In a mouse model of hepatocellular carcinogenesis, highly vascularized tumors develop through two distinct morphologic phases of neovascularization. We show that increased vascular caliber occurs first, followed by extensive vessel sprouting in late-stage carcinomas. To define molecular pathways in tumor neovascularization, endothelial cells were directly purified from normal liver and advanced tumors. Gene expression profiling experiments were then designed to identify genes enriched in the vascular compartment. We report that Cathepsin S is the major protease specifically overexpressed during vessel sprouting. We also show that the CC chemokines CCL2 and CCL3 are secreted by neovessels and stimulate proliferation through their cognate receptors in an autocrine fashion. This suggests that chemokine signaling represents the most prominent signaling pathway in tumor-associated endothelial cells and directly regulates vessel remodeling. Furthermore, high angiogenic activity is associated with attenuated lymphocyte extravasation and correlates with expression of the immunomodulatory cytokine interleukin 10. This is the first comprehensive study addressing liver-specific vascular changes in a murine autochthonous tumor model. These novel insights into liver angiogenesis infer an environmental control of neovascularization and have important implications for the design of antiangiogenic therapies. (Cancer Res 2006; 66(1): 198-211)
- transgenic mice
- endothelial cells
- gene profiling
Carcinogenesis is primarily a consequence of nuclear events within transformed cell clones but equally requires stromal interactions to cause cancer progression ( 1). Tumor-associated stroma provides essential components that promote tumor cell proliferation and formation of a new vascular network. Angiogenesis is indeed a discrete and rate-limiting step that enables small, avascular tumors to develop into hypervascular, rapidly growing cancers ( 2). During neovascularization, endothelial cells, which form the lining of blood vessels, proliferate, invade into surrounding stroma, and finally form new vascular sprouts. Due to constant vessel remodeling, the tumor vasculature develops distinct and stage-specific morphologic features compared with their normal counterparts ( 3, 4). Molecular events that direct angiogenesis have been extensively studied in vitro. For instance, human umbilical vein endothelial cells (HUVEC) incubated with growth factors, such as vascular endothelial growth factor (VEGF) and basic fibroblast growth factor, are classically used to recapitulate tumor angiogenesis ( 5– 9). However, much as these in vitro systems have contributed to our understanding of vessel proliferation, they are selective for single growth factors and poorly mimic the three-dimensional stromal architecture. Furthermore, they do not account for organ specificities of microvessels. Indeed, endothelial cells are morphologically and functionally heterogeneous populations that are highly adapted to their microenvironment ( 10– 12). Importantly, there is increasing evidence that the tumor micromilieu, like its normal counterpart, shapes and imprints unique features onto the neovasculature. In vivo phage-display profiling of murine vessels, for instance, revealed profound molecular differences among endothelial cells of normal organs ( 13, 14) and vascular beds of different tumors ( 15, 16). Therefore, the molecular anatomy of tumor vessels differs between tumor stages as well as tumor types ( 15).
Comprehensive molecular analyses of endothelial cells purified from normal mouse organs are limited ( 12), and even less is known about isolated tumor endothelial cells ( 17). We have established a murine model for hepatocellular carcinoma with the goal of characterizing angiogenic changes during tumorigenesis. Transgenic mice expressing the oncogene SV40 large T antigen (Tag) under the control of the albumin promoter/enhancer (Alb; ref. 18) sequentially develop highly vascularized liver cancers. A major advantage of this model over transplanted tumors is that Tag-induced tumors spontaneously develop within the liver, enabling analysis of the neovasculature in its native architecture and location. To reveal molecular alterations induced in the vasculature of AlbTag tumors, we purified endothelial cells from normal liver and late-stage tumors and did comprehensive gene profiling experiments. Here, we present evidence for novel mechanisms intrinsic to liver tumor endothelial cells (LTEC) that regulate vessel sprouting, migration, and growth.
Materials and Methods
Mice and cell lines. The 3.8 kb Alb promoter/enhancer ( 18) was cloned upstream of a 2.7 kb fragment coding for the early region of Tag ( 19) and injected into (C57BL/6×DBA/2)F1 oocytes. AlbTag mice were backcrossed into the C3HeBFe background for 20 generations. Tag-expressing murine hepatocellular carcinoma cells (Tag-HCC) were established ex vivo from Tag+ tumors by enzymatic digestion, in vitro culture and in vivo passages through Rag-1−/− C3HebFe mice. 6 To collect conditioned medium, Tag-HCC cells were grown in DMEM containing 4.5 g/L glucose, 10% FCS, 2 mmol/L glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin for 3 to 5 days.
Intravital microscopy. The procedure has been described previously ( 3) and modified for liver by placing liver lobes on a rubber stage superfused with Ringer's solution (37°C). Statistical analysis was done by using SPSS software (version 11.5.1, SPSS, Inc., Chicago, IL). Fisher's exact test was used to compare vessel diameters. The differences of adherent leukocytes between groups were compared using Mann-Whitney U test.
Isolation of primary cell populations and fluorescence-activated cell sorting. Isolation of murine liver sinusoidal endothelial cells (LSEC) has been described ( 20). For the isolation of LTECs, AlbTag mice were injected i.p. with 200 units heparin (Sigma, Taufkirchen, Germany) and perfused via the inferior vena cava with Spinner's modified Eagle's medium containing 100 mmol/L EGTA/50 units heparin, followed by 0.025% collagenase A (Roche, Mannheim, Germany) in Williams' medium E. Tumor nodules were digested in 0.025% collagenase A for 30 minutes in a rotary water bath at 37°C. LTECs were filtered, washed, and separated in a Histodenz (nonionic density gradient medium, Sigma) gradient. Cells were further incubated with 0.01% dispase (Roche) for 30 minutes at 37°C in a rotary water bath, washed, and concentrated via a second density gradient centrifugation. For fluorescence-activated cell sorting (FACS) staining, cells were incubated with Fc block (CD16/CD32, 2.4G2, 2.5 μg/μL; BD PharMingen, Heidelberg, Germany) and specifically labeled with anti-CD31-phycoerythrin (rat IgG2a, 4 μg/mL; BD PharMingen) and ME-9F1-FITC (rat IgG2a, 30 μg/mL; ref. 21). Propidium iodide (1 μg/mL) was added to exclude dead cells. Cells were sorted using a FACSVantage SE flow cytomer (Becton Dickinson, Heidelberg, Germany). For in vitro studies, LSECs and LTECs were separated by magnetic cell sorting using biotinylated anti-CD31 antibodies and streptavidin-conjugated magnetic beads (Miltenyi Biotech, Bergisch Gladbach, Germany). Cells were seeded on collagen (Sigma)–coated flasks and grown in DMEM containing 4.5 g/L glucose, 10% FCS, 2 mmol/L glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin. Endothelial cell purity was assessed by incubation with 200 μg/mL acetylated low-density lipoprotein (LDL), conjugated with the fluorochrome 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindo-carbocyanine perchlorate (DiI, Paesel&Lorei, Hanau, Germany) for 4 hours. Tissue-infiltrating lymphocytes (TIL) were prepared as described for endothelial cells but separated on a Percoll gradient. Lymphocytes were washed in PBS and cultured in RPMI 1640, 10% FCS, 2 mmol/L glutamine, 100 units/mL penicillin and 100 μg/mL streptomycin, and 0.05 mmol/L 2-mercaptoethanol. Surface markers were analyzed on a FACScan (BD PharMingen) with the following antibodies purchased from BD PharMingen and used at 10 μg/mL: anti-CD3 (hamster IgG), anti-CD4 (rat IgG2b), anti-CD8 (rat IgG2a), anti-CD45R/B220 (rat IgG2a), anti-DX5 (rat IgM), and anti-CD11b (rat IgG2b).
RNA preparation and microarray analysis. Total RNA from 2 × 105 to 3 × 105 purified LSECs or LTECs was prepared using the Absolutely RNA Microprep kit (Stratagene, Amsterdam, the Netherlands). Total RNA from whole organs was isolated with the RNeasy mini kit (Qiagen, Hilden, Germany). RNA quality was analyzed using RNA 6000 Nano Assays and the Bioanalyzer 2100 Lab-on-a-Chip system (Agilent Technologies, Palo Alto, CA). For all probe syntheses, 250 ng RNA were used. RNA was amplified according to Baugh et al. ( 22) and modified by Kenzelmann et al. ( 23). Five micrograms of biotinylated cRNA were used for hybridization on Affymetrix murine U74Av2 arrays following the instructions of the manufacturer. Three independent probes for LSECs, LTECs, normal liver, and liver tumor were synthesized and three hybridizations per group were done. To collect the material for individual probes, organs from three mice were pooled. Public database references for the U74Av2 gene chip are available on the Affymetrix NetAffx Analysis Center (www.affymetrix.com).
Statistical data evaluation. Statistical analysis was done using the software package R, version 1.9.1 ( 24), together with libraries gcrma and limma of the Bioconductor Project, version 1.4 ( 25). Data preprocessing steps, background adjustment, normalization, and computation of GCRMA gene expression measures were done according to Wu et al. ( 26). For statistical analysis, empirical Bayes inference for linear models with factors tissue type [normal (normal liver, LSEC) or malignant (liver tumor, LTEC)] and sample type (whole tissue or purified endothelial cells) and their interaction was used ( 27). Moderated t statistics based on shrinkage of the estimated sample variance towards a pooled estimate and corresponding P values were calculated for comparisons, LSEC versus normal liver, LTEC versus liver tumor tissue, and LTEC versus LSEC. According to Favre et al. ( 12), candidates for genes/expressed sequence tags (EST) enriched in the endothelial cells were chosen by three criteria: (a) enrichment ratio R ≥ 1.5 (R = ratio of expression in the endothelial cells to expression in whole tissue), (b) difference in fluorescence intensity between the two cell types ≥50 units, and (c) unadjusted P value for the corresponding difference ≤ 0.05. P values for the comparison LTEC versus LSEC were adjusted according to Benjamini and Hochberg ( 28) to control the false discovery rate. A threshold of 0.05 for the adjusted P values and probe sets with a fold change of ≥3 and enrichment in endothelial cells of either normal or tumor tissue were used.
Quantitative reverse transcription-PCR analysis. Quantitative reverse transcription-PCR (RT-PCR) was done using real-time PCR TaqMan technology (Applied Biosystems, Weiterstadt, Germany) as described ( 29). The mouse hypoxanthine phosphoribosyltransferase (Hprt) gene served as an internal control. Primer sequences are available on request.
Immunohistochemistry. The procedure was described previously ( 29). The following reagents were used at 10 μg/mL: anti-CD31 and ME-9F1, biotinylated anti-CD3, and biotinylated anti-B220. The concentration of CCR2B (goat polyclonal IgG, C-20, Santa Cruz Biotechnology, Heidelberg, Germany) and CCR5 (goat polyclonal IgG, M-20, Santa Cruz Biotechnology) was 4 μg/mL. All secondary reagents were used at 3 μg/mL: cyanin-3 or FITC-conjugated IgG F(ab′)2 fragment goat anti-rat (Dianova, Hamburg, Germany), cyanin-3-conjugated donkey anti-goat IgG (Dianova), and streptavidin-phycoerythrin or streptavidin-FITC (BD PharMingen). Histology was analyzed using the Axioplan 2 microscope (Carl Zeiss, Hallbergmoos, Germany) equipped with Plan-Neofluar objective lenses 10×/0.30, 25×/0.08 oil, and 40×/1.30 oil. AxioCAM camera and AxioVision 3.1 (Carl Zeiss) were used for image recording. Images were processed using Adobe Photoshop software (San Jose, CA).
ELISA. Tag-HCC cells, TILs, LSECs, and LTECs were seeded on six-well plates at a density of 4 × 105/cm2. Supernatants were collected for 3 days. ELISAs were done according to the instructions of the manufacturer for mouse MCP-1/CCL2 (BD OptEIA set, BD Biosciences, Heidelberg, Germany) and mouse MIP-1α/CCL3 (DuoSet, R&D Systems, Wiesbaden, Germany) and measured by Multiskan Ascent (Labsystems, Helsinki, Finland).
LSEC proliferation. LSECs were seeded on collagen-coated 12-well plates at a density of 2 × 105/cm2 and treated for 36 hours with reagents from R&D Systems: 5 ng/mL recombinant mouse CCL2, 5 ng/mL recombinant CCL3, antimouse CCL2 antibody (rat IgG2a, clone 123616, 1 μg/mL), and antimouse CCL3 antibody (goat polyclonal IgG, 1.0 μg/mL). For the last 12 hours, 1 μCi/mL [3H]thymidine (Amersham, Freiburg, Germany) was added to the culture. Following incubation with [3H]thymidine, cells were washed in PBS and treated with 10% trichloroacetic acid for 15 minutes. Cells were washed thrice with 90% ethanol, lysed in 0.3 mol/L sodium hydroxide, and collected on nitrocellulose filters (Schleicher & Schüll, Dassel, Germany). Samples were measured in scintillation fluid (Fisher Chemicals, Heidelberg, Germany) in a liquid scintillation counter (Wallac, Turku, Finland).
Vascular transformation occurs late during liver carcinogenesis. Liver carcinogenesis in AlbTag mice is multistep, progressing from preneoplastic foci to nodular adenoma and, finally, carcinoma. Hepatocellular carcinoma eventually arises in all AlbTag mice, which live an average of 16 weeks. Angiogenesis in these mice is central to liver tumor progression, similar to the hypervascularity in poorly differentiated human hepatocellular carcinoma ( 30). We used intravital microscopy to compare transforming, tumorigenic vasculature to normal vessels ( Fig. 1A-F ). The liver has a unique microvascular system consisting of afferent arterioportal and efferent venular blood vessels. Figure 1A shows the typical microarchitecture of normal liver with an interconnecting network of sinusoids and postsinusoidal venules. Sinusoids in the mouse liver are small, usually with a diameter of 5 to 10 μm ( Fig. 1D), whereas postsinusoidal venules range from 20 to 30 μm ( Fig. 1E). Surprisingly, even in macroscopically detectable nodular carcinoma, the principal vascular architecture is not disturbed ( Fig. 1B). Instead, sinusoidal diameter was significantly increased to >10 μm ( Fig. 1D) and postsinusoidal venules to >30 μm ( Fig. 1E). Thus, the earliest vessel transformation in AlbTag liver carcinogenesis is an increase in vessel caliber. In contrast, loss of vessel hierarchy is only evident in more advanced hepatocellular carcinomas, when tumor nodules range from 2 to 10 mm ( Fig. 1C and F). Here, sinusoids and postsinusoidal venules cannot be structurally discriminated. Instead, advanced tumors display a chaotic vascular distribution with marked variability in vessel diameter, generally between 20 and 100 μm, indicative of high angiogenic activity and the formation of a true tumor-specific vasculature.
Ex vivo isolation of normal and tumor-derived liver endothelial cells. Having identified profound morphologic changes in the tumor vasculature during neovascularization, we aimed to characterize molecular differences between LTECs and the corresponding normal LSECs. We isolated these endothelial cells by FACS using specific antibodies. CD31 is an endothelial-specific antibody reacting with normal liver sinusoids ( Fig. 2A ) and heterogeneous vessels of AlbTag tumors ( Fig. 2C) but is unsuitable as a single FACS marker due to cross-reactivity with a subset of immune cells. We therefore used a second endothelial-specific antibody, purified from the hybridoma ME-9F1 ( 21). Although cross-reactive with DX5-positive natural killer (NK) cells, 7 ME-9F1 is highly endothelial specific in histology, staining normal liver sinusoids as well as tumor vessels in a pattern indistinguishable from CD31 ( Fig. 2B and D). Endothelial cells from normal liver ( Fig. 2E) and cancers ( Fig. 2F) were isolated by enzymatic tissue dissociation, enriched by gradient centrifugation, and labeled with CD31 and ME-9F1 antibodies. Cells with the highest fluorescence for both markers were sorted. To confirm identity and viability of endothelial cells, uptake of LDL modified by acetylation and conjugated with the fluorochrome DiI was measured on sorted cells plated on collagen ( Fig. 2E and F, insets). Isolated endothelial cells were at least 95% pure.
Genes involved in angiogenesis during hepatocellular carcinogenesis. To gain insight into molecular mechanisms underlying vessel remodeling, we first identified genes enriched in endothelial cells ( 12) by comparing gene expression between purified endothelial cells and their tissue of origin (normal liver or tumor). RNA from normal liver, isolated LSECs, liver tumor, and isolated LTECs was probed on Affymetrix oligonucleotide microarrays encompassing 12,000 genes and ESTs. Comparing normal liver and its corresponding purified endothelial cells (LTEC), 1,521 genes/ESTs (12.7%) were enriched in LSECs. When whole liver tumors were compared with the corresponding purified LTECs, 892 genes/ESTs (7.4%) were enriched in LTEC. Among genes enriched in both LSECs and LTECs are known endothelial markers, such as vascular endothelial cadherin; claudin 5; von Willebrand factor; VEGF receptors (VEGFR) VEGFR1, VEGFR2, and VEGFR3; angiopoietin 2 (Ang2); and the tyrosine kinase receptor 1 (Tie1), which are not differentially regulated (data not shown; ref. 31). The accumulation of well-known endothelial-specific genes in the enriched gene fractions confirms the validity of our approach.
Differential analysis of endothelial-enriched genes from normal and tumor tissue was subsequently done to show tumor-specific endothelial alterations. We identified 156 genes/ESTs being up-regulated (≥3-fold, adjusted P ≤ 0.05) and 142 genes/ESTs being down-regulated (≥3-fold, adjusted P ≤ 0.05) in LTECs compared with LSECs. Tables 1 and 2 list 100 genes that are up-regulated or down-regulated, respectively. A considerable number of differentially expressed genes cluster in three major signaling pathways, namely, notch [numb gene homologue (Numb) and adaptor protein complex (Ap1b1); ref. 32], sonic hedgehog [ornithine decarboxylase (Odc), cyclin B (Ccnb2), patched homologue2 (Ptch2), and ornithine decarboxylase antizyme 2 (Oaz2); ref. 33), and wnt (transcription factor 3 (Tcf3); casein kinase 1, γ2 (Csnk1g2) and casein kinase, δ (Csnk1d); Syr-box containing gene 17 (Sox17); and tensin-like C1 domain-containing phosphatase (Tenc1); ref. 34), which have crucial roles in cell proliferation, survival, and endothelial differentiation. Moreover, members of the transforming growth factor-β (TGF-β) signaling pathway ( 35), which are essential for normal vascular development, are significantly altered in angiogenic vessels [TG interacting factor (Tgif), bone morphogenetic protein 2 and 6 (Bmp2/Bmp6; ref. 36), latent TGF-β-binding protein 4 (Ltbp4; ref. 37), MAD homologue6 and 7 (Smad6/Smad7), and activin receptor kinase 1 (Alk1; ref. 38)]. Other genes that have previously been implicated in neovascularization and are up-regulated in LTECs include myelocytomas oncogene (myc) and B-cell translocation gene 1 (btg1), both regulators of cell growth and angiogenesis ( 39, 40). Preproenkephalin (Penk1) encodes opioid growth factor, which is important for embryonic vessel development ( 41), as is spleen tyrosine kinase (Syk; ref. 42). The urokinase plasminogen activator system is represented by its receptor (uPar) and is strongly induced in migrating endothelial cells ( 43). Inflammation promotes angiogenesis, which is reflected by up-regulation of interleukin-1β (IL-1β; ref. 44) and tumor necrosis factor receptor 2 (Tnfr2), a mediator of the tumor necrosis factor–induced angiogenic pathway ( 45). The helix-loop-helix transcription factor Dec1 (Bhlhb2) is up-regulated in tumor endothelial cells under hypoxia and may regulate cell death ( 46). In contrast, aminoacyl-tRNA synthetase (TrpRS or Wars) is an antiangiogenic molecule and is down-modulated in LTECs ( 47). Thus, the molecular profile of neovessels in hepatocellular carcinoma shows considerable parallels to previously documented pathways regulating vascular development as well as endothelial cell proliferation and migration.
Prominent features of liver angiogenesis. Vessel remodeling during tumor growth is a multistep process requiring endothelial cell proliferation and tissue invasion through degradation of extracellular matrix. Proteases, in particular matrix metalloproteases (MMP), participate in invasive growth ( 48). Our differential analysis revealed, however, that whereas only a single MMP family member, MMP12, is up-regulated in the tumor vasculature, several members of a family of cysteine proteases, the so-called cathepsins, are highly enriched in LTECs. This suggests a crucial role for cathepsins during liver angiogenesis. Cathepsins are also expressed in immune and tumor cells and may, therefore, play a more general role in tumor-stromal interactions ( 49). To evaluate the relative importance of cathepsins B, C, S, and Z and MMP12 for vascular sprouting, expression levels in different tumor constituents, including tumor cells (Tag-HCC), LTECs, and TILs, were analyzed by quantitative RT-PCR ( Fig. 3A ). Consistent with the microarray analysis, all proteases are up-regulated in LTECs. However, cathepsin S is most dramatically increased and significantly higher in LTECs than in TILs, suggesting an important function during liver neovascularization.
Surprisingly, VEGFR1 and VEGFR2 as well as neuropilin and Tie1 are expressed at similar levels in normal liver and tumor endothelium and not elevated in the liver tumor vasculature. In contrast, endothelial-specific receptor tyrosine kinase (Tie2), a receptor for angiopoietins, is significantly down-regulated in LTECs compared with LSECs ( Fig. 3B). This is an intriguing finding because VEGF and angiopoietins are important growth-promoting and vessel-stabilizing factors during angiogenesis ( 31).
Another unexpected but important finding of our microarray analysis is the abundance of chemokines and G protein–coupled chemokine receptors in LTECs. Most strikingly, a variety of CC chemokines (CCL2, CCL3, CCL4, CCL5, CCL7, and CCL8) and their cognate receptors (CCR2 and CCR5) are up-regulated in tumor endothelium, a finding confirmed by quantitative RT-PCR ( Fig. 3C). Thus, enhanced CC chemokine signaling represents one of the most significant molecular changes in neovessels in AlbTag tumors and is indicative of an inflammatory response of the vessel wall.
Vascular inflammation does not promote leukocyte-tumor endothelium interactions. Chemokines classically recruit leukocytes to inflamed tissue in a paracrine manner. Moreover, LTECs overexpress vascular cellular adhesion molecule 1 as well as endothelial and platelet selectins ( Table 1), which are thought to facilitate leukocyte adhesion. These findings motivated us to analyze leukocyte-endothelial interactions and extravasation of immune cells into liver tumors. We used intravital microscopy to monitor adherence of nonspecific leukocytes to vessels of normal liver and tumors. Postsinusoidal venules of normal liver were compared with vessels of similar diameter in AlbTag tumors (size range, 2-10 mm). The numbers of loosely ( Fig. 4A ) and firmly ( Fig. 4B) adherent leukocytes were evaluated after intravascular labeling of leukocytes. Surprisingly, despite the up-regulation of some adhesion molecules, low- and high-affinity leukocyte-endothelium interactions were not enhanced in tumor vessels. Next, TILs from solid tumors and lymphocytes from normal liver were isolated and quantified using FACS ( Fig. 4C). Normal liver contains large numbers of CD4+ and CD8+ T cells, B cells, monocytes/macrophages, and also NK and NKT cells ( Fig. 4C). These cells do not normally infiltrate the parenchyma ( 50). In contrast, AlbTag tumors contain significantly increased numbers of B cells, NK, and NKT cells, while CD4+ and CD8+ T cells and monocytes/macrophage numbers remain comparable. However, surprisingly few immune cells attach to tumor vessels or infiltrate into tumor parenchyma ( Fig. 4D and E), which sharply contrasts the massive leukocyte recruitment into inflamed liver ( 51). Thus, despite enhanced leukocyte recruitment, liver tumors do not display features of an enhanced inflammatory response. Interestingly, attenuation of lymphocyte migration into tumors coincides with expression of the immune-suppressive cytokine IL-10 ( Fig. 4F) exclusively by tumor endothelial cells.
Chemokine signaling promotes vessel proliferation in an autocrine loop. Chemokines, in particular CXC chemokines, are multifunctional and control leukocyte recruitment and angiogenesis ( 52). The majority of chemokines up-regulated in AlbTag tumor vessels belongs to the CC family. To assess the role of CCL2 and CCL3 as angiogenic growth regulators, protein secretion by cultured tumor cells, TILs, LSECs, and LTECs was analyzed using ELISA. Release of CCL2 ( Fig. 5A ) and CCL3 ( Fig. 5B) in short-term cultured LTECs is ∼6-fold increased compared with LSECs. Moreover, these chemokines are exclusively produced by endothelial cells ( Fig. 5A and B). CCR2 and CCR5, the cognate receptors for CCL2 and CCL3, respectively, are both detectable in the vasculature of AlbTag tumors by immunohistochemistry ( Fig. 5C and E). Vessel staining of CCR2 and CCR5 is noncontinuous, indicating that only subsets of endothelial cells express high levels of these chemokine receptors. Furthermore, mRNA expression of both receptors can be induced in ex vivo purified LSECs upon stimulation with liver tumor-conditioned medium and increases over time ( Fig. 5D and F). Simultaneous expression of chemokines and their cognate receptors on LTECs implies a potential autocrine growth regulation of endothelial cells within the tumor bed. To assess this further, LSECs were induced by conditioned medium to express CCR2 and CCR5, then stimulated with purified CCL2 and CCL3, and the proliferative response was measured. Incubation of LSECs with tumor-conditioned medium alone enhances proliferation, which is in part mediated by VEGF (800 pg/mL in 5-day Tag-HCC-conditioned medium). Addition of recombinant CCL2 or CCL3 induced proliferation by 3- to 4-fold, which was abolished in the presence of neutralizing antibodies. Proliferation was further increased (5-fold) when LSECs were incubated with CCL2 and CCL3 simultaneously ( Fig. 5G). Therefore, expression of chemokines and their cognate receptors on tumor vessels in vivo are part of a synergistic autocrine growth loop, which induces endothelial cell proliferation in vitro and may be crucial for vessel remodeling during liver angiogenesis.
Neovascularization is essential for solid tumor progression and is an integral element of tumorigenesis. Although different tumors frequently share common biological principles, each tumor possesses a unique environment that shapes the phenotype of resident cells, including the vasculature. Indeed, the present study reveals novel aspects of tumor vessel remodeling that have not been apparent by studying in vitro models of angiogenesis.
We show with intravital microscopy of AlbTag tumors that increased vessel caliber was the first detectable abnormality, followed by vessel sprouting. Notably, increasing vessel heterogeneity and chaotic vessel organization were only observed late in tumorigenesis, when the carcinoma had progressed beyond the nodular stage. These findings are similar to those in our earlier studies, using a mouse model of insulinoma ( 3) and rat hepatocellular cancer ( 4), where early phases of angiogenesis are marked by an increase in vessel diameter rather than vessel numbers. Thus, expansion of the endothelial surface area is the first adaptive process of the vasculature during tumorigenesis.
To focus on molecular changes intrinsic to the vasculature, we first identified genes enriched in purified endothelial cells and then compared expression profiles between purified endothelial cells of normal liver and liver tumors, to identify tumor-specific changes. There are currently limited numbers of molecular studies on normal ( 10, 12) or tumor vessels ( 17) because of technical difficulties of ex vivo purification of endothelial cells. However, our microarray studies identified a considerable number of genes shown by gene targeting experiments to be critical, thus confirming the effectiveness of our strategy. These genes include the tyrosine kinase receptors of the VEGF and angiopoietin signaling pathways, hedgehog signaling, and “neuronal” genes, such as neuropilin, ephrins, and members of the notch and delta families. Typically, both up- and down-regulation of these molecules results in vascular defects ( 53, 54) and, indeed, various members of these families are differentially regulated during liver neovascularization ( Tables 1 and 2). Interestingly, some genes that we have now identified as modulators of liver angiogenesis are reported to impair placental or cardiac development. Cysteine-rich protein 61 (Cyr61), for instance, is an extracellular matrix–associated angiogenic inducer strongly expressed in liver neovessels as well as in proliferating endothelial cells in vitro ( 5, 7) and plays an essential role in placental development ( 55). Ltbp4 and bmp2, both molecules of the TGF-β signaling pathway, are essential for heart development ( 36, 37) and are shown to be down-regulated in LTECs. Similarly, the transcriptional cofactor friend of gata-1 (Fog1 or zinc finger protein multitype 1, Zfpm1; ref. 56) and polycystic kidney disease-2 (Pkd2), mutations of which also cause autosomal dominant polycystic kidney disease ( 57), are both necessary for normal cardiac function and are significantly down-regulated in LTECs. Thus, a constellation of genes with known effects on embryonic, placental, and myocardial vascularization is likely to also have overlapping functions in liver angiogenesis.
The Tie2 receptor, which mediates signaling of Ang1, and its antagonist Ang2, is also decreased in AlbTag tumor vessels. Because interaction of Ang1 and Tie2 strengthens cell junctions ( 58, 59), impaired signaling may prevent vessel stabilization during neovascularization. Surprisingly, however, receptors for VEGF, such as VEGFR1, VEGFR2, and neuropilin, prototypic mediators of angiogenic signaling, are expressed at comparable levels in LSECs and LTECs. This provocative finding implies that VEGF signaling is as important for tumor growth as for normal endothelial homeostasis. Indeed, VEGF has been implicated in the homeostasis of lung endothelium ( 12) as well as in neuroprotection ( 60). Alternatively, VEGF might be involved in the early phases of angiogenesis, whereas late-stage tumors rely on different growth factors to promote neovascularization. The fact that liver tumor angiogenesis progresses through at least two different morphologic phases ( Fig. 1) supports this view. Additional evidence that different mechanisms might operate in the early versus late stages of angiogenesis comes from the finding in pancreatic insulinomas that MMP-9 is particularly important in mobilizing VEGF during the “switch” from vascular quiescence to angiogenesis ( 61). Moreover, in this model of islet carcinogenesis, metalloproteases and proteases of the cathepsin family have nonoverlapping functions at different stages of tumor development ( 49). Intriguingly, we have found that cathepsins, and not MMPs, are the dominant proteases in endothelial cells of late-stage hepatocellular carcinoma. In particular, cathepsin S is dramatically increased in LTECs compared with LSECs and may, therefore, play a major role in the control of endothelial cell invasion. Although cathepsin S deficiency impairs wound repair ( 62), this is the first report of cathepsin S being critically involved in tumor angiogenesis.
Increased expression of chemokines and chemokine receptors is also prominent in liver tumor angiogenesis. The CXC chemokine family plays a pivotal role in angiogenesis and, indeed, the CXCR4 receptor is highly overexpressed in LTECs, consistent with results from in vitro angiogenesis assays ( 6, 9) and gene-deficient mice ( 63). Chemokines attract different leukocyte populations that, in turn, may induce angiogenesis. Indeed, increased numbers of NK, NKT, and B cells are recruited into AlbTag tumors. Surprisingly, however, leukocyte adherence in situ is not significantly different between normal and tumor tissue. Moreover, leukocyte extravasation into tumor parenchyma seems attenuated and liver tumors are only diffusely infiltrated. These findings are similar to those in a rat hepatoma model, where chemoattractant-superfused tumors lack leukocyte-endothelium interactions ( 64), and in human hepatocellular carcinoma, where decreased numbers of infiltrating CD8+ T cells were observed in tumors ( 65). Impaired tumor infiltration by immune effector cells may indeed contribute to the immune evasion of tumor and an attractive mechanism for the impaired infiltration is the secretion of inhibitory factors. Interestingly, endothelial cells possess immunomodulatory properties and LSECs, as organ-resident antigen-presenting cells, can induce tolerance ( 11, 20). Intriguingly, LTECs secrete IL-10, which has been shown to be immunosuppressive for circulating dendritic cells in patients with hepatocellular carcinoma ( 66). Furthermore, IL-10 secretion defines a prototype of regulatory, tissue-resident antigen-presenting cells that are able to induce IL-10-producing regulatory T cells with the capacity to suppress immune responses ( 67).
Although chemokines recruit leukocytes, lack of dense infiltrates in AlbTag tumors indicates an additional role for chemokine signaling during liver carcinogenesis. Of particular interest is the finding that CCL2 and CCL3 are coexpressed on ex vivo purified neovessels together with their cognate receptors CCR2 and CCR5, respectively. Although CC chemokines are classically not considered to be angiogenic, there is precedence for CCR2 expression in some vascular cells and recombinant CCL2 induces migration of HUVEC in a dose-dependent manner ( 68, 69). CCR5 is expressed on human brain microvessels ( 70) but growth stimulation by CCL3 has not been assessed. Here, we show functional CCR2 and CCR5 receptor expression on AlbTag tumor vessels and show that this angiogenic effect can be mimicked in vitro by incubating LSECs with tumor-conditioned medium. Subsequent ligand engagement then leads to enhanced endothelial proliferation. Thus, simultaneous expression of CCL2 and CCL3, together with their cognate receptors on LTECs, critically controls vessel remodeling in AlbTag tumors in an autocrine manner.
Our data also show that chemokines may play a broader role in angiogenesis than previously appreciated. In this context, it is of particular interest that late-stage AlbTag tumors incorporate bone marrow–derived endothelial cell precursors into neovessels, a process that involves chemokine signaling and can be blocked by interfering with G protein–coupled receptor signaling. 8 Recruitment of progenitors into tumor lesions is a feature of late angiogenesis and is consistent with our intravital microscopy findings and the notion of multistage angiogenesis. Thus, our novel findings in liver-associated angiogenesis emphasize the importance of tumor-specific as well as stage-specific mechanisms for neovascularization and will confer critical insights for therapeutic interference.
Grant support: Deutsche Forschungsgemeinschaft grant SFB405 and Cooperation Program in Cancer Research of the Deutsches Krebsforschungszentrum and Israel Ministry of Science.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Ludmila Umansky and Christine Schmitt for excellent technical assistance, Klaus Hexel and Manuel Scheuermann for operating the FACSVantage SE flow cytometer, Mark Kenzelmann for advice on Affymetrix probe synthesis, and Manfred Hergenhahn for Affymetrix chip hybridizations.
Note: E. Ryschich and P. Lizdenis contributed equally to this work.
↵6 S. Stahl, unpublished results.
↵7 A. Hamann, personal communication.
↵8 H. Spring, T. Schüler, B. Arnold, G. J. Hämmerling, and R. Ganss. Chemokines direct endothelial progenitors into tumor neovessels. Proc Natl Acad Sci U S A, in press.
- Received May 12, 2005.
- Revision received September 26, 2005.
- Accepted October 18, 2005.
- ©2006 American Association for Cancer Research.