Tumor Angiogenesis and p53 Mutations: Prognosis in Head and Neck Cancer

Objectives  To assess how p53 gene mutations and microvessel density (MVD) may be used as prognostic markers for the study and management of head and neck squamous cell carcinomas and to investigate putative associations between p53 gene mutations and MVD and the relationship of these factors to tumor response to radiotherapy and/or chemotherapy at 6 weeks.

Patients and Design  Thirty-nine patients with squamous cell carcinoma of the head and neck, stages I to IV, who were examined at Rush-Presbyterian-St Luke's Medical Center, Chicago, Ill, and its affiliated hospitals between 1993 and 1995 were monitored. Mutations in the p53 gene were identified by microdissection of tumor cells on frozen sections, followed by single-strand conformation polymorphism analysis of the products of polymerase chain reaction amplification of exons 5 to 9. The microvessels were immunostained with monoclonal antibodies to factor VIII and/or CD31. Microvessel counts were done by 2 investigators blinded to each other's counts and to the p53 gene status. Intratumoral or peritumoral microvascular "hot spots" were assessed and counts were done with an ocular grid in 3×200 fields of hot spots by each investigator. The mean of the highest values was considered. Statistical analysis was done with the Wilcoxon rank sum test, the log-rank test, and proportional hazard models.

Results  Of the 39 patients, 13 had mutations in exons 5 to 9. Mutations in the p53 gene were associated with unfavorable overall (P=.003) and disease-free (P=.02) survival. A strong inverse relationship was seen between MVD and p53 mutations (P=.01). No statistically significant relationship was seen between mean MVD and overall and disease-free survival. The response to therapy differed significantly (P=.03) by p53 mutations, whereas there was no statistical significance with MVD counts.

Conclusion  In this study a strong inverse relationship was seen between MVD and p53 mutations. p53 Mutations in exons 5 through 9 were associated with unfavorable survival, whereas MVD showed no association with survival.

IN CURRENT clinical practice, the treatment of head and neck squamous cell carcinoma (HNSCC) is based on the histological grade and TNM stage. However, these measures have limited predictive value for therapeutic results. Recognition of additional prognostic markers is clearly needed to subdivide the HNSCC patient population into those who respond well and poorly to different modes or combinations of therapy.

Angiogenesis has been recognized as an important factor in tumor growth for more than 2 decades.^1,2 The relationship of angiogenesis to prognosis in various tumors, including HNSCC, has been studied by different workers.^3-8 Some investigators propose that histological evidence of abundant tumor angiogenesis is associated with aggressive tumor behavior in breast, bladder, and prostate cancers, whereas others propose that a high level of vascularity is an indicator of better response to radiotherapy and chemotherapy.

Mutations of the tumor suppressor gene, p53, are found frequently in various human cancers in vivo.^9-11 Mutations in exons 5 to 9 of the p53 gene have been clearly associated with the development and progression¹¹ of HNSCC. Since the wild-type p53 protein functions as a transcription activator, it is possible that 1 or more angiogenesis factors are regulatory targets. For example, it has been reported that, in vitro, wild-type p53 can up-regulate the promoter of thrombospondin-1, an inhibitor of angiogenesis.¹² We undertook the present study to evaluate tumor angiogenesis, as assessed by microvessel density (MVD), and p53 gene mutations as markers for tumor aggressiveness and response to radiotherapy and/or chemotherapy.

We studied 39 patients with HNSCC, stages I to IV, who were examined at Rush-Presbyterian-St Luke's Medical Center, Chicago, Ill, and its affiliated hospitals between 1993 and 1995. The date of the initial biopsy diagnosis was counted as day 0 and the date of last follow-up or the date of death was taken as the end point. Patient follow-up ranged from 2 to 54 months. The patients underwent biopsy and/or resection, followed in most cases by radiotherapy and/or chemotherapy, depending on site, stage, and other clinical factors, in accordance with institutional protocols (Table 1). An informed consent was obtained before surgical and adjuvant treatment was begun. Fresh specimens were sent to the Department of Pathology, where part of the tumor was frozen in a cryostat and 2 consecutive frozen sections were cut. One section was stained with hematoxylin-eosin for confirmation of invasive tumor, and the other section was microdissected for p53 mutation analysis. The rest of the specimen was processed routinely by formalin fixation and paraffin embedding for definitive diagnosis of squamous cell carcinoma. An experienced pathologist (P.U.H.) reviewed hematoxylin-eosin–stained sections to choose the apparently vascularized area of the tumor. Adjacent sections of these paraffin blocks were then processed for immunostaining for assessment of MVD.

Sections 3 to 5 µm thick were stained on an immunohistochemistry slide staining system (Ventana ES 320, Ventana Medical Systems Inc, Tuscon, Ariz) with factor 8/86 mouse monoclonal antibody (anti–human von Willebrand factor, Dako Corp, Carpinteria, Calif) at 1:200 dilution, by avidin-biotin-conjugate immunoperoxidase chemistry. Six sections that showed weak or no staining with factor 8 antibody were stained with anti-CD31 (Dako, JC/70A) at 1:20 dilution (Figure 1). Two sets of coded slides were prepared. The MVD counts in 3 areas of "hot spots" were performed separately by 2 investigators (P.U.H. and A.C.B.) blinded to each other's counts and to the p53 status, with the use of an ocular grid at ×200 magnification. (The hot spots are the areas of maximum vascular density, either within the infiltrating tumor mass or at the tumor stromal interface.) Seven of 39 cases could not be assessed because of poor staining characteristics with both antibodies. The counts of the 2 investigators matched (within a margin of ±5) in all except 6 cases. These were reanalyzed and the discrepancy was resolved by consultation. The mean of the highest figures by each observer was selected as the MVD for the tumor.

Each frozen tumor section was microdissected to ensure that a region containing mostly tumor cells was obtained. Frozen sections 5 µm thick were briefly stained with toluidine blue–methylene blue. Areas of tumor were identified by a pathologist (P.U.H.), and a region containing mostly tumor cells was microdissected. The cells were transferred to a microcentrifuge tube for lysis with 0.5 µg of proteinase K per microliter at 37°C for 2 hours. After heat inactivation of the proteinase K, portions of the lysate were amplified for 35 cycles by means of primers flanking exons 5 to 9. Portions of this product 1.8 kilobases (kb) long were reamplified with primers for each of exons 5 to 9.¹⁰ The sizes of the amplified products were as follows: exon 5, 245 base pairs (bp); exon 6, 183 bp; exon 7, 189 bp; exon 8, 213 bp; and exon 9, 136 bp. In some cases, lysates were amplified directly with exon-specific primers. Single-strand conformation polymorphism analysis of each exon-specific product was carried out on 6% polyacrylamide gels with 5% glycerol at room temperature. Bands were visualized by silver staining (BioRad Laboratories, Hercules, Calif). Reagent blanks were processed in an identical fashion to ensure the absence of polymerase chain reaction carryover. Novel bands were identified by comparison with the wild-type pattern produced by performing identical amplification on human placental DNA, analyzed on the same gels.

The MVD of patients with mutant p53 was compared with the MVD of patients with wild-type p53 by means of the Wilcoxon rank sum test. Kaplan-Meier survival curves were obtained and the log-rank test was used to compare groups of patients with respect to overall and disease-free survival. Hazard ratios for overall survival were estimated from Cox proportional hazard models. The Wilcoxon rank sum test was used to compare the response to therapy at 6 weeks in different groups. Multivariate analysis was performed to assess the independent significance of p53 status controlled for stage, differentiation, and adjuvant chemotherapy or radiotherapy. Significance of p53 status with respect to the tumor site was not analyzed because of the vast heterogeneity and difficulty in grouping sites.

The 39 patients analyzed are summarized in Table 1. The patient population included 30 men and 9 women, with an age range of 34 to 83 years. Thirteen of the 39 patients had mutations in exons 5 to 9. The MVD in the mutant group clustered between 18 and 26 microvessels per ×200 field, with 2 outliers at 46 and 48 microvessels. The MVD in the wild-type group was scattered over a wide range, from 16 to 66 microvessels per ×200 field (Figure 2). A strong inverse relationship was seen between MVD and p53 mutations (Wilcoxon rank sum test, P=.02). The results of the MVD and p53 mutations are summarized in Table 2. Mutations in the p53 gene were associated with unfavorable overall (P=.003) and disease-free (P=.02) survival (Figure 3). The hazard ratio for p53 mutation was 3.83 and the 95% confidence interval was 1.45 to 10.1. There was no association between MVD and overall (P=.93) and disease-free (P=.51) survival. The hazard ratio for MVD was 0.99 and the 95% confidence interval was 0.95 to 1.05. Multivariate analysis showed no significant difference in p53 status controlled for tumor differentiation (P=.10; hazard ratio, 1.81; 95% confidence interval, 0.90-3.62). No significant difference was also noted for tumor stage and chemotherapy or radiotherapy. Multivariate analysis on this small and heterogeneous patient sample was unstable and is not reported.

The therapeutic response after 6 weeks of therapy in patients with wild-type p53 was compared with that in patients with mutant p53. The response data were not available for 5 patients. Of 25 patients with wild-type p53, 19 patients achieved complete remission, 1 patient had a partial response, 2 patients had persistent disease, and 3 patients had disease progression. Of 9 patients with p53 mutations, 3 responded completely, 2 had a partial response, and 4 had progression. These groups differed significantly (Wilcoxon rank sum test, P=.03). The response to therapy at 6 weeks was also compared with respect to MVD, and no significant relationship was observed (Wilcoxon rank sum test, P=.60). These results are summarized in Table 3.

Angiogenesis is a complex process, controlled by a balance of positive and negative regulatory factors. Among the angiogenic regulatory factors produced by tumor cells is thrombospondin, which has shown strong inhibitory effects in experimental models. Secretion of thrombospondin has been further shown to be up-regulated by the wild-type p53, but not mutant gene product, in cultured fibroblasts.¹² The p53 gene mediates a number of cell activities, including the regulation of the cell cycle, cell apoptosis, and DNA repair, and is often mutant in head and neck cancer. These results suggest that p53 mutations might influence the behavior of cancers by blocking the production of a known angiogenesis inhibitor. We therefore examined the level of vascularity in patients as a function of direct determination of p53 mutation status, since expression of p53 protein does not indicate mutation status accurately in this tumor.¹³ Both squamous cell carcinoma cell lines¹⁴ and tumor explants¹⁵ can induce angiogenesis in experimental systems. Unexpectedly, we found that the presence of p53 mutation was associated with low MVD, which suggests that either thrombospondin-1 is not regulated by p53 in these tumors, or neovascularity is regulated by other factors. It is likely that the influence of the p53 gene on angiogenesis is cell and tissue specific.^14,16 In this study, p53 mutations were identified by polymerase chain reaction–single-strand conformation polymorphism analysis of exons 5 to 9, where most mutations are known to occur. It is possible that additional mutations might have been detected by direct sequencing of the entire gene.

Methods for the determination of MVD vary considerably among investigators, which makes direct comparisons of results difficult. Some investigators evaluate hot spots, or areas of highest vessel concentration,^3,17 while others sample tumor areas at random and derive mean values.^6,18 Approaches to determining microvessel counts in the selected area also vary, including several attempts to simplify counting procedures. Both direct counts and subjective vascularity grading scales have been used.^17,18 The Chalkley point counting system has been used to compare data between laboratories.^19-21

Finally, computer image analysis that uses density measurements as a reflection of the intensity of antibody staining has been used.^20-22 Use of an image analysis system (Cell Analysis System 200, Becton Dickinson, San Jose, Calif) proved unsatisfactory with our tumor samples, since background staining was present on some slides. Furthermore, it is necessary to evaluate significant vessel morphological characteristics and caliber to distinguish neovascularization.⁵ These qualitative distinctions require a trained observer.

The range of values (18-66 microvessels per ×200 field) that we obtained by the method outlined by Weidner,³ using hot spots to quantify MVD, were somewhat lower than the range of data of several groups studying breast cancer (30-130 microvessels per ×200 field).^3,5,17,23 One study of nasopharyngeal carcinoma obtained values in the range of 5 to 18 vessels per high-power field.⁷

The question remains, does an MVD count accurately represent neovascularization in HNSCC tumors? Evidence suggests that progression in several tumor types, including breast, prostate, and colon carcinomas, relies on neovascularization to support new growth.^3-5,21,23,24 Other tumors do not require angiogenesis for such expansion because of the pattern of their growth. For example, gliomatosis grows as sheets of cells between membranes without the requirement of neovascularization.²⁵

Neovascularization can influence several descriptors of tumor behavior, including development of local or regional metastases, response to therapy, and survival. While MVD has been shown to have predictive value for metastasis in breast carcinoma and oral cavity tumors,²² no such relationship was found in squamous cell carcinoma of the tongue.¹⁸

In general, studies of breast carcinoma demonstrated a higher recurrence rate and poorer survival in patients whose tumors had higher MVD levels.^5,17,19 On the other hand, poorer survival was associated with lower levels of vascularization in studies of nasopharyngeal⁷ and head and neck⁸ cancers. In the latter study, partial or no response to radiotherapy occurred in patients with a lower MVD score. A similar finding of poorer response to radiotherapy and shorter survival in patients with low MVD was seen in 2 studies of cervical carcinoma.^6,26 These findings are consistent with the widely held view that a high level of tissue oxygenation promotes a favorable response to radiotherapy. Poor response to therapy was associated with the presence of p53 mutations, which may have influenced apoptosis and proliferation rate as well as MVD.

We did not demonstrate an association between MVD and response to therapy or survival. However, many of the tumors in our study were advanced, and angiogenesis, which might have occurred earlier in the development of the tumor and which might have been essential for its establishment, is possibly not reflected in the biopsy specimen. A prospective study of stage T1 lesions would be helpful in answering this question.

In conclusion, no clear relationship was seen in our study between MVD and patient outcome, whereas p53 mutation was strongly associated with an unfavorable response to therapy and survival. Because of the complexity of measuring angiogenic activity and the complexity of the effect of tumor angiogenesis on tumor response to therapy, further study is warranted. With a better understanding of this process, it may be possible to devise improved therapies with the use of carefully targeted angiogenesis inhibitors, such as retinoic acid.¹⁴

Accepted for publication June 4, 1997.

This work was supported in part by the Stanton Friedberg Fund for the Department of Otolaryngology, Rush-Presbyterian-St Luke's Medical Center.

Presented at the Fourth International Conference on Head and Neck Cancer, July 29, 1996, Toronto, Ontario.

Reprints: Poornima U. Hegde, MD, Department of Pathology, Rush-Presbyterian-St Luke's Medical Center, 1653 W Congress Pkwy, Chicago, IL 60612 (e-mail: phegde@sispro.sis.rpslmc.edu).