Expression of Angiogenesis Stimulators and Inhibitors in Human Thyroid Tumors and Correlation with Clinical Pathological Features

Tissue Specimens

We collected 68 consecutive thyroid tumor samples representing the complete spectrum of thyroid neoplasias: 17 adenomas, five papillary microcarcinomas, 26 PTCs, 12 FTCs, six MTCs, and seven UTCs. We also gathered 16 nodal metastatic tissues derived from 12 different PTCs and four distinct MTCs. Moreover, we collected eight normal thyroid tissues resected from patients diagnosed with laryngeal carcinoma. All of the specimens analyzed were reviewed by the same pathologist (S. P.) to confirm the original diagnosis and to classify the tumors according to pathological stage (pTNM) as proposed by the Union Internationale Contre le Cancer. 29

T-tumor sizes: T1, Ø = 1 cm; T2, Ø >1 = 4 cm; T3, Ø > 4 cm; T4, any size extending beyond the thyroid capsule.

N-regional lymph nodes: Nx, regional lymph node cannot be assessed; N0, no regional lymph node metastasis; N1, presence of regional lymph node metastases.

M-distant metastasis: Mx, distant metastasis cannot be assessed; M0, no distant metastasis;

M1, presence of distant metastases.

RNA Isolation and Northern Blotting

Each frozen biopsy was mechanically disrupted in a Mikro-Dismembrator II (B. Braun) containing liquid nitrogen. Total RNA was extracted using Ultraspec II (Biotecx Laboratories) according to the manufacturer’s protocol. After extraction, the poly(A)⁺ mRNA was selected by two passages on oligo (dT)-cellulose columns (USB). Northern blot was carried out using 5 μg of poly(A)⁺ RNA as previously described. 30 The following cDNA fragments were used as probes: 1) the 0.395-kb fragment of bovine cDNA VEGF (kindly provided by Dr. S. J. Mandriota); 2) the 0.409-kb VEGF-C cDNA fragment (kindly provided by Dr. K. Alitalo); 3) the 0.353-kb PCR-amplified Ang2 cDNA fragment; 4) β-actin cDNA. The ³²P signals were quantified with a phosphorimager (Bio-Rad).

Semiquantitative RT-PCR Analysis

RT-PCR was performed as follows: 5 μg of total RNA was reverse transcribed at 42°C for 50 minutes in the presence of 500 ng random examers and Superscript II reverse transcriptase (Gibco-BRL) in a final volume of 20 μl. Two microliters of the cDNA reaction was then subjected to different 28-35 PCR cycles (30 seconds at 95°C, 30 seconds at 62°C, and 1 minute at 72°C), using the AmpliTaq Kit (Perkin Elmer) and 0.4 μmol/L of specific primers. When genomic DNA was used as a template in PCR reactions, fragments longer than the expected bands were observed, thus indicating that intron-spanning primers were selected for amplification. The housekeeping aldolase A mRNA was used as an internal standard and coamplified in the same reaction tube with the other specific markers. Aldolase oligonucleotides (5′-CGCAGAAGGGGTCCTGGTGA-3′ 5′-CAGCTCCTTCTTCTGCTGCGGGGTC-3′) were added to each PCR sample between cycles 3 and 7, depending on the amplified gene. PCR products were electrophoresed onto a 3% agarose gel containing ethidium bromide (0.5 μg/ml) and visualized under UV light. The relative intensities of the bands were quantitated by densitometric analysis and normalized to the coamplified aldolase cDNA fragment. Each PCR product was cloned in pCRII (Invitrogen) and sequenced to confirm the identity of the amplified cDNA fragment. Preliminary experiments were performed to determine the exponential phase of amplification and the exact cycle at which Aldolase primers were to be added to the reaction, using RNA extracted from human normal thyroid tissue as positive control. At least four RT-PCR experiments were performed starting from the same RNA preparation. Gel-digitized images were quantitated with the NIH Image 1.6.1 program (National Institutes of Health, Division of Computer Research and Technology, Bethesda).

Primers Used for RT-PCR

VEGF (5′-TTCCAGGAGTACCCTGATGAG-3′ 3′-CTGTTCGGCTCCGCCACTCGG-5′); KDR (5′-TGTGGTGATTCCATGTCTCGG-3′ 3′-TGCCACCCCCTCGCACAGTC-5′); Flt-1 (5′-TCTTGACCCACATTGGCCAC-3′ 3′-CTATCGCAGTGGTCGTCGCT-5′); VEGF-C (5′-AATGTGGGGCCAACCGAGAA-3′ 3′-GCAACACAGGGAAGTATAACC-5′); Flt-4 (5′-GATCGTGGAGTTCTGCAAGTACG-3′ 3′-TTCGCTGCACCACTTCTAGACA-5′); Ang1 (5′-TCCGGTGAATATTGGCTGGG-3′ 3′-CGGGTCAATGAGGAATGCAAG-5′); Ang2 (5′-GATGGCAGCGTTGATTTTCA-3′ 3′-CGACCACCAAACTACGTACA-5′); Tek (5′-GGCCGCTACCTACTAATGAAG- 3′ 3′-GAAGTAATCATCCATCGCCGG-5′); Tie-1 (5′-ACCCACTACCAGTGGATGT-3′ 3′-GAGTTGCGGTCGTGCGCAG-5′); TSP-1 (5′-CGTCCTGTTCCTGATGCAGT-3′ 3′-AACACGTCCTTCTGTCCCGG-5′); CD31 (5′-GGACCAAGCAGAAGGCTAGC-3′ 3′-GAAGACTTGAGGTTGTTGCTC-5′).

Immunohistochemistry and Vessel Counting

Tissue samples were obtained from the archives of our institute. Immunohistochemistry was performed on a limited number of samples (four normal tissues, 10 adenomas, three papillary microcarcinomas, 12 PTCs, five FTCs, four UTCs, and four MTCs). Tissue preparation and antigen retrieval procedures were performed as previously described. 31 Immunostaining was carried out by streptavidin peroxidase (Dako) and 3-amino-9-ethyl carbazole (AEC) as previously reported. 32 For negative controls, the primary antibody was replaced with normal mouse, goat, or rabbit sera. TSP-1 (P10; Immunotech) immunostaining was carried out on acetone-fixed frozen sections. VEGF (BioGenex), CD31 (Dako), VEGF-C (SC1881; Santa Cruz), and Ang2 (C-19 SC7015; Santa Cruz) immunostainings were performed on formalin-fixed paraffin-embedded sections.

Immunohistochemistry for CD31 was assessed as described above. Anti-CD31 mAb highlights vascular endothelial cells of blood vessels of formalin-fixed tumor samples, so it is possible to identify easily the most intense areas of neovascularization in the samples. We considered as a single microvessel any brown, immunostained endothelial cells separated from the adjacent microvessels, tumor cells, and other connective tissue elements. Nonspecific staining was evaluated by use of a control section for each case, replacing the primary antibody with normal serum. Microvessels were carefully counted in the most intense areas of neovascularization on a ×200 field.

Statistical Analyses

Main pathological tumor characteristics in the available patient sample were described by means of contingency tables. The possible association between pathological tumor characteristics was tested by Fisher’s exact test for cross-classified data.

For each molecule considered (VEGF, VEGF-C, Ang1, Ang2, KDR, Flt-4, Tek, TSP-1) the distribution according to the type of tumor (benign or malignant), pTNM parameters, and histology of tissue samples was mainly described in terms of a median value.

To investigate the possible association between different angiogenic factors, Spearman correlation coefficients were computed. We compared with nonparametric methods, such as the Wilcoxon signed rank test (for paired data), the Wilcoxon rank sum test (two independent groups), or the Kruskal-Wallis test (more than two groups), as appropriate, the mean levels of expression of these factors in different types of tumor tissue. Considering only the cases with follicular adenoma or carcinoma, logistic discriminant analyses were carried out to investigate whether the singly taken angiogenic factor allows reliable discrimination between benign and malignant lesions. The results obtained are reported in terms of Somer’s D coefficient and the likelihood ratio P for the regression coefficient of the fitted logistic models. The conventional 5% significance level was adopted. The analyses were carried out with SAS software.

Increases of VEGF and KDR in Malignant Thyroid Carcinomas

To identify VEGF isoforms 5 expressed in thyroid tumors, we chose specific primers able to detect the various mRNA transcripts generated by alternative splicing. In most benign adenomas and in papillary microcarcinomas we did not observe any increase in VEGF expression with respect to normal thyroid (Figure 1A ▶ , Figure 5A ▶ , and Table 4A ▶ ). In contrast, all other malignant carcinomas showed a marked VEGF up-regulation. In particular, PTCs and most of aggressive FTCs presented a stronger increase, ranging from three- to 10-fold (an example of the results is shown in Figure 1A ▶ ). Finally, a two- to ninefold increase in VEGF expression was detected in 70% of UTCs. A similar VEGF increase was also observed in 50% of MTCs (Figure 1A ▶ and Table 4A ▶ ). Provocatively, most of the lymph nodal metastases analyzed presented a further VEGF increase with respect to primary tumors (data not shown). Interestingly, the up-regulated VEGF isoforms (VEGF₁₂₁, VEGF₁₄₅, and VEGF₁₆₅) were essentially the same, thus suggesting that thyroid tumors preferentially express efficiently secreted isoforms of VEGF. 33 Northern blot analysis of different thyroid tumors showed up-regulated VEGF mRNA levels ranging from threefold (FTC 250) to ninefold (UTC 325), whereas in the adenoma we failed to find any increase (Figure 5B) ▶ . Immunohistochemical analysis completely paralleled the data obtained with semiquantitative RT-PCR. In Figure 1B ▶ is shown VEGF staining in both normal and PTC samples. In PTC specimens epithelial follicular cells were strongly stained, whereas normal tissues showed a weak staining in addition to the expected reactivity in vascular endothelial cells (feature not shown).

With the exception of papillary microcarcinomas, increased KDR mRNA levels could be detected throughout the considered lesions (Table 4A) ▶ . In particular, epithelial-derived thyroid tumors displayed consistent KDR increases ranging from 24-fold in PTCs to three- to sixfold in benign adenomas, malignant FTCs, and UTCs (Figure 1A) ▶ . These data were also confirmed by Western blot of a representative panel of tumors (data not shown). The C-cell-derived MTCs showed a slight KDR up-regulation. Because KDR expression levels are different in thyroid tumors, differences in KDR expression levels in normal tissues are due to differences in film exposure time. For densitometric analysis we needed a nonsaturated exposition of the photograph. Furthermore, we did not observe any modulation of Flt-1 mRNA expression among the analyzed specimens (Figure 1A) ▶ .

Strong Increases of Angiopoietin-2 Levels in Thyroid Neoplasias

Compared to normal thyroid tissues we observed a marked increase in Ang2 mRNA levels in the vast majority of thyroid tumor specimens analyzed (Table 4A) ▶ . Interestingly, whereas papillary microcarcinomas showed no Ang2 up-regulation (Figure 5A) ▶ , almost all of the other PTCs and their nodal metastases displayed a strong (over 10-fold) Ang2 increase (Figure 2) ▶ . Indeed, with few exceptions, all thyroid adenomas, FTCs, UTCs, and MTCs presented Ang2 mRNA levels that rose eight- to 20-fold. These results obtained by semiquantitative RT-PCR were consistent with Northern blot analysis (Figure 5B) ▶ .

We also found a moderate up-regulation of angiopoietin tyrosine kinase receptor Tek in a few PTCs and in most primary and lymph node metastases of MTCs (Figure 2 ▶ and Table 4A ▶ ).

In our thyroid tumors, we also analyzed the expression of both Ang1 and the orphan receptor Tie-1. Ang1 mRNA expression was slightly down-regulated or up-regulated in a few carcinomas (Figure 2) ▶ . No significant differences in the expression of the Tie-1 receptor could be detected (data not shown).

Up-Regulation of VEGF-C and Flt-4 in Lymph Node Invasive Thyroid Tumors

Compared to normal thyroid tissues, we observed a generalized up-regulation of VEGF-C mRNA levels in thyroid neoplasias that frequently present lymph node involvement (Table 4A) ▶ . In fact, only one adenoma and a few FTCs displayed a significant increase in VEGF-C mRNA expression (Figure 3A ▶ and Table 4A ▶ ). Conversely, a two- to fourfold VEGF-C up-regulation was detected in almost 80% of PTCs, which frequently invade cervical nodes, and in five of seven highly invasive UTCs (Figure 4A) ▶ . The papillary microcarcinoma variant that did not show any variation of expression of the angiogenic molecules analyzed presented up-regulation of VEGF-C (Figure 5A) ▶ . Almost all of the tumors arising from neural crest-derived C cells (MTCs) showed a marked increase in VEGF-C expression, up to 20-fold (Figure 3A) ▶ . Northern and Western blot analyses of a representative panel of tumors confirmed these results (Figure 5B ▶ and data not shown). Interestingly, VEGF-C up-regulation was evident specifically in pN1 cases. The augmented expression of VEGF-C in thyroid tumors was also confirmed by immunohistochemical analysis. As shown in Figure 3B ▶ , with respect to normal thyroid tissue, we observed a significant increase in VEGF-C reactivity in epithelial follicular cells of PTCs. A positive reaction was also detected in the endothelial vessel lining cells of both normal and neoplastic tissues (data not shown). This result confirmed that the increment in mRNA expression levels observed coincided with protein increase.

Compared to normal thyroid tissues, few PTCs and 50% of their lymph node metastases showed a moderate increase in the VEGF-C tyrosine kinase receptor Flt-4 (Figure 3A ▶ and Table 4A ▶ ). This receptor was also slightly up-regulated in few benign adenomas, FTCs, and UTCs. As for VEGF-C, Flt-4 overexpression was strongly evident in the vast majority of MTCs.

Down-Regulation of Thrombospondin-1 Expression in Aggressive Thyroid Carcinomas

In our cases, TSP-1 expression analysis showed no reduction in mRNA levels in benign adenomas and papillary microcarcinomas (Table 4B ▶ , Figures 4 and 5A ▶ ▶ ). Conversely, 20% of the other PTCs presented a twofold TSP-1 down-regulation that become much more significant in all PTC nodal metastases analyzed (Table 4B ▶ and Figure 4A ▶ ). In contrast, 75% of FTCs presented a remarkable TSP-1 mRNA down-regulation. As for UTCs, in seven of seven neoplasias, TSP-1 expression could not be detected by semiquantitative RT-PCR analysis (Figure 4A) ▶ . Because five of seven undifferentiated tumors presented a mutated form of TP53, 34 these data are in keeping with the previous observation that TP53 regulates TSP-1 transcription. 35 Sixty-seven percent of MTCs showed a TSP-1 reduction, whereas all of their metastatic tissues presented a further TSP-1 down-regulation (Table 4B) ▶ . The same results were also evident by immunohistochemical staining with anti-TSP-1 antibody. In fact, normal thyroid tissues showed a strong immunoperoxidase staining for TSP-1 confined to the interfollicular stroma (Figure 4B) ▶ and, to some extent, to the vascular endothelium (feature not shown). A similar degree of reactivity was found in benign follicular adenomas, although the tissue architecture was altered. In PTCs we observed a slight decrement of TSP-1 staining with respect to normal tissues. In contrast, the malignant follicular carcinomas showed a barely detectable reaction to the anti-TSP-1 antibody (Figure 4B) ▶ . These data were also confirmed by Western blot analysis (data not shown).

Statistical Analysis

We performed statistical analyses to verify the possible associations between the expression levels of angiogenic molecules and clinical pathological features of thyroid tumors, like tumor size (T), presence of lymph node metastases (N1), and presence of distant metastases (M1).

The analysis of clinical pathological characteristics (pTNM) and the expression levels of angiogenic factors in thyroid neoplasias compared to normal tissues showed that two factors were strongly associated with the progression of thyroid tumors from stage T1 to stage T4. In particular, Table 5 ▶ shows a significant association between T-stage progression with high levels of VEGF (P < 0.001) and Ang2 (P < 0.001). Also in the case of KDR, statistical tests yielded significant results showing a clear trend according to T-stage progression (P = 0.037). Thus the up-regulation of the ligand-receptor couple (VEGF/KDR) and Ang2 showed statistical significance when associated with tumor size, suggesting that these factors (as hypothesized by other authors) are associated with neovascularization and tumor growth.

Regarding the ability of tumors to invade regional lymph nodes (pN1), we found that VEGF-C overexpression was associated with lymph node invasion, comparing the levels of VEGF-C mRNA in pN0 and pN1 tumors. Statistical analyses showed that VEGF-C mRNA was up-regulated prominently in node-positive tumors (P < 0.001, Table 5 ▶ ).

We also examined the factor(s) associated with the tumor’s ability to spread by hematic vessels to give distant metastases (pM1). The only significant association between pN1 tumors and angiogenic factors was observed for TSP-1, which was significantly decreased only in tumors able to generate distant metastases (P < 0.001, Table 5 ▶ ).

Regarding discrimination between follicular adenoma and carcinoma, the analysis base logistic discriminant yielded significant results for VEGF (P = 0.011) and TSP-1 (P < 0.001). The discrimination obtained based on TSP-1 expression was relatively high (Somer’s D coefficient = 0.75). In particular, by adopting for this factor a 0.8 cutoff value of expression, all malignant lesions showed TSP-1 down-regulation below the threshold against only five of the 17 adenomas. The corresponding figures of sensitivity and specificity (95% exact confidence intervals) were 100% (74–100) and 71% (44–90), respectively.