Clear cell renal cell carcinoma (CCRCC) is a highly vascularized malignancy. As evaluated by 2-phase multidetector computed tomography angiography, CCRCC is found to be able to develop an aberrant capsular vascular supply with outflow through an ovarian vein or testicular vein,1 although positron emission tomography shows that blood perfusion of the tumor is reduced compared with normal kidney tissues.2 Our expression profiling study also indicated that the expression level of vascular endothelial growth factor (VEGF; a potent proangiogenesis factor) in CCRCC is 5-fold higher than in other types of kidney neoplasm.3 To our knowledge, the complexity of the tumor vasculature in CCRCC has been noted previously,4 but its underlying biological mechanism remains unknown. The resistance of CCRCC cells to most of the existing cytotoxic drugs has prompted the use of novel agents targeting angiogenesis, several of which are either recently approved for therapeutic use or undergoing preclinical or clinical evaluations.5-7 The promising yet limited efficacy of the current antiangiogenic strategy points to the need to further elucidate the complexity of CCRCC vasculature. Herein, we discuss several important issues regarding the vasculature of, and angiogenesis in, CCRCC.
Clear cell renal cell carcinoma (CCRCC) is a highly vascularized cancer resistant to conventional chemotherapy and radiotherapy. Antiangiogenic therapy has achieved some effectiveness against this unique malignancy. The complexity of the tumor vasculature in CCRCC has led to differences in correlating tumor microvessel density with patient prognosis. The authors' recent findings demonstrated that there were at least 2 major categories of tumor vessels in CCRCC—namely, undifferentiated and differentiated—correlating with patient prognosis in contrasting ways, with higher undifferentiated vessel density indicating poorer prognosis, and higher differentiated vessel density correlating with better prognosis. Furthermore, the presence of pericytes supporting the differentiated vessels varied in CCRCC. The distributions of pericyte coverage and differentiated vessels in CCRCC were uneven. The tumor margin had a higher pericyte coverage rate for differentiated vessels than did the inner tumor area. The uneven distributions of pericyte coverage and differentiated vessels in CCRCC prompted the authors to revisit the mechanism of tumor central necrosis, which was also known to be a prognostic indicator for CCRCC. The discrepancy of prognostic correlation between protein and messenger RNA levels of vascular endothelial growth factor in CCRCC was discussed. The complexity of the tumor vasculature in CCRCC also led the authors to begin to re‒evaluate the therapeutic effects of antiangiogenic agents for each type of tumor vessel, which will in turn significantly broaden understanding of tumor angiogenesis and improve therapeutic effect. Cancer 2009;115(10 suppl):2282-9. © 2009 American Cancer Society.
Differential Analysis of Tumor Vasculature
Microvessel density (MVD) is usually used as an indicator of tumor angiogenesis. In the majority of solid tumors, higher MDC is correlated with poorer prognosis.8 In the patients with hematologic malignancies, increased bone marrow MVD has been found.9-11 Higher bone marrow MVD is reported to correlate with more advanced hematologic malignancies12 and shorter progression-free survival in patients with multiple myeloma.13 However, intratumoral MVD has been a controversial prognostic predictor for CCRCC. As summarized in Table 1, higher MVD has been reported in many studies to be a favorable prognostic factor (eg, a higher blood vessel density in CCRCC is correlated with a better prognosis or longer patient survival).4, 14-22 However, other studies have reported opposite results correlating higher MVD with poorer prognosis,23-28 whereas others have been unable to find a significant prognostic role for MVD.29-33 The controversy could result from many nonmechanistic factors, including sample size, sampling bias, the different blood vessel markers chosen for immunohistochemical (IHC) characterization, the quality of IHC staining, the methods of vasculature quantification, and the methods of interpretation. However, one other important cause may be that, without differential analysis, blood vessels in CCRCC may be regarded as a single type of vessel having presumably similar functions throughout, which might not accurately reflect the functional diversity of different types of microvessels.
|Study||Blood Vessel Marker||No. of Patients||Higher Microvessel Density Correlates With|
|Anastassiou 199614||CD31||23||Longer survival|
|Delahunt 199716||Factor VIII||150||Longer survival|
|Sabo 20014||CD34||49||Longer survival|
|Rioux-Leclercq 200119||CD34||73||Longer survival|
|Yagasaki 200321||CD105||84||Longer survival|
|Imao 200417||CD34||70||Longer survival|
|Sandlund 200620||CD105||168||Longer survival|
|Mertz 200718||CD34||284||Longer survival|
|Baldewijns 200715||CD34||150||Lower Fuhrman grade|
|Yildiz 200822||CD34||54||Longer survival, less metastasis|
|Yoshino 199527||Factor VIII||84||Shorter survival and higher metastasis rate|
|Nativ 199826||Factor VIII||36||Shorter survival|
|Zhang 200228||CD31||70||Advanced tumor stage|
|Joo 200424||CD34||67||Higher metastasis rate and worse prognosis|
|Fukata 200523||CD34||54||Higher metastasis rate|
|Kavantzas 200725||Factor VIII||53||Higher pathologic grade|
|MacLennan 199530||Factor VIII||97||No correlation with prognosis|
|Gelb 199729||Factor VIII and CD31||52||No correlation with prognosis|
|Schraml 200233||CD34||77||No correlation with survival time|
|Minardi 200531||CD34||48||No correlation with prognosis|
|Sandlund 200732||CD31||167||No correlation with prognosis|
Our recent study demonstrated that within CCRCC there are at least 2 major categories of blood vessels having contrasting prognostic implications: the undifferentiated vessels (expressing CD31 but not CD34) and the differentiated vessels (expressing both CD31 and CD34).34 The morphologic characteristics of undifferentiated vessels include no (or a small) lumen, a thicker vessel wall, and smaller size when compared with differentiated vessels. A higher density of undifferentiated vessels is correlated with shorter patient survival, higher grade tumors, and a more advanced primary tumor. Conversely, a higher density of differentiated vessels in CCRCC is correlated with longer patient survival, low tumor grade, and earlier T classification. These findings suggest that 1) the vasculature in CCRCC should not be regarded as 1 single type of vessel with presumably similar biologic roles in tumor progression; and 2) by differential analysis of the vasculature, we might be able to identify the key types of tumor vessels corresponding to the aggressiveness of tumor behaviors and perhaps response or resistance to antiangiogenic agents.
Tumor blood volume can be measured using contrast-enhanced magnetic resonance imaging (MRI).35 Using gadopentetate dimeglumine, which can provide contrast enhancement in tumors with functioning abnormal vasculature, Yabuki et al reported a significant tendency for higher grade CCRCCs to present with less enhancement and lower grade tumors to present with better enhancement.36 In this study, 43 of 54 patients had CCRCCs. Bolus injection of gadopentetate dimeglumine was performed for MRI. The enhancement of the solid part of the tumor was compared with the renal cortex of the same kidney. The nuclear grading system was used with grade 1 indicating a tumor with a nucleus smaller than that of a normal tubular cell nucleus, grade 2 indicating a tumor with a nucleus of similar size, and grade 3 indicating a tumor with a larger nucleus. The authors found a significant correlation between grade 3 tumors and “no or very little enhancement” in both dynamic early-phase and late-phase images, suggesting less blood volume in high-grade tumors. This finding supports the notion that the differentiated vessels, which are less common in high-grade tumors and more frequent in low-grade tumors, are the primary blood supplier to CCRCC.
Impact of Pericytes on Prognosis
Mature blood vessels are usually covered/bounded by pericytes. The recruitment of pericytes is essential for the formation and stabilization of mature blood vessels.37-39 Therefore, pericyte coverage is regarded as an indicator of vascular maturation. The absence of pericyte coverage in tumor vasculature has been found to be associated with metastasis and poorer prognosis in patients with colorectal cancer.40 Our study of CCRCC revealed that the undifferentiated microvessels, which are correlated with poor prognosis, are not covered by pericytes.34 Irregular pericyte coverage is found in a portion of the differentiated vessels. These findings indicate that, in terms of vascular maturation in CCRCC, undifferentiated vessels are completely immature, whereas differentiated vessels could be further divided into differentiated mature vessels and differentiated immature vessels, depending on the presence of pericyte coverage.
In our most recent study using quantitative analyses of double IHC-stained images, the presence of pericytes supporting the differentiated vessels in CCRCC has been further analyzed. The distribution of pericytes is uneven in the tumor mass, with abundant pericytes noted in the peripheral tumor area and fewer pericytes in the inner tumor area, resulting in a higher pericyte coverage rate on the differentiated vessels in the tumor rim. Consistent with the supporting role of pericytes to mature blood vessels, a higher density of differentiated vessels is found in the peripheral tumor area in contrast to the inner tumor area. These results suggest that pericyte coverage might play an important role in tumor progression, and refined differential analysis of CCRCC vasculature might disclose more useful targets for therapeutic intervention.
Revisiting the Concept of Central Necrosis in CCRCC
Tumor necrosis is present in approximately 50% of all CCRCCs, with a prevalence in high-grade tumors,41 and is recognized as an important prognostic indicator.4, 41-43 Recently, a scoring system including the stage, size, grade, and necrosis of the tumor (SSIGN model) has been reported to better predict the outcome of patients with CCRCC.44 Although the association between an increased number of infiltrating lymphocytes and larger necrotic areas suggests a possible role of the host immune response in CCRCC necrosis, its precise mechanism remains unclear.45
Traditionally, tumor necrosis is believed to be present when tumors outgrow their blood supply. However, the study by Hemmerlein et al revealed that highly proliferative renal cell carcinomas outgrow their vascular supply, resulting in chronic hypoxia, which in turn induces a decrease in proliferation and an increase in VEGF expression without significant necrosis or apoptosis.46 Therefore, it is suggested that tumor necrosis in CCRCC is more likely induced by acute hypoxia due to immature microvessels. In addition, challenging the conventional concept, Holash et al demonstrated in their animal model that glioma cells can co-opt normal host vessels for initial tumor growth.47 The co-opted vessels will regress along with the out-growth of the tumor, resulting in central necrosis of the expanding tumor. They even suggest that angiopoietin-2 and VEGF cooperate in the regression of the co-opted vessels. The integration of normal high endothelial venules into the tumor vasculature in breast cancer metastatic to the axillary lymph nodes has also been demonstrated in our study.48 It is therefore reasonable to hypothesize that the outgrowth of CCRCC, which is initiated in blood vessel-rich kidney tissue, may also be involved in integrating the pre-existing normal vessels into its tumor vasculature. If this hypothesis is true, then the fate of the integrated pre-existing vessels in the tumor vasculature could be the key factor for the presence of necrosis in CCRCC.
From our observation, between the tumor rim with a high density of differentiated vessels and the area of central necrosis, there usually is a transitional area with fewer or no pericytes. As the necrotic area expands outward, the pericyte-free zone will eventually be overtaken by the necrotic area. In animal models, pericytes migrate away from brain microvessels in response to hypoxia or brain trauma.49, 50 It has been found that loss of pericytes precedes damage to endothelial cells,51 and microvascular injury precedes tumor necrosis.52 Therefore, we hypothesized that loss of pericyte coverage of differentiated vessels causes dysfunction of the differentiated vessels, which in turn precedes necrosis in CCRCC. Obviously, further studies are needed to prove these hypotheses and fully elaborate the mechanisms of tumor necrosis in CCRCC.
VEGF-A and Other Proangiogenic Factors in CCRCC
Recently, angiogenic factors have gained increasing prognostic interests in a variety of solid tumors. The most important of these factors is VEGF-A. The correlation between the VEGF-A protein level and prognostic parameters in CCRCC patients has been confirmed in several studies.15, 46, 53 Higher levels of VEGF-A protein in both the tumor tissue and patient plasma have been shown to be correlated with the presence of tumor necrosis, higher tumor grades, higher tumor cell proliferation rates, and higher metastatic rates. However, VEGF-A mRNA levels in CCRCC indicate a different correlation with tumor aggressiveness and patient prognosis. In 1 study, a significantly higher VEGF-A mRNA level was found in a tumor having a lower proliferation rate.46 In another study, a higher VEGF-A mRNA level was found to be significantly correlated with longer patient survival.54 Moreover, the mRNA levels of some other proangiogenic genes are also up-regulated in tumors with less aggressiveness and are correlated with a better patient prognosis; these include the genes for angiopoietin‒2, VEGF receptor (VEGFR)‒1, and VEGFR‒2.15, 54
The striking discrepancy in the correlation to patient prognosis between the VEGF-A protein level and VEGF-A mRNA level could result from the tumor sample collection procedure, meaning that an artificially greater up-regulation of the mRNA levels of proangiogenic factors could take place in the lower grade tumor with better prognosis. The explanatory assumption is as follows. We have reported that low-grade CCRCCs have a higher density of differentiated vessels.34 Consistent with our finding, Yabuki et al reported that low-grade CCRCC tumors demonstrate more enhancement on MRI, indicating larger blood volume in the tumor.36 During nephrectomy or other related surgeries, clamping and ligations of the renal artery, renal vein, and some other draining vessels always precede the isolation of the kidney and/or the resection of the tumor. From the clamping of the blood vessels to the removal of the kidney/tumor, which is usually at least 20 minutes, a low-grade tumor will undergo a reduction of blood supply greater than that of a high-grade tumor. This sudden deprivation of blood supply could induce a more dramatic reduction in the oxygen level inside the low-grade tumor, and consequently trigger more transcription activity from proangiogenic genes before the tumor samples were frozen and the proteins were expressed. To test this idea, we mimicked surgical conditions of human kidney tumor removal using an orthotopic human CCRCC xenograft model. In preliminary experiments, we found that the duration between blood supply depletion and freezing of the tumor tissue in liquid nitrogen is critical to tumor VEGF mRNA levels. As shown in Figure 1, prolongation of this time artificially up-regulates VEGF mRNA levels, but VEGF protein levels remain the same. Although our results are highly preliminary, they are illustrative of our model that prolonged depletion of blood supply resulting from the tumor sample collection procedure may artificially alter VEGF mRNA levels.
Somatic von Hippel-Lindau (VHL) gene inactivation has been shown in up to 70% of CCRCC cases.55 The VHL gene product (pVHL), together with other component proteins, forms a complex that causes destruction of hypoxia-inducible factor‒1α (HIF-1α) under normoxic conditions. Lack of VHL protein leads to stabilization of HIF-1α, which activates a cascade of pathways, including the VEGF-A pathway. Theoretically, tumor cells with aberrant VHL genes might not produce more VEGF in response to the sudden blood deprivation during surgery. However, VEGF-A can also be produced by macrophages, fibroblasts, endothelial cells, and other cell types,56-58 which are normally able to respond to hypoxic conditions.
Future Targets for Antiangiogenic Therapy
The complexity of the tumor vasculature in CCRCC, especially the characterization of the differentiated versus undifferentiated microvessels, raises concerns about the variation of therapeutic effects on tumor endothelial cells compared with those in normal endothelial cells. Moreover, endothelial cells with pericyte coverage are usually resistant to antiangiogenic drugs because of the presence of the survival factors secreted by the pericytes.59, 60 The limitation of the therapeutic effects from current antiangiogenic drugs in CCRCC clinical trials further strengthens the notion that more realistic screening assays using tumor-derived endothelial cells and even pericytes should be developed and applied toward the identification of more effective antiangiogenic agents. Recently, Maciag et al reported that targeting the pericytes in tumor vasculature can inhibit the growth of tumors (including renal carcinoma) in animal models.61 In fact, combined therapy targeting both endothelial cells and pericytes has been proposed for clinical trials.62
In conclusion, CCRCC is an important malignancy, not only because of its frequently fatal progression, but also due to its unique pathologic and biologic characteristics, from which we might be able to discern some critical steps of tumor angiogenesis and metastasis. Differential analysis of the CCRCC vasculature is necessary for drawing prognostic implications and for re‒evaluating the therapeutic effects of antiangiogenic agents on each type of tumor vessel. Many unanswered questions remain regarding the establishment and transformation of different types of tumor vessels and the underlying molecular mechanisms. A high VEGF protein level is reportedly correlated with tumor necrosis in CCRCC and poorer patient survival; however, prudent interpretation of the mRNA levels of VEGF and other proangiogenic factors is needed. Novel screening methods for antiangiogenic drugs involving tumor-derived endothelial cells should be developed.
The questions and discussion below follow from the oral presentation given at the Third Cambridge Conference on Innovations and Challenges in Renal Cancer and do not correspond directly to the written article, which is a more general review.
Dr. Michael Atkins: Thinking about this therapeutically, I would hypothesize that less organized blood vessels with less pericyte coverage might be more sensitive to antiangiogenic therapies. Do you have any sense of how this differs between patients with just primary tumors and those who develop metastases?
Dr. Bin Tean Teh: There is a difference between the lower grade and the metastatic group, but not much difference between the higher grade and the metastatic group.
Dr. Atkins: So, the metastatic group appears to behave like the high-grade tumors.
Dr. Teh: Yes. Also, if you look at expression of c-Met, it is definitely higher in the papillary type 1 than the type 2.
Dr. William Kaelin: It would be interesting to know whether VHL mutations are more prevalent in the approximately 50% that you are calling good risk that have up-regulated VEGF.
Dr. Teh: It appears that there are more mutations in this group.
Dr. Kaelin: Treating endothelial cells in culture with VEGF or blocked VEGF signaling would be a nice place to use your gene expression profiling to try to get a robust signature or find some biomarkers.
Dr. Teh: We are doing that.
Dr. Atkins: I have seen some data presented about expression profiling where the biggest difference between the sample sets is the institution that does the expression profile rather than the actual changes within the tumor. To what extent are we now at a place where this expression profiling is standard enough that the impact of institutional variation is minimized enabling this work to be both reproducible and decentralized?
Dr. Teh: It depends on the purpose. For a diagnostic purpose, to distinguish between cell types, some of this study may be too subtle within the same type, so we probably need better control. Among those 200 genes for which we have prognostic value, we do not understand the roles of a lot of them in RCC.
We thank Vanessa Fogg for critically reading the article and Sabrina Noyes for article preparation and submission.
Conflict of Interest Disclosures
The program was made possible by educational grants provided by Genentech, Novartis Pharmaceuticals, Pfizer, Inc., and Wyeth Pharmaceuticals. Program management and CME sponsorship were provided by InforMEDical Communications, Inc., Carlisle, Massachusetts.
Funding was received from The Gerber Foundation, Hauenstein Foundation, and Schregardus Family Foundation