As a tumor grows, the existing blood supply becomes inefficient at supporting the tissue, and areas of the tumor become hypoxic. The hypoxic condition triggers the tumor to enhance the expression of angiogenic factors, triggering the formation of new blood vessels to support the growing tissue 12. Angiogenesis is required for tumor survival as well as further growth, progression and metastasis 3. In fact, high tumor vascular density is correlated with negative patient outcomes, including shorter progression-free interval and reduced overall survival 145.

One angiogenic factor whose expression is induced in response to a hypoxic environment is vascular endothelial growth factor (VEGF). VEGF belongs to the cystine-knot family of growth factors 4. Four homologous polypeptides for VEGF exist, derived by alternative splicing of mRNA 2. VEGF is secreted by cancer cells as well as supporting stromal cells, including fibroblasts, especially during conditions of hypoxia 1. In vitro studies have shown that stromal cells cultured in hypoxic growth conditions secrete higher levels of critical angiogenesis-inducing factors than cells cultured in normoxic conditions 3. High expression of VEGF is observed in many tumor types and is correlated with aggressive tumor growth and metastasis 567.

Regulation of VEGF is complex, occurring at both the transcription and translation stages of protein synthesis, with many ligand-receptor interactions 238. Expression of VEGF is up-regulated by hypoxia inducible factor-1 (HIF-1), which binds to the VEGF promoter, increasing transcription of VEGF 910. Once expressed, VEGF has the ability to bind to two endothelial cell-specific receptors, kinase domain receptor (KDR, VEGFR2) and fms-like tyrosine kinase (Flt-1, VEGFR1) to initiate angiogenesis among other survival signals 411. While VEGF binds to Flt-1 with 50-fold higher affinity, KDR binding is more important for angiogenic responses 4. Brogi et al. found hypoxia induced a 13-fold increase in the number of KDR receptors per endothelial cell in vitro, which may be the mechanism of action for the pronounced effect of hypoxia and VEGF in vivo 12. In addition to simulating endothelial cell proliferation and migration, VEGF increases vasculature permeability, earning its other name as vascular permeability factor (VPF) 81013. This vascular leakage is critical for initiating angiogenesis as it allows proteins, such as matrix metalloproteases (MMPs), to be deposited in the extracellular fluid 9. MMPs break down the extracellular matrix to enable endothelial cells to migrate and invade areas in close proximity to the tumor 9.

In addition to VEGF, a number of cytokines, chemokines, and growth factors are involved in angiogenesis. The eleven factors tested in this study, summarized in Table 1, were chosen because of their implication in altering vascular structure and the availability of Enzyme-Linked ImmunoSorbent Assays (ELISAs) for quantitative measurement. These angiogenesis-related factors fall into a number of general categories. Some work by mediating VEGF production, such as basic Fibroblast Growth Factor (bFGF/FGF-2) and Epidermal Growth Factor (EGF) 691114. Others work by modifying the extracellular environment of the tumor, including bFGF, Interleukin-8 (IL-8/CXCL8), and Platelet-derived Growth Factors-AA and -AA/BB (PDGFs) 6151617. Induction of endothelial cell growth is accomplished by IL-8, Fms Related Tyrosine Kinase (Flt-3 Ligand), and PDGFs, while EGF and Transforming Growth Factors-β1, β2, and β3 (TGFs) are involved in tumor growth and proliferation 915181920. Lastly, IP-10/CXCL10 inhibits tumor and endothelial cell growth and is inversely correlated with VEGF production 21.

VEGF production by tumor cells is of particular interest as this growth factor is being targeted by new anticancer agents. Specifically, Bevacizumab (Avastin^®, Genentech) is a recombinant humanized monoclonal antibody, approved for the treatment of colorectal cancer and non-small cell lung cancer treatment by the FDA 5722. This drug binds VEGF with high specificity, neutralizing the growth factor and preventing the interaction of VEGF with its receptors. Therefore, proliferation of endothelial cells is inhibited and tumor progression is hindered 811.

Based on physiological in vivo conditions, it was hypothesized that cells grown in a hypoxic in vitro environment will express angiogenic factors, including VEGF, at higher levels than those grown under normoxic conditions. A secondary goal of this study was to determine whether primary tumors exhibit differential expression of angiogenic-related factors, a phenomenon which may be useful in predicting patient response to anti-angiogenic anticancer agents.

This study addressed a number of topics related to the expression of angiogenesis-related factors in normoxic versus hypoxic environments. Specifically, (1) linear correlations exist for a number of angiogenesis-related factors, (2) linear correlations for VEGF exist and group by tumor type, and (3) primary expression levels vary between samples and across factors.

Linear correlations between protein expression in normoxic and hypoxic environments exist for eight of the eleven angiogenesis-related factors tested in this study (Table 2). Hypoxic expression levels were generally higher than normoxic for IL-8 (r² > 0.95) and VEGF (r² > 0.60), though only modestly. Both of these factors are expressed to induce vascular growth due to hypoxia in vivo, and appear to do the same in vitro. The degree of difference was surprising, as both IL-8 and VEGF have been reported to be up-regulated in response to hypoxic conditions. IL-8 regulates angiogenesis by promoting survival of endothelial cells, stimulating matrix metalloproteinases, and increasing endothelial permeability 1516. VEGF is a major signalling protein for angiogenesis secreted in higher levels when cells experience hypoxia 4. IP-10, an anti-angiogenic factor, had lower expression levels in the hypoxic condition than in the normoxic condition (r² > 0.80). This was expected, as this protein inhibits tumor growth by regulating lymphocyte chemotaxis and inhibiting endothelial growth 21.

Trends in the expression levels of other growth factors were variable. Lower expression levels were observed in the hypoxic condition for PDGF-AA (r² > 0.95) and similar levels were observed for PDGF-AA/BB (r² > 0.80). These results are not surprising as platelet populations are minimal in culture. These cells are non-adherent to flask surfaces and are rinsed away during routine media changes. Different results were observed for each of the transforming growth factors, likely related to the specific role each plays in cancer pathogenesis 18. Lower expression levels were observed in the hypoxic condition for TGF-β1, while similar expression levels were observed in both conditions for TGF-β2 (r² > 0.80) and no correlation existed for TGF-β3 (r² < 0.25). Similar expression levels were observed in both conditions for EGF, which may be due to the fact that EGF induces VEGF, IL-8, and bFGF release by tumor cells, and is transformed in the process 9. A correlation did not exist for bFGF, which mediates VEGF production and induces extracellular matrix formation. Another in vitro study showed bFGF was unaffected by hypoxia in cell lines 3. In all, the correlations between the conditions in vitro suggest the expression levels may be linked to in vivo expression of each angiogenesis-related factor, whether measured in normoxic or hypoxic conditions.

The combined results of all cell sources analyzed for VEGF showed a moderate correlation between normoxic and hypoxic expression levels. Stronger linear correlations were observed for breast and lung samples specifically. Breast and lung samples are cultured in unique culture media as compared to ovarian, CNS, and colon samples. Primary breast tumors are cultured in Mammary Epithelial Growth Media (MEGM), while lung tumors are cultured in Bronchial Epithelial Growth Media (BEGM). These media require addition of SingleQuots^® to basal media that include EGF. Significantly, EGF induces VEGF, IL-8, and bFGF release by tumor cells 9. While this SingleQuots^® may have contributed to the VEGF production in these tumor types, the other analytes (IL-8 and bFGF) induced by EGF did not correlate by tumor type (data not shown). Therefore, culture media is probably not responsible for the differential expression levels of the ten evaluable angiogenesis-related proteins and a unique fingerprint for each sample. In general, these data suggest in vitro expression levels of VEGF can be measured in either a normoxic or a hypoxic condition, since a linear correlation exists between expression levels in both conditions.

Differential protein expression levels existed for each factor tested in this study, as is evident in Figure 3. In vitro studies show differential degrees of primary tumor response to chemotherapy agents. These response rates correlate with progression-free interval in ovarian cancer patients, which indicates in vitro tests performed on primary cultures may be used to enhance the probability of choosing the best treatment regimen for the patient 24. Similarly, differential protein expression levels were observed across patients in this study for each of the factors. This suggests it may be possible to build a predictor for angiogenesis-related anticancer agents using an array of protein expression levels observed in vitro.

There are limits to the application of these in vitro results to the in vivo condition. The tumor microenvironment in vivo is unique both in its three-dimensional structure and the chemical environment 25. This affects cellular behaviour, including response to chemotherapeutic agents. Some researchers have successfully developed culture systems that replicate this three-dimensional interaction of cells 26. This study, however, employed a monolayer culture system specifically designed to enrich the population of malignant epithelial cells 27. While toxicity, delivery, metabolism, and clearance affect patient response to therapeutics in vivo, in vitro studies are commonly used in initial testing of novel treatments and have clinical potential when applied 28.

Although extreme hypoxic conditions may compromise the health of the cells and lead to cell death, similar confluencies between the normoxic and hypoxic condition at the conclusion of testing suggest that the 48 hour incubation prior to testing was sufficient for cell adherence and equilibration (Figure 1). To support this observation, Pilch et al. found that hypoxia did not cause cell death/decreased confluency, as dead cells were not observed in supernatant post-hypoxia 1. Furthermore, the hypoxic oxygen concentration used in the study in vitro is similar to that reported by Hockel and Vaupel, 2001, in the core of solid tumors in vivo 29.

Also, in addition to the 11 angiogenesis-related analytes chosen for testing, a negative control unrelated to angiogenesis was also assessed. A chemotactic cytokine, RANTES, is responsible for recruiting leukocytes and activating natural killer cells 23. This cytokine was not expected to vary in a normoxic versus hypoxic environment, and we found similar expression levels of RANTES for both conditions validating our technical approach.

Multiple techniques are available to assess VEGF expression. Some laboratories employ immunohistochemical (IHC) analysis to determine VEGF receptor levels 5, usually for diagnostic and prognostic purposes. However, this study employed the Beadlyte^® CytokineProfiler™ Testing Service for two reasons. First, this service provides quantitative analysis of the expression levels of the angiogenesis-related factors, including VEGF. Second, testing was performed on malignant epithelial cell cultures, rather than tissue sections. Intact tissue sections that contain tumor cells as well as support tissue and vasculature are generally stained using IHC. VEGF receptors on endothelial cells and monocytes fluoresce. The described culture process selects specifically for malignant epithelial cells 27. Endothelial cells are selected against by culture conditions, as the media employed do not promote the growth of these cells; monocytes are non-adherent, so are rinsed away in routine media changes 27. Neither of these cell types is present in the described samples so IHC of the VEGF receptors was not possible.

VEGF production was of most interest to this study due to its role in the mechanism of action of bevacizumab. The testing conditions were optimized to ensure that VEGF production was measurable, so VEGF results were available for all samples tested. Table 3 includes the summary of all data in the "All Samples" field, a total of 50 samples. Results for the other ten analytes had detection levels out of range of the standard curve for at least two samples, if not more. As a result, the sample size for most of these angiogenesis-related factors was less than 50 (Table 2). However, nine of these ten factors had at least 20 samples available for analysis, and were considered evaluable in the study.

As with any anticancer therapeutic agent, there is clinical ambiguity regarding individual patient response. Some agents directly target VEGF, such as bevacizumab, a humanized monoclonal antibody, while others indirectly target receptors and downstream regulators, such as sunitinib and rituximab 8. While the regulation and metabolism are unique in vivo, the protein expression levels produced by individual patient cells may provide information on how each patient will respond clinically to a given anticancer agent. The heterogeneity of protein expression demonstrated in this study may provide information to enable the prediction of the efficacy of anti-angiogenic factors. Further studies correlating the in vitro expression levels with patient outcome are warranted.

Linear correlations exist between expression levels of angiogenesis-related factors under normoxic and hypoxic conditions. This suggests the behaviour of primary cells derived from patient tumors grown under in vitro normoxic conditions may provide a correlation to the in vivo hypoxic environment. Differential expression for each sample across all factors suggests predictive value for angiogenesis-related anti-cancer agents, using not only VEGF, but an array of angiogenesis-related proteins. These data suggest further studies should be considered to correlate in vitro expression of these proteins with in vivo patient response to anti-angiogenesis therapeutics.

Drs. Brower and Bush, and Ms. Heinzman are employees of Precision Therapeutics, Inc. and have declared a financial interest in the company, which supported this work.

JH performed the cell culture studies and drafted the manuscript, SB participated in the design of the study and review of the manuscript, and JB coordinated the manuscript and assisted in statistical analysis.