Overexpression of CXCL5 Mediates Neutrophil

Infiltration and Indicates Poor Prognosis for

Hepatocellular Carcinoma
Shao-Lai Zhou,1* Zhi Dai,1* Zheng-Jun Zhou,1,2* Xiao-Ying Wang,1 Guo-Huan Yang,1 Zheng Wang,1
Xiao-Wu Huang,1 Jia Fan,1,2 and Jian Zhou1,2

CXCL5 (epithelial neutrophil-activating peptide-78) is a member of a proangiogenic sub-

group of the CXC-type chemokine family of small, secreted proteins. Recently, evidence
that CXCL5 is involved in carcinogenesis and cancer progression has emerged. To investi-
gate the role of CXCL5 in tumor growth, invasion, and prognosis of hepatocellular carci-
noma (HCC), we examined CXCL5 messenger RNA (mRNA) and protein levels in HCC
cell lines with various metastatic potentials and in three independent cohorts of 919 HCC
patients. We found that CXCL5 expression was increased in the highly metastatic HCC
cell lines and in tumor tissues from patients with recurrent HCC compared to controls.
CXCL5 activated the PI3K-Akt and ERK1/2 signaling pathways in HCC cells and pro-
moted proliferation, migration, and invasion. Furthermore, we found that CXCL5 had a
direct chemoattractant effect on neutrophils in vitro. In animal studies, the up-regulation
of CXCL5 in HCC cells promoted tumor growth, lung metastasis, and intratumoral neu-
trophil infiltration. Conversely, down-regulation of CXCL5 in HCC cells reduced tumor
growth, metastasis, and intratumoral neutrophil infiltration. Immunohistochemical analy-
sis in HCC samples showed that overexpression of CXCL5 was well correlated with intra-
tumoral neutrophil infiltration, shorter overall survival, and tumor recurrence.
Multivariate analysis revealed that CXCL5 overexpression alone, or combined with the
presence of intratumoral neutrophils, was an independent prognostic indicator for overall
survival and cumulative recurrence. Conclusion: CXCL5 promotes HCC cell proliferation,
invasion, and intratumoral neutrophil infiltration. CXCL5 overexpression, alone or com-
bined with intratumoral neutrophil presence, is a novel prognostic predictor, and CXCL5
is a potential therapeutic target for HCC. (HEPATOLOGY 2012;56:2242-2254)

H
epatocellular carcinoma (HCC) is one of the fully understood. Although several molecular markers
most prevalent tumor types, and both the for the risk of recurrence and metastatic potential of
incidence and mortality rates of HCC have HCC have been proposed, none have been approved
increased in recent years.1 Although survival of for routine clinical use.4
patients with HCC has improved due to advances in Inflammation has emerged as the seventh hallmark
surgical techniques and perioperative management, of cancer.5 Over the last decade it has been established
long-term survival after surgical resection remains low that cancer-related inflammation is involved in many
due to the high rate of recurrence and metastasis.2,3 aspects of malignancy, and in particular enhances tu-
The molecular pathogenesis and complicated signal mor cell survival, proliferation, and metastasis.6–8
transduction pathways implicated in HCC are not Most HCCs occur in an inflamed liver, often observed

Abbreviations: HCC, hepatocellular carcinoma; HR, hazard ratio; OS, overall survival; qRT-PCR, quantitative reverse transcription-polymerase chain reaction;
TMA, tissue microarray.
From the 1Liver Cancer Institute, Zhongshan Hospital, Fudan University; Key Laboratory of Carcinogenesis and Cancer Invasion (Fudan University), Ministry of
Education, Shanghai 200032, China; and 2Institutes of Biomedical Sciences, Fudan University, Shanghai 200032, China.
Received January 16, 2012; accepted June 3, 2012.
Jointly supported by National Natural Science Funds of China (No. 30972949; No. 30972906), the National Basic Research Program of China (973
Program) (2011CB504001), Shanghai Key-Tech Research & Development Program (No. 09411951700), and Program of Shanghai Excellent Subject Leaders
(No. 10XD1401200).
*These authors contribute equally to this work.

2242
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2243

in Chinese patients infected with hepatitis B virus and enced neutrophil infiltration. Finally, using tissue
in Western populations with hepatitis C virus, suggest- microarrays (TMAs) and quantitative reverse transcrip-
ing a possible crosstalk between inflammation and tion-polymerase chain reaction (qRT-PCR) in HCC
HCC development.9 Therefore, studies on the mecha- samples, we determined the relationship between
nisms of inflammation-associated progression and CXCL5 expression and neutrophil infiltration and
prognosis in HCC are urgently needed. evaluated the prognostic significance of CXCL5 expres-
Chemokines and chemokine receptors are down- sion and neutrophil infiltration.
stream of genetic events and are components of can-
cer-related inflammatory conditions, which predispose
one to cancer and promote cancer progression.10 Che- Materials and Methods
mokines and their receptors affect multiple pathways
Cell Lines and Animals. Six HCC cell lines, one
that contribute to tumor progression in both cell au-
normal liver cell line, and nude mice were used in this
tonomous and nonautonomous ways, including: leuko-
study and are described in the Supporting Information.
cyte recruitment and function, cellular senescence, tu-
Patients and Follow-up. Three independent cohorts
mor cell proliferation and survival, and invasion and
totaling 919 HCC patients were enrolled in this study.
metastasis.11 Recently, CXCL5 (epithelial neutrophil-
The detailed information and follow-up procedures are
activating peptide-78) has been the focus of studies
described in the Supporting Information.
examining the role(s) of chemokines in carcinogenesis
Statistical Analysis. Statistical analyses were done
and tumor progression. Like other chemokines that
using SPSS 16.0 for Windows. Quantitative data
recognize and bind the G-protein-coupled receptor
between groups were compared using Student’s t test.
CXCR2, CXCL5 is a proangiogenic CXC-type chemo-
Categorical data were analyzed by the chi-square test
kine that is an inflammatory mediator and a powerful
or Fisher’s exact test. Correlation analysis was per-
attractant for neutrophils.12–14 CXCL5 is overex- formed between CXCL5 and CD66b. Overall survival
pressed in gastric,15 prostate,16,17 endometrial,18 squa- (OS) and cumulative recurrence rates were calculated
mous cell,19 and pancreatic cancer.20–22 Its increased by the Kaplan-Meier method and differences were ana-
expression is associated with advanced tumor stages, lyzed by the log-rank test. Univariate and multivariate
local invasion, and metastatic potential. These studies analyses were performed using the Cox proportional
demonstrated that CXCL5 directly stimulates cancer hazards regression model. P < 0.05 was considered
cell proliferation and invasion.17,19 Furthermore, statistically significant.
CXCL5 directly induces endothelial cell proliferation For other descriptions of the materials and methods
and invasion in vitro23,24 and promotes tumor angio- used in this study, see the Supporting Information.
genesis in nonsmall cell lung carcinoma and pancreatic
cancer.22,25 However, the role of CXCL5 in HCC and
the relationship between CXCL5 and cancer-related Results
inflammation is largely unknown. CXCL5 Expression Level Is Associated with the
In the present study we investigated the expression Metastatic Potential of HCC Cell Lines. Fluorescence
of CXCL5 in a series of different metastatic HCC cell microscopic analysis for CXCL5 expression showed
lines. We then explored how invasive and metastatic that HCC cells displayed immunostaining in the cyto-
ability changed with changes in CXCL5 expression. plasm (Fig. 1A-C). qRT-PCR, enzyme-linked immu-
Cell lines with low metastatic potential (HepG2 and nosorbent assay (ELISA), and western blot showed
PLC/PRF/5) were transfected with CXCL5 comple- that both the mRNA and protein expression levels of
mentary DNA (cDNA) and highly metastatic cell lines CXCL5 in six established HCC cell lines were signifi-
(HCCLM3 and MHCC97H) were transfected with cantly increased in comparison to the nontransformed
specific short hairpin RNAs (shRNAs). We also inves- hepatic cell line L-02 (P < 0.005; Fig. 1D-F). In addi-
tigated how these changes in CXCL5 expression influ- tion, we observed that CXCL5 expression in the highly
Address reprint requests to: Jian Zhou, Liver Cancer Institute, Zhongshan Hospital, Fudan University, 136 Yi Xue Yuan Road, Shanghai 200032, China.
E-mail: [email protected] fax: þ86-21-64037181; or Jia Fan, Liver Cancer Institute, Zhongshan Hospital, Fudan University, 136 Yi Xue Yuan Road,
Shanghai 200032, China. E-mail: [email protected]
Copyright VC 2012 by the American Association for the Study of Liver Diseases.

View this article online at wileyonlinelibrary.com.

DOI 10.1002/hep.25907
Potential conflict of interest: Nothing to report.
Additional Supporting Information may be found in the online version of this article.
2244 ZHOU ET AL. HEPATOLOGY, December 2012

Fig. 1. CXCL5 expression in six HCC cell lines (HepG2, PLC/PRF/5, Huh7, MHCC97L, MHCC97H, and HCCLM3) with low and high invasive
potentials, and a nontransformed hepatic cell line, L-02. (A-C) Fluorescence microscopic analysis for CXCL5 expression. MHCC97L, MHCC97H,
and HCCLM3 cells displayed strong immunostaining, especially in the cytoplasm, compared with HepG2, PLC/PRF/5, Huh7, and L-02, which dis-
played little fluorescence. (D) qRT-PCR confirmed different CXCL5 mRNA expression in HCC cell lines. Data shown are mean (6SD) from three
independent experiments. (E) ELISAs showed significantly increased levels of CXCL5 protein in the highly invasive cells lines MHCC97L,
MHCC97H, and HCCLM3, compared with HepG2, PLC/PRF/5, Huh7, and L-02. Data shown are means (6SD) from three independent experi-
ments. (F) Western blot analysis showed a significant increase of CXCL5 in the highly invasive cells lines, MHCC97L, MHCC97H, and HCCLM3.

metastatic HCC cell lines (MHCC97-L, MHCC97- CXCL5 Promotes Proliferation, Migration, and
H, and HCCLM3) was significantly higher than in Invasion of HCC Cells. Stable up- or down-regulation
the HCC cell lines with low metastatic potential of CXCL5 expression in transfected HCC cell lines
(HepG2, PLC/PRF/5, and Huh7; P < 0.005). These was confirmed by qRT-PCR and western blot. The
data indicate that CXCL5 is overexpressed in HCC results were consistent with those obtained using an
cell lines and its increased expression is positively cor- ELISA to determine the levels of CXCL5 in cell cul-
related with the metastatic potential of HCC cells. ture supernatants (Fig. 2A). Down-regulation of
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2245

Fig. 2. Characterization of CXCL5 expression, proliferation, migration, and invasion ability in HCC cells. (A) qRT-PCR, western blot, and ELISAs
confirmed CXCL5 expression in stably transfected and parent cells. (B) CXCL5 induced phosphorylation of PI3K-Akt and ERK1/2 in HCC cells in
a CXCR2-dependent manner. (C) Cell proliferation was detected by CCK8 assay. (D) Cell monolayers were wounded and then monitored at 0
and 24 hours for wound channel closure. The cleaned area was measured and plotted as the percentage of the original timepoint (0 hours). (E)
Invasive behavior tested by transwell Matrigel invasion assays. Data shown are mean (6SD) from three independent experiments.

CXCL5 by shRNA in MHCC97H and HCCLM3 HCCLM3-Mock and MHCC97H-Mock cells were
cells caused significant suppression of cell proliferation 42.4 6 14.7 and 31.9 6 12.8, respectively, signifi-
after 48 hours (P < 0.05). Similarly, cell proliferation cantly higher than those of HCCLM3-shRNA-CXCL5
of HepG2-CXCL5 and PLC/PRF/5-CXCL5 cells was and MHCC97H-shRNA-CXCL5 cells (25.5 6 9.9
significantly higher than HepG2-Mock and PLC/PRF/ and 15.7 6 9.2, respectively; P < 0.005). Similarly,
5-Mock cells (P < 0.05; Fig. 2C). In wound-healing the numbers of invasive HepG2-CXCL5 and PLC/
migration assays, microscopic examination at 24 hours PRF/5-CXCL5 cells were 25.5 6 10.4 and 29.7 6
postwounding revealed a significant delay in the 11.7, respectively, significantly higher than those of
wound closure rate of HCCLM3-shRNA-CXCL5 and HepG2-Mock and PLC/PRF/5-Mock cells (15.1 6
MHCC97H-shRNA-CXCL5 cells compared with the 5.9 and 18.2 6 8.5, respectively; P < 0.05; Fig. 2E;
control cell lines, whereas HepG2-CXCL5 and PLC/ Supporting Fig. S1C,D).
PRF/5-CXCL5 cells had a significant increase in the CXCL5 Activates PI3K-Akt and ERK1/2 Signaling
wound closure rate compared with control cell lines (P Pathways in HCC Cells. To explore which signaling
< 0.0005; Fig. 2D; Supporting Fig. S1A,B). In vitro pathways contributed to HCC proliferation and
invasive assays showed that the numbers of invasive invasion induced by CXCL5, the stably transfected
2246 ZHOU ET AL. HEPATOLOGY, December 2012

and parent cells were assessed for the activation status 0.16 cm3, respectively, P < 0.0005). Similarly, the tu-
of multiple pathways. Activation was determined by mor sizes of HepG2-CXCL5-derived and PLC/PRF/5-
the analyzing the phosphorylation state of Akt (PI3K- CXCL5-derived xenografts were 1.84 6 0.65 cm3 and
Akt pathway), ERK1/2 (ERK1/2 pathway), p38 2.71 6 0.49 cm3, respectively, markedly larger than
MAPK (p38 MAPK pathway), JNK (JNK pathway), those of HepG2-Mock-derived and PLC/PRF/5-
p65 (nuclear factor kappa B [NF-jB] pathway), and Mock-derived tumors (0.53 6 0.38 cm3 and 0.73 6
STAT3 (JAK/STAT pathway). All of these proteins, 0.40 cm3, respectively; P < 0.0005; Fig. 3A,B; Fig.
and the associated pathways, can be activated through S3A). Pulmonary metastasis occurred in 100% (8/8)
chemokine receptors. As shown in Fig. 2B (left panel), of the HCCLM3-Mock and MHCC97H-Mock mice,
CXCL5 did not affect p38 MAPK, JNK, p65, or a higher rate than observed in the HCCLM3-shRNA-
STAT3 phosphorylation in HCC cells. Intriguingly, CXCL5 (2/8) and the MHCC97H-shRNA-CXCL5
down-regulation of CXCL5 by shRNA in MHCC97H mice (1/8). The number of metastatic nodules of each
and HCCLM3 cells caused a significant decrease in grade was also greater in the HCCLM3-Mock and
the phosphorylation levels of Akt and ERK1/2. Phos- MHCC97H-Mock mice (Fig. 3D; Fig. S3E,F). In the
phorylated Akt and ERK1/2 were also obviously HepG2-CXCL5 and the PLC/PRF/5-CXCL5 mice,
increased in HepG2-CXCL5 and PLC/PRF/5-CXCL5 pulmonary metastasis occurred in 75% (6/8) and
cells when compared with HepG2-Mock and PLC/ 87.5% of mice (7/8), respectively, but was 0% (0/8) in
PRF/5-Mock cells. Furthermore, we confirmed that PLC/PRF/5-Mock and HepG2-Mock mice (Fig. 3C;
the effect of CXCL5 on Akt and ERK1/2 phosphoryl- Fig. S3C,D). Notably, we found that in the
ation in HCC cells was dependent on the CXCL5 re- HCCLM3-Mock-derived and MHCC97H-Mock-
ceptor, CXCR2 (Fig. 2B, right panel). derived xenografts, intratumor Gr1þ neutrophil infil-
CXCL5 Is a Direct Chemoattractant for tration was significantly more than that observed in
Neutrophils In Vitro. The effect of CXCL5 on neu- HCCLM3-shRNA-CXCL5-derived and MHCC97H-
trophil chemotaxis was assayed using a transwell sys- shRNA-CXCL5-derived xenografts. In the HepG2-
tem. As shown in Fig. S2A, CXCL5 induced concen- Mock-derived and PLC/PRF/5-Mock-derived xeno-
tration-dependent migration of neutrophils in vitro. To grafts, intratumor Gr1þ neutrophil infiltration was
analyze the mechanism underlying CXCL5-mediated significantly less than that observed in HepG2-
chemotaxis of neutrophils, we performed a phos- CXCL5-derived and PLC/PRF/5-CXCL5-derived xen-
phoexplorer antibody array on untreated human neu- ografts (Fig. 4; Figs. S3B, S4).
trophils or those treated with 10 nM of CXCL5. CXCL5 is Up-regulated in HCC Tissues and
Among the 19 phosphorylated proteins that were up- Coincides with Increased Intratumoral Neutrophil
regulated more than 1-fold (Table S1), we found seven Infiltration. Ninety-four tumors (cohort 1) were ana-
that participate in the PI3K/Akt pathway, including p- lyzed by immunohistochemistry, qRT-PCR, and west-
AKTSer473. In addition, two phospho-proteins (p-IKK- ern blot. qRT-PCR revealed that CXCL5 was signifi-
a/b and p-p65) involved in NF-jB signaling were cantly overexpressed in tumors when compared to
increased when compared with control. To validate corresponding peritumoral liver tissue (Fig. 5A). Sam-
these data we performed western blots to detect the ples from patients with tumor recurrence (59 of 94)
phosphorylation of Akt and the p65 subunit of NF- had higher levels of CXCL5 than those without recur-
jB in cells stimulated with CXCL5 (Fig. S2B). These rence (35 of 94; Fig. 5B). These results were con-
data showed that CXCL5 activates several signaling firmed by western blot analyses on 54 HCC samples
pathways in human neutrophils, including the PI3K/ randomly selected from the 94 HCC cases using com-
Akt and NF-jB pathways. puter-generated random numbers with SPSS software
CXCL5 Mediates Neutrophil Infiltration and (Fig. 5C). Tumor tissues that expressed high levels of
Promotes HCC Progression In Vivo. After ortho- CXCL5 showed more neutrophil infiltration. A scatter-
tropic transplantation of all cell lines into nude mice, plot of CXCL5 expression and intratumoral neutrophil
all the groups successfully formed liver tumors. The presence revealed a significant positive correlation
tumor sizes of HCCLM3-Mock-derived and between CXCL5 expression (at mRNA and protein
MHCC97H-Mock-derived xenografts were 5.83 6 levels) and neutrophil infiltration in cancerous tissues
1.29 cm3 and 3.39 6 0.72 cm3, respectively, signifi- (Fig. 5D,E). Similar results were also observed in the
cantly larger than the tumor sizes of xenografts derived other two cohorts of patients (paraffin-embedded tis-
from HCCLM3-shRNA-CXCL5 and MHCC97H- sues, cohort 2, n ¼ 323; cohort 3, n ¼ 502) by
shRNA-CXCL5 cells (1.30 6 0.49 cm3 and 0.56 6 immunohistochemistry.
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2247

Fig. 3. CXCL5 promotes HCC progression in a xenograft nude mice model. (A,B) Macrograph of tumors in all groups. (C,D) Hematoxylin and
eosin (H&E)-stained images of metastatic nodules in lungs from all groups with magnification of the selected areas.

Expression of CXCL5 and Combined Expression Representative cases of immunohistochemical staining

of CXCL5 and Intratumoral Neutrophils Correlates are shown in Fig. 6A-D. We found that CXCL5
with Poor Prognosis in HCC Patients. In the cohort expression was significantly correlated with tumor size
of 94 patients, high CXCL5 level indicated worse (P ¼ 0.005), tumor encapsulation (P ¼ 0.017), vascu-
patient outcome (Fig. S5). The cumulative recurrence lar invasion (P ¼ 0.039), and tumor-node-metastasis
rates were significantly higher among patients with staging (P ¼ 0.030). Other clinical characteristics,
high CXCL5 protein levels (72.3% versus 53.2%, P < including age, sex, preoperative serum alpha-fetopro-
0.05) or high CXCL5 mRNA levels (70.2% versus tein, tumor number, and tumor differentiation,
55.3%, P < 0.05) when compared to those of HCC were not directly related to the expression of CXCL5
patients with low CXCL5 expression. The median OS (Table 1).
was substantially reduced among patients with high By the last follow-up, in March of 2009, 54.2%
CXCL5 protein expression (24.4 months versus 48.0 (175/323) of the patients had suffered a recurrence
months, P < 0.005) or patients with high CXCL5 and 51.1% (165/323) had died. The 1-, 3-, and 5-
mRNA expression (24.4 months versus 48.0 months, year OS rates and cumulative recurrence rates in the
P < 0.005). group (cohort 2) were 85.4% and 25.4%, 62.2% and
We investigated the expression of CXCL5 and 50.2%, and 50.7% and 59.7%, respectively. In addi-
CD66b (a marker which is thought to be uniquely tion, we found that the 1-, 3-, and 5-year survival rates
expressed by human neutrophils) by immunohisto- of the CXCL5Low patients were significantly higher
chemical staining in a tissue microarray composed of than the survival rates of those of the CXCL5High
primary tumors from 323 HCC patients in cohort 2. group (92.0% versus 79.0%, 73.5% versus 50.9%,
2248 ZHOU ET AL. HEPATOLOGY, December 2012

Fig. 4. CXCL5 mediates neutrophil infiltration in a xenograft nude mice model. (A-D) Representative images from tumor sample serial sections
stained with H&E, CXCL5, and Gr1. Magnification, 100
(A-D).

and 62.9% versus 38.5%, respectively; Fig. 6F). Simi- tion, microvascular invasion, and tumor-node-metasta-
larly, CXCL5High HCC patients had the poorest prog- sis stage, CXCL5 expression and the coindex (CXCL5/
nosis at 1-, 3-, and 5-years, with higher cumulative re- intratumoral neutrophils) were independent prognostic
currence rates than CXCL5Low patients (32.7% versus factors for both OS (P < 0.001, HR ¼ 1.817 and P
18.4%, 59.8% versus 41.2%, and 72.2% versus < 0.001, HR ¼ 2.456) and TTR (P ¼ 0.001, HR ¼
48.4%, respectively; Fig. 6F). The presence of intratu- 1.697 and P < 0.001, HR ¼ 2.334; Table 2).
moral neutrophils significantly correlated with OS (P Independent Validation. The prognostic value of
< 0.001, hazard ratio [HR] ¼ 1.896) and time to re- CXCL5 protein expression and intratumoral neutro-
currence (TTR) (P < 0.001, HR ¼ 1.846; Table 2), phil presence was validated in an independent cohort
consistent with previously reported results.9 of 502 HCC patients by immunohistochemical stain-
HCC patients with high CXCL5 expression had ing. Similarly, the expression of tumor-derived CXCL5
more intratumoral neutrophil infiltration than patients had a significant positive correlation with intratumoral
with low CXCL5 expression (n ¼ 323, r ¼ 0.569, P neutrophil presence (n ¼ 502, r ¼ 0.632, P < 0.001).
< 0.001) (Fig. 6E). When evaluating the combined Univariate and multivariate analyses revealed that
effect of CXCL5 expression and intratumoral neutro- CXCL5 and the coindex (CXCL5/intratumoral neutro-
phil presence on HCC prognosis, we found that the phils) were independent prognostic factors for both
1-, 3-, and 5-year OS rates in the CXCL5high/intratu- OS (P < 0.001, HR ¼ 1.895 and P < 0.001, HR ¼
moral neutrophilshigh patients were 78.9%, 46.5%, 2.557) and TTR (P < 0.001, HR ¼ 1.688 and P <
and 33.3%, respectively; significantly lower than that 0.001, HR ¼ 2.068) (Fig. S6, Table S6).
in CXCL5low/intratumoral neutrophilslow patients
(92.2%, 79.1%, and 67.8%, respectively; Fig. 6F).
The 1-, 3-, and 5-year cumulative recurrence rates in
Discussion
the CXCL5high/intratumoral neutrophilshigh patients In this study we confirmed that the expression of
were 34.9%, 65.0%, and 77.2%, respectively, signifi- CXCL5 was markedly increased in HCC tissues com-
cantly higher than those in the CXCL5low/intratumoral pared with that in adjacent nontumor liver tissues.
neutrophilslow patients (17.0%, 36.0%, and 43.2%, Moreover, our results indicate that overexpression of
respectively; Fig. 6F). Univariate and multivariate anal- CXCL5 contributes to the invasion and metastasis of
yses revealed that, along with tumor size, encapsula- HCC cells, and the activation of PI3K-Akt and
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2249

Fig. 5. CXCL5 is up-regulated in HCC tissues (cohort 1, n ¼ 94) and coincides with increased intratumoral neutrophil infiltration. (A) qRT-PCR
analysis of CXCL5 expression in 94 HCC tumors and peritumoral liver tissues. (B) qRT-PCR analysis of CXCL5 expression in 94 HCC tumors with
recurrence or no recurrence. (C) Western blot analysis showed CXCL5 in 54 HCC samples with recurrence or no recurrence. (D,E) Scatterplots
show a positive correlation between CXCL5, at the mRNA or protein level, and intratumoral CD66bþ neutrophils in 94 HCCs.

ERK1/2 signaling pathways by way of its receptor prostate, squamous, and pancreatic cancer,17,19,22 our
CXCR2. Although CXCL5 has been found to partici- results revealed that HCC cells in which CXCL5 was
pate in various cellular functions, including activation, down-regulated have decreased proliferation, invasion,
proliferation, migration, invasion,16,17,19,22 and patho- and metastasis in vitro and in vivo. The above results
logical angiogenesis,22,25 to the best of our knowledge, support the notion that CXCL5 makes a substantial
this is the first study to demonstrate that expression of contribution to tumor progression.
CXCL5 is positively related to the metastatic potential CXCL5 was first identified as a neutrophil-activating
of HCCs. In line with recent studies on CXCL5 in inflammatory peptide with homology to interleukin 8.12
2250 ZHOU ET AL. HEPATOLOGY, December 2012

Fig. 6. Expression and prognostic value of CXCL5 and intratumoral CD66bþ neutrophils in HCC samples (cohort 2, n ¼ 323). (A-D) Repre-
sentative HCC tumor samples show the expression of CXCL5 (brown staining in the cytoplasm of HCC cells) and intratumoral CD66b (brown
staining in the cell membrane of the intratumoral neutrophils). (A,B) High expression of both CXCL5 and intratumoral CD66b. (C,D) Low-level
expression of CXCL5 and intratumoral CD66b. (E) Scatterplot indicates a significant positive correlation between CXCL5 and intratumoral CD66b
in cancerous tissues. (F) Prognostic values of CXCL5 and intratumoral CD66b using Kaplan-Meier analysis. I, CXCL5Low/intratumoral CD66bLow; II,
CXCL5Low/intratumoral CD66bHigh and CXCL5High/intratumoral CD66bLow; III, CXCL5High/intratumoral CD66bHigh.

In a study of rheumatoid arthritis14 and chronic obstruc- site in vivo. Finally, we validated this relationship
tive pulmonary disease,13 CXCL5 was shown to have a between CXCL5 and neutrophil infiltration using clini-
strong effect on neutrophil recruitment. However, its cal HCC samples. In 94 tumors (cohort 1), we found a
relationship to infiltrating neutrophils in tumors, espe- strong correlation between CXCL5 expression (both at
cially in HCC, is largely unknown. In this study we protein and mRNA levels) and intratumoral neutrophils;
revealed a striking correlation between CXCL5 produc- in two other independent cohorts of patients (paraffin-
tion and neutrophil infiltration in HCC. First, we found embedded tissues, cohort 2, n ¼ 323; cohort 3, n ¼
that CXCL5 acts as a direct chemoattractant on neutro- 502), similar results were obtained. The migration of
phils in vitro. Then, through CXCL5 overexpression or neutrophils across the tumor vasculature is mainly medi-
knockdown in HCC cells, we further investigated the ated by CXC chemokines, especially glutamic acid-leu-
role of CXCL5 in recruiting neutrophils to the tumor cine-arginine motif positive-CXC chemokines that bind
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2251

Table 1. Correlation Between the Factors and Clinicopathologic Characteristics in HCC (Cohort 2, n 5 323)
CXCL5 CD66b

Clinicopathological Indexes Low High P Low High P

Age(year) 50 84 82 0.869 88 78 0.291

>50 78 79 74 83
Sex Female 26 20 0.351 23 23 0.982
Male 136 141 139 138
HBsAg Negative 24 21 0.646 20 25 0.409
Positive 138 140 142 136
HCV Negative 157 159 0.448* 160 156 0.283*
Positive 5 2 2 5
Alcohol intake history No 149 144 0.433 150 143 0.243
yes 13 17 12 18
AFP (ng/ml) 20 52 44 0.348 53 43 0.237
>20 110 117 109 118
GGT (U/L) 54 71 60 0.23 69 62 0.455
>54 91 101 93 99
Liver cirrhosis No 21 15 0.298 21 15 0.298
yes 141 146 141 146
Tumor size(cm) 5 95 69 0.005 84 80 0.698
>5 67 92 78 81
Tumor number Single 136 140 0.444 143 133 0.149
Multiple 26 21 19 28
Microvascular invasion absence 97 78 0.039 89 86 0.784
present complete 65 83 73 75
Tumor encapsulation none 93 71 0.017 85 79 0.541
I þ II 69 90 77 82
Tumor differentiation III þ IV 126 129 0.605 130 125 0.566
I 36 32 32 36
TNM stage II þ III 87 67 0.03 86 68 0.051
75 94 76 93

AFP, alpha-fetoprotein; GGT, gamma glutamyl transferase; TNM, tumor-node-metastasis.

*Fisher’s exact tests; chi-square tests for all other analyses.

to and activate CXCR1 and/or CXCR2.26 In this study that neutrophils mediate cancer cell adhesion to he-
we demonstrated that CXCL5 activates several signaling patic sinusoids, which could promote the implantation
pathways in human neutrophils, including the PI3K-Akt of circulating tumor cells and formation of hepatic
and NF-jB signaling pathways. Activation of Akt is cru- metastases. In our study the HCCLM3-Mock-derived
cial for the chemotaxis and migration of neutrophils and MHCC97H-Mock-derived xenografts with high
induced by chemoattractants.27,28 Recent studies have expression of CXCL5 had more intratumoral neutro-
also revealed that NF-jB activation contributes to the phils, larger tumors, and more pulmonary metastases
recruitment of neutrophils that is induced by chemotac- compared to HepG2-Mock-derived and PLC/PRF/5-
tic cytokines.29 Thus, we suggest that CXCL5 plays a Mock-derived xenografts with low CXCL5 expression.
crucial role in recruiting neutrophils to the tumor site of After CXCL5 expression was inhibited by shRNA in
HCCs, potentially through the activation of AKT and the HCC cells, the number of intratumoral neutro-
NF-jB. phils, tumor volume, and pulmonary metastasis
Neutrophils, initially recognized as short-lived effec- decreased, and vice versa. These results suggest that tu-
tor cells providing the first line of defense against mor-derived CXCL5 and the accompanying neutrophil
invading pathogens, are the most abundant subpopula- infiltration substantially contribute to HCC progres-
tion of leukocytes.30 Neutrophils can potentiate cancer sion in nude mice models.
cell migration, invasion, and dissemination by secret- To determine whether CXCL5 could be an impor-
ing immunoreactive molecules such as hepatocyte tant factor related to clinical outcome of HCC patients
growth factor,31 oncostatin M,32 b2-integrins,33 or after curative resection, we examined CXCL5 mRNA
neutrophil elastase.34 Imai et al.35 reported that neu- expression in a cohort of 94 HCC patients by qRT-
trophils enhance the invasion activity of the HepG2 PCR. Our results indicated that patients with high
cell line in vitro by way of the paracrine regulation of CXCL5 mRNA levels have a higher recurrence rate
hepatocyte growth factor. McDonald et al.36 reported and shorter survival time. Subsequently, we detected
2252 ZHOU ET AL. HEPATOLOGY, December 2012

Table 2. Univariate and Multivariate Analyses of Prognostic Factors in HCC (Cohort 2, n 5 323)
TTR OS

Variable HR (95% CI) P HR (95% CI) P

Univariate analysis
Age, year (50 versus >50) 1.023 (0.760-1.377) 0.879 0.828 (0.610-1.125) 0.228
Sex (female versus male) 1.863 (1.143-3.036) 0.013 1.757 (1.049-2.942) 0.032
HBsAg (negative versus positive) 1.007 (0.659-1.538) 0.974 1.039 (0.669-1.614) 0.866
Alcohol intake history (no versus yes) 1.021 (0.619-1.683) 0.936 1.001 (0.598-1.677) 0.996
AFP, ng/ml (20 versus >20) 1.155 (0.835-1.597) 0.385 1.548 (1.083-2.211) 0.016
GGT, U/L (54 versus >54) 1.329 (0.978-1.805) 0.069 1.737 (1.253-2.409) 0.001
Liver cirrhosis (no versus yes) 1.102 (0.677-1.795) 0.696 1.308 (0.769-2.224) 0.332
Tumor size, cm (5 versus >5) 1.817 (1.346-2.452) 0.000 2.482 (1.806-3.412) 0.000
Tumor number (single versus multiple) 1.362 (0.907-2.044) 0.136 1.517 (1.025-2.243) 0.037
Microvascular invasion (no versus yes) 1.915 (1.420-2.583) 0.000 2.479 (1.815-3.388) 0.000
Tumor encapsulation (complete versus none) 1.679 (1.246-2.263) 0.001 1.715 (1.260-2.336) 0.001
Tumor differentiation (I þ II versus III þ IV) 1.238 (0.870-1.763) 0.236 1.581 (1.119-2.233) 0.009
TNM stage (I versus II III) 2.072 (1.527-2.814) 0.000 2.683 (1.935-3.720) 0.000
CXCL5 (low versus high) 1.932 (1.428-2.612) 0.000 2.100 (1.533-2.877) 0.000
Intratumor CD66b (low versus high) 1.846 (1.363-2.499) 0.000 1.896 (1.386-2.593) 0.000
Combination of CXCL5 and CD66b* 0.000 0.000
I versus II 1.592 (1.070-2.368) 0.022 1.727 (1.134-2.629) 0.011
II versus III 1.562 (1.097-2.224) 0.013 1.556 (1.085-2.233) 0.016
I versus III 2.503 (1.731-3.620) 0.000 2.696 (1.836-3.958) 0.000
Multivariate analysis
Sex (female versus male) 1.558 (0.965-2.515) 0.070 1.325 (0.796-2.206) 0.279
AFP, ng/ml (20 versus >20) NA NA 1.242 (0.857-1.802) 0.252
GGT,U/L (54 versus>54) NA NA 1.434 (1.020-2.015) 0.038
Tumor size, cm (5 versus >5) 1.622 (1.191-2.210) 0.002 2.035 (1.459-2.839) 0.000
Tumor number (single versus multiple) NA NA 1.311 (0.877-1.959) 0.187
Tumor encapsulation (complete versus none) 1.614 (1.186-2.197) 0.002 1.583 (1.148-2.184) 0.005
Microvascular invasion (no versus yes) 1.513 (1.106-2.070) 0.010 1.997 (1.445-2.760) 0.000
Tumor differentiation (I þ II versus III þ IV) NA NA 1.491 (1.040-2.140) 0.030
CXCL5 (low versus high) 1.697 (1.249-2.304) 0.001 1.817 (1.323-2.496) 0.000
Intratumor CD66b (low versus high) 1.938 (1.428-2.629) 0.000 2.000 (1.455-2.749) 0.000
Combination of CXCL5 and CD66b* 0.000 0.000
I versus II 1.521 (1.019-2.272) 0.040 1.623 (1.060-2.485) 0.026
II versus III 1.470 (1.027-2.105) 0.035 1.515 (1.052-2.182) 0.026
I versus III 2.334 (1.611-3.382) 0.000 2.456 (1.669-3.612) 0.000

Cox proportional hazards regression model.

*I, CXCL5Low/CD66bLow; II, CXCL5Low/CD66bHigh and CXCL5High/CD66bLow; III, CXCL5High/CD66bHigh. AFP, alpha-fetoprotein; GGT, gamma glutamyl transferase;
TNM, tumor-node-metastasis.
HR, hazard ratio; CI, confidential interval; NA, not adopted.

the protein expression of CXCL5 in TMAs containing HCC patients. We made direct comparisons of prog-
this cohort of 94 HCC patients and two other cohorts nosis among three subgroups (CXCL5low/intratumor
(323 HCC patients and 502 HCC patients) by immu- neutrophilslow, CXCL5low/intratumor neutrophilshigh
nohistochemistry. Multivariate analysis confirmed that or CXCL5high/intratumor neutrophilslow, and
CXCL5 (either at the protein or mRNA level) is a sig- CXCL5 /intratumor neutrophils high) in the two
high

nificant independent predictor for OS and TTR. cohorts (323 HCC patients and 502 HCC patients).
Because CXCL5 acted as a strong chemotactic cyto- In the TMA analysis, HCC patients who had both
kine for neutrophils in our xenograft model and there high CXCL5 expression and high levels of intratumor
was a significant positive correlation between CXCL5 neutrophil infiltration were more prone to recurrence
expression and intratumoral neutrophils in clinical of HCC following curative resection. Conversely,
HCC samples, we next investigated the prognostic sig- HCC patients who expressed both low CXCL5 and
nificance of intratumoral neutrophils in TMA by had low intratumor neutrophil infiltration had the best
immunohistochemistry. In accordance with our previ- prognosis. Although the expression of CXCL5 was also
ous study,9 the presence of intratumoral neutrophils an independent predictor for OS and TTR, the pre-
was a poor prognostic indicator for HCC after resec- dictive range of CXCL5high/intratumor neutrophilshigh
tion. Thereafter, we evaluated the prognostic value of was more sensitive than that of CXCL5high or intratu-
both CXCL5 expression and intratumor neutrophils in mor neutrophilshigh alone.
HEPATOLOGY, Vol. 56, No. 6, 2012 ZHOU ET AL. 2253

Tumor growth and dissemination is the result of 3. Yang Y, Nagano H, Ota H, Morimoto O, Nakamura M, Wada H,
et al. Patterns and clinicopathologic features of extrahepatic recurrence
dynamic interactions between tumor cells themselves of hepatocellular carcinoma after curative resection. Surgery 2007;141:
and with components of the tumor inflammatory envi- 196–202.
ronment.37 The inflammatory environment includes 4. Llovet JM, Bruix J. Molecular targeted therapies in hepatocellular carci-
inflammatory cells and inflammatory mediators (for noma. HEPATOLOGY 2008;48:1312–1327.
5. Mantovani A. Cancer: Inflaming metastasis. Nature 2009;457:36–37.
example, chemokines, cytokines, and prostaglandins).6 6. Mantovani A, Allavena P, Sica A, Balkwill F. Cancer-related inflamma-
Chemokines are emerging as key mediators not only in tion. Nature 2008;454:436–444.
the homing of cancer cells to metastatic sites but also in 7. Karin M. Nuclear factor-kappaB in cancer development and progres-
sion. Nature 2006;441:431–436.
the recruitment of a number of different cell types to 8. DeNardo DG, Johansson M, Coussens LM. Immune cells as mediators
the tumor microenvironment. Recent studies suggest of solid tumor metastasis. Cancer Metastasis Rev 2008;27:11–18.
that chemokines are produced by epithelial cancer cells, 9. Li YW, Qiu SJ, Fan J, Zhou J, Gao Q, Xiao YS, et al. Intratumoral
leading to the recruitment of tumor-associated macro- neutrophils: a poor prognostic factor for hepatocellular carcinoma fol-
lowing resection. J Hepatol 2011;54:497–505.
phages, tumor-associated neutrophils, lymphocytes, can- 10. Balkwill F. Cancer and the chemokine network. Nat Rev Cancer 2004;
cer-associated fibroblasts, mesenchymal stem cells, and 4:540–550.
endothelial cells to the tumor microenvironment. These 11. Mantovani A, Savino B, Locati M, Zammataro L, Allavena P, Bonecchi
R. The chemokine system in cancer biology and therapy. Cytokine
infiltrating cells provide a secondary source of chemo- Growth Factor Rev 2010;21:27–39.
kines that could affect tumor growth, cell survival, se- 12. Walz A, Burgener R, Car B, Baggiolini M, Kunkel SL, Strieter RM.
nescence, angiogenesis, and metastasis.37 Considering Structure and neutrophil-activating properties of a novel inflammatory
that the inflammatory cells in HCC are usually mixed peptide (ENA-78) with homology to interleukin 8. J Exp Med 1991;
174:1355–1362.
with intratumoral lymphocytes accounting for a signifi- 13. Qiu Y, Zhu J, Bandi V, Atmar RL, Hattotuwa K, Guntupalli KK,
cant portion, we investigated the relationship between et al. Biopsy neutrophilia, neutrophil chemokine and receptor gene
intratumoral infiltration of lymphocytes and neutro- expression in severe exacerbations of chronic obstructive pulmonary dis-
ease. Am J Respir Crit Care Med 2003;168:968–975.
phils. As shown in Fig. S7 and Table S8, intratumoral 14. Koch AE, Kunkel SL, Harlow LA, Mazarakis DD, Haines GK, Burdick
neutrophils positively correlated with most types of MD, et al. Epithelial neutrophil activating peptide-78: a novel chemo-
intratumoral lymphocytes, including CD3þ, CD4þ, tactic cytokine for neutrophils in arthritis. J Clin Invest 1994;94:
1012–1018.
CD8þ, and Foxp3þ lymphocytes. However, we did
15. Park JY, Park KH, Bang S, Kim MH, Lee JE, Gang J, et al. CXCL5
not observe a significant correlation between tumor- overexpression is associated with late stage gastric cancer. J Cancer Res
derived CXCL5 and intratumoral infiltration of lym- Clin Oncol 2007;133:835–840.
phocytes, indicating that CXCL5 is a strong and specific 16. Begley LA, Kasina S, Mehra R, Adsule S, Admon AJ, Lonigro RJ,
et al. CXCL5 promotes prostate cancer progression. Neoplasia 2008;
chemotactic cytokine for neutrophils in HCC. Tumor- 10:244–254.
derived CXCL5 and accompanying intratumoral neu- 17. Kuo PL, Chen YH, Chen TC, Shen KH, Hsu YL. CXCL5/ENA78
trophils may cooperatively contribute to HCC growth, increased cell migration and epithelial-to-mesenchymal transition of
hormone-independent prostate cancer by early growth response-1/snail
invasion, and metastasis. HCC cells with higher signaling pathway. J Cell Physiol 2011;226:1224–1231.
expression of CXCL5 are prone to recruit more 18. Wong YF, Cheung TH, Lo KW, Yim SF, Siu NS, Chan SC, et al.
neutrophils to the tumor site, establishing a tumor- Identification of molecular markers and signaling pathway in endome-
promoting microenvironment and amplifying the trial cancer in Hong Kong Chinese women by genome-wide gene
expression profiling. Oncogene 2007;26:1971–1982.
inflammatory response, thus facilitating HCC invasion 19. Miyazaki H, Patel V, Wang H, Edmunds RK, Gutkind JS, Yeudall
and metastasis. WA. Down-regulation of CXCL5 inhibits squamous carcinogenesis.
In conclusion, CXCL5 promotes HCC cell prolifer- Cancer Res 2006;66:4279–4284.
20. Frick VO, Rubie C, Wagner M, Graeber S, Grimm H, Kopp B, et al.
ation, invasion, and intratumoral neutrophil infiltra- Enhanced ENA-78 and IL-8 expression in patients with malignant pan-
tion. CXCL5 expression alone or in combination with creatic diseases. Pancreatology 2008;8:488–497.
intratumoral neutrophils serves as a novel prognostic 21. Wente MN, Keane MP, Burdick MD, Friess H, Buchler MW, Ceyhan
indicator for HCC patients undergoing curative resec- GO, et al. Blockade of the chemokine receptor CXCR2 inhibits pan-
creatic cancer cell-induced angiogenesis. Cancer Lett 2006;241:
tion. These features of CXCL5 make it a potential 221–227.
therapeutic target as well. 22. Li A, King J, Moro A, Sugi MD, Dawson DW, Kaplan J, et al. Over-
expression of CXCL5 is associated with poor survival in patients with
pancreatic cancer. Am J Pathol 2011;178:1340–1349.
References 23. Takahashi H, Numasaki M, Lotze MT, Sasaki H. Interleukin-17
enhances bFGF-, HGF- and VEGF-induced growth of vascular endo-
1. Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J thelial cells. Immunol Lett 2005;98:189–193.
Clin 2010;60:277–300. 24. Matsuo Y, Raimondo M, Woodward TA, Wallace MB, Gill KR, Tong
2. Tang ZY, Ye SL, Liu YK, Qin LX, Sun HC, Ye QH, et al. A decade’s Z, et al. CXC-chemokine/CXCR2 biological axis promotes angiogenesis
studies on metastasis of hepatocellular carcinoma. J Cancer Res Clin in vitro and in vivo in pancreatic cancer. Int J Cancer 2009;125:
Oncol 2004;130:187–196. 1027–1037.
2254 ZHOU ET AL. HEPATOLOGY, December 2012

25. Arenberg DA, Keane MP, DiGiovine B, Kunkel SL, Morris SB, Xue bronchioloalveolar subtype pulmonary adenocarcinoma: role in tumor
YY, et al. Epithelial-neutrophil activating peptide (ENA-78) is an im- progression and death. Cancer Res 2003;63:1405–1412.
portant angiogenic factor in non-small cell lung cancer. J Clin Invest 32. Queen MM, Ryan RE, Holzer RG, Keller-Peck CR, Jorcyk CL. Breast
1998;102:465–472. cancer cells stimulate neutrophils to produce oncostatin M:
26. Murdoch C, Muthana M, Coffelt SB, Lewis CE. The role of myeloid potential implications for tumor progression. Cancer Res 2005;65:
cells in the promotion of tumour angiogenesis. Nat Rev Cancer 2008; 8896–8904.
8:618–631. 33. Strell C, Lang K, Niggemann B, Zaenker KS, Entschladen F. Neutrophil
27. Williams MR, Azcutia V, Newton G, Alcaide P, Luscinskas FW. Emerg- granulocytes promote the migratory activity of MDA-MB-468 human
ing mechanisms of neutrophil recruitment across endothelium. Trends breast carcinoma cells via ICAM-1. Exp Cell Res 2010;316:138–148.
Immunol 2011;32:461–469. 34. Houghton AM, Rzymkiewicz DM, Ji H, Gregory AD, Egea EE, Metz
HE, et al. Neutrophil elastase-mediated degradation of IRS-1 acceler-
28. Heit B, Tavener S, Raharjo E, Kubes P. An intracellular signaling hier-
ates lung tumor growth. Nat Med 2010;16:219–223.
archy determines direction of migration in opposing chemotactic gra-
35. Imai Y, Kubota Y, Yamamoto S, Tsuji K, Shimatani M, Shibatani N,
dients. J Cell Biol 2002;159:91–102.
et al. Neutrophils enhance invasion activity of human cholangiocellular
29. Choi M, Salanova B, Rolle S, Wellner M, Schneider W, Luft FC, et al. carcinoma and hepatocellular carcinoma cells: an in vitro study. J Gas-
Short-term heat exposure inhibits inflammation by abrogating recruit- troenterol Hepatol 2005;20:287–293.
ment of and nuclear factor-{kappa}B activation in neutrophils exposed 36. McDonald B, Spicer J, Giannais B, Fallavollita L, Brodt P, Ferri LE. Sys-
to chemotactic cytokines. Am J Pathol 2008;172:367–777. temic inflammation increases cancer cell adhesion to hepatic sinusoids by
30. Pham CT. Neutrophil serine proteases: specific regulators of inflamma- neutrophil mediated mechanisms. Int J Cancer 2009;125:1298–1305.
tion. Nat Rev Immunol 2006;6:541–550. 37. Lazennec G, Richmond A. Chemokines and chemokine receptors: new
31. Wislez M, Rabbe N, Marchal J, Milleron B, Crestani B, Mayaud C, insights into cancer-related inflammation. Trends Mol Med 2010;16:
et al. Hepatocyte growth factor production by neutrophils infiltrating 133–144.