DakoCytoma-
based on the availability of adequate full tissue sections from the tion) at 1:40 dilution, and monoclonal mouse anti-human a-SMA
primary tumor. Sections from 51 primary neuroblastoma tumors or (clone 1A4, DakoCytomation) at 1:50 dilution. The EnVision+/
ganglioneuromas were obtained from CMH at the time of diagnosis, Horseradish Peroxidase antimouse and antirabbit detection systems
before administration of chemotherapy. Patients were diagnosed (DakoCytomation) were used to visualize antibody binding sites with
between 1986 and 2005. Medical records were reviewed to obtain 3,3¶-diaminobenzidine (DakoCytomation) as a chromogen. Sections
information about patient age, sex, tumor stage, histology, MYCN gene were counterstained with Gill’s hematoxylin.
status, and outcome. The CMH Institutional Review Board approved For hypoxia-inducible factor (HIF)-1a, antigen retrieval was done in
this study. pH 6 citrate buffer in a pressure cooker. Primary mouse monoclonal
Histology sections of 154 different neuroblastoma tumors on a TMA antibody to HIF-1a (ESEE122; Novus Biologicals, Abcam) was
constructed at CHOP were also examined (one to four cores per tumor). visualized using the Catalyzed Signal Amplification System according
All the samples used were from the initial biopsy or surgery before to the manufacturer’s instructions (DakoCytomation).
administration of chemotherapy. The cores were linked to clinical Immunofluorescence. Sections were deparaffinized and heat-induced
information, including patient age, stage, sex, histology, MYCN status, antigen retrieval was carried out in a steamer for 20 min in citrate buffer
and outcome. The CHOP Institutional Review Board approved this pH 6. Nonspecific staining was blocked by preincubation in PBS con-
study. taining 10% donkey serum. Primary antibodies for anti-CD31 (platelet/
For both patient cohorts, tumors were staged according to the endothelial cell adhesion molecule 1, M-20, Santa Cruz Biotechnology)
International Neuroblastoma Staging System (23). MYCN gene and anti-a-SMA (clone 1A4, DakoCytomation) were used at 1:100
status was determined in the Children’s Oncology Group Neuro- dilution. Immunocomplexes were visualized with corresponding FITC-
blastoma Reference Laboratory using fluorescence in situ hybridiza- donkey anti-mouse and R-PE donkey anti-goat – labeled secondary
tion (24). Tumors were histologically classified as favorable or antibody (Jackson ImmunoResearch Laboratories).
unfavorable histology according to the criteria described by Shimada Statistical analysis. Associations of MVP with various known
et al. (2). clinicopathologic prognostic factors of neuroblastoma were analyzed
Tissue microarray. The TMA was constructed from formalin-fixed, using the m2 or Fisher’s exact tests. All degrees of MVP, including GMP,
paraffin-embedded archival tissue specimens accessioned at CHOP were considered MVP positive. CMH and CHOP cohorts were analyzed
from 1974 to 2004. All tumors were reviewed by a pediatric pathologist separately due to smaller size of the tissue sections on the TMA from the
(B. Pawel). One to four samples (0.6 mm cores) of representative tumor CHOP series. Ganglioneuromas were included only in analyses related
tissue from each case and normal control tissues were included using a to stroma histology. Patients were stratified into two risk groups based
manual arrayer (Beecher Instruments, Inc.). on stage, age, and MYCN status. The non – high-risk group included
Clin Cancer Res 2007;13(12) June 15, 2007 3500 www.aacrjournals.org
Vascular Architecture in Neuroblastoma
0.001) were seen. However, MVP did not correlate with age
Table 1. Clinical and biological characteristics of (P = 0.82). In the smaller CMH series, the correlation between
the CMH cohort MVP and stage was not significant (P = 0.6), and MVP was seen
in only 5 of the 10 MYCN– amplified tumors. Interestingly,
Characteristic No. patients OS rate P
(%) (95% CI) four of these five patients have died, whereas there were no
deaths in the subset of patients with MYCN-amplified tumors
Age at diagnosis (mo)
<12 16 (34.7) 85.9 (71.1-100) NS
that lacked MVP (n = 5).
z12 30 (65.3) 79 (60.6-97.4) MVP is associated with Schwannian stroma-poor histology.
Pathology The presence of MVP was significantly associated with
NB 33 (64.7) 81.3 (67.8-94.9) NS* Schwannian stroma-poor histology in the CMH series (P =
GNB 13 (25.4) 83.3 (53.5-100)
0.008) and CHOP series (P < 0.001). All of the Schwannian
GNR 5 (9.8)
Stage stroma-dominant ganglioneuromas (CMH, n = 5; CHOP,
1 18 (39.1) 0.002 n = 12) had thin-walled vessels with no evidence of MVP
2 8 (17.4) 92.4 (82.3-100) (Fig. 1A-C). Immunohistochemistry showed a single layer of
3 8 (17.4) CD31-positive endothelial cells (Fig. 1B) and good coverage by
4 12 (26.1) 55.5 (26.3-84.8)
a-SMA – positive pericytes (Fig. 1C). Similarly, blood vessels in
Non – high-risk 31 (60.8) 100 <0.001 the Schwannian stroma-rich ganglioneuroblastoma intermixed
High-risk 15 (29.5) 50.5 (24.4-76.7) tumors were thin walled and MVP was not seen (CMH, n = 6;
Histology CHOP, n = 13; Fig. 1D-F and Fig. 2, left).
Unfavorable 24 (55.8) 61.9 (35.6-88.2) 0.079
In contrast, MVP was detected in 65% of the Schwannian
Favorable 19 (44.1) 94.1 (82.9-100)
MYCN stroma-poor tumors, and the incidence was inversely correlated
Amplified 10 (23.2) 60 (29.6-90) 0.041 with neuroblast differentiation. In the CMH series, all four
Nonamplified 33 (76.7) 89.2 (77.7-100) (100%) undifferentiated tumors had MVP. MVP was seen in 6
MVP of the 11 (54.5%) differentiating neuroblastomas (Fig. 1J-L)
Present 26 (56.5) 70.1 (51.3-88.9) 0.017
Absent 20 (43.4) 100
and11 of the 16 (68.7%) poorly differentiated tumors
(Fig. 1M-O). Of the 118 neuroblastoma tumors analyzed on
the TMA, all eight (100%) of the undifferentiated neuroblas-
Abbreviations: 95% CI, 95% confidence interval; GNB, ganglio-
neuroblastoma; GNR, ganglioneuroma; NS, not significant.
toma had evidence of MVP. Fifty-three of the 82 (64.6%)
*Only neuroblastoma and ganglioneuroblastoma were used in poorly differentiated tumors and 17 of the 28 (60.7%)
survival analysis. differentiating neuroblastomas had MVP. Similarly in the
Table 2. Clinical and biological characteristics of
patients with nonamplified MYCN stage 1, 2, and 3 tumors and infants
the CHOP cohort
with stage 4 and 4s neuroblastoma that lacked MYCN amplification.
Patients with stage 3 MYCN-amplified tumors and children older than
Characteristic No. patients OS rate P
1 year of age with stage 4 disease were classified as high risk. Survival (%) (95% CI)
estimates were described using the Kaplan-Meier method, and survival
curves were compared among clinical and biological subgroups using Age at diagnosis (mo)
the log-rank test. Five-year overall survival estimates are reported with <12 51 (36.1) 97.3 (93.6-100) 0.014
z12 90 (63.8) 77.2 (66.7 -87.8)
corresponding SE. Cox proportional hazards regression analysis was
used to test the association between risk factors and survival. For the NB 118 (75.6) 83.7 (76.8-90.6) NS*
CMH cohort, sample size limited the analyses to single predictor GNB 24 (15.3) 88.9 (74.1-100)
models. The best two-predictor model was selected based on the score GNR 12 (7.6)
statistic for the CHOP cohort. Statistical analyses were conducted using Stage
SAS statistical software version 9.1 and S-Plus version 6.2. 1 45 (32.8) <0.001
2 39 (28.4) 90.2 (84.4-96)
3 25 (18.2)
Results 4s 1 (0.7)
4 27 (19.7) 64.4 (45.6-83.2)
Clinical and biological characteristics of the patient Risk group
cohorts. Tables 1 and 2 list the clinical characteristics of Non – high-risk 104 (73.2) 95.5 (91.2-99.9) <0.001
CMH and CHOP cohorts. In the CMH series, 15 patients had High-risk 38 (26.7) 55.5 (39.2-71.9)
high-risk disease and 31 patients had non – high-risk disease. Histology
Unfavorable 58 (42) 66.6 (53.9-79.4) <0.001
Overall survival was 82% F 6 with a median follow-up time of Favorable 80 (58) 98.6 (78.3-100)
5 years and 10 months. The CHOP series consisted of 38 high- MYCN
risk and 104 non – high-risk patients, and the overall survival Nonamplified 85 (88.5) 87.2 (79.8-94.7) 0.003
was 84% F 3. Amplified 11 (11.5) 46.6 (14.2-79.2)
MVP is associated with widely disseminated disease and high-
Present 83 (58) 77.9 (68.6-87.3) 0.014
risk group classification. Table 3 shows the relationships Absent 60 (42) 94.7 (89-100)
between MVP and stage, age, MYCN status, and risk group
classification. In the CHOP series, statistically significant *Only neuroblastoma and ganglioneuroblastoma were used in
associations between MVP and stage (P = 0.008), MYCN survival analysis.
amplification (P = 0.006), and risk group classification (P <
www.aacrjournals.org 3501 Clin Cancer Res 2007;13(12) June 15, 2007
Human Cancer Biology
CMH series, all four (100%) undifferentiated tumors had MVP. stroma-poor regions. Further evidence of hypoxic-ischemic
MVP was seen in 6 of the 11 (54.5%) differentiating changes, including nuclear pyknosis, hypereosinophilic cyto-
neuroblastomas (Fig. 1J-L) and 11 of the 16 (68.7%) poorly plasm, and ghost cells, were also noted in the populations of
differentiated tumors (Fig. 1M-O). Both immunohistochemical differentiating neuroblasts closest to the abnormal vessels. In
and immunofluorescent staining with a-SMA revealed at least response to hypoxia, tumor cells commonly adapt by up-
one but sometimes multiple layers of pericytes (Fig. 1L and O regulating HIF-1a, a major regulator of the proangiogenic factor
and Fig. 2, right). In the CMH series, GMP was extensive in seven vascular endothelial growth factor. To investigate whether the
of the differentiating and poorly differentiated neuroblastomas, HIF-1a – dependent mechanism had a role in the induction of
and staining with a-SMA revealed a prominent pericytic MVP, we stained representative sections from four neuroblas-
component (Fig. 1L and O). In undifferentiated neuroblastoma toma tumors that had extensive necrosis for HIF-1a. In all four
tumors, a slightly different pattern of vasculature was noted neuroblastoma tumor sections, there was focal up-regulation of
characterized by continuous networks of MVP (Fig. 1P-R). Tiny nuclear HIF-1a positivity in the neuroblasts intervening
microvessels were lined by endothelial cells that were small and between necrosis and MVP (Fig. 3B).
weakly positive for CD31 (Fig. 1Q). a-SMA – positive pericytes MVP is associated with poor outcome. MVP correlated
were also detected in close proximity to the microvessels in strongly with poor outcome in both cohorts. In the CMH
undifferentiated neuroblastoma tumors (Fig. 1R). series, MVP was seen in the tumors from all seven children who
The difference in the vascular architecture was most died from disease. The 5-year survival in the cohort of children
prominent in the composite ganglioneuroblastomas of the with tumors that lacked MVP (n = 20) was 100% (Fig. 4B). In
nodular type. There were five ganglioneuroblastoma nodular contrast, a survival rate of 70% F 9.6 was seen in the subset of
tumors in the CMH series for which both stroma-rich and patients (n = 26) with tumors with MVP (P = 0.017). Of the 12
stroma-poor regions were analyzed. Of the 10 nodular patients with stage 4 disease, 5 died, and all had tumors with
ganglioneuroblastomas in the CHOP series, cores from 6 evidence of MVP. Similarly, in the CHOP series, 18 of the 22
tumors had only stroma-rich tissue, whereas cores from 4 were (81.8%) patients who died had tumors with MVP (Fig. 4D).
stroma-poor areas. In both series, vessels were MVP negative in The 5-year survival in the groups with (n = 83) and without
Schwannian stroma-rich regions and MVP positive in Schwan- (n = 60) MVP was 77.9% F 4.7 versus 94.7% F 2.9,
nian stroma-poor areas (Fig. 1G-I). Pericytes were present but respectively (P = 0.014). Of the 27 patients with stage 4 disease
provided poor coverage in the vessels with MVP as highlighted in the CHOP series, there were 10 deaths, and 9 (90%) of these
by the anti-a-SMA antibody. patients had tumors with MVP.
MVP is spatially related to regions of necrosis. Interestingly,
necrosis was not detected in tumors that lacked MVP, whereas Discussion
12 of 17 (70%) of neuroblastoma tumors with MVP exhibited
frank necrosis in close proximity to the abnormal vessels. The Tumor vessels are frequently disorganized and tortuous due
neuroblasts surrounding the necrotic areas and leading up to to dysregulated angiogenesis. Structural abnormalities, such
the MVP exhibited a pseudopalisading pattern (Fig. 3A). In as GMP, are characteristic of an angiogenic tumor phenotype
nodular ganglioneuroblastomas, frank necrosis was not ob- and are associated with poor prognosis in some types of cancer
served, although hypocellular and acellular islands of neuropil (6 – 8). However, much less is known about the role of
were noted in the vicinity of MVP within the Schwannian angiogenesis in pediatric cancers, and to our knowledge, this
Table 3. Associations between clinical factors and MVP in CMH and CHOP cohorts
Characteristic CMH Total P CHOP Total P
MVP Present MVP Absent MVP Present MVP Absent
<12 14 8 26 0.35 29 22 51 0.82
z12 12 12 22 53 37 90
1, 2, 3, 4s 20 14 34 0.6 59 54 113 0.008
4 6 6 12 22 5 27
High-risk 8 7 15 0.76 31 7 38 <0.001
Non – high-risk 18 13 31 51 53 104
Amplified 5 5 10 0.73 11 0 11 0.006
Nonamplified 19 14 33 50 35 85
Dead 7 0 7 0.017 18 4 22 0.014
Alive 19 20 39 65 56 121
Unfavorable 7 6 13 0.39 45 13 58 <0.001
Favorable 8 11 19 36 44 80
Stroma-poor 23 14 37 0.008 81 42 123 <0.001
Stroma-rich 3 11 14 2 18 20
Clin Cancer Res 2007;13(12) June 15, 2007 3502 www.aacrjournals.org
Vascular Architecture in Neuroblastoma
Fig. 1. Representative sections of
human neuroblastoma (NB),
ganglioneuroblastoma (GNB), and
ganglioneuroma (GNR) stained with H&E
(A, D, G, J, M, and P), CD31 (B, E, H, K, N,
and Q), and a-SMA (C, F, I, L, O, and R).
Blood vessels are thin walled in
ganglioneuroma (A-C) and Schwannian
(D-F). In contrast, blood vessels are
structurally abnormal and show MVP in
Schwannian Stroma-poor regions of
nodular ganglioneuroblastoma (G-I) and
neuroblastoma (J-R). CD31and a-SMA
immunostaining highlight the endothelial
cells and pericytes, respectively.
Magnifications, Â400 (A-Q) and
is the first study evaluating the clinical significance of MVP in ing advanced stage disease and MYCN amplification. In both
neuroblastoma. We initially examined tumor MVP in tumor study cohorts, we also found that MVP was significantly
sections from 51 patients from a single institution and then associated with decreased survival.
confirmed our findings using a TMA that contained 154 tumor Angiogenesis has been extensively studied in various adult
samples. Other series have shown an overall survival in patients cancers, and it is well established that microvessel density
with neuroblastoma in the United States to be 57% (51-63%; is associated with prognosis in many types of neoplasms
ref. 26). Thus, the overall survival of both study groups was (27 – 29). Much less is known about angiogenesis in pediatric
more favorable than expected (CMH series, 82% F 6; CHOP cancers, but there is evidence that angiogenesis also plays a
series, 84% F 3), most likely due to a bias away from high-risk critical role in the pathogenesis of neuroblastoma and Wilms’
patients for whom tumor sections were not available because tumors (11 – 13). We showed previously that high vascular
the diagnosis was based on bone marrow studies. However, as index correlated with MYCN amplification, metastases, and
expected, worse outcome was associated with established poor outcome (11). Advanced-stage neuroblastoma has also
unfavorable prognostic clinical and biological factors, includ- been associated with high levels of angiogenic stimuli and avh3
www.aacrjournals.org 3503 Clin Cancer Res 2007;13(12) June 15, 2007
Human Cancer Biology
Fig. 2. Sections of human
ganglioneuroblastoma (left) and
neuroblastoma (right) double stained for
CD31 (red) and a-SMA (green). Thin
vessels showing flat, spindle-shaped
endothelial cells and one layer of pericytes
(left). MVP seen in vessels with enlarged
endothelial cells and multiple layers of
a-SMA ^ positive vascular mural cells
(right). Magnification, Â100 (left and right).
and avh5 integrins, both markers of active angiogenesis evaluated was seen (Fig. 3B). Recently, Holmquist-Mengelbier
(13, 30). However, vascular variables were not predictive of et al. (34) have reported an association between high levels of
survival in a cohort of neuroblastoma patients analyzed by HIF-2a expression and poor outcome in neuroblastoma.
Canete et al. (21). These investigators used a computerized However, in contrast to our results, only low to undetectable
system to assess CD34-stained sections in the richest vascular HIF-1a staining was seen in well-vascularized areas of the
area. The contrasting studies highlight the effect different tumors. The reasons for the conflicting results are unclear but
techniques can have in quantifying vessel number and show may indicate that different HIF-a proteins may be present in
how difficult obtaining reproducible results can be. In our highly vascularized versus necrotic regions of the tumor.
study, all degrees of MVP were identified readily in H&E as well Similar to other types of cancers, HIF-1a stimulates vascular
as immunostained neuroblastoma tumor sections, first on the endothelial growth factor mRNA and protein expression in
CMH samples and then reproduced in the smaller TMA cores, neuroblastoma cells (35). Interestingly, recent studies indicate
suggesting that assessing tumors for MVP may prove to be an that serum-derived growth factors, insulin-like growth factor-1,
effective method for identifying an angiogenic phenotype in and high levels of brain-derived neurotrophic factor and its
neuroblastoma tumors. tyrosine kinase receptor TrkB, also stimulate HIF-1a and
In glioblastoma multiforme, where glomeruloid microvessel vascular endothelial growth factor expression in neuroblastoma
proliferations and MVP were first described, a spatial relation- cells (36).
ship between GMP and necrosis is observed (31, 32). With the exception of composite ganglioneuroblastomas
Furthermore, the hypoxic conditions associated with necrosis (nodular), high rates of survival are associated with neuro-
leads to up-regulation of HIF-1a, a major regulator of the blastic tumors with abundant Schwannian stroma, suggesting
proangiogenic factor vascular endothelial growth factor, and that Schwann cells are capable of influencing neuroblastoma
chaotic blood vessel growth (33). In our study of neuroblas- tumor biology (37). In support of this hypothesis, Schwann
toma tumors, we also found a spatial relationship between cells are known to produce neurotrophic factors as well as a
MVP and necrosis, and nuclear HIF-1a expression in neuro- spectrum of angiogenesis inhibitors (14, 15, 38). Furthermore,
blasts closest to the necrotic areas in the four tumor sections we Schwann cell – conditioned medium is capable of inducing
Fig. 3. Sections of human neuroblastoma
stained with H&E (A) and HIF-1a (B).
Blood vessels with MVP are seen adjacent
to the areas of necrosis (A). Up-regulation
of HIF-1ais seen in the area of necrosis with
adjoining blood vessels exhibiting MVP.
N, necrosis; T, tumor cells; V, blood vessels.
Magnifications, Â50 (A) and Â600 (B).
Clin Cancer Res 2007;13(12) June 15, 2007 3504 www.aacrjournals.org
Vascular Architecture in Neuroblastoma
Fig. 4. Kaplan-Meier analysis of the two cohorts of neuroblastoma patients. A, overall survival in CMH series. B, survival by MVP in CMH series (P = 0.017). C, overall survival
in CHOP series. D, survival by MVP in CHOP series (P = 0.014).
neuroblastoma differentiation in vitro (39), and we have shown Recently, significantly increased survival has been reported in
that infiltrating mouse Schwann cells can induce differentiation patients with colon, breast, and lung cancer following
and inhibit angiogenesis in human neuroblastoma xenografts treatment with antiangiogenic agents in combination with
in vivo (40). The current study indicates that vessel structure is chemotherapy (41, 42). Emerging evidence suggests that
also influenced by Schwann cells as MVP was not detected in antiangiogenic therapy can normalize blood vessel architecture
any of the Schwannian stroma-dominant ganglioneuromas or leading to more efficient drug delivery to the tumor (43, 44).
Schwannian stroma-rich intermixed ganglioneuroblastomas. In Our results correlating MVP with poor survival in children with
contrast, structurally abnormal blood vessels were seen in 65% neuroblastoma provide further rationale for using antiangio-
of the stroma-poor neuroblastomas. Similarly, in the nodular genic strategies in this cohort of patients. Additional clinical
ganglioneuroblastomas, prominent MVP was seen in the studies testing the activity of angiogenic inhibitors alone or in
Schwannian stroma-poor areas, whereas only thin vessels were combination with cytotoxic therapy in children with clinically
present areas of the tumor that were Schwannian stroma rich. aggressive neuroblastomas are warranted.
1. Brodeur GM, Maris JM. Neuroblastoma. In: Pizzo PA, 5. Gijtenbeek JM,Wesseling P, Maass C, Burgers L, van nostic importance of glomeruloid microvascular prolif-
Poplack DG, editors. Principles and practice of der Laak JA. Three-dimensional reconstruction of tu- eration indicates an aggressive angiogenic phenotype
pediatric oncology, 4 ed. Philadelphia: Lippincott- mor microvasculature: simultaneous visualization of in human cancers. Cancer Res 2002;62:6808 ^ 11.
Raven, 2001:895 ^ 937. multiple components in paraffin-embedded tissue. 9. Wesseling P, Schlingemann RO, Rietveld FJ, et al.
2. Shimada H, Ambros IM, Dehner LP, et al. The interna- Angiogenesis 2005;8:297 ^ 305. Early and extensive contribution of pericytes/vascular
tional neuroblastoma pathology classification (the 6. Kim TS, Halliday AL, Hedley-Whyte ET, Convery K. smooth muscle cells to microvascular proliferation in
Shimada System). Cancer 1999;86:364 ^ 72. Correlates of survival and the Daumas-Duport grading glioblastoma multiforme : an immuno-light and
3. Folkman J. What is the evidence that tumors are system for astrocytomas. JNeurosurg1991 ;74:27 ^ 37. immuno-electron microscopic study. J Neuropathol
angiogenesis dependent? J Natl Cancer Inst 1990; 7. Brat DJ,Van Meir EG. Glomeruloid microvascular prolif- Exp Neurol 1995;54:304 ^ 10.
82:4 ^ 6. eration orchestrated by VPF/VEGF: a new world of an- 10. Cavenee WK, Furnari FB, Nagane M. Diffusely infil-
4. Folkman J. Angiogenesis in cancer, vascular, rheuma- giogenesis research. AmJPathol 2001 ;158:789 ^ 96. trating astrocytomas. In: Kleihues P, CaveneeWK, edi-
toid and other disease. Nat Med 1995;1:27 ^ 31. 8. Straume O, Chappuis PO, Salvesen HB, et al. Prog- tors. Pathology and genetics of tumors of the nervous
www.aacrjournals.org 3505 Clin Cancer Res 2007;13(12) June 15, 2007
Human Cancer Biology
system: WHO classification of tumors, 2nd ed. Lyon vessel density in human solid tumours. Br J Cancer 34. Holmquist-Mengelbier L, Fredlund E, Lofstedt T,
(France): IARC Press; 2000:9 ^ 54. 2002;86:1566 ^ 77. et al. Recruitment of HIF-1a and HIF-2a to common
11. Meitar D, Crawford SE, Rademaker AW, Cohn SL. 23. Brodeur GM, Seeger RC, Barrett A, et al. Interna- target genes is differentially regulated in neuroblas-
Tumor angiogenesis correlates with metastatic dis- tional criteria for diagnosis, staging, and response to toma: HIF-2a promotes an aggressive phenotype.
ease, N-myc amplification, and poor outcome in hu- treatment in patients with neuroblastoma. J Clin Oncol Cancer Cell 2006;10:413 ^ 23.
man neuroblastoma. J Clin Oncol 1996;14:405 ^ 14. 1988;6:1874 ^ 81. 35. Beppu K, Nakamura K, Linehan WM, Rapisarda A,
12. Ribatti D, Surico G,Vacca A, et al. Angiogenesis ex- 24. Mathew P,Valentine MB, Bowman LC, et al. Detec- Thiele CJ. Topotecan blocks hypoxia-inducible factor-
tent and expression of matrix metalloproteinase-2 and tion of MYCN gene amplification in neuroblastoma by 1a and vascular endothelial growth factor expression
-9 correlate with progression in human neuroblasto- fluorescence in situ hybridization: a pediatric oncolo- induced by insulin-like growth factor-I in neuroblasto-
ma. Life Sci 2001 ;68:1161 ^ 8. gy group study. Neoplasia 2001 ;3:105 ^ 9. ma cells. Cancer Res 2005;65:4775 ^ 81.
13. Eggert A, Ikegaki N, Kwiatkowski J, et al. High-level 25. Rojiani AM, Dorovini-Zis K. Glomeruloid vascular 36. Nakamura K, Martin KC, Jackson JK, et al. Brain-
expression of angiogenic factors is associated with structures in glioblastoma multiforme: an immunohis- derived neurotrophic factor activation of TrkB induces
advanced tumor stage in human neuroblastomas. Clin tochemical and ultrastructural study. J Neurosurg vascular endothelial growth factor expression via
Cancer Res 2000;6:1900 ^ 8. 1996;85:1078 ^ 84. hypoxia-inducible factor-1a in neuroblastoma cells.
14. Chlenski A, Liu S, Crawford SE, et al. SPARC is a key 26. Gatta G, Capocaccia R, Coleman MP, Ries LA,Berrino Cancer Res 2006;66:4249 ^ 55.
Schwannian-derivedinhibitor controllingneuroblastoma F. Childhood cancer survival in Europe and the United 37. Ambros IM, Zellner A, Roald B, et al. Role of ploi-
tumor angiogenesis. Cancer Res 2002;62:7357 ^ 63. States. Cancer 2002;95:1767 ^ 72. dy, chromosome 1p, and Schwann cells in the matu-
15. Crawford SE, Stellmach V, Ranalli M, et al. Pigment 27. Gasparini G, Weidner N, Bevilacqua P, et al. Tumor ration of neuroblastoma. N Engl J Med 1996;34:
epithelium-derived factor (PEDF) in neuroblastoma: a microvessel density, p53 expression, tumor size, and 1505 ^ 11.
multifunctional mediator of Schwann cell antitumor peritumoral lymphatic vessel invasion are relevant 3 8. Huang D, Rutkowski JL, Brodeur GM, et al.
activity. J Cell Sci 2001;114:4421 ^ 8. prognostic markers in node-negative breast carcino- Schwann cell-conditioned medium inhibits angiogen-
16. Kaicker S, McCrudden KW, Beck L, et al. Thalido- ma. J Clin Oncol 1994;12:454 ^ 66. esis. Cancer Res 2000;60:5966 ^ 71.
mide is anti-angiogenic in a xenograft model of neuro- 28. Depasquale I, Thompson WD. Microvessel density 39. Kwiatkowski JL, Rutkowski JL, Yamashiro DJ,
blastoma. Int J Oncol 2003;23:1651 ^ 5. for melanoma prognosis. Histopathology 2005;47: Tennekoon GI, Brodeur GM. Schwann cell-condi-
17. Ribatti D, Raffaghello L, Marimpietri D, et al. Fenreti- 186 ^ 94. tioned medium promotes neuroblastoma survival and
nide as an anti-angiogenic agent in neuroblastoma. 29. Offersen BV, Borre M, Overgaard J. Immunohisto- differentiation. Cancer Res 1998;58:4602 ^ 6.
Cancer Lett 2003;197:181 ^ 4. chemical determination of tumor angiogenesis mea- 40. Liu S, Tian Y, Chlenski A, et al. ‘‘Cross-talk’’ be-
18. Streck CJ, Ng CY, ZhangY, et al. Interferon-mediated sured by the maximal microvessel density in human tween Schwann cells and neuroblasts influences the
anti-angiogenic therapy for neuroblastoma. Cancer Lett prostate cancer. APMIS 1998;106:463 ^ 9. biology of neuroblastoma xenografts. Am J Pathol
2005;228:163 ^ 70. 30. Erdreich-Epstein A, Shimada H, Groshen S, et al. 2005;166:891 ^ 900.
19. Katzenstein HM, Rademaker AW, Senger C, et al. Integrins avh3 and avh5 are expressed by endothelium 41. Hurwitz H, Kabbinavar F. Bevacizumab combined
Effectiveness of the angiogenesis inhibitor TNP-470 of high-risk neuroblastoma and their inhibition is asso- with standard fluoropyrimidine-based chemotherapy
in reducing the growth of human neuroblastoma in ciated with increased endogenous ceramide. Cancer regimens to treat colorectal cancer. Oncology 2005;
nude mice inversely correlates with tumor burden. Clin Res 2000;60:712 ^ 21. 69 Suppl 3:17 ^ 24.
Cancer Res 1999;5:4273 ^ 8. 31. Zagzag D, Zhong H, Scalzitti JM, et al. Expression of 42. Miller JC, Pien HH, Sahani D, Sorensen AG, Thrall
20. Shusterman S, Grupp SA, Barr R, et al. The angio- hypoxia-inducible factor 1a in brain tumors: associa- JH. Imaging angiogenesis: applications and poten-
genesis inhibitor TNP-470 effectively inhibits human tion with angiogenesis, invasion, and progression. tial for drug development. J Natl Cancer Inst 2005;
neuroblastoma xenograft growth, especially in the Cancer 2000;88:2606 ^ 18. 97:172 ^ 87.
setting of subclinical disease. Clin Cancer Res 2001 ; 32. Sondergaard KL, Hilton DA, Penney M, Ollerenshaw 43. Jain RK. Normalization of tumor vasculature: an
7:977 ^ 84. M, Demaine AG. Expression of hypoxia-inducible emerging concept in antiangiogenic therapy. Science
21. Canete A, Navarro S, Bermudez J, et al. Angiogene- factor 1a in tumours of patients with glioblastoma. 2005;307:58 ^ 62.
sis in neuroblastoma: relationship to survival and other Neuropathol Appl Neurobiol 2002;28:210 ^ 7. 44. Yang Q,Tian Y, Liu S, et al. Thrombospondin-1 pep-
prognostic factors in a cohort of neuroblastoma 33. Fischer I, GagnerJP, Law M, Newcomb EW, Zagzag tide ABT-510 combined with valproic acid is an effec-
patients. J Clin Oncol 2000;18:27 ^ 34. D. Angiogenesis in gliomas: biology and molecular tive antiangiogenesis strategy in neuroblastoma.
22. Hasan J, Byers R, Jayson GC. Intra-tumoural micro- pathophysiology. Brain Pathol 2005;15:297 ^ 310. Cancer Res 2007;67:1716 ^ 24.
Clin Cancer Res 2007;13(12) June 15, 2007 3506 www.aacrjournals.org