Abstract
Background: the vascular endothelial growth factor (VEGF) pathway plays a prominent role in the growth and progression of human cancer, including non-small cell lung carcinoma (NSCLC). The key mediators of VEGF signaling are a family of related receptor tyrosine kinases that include VEGFR1, VEGFR2, and VEGFR3. The relative expression levels, activity, and cross-talk among these receptors may contribute to response of NSCLC to anti-angiogenic therapies, and a better systematic, translatable approach to categorizing tumors is needed. Materials and Methods: We comparatively evaluated immunohistochemical expression of the three VEGFRs in archival primary NSCLC tissues (n=96). Results: VEGFR1 and VEGFR2 were localized both in vessels and tumor cells, while VEGFR3 was only localized in tumor vessels. A set of eight VEGFR staining subclasses were identified: Triple VEGFR positive (n=11, 11.5%), VEGFR1 predominant (n=22, 22.9%), VEGFR2 predominant (n=9, 9.4%), VEGFR3 predominant (n=3, 3.1%), VEGFR1/2 predominant (13, 13.5%), VEGFR1/3 predominant (2, 2.1%), VEGFR2/3 predominant (n=8, 8.3%), and triple VEGFR negative (n=28, 29.2%). An objective categorization based on K-means clustering revealed four clusters, three of which showed high VEGFR2 compared to VEGFR3 (30.7% of cases), cases high in both VEGFR2 and VEGFR3 (18.2%), and cases that were negative/low for both VEGFR2 and VEGFR3 (45.4%). A positive association between VEGFR2 and VEGFR3 was found, however no associations were observed between VEGFR1 and VEGFR2, nor VEGFR1 and VEGFR3. Conclusion: The proposed subclasses of NSCLC are an approach for complementing lines of investigation of anti-angiogenic therapies beginning with systematic characterization of the disease.
The molecular mechanisms of angiogenesis are exploited by numerous human cancer types to stimulate tumor growth and progression (1, 2). These include cancer of the lung, the leading cause of cancer death worldwide (3, 4). In non-small cell lung carcinoma (NSCLC), several anti-angiogenic agents have shown clinical efficacy in various histological subtypes (5-9), however NSCLC continues to be one of the most frequent and aggressive forms of human cancer (10, 11). In order to better understand and to target resistance to anti-angiogenic mechanisms, an improvement in reliable characterization of the most appropriate biomolecular factors that govern the complex architecture of angiogenic signaling in NSCLC may be needed (12).
Many factors work in concert to elicit effects on angiogenic processes, but the most prominent factors are vascular endothelial growth factor (VEGF) and its cognate receptor, VEGF receptor 2 (VEGFR2) (2, 13). VEGFR2 belongs to a class of receptor tyrosine kinases exhibiting seven immunoglobulin-like domains in their extracellular regions (type V) and includes family members VEGFR1 and VEGFR3 (14). These receptors are commonly located in the vascular and lymphatic endothelium within tumor stroma and are involved in the differentiation, maintenance, proliferation, and/or migration of endothelial cells (15, 16). A number of studies have evaluated expression of the VEGFRs with or without parallel assessment of vascular density and various VEGF ligands in NSCLC tissues (17). These include more focused analyses of the VEGFA growth factor and its two receptors, VEGFR1 and VEGFR2 (18), as well as profiling of VEGFC and/or VEGFR2 and VEGFR3 in the tumor vasculature (19-22). While individual VEGFRs have been shown to have prognostic relevance (23), systematic analysis of vascular expression of the three VEGFRs using a quantifiable scoring approach has not been uniformly applied in these studies.
Recently, we implemented immunohistochemical (IHC) assays for two members of the VEGFR family following a rigorous development paradigm (24, 25). Here, we developed an IHC for a third VEGFR family member and applied these assays to an independent series of archival primary NSCLC with the following aims: i) characterization of the heterogeneity of IHC-based expression of the three VEGFRs in NSCLC tissues, ii) determination of associations among the prevalence of these three VEGFRs in NSCLC, and iii) identification of prominent subsets of patients with NSCLC based on relative expression of the three VEGFRs. As the relative expression and interaction among these receptors may allow a better understanding of the intricacies of VEGF signaling, we propose this systematic VEGFR protein profiling approach to complement established disease characterization methodologies in NSCLC (26).
Materials and Methods
Human tissue specimens. Commercially available human tumor tissue microarrays (TMAs) were purchased from TriStar Technology Group (Rockville, MD, USA) for this study. Use of human tissues in this study followed standard operating procedures of TriStar, and considered informed donor consent, Institutional Review Board approval, patient anonymity, compliance with current international and European Union regulations.
A human NSCLC TMA comprising samples from 111 patients was made from formalin-fixed paraffin-embedded (FFPE) tissue blocks collected by a standard procedure that considered tissue preservation (TriStar). Clinical data were reviewed by team of pathologists and oncologists. Under supervision of qualified consultant pathologists, NSCLC tissues were collected from surgically resected primary tumor tissues from adult patients, processed, and used to construct TMAs. A single 1 mm core was obtained from the most representative area of the donor block from each patient. The selection of cases represented a random sample of various histopathological subtypes, histological grades and stages of NSCLC. Each case was histologically typed by the Tristar Pathologists, based on the histopathological criteria proposed in the World Health Organization (WHO) Classification of Lung Tumors (27, 28). The hematoxylin and eosin-stained TMA slide was also reviewed by a board-certified anatomic pathologist (AN) for confirmation of histopathological diagnoses and to update the submitted diagnoses of bronchioloalveolar carcinomas (n=5) to adenocarcinoma (ADC) with lepidic pattern, using the latest (2015) WHO criteria (27, 28). Any cores featuring equivocal histology, technical failure, attrition of sampled cores, or which had missing clinicopathological data (e.g. demographic information) were excluded, resulting in 96 evaluable NSCLC specimens. Patients were followed-up for up to 14 years or until death. Clinicopathological characteristics of patients and their specimens are summarized in Table I. For VEGFR3 assay development, additional human tumor blocks were obtained from commercial sources (Asterand, Detroit, MI, USA; and Tristar), including angiosarcoma, and colonic adenocarcinoma (CRC) specimens. Acquisition and processing of these tissues was confirmed to be in line with rigorous human tissue-acquisition protocols that ensure collection and supply of quality human tissues for novel biomarker studies. Asterand states that collaborators obtained the following: approval of institutional and independent review board, authorization from a privacy officer, and necessary government licenses and industry accreditations. Informed consent documentation used by Asterand was subjected to review and approval by regulatory and ethics authorities. If necessary, Asterand obtained a waiver of informed consent from an institutional review board to enable the use of the tissues and clinical information for exploratory research.
Cell culture and processing of cell lines. All cells were obtained from the American Type Culture Collection (Manassas, VA, USA). All media were supplemented to include 1% (v/v) penicillin/streptomycin and 10% fetal bovine serum (FBS). HEL (erythroleukemia) cells were expanded using RPMI-1640 with 10 mM HEPES, and 1 mM sodium pyruvate. A549 (lung carcinoma) cells were propagated with F-12K medium. U87MG (glioblastoma) with Eagle's minimum essential medium with Earle's balanced salt solution supplemented with both 0.5 mM sodium pyruvate and 1% non-essential amino acids. All cells were propagated at 37°C and 5% CO2, then harvested by incubation in 0.25% trypsin with 0.53 mM EDTA at 37°C, and then collected by pipetting. Cells were pooled and then aliquoted for: i) paraffin embedding by fixation in 10% neutral buffered formalin (NBF) for 24 h, or ii) western blotting by pelleting and lysing as described below. Histotechnological preparation of cell lines was performed as described previously (24).
Western blots. Whole-cell extracts were prepared by re-suspending in protease and phosphatase inhibitor-supplemented RIPA buffer (Thermo Scientific, Waltham MA, USA). Jurkat (acute T-cell leukemia) whole cell lysate was included (Abnova, Walnut, CA, USA). Samples were combined with loading buffer containing sodium dodecyl sulfate and dithiothreitol, placed in a 95°C heat block for 3 min, and then separated on NuPage 4-12% bis-tris polyacrylamide gels (Thermo Scientific). Recombinant proteins were added as controls: a dilsufide-linked homodimer of VEGFR3 N-terminal region/Fc chimera (P35916_a.a. 25-776 plus 473-776; R&D Systems, Minneapolis, MN, USA; rECD-A) and HIS-tagged N-terminal region of VEGFR3 (NP_002011.2_a.a. 1-775; rECD-B) produced by transient transfection, then purified on nitrilotriacetic acid affinity and size-exclusion chromatography columns. Gels were transferred to nitrocellulose membranes and probed with a primary antibody to VEGFR3 (9D9F9; EMD Millipore, Billerica, MA, USA) overnight at 4°C with agitation. Blots were incubated with species-specific, horseradish peroxidase-conjugated secondary antibodies for 1 h then visualized using enhanced chemiluminescence detection (Pierce, Thermo Scientific). A glyceraldehyde 3-phosphate dehydrogenase-directed primary antibody (clone 14C10; Cell Signaling Technologies, Danvers, MA, USA) was used to verify equal protein loading on gels. All primary antibodies were applied at a working dilution of 1:1000.
Immunohistochemistry. We developed and optimized specific and selective assays for evaluation of VEGFR1, VEGFR2, and VEGFR3 in archival human tissues. Staining methodologies for VEGFR1 (using a rabbit monoclonal IgG, 1303-1; Abcam/Epitomics, Burlingame, CA, USA) and VEGFR2 (using a rabbit monoclonal IgG, 55B11; Cell Signaling Technologies) IHC were performed as previously described (24, 25, 29). To detect VEGFR3, FFPE sections of the NSCLC TMA were cut, dried, and then baked at 60°C prior to staining. Slides were deparaffinized and rehydrated on the Bond III automated stainer (Leica Biosystems, Buffalo Grove, IL, USA), and antigens were retrieved at 100°C for 40 min in EDTA-based buffer at pH 9 (ER2). Endogenous peroxidases were treated with peroxide block for 5 min, and then a protein block (PowerVision Super Block; Leica Biosystems) was applied for 10 min. Anti-VEGFR3 mouse monoclonal IgG (clone 9D9F9; Millipore) was applied at ~1.0 μg/ml (based on a 1:3000 dilution of stock ascites fluid) in antibody diluent for 15 min. A post-primary reagent (Leica Biosystems) was applied for 8 min to enable detection of mouse primary antibodies. Refine HRP Polymer (Leica Biosystems) was next applied for 8 min, 3,3’-diaminobenzidine (DAB Refine Chromogen; Leica Biosystems) was applied for 10 min, and then slides were counterstained with hematoxylin for 5 min. Slides were removed from the stainer, dehydrated by sequential submersion in 95% ethanol, 100% ethanol, and xylene, and coverslipped following routine procedures. Reagent negative controls and isotype-specific IgG were used to assess non-specific staining for each tissue. Serial sections of tissues were stained at a Clinical Laboratory Improvement Amendments-certified reference lab (Clarient, Aliso Viejo, CA, USA) using the well-established IHC assay protocols for podoplanin (D2-40), and CD34 (QBEnd/10), for lymphatic and blood vessels, respectively (30, 31). To represent various individual or combined VEGFR staining patterns, images were obtained from high-resolution digital scans at ×200 original magnification (Scanscope XT; Aperio Technologies, Vista, CA, USA).
Pre-absorption assays. For initial selectivity assays, N-terminal/Fc chimera recombinant proteins for VEGFR1 (AAC50060_a.a. 27-687), VEGFR2 (AAC16450_ a.a. 20–764), and VEGFR3 (rECD-A, as above) were obtained from R&D Systems. When performing pre-absorption assays, the diluted anti-VEGFR3 was combined with a 25-fold molar excess of recombinant proteins in antibody diluent. Mixtures were incubated overnight (12-15 h) at 4°C with rocking before application in IHC.
Brightfield in situ hybridization (BRISH). FFPE sections were cut and dried as described above and used for automated BRISH. Dry slides were loaded on a Bond RX instrument (Leica Biosystems), and deparaffinization, rehydration, and pre-treatment using EDTA buffer solution (ER2, pH 9; Leica Biosystems) was performed at 100°C for 15 min for FFPE cell lines and tumor tissues. RNAscope LS reagent kit (Advanced Cell Diagnostics, Hayward, CA, USA) was then used for detection of mRNA. Briefly, protease was applied for 20 min at 40°C, then peroxide was applied for 10 min. FLT4 (VEGFR3) RNAscope LS probe 2.0 (Advanced Cell Diagnostics) was hybridized at 40°C for 2 h. The signal amplification steps were performed per manufacturer's instructions for amplifications 1-6, with the addition of a 2× saline sodium citrate stringency wash performed after the first amplification. DAB and hematoxylin were applied as part of the modified Bond Polymer Refine Detection kit (Leica Biosystems). Slides were dehydrated and coverslipped as described above.
Interpretation of IHC. Stained TMA slides were evaluated by a Board-certified pathologist (AN). For each evaluable TMA core, based on unequivocal results, vascular VEGFR1 immunoreactivity was interpreted as present (based on the presence of any number of VEGFR1-positive vessels in tumor stroma) or absent. Tumor cell VEGFR1 expression was interpreted as negative, low, medium, or high based on the intensity and proportion of stained tumor cells. For VEGFR2 positivity in tumor cells, a core was considered positive if at least 50% of the cells exhibited weak, moderate or strong immunoreactivity. For both VEGFR2 and VEGFR3, the numbers of tumor stromal vessels showing unequivocal positive immunoreactivity were counted in each TMA core. Medians and 25th and 75th percentiles for the range of non-zero values for IHC-positive tumor stromal vessel counts were determined for VEGFR2 and separately for VEGFR3. Based on the 25th percentile as the cutoff between low and medium values and the 75th percentile as the cutoff between medium and high values, vascular VEGFR2 counts were categorized as negative/low (≤2 VEGFR2-positive vessels/TMA core), medium (3-10 VEGFR2-positive vessels/TMA core), and high (>10 VEGFR2-positive vessels/TMA core). Similarly, VEGFR3 vessel counts were categorized as negative/low (≤1 positive vessels/TMA core), medium (2-9 positive vessels/TMA core), and high (>9 positive vessels/TMA core). Since tumor cell expression of VEGFR1, when present, was widespread in a given NSCLC tissue while vascular expression of VEGFR1 was variable and frequently focal, we utilized tumor cell expression of VEGFR1 along with vascular expression of VEGFR2 and VEGFR3 in determining the final VEGFR profiles of NSCLC cases in this study.
For determination of subclasses, scores were dichotomized into either positive (medium or high categories) or low/negative. Predominance was defined as positive for one or two VEGFRs and low/negative for the remainder.
The 3-Venn diagram was generated using eulerAPE software (32). Statistical methods. Associations were determined among the IHC expression levels (negative/low, medium, and high) of the three VEGFRs. The histological subtypes of NSCLC cases (n=88) included in these association analyses were ADC, squamous cell carcinoma (SCC), ADC with lepidic pattern, and large cell carcinoma. VEGFR2 and VEGFR3 were dichotomized as positive (count>0) and negative (count=0). Pearson's chi-squared test was performed on categorical values for VEGFR2 vs. VEGFR3. For VEGFR1 vs. VEGFR2 and VEGFR1 vs. VEGFR3, Fisher's exact test was performed. Since in examining the associations among VEGFRs we conducted these three hypothesis tests, a statistically significant p-value was corrected for multiplicity to be p<0.0167 (i.e. 0.05/3). Additionally a correlation test (Pearson product-moment) was performed on original, continuous expression scores for VEGFR2 and VEGFR3.
To find any association between expression of VEGFR1, VEGFR2, or VEGFR3 and histology, permutation testing based on 10000 random permutation samples was performed. Preliminary observations on association with histological subtypes found no association for VEGFR1 (permutation Fisher's exact p=0.48), marginal association for VEGFR2 (permutation Wilcoxon p=0.09), and strong association for VEGFR3 (permutation Wilcoxon p=0.01). Based on these findings, an objective form of clustering analysis (K-means) was carried out, using relative vascular expression levels of VEGFR2 and VEGFR3 as a continuous variable. This analysis was based on log-transformed values for VEGFR2- and VEGFR3-positive vascular counts. To determine the number of clusters and control overfitting to noise, Bayesian Information Criteria (BIC) was used to compare multiple models. Our model selection, which was based on best BIC values, resulted in four clusters.
The VEGFR1 expression levels (negative, low, medium and high) were recoded as 1, 2, 3, and 4, respectively, then log-transformed. Based on the availability of analyzable data for all three VEGFRs in the TMA, VEGFR expression values (n=88) were standardized to zero-mean and unit standard deviation and plotted as a heatmap. NSCLC cases with mixed, not otherwise specified (NOS), and adenosquamous histology were excluded from clustering analysis due to the small numbers of cases.
Results
Performance verification of VEGFR3 IHC. Following a comprehensive assay development paradigm, we demonstrated the specificity, selectivity, and appropriate immunoreactivity pattern of VEGFR3 IHC using a similar approach as reported for VEGFR1 and VEGFR2 IHC (24, 29). The mouse monoclonal IgG was first profiled by immunoblot (Figure 1A). In HEL cell lysates, bands were observed at approximately 195 kDa and 130 kDa, consistent with expected molecular weights of uncleaved and cleaved receptor, respectively (33). No bands were observed in lysates from Jurkat, U87MG, or A549 cell lines. Consistent with the results of the western blot, VEGFR3 immunoreactivity was absent in FFPE U87MG cells and positive in HEL cells, showing cytoplasmic and membranous localization (Figure 1B). Dilution of the primary antibody resulted in a decrease of immunoreactivity, and conversely an increase resulted in greater immunoreactivity and possible saturation of the signal. This suggests the selected antibody concentration used was within the linear range of detection. In FFPE human CRC tissue and in squamous cell carcinoma of the lung, VEGFR3 was restricted to the vasculature, consistent with localization of FLT4 (VEGFR3) mRNA (Figure 1C). Recombinant proteins representing the extracellular domains of VEGFR1 and VEGFR2 failed to abolish immunoreactivity of the VEGFR3 antibody in angiosarcoma tissues, suggesting that it did not cross-react with the other VEGFR family members (Figure 1D). Finally, in CRC, VEGFR3 immunoreactivity was observed in D2-40-positive vessels as well as CD34-positive/D2-40-negative vessels, consistent with the observation that VEGFR3 is expressed in both blood and lymphatic vessels (Figure 1E). After confirming the performance of the VEGFR3 IHC assay, we expanded our analyses into a larger cohort of NSCLC tissues, along with IHC for VEGFR1 and VEGFR2.
VEGFR expression in NSCLC. The results of IHC expression scoring and representation of VEGFR expression heterogeneity are shown in Figure 2. Immunoreactivity for VEGFR1 was localized mainly to tumor cell cytoplasm, tumor stromal vasculature, and, where present, macrophages. Forty-eight out of 96 cases (50.0%) were positive for tumor cell expression of VEGFR1, while 87 cases (90.6%) cases were positive for vascular expression. Overall, when present, tumor cell expression of VEGFR1 was distinct, widespread and diffuse with a given case (core) of NSCLC, while vascular expression was more variable, focal and at times weak in intensity. Out of the 10 cases that were negative for vascular VEGFR1 expression, six were positive for VEGFR1 expression in the tumor cells. Because of the unique expression patterns of VEGFR1 compared to VEGFR2 and VEGFR3, we decided to represent cases as VEGFR1-positive based on an assessment of cytoplasmic expression in tumor cells. Of the cases that were VEGFR1-positive in tumor cells, 33 out of 48 (68.8%) showed a moderate level of expression and 15 (31.3%) a high level. There was no apparent association between VEGFR1 and expression of the other receptors. Out of cases that were negative/low for VEGFR1 (n=48), 17 (35.4%) and 11 (22.9%) expressed VEGFR2 and VEGFR3, respectively.
Immunoreactivity for VEGFR2 was localized mainly in tumor stromal vasculature, with variable expression in tumor cells (17.7% of cases positive). Forty-one cases (42.7%) were positive for vascular VEGFR2 expression.
Twenty-four NSCLC cases (25.0%) were positive for vascular VEGFR3 expression. A representation of the distribution of VEGFRs among all cases is shown in Figure 3. As observed in CRC (Figure 1), immunoreactivity for VEGFR3 in NSCLC was localized exclusively to tumor stromal vasculature, including both lymphatic vessels and blood vessels.
Classification of NSCLC cases based on VEGFR profiling. In line with a previous approach (25), the NSCLC cases were grouped into eight different subclasses based on relative (negative/low, medium, high) expression of the three VEGF receptors (Table II). Among these, 28 (29.2%) cases were triple-VEGFR-negative, 22 cases (22.9%) were VEGFR1-predominant, 11 cases (11.5%) were triple-VEGFR-positive, and nine (9.4%) were VEGFR2-predominant.
VEGF receptor associations and clustering. Results of clustering analysis of the level of expression of VEGFR1, VEGFR2 and VEGFR3 in various histological subtypes of NSCLC cases are shown in Figure 4A. Based on differential expression of VEGFR2 and VEGFR3 in this cohort, (n=88) four distinct clusters were identified as follows: Cluster 1 (n=27, 30.7%) included cases with high vascular VEGFR2 and low or negative vascular VEGFR3 levels. Cluster 2 (n=16, 18.2%) cases were high in levels of both vascular VEGFR2 and VEGFR3. This cluster was dominated by squamous lung carcinoma histology. The largest cluster in this analysis was cluster 3 (n=40, 45.4%), which included cases that were low or negative both for VEGFR2 and VEGFR3 and dominated by ADC histology (Figure 4B). Although small, cluster 4 (n=5, 5.7%) included cases that showed high expression of VEGFR3 but were low or negative for VEGFR2.
Associations between VEGFR IHC scores were determined. Categorical scores for VEGFR2 and VEGFR3 showed a positive association; a p-value of 0.0176 resulted from the chi-squared test for categorical data (Figure 4C), and p<0.0001 resulted from the correlation test for original continuous values. No association was observed between VEGFR1 and VEGFR2 (correlation test p=0.5590) or VEGFR1 and VEGFR3 (correlation test p=0.8538). Interestingly, most cases with moderate to high vascular VEGFR3 positivity also had similarly high expression of VEGFR2 (Figure 4D). Cases representing these four clusters are illustrated in Figure 5, which also includes relative expression of VEGFR1 in tumor cells in the NSCLC cases grouped in each of the four clusters.
Discussion
Recent advances in the treatment of NSCLC have been attributed to incremental progress in the understanding of it as a heterogeneous disease and include efforts to classify the disease into subtypes (26). Several anti-angiogenic agents are currently approved for the treatment of advanced NSCLC, including therapeutic antibodies and tyrosine-kinase inhibitors (TKIs). Both have led to increased progression-free survival in clinical trials of NSCLC, however, only the use of therapeutic monoclonal antibodies resulted in better overall survival (34), with the exception of the TKI nintedanib in combination with docetaxel (35). The large phase III trials that resulted in the approval of bevacizumab, a monoclonal antibody against VEGF (36, 37), and ramucirumab, a monocloncal antibody targeting VEGFR2 (38), were based on unselected patient populations. As characterization of the different patient populations that exist in NSCLC may depend on biomarker approaches that better represent complex angiogenesis signaling processes (12), we designed a biomarker approach that is amenable to clinical translation, as well as being informative on the design of pre-clinical studies: we characterized the relative expression of the prominent mediators of angiogenesis, the VEGFRs, to better understand the range and heterogeneity of expression in various histological subtypes of human NSCLC. We first characterized the relative VEGFR expression patterns and distribution in various human tissues (24, 25), and then applied these assessments to an archival cohort of primary NSCLC.
Our findings show immunoreactivity of VEGFR2 was localized to both the tumor vasculature and tumor cells, in line with findings in some recent studies (13, 24). In some of the recent publications using the same VEGFR2 antibody used in the current study (clone 55B11), VEGFR2 staining was reported to be localized only to tumor vasculature (19). Such a discrepancy may have a number of potential explanations as we and others have discussed previously (19, 24). Of note, 42.7% of NSCLC cases in our analysis were moderate to high for vascular expression of VEGFR2. This finding merits further investigation in larger, well-characterized series of NSCLC tissues.
VEGFR1 showed more widespread distribution in tumor cells, vessels, and inflammatory cells in the NSCLC tissues analyzed. Although we noted VEGFR1 expression in tumor vasculature, the overall intensity of VEGFR1 staining in tumor vasculature was not as high as in tumor cells. Not infrequently, the vascular expression of VEGFR1 was weak or focal, while tumor cell expression was generally diffuse. In addition, vascular expression of VEGFR1 did not correlate with the vascular expression of VEGFR2 or VEGFR3. In an IHC-based analysis, using polyclonal anti-VEGFR1 and monoclonal anti-VEGFR2, Decaussin et al. showed strong expression of both VEGFR1 and VEGFR2 in tumor cells, endothelial cells, and stromal fibroblasts, sometimes with membranous accentuation (18). These observations are in contrast to our findings and also to a previous report showing VEGF-induced angiogenesis requires an inverse regulation between VEGFR1 and VEGFR2 in tumor-associated endothelial cells (39). Interestingly, the anti-angiogenic effect of the anti-VEGF bevacizumab was shown to cause inverse reciprocal regulation of VEGFR1 and VEGFR2 expression levels in vascular endothelial cells in SCC (39), suggesting the relevance of systematic VEGFR profiling in this tumor type in the context of pathway inhibition.
In the current study, VEGFR3 was restricted to tumor vasculature, while Li et al. showed that VEGFR3 expression was localized to tumor cell cytoplasm and membrane with a frequency of 52.3% (21). Such discrepancy in prevalence of VEGFR3 may be due to different VEGFR3 assay conditions and their ability to detect distinct subcellular localization patterns of VEGFR3. In a recent IHC-based study on a cohort of 62 patients with NSCLC, 74.1% of cases were found to be positive for vascular expression of VEGFR2 and 64.5% for VEGFR3, respectively (20), much higher than in our study. Also previously reported, the percentages of patients with positive VEGFR expression in cancer cells were 100% for VEGFR2 and 77.4% for VEGFR3 (20). Such variation in prevalence and distribution of VEGFRs in different studies could be explained by differences in antibody selection and assay specificity, especially as the authors found all patients exhibited tumor cell staining of VEGFR2 and a majority of patients expressing VEGFR3 in the tumor epithelium. Such profound differences in the performance of various VEGFR assays highlight the need for technical verification of the results by standardized approaches including selection and demonstration of truly specific antibodies, use of high-quality tissues, technically stringent assay optimization protocols, and careful interpretation of IHC staining results, as have been put forward over the years for standardization of IHC assays (40, 41).
The association between VEGFR2 and VEGFR3 expression in NSCLC is a finding that suggests communication between these receptors. Previous reports show VEGFR2 and VEGFR3 can heterodimerize and function to positively regulate angiogenic sprouting (42), and that they also must cooperate to promote lymphatic migration and proliferation (43). Additionally, in the absence of VEGFA, binding of VEGFC and VEGFD to VEGFR2 and VEGFR3 may be sufficient to induce angiogenesis and tumor progression (44). It will be interesting to see if our VEGFR subclass VII (mixed, VEGFR2/3 predominant) is more responsive to inhibition of VEGFR2 or VEGFR3 than to inhibition of VEGFA ligand. This could be tested in relevant pre-clinical models of NSCLC.
In addition to manual selection of NSCLC cases in the form of various receptor-high or receptor-low subsets of cases, we also wanted to use an automated and objective method for clustering analysis of our VEGFR profiling. The clustering approach for various histological subtypes of NSCLC revealed four subsets of cases, of which the most important clusters were: cluster 2 (high in vascular VEGFR2 and VEGFR3), consisting predominantly of SCCs, and cluster 3 (low or negative for VEGFR2 and VEGFR3) comprising mainly ADC histology. These data also reveal interesting trends between relative expression of VEGFR2 and VEGFR3 in the tumor vasculature and histology of NSCLC. These findings are in line with analysis from Saintigny et al on a set of 92 NSCLC tissues in which 72% of VEGFR3-positive cases in the cohort were of squamous histology (22). Our observations regarding differential expression of VEGFR3 in various lung histologies are in line with those of Saintigny et al and could reasonably be corroborated with data from pertinent preclinical models, ideally those using various anti-angiogenic agents.
We recognize that limitations of our study include a relatively small NSCLC cohort. In addition, we did not include profiling of the expression of the various VEGF ligands, as soluble proteins produce a more diffuse staining pattern that is more challenging to score. We are investigating use of in situ-hybridization methods to more reliably report VEGF ligand expression. Moreover, it is clear that additional analytes will need to be added to our profiling panel to represent the contributions to angiogenic signaling by the platelet-derived growth factor receptors and fibroblast growth factor receptors (45). Finally, in our experience, IHC expression of the three VEGFRs was not uniform (monotonous) within a given NSCLC case, suggesting the frequent existence of intra-tumoral expression of the three VEGFRs in the NSCLC tissues analyzed. It will be interesting to see how the observed intra-tumoral heterogeneity of expression of the three VEGFRs will translate in future analyses of NSCLC, especially using conventional tissue sections.
While the characterization of NSCLC into subtypes has typically been based on genotype and histology (26), we propose that profiling of VEGFR protein expression by IHC may provide valuable understanding of disease heterogeneity in human NSCLC. We have already begun to test other tumor types in the same manner and observed some tumor-specific differences in relative expression of the three VEGFRs. For example, the NSCLC triple-VEGFR-negative subclass comprises one-third of the cases, while in CRC this subclass comprised only 3.6% of the cases (25). The unique VEGFR profiles in NSCLC compared with CRC suggest differences in VEGFR biology in these two cancer types, and raises the possibility that other human cancer types such as breast, gastric, and pancreatic cancer, may also exhibit unique VEGFR profiles.
We conclude that the heterogeneity of VEGFR profiles in NSCLC tissues represents some of the intricacies of angiogenic signaling and that the proposed subclasses of NSCLC are an approach to complement lines of investigation beginning with systematic characterization of the disease. These observations allow for the formation of data-driven hypotheses to test the status of pathological angiogenesis in human NSCLC as well as relevant pre-clinical tumor models. For instance, one can hypothesize that tumor model tissues that exhibit high levels of all three receptors (triple-VEGFR-positive) or VEGFR2 and VEGFR3 (VEGFR2/VEGFR3-predominant) may respond differently to anti-angiogenic treatment as compared to those which lack all three VEGFRs (triple-VEGFR-negative). We are currently investigating the link between IHC-based VEGFR subclasses and canonical pathways involved in proliferation, survival, angiogenesis, and metastasis using gene expression profiling.
Acknowledgements
The Authors gratefully acknowledge expert pathology support from Kelly Credille, DVM, Ph.D., Dip. ACVP, as well as the Experimental Pathology Laboratory scientists and histotechnologists at Eli Lilly and Company. Appreciation goes to Ashley Bay and Amanda Estelle (Lilly) for work on in situ hybridization. Thanks to Karen Yamamoto, Ph.D. (Clarient) for support on IHC, and to Yuewei Qian, Ph.D. and Xiaohua Huanghe (Lilly) for generating recombinant protein. They also recognize Milan Bhagat (TriStar) for critical support of TMA sample and patient data.
Footnotes
This article is freely accessible online.
- Received April 13, 2016.
- Revision received May 26, 2016.
- Accepted May 31, 2016.
- Copyright© 2016 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved