Potential conflict of interest: Nothing to report.
Angiogenesis is a critical step in the development and progression of hepatocellular carcinoma (HCC). Myeloid lineage cells, such as macrophages and monocytes, have been reported to regulate angiogenesis in mouse tumor models. TIE2, a receptor of angiopoietins, conveys pro-angiogenic signals and identifies a monocyte/macrophage subset with pro-angiogenic activity. Here, we analyzed the occurrence and kinetics of TIE2-expressing monocytes/macrophages (TEMs) in HCC patients. This study enrolled 168 HCV-infected patients including 89 with HCC. We examined the frequency of TEMs, as defined as CD14+CD16+TIE2+ cells, in the peripheral blood and liver. The localization of TEMs in the liver was determined by immunofluorescence staining. Micro-vessel density in the liver was measured by counting CD34+ vascular structures. We found that the frequency of circulating TEMs was significantly higher in HCC than non-HCC patients, while being higher in the liver than in the blood. In patients who underwent local radio-ablation or resection of HCC, the frequency of TEMs dynamically changed in the blood in parallel with HCC recurrence. Most TEMs were identified in the perivascular areas of tumor tissue. A significant positive correlation was observed between micro-vessel density in HCC and frequency of TEMs in the blood or tumors, suggesting that TEMs are involved in HCC angiogenesis. Receiver operating characteristic analyses revealed the superiority of TEM frequency to AFP, PIVKA-II and ANG-2 serum levels as diagnostic marker for HCC. Conclusion: TEMs increase in patients with HCC and their frequency changes with the therapeutic response or recurrence. We thus suggest that TEM frequency can be used as a diagnostic marker for HCC, potentially reflecting angiogenesis in the liver. (HEPATOLOGY 2013)
Hepatocellular carcinoma (HCC) is one of the most prevalent malignancies and the third leading cause of cancer-related deaths worldwide.1 Clinically, HCC frequently develops from liver cirrhosis, with most etiologies involving hepatitis B and C virus (HBV and HCV) infection.2, 3 Since the majority of HCCs are characterized by a florid intra-tumoral vasculature, angiogenesis is deemed to be a critical step in the development and progression of HCC.4 Some clinical studies have demonstrated that the degree of vascularity in HCC tissues correlates with the severity of the disease condition,5 suggesting that prevention of this process could have beneficial impact on patient prognosis. However, the precise mechanisms of HCC-related angiogenesis in the liver remain obscure.
In general, two types of components are cooperatively involved in the progression of angiogenesis: humoral angiogenesis factors and vascular progenitor cells.4 Many studies have reported that angiogenesis factors produced from HCC drive tumor vascularization, which supports the development and progression of liver cancer, including invasion and metastasis.6 Among such factors, serum levels of angiopoietin-2 (ANG-2), macrophage migration inhibitory factor (MIF), vascular endothelial cell growth factor (VEGF), and soluble vascular endothelial cell growth factor receptor-1 (sVEGFR-1) have been reported to be higher in HCC patients than non-HCC subjects and to correlate with poorer prognosis or survival.7-11 However, such angiogenesis molecules have failed to show any advantage over other clinically available markers for HCC diagnosis.12
Hematopoietic lineage cells, including hematopoietic progenitors and myeloid lineage cells, have been implicated in the promotion of tumor angiogenesis.4 Tyrosine kinase with Ig and EGF homology domains 2 (TIE2) is a receptor of angiopoietins (ANGs); it is primarily expressed on endothelial cells and is capable of binding to all the known ANGs (ANG-1, ANG- 2 and ANG-3/ANG-4). TIE2-expressing monocytes (TEMs) are a recently described subpopulation of peripheral and tumor-infiltrating myeloid cells presumed to be equipped with profound pro-angiogenic activity; these cells are found both in mice and humans.13-15
Among human cancers, TEMs have been reported in tumors of the kidney, colon, pancreas and lung, as well as in soft tissue sarcomas,15 where angiogenesis is known to be important for tumor progression. However, it is not known whether TEMs are present in HCC, and their significance for the pathophysiology of the disease remains to be investigated.
In this study, we analyzed TEMs in HCC patients by investigating their frequency, localization and correlation with microvessel density and other clinical parameters in HCC patients. Our findings suggest that TEMs could serve as a diagnostic marker of HCC.
Materials and Methods
Among chronically HCV-infected patients who had been followed at Osaka University Hospital, 168 patients were enrolled (Table 1). They were categorized into three groups according to the stage of the liver disease: chronic hepatitis (CH), liver cirrhosis (LC) and hepatocellular carcinoma (HCC). The clinical stage of HCC was determined according to the TNM classification system of the International Union against Cancer (7th edition) or the BCLC staging classification system. The protocol of this study was approved by the ethical committee of Osaka University Hospital and Osaka University Graduate School of Medicine. At enrollment, written informed consent was obtained from all patients and volunteers. Some of the HCC patients in this study received radiofrequency ablation (RFA) therapy based on the therapeutic guidelines for HCC promoted by the Japan Society of Hepatology.16 After the RFA sessions, the efficacy of tumor ablation or HCC recurrence was evaluated by computed tomography (CT) or magnetic resonance imaging (MRI) scanning. With some of the HCC patients who underwent surgical resection, cancerous and adjacent non-cancerous tissues were obtained at operation for further analyses of TEMs. As controls, we examined healthy subjects (HS) without history of liver disease, HCC patients with HBV infection (HBV-HCC group) and those without HBV or HCV (non-B, non-C [NBNC]-HCC group). The clinical backgrounds of the subjects are shown in Table 1.
The fluorescence-labeled mouse or rat monoclonal antibodies against relevant molecules used in this study were: CD14 (M5E2), CXCR4 (12G5), CD40 (5C3), CD16 (3G8), CD34 (563),
CD11b (ICRF44), CD49d (9F10), CD80 (L307.4), CD86 (2331), CD33 (WM53), CCR4 (1G1), HLA-DR (L243) and CCR5 (2D7/CCR5), which were purchased from Becton Dickinson (BD) Biosciences, San Jose, CA. Anti-human VEGFR2 (89106) or TIE2 (83715) Abs were purchased from R&D SYSTEMS, Minneapolis, MN; anti-human CD45 (HI30) from BioLegend, San Diego, CA; anti-human CX3CR1 (2A9-1) was from Medical & Biological Laboratories (MBL), Nagoya, Japan, and anti-AC133 (AC133) was from Militenyi Biotec.
Phenotype and frequency analysis of peripheral and tumor-infiltrating TEMs
After peripheral blood mononuclear cells (PBMC) had been separated from heparinized venous blood by Ficoll-Hypaque (Nacalai tesque, Kyoto, Japan) density gradient centrifugation, they were stained with a combination of fluorescence-labeled anti-human mouse mAbs against CD14, CD16 and TIE2. For the analyses of liver-infiltrated cells, fresh liver specimens were washed twice with phosphate-buffered saline (PBS) and then diced into 5-mm pieces. After these pieces had been passed through a nylon mesh (BD Falcon, San Jose, CA), tumor- infiltrating and non-cancerous tissue-infiltrating leukocytes were isolated by density gradient centrifugation as described above. These cells were stained with fluorescence-labeled Abs (CD14, CD16 and TIE2) as done for PBMC. The stained cells were analyzed using FACS CantoII (BD) and FCS Express software (De Novo, Los Angeles, CA, USA).
CD16+ and CD16- monocytes were sorted using a FACS sorter. The sorted cells (105-5x105) were subjected to Western blot analysis for TIE2 expression as described elsewhere.15
Immunofluorescence staining analysis
Tissue specimens were obtained from surgical resections of HCCs. Five-micrometer sections were fixed in 4% paraformaldehyde (PFA) for 15 minutes and immunostained. Briefly, the sections were incubated with the following antibodies by detection with a polymeric labeling 2- step method as described:15 rabbit anti-human CD14 antibody (clone, HPA001887; Sigma), mouse anti-human CD16 (2H7; MBL) and mouse anti-human TIE2 (AB33; Upstate Biotechnology) antibodies and subsequently with secondary goat anti-rabbit Alexa Fluor®488 or goat anti-mouse Alexa Fluor®594 (Invitrogen, Molecular Probes) antibodies. Cell nuclei were counterstained with Dapi-Fluoromount-GTM (SouthernBiotech, Birmingham, AL). The stained tissues were analyzed by fluorescence microscopy (Model BZ-9000; Keyence, Osaka, Japan).
Immunohistochemical analysis and assessment of microvessel density
To evaluate microvessel density (MVD), immunohistochemical analyses were performed with anti-CD34 antibody (1/50 dilution; QB-END/10, Novo-castra, Newcastle, UK) using the avidin- biotin complex (ABC) method (Vectastain) as described.17
Single microvessels were detected as any brown CD34-immunostained endothelial cell structures. MVD was evaluated according to the method described by Poon et al.17 Sections were read by two double-blinded pathologists according to stainingintensity.
Differences between two groups were assessed by the Mann-Whitney nonparametric U test, and multiple comparisons between more than two groups by the Kruskal-Wallis nonparametric test. Paired t tests were used to compare differences in paired samples using GraphPad Prism software (GraphPad Prism, San Diego, CA, USA). To differentiate HCC and LC, receiver operating characteristics (ROC) analyses were done using JMP software (SAS, Cary, NC, USA). The correlation between two groups was assessed by Pearson's analysis. The recurrence-free survival rate in patients with HCC who underwent the treatment was compared using the Kaplan-Meier method, with the log-rank test for comparison. Associations among the variables were determined by %2 test of Fisher exact test and Student t test. All tests were two-tailed, and a P value of less than 0.05 was considered statistically significant.
TIE2 is selectively expressed on CD14+CD16+ monocytes
In order to examine which population of cells expresses TIE2, we stained PBMC obtained from HCC patients with relevant Abs. Among PBMC, CD14+HLA-DR+ monocytes we found to express TIE2 (Fig. 1A), whereas CD14-HLA-DR- cells did not (Fig. 1A). In particular, T-cells, B-cells, natural killer (NK) cells, NKT cells, and dentritic cells did not express detectable TIE2 (data not shown).
Monocytes can be divided into two distinct subsets according to the expression of CD14 and CD16: CD14++CD16- and CD14+CD16+ monocytes, respectively (Fig. 1A). CD14+CD16+ monocytes express TIE2 to a higher degree than CD14++CD16- monocytes (Fig. 1B). By Western blot analysis, TIE2 was expressed in CD14+CD16+ cells to a lesser extent than in HUVEC (used as a positive control); however, TIE2 expression was higher in CD14+CD16+ than CD14++CD16- cells (Fig. 1C), in agreement with the flow cytometry data. From here on, we refer to CD14+CD16+TIE2+ cells as TIE2-expressing monocytes (TEMs).
TEMs are phenotypically and functionally distinct from TIE2- monocytes or endothelial progenitor cells in myeloid lineage
TEMs were found to express CD45, CD11b, CCR4, CCR5, CX3CR1, CD40 and CD86, the expression of which was generally higher than in TIE2- monocytes (Fig. 1D, 1E). The expression of CD33, HLA-DR, CD49d and CXCR4 was comparably high in monocytes regardless of CD16 or TIE2 expression. Since TEMs have been reported to be involved in the promotion of angiogenesis,4, 13, 18 we analyzed the expression of endothelial progenitor cell (EPC) markers.19 We found that TEMs do not express the EPC markers AC133, VEGFR2 or CD34 (Fig. 1D). Together, these results confirm that TEMs are phenotypically and functionally distinct from TIE2- monocytes or EPCs.
TEMs are significantly increased in the peripheral blood of HCC patients and their increase is associated with cancer occurrence and recurrence
We compared the frequency of TEMs in PBMC among healthy subjects and chronically HCV- infected patients with various stages of liver disease. With respect to the demographics of the subjects, no difference was found in the clinical and pathological characteristics among patient groups (Table 1). In HCC patients, the frequency of TEMs in the blood was significantly higher than that in all other groups (Fig. 2A). However, the frequency of TEMs did not differ between patients at advanced HCC stages (TNM stages III and IV) and those at early HCC stages (stages I and II) (Fig. 2B). Similar results were obtained with the classification according to the BCLC staging system. Indeed, the frequency of TEMs did not differ between patients with advanced HCC stages (BCLC, C and D) and those with early stages (A and B) (3.6 ± 2.2% vs. 3.3 ± 2.3%). These results show that the increase of TEMs is closely related to the presence of HCC, irrespective of the stage of cancer. Furthermore, we observed higher TEM frequency in non- HCV-infected HCC patients (NBNC-, alcoholic- and HBV-HCC patients),than non-HCC subjects (Supplementary figure 1), suggesting that the increase of circulating TEMs is influenced by HCC, not by infection with hepatitis viruses.
We then serially examined the frequency of TEMs in HCC patients who underwent RFA therapy or tumor resection. We assessed the viability of HCC by CT or MRI scanning every 3 to 6 months after the treatment. In patients without HCC recurrence, the frequency of TEMs dramatically decreased after successful HCC ablation or resection (Fig. 2C). By contrast, in patients with subsequent HCC recurrence, TEMs increased before the apparent radiological identification of HCC (Fig. 2C). Therefore, TEM frequency dynamically changes in patients in correlation with the presence or recurrence of HCC.
In order to assess the clinical significance of TEMs as tumor biomarkers, we compared various clinical parameters in patients with either high or low TEM frequency. We fractionated HCC patients according to the frequency of circulating TEMs, defined as higher (TEMhigh) or lower (TEMlow) than the median value (cut off value = 2.75%). We found that HCC patients in the TEMhigh group displayed (i) a more advanced Child-Pugh grade (B); (ii) a higher model for endstage liver disease (MELD) score; (iii) a lower prothrombin time; (iv) a lower serum albumin level (Table 2). These results suggest that elevated TEM frequency in the peripheral blood is associated with a deterioration of liver function in HCC patients. Furthermore, patients in the TEMhigh group showed significantly shorter recurrence-free survival rates than those in the TEMlow group (assessed before RFA treatment or resection of HCC), suggesting that the assessment of TEM frequency in the blood holds prognostic value (Fig. 2D)
Table 2. Comparison of Clinical Parameters of HCC Patients Between Those with Higher Frequency of TEMs and Those with Lower Frequency
High (n = 45)
Low (n = 44)
P < 0.05 in bold.
χ2 test or Fisher's exact test;
Student t test;
Mann-Whitney U test.
TEMs, TIE2-expressing monocytes; MELD, model for endstage liver disease.
We found that most TEMs, identified as CD14+TIE2+ cells by immunofluorescence staining, were located in the perivascular areas of HCC (Fig. 3A-[A], [B], [D]). Most of the CD16+ cells co-stained with CD14+TIE2+ cells (Fig. 3A-[C]). Some of the TEMs were observed in the lumen of intra-tumoral blood vessels, identified as CD14-TIE2+ vascular structures (Fig. 3A- [D]). However, TEMs were scarce in the adjacent non-cancerous liver tissue (not shown).
TEMs accumulate in HCC tissue
The frequency of TEMs among tumor-infiltrating leukocytes (TIL) was higher than nontumor- infiltrating leukocytes (NIL) and PBMC (Fig. 3B), suggesting that TEMs preferentially accumulate in the HCC tissue as compared with normal liver tissue. Moreover, in patients with HCC, the frequency of TEMs in PBMCs correlated with that in TIL (Fig. 3C), suggesting that TEM frequency in the blood may represent a surrogate biomarker of TEM infiltration in the tumors.
TEM frequency correlates with microvessel density (MVD) in HCC
To study whether the frequency of circulating and intra-hepatic TEMs correlate with tumor angiogenesis, we measured microvascular density by anti-CD34 staining in liver tissue obtained from 12 HCC patients. The expression of CD34 was predominantly confined to the cytoplasm of vascular endothelial cells. Microvessels were identified as brow/yellow capillaries or small cell clusters. The CD34+ microvessels were located mainly in tumor cell areas (Fig. 4A). MVD tended to be higher in HCC (67.0 ± 57.8) than in non-cirrhotic (26.7 ± 7.5) or cirrhotic non- cancerous tissues (32.1 ± 11.6). Furthermore, MVD in HCC correlated with both circulating and
intra-hepatic TEM frequency (Fig. 4B). These results might suggest that TEMs are involved in the promotion of neo-vascularization in HCC.
Blood TEM frequency is superior to AFP and PIVKA-II serum levels as diagnostic marker in HCC
In order to evaluate the significance of TEM frequency as a diagnostic marker of HCC, we examined whether TEM frequency correlates with various clinical parameters in HCC patients. No correlation was found between TEM frequency and other HCC-specific markers such as a- fetoprotein (AFP) or protein induced by the absence of vitamin K or antagonist II (PIVKA-II) (Fig. 5A). In addition, circulating TEM frequency did not correlate with the levels of examined angiogenic factors, such as VEGF, ANG-2, sVEGFR-1 and MIF (Supplementary figure 2). As for the diagnostic value of TEM frequency for differentiating HCC from chronic liver disease (CLD; chronic hepatitis and liver cirrhosis patients) or liver cirrhosis, its sensitivity and specificity were 86.1 and 71% and 81.3 and 90%, respectively (Table 3). ROC analyses revealed that TEM frequency was superior to AFP, PIVKA-II and ANG-2 levels as a diagnostic marker for HCC (Fig. 5B and Table 3).
Table 3. Assessment of Diagnostic Value of TEMs, AFP, and PIVKA-II by ROC Analyses
Positive Predictive Values (%)
Negative Predictive Values (%)
The optimal cutoff point was determined as those yielding the minimal value for (1–sensitivity)2 + (1–specificity)2. Such points with those sensitivity and specificity values are the closest to the (0, 1) point on the ROC curve.
AUC, area under the curve; TEMs, TIE2-expressing monocytes; CLD, chronic liver disease; LC, liver cirrhosis; ANG-2, angiopoietin-2.
HCC and CLD
HCC and LC
In this study, we defined TEMs as CD14+CD16+TIE2+ cells and examined their frequency, localization and correlation with micro-vessel density in HCC. We show that: (i) TEM frequency is significantly increased both in PBMC and tumors of HCC patients, and positively correlates with MVD in the HCC tissue; (ii) TEM frequency dynamically changes in relation to tumor ablation or recurrence; (iii) The frequency of TEMs in the peripheral blood may serve as a better diagnostic marker of HCC than AFP, PIVKA-II and ANG-2 serum levels. These results may suggest that certain HCC-derived factors stimulate the differentiation of TEMs, whose abundance in HCC correlates with the degree of vascular densiry.
According to the pattern of CD16 and CD14 expression, it has been reported that monocytes can be categorized into distinct subsets, such as classical (CD14+CD16-) and non-classical monocytes (CD14+CD16+).20 Such populations are regarded as functionally distinct, since the frequency of CD14+CD16+ cells predominantly increases under inflammatory conditions such as chronic hepatitis and inflammatory bowel disease.21, 22 We identified TEMs within the CD14+CD16+ monocyte subset, suggesting that TIE2 is predominantly expressed/induced in the CD16+ monocytes. However, the precise mechanisms regulating TIE2 induction in monocytes remain largely unknown. Furthermore, it is yet to be clarified whether CD16+ monocytes derive from CD16- monocytes. Multiple factors are reported to enhance CD16 expression on monocytes, such as macrophage-colony-stimulating factor (M-CSF), IL-10 and transforming growth factor (TGF)-B1.23 Recent studies have suggested that hypoxia and MIF increase TIE2 expression on monocytes in vitro.14, 24 Cumulative data have been published showing that cancer cells, including HCC, are dichotomously capable of releasing various inflammatory (TNF-a, IL-1P) and anti-inflammatory (TGF-P and IL-10) cytokines, as well as hematopoietic factors (M-CSF and MIF).7, 25-29 However, no correlation was found between the frequency of TEMs and that of angiogenesis factors in the serum of HCC patients (Supplementary figure 2). It is thus plausible that a combination of HCC-derived factors contributes to the induction/expansion of TEMs, as suggested by preliminary studies (manuscript in preparation).
Although the presence of TEMs has been reported in cancer tissues from patients with colorectal, pancreatic or renal cancer,15 the actual role of TEMs in the angiogenic process of HCC remains to be elucidated. In this study, we showed that blood and intra-hepatic frequency of TEMs positively correlate with the density of microvascular structures in the liver. Additionally, we found that TEMs accumulate in the liver and are mostly located in the perivascular HCC areas. In support to our observations, Venneri et al. reported that TEMs preferentially localize in the vicinity of tumor blood vessels in human cancer specimens, but are not found in non-neoplastic tissues adjacent to tumors.15 These results suggest that TEMs may have a role in HCC-induced angiogenesis. Further studies are now needed to identify the molecular mechanisms that regulate TEM accumulation in the liver. Interactions between chemokine and their receptors, such as CCR5 or CX3CR1 (which are expressed in TEMs), may be involved in such process.22
Thus far, many studies have reported that the histological degree of angiogenesis in the liver closely correlates with prognosis or survival of HCC patients. Accordingly, several serum angiogenesis factors, such as VEGF, ANG-2, sVEGFR-1 and MIF, were reported as markers of prognosis, invasiveness and/or post-therapeutic recurrence in HCC patients.7-11 Multivariate analysis showed that the degree of angiogenesis in the liver, as assessed by MVD, is an independent predictivefactor for disease-free survival in patients with resectable HCC.17 Therefore, the positive correlation between circulating/intra-hepatic TEM frequency and MVD in the liver observed in our supports the notion that TEMs could represent a prognostic marker in HCC patients. In support of this, we found that patients with higher TEM frequency were at a more advanced Child-Pugh stage, had poorer liver function and showed higher rate of post¬therapy recurrence.
In order to diagnose HCC, serum AFP and PIVKA-II levels have been commonly used in the clinical practice. However, the sensitivity and specificity of both markers are unsatisfactory for the detection of HCC.12 In the present study, ROC analysis revealed that the frequency of TEMs in the peripheral blood has higher sensitivity and specificity than serum AFP, PIVKA-II, ANG-2 levels in differentiating HCC from CLD or LC. However, evaluation of the clinical value of TEMs in the long-term prognosis of HCC patients is needed as the observation periods in the current study were limited to a maximum of 2 years.
In summary, we found that TEMs are significantly increased both in the peripheral blood and liver of HCC patients, thus holding diagnostic value for HCC. Furthermore, the frequency of TEMs correlated with the degree of angiogenesis in HCC tissue. Thus, TEMs might represent a novel diagnostic cellular marker for HCC, potentially reflecting angiogenesis in the liver.