The presence of clustered circulating

tumor cells (CTCs) and circulating

cytokines define an aggressive phenotype in
metastatic colorectal cancer
Rosa Divella, Antonella Daniele, Ines
Abbate, Antonia Bellizzi, Eufemia
Savino, Giovanni Simone, Grazia
Giannone, Francesco Giuliani, et al.
Cancer Causes & Control
An International Journal of Studies of
Cancer in Human Populations
ISSN 0957-5243
Cancer Causes Control
DOI 10.1007/s10552-014-0457-4

1 23

Your article is protected by copyright and

all rights are held exclusively by Springer
International Publishing Switzerland. This eoffprint is for personal use only and shall not
be self-archived in electronic repositories. If
you wish to self-archive your article, please
use the accepted manuscript version for
posting on your own website. You may
further deposit the accepted manuscript
version in any repository, provided it is only
made publicly available 12 months after
official publication or later and provided
acknowledgement is given to the original
source of publication and a link is inserted
to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com.

1 23

Author's personal copy

Cancer Causes Control
DOI 10.1007/s10552-014-0457-4

ORIGINAL PAPER

The presence of clustered circulating tumor cells (CTCs)

and circulating cytokines define an aggressive phenotype
in metastatic colorectal cancer
Rosa Divella Antonella Daniele Ines Abbate Antonia Bellizzi
Eufemia Savino Giovanni Simone Grazia Giannone Francesco Giuliani
Vito Fazio Gennaro Gadaleta-Caldarola Cosimo Damiano Gadaleta
Ivan Lolli Carlo Sabba` Antonio Mazzocca

Received: 10 December 2013 / Accepted: 6 August 2014

Springer International Publishing Switzerland 2014

Abstract
Purpose Colon carcinoma is a malignant tumor showing
a marked preference to metastasize to distant organs. The
presence of circulating tumor cells (CTCs) in the peripheral
blood is a prerequisite for the formation of distant metastases. However, whether circulating cytokines are linked to
the circulation of tumor cells, as individual cells or clusters,
remain unclear. In this study, we investigated the circulating levels of TGF-beta, CXCL1, VEGF and PAI-1 as
potential bioindicators of the presence of CTCs in patients
with metastatic colon cancer.
R. Divella (&)
A. Daniele
I. Abbate
A. Bellizzi

E. Savino
Clinical Pathology Laboratory, Department of Experimantal
Oncology, National Cancer Institute, Istituto Tumori Giovanni
Paolo II, Viale Orazio Flacco, 65, 70100 Bari, Italy
e-mail: rosadive@inwind.it
G. Simone
G. Giannone
Department of Pathology, National Cancer Institute Giovanni
Paolo II, Viale Orazio Flacco 65, 70124 Bari, Italy
F. Giuliani
Medical Oncology Department, National Cancer Institute
Giovanni Paolo II, Viale Orazio Flacco 65, 70124 Bari, Italy
V. Fazio
G. Gadaleta-Caldarola
C. D. Gadaleta
Interventional Radiology and Medical Oncology Unit, National
Cancer Institute Giovanni Paolo II, Viale Orazio Flacco 65,
70124 Bari, Italy
I. Lolli
C. Sabba`
A. Mazzocca (&)
Interdisciplinary Department of Medicine, University of Bari
School of Medicine, Piazza G. Cesare, 11, 70124 Bari, Italy
e-mail: a.mazzocca@intmed.uniba.it
A. Mazzocca
National Institute for Digestive Diseases, IRCCS Saverio De
Bellis, Via Turi 27, 70013 Castellana Grotte, Bari, Italy

Methods Circulating tumor cells (CTCs) were isolated

from peripheral blood by immunomagnetic separation and
phenotypically characterized in a cohort of 103 patients
with metastatic colon cancer. TGF-beta, CXCL1, VEGF
and PAI-1 concentrations were determined by immunoassay in plasma samples from the same patients.
Results We detected two different populations of CTCs,
single cells or clusters in patients with metastatic colon
cancer. Importantly, we found that the presence of clustered CTCs is significantly associated with elevated circulating levels of TGF-beta and CXCL1 and with reduced
overall survival. Finally, we observed that circulating levels of cytokines are differently associated with the two
populations of CTCs.
Conclusions Taken together, these findings show that
detection of clustered CTCs represents a negative prognostic factor in patients with metastatic colon cancer. The
presence of clustered CTCs is associated with elevated
circulating levels of cytokines such as TGF-beta and
CXCL1. This suggests an additional role for circulating
cytokines as predictive tool for cancer prognosis and
diagnosis of minimal residual disease as well as assessment
of tumor sensitivity to anticancer therapy.
Keywords CTCs
Cell clusters
Metastasis
CXCL1

TGF-beta
VEGF
PAI-1
Abbreviations
CRC
Colon rectal cancers
CTCs
Circulating tumor cells
CXCL1
Chemokine (C-X-C motif) ligand 1
(melanoma growth stimulating activity, alpha)
VEGF
Vascular endothelial growth factor
TGF-beta Transforming growth factor beta
PAI-1
Plasminogen activator inhibitor-1

123

Author's personal copy

Cancer Causes Control

Introduction
Colorectal cancer (CRC) is a type of cancer arising in the
colon or rectum portion of large intestine in the gastro
intestinal tract. The mechanisms underlying the pathogenesis of CRC remain subject of an extensive investigation in
the field of cancer biology. It is known that CRC results
from accumulation of both genetic and epigenetic alterations leading to transformation of the normal glandular
epithelium into adenocarcinoma. The acquisition of the
invasive and metastatic phenotype allows cancer cells to
colonize distant organs, primarily to the liver and the lung
[1]. Surgical resection of liver and lung metastases is the
standard treatment of advanced CRC [24]. The potential
of a tumor cell (seed) to become metastatic depends on its
interactions with homeostatic factors in a target organ (soil)
that promote cell survival, angiogenesis and tumor growth.
According to the seed and soil theory, postulated by
Stephen Paget in 1889, there is a propensity for certain
tumors to seed in particular organs [5, 6]. To metastasize
and form secondary tumors in distant organs, tumor cells
must overcome a series of complicated steps [7]. The
success of colonization by circulating colon carcinoma
cells is mediated in part by specific adhesive interactions
between these cells and the host organ microvasculature
[8]. In fact, an important early step during hematogenous
formation of distant metastases is the arrest of circulating
tumor cells (CTCs) within the host organ [9, 10]. Recent
studies have confirmed that the target organs do express
specific chemokines, adhesion molecules and growth factors that cooperate with receptors or ligands present on
tumor cells and together contribute to organ-specific
metastasis formation [11, 12]. Chemokines and growth
factors such as chemokine (C-X-C motif) ligand 1 (melanoma growth stimulating activity, alpha) (CXCL1) and
TGFB1 are intricately associated with cellular transformation, tumor growth and increase of invasive potential
[1315]. Many chemokines, including CXCL1, are strong
inducers of chemotaxis, and there are several lines of
evidence showing that they play a role in tumor progression, for example, by increasing metastases formation in
preferential target organs [16]. In patients with late stages
of tumorigenesis, TGFB1 may favor a more aggressive
phenotype by promoting tumor growth and resistance to
apoptosis or by enhancing tumor cell motility and eventually the outgrowth of distant metastases [17]. The cellular
invasion and migration can be induced by SERPINE1 that
encoded a serine protease protein PAI-1, inhibitor and a
component of the plasminogen activator system (uPA)
[18]. SERPINE1, known as regulator of cell migration and
proliferation of endothelial cells, is of particular clinical
relevance in CRC since elevated serum and tissue levels

123

have been shown to correlate with poor prognosis [19].

Another important aspect in metastasis is the establishment
of tumor vasculature. VEGF encoded by VEGFA gene is a
key protein involved in the angiogenic switch in tumors,
and it is also known as a vascular permeability factor. In
fact, VEGF has been reported to decrease biological barrier
function, promoting vascular permeability and extravasation via VEGFVEGFR interactions [20, 21].
In this study, we investigated whether levels of circulating plasma cytokine such as TGF-beta, CXCL1, VEGF
and PAI-1 are associated with a higher detection rate of
CTCs in patients with metastatic CRC and, additionally,
whether these mediators have a role in favoring the colonization of metastatic cells to preferential sites. We finally
investigated whether the expression levels of these cytokines influence the circulation of tumor cells in an autonomous or in a collective manner. Last, we analyzed the
differences in patient survival according to the type of
CTCs detected in the peripheral blood.

Patients and methods

Patients and blood samples
In this analysis, 103 patients, 65 males and 38 females with
median age 65 years (range 3982) who underwent curative hepatic and pulmonary metastases from CRC treatable
with resection or tumor ablation, were enrolled in this
prospective study at Giovanni Paolo II National Cancer
Institute (NCI), Bari and at IRCCS Saverio DeBellis,
Castellana Grotte, Italy. Patients with synchronous liver or
lung metastases were included if the resection of metastases was performed together with that of primary tumor.
Sixty-nine patients (67 %) had liver metastases, while 34
(33 %) had lung metastases. Thirty patients (29.1 %) had
synchronous metastasis, while 73 patients (70.9 %) had
metachronous metastasis. Only five patients (4.8 %)
had synchronous presentation of liver or lung metastasis. A
total of 47 patients (45.6 %) had solitary metastasis, and 58
patients (56.3 %) had unilobular distribution of metastasis.
The clinical characteristics of patient (age, sex, therapeutic
interventions, etc.) were obtained from medical records.
In according to national and institutional standard procedures, all patients received systemic therapy. Thirty-five
patients underwent neoadjuvant chemotherapy, and 20
patients received neoadjuvant bevacizumab treatment.
Patient evaluation consisted of imaging studies (computed
tomography, positron emission tomographycomputed
tomography, chest X-ray, abdominal ultrasound) and biochemical analyses and was performed at regular intervals
depending on treatment and schedule. Tumor stage was

Author's personal copy

Cancer Causes Control

classified according to the current edition of the TNM

classification of the UICC (International Union Against
Cancer). Blood samples were taken at the day of diagnosis
from newly diagnosed patients with synchronous metastasis or from patients who developed metastases metachronously after R0-resection of the primary tumor. For
detection of CTCs, 15 ml of peripheral blood were collected from each participant in a vacutainer system with
EDTA. For analysis of cytokine, 5 ml of peripheral blood
were collected in a vacutainer system with EDTA and
plasma was immediately separated from the cellular fraction by centrifugation at 1,5009g for 10 min and frozen at
-20 C. Blood samples from 20 healthy individuals were
used as negative control. Written consent was obtained
from all patients prior to enrollment in the study, and the
Ethical Committee of the NCI approved the protocol in
accordance with once the ethical guidelines of the 1975
Declaration of Helsinki.
Enumeration and characterization of CTCs
This procedure has been previously described [22]. Briefly,
15 ml of anti-coagulated blood were centrifuged at
4009g for 35 min and buffy coats were collected into
50-ml conical tubes. Enrichment of disseminated carcinoma cells from peripheral blood was performed by positive selection of cytokeratin-7/8 expressing cells. For direct
immunomagnetic labeling of intracellular cytokeratin-7/8,
cells were permeabilized with MACS CellPerm Solution,
fixed with MACS CellFix Solution and incubated with
MACS Cytokeratin MicroBeads in MACS CellStain
Solution (Carcinoma Cell Enrichment and Detection Kit,
Miltenyi Biotec Inc., Bergish Gladbach, Germany). The
magnetically labeled cells were enriched on a positive
selection column in the magnetic field of a MACS separator. For immunocytochemical detection of carcinoma
cells in the MACS-enriched cell fraction, the cells were
first incubated with fluorescein isothiocyanate (FITC)conjugated antibody to cytokeratin and then with anti-FITC
antibody conjugated to alkaline phosphatase. These staining steps were performed in suspension before magnetic
enrichment. After MACS enrichment, cells of the magnetic
fraction were spun on slides and incubated with alkaline
phosphatase substrate.
TGF-beta, CXCL1, VEGF and PAI-1 enzyme-linked
immunosorbent assay (ELISA)
Plasma samples from patients and healthy donors were
assayed for levels of TGF-beta, CXCL1, VEGF and PAI-1
by a sandwich ELISA (Quantikine Human TGF-beta,
CXCL1, VEGF and PAI-1 Immunoassay; R&D Systems,
Inc., Minneapolis, USA) according to the manufacturers

recommendations. The absorbance of the solution produced was measured at 490 nm. The absorbance is directly
proportional to the amount of TGF-beta and CXCL1
present in the sample. A standard curve was constructed by
plotting the mean absorbance value measured for each
standard versus its corresponding concentration. The minimal detection limit was 4.61 pg/ml for TGF-beta, 10 pg/
ml for CXCL1, 5 pg/ml for VEGF and 0.059 ng/ml for
PAI-1. Each sample, standard and control, was analyzed in
duplicate. Intra-assay and inter-assay coefficients of variation for each plasma assays were less than 10 %.
Statistical analysis
The association between the detection rate of CTCs and
clinical parameters was investigated with chi-square test,
whereas the association between CTCs and plasma levels
of TGF-beta, CXCL1, PAI-1 and VEGF was analyzed with
unpaired t test and ANOVA test. Spearmans correlation
was used for the correlation analysis, and the survival
analysis was evaluated with KaplanMeier and log-rank
test. Overall survival (OS) was defined as the time between
the date of the blood sample drawing and the date of death
or the last follow-up examination. TGF-beta, CXCL1, PAI1 and VEGF cutoff values for survival analyses were
determined by ROC curve. A p value B 0.05 indicates
statistical significance. Multivariate analysis was performed with the Coxs proportional hazards model. The
four cytokines such as TGF-beta, CXCL1, PAI-1 and
VEGF, age, gender, tumor stage, metachrone versus synchron metastases, lung versus liver metastases, unilobular
versus bilobular metastases and CTCcluster versus CTClow
versus CTChigh, were the covariates included in the multivariate analysis.
All statistical analyzes were performed by Number
Cruncher Statistical System-Power Analysis and Sample
Size Software 2007 (NCSS-PASS, 329 North 1000 East
Kaysville, UT, USA).

Results
Two populations (single cells or clusters) of CTCs are
present in patients with metastatic colon carcinoma
To evaluate the presence of CTCs in patients with metastatic colon carcinoma, blood samples were assayed with
immunemagnetic beads and isolated carcinoma cells were
characterized immune phenotypically. For all patients, the
median number of CTCs was of 105 tumor cells (range
2212) per 107 leukocytes. We found two different populations of CTCs, those circulating as single cells (Fig. 1a, b)
and those circulating as clusters (Fig. 1c, d). According to

123

Author's personal copy

Cancer Causes Control

Cell Population

Fig. 1 Representative CTCs are found in the peripheral blood of

patients with metastatic colon cancer. a, b are the representative
images of single CTCs, whereas c, d are the representative images of
CTC clusters (940). Within the single-cell population of CTCs and

according to the number of cells, two groups of patients are identified:

patients (n = 27) with a low number of CTCs (CTCsLow) and patients
(n = 43) with a high number of CTCs (CTCsHigh), e

the number of CTCs found within the single-cell population, we divided two groups of patients: (1) patients
(n = 27) with a low number of CTCs, namely CTCslow and
(2) patients (n = 43) with high number of CTCs, namely
CTCshigh (Fig. 1e). By contrast, we were unable to make
any distinction within the clustered population of CTCs,
and therefore, 33 patients with clustered CTCs are represented as a unique group, namely CTCscluster. In these
patients, the number of CTCs was determined by the sum
of the individual CTCs as those present in the cluster. In
each cluster, it is present from a minimum of three to a
maximum of 25 cells. We also observed an increased
number of CTCs that were significantly associated with the
initial IIIIV TNM stages (p \ 0.0001, chi-square test),
with the number of metastasis (p \ 0.0001, chi-square test)
and with degree of cell differentiation (p \ 0.0001, chisquare test) (Table 1).

different cytokines TGF-beta, CXCL1, PAI-1 and VEGF in

patients with single-cell and clusters populations. We found
that high levels of circulating TGF-beta and CXCL1 are
associated with the presence of CTCscluster (p B 0.0001,
ANOVA) (Fig. 2a, b), whereas elevated levels of PAI-1 and
VEGF are associated with individual CTCshigh (p = 0.004
and p = 0.04 ANOVA, respectively) (Fig. 2c, d). In a preliminary analysis aimed at evaluating the correlation
between the four cytokines and the number of CTCs, we
observed that only TGF-beta is correlated with the number
of CTCs (Spearmans correlation r = 0.67, p = 0.0001)
(Fig. 3). Therefore, we carried out experiments to evaluate
whether levels of circulating TGF-beta are linked to levels
of CXCL1, PAI-1 and VEGF and found that the levels of
CXCL1 and VEGF are correlated with TGF-beta levels in
patients with metastatic colon carcinoma (Spearmans correlation r = 0.3, p = 0.001; r = 0.35, p = 0.001, respectively) (Fig. 4a, b). Moreover, we observed that the levels
of VEGF and PAI-1 are correlated with CXCL1 levels
(Spearmans correlation r = 0.39, p = 0.001; r = 0.62,
p = 0.001, respectively) (Fig. 4c, d). Taken together, these
results emphasize the correlation of CXCL1 with other
cytokines such as TGF-beta, VEGF and PAI-1 as well as the
association with the presence of both single and clustered

Circulating levels of cytokines are differently

associated with the two populations of CTCs in patients
with metastatic colon carcinoma
To study the influence of the circulating levels of cytokines
on CTCs, we measured the plasma concentration of four

123

Author's personal copy

Cancer Causes Control
Table 1 Clinicopathologic
characteristics of patients
(n = 103) and association with
the presence of CTCs

CTCslow
n = 27 (26 %)
median 4 (range
59)

CTCshigh
n = 43 (41 %)
median 80 (range
10212)

CTCsCluster
n = 33
(32 %)

65 (63.1)
38 (36.9)

17
10

26
17

22
11

5 (4.8)

II

20 (19.4)

14

III

33 (32)

15

13

IV

45 (43.6)b

20

20

5 (4.7)

98 (93.3)

22

43

33

n (%)

Age (year)

65 (3982)

Gender
Male
Female

0.8

TNM stages

0.0001

\0.0001

Cell differentiation
Well
Moderate
Site of metastasis

0.5

Liver

69 (67)

16

29

24

Lung

34 (33)

11

14

Synchronous

30 (29.1)

12

10

Metachronous
No. of metastasis

73 (70.9)

15

33

25

47 (45.6)

23

15

[1

56 (54.3)

28

24

Time of metastasis

0.1

\0.0001

a
All p values were calculated
with chi-square test

Distribution of
metastasis

Unilobular

58 (56.3)

20

22

16

Bilobular

45 (43.7)

21

17

Fifteen missing patients are in

the group of patients with
metachronous metastases

p valuea

0.09

CTCs in the blood circulation of patients with metastatic

colon carcinoma.
Correlation between circulating levels of cytokines
and clinicopathological parameters
The circulating plasma levels (mean SD) of TGF-beta,
CXCL1, VEGF and PAI-1 are reported in Table 2. The
mean values of these cytokines were much higher in
patients than in healthy donors (p = 0.0001, ANOVA test).
We found that plasma levels of PAI-1 are significantly
correlated with TNM stage (p \ 0.0001). A significant
association between female gender and circulating levels of
VEGF was also found (p \ 0.0001). In addition, plasma
levels of VEGF were inversely associated with TNM stage
(p \ 0.0001), with metachronous metastasis (p \ 0.0001)
and with the number of metastases (p \ 0.0001), as well as
neoadjuvant therapy with (p \ 0.0001) and without bevacizumab (p \ 0.0001). The levels of TGF-beta and

CXCL1 were much higher in patients with lung metastases

than that in patients with liver metastases (p = 0.02 and
p = 0.001, respectively). Lastly, TGF-beta and CXCL1 were
positively associated with 34 TNM stages (p \ 0.0001),
with bilobular distribution of metastases (p \ 0.0001)
and with number of metastases [1 and inversely associated
with synchronous metastases (p = 0.01 and p \ 0.008,
respectively).
The presence of clusters within the CTC populations is
associated with worse prognosis in patients
with metastatic colon carcinoma
To evaluate the OS in the three groups of patients (i.e.,
CTCsclusters, CTCshigh and CTCslow), we conducted a KaplanMeier analysis and found that the presence of clustered
CTCs in patients with metastatic colon cancer is associated
with reduced OS. The median OS was 12, 27 and 54 months
in CTCsclusters patients group, CTCshigh and CTCslow

123

Author's personal copy

Cancer Causes Control

TGF-

p<0,001*

CTCsCluster

CTCsHigh

CTCsLow

Amount (pg/ml)

Amount (ng/ml)

CTCsHigh

CTCsLow

250

CTCs in 10 7 leukocytes

r= 0.6, p<0,0001

200

150

100

50

0
500

1000

1500

2000

TGF- (pg/ml)

Fig. 3 Scatter plots showing the correlation between CTCs counts

and plasma levels of TGF-beta. p \ 0.0001

patients group (p \ 0.0001, log-rank test), respectively

(Fig. 5a). The median follow-up was 28 months (range
368 months) from blood sampling. For all patients, the
median OS was 26 months. Multiple variables such as gender, time of metastasis and metastatic site affecting survival
were examined by univariate analysis using the log-rank test.
We therefore established the cutoff value for TGF-beta

123

CTCsLow

p=0,04**

CTCsCluster

Fig. 2 Box plots showing the median value of circulating levels of

different cytokines in relation to the presence of CTCs. a, b show that
the amount of circulating TGF-beta and CXCL1 are associated with
the presence of clustered CTCs, whereas c,d show that the amount of

CTCsHigh

VEGF

p=0,004**

CTCsCluster

p<0,001*

CTCsCluster

PAI-1

CXCL1

Amount (pg/ml)

Amount (pg/ml)

CTCsHigh

CTCsLow

circulating are associated with the presence of CTCsHigh p *

compared to circulating tumor cell clusters; p ** compared to
CTCsHigh

(800 pg/ml), CXCL1 (410 pg/ml), PAI-1 (46 ng/ml) and

VEGF (500 pg/ml) by ROC curve. Univariate log-rank
analysis showed that shorter survival was observed in males
than in females with a median OS of 20 and 38 months
(p \ 0.0001 log-rank test), respectively (Fig. 5b). Also,
patients with metachronous metastases had a poor prognosis
compared with patients with synchronous metastasis, with
median OS of 24 months versus 44 months (p \ 0.0001,
log-rank test) (Fig. 5c). Finally, a shorter survival was
associated with hepatic metastases with a median OS of
25.5 months versus 32 months (p = 0.02, log-rank test)
observed in patients with pulmonary metastasis (Fig. 5d).
Multivariate regression analyses using Coxs proportional
hazards model revealed that CTCs (p = 0.00001,
HR = 5.867), gender (p = 0.003, HR = 2.61) and time of
metastases (p = 0.002, HR = 2.65) were independent
prognostic factors for survival of metastatic CRC patients
(Table 3).

Discussion
In this study, we have shown that levels of circulating
cytokines are useful indicators to predict both the presence

Author's personal copy

Cancer Causes Control

2000

2000

r=0,35 p=0,001

1800
1600

r=0,3 p=0,001

TGF- (pg/ml)

TGF- (pg/ml)

1600
1200
800

1400
1200
1000

800
600
400

400

200
0

0
100

200

300

400

500

600

700

500

CXCL1 (pg/ml)
2000

r=0,39 p=0,001

1500

2000

100

r=0,62 p=0,001

80

1600

PAI-1 (ng/ml)

VEGF (pg/ml)

1000

VEGF (pg/ml)

1200
800
400

60
40
20

0
100

300

500

700

CXCL1 (pg/ml)

100

200

300

400

500

600

700

CXCL1 (pg/ml)

Fig. 4 Scatter plots showing the correlation between plasma levels of a VEGF and CXCL1 b VEGF and TGF-beta, c TGF-beta and CXCL1 and
d between PAI-1 and CXCL1 in patients with metastatic colon cancer

of CTCs and their capacity to metastasize to preferential

sites in patients with metastatic CRC. In particular, we
have shown that elevated plasma levels of CXCL1 and
TGF-beta are predictors of the presence of CTC clusters in
peripheral blood, occurrence of metachronous metastases
and predictor of lung metastasis.
Most malignant tumors display an organ-specific metastatic pattern. For instance, the liver and the lungs are the
preferential metastatic sites of colon cancer, which rarely
metastasizes to bone, skin, brain, and almost never gives
rise to metastases to kidney, intestine and muscle. In contrast, other tumor entities, such as breast carcinomas, frequently form metastases in most of these organs. In a study
by Schluter et al. [23], it has been shown that CTCs adhere
to the inside of the hepatic and pulmonary microvascular
system, leaving remaining perfused vessel lumen. In contrast, in organs such as kidney, mesenteric muscle or skin,
only single cells are able to arrest within the microvasculature. For cancer cells that survive in these sites, the
crosstalk with the stromal component becomes favorable
for the outgrowth of metastatic foci [24]. Molecular factors
such as cytokines and chemokines released into the
microenvironment by both host and tumor cells affect

organ selectivity [25, 26]. Several studies have shown that

the tropism of metastatic tumor cells for specific organs can
occur through interactions between chemokine receptors
on cancer cells and their ligands in the target organs [27].
Constitutive expression of CXCL1 and its receptor CXCR2
was reported to regulate cell proliferation and the invasive
phenotype in CRC cell lines [28]. By comparing gene
signatures of breast cancer cell lines with weak or strong
metastatic tropism for the lung, Massague and colleagues
have shown that CXCL1 is one of the gene products that
promote lung metastasis [29]. In our study, we confirmed
that the expression of chemokine CXCL1 facilitates cell
seeding and outgrowth of metastases at distant sites and
that CXCL1 also plays an important role in the recruitment
of CTCs to the lung in patients with metastatic colon
cancer. Our results highlight another important aspect that
is the presence of CTC clusters in patients with metastatic
colon cancer. In fact, our study also shows that the CTC
clusters have a higher metastatic potential than single
CTCs and that the presence of clusters is a negative
prognostic factor for patients. In fact, the presence of CTC
clusters is suggested to be a potential marker of metastatic
dissemination [30]. An explanation to this is that the

123

Author's personal copy

Cancer Causes Control
Table 2 Correlation between circulating cytokines and clinicopathologic variables in patients with metastasis from CRC
n

Control group
Patients

PAI-1
(ng/ml)

p value

20

10 5

\0.00

103

45 25

01

Gender

VEGF
(pg/ml)
88 14
500 391

38

51 25

Male

65

42 25

01

646 520

60 12
410 146

p value
\0.00
01

TGF-b
pg/ml
40 10
822 390

n.s
411 126

365 108
\0.0001

TNM stages

\0.00

CXCL1
(pg/ml)

\0.0001

n.s

Female

p value

883 344
\0.0001

200 152

5
20

22 10
26 12

700 300
642 254

134 100
253 210

300 210

III

33

48 20

352 200

300 152

665 325

IV

45

57 25

256 124
n.s

420 200
n.s

0.02

Liver

69

52 26

384 163

325 123

869 246

34

54 26

453 196

415 136

1,049 374

Synchronous

30

58 20

Metachronous

73

56 25

\0.0001

n.s

No. of metastasis

458 256

\0.008
325 205

254 200

0.01
668 356

423 150
\0.0001

n.s

869 254
\0.0001

\0.0001

47

55 30

560 230

200 188

623 321

[1

56

54 25

365 245

356 123

956 325

Unilobular

58

48 26

658 260

235 158

580 260

Bilobular
Neoadjuvant therapy

45

52 27

349 230

452 122

865 245

Yes

35

51 13

No

68

52 15

Distribution of metastasis

\0.0001

n.s

Neoadjuvant bevacizumab

\0.0001

\0.0001

n.s
258 125

\0.0001

n.s
430 123

542 200

n.s
750 433

400 136
\0.0001

n.s

\0.0001

958 300
0.001

Lung
Time of metastasis

01
n.s

I
II

Metastatic site

\0.00

755 431

409 167
\0.0001

p value*

830 344
n.s

n.s

Yes

20

48 20

260 165

410 123

745 431

No

83

52 21

600 130

409 167

838 350

* All p values were calculated with ANOVA test

clustered cells are more capable of single CTCs to suppress

the anoikis and that an enhanced survival advantage of
cluster might also be afforded by the continued production
of autocrine promigratory factors such as proteases and
chemokine [31]. In this context, TGF-beta signaling may
enhance the capacity of some CRCs to invade and colonize
secondary sites increasing resistance to anoikis [32, 33].
The absence of proliferating cells within clusters suggests
that these cells may have a survival advantage, and theoretically, relatively resistant to chemotherapy compared
with proliferating single CTCs. Because of the positive
correlation found between CXCL1 and TGF-beta, we
hypothesized that these two cytokines may cooperate to the
formation of CTC clusters and the resulting collective
migration toward metastatic organs in a more efficient way
than single CTCs. Indeed, it has been shown that circulating tumor clusters may give rise to metastases without

123

extravasation, by attaching to vessel walls of arterioles and

capillaries, and proceeding to cell proliferation within the
vasculature, rupture of capillary walls and formation of
micro- or macrometastases [34]. Thus, CTC clusters may
mediate a shortcut to metastases. The spontaneous circulation of tumor cells and/or tumor cell clusters is the
hallmark of the invasive behavior of a proportion of
cancer cells. In line with this concept, the analysis of circulating TGF-beta and CXCL1 we performed in this study
has a significant impact on the recovery of CTCs and
allowed us to demonstrate that elevated plasma levels of
both TGF-beta and CXCL1 are predictive of detection of
CTC clusters in patients with metastatic CRC.
Colussi and coworkers have recently emphasize the role
of chromosomal instability, microsatellite instability and
CpG island methylator phenotype as three fundamental
mechanisms involved in initiation and development of

Author's personal copy

Cancer Causes Control

Gender

Circulating tumor cells

1,000

Probability of overall survival (%)

Probability of overall survival (%)

CTCsCluster
CTCs High
CTCs Low

0,750

0,500

All log rank

P<0,0001

0,250

1,000

Famele
Male
0,750

P<0,0001
0,500

0,250

0,000

0,000
0,0

16,3

32,5

48,8

65,0

0,0

16,3

Survival (months)

Time of metastasis
1,000

Synchron
Metachr
0,750

P<0,0001

0,500

0,250

0,000
0,0

16,3

32,5

48,8

65,0

Survival (months)

Probability of overall survival (%)

Probability of overall survival (%)

32,5

48,8

65,0

Survival (months)
1,000

Metastatic site

0,750

Lung
Liver
P=0,02

0,500

0,250

0,000
0,0

16,3

32,5

48,8

65,0

Survival (months)

Fig. 5 Cumulative survival curves comparing a clusters versus CTCsHigh and versus CTCsLow CTCs subgroups, b male and female subgroups,
c synchronous versus metachronous groups and d the pulmonary versus the hepatic metastatic site

CRCs. In addition, these mechanisms are related to

patients response to drugs, such as aspirin, which could be
used for both chemoprevention and treatment in specific
settings [35]. This is strengthen by the study of Liao et al.
[36] showing that the regular use of aspirin is efficient in
increasing survival only in colorectal patients with mutated-PIK3CA indicating that the molecular status of the
tumor is an important determinant in response to drug
treatment or chemoprevention. An interesting hypothesis,
namely the colorectal continuum, supporting the gut
biogeography (meaning the proximal versus distal colorectum) as another important determinant for the response
to therapy, has recently been proposed by Yamauchi et al.
[37, 38]. The frequency of microsatellite instability and
extensive CpG island methylation, as well as BRAF and
PIK3CA mutations, appear to increase continuously from
the rectum to ascending colon. Another evidence is provided in a recent study by Rosty et al. [39] showing the

highest incidence of site-specific KRAS-mutated tumors in

the cecum with the frequency of KRAS-mutated carcinoma
significantly higher in the cecum (39 %) than in the rest of
the colorectum (26 %). How our findings regarding the
CTCs presented in this study are correlated with the frequency of mutations and the gut biogeography would be an
interesting question to be addressed by future studies.
In conclusion, we have shown that two distinct populations of CTCs (single or clusters) can be detected in
patients with metastatic colon carcinoma and that the
presence of clusters within the CTC populations is associated with a worse prognosis in these patients. In addition,
the presence of clustered CTCs is associated with elevated
circulating levels of TGF-beta and CXCL1. Taken together, these findings may provide a powerful predictive tool
for cancer prognosis, diagnosis of minimal residual disease,
assessment of tumor sensitivity to anticancer drugs, and
personalization of anticancer therapy.

123

Author's personal copy

Cancer Causes Control
Table 3 Univariate and multivariate analysis for OS in 103 patients
with metastatic colon cancer
HR

95 % CI.

Univariate
p value**

Multivariate
p value**

PAI-1 \ 6 vs.
C46 ng/ml*

0.8

0.781.90

0.2

0.6

VEGF \500 vs.

C500 pg/ml*

0.75

0.125.2

0.5

0.7

CXCL1 \410 vs.

C410 pg/ml*

0.315

0.050.87

0.06

0.3

TGF-b \800 vs.

C800 pg/ml*

1.041

0.392.88

0.058

0.1

Age \65 vs.

C65 years

0.95

0.772.35

0.2

0.2

Gender female vs.

male

2.61

1.843.70

0.0001

0.003

Tumor stages III

vs. IIIIV

0.45

0.07

0.1

Metachrone vs.
synchron
metastases

2.65

1.923.50

0.0001

0.002

Lung vs. liver

metastases

1.53

1.092.20

0.02

0.27

Multiple vs. single

metastases

0.9

0.650.98

0.2

0.7

Unilobular vs.
bilobular
metastases

0.44

0.030.78

0.06

0.1

CTCcluster vs.
CTClow vs.
CTChigh

5.867

2.8686.20

0.0001

0.0001

0.0480.9

** Multivariate analysis was performed with the Coxs proportional

hazards model
* Cutoff value by ROC curve
Conflict of interest
peting interests.

The authors declared that they have no com-

References
1. Bardhan K, Liu K (2013) Epigenetics and colorectal cancer
pathogenesis. Cancers 5(2):676713
2. Marudanayagam R, Ramkumar K, Shanmugam V, Langman G,
Rajesh P, Coldham C, Bramhall SR, Mayer D, Buckels J, Mirza
DF (2009) Long-term outcome after sequential resections of liver
and lung metastases from colorectal carcinoma. HPB (Oxford)
11(8):671676
3. Shah SA, Haddad R, Al-Sukhni W, Kim RD, Greig PD, Grant
DR, Taylor BR, Langer B, Gallinger S, Wei AC (2006) Surgical
resection of hepatic and pulmonary metastases from colorectal
carcinoma. J Am Coll Surg 202(3):468475
4. Miller G, Biernacki P, Kemeny NE, Gonen M, Downey R, Jarnagin WR, DAngelica M, Fong Y, Blumgart LH, DeMatteo RP
(2007) Outcomes after resection of synchronous or metachronous
hepatic and pulmonary colorectal metastases. J Am Coll Surg
205(2):231238
5. Paget S (1889) The distribution of secondary growth in cancer of
the breast. Lancet 1:571573

123

6. Fidler IJ (2003) The pathogenesis of cancer metastasis: the seed

and soil hypothesis revisited. Nat Rev Cancer 3:453458
7. Balzer EM, Konstantopoulos K (2012) Intercellular adhesion:
mechanisms for growth and metastasis of epithelial cancers.
Wiley Interdiscip Rev Syst Biol Med 4(2):171181
8. Brooks SA, Lomax-Browne HJ, Carter TM, Kinch CE, Hall DM
(2010) Molecular interactions in cancer cell metastasis. Acta
Histochem 112(1):325
9. Gassmann P, Haier J (2008) The tumor cell-host organ interface
in the early onset of metastatic organ colonization. Clin Exp
Metastasis 25(2):171181
10. Gassmann P, Enns A, Haier J (2004) Role of tumor cell adhesion
and migration in organ specific metastasis formation. Onkologie
27(6):577582
11. Richmond A (2010) Chemokine modulation of the tumor
microenvironment. Pigment Cell Melanoma Res 23(3):312313
12. Spano D, Heckc C, De Antonellis P, Christofori G, Zollo M
(2012) Molecular networks that regulate cancer metastasis. Semi
Cancer Biol 22:234249
13. Balkwill F (2004) Cancer and the chemokine network. Nat Rev
Cancer 4:540550
14. Mazzocca A, Fransvea E, Dituri F, Lupo L, Antonaci S, Giannelli
G (2010) Down-regulation of connective tissue growth factor by
inhibition of transforming growth factor beta blocks the tumorstroma cross-talk and tumor progression in hepatocellular carcinoma. Hepatology 51(2):523534
15. Divella R, Daniele A, Gadaleta C, Tufaro A, Venneri MT,
Paradiso A, Quaranta M (2012) Circulating transforming
growth factor-b and epidermal growth factor receptor as related
to virus infection in liver carcinogenesis. Anticancer Res 32(1):
141145
16. Ogata H, Sekikawa A, Yamagishi H, Ichikawa K, Tomita S,
Imura J, Ito Y, Fujita M, Tsubaki M, Kato H, Fujimori T, Fukui H
(2010) GROa promotes invasion of colorectal cancer cells. Oncol
Rep 24(6):14791486
17. Imamura T, Hikita A, Inoue Y (2012) The roles of TGF-b signaling in carcinogenesis and breast cancer metastasis. Breast
Cancer 19(2):118124
18. Divella R, Lacalamita R, Tommasi S, Coviello M, Daniele A,
Garrisi VM, Abbate I, Simone G, Gadaleta C, Paradiso A, Quaranta M (2008) PAI-1, t-PA and circulating hTERT DNA as
related to virus infection in liver carcinogenesis. Anticancer Res
28(1A):223228
19. Halamkova J, Kiss I, Pavlovsky Z, Tomasek J, Jarkovsky J, Cech
Z, Tucek S, Hanakova L, Moulis M, Zavrelova J, Man M, Benda
P, Robek O, Kala Z, Penka M (2011) Clinical significance of the
plasminogen activator system in relation to grade of tumor and
treatment response in colorectal carcinoma patients. Neoplasma
58(5):377385
20. Xu B, Shen F, Cao J, Jia L (2013) Angiogenesis in liver metastasis of colorectal carcinoma. Front Biosci (Landmark Ed)
18:14351443
21. Liu W, Xu J, Wang M, Wang Q, Bi Y, Han M (2011) Tumor
derived vascular endothelial growth factor (VEGF-a) facilitates
tumor metastasis through the VEGF-VEGFR1 signaling pathway.
Int J Oncol 39:12131220
22. Bellizzi A, Sebastian S, Ceglia P, Centonze M, Divella R,
Manzillo EF, Azzariti A, Silvestris N, Montemurro S, Caliandro
C, De Luca R, Cicero G, Rizzo S, Russo A, Quaranta M, Simone
G, Paradiso A (2013) Co-expression of CD133 (?)/CD44 (?) in
human colon cancer and liver metastasis. J Cell Physiol
228(2):408415
23. Schluter K, Gassmann P, Enns A, Korb T, Hemping-Bovenkerk
A, Holzen J, Haier J (2006) Organ-specific metastatic tumor cell
adhesion and extravasation of colon carcinoma cells with different metastatic potential. Am J Pathol 169(3):10641073

Author's personal copy

Cancer Causes Control
24. Sleemana JP, Christofori G, Fodded R, Collarde JG, Berx G,
Decraeneh C, Ruegg C (2012) Concepts of metastasis in flux: the
stromal progression model. Semin Cancer Biol 22:174186
25. Lorusso G, Ruegg C (2012) New insights into the mechanisms of
organ-specific breast cancer metastasis. Semin Cancer Biol
22:226233
26. Catalano V, Turdo A, Di Franco S, Dieli F, Todaro M, Stassi G
(2013) Tumor and its microenvironment: a synergistic interplay.
Semin Cancer Biol 23:522532
27. Divella R, Daniele A, Savino E, Palma F, Bellizzi A, Giotta F,
Simone G, Lioce M, Quaranta M, Paradiso A, Mazzocca A
(2013) Circulating levels of transforming growth factor-beta
(TGF-beta) and chemokine (C-X-C motif) ligand-1 (CXCL1) as
predictors of distant seeding of circulating tumor cells in patients
with metastatic breast cancer. Anticancer Res 33(4):14911497
28. Lazennec G, Richmond A (2010) Chemokines and chemokine
receptors: new insights into cancer-related inflammation. Trends
Mol Med 16(3):133144
29. Acharyya S, Oskarsson T, Vanharanta S, Malladi S, Kim J,
Morris PG, Manova Todorova K, Leversha M, Hogg N, Seshan
VE, Norton L, Brogi E, Massague J (2012) A CXCL1 paracrine
network links cancer chemoresistance and metastasis. Cell
150(1):165178
30. Hou JM, Krebs MG, Lancashire L, Sloane R, Backen A, Swain
RK, Priest LJ, Greystoke A, Zhou C, Morris K, Ward T,
Blackhall FH, Dive C (2012) Clinical significance and molecular
characteristics of circulating tumor cells and circulating tumor
microemboli in patients with small-cell lung cancer. J Clin Oncol
30(5):525532
31. Kochetkova M, Kumar S, McColl SR (2009) Chemokine receptors CXCR4 and CCR7 promote metastasis by preventing anoikis
in cancer cells. Cell Death Differ 16(5):664673

32. Munoz NM, Baek JY, Grady WM (2008) TGF-beta has paradoxical
and context dependent effects on proliferation and anoikis in
human colorectal cancer cell lines. Growth Factors 26(5):254262
33. Mazzocca A, Fransvea E, Lavezzari G, Antonaci S, Giannelli G
(2009) Inhibition of transforming growth factor beta receptor I
kinase blocks hepatocellular carcinoma growth through neoangiogenesis regulation. Hepatology 50(4):11401151
34. Friedl P, Wolf K (2003) Tumour-cell invasion and migration:
diversity and escape mechanisms. Nat Rev Cancer 3:362374
35. Colussi D, Brandi G, Bazzoli F, Ricciardiello L (2013) Molecular
pathways involved in colorectal cancer: implications for disease
behavior and prevention. Int J Mol Sci 14(8):1636516385
36. Liao X, Lochhead P, Nishihara R, Morikawa T, Kuchiba A,
Yamauchi M, Imamura Y, Qian ZR, Baba Y, Shima K, Sun R,
Nosho K, Meyerhardt JA, Giovannucci E, Fuchs CS, Chan AT,
Ogino S (2012) Aspirin use, tumor PIK3CA mutation, and
colorectal-cancer survival. N Engl J Med 367(17):15961606
37. Yamauchi M, Morikawa T, Kuchiba A, Imamura Y, Qian ZR,
Nishihara R, Liao X, Waldron L, Hoshida Y, Huttenhower C,
Chan AT, Giovannucci E, Fuchs C, Ogino S (2012) Assessment
of colorectal cancer molecular features along bowel subsites
challenges the conception of distinct dichotomy of proximal
versus distal colorectum. Gut 61(6):847854
38. Yamauchi M, Lochhead P, Morikawa T, Huttenhower C, Chan
AT, Giovannucci E, Fuchs C, Ogino S (2012) Colorectal cancer:
a tale of two sides or a continuum? Gut 6:794797
39. Rosty C, Young JP, Walsh MD, Clendenning M, Walters RJ,
Pearson S, Pavluk E, Nagler B, Pakenas D, Jass JR, Jenkins MA,
Win AK, Southey MC, Parry S, Hopper JL, Giles GG, Williamson
E, English DR, Buchanan DD (2013) Colorectal carcinomas with
KRAS mutation are associated with distinctive morphological and
molecular features. Mod Pathol 26(6):825834

123