The authors previously demonstrated that fibroblast growth factor 2 (FGF2) expression levels in tumor cells (FGF2-T) may be an indicator of the efficacy of radiotherapy in patients with cervical cancer (CC). In the current study, this finding was extended in newly enrolled patients and was investigated further in stromal FGF2 (FGF2-S) expression.
Sixty-nine patients with CC were recruited as a validation set for the immunohistochemical detection of FGF2-T from biopsy samples that were taken before (pretreatment) or 1 week after the initiation of radiotherapy (midtreatment). The authors also investigated the expression of FGF2 in tumor FGF2-S and investigated vascular endothelial growth factor (VEGF), and cluster of differentiation 31 (CD31) (also called platelet endothelial cell adhesion molecule) in these patients and in an additional 35 patients from a previous study.
FGF2 expression was detected in tumor cells from all patients and in stromal cells from 87% of patients. FGF2-T was significantly higher in midtreatment samples (P = .0002), and a high ratio of midtreatment/pretreatment FGF2-T was related significantly to a better prognosis (P = .025). Increased VEGF expression after the initiation of radiotherapy was related significantly to positive FGF2-S in pretreatment samples (P = .035); however, it was not related to prognosis or microvessel density detected by CD31 expression.
Gynecologic cancer is a common cause of death in women, and cervical cancer (CC) is the second most common malignancy among women.1 CC is managed successfully by surgery or radiotherapy (RT) if it is detected at an early stage, however, treatment outcomes for patients with locally advanced CC are suboptimal despite the recent success of first-line therapy with concurrent chemoradiation (CRT). Increasingly, clinical trial designs involving combinations of drugs that target different critical tumor components or compartments are being discussed to enhance efficacy and to reduce the likelihood of an emergence of resistance to treatment.1, 2 It would be useful for radiation oncologists to know the efficacy of standard-of-care modalities during treatment to sequence different treatment modalities as an adjuvant to RT.3-7
We previously identified specific RT-responsive genes using comprehensive transcriptome analysis with microarrays in sequential biopsies from patients with CC who were receiving CRT.8 We demonstrated that CRT itself alters the molecular status of tumors and stroma during a fractionated irradiation schedule. This makes it possible to identify patients who require further treatment in addition to standard therapy using sequential biopsies during RT. We identified dozens of CRT-responsive genes, including well known radiation-responsive apoptosis/cell cycle-related genes (such as cyclin dependent kinase inhibitor 1A [CDKN1A], carbamoyl-phosphate synthetase 2 [CAD], and BCL2-associated X protein [BAX]), several extracellular matrix genes (such as heparanase [HPSE], cadherin 3 [CDH3], and cluster of differentiation 44 [CD44]), and several cytokine genes (such as fibroblast growth factor 2 [FGF2]).8
FGF2 is closely associated with the activation of fibroblasts and can be produced by many cell types. FGF2 plays a part in various steps of neoangiogenesis by promoting angioblast differentiation, cell growth, and invasion. When secreted from tumor cells, FGF2 also is responsible for basement membrane dissolution, migration, and metastasis.9, 10
There is increasing evidence of an angiogenic response of irradiated tumors resulting in decreased radiation sensitivity11-13 that may be regulated in part by vascular endothelial growth factor (VEGF) and FGF2.14 It has been demonstrated that FGF2 has prognostic value in several malignancies, including esophageal cancer, renal cell carcinoma, ovarian cancer, lung cancer, and uterine endometrial cancer.15-20 The down-regulation of FGF2 messenger RNA expression may be associated with increased severity of CC.21
FGF2 can be released from both tumor cells and stromal cells. When the contribution of irradiation-induced VEGF and FGF2 release to the individual level of tumor radioresistance was examined using 6 epithelial tumor cell lines before and after irradiation, irradiation led to a several-fold increase in VEGF expression and a several hundred-fold increase in expression levels of FGF2, but these changes were not correlated with the survival of tumor cells.22 FGF2 also is up-regulated by irradiation of normal fibroblasts.23
Hase et al reported that FGF2 in fibroblasts, which is detected in 67% of oral squamous cell carcinoma (SCC) pretreatment biopsy specimens, was related significantly to a poor prognosis.24 Furthermore, there is an interaction between cancer cells and fibroblasts around the tumor through the mediation of cytokines produced by the cancer cells and/or fibroblasts.25
Recently, several drugs that inhibit neoangiogenesis, such as imatinib, have been developed as possible therapeutic options for CC through a preclinical phase 1 study.26 By using an in vivo model, other research has revealed that the pharmacologic targeting of these drugs involves the suppression of FGF2 expression.27 These accumulated research articles highlight the urgent need to investigate the status of FGF2 in the clinical setting of CC during RT.
In a previous study, we observed that FGF2 protein expression in midtreatment samples was significantly higher than in pretreatment samples during a fractionated schedule of CRT and was related to a better prognosis.28 However FGF2 expression in stromal components was not investigated in that study.
In the current study, we enrolled 69 new patients to confirm our previous finding of FGF2 as a radioresponsive marker and an indicator for the efficacy of RT. We also investigated the localization of FGF2 in stromal fibroblasts to examine whether FGF2 in the stroma was related to radioresistance behavior of tumors using pretreatment and midtreatment biopsy samples from 104 patients, including 35 patients from our previous study.28
MATERIALS AND METHODS
Study Population and Biopsy Samples
The full study included 104 patients with CC who were treated at the National Institute of Radiological Sciences, Chiba, Japan. These patients gave appropriate informed consent to allow examination of their tissues and medical records, and the study protocols were approved by the institutional review board. Thirty-five patients were the included in previous research,28 and 69 patients with CC were newly enrolled as a validation set for the RT-responsive molecular marker FGF2. All biopsy specimens were obtained from cervical tumors before radiation therapy and at 6 hours after the fifth fraction.29 Each sample was fixed in 10% formalin and embedded in paraffin.
Clinical staging of CC was performed according to the International Federation of Gynecology and Obstetrics (FIGO) classification.30 After the completion of therapy, patients were followed monthly for 1 year then bimonthly during the second year. The effect of treatment was evaluated in terms of local control and disease-free survival. Local control was defined as showing no evidence of tumor regrowth or recurrence in the treatment volume according to pelvic examination along with computed tomography scans, magnetic resonance imaging, and biopsy. We classified the RT response as a complete response (CR), a partial response (PR), or progressive disease (PD). A PR was defined as a reduction >50% in tumor size 6 months after the completion of RT and the absence of any new lesions. PD was defined as the appearance of any new lesion during treatment or an increase >25% in the size of the local tumor. Once treatment ended, patients were evaluated for disease status and late toxicity every 3 months for the first 3 years and then every 6 months for an additional 2 years.
Two methods were used to evaluate prognosis: disease-free survival using the Kaplan-Meier method31 and disease-free status (good/poor responders) at 2 years after treatment. Patients who did not develop any evidence of disease were defined as good responders within 2 years. Patients who were alive with either a recurrent tumor or newly developed metastasis and patients who died because of their tumor were defined as poor responders within 2 years.
Expression levels of FGF2, VEGF, and cluster of differentiation 31 (CD31) (platelet endothelial cell adhesion molecule) were analyzed using a streptavidin-biotin immunoperoxidase technique32 with the automated Ventana Discovery System (Ventana Medical Systems, Tucson, Ariz) according to the manufacturer's instructions. VEGF and CD31 were investigated in only 77 patients who received either RT alone or CRT. From formalin-fixed, paraffin-embedded tissue samples, 3-μm thick sections were cut serially and mounted on precoated slides. The tissue section slides were incubated at 42°C with the following different antibodies: an antihuman FGF2 polyclonal antibody (1:100 dilution; Santa Cruz Biotechnology, Santa Cruz, Calif), an antihuman VEGF antibody (1:100 dilution; Santa Cruz Biotechnology), or an antihuman CD31 antibody (1:500 dilution; DakoCytomation, Glostrup, Denmark). Next, the slides were incubated with universal secondary antibody (Ventana Medical Systems) and counterstained with hematoxylin. For positive controls, we used human colon cancer tissue for FGF2. For VEGF and CD31, human cervical cancer tissues that had demonstrated a high level of immunoreactivity in preliminary experiments were used as positive controls. For negative controls, we followed the same procedure but omitted the primary antibody. Sections were photographed using a BX51 microscope (Olympus, Tokyo, Japan).
Immunostained tissue sections were evaluated by 2 observers in a coded manner without knowledge of the clinical or pathologic parameters. The magnitude of expression of FGF2 in tumor cells (FGF2-T) and VEGF was measured by computerized quantitative image analysis as described elsewhere.28 Briefly, 5 fields at ×400 magnification in non-necrotic tumor cell areas were captured randomly as digital images and then analyzed using the WinROOF image analysis system (Mitani, Fukui, Japan) with macroinstructions. Analysis was performed after transformation of color information to hue-saturation-intensity information. After evaluating several fields on the positive control slides, an intensity threshold for the immunostaining (brown) was set. The proportion of stained area relative to the total area of tumor cells was defined as the positive area (%). The ratio was calculated as the percentage positive area in midtreatment tissues divided by the percentage positive area in pretreatment tissues. For statistical analysis, the FGF2 ratio was used to separate patients into a high ratio group (ratio >3) and a low ratio group (ratio <3) as described previously.28
FGF2 in tumor stroma (FGF2-S) was rated semiquantitatively on a 3-grade scale (negative, medium, or strong) according to the intensity of the reaction and the extent of staining, and patients were divided into 2 groups, positive (which included medium and strong ratings) and negative.33 CD31-stained sections were scanned at low magnification (×40) to determine areas with the highest number of microvessels (hot spots). Microvessels were counted at a ×200 magnification in 2 hot spots on each section, and the average number of microvessels was defined as the microvessel density (MVD).34
Animal Study (Mouse Experiment)
C3H/HeMsNrs male mice aged 9 weeks were used in this study (n = 5 mice in each group). The animals were bred and maintained in specific pathogen-free facilities in our institute and were housed in groups. Two murine SCCs were studied, NR-S1 and SCCVII. Single-cell tumor suspensions were prepared by enzymatically digesting tumors and transplanting the cells subcutaneously into the right hind leg. The study protocol was reviewed and approved by the National Institute of Radiological Sciences Institutional Animal Care and Use Committee. An average greatest tumor dimension of 7.5 ± 0.5 mm (mean ± standard deviation) was attained 7 days after the subcutaneous inoculation of tumor cells into the right hind leg.
Radiation using cesium-137 gamma rays at a dose rate of 1.3 grays (Gy) per minute and with a focus-to-skin distance of 21 cm was administered at 30 Gy and 50 Gy as a single dose for the tumor growth delay assay and at 2 Gy daily for 4 days for immunohistochemical analyses.35 Tumor size was measured every day after irradiation. The tumor volume was estimated using the following formula: volume = 4/3 × a × b × c × π.
Tumor growth time, ie, the time required for each tumor volume to become 5 times the initial volume, was calculated from the first irradiation day in each group. The difference between the tumor growth time for the treatment group and for a nonirradiated control group was defined as the tumor growth delay.36 Tumor growth delay induced by radiation treatment was used as a parameter to measure treatment effect.
For immunohistochemical analysis, the animals were killed, and tumors were extracted 3 hours after the last irradiation. Formalin-fixed, paraffin-embedded tumor samples were sectioned at an average thickness of 3 μm, deparaffinized, and stained with hematoxylin and eosin. The expression of FGF2 was detected as described above with the same antibody that was used for patient samples (dilution, 1:200).
For the metastasis assay, the mice were killed 2 weeks after irradiation, and the lungs were removed and placed in Bouin solution. Metastatic nodules on the surface of all of pulmonary lobes were counted macroscopically by 2 investigators.37
The Fisher exact test38 and the Kruskal-Wallis test39 were used to analyze correlations between expression variances and clinical factors. Wilcoxon signed-rank tests40 were used to analyze the percentage positive area of protein expression between pretreatment and midtreatment samples. The Mann-Whitney U test41 was used to analyze patient age and the percentage positive area of protein expression between good responders and poor responders in clinical samples and between nonirradiated and irradiated samples in the murine experiment. Survival curves were plotted according to the Kaplan-Meier method, and the log-rank test42 was used to determine significant differences. P values <.05 were considered significant.
Patient characteristics of newly enrolled patients and full study patients are listed in Table 1. First, we evaluated the changes in FGF2-T expression after RT in the 69 newly enrolled patients as a validation set to test our previous findings. Second, we investigated the expression of FGF2-S and related molecules in all 104 patients, including 35 patients from our previous study. In addition, we performed an animal experiment using 2 different tumors, 1 radioresistant tumor and 1 radiosensitive tumor, and investigated the change in FGF2 expression induced by fractionated irradiation.
Table 1. Patient Characteristics
Newly Enrolled (%)
Full Study (%)
SD indicates standard deviation; FIGO, International Federation of Gynecology and Obstetrics; SCC, squamous carcinoma; AD, adenocarcinoma including adenosquamous carcinoma; CRT, chemoradiotherapy; RT, conventional radiotherapy; C-ions, carbon-ion beam radiotherapy; CR, complete response; PR, partial response; PD, progressive disease; HPV, human papilloma virus.
Local control was defined as having no evidence of tumor regrowth or disease recurrence in the treatment volume according to pelvic examination along with computed tomography scans, magnetic resonance imaging, and biopsy.
Status at 2 years after radiotherapy: Good responders had no evidence of disease, and poor responders were either alive with disease or had died of disease.
In the validation set of 69 samples, FGF2-T was detected immunohistochemically in pretreatment and midtreatment samples (Fig. 1A,B), as reported in our previous study.28 After the initiation of RT, FGF2-T expression was significantly higher in midtreatment samples than in pretreatment samples during the fractionated RT schedule (P = .0002) (Fig. 1C), confirming our previous findings.28
We also examined the relation between the FGF2-T ratio and RT response. The midtreatment/pretreatment ratio of FGF2-T tended to be higher in patients who achieved a CR after RT compared with patients who had a PR or PD (Fig. 1D); however, the difference was not significant. Next, we investigated the prognostic value of FGF2-T in the 69 newly enrolled patients. Patients with stage IV cancer according to the FIGO classification and patients who were not followed for the full 2 years were excluded from this prognostic analysis. The remaining 49 patients were classified into 2 groups: good responders (n = 26) and poor responders (n = 23). Our results indicated that the FGF2-T ratio was related significantly to good responders (P = .025) (Fig. 1E), thus confirming the findings from our previous study.28
RT-Induced Change in FGF2-S Expression
We were able to analyze FGF2-S expression in 98 of the 104 patients (Fig. 2A,B). Six patients who had specimens that were not wide enough to evaluate the extent of positive staining were excluded from this analysis. Eighty-five patients (87%) presented with positive FGF2-S pretreatment samples (medium expression, 28 patients; strong expression, 57 patients), and 13 patients had negative pretreatment samples. FGF2-S expression was not changed in 57 of 98 patients after 1 week of RT but was increased in 23 patients (Fig. 2C). There was no correlation between FGF2-S expression and FGF2-T expression. Next, we investigated the prognostic value of FGF2-S expression in the full set of 104 patients omitting 6 patients because they could not be evaluated when we applied the same exclusion criteria that were applied for analyzing FGF2-T expression. Of the remaining 78 patients, those who were positive for FGF2-S (n = 67) in pretreatment samples had a trend toward a poorer prognosis than patients who were negative FGF2-S (n = 11), although the difference was not significant (Fig. 2D).
VEGF and CD31
We investigated VEGF expression and CD31-staining for MVD as angiogenic molecular markers43 in 77 patients who received either CRT or RT alone (Fig. 3A). One patient whose specimens were not wide enough to evaluate the extent of positive staining was excluded from this analysis. VEGF expression was decreased significantly after 1 week of RT (P = .0029); however, 25 tumors had increased VEGF expression after RT (Fig. 3B). All tumors that had up-regulated VEGF expression after RT were positive for FGF2-S at pretreatment (P = .035) (Fig. 3C).
MVD was observed clearly and measured (Fig. 3D). It was not changed after 1 week of RT in most patients (data not shown). No correlation between MVD and FGF2-S/VEGF expression was observed.
Change in FGF2 Expression in Murine Models
We also examined FGF2 expression in 2 murine tumor models using NR-S1 and SCCVII SCCs. Tumor growth delay for 2 tumors that were irradiated with gamma rays revealed differences in tumor radiosensitivity (Fig, 4A,B). In the study with single high doses, SCCVII tumors responded better to irradiation and had a longer growth delay after irradiation than NR-S1 tumors, as reported previously.35, 44 Spontaneous lung metastases were observed with both tumor types in mice that were not irradiated; NR-S1 tumors were highly metastatic to the lung, whereas SCCVII tumors had lower potential for metastases (Fig. 4C).37 Representative immunohistochemical staining of FGF2 expression and FGF2-T–positive area (%) in control and fractionated irradiation samples are shown in Figure 4D-G. There was low FGF2-T expression in control samples for both tumor types (Fig. 4D,F). The FGF2-T–positive area was increased in SCCVII tumors after irradiation (P = .005) (Fig. 4G), but there was no change in NR-S1 tumors (Fig. 4E). It is important to mention that the murine SCC tumors NR-S1 and SCCVII were not CCs; furthermore, male mice were used in this experiment.
RT plays a critical role in the management of CC. Despite the major therapy achievements have been made by with radiation physics in the past, treatment remains suboptimal in many patients, and future improvements are likely to come from biologic approaches. To differentiate patients who need further therapy and who can be observed after standard-of-care treatment, there is clearly a need to identify additional markers to aid in refining treatment strategies and improving outcomes.43 While searching for new markers, we need to remember that radiation causes severing of the normal or cancerous associations with adjacent cells, and it also changes the compositional adhesion of the extracellular matrix.14, 44, 45 Therefore, it is necessary to investigate not only the pretreatment status of tumors but also the modified tumor structures during fractionated RT.
Several reports have identified potential molecular markers, such as epidermal growth factor receptor in SCC of the head/neck and glioblastoma multiforme; Her-2/neu in breast cancer; p53 in SCC of the head and neck, breast, cervix, lung, prostate, colorectal, and brain tumors; BCL-2/BAX in cervical, prostate, head/neck, and bladder tumors; insulin-like growth factor-1 receptor in breast cancer; cyclin D1 in SCC of the head and neck and in breast tumors; VEGF in head/neck tumors; and cyclooxygenase-2 in cervical tumors.43, 46, 47 These markers were identified mostly using pretreatment samples.
The results of the current study using sequential biopsy samples of CC suggest that FGF2-T expression may be a marker of radiation efficacy. We also observed an increase in FGF2-T in radiosensitive murine tumors after local fractionated irradiation. FGF2 is released from in vitro tumor cells, and the amount is increased several to several hundred times by irradiation.22, 48 FGF2 also is up-regulated by irradiation of normal fibroblasts.22, 23 Brieger and colleagues reported a marked increase in time-dependent and dose-dependent FGF2 and VEGF release after the irradiation of oropharyngeal SCC cells.48 Subsequent irradiation of cells cultured with conditioned media, including released FGF2 and VEGF, decreased colony formation by about 50%.48 Costa et al reported that FGF2 restrained the proliferation of malignant cells by triggering senescence.49 However, some groups have questioned these findings. In vitro studies of FGF2 release after irradiation have concentrated on its relation with radioresistance through or with VEGF.11, 50 Those investigators suspected that increased FGF2 after irradiation would induce neoangiogenesis, which would protect tumor cell death, suggesting that FGF2 can be considered an oncogenic factor. Together, these findings suggest that FGF2 is a radiation-responsive marker with various functions.
We observed that the midtreatment/pretreatment FGF2-T expression ratio in good responders (patients with no evidence of disease during 2 years of follow-up) was significantly higher than that in poor responders (patients who died or developed recurrent disease). The midtreatment/pretreatment FGF2-T expression ratio was high in patients who achieved a CR after RT, and this significant difference was reflected in the disease-free survival curve that combined data from our previous work and the current study (Fig. 5). The predictive value of FGF2-T was 71% (specificity, 79%, sensitivity, 45%). These findings indicate that FGF2-T is an independent RT-efficacy marker.
To investigate another functional role of FGF2 in the clinical setting during RT, we hypothesized that there was a subpopulation of patients whose stromal fibroblasts were responsible for positive FGF2 expression with enhanced neoangiogenesis, which may cause the treatment resistance behavior observed in these tumors. We noted that patients who had positive pretreatment FGF2-S expression had a lower survival rate, and a subpopulation of these patients with positive FGF2-S expression had increased VEGF expression after RT. These patients may be potential candidates for targeted therapy with anti-FGF2 treatment. We need to continue investigating FGF2-S expression after the completion of RT or in patients with recurrent tumors to reveal a predictive value for FGF2-S and to prove our hypothesis. We also need to investigate samples taken at more time points during fractionation to determine whether FGF2-T expression increases in a time-dependent or dose-dependent manner.
The finding that a change in the expression ratio (midtreatment/pretreatment) of FGF-2, and not the level of FGF2-T expression midtreatment, has use as a marker suggests that FGF2-T may be a surrogate marker for other molecules. A transcription factor database search (TRANSFAC Professional 2008.3; BIOBASE Biological Databases, Wolfenbuttel, Germany) revealed several possible FGF2 binding factors: Early growth response-1 (Egr-1) is among the most promising of these molecules.51-53
In conclusion, to our knowledge, this is the first study using sequential clinical samples to assess changes in FGF2 expression in both tumor cells and stroma from patients with CC during RT. The increased protein expression of FGF2 during RT highlighted the important complexity of the fractionated radiation effect on tumor cells and the tumor microenvironment. For monitoring tumor response to RT, we suggest that immunohistochemical staining for FGF2-T is a useful marker.