Tumor-associated carbonic anhydrases are linked to metastases in primary cervical cancer

Abstract

Purpose: Carbonic anhydrase IX (CA9) has emerged as an important surrogate marker for hypoxia in solid tumors. CA12 shares a role with CA9 in acidification of micromilieu but it is less strictly regulated by hypoxia than CA9. In this study, we investigated expression of CA9 and CA12 mRNA in primary cervical cancer. We also examined whether CA9 expression can be an indicator of reoxygenation of tumor by measuring its mRNA expression during fractionated radiotherapy. Methods: Tumor tissues were obtained from 59 patients with uterine cervical cancer who underwent radiotherapy, and a second biopsy was taken after patients had received either 10 or 20 Gy of radiation. The follow-up period ranged from 2.4 to 75 months (median=23 months). The ratio of CA9 and β-actin mRNA expression was determined both pre- and during radiation treatment by RT-PCR. Results: CA9 and CA12 mRNA expression was detected in 62.7 and 88.1% of tumors (i.e. patients), respectively, and co-expression was observed in 61% of patients. Multivariate analysis revealed that CA9 expression was the most significant factor associated with metastasis-free survival (P=0.008, hazard ratio 34.8), whereas CA12 mRNA expression was linked to a lower risk of metastasis (P=0.007, hazard ratio of 0.07). Tumor CA9 expression was not altered following either 10 or 20 Gy of radiotherapy. Conclusion: The strong correlation between CA9 expression and metastasis suggests that CA9 expression might be an important indicator for identifying patients who require more aggressive systemic therapy.

Introduction

The presence of hypoxic conditions within solid tumors is one of the most important factors affecting treatment outcome. As hypoxia-regulated molecular pathways have been unveiled in recent years, targeting hypoxia is increasingly considered a sound and important strategy in treating cancer. Various experimental and clinical studies have shown that tumor cells under hypoxic conditions are resistant to radiation-induced cell killing compared to cells in well-oxygenated conditions. Early animal experiments showed that as the reoxygenation process occurs, the hypoxic proportion of a tumor can remain constant or even fall after radiation (Hall 2000; Van Putten and Kallman 1968). In clinical radiotherapy practice, it has been postulated that many types of tumors can be eradicated by step-wise killing of the reoxygenated proportion of a tumor using a fractionated radiotherapy regimen over 6 or more weeks. However, reoxygenation has not been verified to occur in clinical tumor samples.

Following sporadic efforts to find useful hypoxia markers, a recent study has identified 12 hypoxia-overexpressed genes (HOG) using serial analysis of gene expression in response to hypoxic stimuli (Lal et al. 2001). Among these genes, the isoenzymes of carbonic anhydrase (CA) families, CA9 and CA12, are shown as being induced under hypoxic conditions in many cancer cell lines. CA9 protein is expressed in the perinecrotic region of solid tumors and co-localizes with pimonidazole staining, a well-known hypoxia indicator (Vordermark and Brown 2003). In addition, CA9 expression is more specifically and rapidly regulated by oxygen concentration compared to other HOG such as glucose transporter-1 (GLUT-1) and -3 and vascular endothelial growth factor (VEGF), the expression of which is heavily regulated by a number of other signaling pathways (Harris 2002; Lal et al. 2001; Semenza 2003). Although both CA9 and CA12 are downregulated by VHL, the transcription of which are considered to be under control of HIF-1α, the expression and the tissue distribution of CA12 gene are not associated with the expression of CA9 (Ivanov et al. 1998). Unlike to CA9, which is usually found in tumor tissues, CA12 is distributed in a variety of normal tissues and its expression becomes stronger in tumors originating from the same tissues (Ivanov et al. 2001).

Identifying CA9 expression might be particularly meaningful in cervical cancer due to the possible link between CA9 induction and human papilloma virus (HPV) infection. CA9 was cloned from HeLa cells in 1994 and has been known to be the only tumor-associated member of the CA family, which catalyzes the carbon dioxide to carbonic acid conversion (Opavsky et al. 1996). CA9 is subsequently shown to be involved in cell–cell and cell–matrix interactions that affect loss of contact inhibition and anchorage-independent cell growth in MN/CA9 gene-transfected C33A and SiHa cervical cancer cell lines (Lieskovska et al. 1999). In addition, CA9 is considered as a biomarker of HPV infection, along with P16 and cyclin E (Keating et al. 2001; Resnick et al. 1996), and these molecules can be used as novel indicators for screening and diagnosis of HPV infection.

In the present study, we examined whether expression of CA9 and CA12 were prognostic indicators for patients with uterine cervical cancer. As a potential measure of reoxygenation, we determined the kinetics of CA9 expression in individual tumors during fractionated radiotherapy. The results in this paper demonstrated that CA9 expression is linked to distant metastasis in patients with uterine cervical cancer. The data further suggest that patients with high CA9 expression potentially have a considerable hypoxic tumor burden and are prone to fail at distant sites and hence should be administered more aggressive systemic treatments.

Methods and materials

Patients and clinical specimen

Pre-treatment cervical cancer tissues were obtained from the 59 patients who underwent radiotherapy as their main treatment modality. Our study was approved by the Institutional Review Board on the experimental studies and informed consent was obtained from each patient or a responsible relative accordingly. All patients had newly diagnosed, histologically proven squamous cell carcinoma or adenocarcinoma of the cervix. In these patients, 12 and 32 patients undertook an additional biopsy at a cumulative radiation dose of 10 and 20 Gy, respectively. We used 3–4 mm width tumor tissue and observed the frozen cut of each tissue under microscope before RNA is extracted to minimize the possibility of normal tissue contamination. This second biopsy was obtained to detect any change in hypoxia-related genes during fractionated radiotherapy. Multiple biopsies were taken avoiding the grossly necrotic portion and the tumor tissues thus obtained were fixed in the 4% neutral formalin and were snap-frozen in liquid nitrogen. Chemical coagulation was avoided as much as possible for later biopsy procedures. All tumor sections were H-E stained and were subsequently examined for the presence of tumor cells by a pathologist.

RNA preparation

About 3–4 mm length of the tumor tissues were taken from the biopsy specimens and were kept at −80°C in Trizol 500 μl until it was further processed. For RNA preparation, samples were chopped and minced, followed by addition of 200 μl of chloroform. The samples were incubated in ice for 5 min and then centrifuged to separate the aqueous phase. The upper phase was carefully removed and RNA was precipitated by adding 1 ml of isopropanol followed by centrifugation. The RNA pellet thus obtained was rinsed with 70% alcohol in DEPC water. Finally, the pellet was air-dried and was then solubilized with RNase-free water. RNA concentration was measured by spectrophotometry with optical density read at 260 nm.

RT-PCR of the CA9, CA12, and VEGF genes

Total RNA (2 μg) was reverse transcribed for 1 h at 37°C in a reaction mixture containing 5 U RNase (Amersham, Piscataway, NJ, USA), 0.5 mM dNTP (Böeringer Mannheim, Indianapolis, IN, USA), 2 μM random hexamer (Stratagene, La Jolla, CA, USA), RT buffer, and 5 U reverse transcriptase (Quiagen). The first-strand cDNA synthesis was performed with the oligo (dT) primed first-strand synthesis in a total volume of 10 μl with a reverse transcriptase. The product of the RT-PCR 1 μl was mixed with 10× buffer (MgCl₂), 3 μl of 2.5 mM dNTP, 1 μl of 10 pmol forward and reverse primer, 1 μl of Taq polymerase, and finally RNase and DNase-free water 20 μl was mixed to make a total volume of 30 μl. The reaction mixture was subjected to PCR at 94°C 5 min, 94°C 30 s, 62°C 30 s (57°C for CA 12 and 55°C for VEGF), 72°C 30 s for 30 cycles, and 72°C 7 min. The primers were CA9, forward, 5′-TAAGCAGCTCCACACCCTCT-3′, and reverse, 5′-TCTCATCTGCACAAGGAACG-3′; CA12, forward, 5′-ATGGCAGGTTCAAGTTCCAC-3′, and reverse, 5′-TCGGAACTCATGTCTCCTCC-3′; VEGF, forward, 5′-GTGGACATCTTCCAGGAGTA-3′ and reverse, 5′-ATCTGCAAGTACGTTCGTTT-3′) and GAPDH, forward, 5′-ATGTACGTAGCCATCCATCCAGGC-3′, reverse, 5′-AGGAAGGAAGGCTGGAAGAG-3′. Analysis of the resulting PCR products on 1% agarose gels showed single-band amplification products with expected sizes.

Real-time PCR of CA9 from the serially obtained tumor tissues

CA9 DNA amplification was carried out using CA9 cDNA template obtained from the mRNA produced from the RT-PCR procedure. Real-time PCR reactions were set up in a reaction volume of 15 μl using the TaqMan Universal PCR Master Mix (PE Applied Biosystems, CA, USA). A reporter dye FAM (6-carboxy-fluorescein) was covalently attached to the 5′ end and a quencher dye TAMRA (6-carboxy-tetramethyl-rhodamine) was incorporated into the 3′ end of TaqMan probe with sequence of TGAACTTCCGA GCGACGCAGCC for CA9. PCR primers of CA9 was 5′-ACCTGGTGACTCTC GGCTACAG-3′as a forward primer and 5′-TTTGAATGG GCGAGTGATTG-3′as a reverse primer. As an endogenous control, real-time PCR analysis was performed on the β-actin gene in the same cDNA samples for relative gene expression quantification. The β-actin gene primers and probe labeled with VIC^TM dye and MGB/non-fluorescent quencher were provided by Applied Biosystems. DNA amplification was done in a 96-well reaction plate format in an ABI PRISM^® 7900HT Sequence Detection system (PE Applied Biosystems). Both CA9 and β-actin PCR reactions were carried out in triplicate. Multiple negative water blanks were included in every analysis. Relative standard curves were established from the control DNA extracted from HeLa cells (1 μg cDNA/μl). The control cDNA was provided as 1 μg/ul and was diluted from 1,000 to 0.01 ng/μl. Two-hundred ng/μl of sample cDNA was loaded and the reactions were run on ABI PRISM^® 7900HT Sequence Detection system programmed to 5-min initial step at 95°C followed by 40 cycles of 15 s at 95°C and 60 s at 60°C. The mean Ct values and quantities in log linear range were used for quantification of CA9 gene and control gene.

Statistical method

Overall survival, recurrence-free survival, and metastasis-free survival were analyzed by Kaplan–Meier method. Univariate analysis was made on the relationship of CA9, CA12, VEGF, and other individual variables with respect to the above survival parameters using the log-rank test. Multivariate analysis for the same variables was also undertaken with the Cox model using the general strategy for model selection involving the hierarchic principle with the likelihood ratio chi-square test. The correlations of the expression of CA9, CA12, and VEGF were made by Spearmann rank correlation test. Pre-treatment and post-treatment CA9 expression was compared by paired T-test. All statistical analyses were performed with SAS programs (version 8.01, SAS Institute Inc., Cary, NC, USA).

Results

Association between CA9, CA12, and VEGF gene expression and clinical outcome

The 5-year overall survival, local recurrence-free survival, and metastasis-free survival rates for all patients were 73.4, 69, and 82%, respectively. Distant metastasis developed in 14 of 59 patients (23.7%). Using RT-PCR, biopsy tissues from individual tumors were measured for CA9, CA12 and VEGF mRNA expression, and were analyzed in terms of clinical outcomes. CA9 and CA12 mRNA was detected in 63% (37/59) and 88% (52/59) of tumors (i.e. patients), respectively, and VEGF expression was detected in 58% (54/59) of tumors. CA9 expression was closely correlated to that of CA12 and VEGF (r=0.75, P<0.001). Analysis of clinical outcomes showed that overall survival and local recurrence were not linked to mRNA expression of any of the three genes. Univariate analysis showed tumor size and clinical stage were significant factors related to overall survival of patients (Tables 1, 2, 3), while multivariate analysis showed tumor size was the only significant factor influencing local control of the disease (P=0.008, hazard ratio 3). In addition, multivariate analysis showed that CA9 mRNA expression was the single most important factor associated with the occurrence of distant metastasis (P=0.008, hazard ratio of 34.8; Fig. 1a). In remarkable contrast, CA12 mRNA expression was associated with less likelihood of metastasis (P=0.007, hazard ratio of 0.07; Figs. 1b). VEGF mRNA expression was not related to any clinical end-point.

Table 1.

Patients characteristics (N=59)

Characteristics	Number (%)
Age
≤60	23 (39)
>60	36 (61)
Tumor size
≥4 cm	30 (49)
<4 cm	29 (51)
Stage
I	6 (10)
IIA	16 (27)
IIB	25 (42)
III	6 (10)
IV	6 (10)
CA IX
−	22 (37)
+	37 (63)
CAXII
−	7 (12)
+	52 (88)
VEGF
−	5 (8.3)
+	54 (90)
CA IX and VEGF
++	33 (56)
Others	26 (34)

Table 2.

Univariate analysis of clinical outcome with variables

Factors	OS		RFS		MFS
	5-year OS	P value	5-year RFS	P value	5-year MFS	P value
Age
≤60	63	0.25	64	0.40	76	0.76
>60	79		70		75
Tumor size
>4 cm	83	0.04	57	0.02	60	0.15
≤4 cm	62		80		86
Stage
I	100	0.02	100	0.003	100	0.03
IIA	61		82		58
IIB	84		60		88
III	25		44		83
IV	0		0		0
CA IX
−	74	0.83	66	0.98	90	0.07
+	72		70		65
CA XII
−	69	0.82	100	0.13	57	0.20
+	73		63		78
VEGF
−	74	0.99	66	0.89	82	0.40
+	70		71		67

OS overall survival, RFS recurrence-free survival, MFS metastasis-free survival

Table 3.

Multivariate analysis of clinical outcome with various parameters

Factors	Overall survival		Recurrence-free survival		Metastasis-free survival
	P value	Hazard ratio	P value	Hazard ratio	P value	Hazard ratio
Age (≤60 vs. >60)	0.54	0.70	0.10	0.10	0.85	0.89
Tumor size (≤4 cm vs. >4 cm)	0.18	2.45	0.08	3.14	0.25	2.31
Stage (I–IV)	0.29	1.40	0.11	1.64	0.46	1.3
CA IX (− vs. +)	0.98	1.52	0.74	0.78	0.008	34.80
CA XII (− vs. +)	0.47	2.84	0.99	1.00	0.007	0.07
VEGF (− vs. +)	0.44	0.56	0.49	0.60	0.28	0.41

Fig. 1.

CA9 and CA12 mRNA expression was examined by RT-PCR in pre-treatment tumor specimens of 59 patients and the result was analyzed with respect to the survival parameters. A Metastasis-free survival of the patients with positive CA9 expression (5-year rate 67%, n=37) versus negative CA9 expression (5-year rate 82%, n=21). B Metastasis-free survival of the patients with positive CA12 expression (5-year rate 83%, n=52) versus negative CA12 expression (5-year rate 57%, n=7)

The effect of radiotherapy on CA9 gene expression

Using RT-PCR and real-time quantitative PCR, CA9 mRNA expression in tumor samples was measured quantitatively both prior to radiotherapy and after receiving either 10 or 20 Gy radiation (two separate patient groups). We found that after receiving 10 Gy treatment there was no difference in CA9 mRNA levels in biopsy samples compared to levels in tissue taken prior to treatment. Although there was considerable decrease in tumor CA9 mRNA expression following 20 Gy irradiation, this change was similar to that observed in expression of the comparison control gene β-actin. The mean ratios of CA9 to β-actin mRNA expression were 1.18 at pre-treatment and 1.03 at the completion of 20 Gy fractionated radiation therapy (P=0.22). The corresponding values for 10 Gy were 1.1 and 1.0, respectively (Fig. 2).

Fig. 2.

Change in the expression of CA9 mRNA. A second biopsy was obtained at 10 or 20 Gy of cumulative radiation dose in 12 and 32 patients, respectively, and their CA9 mRNA expression was examined by RT-PCR. Open box pre-treatment, shaded box CA9 mRNA at 10 Gy (Lt) and 20 Gy (Rt), respectively. The bottom and the top of each box represent 25 percentile and 75 percentile of the observed CA9/β-actin mRNA ratio. The figure and the marks in each box represent the median value of each group

Discussion

The aims of this study were to investigate whether expression of CA9 and/or CA12 in tumors was a prognostic indicator in primary cervical cancer, and whether CA9 could be used as a molecular marker for tissue reoxygenation. The study undertook a retrospective analysis of 59 patients with cervical carcinoma and analyzed CA9, CA12, and VEGF mRNA expression by RT-PCR in both pre- and post-irradiated tumor biopsy tissues. Expression of VEGF, a powerful mitogen for endothelial cell proliferation, is induced by hypoxia and was used as a comparison. Among a variety of parameters examined, we found that CA9 was the most significant factor predicting development of distant metastases. Studies on the clinical implication of CA9 expression in various tumor tissues show a variety of relationships with different clinical end-points (Vordermark and Brown 2003). A study similar to the present one examined CA9 expression in uterine cervical cancer and reported results consistent with those reported here in that positive CA9 immunohistochemical staining is strongly associated with disease-free and metastasis-free survival (Loncaster et al. 2001).

The reasons why CA9 expression affects uterine cervical cancer metastasis are currently unknown. Olive et al. (2001) showed a direct relationship between CA9 expression and the hypoxic status of cells using in vitro and in vivo experimental system, and that cells with high CA9 expression are radioresistant. If the high CA9 expression mediates the radioresistance of tumor cells by a certain mechanism, it might be expected to predict local recurrence. However, our results suggest that CA9 expression is strongly related to the enhanced metastatic potential of hypoxic tumors and not to the local control of these tumors by radiotherapy. It is highly likely that CA9 expression may have an important function in modulating tumor invasion and metastasis. The role of CA9 in acidifying extracellular matrices can be considered as a reason why CA9-expressing tumor cells are more metastatic; a low pH microenvironment assists in extracellular matrix degradation, which increases the likelihood of blood vessel invasion. Identification of the exact mechanisms underlying metastases associated with CA9-overexpressing cervical cancers awaits further studies.

Intriguingly, the present study showed that CA12 and CA9 expression had very different associations with metastasis. It has been reported that CA12 expression is linked to more differentiated, less invasive status of the tumor cells of the human breast cancer tissues (Wykoff et al. 2001). Besides, CA12 expression became stronger with increasing age of mouse embryos, indicating developmental regulation of this protein (Halmi et al. 2004). Usually, CA12 is distributed in a variety of normal tissues and its expression becomes stronger in tumors originating from those same tissues (Ivanov et al. 2001). It is likely that CA12 has an unknown role in tumor progression and differentiation in its own right, which is not related to hypoxia-associated pathways. Despite the above evidences, which support our findings on the inverse correlation of CA12 expression and tumor metastasis, our data need to be interpreted with caution when considering the relatively small number of patients with CA12 negative tumors in the present study.

Another important aim of this study was to define the CA9 mRNA levels within the tumor during radiotherapy. We have observed that CA9 and CA12 mRNA began to be expressed once oxygen concentrations fell below 5% O₂ in the cell culture system (data not shown). Five percent oxygen level is often found in normal tissues with low oxygen pressure as well as in intermediately hypoxic tumor tissues. Hence, CA9 mRNA expression can represent a broad range of hypoxia. Radiobiological modeling shows that the presence of cells exposed to the intermediate hypoxia between 0.5 and 20 mmHg O₂ can be very important in low dose (1.8–2 Gy) fractionated radiotherapy as opposed to a large single dose of radiation (Wouters and Brown 1997). By using RT-PCR rather than immunohistochemistry for CA expression, we expected to measure its overall level, rather than the hypoxic subvolume within the tumors. It has been hypothesized that the hypoxic fraction of tumors remains constant, or even decreases, during fractionated radiotherapy, with each radiotherapy fraction killing only the well-oxygenized proportion of tumor cells (Hall 2000; Van Putten and Kallman 1968). The present study shows that while CA9 expression decreased significantly during radiotherapy, this reduction was the same as that for the β-actin control gene. The reduction in CA9 expression appeared to be mostly due to the reduced tumor cell burden following radiation. Comparison between the expression levels in biopsies taken at two different times is not likely to represent the whole tumor due to inherent limitation using human tissue samples. An examination of a comprehensive set of HOGs might be required to explore the changes in hypoxia levels in human tumors. Furthermore, analysis of heavily irradiated human tumor specimens can be affected by the number of residual tumor cells as well as the changes in structure and integrity of subcellular macromolecules after irradiation. In our experiments, many biopsies collected following 20 Gy treatment had to be excluded from analysis as they did not contain enough tumor cells or did not allow for the extraction of enough RNA of good quality.

In conclusion, we examined the expression of CA9 and CA12 in human cervical tumors. Measuring CAs mRNA expressions using RT-PCR was sensitive, reproducible and simple, and yielded significant prognostic information. CA9 expression was strongly associated with the development of distant metastases, while CA12 expression was associated with less such metastases. A larger study would probably find that CA9 expression is linked to overall survival in uterine cervical cancer patients and may also provide further information regarding the interesting observation that CA12 expression appears to be associated with less metastasis in these patients. We propose that CA9 expression be used to identify patients who require more intensive systemic therapies due to their increased likelihood of failure at distant sites.