Prognostic role of vasculogenic mimicry in colorectal cancer.

Abstract

Angiogenesis, as measured by degree of microvessel density, has been associated with tumor progression in many tumor types but does not always correlate with clinical outcome. In 1999, aggressive tumor cells were shown to form blood-conducting tubes not lined by endothelial cells, and this phenomenon was termed vasculogenic mimicry. We investigated angiogenesis and the presence of vasculogenic mimicry in colorectal carcinoma in relation to tumor stage, patient survival, and genetic indicators of tumor cell plasticity.
Paraffin-embedded tissue samples were examined from a study of 117 patients with colorectal carcinoma with a 12-year follow-up. Immunohistochemical analysis was used to measure microvessel density and proliferating endothelial cells and to detect vasculogenic mimicry (scored by 3 independent observers). Cell cultures from 7 colon cell lines, real-time polymerase chain reaction (PCR) on cell lines, frozen tissue material from 4 colorectal cancer patients with and 4 without vasculogenic mimicry, and fresh colorectal cancer tissue from 2 patients were used to investigate the relationship between vasculogenic mimicry and tumor cell plasticity.
Microvessel density was not a prognostic marker in our patients. We found vasculogenic mimicry in 23 (19.7%) of 117 colorectal tumor samples. Cell culture experiments and real-time PCR on human colorectal carcinoma material showed evidence for vasculogenic mimicry with overexpression of EPHA2 and LAMC2, known to be important for the tube-forming capacity of melanoma tumor cells. The presence of vasculogenic mimicry was associated with significantly shortened survival, both overall (P < 0.0001) and within intermediate cancer stages (Dukes B, P = 0.0277; Dukes C, P < 0.0001).
Vasculogenic mimicry can occur in colorectal carcinoma and appears to be comparable to vasculogenic mimicry described in other tumors. Moreover, vasculogenic mimicry in colorectal carcinoma may be a strong independent prognostic marker for survival.

Full text

Page 1

ORIGINALCONTRIBUTION
Prognostic Role of Vasculogenic Mimicry in
Colorectal Cancer
Coen I. M. Baeten, M.D., Ph.D.1• Femke Hillen, Ph.D.1
Patrick Pauwels, M.D., Ph.D.2• Adriaan P. de Bruine, M.D., Ph.D.2
Cor G. M. I. Baeten, M.D., Ph.D.3
1 Angiogenesis Laboratory, Research Institute for Growth and Development (GROW), University Hospital Maastricht,
Maastricht, The Netherlands
2 Department of Pathology, University Hospital Maastricht, Maastricht, The Netherlands
3 Department of Surgery, University Hospital Maastricht, Maastricht, The Netherlands
PURPOSE: Angiogenesis, as measured by degree of
microvessel density, has been associated with tumor
progression in many tumor types but does not always
correlate with clinical outcome. In 1999, aggressive
tumor cells were shown to form blood-conducting tubes
not lined by endothelial cells, and this phenomenon was
termed vasculogenic mimicry. We investigated
angiogenesis and the presence of vasculogenic mimicry in
colorectal carcinoma in relation to tumor stage, patient
survival, and genetic indicators of tumor cell plasticity.
METHODS: Paraffin-embedded tissue samples were
examined from a study of 117 patients with colorectal
carcinoma with a 12-year follow-up.
Immunohistochemical analysis was used to measure
microvessel density and proliferating endothelial cells
and to detect vasculogenic mimicry (scored by 3
independent observers). Cell cultures from 7 colon cell
lines, real-time polymerase chain reaction (PCR) on cell
lines, frozen tissue material from 4 colorectal cancer
patients with and 4 without vasculogenic mimicry, and
fresh colorectal cancer tissue from 2 patients were used to
investigate the relationship between vasculogenic
mimicry and tumor cell plasticity.
RESULTS: Microvessel density was not a prognostic
marker in our patients. We found vasculogenic mimicry
in 23 (19.7%) of 117 colorectal tumor samples. Cell
culture experiments and real-time PCR on human
colorectal carcinoma material showed evidence for
vasculogenic mimicry with overexpression of EPHA2 and
LAMC2, known to be important for the tube-forming
capacity of melanoma tumor cells. The presence of
vasculogenic mimicry was associated with significantly
shortened survival, both overall (P ? 0.0001) and within
intermediate cancer stages (Dukes B, P ? 0.0277; Dukes
C, P ? 0.0001).
CONCLUSIONS: Vasculogenic mimicry can occur in
colorectal carcinoma and appears to be comparable to
vasculogenic mimicry described in other tumors.
Moreover, vasculogenic mimicry in colorectal carcinoma
may be a strong independent prognostic marker for
survival.
KEY WORDS: Colorectal cancer; Angiogenesis;
Vasculogenic mimicry; Prognosis.
B
cal stage. The main predictor of outcome is the stage of
disease at time of diagnosis. The earlier Dukes classifica-
tion system1has now generally been replaced by the virtu-
allyequivalentbutmoredetailedTNMsystem2asthemost
reliableprognosticfactorinCRCtodate.Clinicaloutcome
can be predicted reliably for early (stage I/Dukes A) and
advanced stages of CRC (stage IV/Dukes D). However,
clinical outcome varies greatly for intermediate stages. Af-
ter potentially curative surgical excision, five-year survival
can range from 60 to 85% for patients with stage II/Dukes
B and 25 to 65% for those with stage III/Dukes C.2,3Alter-
native prognostic methods such as molecular predictors4,5
are not yet able to conclusively differentiate aggressive
from nonaggressive tumors in the intermediate stages
ecause of localization and late occurrence of symp-
toms, colorectal cancer (CRC) is often diagnosed
only after it has reached a relatively advanced clini-
Addressofcorrespondence:Dr.C.I.M.Baeten,DepartmentofSurgery,
Groene Hart Hospital, 2803 HH, Gouda, The Netherlands. E-mail:
Coen.Baeten@path.unimaas.nl
Dis Colon Rectum 2009; 52: 2028–2035
DOI: 10.1007/DCR.0b013e3181beb4ff
©The ASCRS 2009
2028
DISEASES OF THE COLON & RECTUM VOLUME 52: 12 (2009)

Page 2

CRC. Therefore, more specific and more sensitive prog-
nostic markers are needed.
Angiogenesis, the formation of new blood vessels via
sprouting of existing capillaries, is a process of major im-
portance in tumor growth and metastasis. Because diffu-
sion limits prevent a tumor from growing larger than ap-
proximately 2 mm in diameter without new blood vessel
formation,6tumor cells release angiogenic stimuli such as
vascularendothelialgrowthfactor(VEGF)andbasicfibro-
blast growth factor (bFGF).7,8Until recently, microvessel
density was the gold standard for measurement of angio-
genesisintumortissuesandwasthoughttobeapromising
prognostic variable. In various studies, high microvessel
density was associated with shorter survival.9,10,11How-
ever, conflicting reports also exist,12,13and the usefulness
of microvessel density as a prognostic factor is unclear.
In addition to using regular blood vessels for circula-
tion, some tumors appear able to use a means of circula-
tion that is independent of angiogenesis. In 1999, a novel
phenomenonofbloodvesselformationbymelanomacells
oncollagenwasreportedandtermedvasculogenicmimic-
ry.14In vasculogenic mimicry, fluid-conducting channels
are formed by highly invasive and genetically dysregulated
tumor cells with a capacity to form tube-like structures
which can be identified by microscopy.15These blood-
conducting structures are not lined by endothelial cells.
However, dedifferentiation of aggressive tumor cells can
lead to acquisition of endothelial characteristics.16,17This
processoftumorcellplasticityappearstocontributetothe
circulatory system in the tumor. In melanoma and other
tumors, vasculogenic mimicry was shown to be associated
with expression of endothelial markers such as Tie-1,
EphA2, LamC2, and anticoagulant molecules such as tis-
sue factor pathway inhibitors.14,17,18
In an unpublished pilot study, we observed that areas
with channels and lakes or pools of extravasated erythro-
cytesarepresentinCRCtissues.Inaccordwithotherstud-
ies,17,19,20,21,22we identified this phenomenon as vasculo-
genic mimicry and found evidence that it may be causally
related to mechanisms of tumor cell plasticity in CRC in
the same way as previously shown in melanoma. The
present study was undertaken to investigate the phenom-
enon of vasculogenic mimicry and its potential relation-
shiptotumorstageandsurvivalinaseriesofpatientswith
CRC, and to investigate whether behavior of a small sub-
group of aggressive colorectal tumors might be related to
endothelial genetic markers similar to those observed in
other tumors.
MATERIALSANDMETHODS
Patients
All samples used for these studies were archival material
fromthetissuebankoftheDepartmentofPathologyofthe
University Hospital Maastricht. They were made available
accordingtotheguidelinesforadditionalusageofencoded
remnant patient material formulated in the Code of the
DutchFederationofMedicalScientificSocieties,withper-
mission from the Medical Ethical Committee of the Uni-
versity Hospital/University of Maastricht. Three groups of
tissue samples were studied.
Group1.Toexploretherelationshipofvasculogenicmim-
icry to cancer stage and survival, we studied colorectal ad-
enocarcinoma tissue samples from 117 patients who had
receivedthediagnosisbetween1979and1982.23Thesepa-
tients had been part of 2 multicenter prospective clinical
trialsconductedtoinvestigate“notouch”surgicalremoval
(96 patients) and preoperative radiotherapy (21 patients,
only10ofwhomhadactuallyreceivedpreoperativeradio-
therapy).24,25Cancer stage was determined at the time of
diagnosisusingtheDukesclassificationsystem,whichwas
therefore also used in the current study. Because the sam-
pleswerefromstudiesconductedfrom1979through1982,
the Dukes classification system was used throughout the
present study.
In these trials, no postoperative adjuvant treatment
was applied. Cross-sections through the center of the tu-
mor were sampled, and clinical data were collected during
12 years of follow up. Current standard procedures for
determination of markers such as microsatellite instability
(MSI) and K-ras were not available at the time of research
and were therefore not performed.
Group 2. To explore whether a relationship between vas-
culogenicmimicryandgeneexpressionexistsinCRCsim-
ilar to that found in melanoma and other tumors, we in-
vestigated differences in gene expression between CRC
tumor cells with and those without vasculogenic mimicry.
We therefore examined frozen material from 30 patients
who had received a diagnosis of CRC between 2002 and
2004andhadnotundergonepreoperativetreatment.Four
of these patients had vasculogenic mimicry inside the tu-
mor tissue and were compared with 4 patients randomly
selectedfromthosewithoutvasculogenicmimicry.Wean-
alyzed gene expression in normal mucosa as well as in tu-
mor tissue from these 8 patients.
Group 3. For the specific isolation of tumor cells and en-
dothelial cells to compare gene expression patterns, we
used fresh tissue material from 2 patients who had been
identified as having vasculogenic mimicry.
Immunohistochemistry(Group1)
Serial4-?msectionsweremadefromtheparaffin-embed-
ded specimens selected from the tissue bank. The sections
weredouble-stainedfortheendothelialmarkersCD31and
CD34, and for an antibody for human Ki-67 as a prolifer-
ation marker, as previously described.23In short, the sec-
tionsweredeparaffinized,rehydrated,andblockedin0.3%
DISEASES OF THE COLON & RECTUM VOLUME 52: 12 (2009)
2029

Page 3

H2O2inmethanol.Subsequently,sectionswerecookedfor
10 minutes in citrate buffer (pH 6.0) and blocked with 5%
bovine serum albumin in phosphate-buffered saline
(PBS). The sections were incubated for one hour at room
temperaturewithKi67(1:200dilution,NeoMarkers,Free-
mont, CA), then incubated for 30 minutes with a biotin-
labeled swine anti-rabbit IgG (DAKO, Glostrup, Den-
mark),furtherincubatedfor30minuteswithahorseradish
peroxidase–labeled avidin-biotin complex (DAKO), and
finally incubated with diaminobenzidine (Sigma, Zwijn-
drecht, The Netherlands). After this first staining, a sec-
ondary cocktail of primary antibodies against endothelial
cells (anti-CD31, 1:100, DAKO; and anti-CD34, 1:50,
Monosan, Uden, The Netherlands) was applied to the sec-
tions.Thisstepwasfollowedbyincubationfor30minutes
withbiotin-labeledgoatanti-mouseIgG(DAKO)andavi-
din-biotincomplexAP(DAKO).Theboundantibodywas
visualized by the Vector Blue Alkaline Phosphatase Sub-
strate Kit III (Vector Laboratories, Burlingame, CA). In all
immunohistochemical experiments, a separate colon tis-
sue sample served as a negative control, with staining ac-
cordingtotheproceduresdescribedaboveexceptthatrab-
bit,mouse,orswineserumwasusedinsteadoftheprimary
or secondary antibody. We also performed single staining
with the endothelial markers. However, the combination
of staining for endothelial markers and Ki-67 showed the
best results. Staining with cytokeratin 20 (DAKO), a pro-
cedure routinely performed at our institution, was used as
a positive control. After being covered with Imsol-mount
(Klinipath, Duiven, The Netherlands), the slides were
mounted in Entellan (Merck, Darmstadt, Germany).
The number of vessels was measured at 200? magni-
fication in 3 random areas of 0.22 mm2each (Zeiss, Ger-
many). Vasculogenic mimicry was considered to be
present when areas with extravascular erythrocytes were
observed in between or in direct contact with tumor cells,
withoutanendothelialcelllining,andwithoutanyformof
glandularformationbythetumorcells.Thisphenomenon
is best described as the presence of intratumoral extravas-
cular erythrocytes and is in accordance to previously
shown vasculogenic mimicry.17The hematoxylin and eo-
sin–stained sections and the adjacent sections stained for
endothelialcellswerescoredforvasculogenicmimicryby3
independent observers: a pathologist (AdB), the principle
investigator (CB), and a molecular biologist; all were
blinded to clinical data and other results.
FrozenandFreshCRCTissuePreparation(Groups2
and3)
Single-cell suspensions were generated by incubation with
collagenase/dispase (Invitrogen, Breda, The Netherlands)
5%inPBSat37°Cfor30minutesinthepresenceofDNase
(1mg/ml; Qiagen, Venlo, The Netherlands). The cell sus-
pension was subsequently filtered through a nylon filter,
resuspended and put on a Ficoll Paque (Amersham Bio-
sciences, Uppsala, Sweden), followed by centrifugation at
400?g. The single cell suspension was subsequently incu-
bated with CD31 and CD34 antibody dilutions for one
hour, then incubated with anti-mouse labeled Dynabeads
(Dynal Biotech, Oslo, Norway). After the endothelial cells
were shifted from the tumor suspension with a magnet,
and after immunohistochemical control with a cytospin
andstainingwith3,3?-diaminobenzidine(DAB),bothsus-
pensions were used for RNA isolation.26
RNAIsolationandcDNASynthesis
Total RNA was isolated from the cell lines listed below, as
well as from tissues and cell suspensions, by use of the
RNeasy RNA isolation kit (Qiagen) according to the sup-
plier’s protocol. Possible genomic DNA contaminations
were removed by DNase treatment with the RNase-free
DNase set (Qiagen). The purity and integrity of the RNA
was checked on a Nanodrop spectrophotometer (Thermo
Scientific, Wilmington, DE) and by gel electrophoresis ac-
cording to standard procedures. RNA was reverse tran-
scribed with M-MLV reverse transcriptase (Promega, Lei-
den, The Netherlands) according to standard procedures.
cDNA was stored at –20°C.
PCRandQuantitativeReal-TimePCR
Semi-quantitativepolymerasechainreaction(PCR)ofthe
genes for beta-actin (ACTB), tyrosine kinase Tie-1 (TIE1),
tissue factor pathway inhibitor 1 and -2 (TFPI1 and
TFPI2), ephrin-A2 (EPHA2), and laminin-5 gamma 2
(LAMC2)wasperformed.17Thegeneswereamplifiedover
30 cycles. The PCR products were analyzed by electro-
phoresis on a 1.5% agarose gel.
Quantitative real-time PCR was performed on an ABI
PRISM 7700 Sequence Detection System apparatus (Ap-
pliedBiosystems,FosterCity,CA).SYBRGreenPCRMas-
ter Mix (Applied Biosystems) was used in admission of 40
ng cDNA and 800 nM of forward and reverse primer sets,
with a standard two-step amplification protocol (Ta at
60°C)followedbyameltingcurveanalysis.Theexpression
ofeachtargetgenewasnormalizedtothebeta-actinhouse-
keepinggene.Allexperimentsweredonetwiceinduplicate
on separate plates and a negative control with sterile water
was used.
CellCulturesandThree-DimensionalCulture
We used 7 colon carcinoma cell lines (HT29, CaCo2,
SW480, SW742, HCT116, RKO, and HCA-7) and 4 mela-
noma cells lines (C8161, C81–61, MUM2-B, and MUM2-
C). From each cell line, 3 ? 104cells were seeded on a
96-well plate coated with rat tail collagen. Each day, the
medium (DMEM/RPMI-1640 with 10% FSC and 1% glu-
tamine) was replaced; after 8 days, photographs were
taken.
2030
BAETEN ET AL: VASCULOGENIC MIMICRY IN COLORECTAL CANCER

Page 4

RNA was isolated from all cell lines and used for PCR
and real time PCR as described above.
StatisticalAnalysis
We used the Student t test to determine the significance of
differences between groups regarding microvessel density
anddifferencesbetweentumorareaswithandwithoutvas-
culogenic mimicry in regard to microvessel density.
Kaplan-Meier analyses and log rank tests were used for
comparison of survival curves. The quantitative real-time
PCR data were analyzed using the Mann-Whitney U test.
Thekappastatisticwasusedtomeasurenonrandomagree-
ment among observers.
All calculations were performed using SPSS (version
11), and P values ? 0.05 were considered statistically sig-
nificant.
RESULTS
PatientCharacteristics
Group1consistedof59menand58womenwithamedian
age at diagnosis of 70 (range, 29 to 87) years. The median
survival was 4.5 (range 0.1 to 12) years. Two patients pre-
sentedwithDukesA,72patientswithDukesB,32patients
withDukesC,and11patientswithDukesDstagedcancer.
Themeantumorsize?SD(SD)was46.4?22.3mm.The
primary CRC location of 113 tumors was known: 18 tu-
mors were located in the ascending colon, 16 in the colon
transversum, 32 in the descending colon, and 47 in the
rectosigmoid(17intherectum).Atotalof79patientsdied
duringthe12yearsoffollow-up(meansurvival,3.2years);
49 deaths were due to metastatic disease (mean disease-
specific survival, 2.7 ? 1.6 years). Mean disease-free sur-
vival was 7.9 ? 4.2 years.
VasculogenicMimicryinColorectalCarcinomas:Group1
AnalysisofGroup1showedvasculogenicmimicrywithno
signs of coagulation in 23 (19.7%) of the 117 tumor tissue
samples (three independent observers, kappa 0.884, P ?
0.001) (Fig. 1). Of the 23 samples with vasculogenic mim-
icry, 4 were from rectal and 19 from colon carcinomas.
Numerous proliferating tumor cells were found in the ar-
eas of vasculogenic mimicry, as indicated by brown stain-
ing with DAB showing Ki-67–positive cells (Fig. 1). This
phenomenon occurred mostly in central parts of the tu-
mor.Becauseasubgroup(10ofthe21patientsincludedin
the radiotherapy trial) had received preoperative radio-
therapy,weanalyzedtheresultsofthissubgroupseparately
to determine whether they biased the overall data. Of the
10 patients who received radiotherapy, 2 showed vasculo-
genic mimicry and did not influence the overall results of
the group of 117 patients.
The presence of vasculogenic mimicry was strongly
correlated (Spearman rho test; P ? 0.001) with the Dukes
stage of disease (Fig. 2F): Whereas vasculogenic mimicry
was absent in the 2 patients with Dukes A, it was observed
in 5 of 72 patients (6.9%) with Dukes B, in 9 of 32 patients
(28.1%)withDukesC,andin9of11patients(81.8%)with
Dukes D (Fig. 2F). Overall survival was significantly
shorter in patients with vasculogenic mimicry (P ?
0.0001) (Fig. 2B), with an overall five-year survival rate of
40% in patients without vasculogenic mimicry compared
with 0% survival in patients with vasculogenic mimicry.
Moreover, within the groups of intermediate-stage tu-
mors,survivalratesweresignificantlylowerwhenvasculo-
genic mimicry was present (Dukes B, P ? 0.0277, Fig. 2C;
Dukes C, P ? 0.0001, Fig. 2D).
Nosignificantcorrelationswerefoundbetweenvascu-
logenic mimicry and patient age or gender, tumor loca-
tion, or tumor size.
FIGURE 1. Vasculogenic mimicry in colorectal cancer tissue
indicated by the presence of intratumoral extravascular erythrocytes
(next to regular endothelial cell–lined blood vessels. Hematoxylin
and eosin–stained colon tumor sections with vasculogenic mimicry
(A, C, and E), and Ki-67 (DAB) double-stained with CD31/CD34
alkaline phosphatase blue adjacent tissue sections (B, D, and F).
Magnification of panels A-D, 200?; panels E and F, 400?
(enlargements of C and D, respectively) (bar represents 100 ?m).
Regular blood vessels are sharp by arrowheads, and areas of
vasculogenic mimicry are identified by arrows.
DISEASES OF THE COLON & RECTUM VOLUME 52: 12 (2009)
2031

Page 5

Furthermore, in this group of 117 patients we did not
find a significant correlation between microvessel density
(as a measure of angiogenesis) and survival (Fig. 2A).
When investigating the relationship between vasculogenic
mimicry and microvessel density, we did not find a higher
overall microvessel density (bloodvessels/mm2) in CRC
samples with vasculogenic mimicry compared with CRC
samples without vasculogenic mimicry: 63 vs. 61 (P ?
0.512). However, within the tumors that showed vasculo-
genicmimicry,themean(?SD)numberofbloodvessels/
mm2in the areas with vasculogenic mimicry was signifi-
cantly lower than in areas without vasculogenic mimicry:
41.24 ? 14.65 vs. 60.93 ? 17.37 (P ? 0.05).
EvidenceforTumorCellPlasticityinColorectalCancer:
ExplanationforthePresenceofVasculogenicMimicry
Analysis of the Group 2 samples (Fig. 3A) showed that
EPHA2 and LAMC2 were significantly overexpressed in
colon cancer tissues positive for vasculogenic mimicry
compared with those negative for vasculogenic mimicry,
and compared with normal mucosa (all P ? 0.05).
To exclude the possibility that the measured gene ex-
pression was due to expression levels in endothelial cells,
tumor cells and endothelial cells from Group 3 samples
were separated by magnetic sorting,26and the expression
levels of the genes in the subpopulations were compared.
Both tumor cells and endothelial cells expressed the genes
of interest (Fig. 3B), with higher expression levels of
EPHA2 and LAMC2 in the tumor cells.
Since we found overexpression of EPHA2 and
LAMC2, both known to be of major importance for the
tube-formingcapacityofmelanomatumorcells,14,27,28we
studiedtheexpressionofthesemoleculesinthecoloncan-
cer cell lines RKO, SW742, SW480, CaCo-2, HCA-7,
HT-29andHCT116.EPHA2andLAMC2werefoundtobe
expressedinSW742,SW480,HCT-116,HCA-7,RKO,and
CaCo-2,butnotinHT29(Fig.3C).Thisfindingcorrelated
with tube-forming capacity of these cell lines on a three-
dimensionalcollagenmatrixinvitroandwasconfirmedby
melanomacelllines(Fig.3D-I).Onecellline,RKO,which
expressed low levels of LAMC2 (Fig. 3C), was found to
haveanintermediateactivityattheleveloftubeformation
(not shown).
Expression of TFPI1 and TFPI2 was higher in tumor
cells of patients with vasculogenic mimicry than in tumor
cells and normal mucosa of patients without vasculogenic
FIGURE 2. Comparison of Kaplan-Meier survival curves (log-rank test) in patients with colorectal cancer (CRC) staged according to the Dukes
system (Group 1). A. Patients with microvessel density (MVD) values above the group mean vs. those with values below the group mean (P ?
0.5156). B.-E. Patients with tumors positive for vasculogenic mimicry (VM?) vs. those with no vasculogenic mimicry (VM–): B. All cancer stages
(P ? 0.0001). C. Dukes B (P ? 0.0277); D. Dukes C (P ? 0.0001); E. Dukes D (P ? 0.9431). PAR ? number of patients at risk. F. Number of patients
and mean survival with and without vasculogenic mimicry according to Dukes cancer stage.
2032
BAETEN ET AL: VASCULOGENIC MIMICRY IN COLORECTAL CANCER

Page 6

mimicry, and the differences for TFPI1 were significant
(P?0.05)(Fig.3A).TFPI1wasalsooverexpressedintube-
forming tumor cell lines (Fig. 3C).
DISCUSSION
Currently, adjuvant treatment is not generally recom-
mended for the large group of patients with Dukes B/stage
II colorectal tumors.29However, because the recurrence
rate for such tumors is approximately 25%, there is a great
need for further discrimination of aggressive tumors
within this stage. Although detection of MSI, K-ras, and
othermarkershasshownclinicalrelevance,30thereisstilla
need for new immunohistochemical approaches and gene
profiling of colon carcinomas to identify appropriate ad-
ditional prognostic markers.31,32In agreement with Bossi
etal.12andGaoetal.,13wedidnotfindmicrovesseldensity,
a parameter that presumably reflects angiogenic potential,
tobecorrelatedwithcancerprogressionorpatientsurvival
in our colorectal cancer patients.
Maniotis et al.14found that highly aggressive mela-
noma tumor cells were characterized by the formation of
vascular-like tubes, which they termed vasculogenic mim-
icry. A similar process has been described in various carci-
nomas, such as breast33and ovarian carcinoma,34Ewing
sarcoma,17and gastrointestinal stromal tumors (GIST),20
We observed vasculogenic mimicry in our patients with
CRC. To our knowledge, vasculogenic mimicry has not
been previously described in colorectal tumors. The pres-
ence of vasculogenic mimicry significantly predicted poor
prognosis overall in our patients. This finding may have
beenrelatedtothecorrelationbetweenvasculogenicmim-
icry and cancer stage. However, the presence of vasculo-
genic mimicry also lowered survival rates within the inter-
mediate stages (Dukes B and C), with significant
differences in long-term survival between patients with
and those without vasculogenic mimicry within Dukes
stage B, as well as C. This was true even although in Dukes
stage B the detrimental effect of vasculogenic mimicry was
not apparent until the second year of follow-up, possibly
because of the small number of patients with vasculogenic
mimicry. These findings suggest an independent effect of
vasculogenic mimicry on prognosis.
Itmightbearguedthatthepresenceofareas(mostlyin
central parts of the tissue) of intratumoral extravasated
erythrocytes(vasculogenicmimicry)isaresultofmechan-
icaldamage.Itisunlikelythatthepresenceofvasculogenic
mimicry in our study was due to bleeding caused by the
surgical procedure, because most tissues (96 of 117) were
from patients who underwent no-touch surgery.24
FIGURE 3. Tumor cell plasticity as an explanation for vasculogenic mimicry in colorectal cancer (CRC). A. Relative gene expression of EPHA2,
LAM2C, TIE1, TFPI1, and TFPI2 corrected for beta-actin expression in normal mucosa from CRC patients who had tumors with or without
vasculogenic mimicry (n ? 8), in tumor samples without vasculogenic mimicry (VM–) (n ? 4) and in tumor samples with vasculogenic mimicry
(VM?) (n ? 4) (Group 2). # ? significant difference between VM? and normal mucosa (P ? 0.05); ? ? significant difference between VM? and
VM– (P ? 0.05). B. Gene expression determined by quantitative real-time PCR in isolated tumor cells (TC) and endothelial cells (EC) from 2 fresh
VM? colon tissue samples (Group 3). C. Gene expression examined with semiquantitative PCR in 7 colon cancer cell lines (SW-742, SW-480,
CaCo-2, HCA-7, RKO, HT-29, and HCT-116) and 4 melanoma cell lines (MUM2-B, MUM2-C, C8161, and C81–61). The MUM2-B and C8161
represent positive control cell lines and MUM2-C and C81–61 represent negative control cell lines. D.-I. Tube formation by colon cancer cell
lines examined on three-dimensional cultures, with MUM2-B and MUM2-C as positive and negative controls.
DISEASES OF THE COLON & RECTUM VOLUME 52: 12 (2009)
2033

Page 7

Furthermore,iftheextravascularbloodhadbeentheresult
of vascular damage, signs of coagulation would have been
expected.Wedidnotobserveanycoagulationorthrombus
formation in the areas of vasculogenic mimicry. Similar to
what was described in melanoma,35we observed the over-
expression of the tissue factor pathway inhibitor gene
TFPI1, and to a lesser extent TFPI2, in both vasculogenic
mimicry–positive tumor tissues and in tube-forming tu-
mor cell lines, explaining the lack of coagulation activity.
The previous studies14,17,18described a process in
which plasticity and dedifferentiation led tumor cells to
acquire the characteristics genetic profile of endothelial
cells, resulting in the formation of vascular-like tubes. The
areas of vasculogenic mimicry found in CRC resembled
thevascular-liketubeformationobservedinthosestudies.
Blood lakes similar to those observed in breast cancer33
and Ewing sarcoma17were found in CRC, although the
lining of tumor cells was more uneven. Colorectal tumors
do not show periodic acid Schiff reagent (PAS)–positive
patterns as observed in melanoma,14probably because the
normal histology of the colorectum is different from skin
histology. However, the overlapping gene profiles among
CRC, melanoma, and Ewing sarcoma cells (i.e., LAMC2,
EPHA2, and TFPIs) show great similarity. We therefore
favor the view that the vasculogenic mimicry observed in
CRC is induced by a similar phenomenon, based on plas-
ticityoftumorcellsanddedifferentiationtoamoreaggres-
sive phenotype.
Because of the differences in survival observed be-
tween vasculogenic mimicry–positive and vasculogenic
mimicry–negative tumors, vasculogenic mimicry detec-
tion may have prognostic relevance in CRC, especially
within the group of Dukes B–staged tumors. Since we
know that vasculogenic mimicry is associated with meta-
static disease in melanoma, further research on vasculo-
genic mimicry and metastasis in CRC is needed. Further-
more, increased knowledge may affect treatment and
follow-upofCRCpatientswithvasculogenicmimicry.The
current treatment of CRC with angiostatic (anti-VEGF)
agents may need to be reconsidered, since we know from
melanoma that such treatment can potentially facilitate
vasculogenic mimicry.36,37
In conclusion, the study presented here showed that
vasculogenicmimicrycanoccurinCRCanditspresenceis
associated with poor prognosis. Vasculogenic mimicry
may become a useful prognostic marker in addition to
TNM classification, although future studies on vasculo-
genic mimicry and metastasis should be performed.
ACKNOWLEDGMENTS
We are grateful to Prof. Dr. T. Wiggers, founder of the colo-
rectal cancer tissue bank of the Department of Pathology of
the University Hospital Maastricht.
REFERENCES
1. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer
screening: clinical guidelines and rationale. Gastroenterology
1997;112:594–642.
2. GreeneF,PageD,FlemingI,etal.AJCCCancerStagingManual;
New York, NY, Springer, 2002.
3. Hermanek P. pTNM and residual tumor classifications: prob-
lems of assessment and prognostic significance. World J Surg
1995;19:184–90.
4. McLeodHL,MurrayGI.Tumourmarkersofprognosisincolo-
rectal cancer. Br J Cancer 1999;79:191–203.
5. Graziano F, Cascinu S. Prognostic molecular markers for plan-
ningadjuvantchemotherapytrialsinDukes’Bcolorectalcancer
patients: how much evidence is enough? Ann Oncol 2003;14:
1026–38.
6. Folkman J. What is the evidence that tumors are angiogenesis
dependent? J Natl Cancer Inst 1990;82:4–6.
7. Shing Y, Folkman J, Sullivan R, Butterfield C, Murray J, Klags-
brun M. Heparin affinity: purification of a tumor-derived cap-
illary endothelial cell growth factor. Science 1984;223:1296–9.
8. Gospodarowicz D, Bialecki H, Greenburg G. Purification of the
fibroblastgrowthfactoractivityfrombovinebrain.JBiolChem
1978;253:3736–43.
9. HillenHF,HakLE,Joosten-AchjanieSR,ArendsJW.Microves-
sel density in unknown primary tumors. Int J Cancer 1997;74:
81–5.
10. WeidnerN,FolkmanJ.Tumoralvascularityasaprognosticfac-
tor in cancer. Important Adv Oncol 1996;167–90.
11. Choi HJ, Hyun MS, Jung GJ, Kim SS, Hong SH. Tumor angio-
genesis as a prognostic predictor in colorectal carcinoma with
special reference to mode of metastasis and recurrence. Oncol-
ogy 1998;55:575–81.
12. BossiP,VialeG,LeeAK,AlfanoR,CoggiG,BosariS.Angiogen-
esis in colorectal tumors: microvessel quantitation in adenomas
and carcinomas with clinicopathological correlations. Cancer
Res 1995;55:5049–53.
13. GaoJ,KnutsenA,ArbmanG,CarstensenJ,FranlundB,SunXF.
Clinical and biological significance of angiogenesis and lymph-
angiogenesisincolorectalcancer.DigLiverDis2009;41:116–22.
14. Maniotis AJ, Folberg R, Hess A, et al. Vascular channel forma-
tionbyhumanmelanomacellsinvivoandinvitro:vasculogenic
mimicry. Am J Pathol 1999;155:739–52.
15. FolbergR,HendrixMJ,ManiotisAJ.Vasculogenicmimicryand
tumor angiogenesis. Am J Pathol 2000;156:361–81.
16. Hillen F, Kaijzel EL, Castermans K, oude Egbrink MG, Lowik
CW, Griffioen AW. A transgenic Tie2-GFP athymic mouse
model;atoolforvascularbiologyinxenografttumors.Biochem
Biophys Res Commun 2008;368:364–7.
17. van der Schaft DW, Hillen F, Pauwels P, et al. Tumor cell plas-
ticity in Ewing sarcoma, an alternative circulatory system stim-
ulated by hypoxia. Cancer Res 2005;65:11520–8.
18. Sharma N, Seftor RE, Seftor EA, et al. Prostatic tumor cell plas-
ticity involves cooperative interactions of distinct phenotypic
subpopulations:roleinvasculogenicmimicry.Prostate2002;50:
189–201.
19. Sun B, Zhang D, Zhang S, Zhang W, Guo H, Zhao X. Hypoxia
influences vasculogenic mimicry channel formation and tumor
invasion-related protein expression in melanoma. Cancer Lett
2007;249:188–97.
2034
BAETEN ET AL: VASCULOGENIC MIMICRY IN COLORECTAL CANCER

Page 8

20. SunB,QieS,ZhangS,etal.Roleandmechanismofvasculogenic
mimicry in gastrointestinal stromal tumors. Hum Pathol 2008;
39:444–51.
21. Pisacane AM, Picciotto F, Risio M. CD31 and CD34 expression
as immunohistochemical markers of endothelial transdifferen-
tiation in human cutaneous melanoma. Cell Oncol 2007;29:
59–66.
22. Shirakawa K, Kobayashi H, Sobajima J, Hashimoto D, Shimizu
A,WakasugiH.Inflammatorybreastcancer:vasculogenicmim-
icry and its hemodynamics of an inflammatory breast cancer
xenograft model. Breast Cancer Res 2003;5:136–9.
23. Baeten CI, Castermans K, Hillen HF, Griffioen AW. Proliferat-
ing endothelial cells and leukocyte infiltration as prognostic
markers in colorectal cancer. Clin Gastroenterol Hepatol 2006;
4:1351–7.
24. Wiggers T, Jeekel J, Arends JW, et al. No-touch isolation tech-
nique in colon cancer: a controlled prospective trial. Br J Surg
1988;75:409–15.
25. de Bruine AP, Wiggers T, Beek C, et al. Endocrine cells in colo-
rectal adenocarcinomas: incidence, hormone profile and prog-
nostic relevance. Int J Cancer 1993;54:765–71.
26. vanBeijnumJR,RouschM,CastermansK,vanderLE,Griffioen
AW. Isolation of endothelial cells from fresh tissues. Nat Protoc
2008;3:1085–91.
27. HendrixMJ,SeftorEA,HessAR,SeftorRE.Vasculogenicmim-
icryandtumour-cellplasticity:lessonsfrommelanoma.NatRev
Cancer 2003;3:411–21.
28. Hess AR, Margaryan NV, Seftor EA, Hendrix MJ. Deciphering
the signaling events that promote melanoma tumor cell vascu-
logenicmimicryandtheirlinktoembryonicvasculogenesis:role
of the Eph receptors. Dev Dyn 2007;236:3283–96.
29. Benson AB, III, Schrag D, Somerfield MR, et al. American
Society of Clinical Oncology recommendations on adjuvant
chemotherapy for stage II colon cancer. J Clin Oncol 2004;
22:3408–19.
30. KlumpB,NehlsO,OkechT,etal.Molecularlesionsincolorectal
cancer: impact on prognosis? Original data and review of the
literature. Int J Colorectal Dis 2004;19:23–42.
31. O’DwyerPJ,EckhardtSG,HallerDG,etal.Prioritiesincolorec-
talcancerresearch:recommendationsfromtheGastrointestinal
Scientific Leadership Council of the Coalition of Cancer Coop-
erative Groups. J Clin Oncol 2007;25:2313–21.
32. Laird PW. The power and the promise of DNA methylation
markers. Nat Rev Cancer 2003;3:253–66.
33. ShirakawaK,WakasugiH,HeikeY,etal.Vasculogenicmimicry
and pseudo-comedo formation in breast cancer. Int J Cancer
2002;99:821–28.
34. Sood AK, Seftor EA, Fletcher MS, et al. Molecular determinants
of ovarian cancer plasticity. Am J Pathol 2001;158:1279–88.
35. SeftorEA,MeltzerPS,SchattemanGC,etal.Expressionofmultiple
molecularphenotypesbyaggressivemelanomatumorcells:rolein
vasculogenicmimicry.CritRevOncolHematol2002;44:17–27.
36. vanderSchaftDW,SeftorRE,SeftorEA,etal.Effectsofangiogen-
esisinhibitorsonvascularnetworkformationbyhumanendothe-
lialandmelanomacells.JNatlCancerInst2004;96:1473–77.
37. Dome B, Hendrix MJ, Paku S, Tovari J, Timar J. Alternative
vascularizationmechanismsincancer:Pathologyandtherapeu-
tic implications. Am J Pathol 2007;170:1–15.
DISEASES OF THE COLON & RECTUM VOLUME 52: 12 (2009)
2035