Hypoxia-regulated Carbonic Anhydrase-9 (CA9) Relates to Poor Vascularization and Resistance of Squamous Cell Head and Neck Cancer to Chemoradiotherapy Free

Hypoxia has been long considered a major cause of the failure of radiotherapy (1). The development of simple and reliable tests to estimate the clinically relevant hypoxia is of importance for the identification of subgroups of patients that could benefit from hypoxia targeting therapeutic policies (2). Assessment of tissue oxygen tension with polarographic electrodes or immunohistochemical detection of nitroimidazole accumulation is quite cumbersome to find a place in the clinical routine. Scintigraphic and magnetic resonance imaging of hypoxia are promising approaches, although as yet experimental. Nevertheless, all of these hypoxia detecting methods can only be applied in vivo and on a prospective basis.

In search of a simple test that would detect hypoxia even on archival tissue material, immunohistochemical detection of proteins induced by clinically relevant levels of hypoxia represents an appealing option. The transmembrane carbonic anhydrase group of proteins involved in the catalytic hydration of carbon dioxide to carbonic acid (3, 4) is up-regulated by hypoxia, probably through the HIF-1² pathway (5). In a previous study, Wykoff et al. (5) reported that CA9 localization showed a substantial, though incomplete, overlap with pimonidazole accumulation in patients with skin and bladder cancer. Therefore, the frame of hypoxia levels for CA9 induction may be of clinical relevance, at least in those tumors that do not express constitutively CA9 (i.e., renal clear cell carcinoma; Ref. 5).

In the present study, we assessed the expression of CA9 in archival tissue material from SCHNC. Our findings suggest that CA9 expression by cancer cells occurs mainly in very poorly vascularized tumors, is located around areas of necrosis, and strongly associates with resistance to radiotherapy.

Archival paraffin-embedded biopsy material from 75 primary SCHNCs and 15 normal tongue, pharyngeal, and laryngeal mucosa were retrieved, and 2-μm tissue sections were cut on slides. All of the patients had locally advanced inoperable cancer (Tx/N2b-3 or T3–4/Nx stage) and were treated with conventionally fractionated radiotherapy (2 Gy/fraction, 5 fractions/week; total dose 68–72 Gy; LINAC 6 MV X-ray irradiation) concurrently with carboplatin (area under the curve, 5 every 4 weeks or area under the curve 1 from day 1 to day 9 and from day 31 to day 39 of radiotherapy). None of the patients underwent surgery. The follow-up of patients ranges from 1–108 months (median, 14 months). For patients alive the follow-up ranges from 6–108 months (median, 48 months). Thirty-one patients had pharyngeal cancer (oropharynx and hypopharynx), 32 had laryngeal cancer, and 12 patients had cancer of the maxillary antrum or well-differentiated squamous cell cancer of the nasopharynx.

Response to treatment was assessed with a computed tomography scan of the head-neck area 45–60 days after radiation treatment completion. Complete response was defined as the disappearance of all of the measurable lesions within 2 months after treatment completion, whereas partial response refers to a 50–95% reduction in tumor size. In the present study, minimal responders (tumor reduction between 25–50%) or patients with stable and progressive disease were considered in one group as “nonresponders.”

For the detection of CA9, we used the APAAP procedure and the mouse monoclonal antihuman CA9 antibody M75 (5). Briefly, sections were dewaxed and rehydrated. After microwaving (4 min × 2) the primary antibody (dilution 1:50) was applied at room temperature for 90 min, and slides were washed in Tris-buffered saline. Rabbit antimouse antibody 1:50 (v/v) was applied for 30 min, followed by application of APAAP complex (1:1, v/v; Dako, Copenhagen, Denmark) for 30 min. After washing in Tris-buffered saline, the last two steps were repeated for 10 min each. The color was developed by 15-min incubation with new fuchsin solution. The specimens were scanned at low optical power (×40 and ×100), and the percentage of cells with positive CA9 reactivity was assessed on all of the ×200 fields (3–7 × 200 fields/case).

The JC70 monoclonal antibody (Dako) recognizing CD31 (platelet/endothelial cell adhesion molecule; PECAM-1) was used for microvessel staining using the APAAP method, as described previously (6). The specimens were scanned at low optical power (×40 and ×100), and microvessel counting was performed on ×200 fields. Three areas (per case) of high vascularization were chosen for microvessel counting. The final MVD for a case was the mean value of the three appraised fields. Vessels with a clearly defined lumen or well-defined linear vessel shape but not single endothelial cells were taken into account for microvessel counting. Interobserver variability of MVD assessment was minimal as analyzed previously (6). The MVD grouping of our cases was based on a previous study (6). The 25^th, 50^th, and 75^th percentile of MVD was used to define four MVD categories: MVD1 (very low); MVD2 (low); MVD3 (medium); and MVD4 (high).

Statistical analysis and graphs were performed using the GraphPad Prism 2.01 and the Instat 3.0 packages (San Diego, CA).³ The Fisher’s exact t test or the Yates’ continuity corrected χ² test was used for testing relationships between categorical tumor variables as appropriate. Nonparametric analysis was used to assess interobserver variability. Survival curves were plotted using the Kaplan-Meier method, and the log-rank test was used to determine statistical differences between life tables. The end points for analysis were the response rate, the local progression-free survival, and the OS starting from the last day of radiotherapy. Complete response rate was separately assessed from partial response rate, because head and neck cancer is a curable disease, and, thus, the partial response been always indicative of treatment failure. All of the Ps are two sided and Ps <0.05 were used for significance.

Strong cancer cell membrane and cytoplasmic CA9 expression was noted in 20/75 (26.6%) cases (Fig. 1). These cases were considered as positive. In the remaining “negative” cases, CA9 reactivity was either absent or very weak cytoplasmic without membrane reactivity, confined in focal tumor areas. Positive staining, when present, was located around areas of focal necrosis. Because the specimens analyzed were bioptical, a direct analysis of CA9 expression with the extension of necrosis would not be reliable and was not attempted. A direct association of CA9 expression with necrosis in squamous cell lung cancer has been noted in a submitted study.⁴ Overall, the percentage of CA9-positive cells in tumor samples ranged from 0 to 60% (median, 0%; mean, 7.7%). In CA9-positive cases, the percentage of cells with CA9 reactivity ranged from 10 to 60% (median, 30%; mean, 29%). The percentage of cancer cells with positive cases (strong membrane and cytoplasmic staining) was recorded by two observers independently. Nonparametric analysis revealed overlapping results (P = 0.0001; r = 0.94), which is explained by the very clear membrane staining provided by the antibody. Normal head and neck tissues never expressed CA9, although a rather weak cytoplasmic expression of the normal mucosa cells located adjacent to tumors was noted occasionally (independently of the tumor CA9 expression), showing that normal tissues adjacent to tumors may suffer from hypoxia. Tumoral vessels and fibroblasts were occasionally positive.

The expression of CA9 was not related to the primary site, the T and N-stage (data not shown). Tumors (34%; 17/50) with histological grade 1/2 were positive versus 12% (3/25) of grade 3 tumors (P = 0.054).

The expression of CA9 was mainly confined in tumors with very low MVD (MVD1 group; Fig. 2). The poor blood supply in these tumors probably accounts for a hypoxia-mediated induction of CA9. This finding also suggests that, in the majority of cases, quite a low number of vessels is adequate to provide a blood supply and oxygen tension enough to avoid the induction of CA9. The fact that: (a) some cases (7/19) with very low MVD did not express CA9 and that (b) some cases (8/56) with and intermediate or even high MVD expressed CA9 may show that, apart from the blood supply, the individual cancer cell metabolic demands for oxygen may also define the tumor oxygenation status. Moreover, intratumoral conditions that favor vascular collapse and reduced blood flow despite the high vascular density may also underlie the induction of CA9 in a subset of well-vascularized tumors.

On the other hand, as CA9 expression depends on the HIF-1 pathway (5), and as HIF-1 may be constitutively overexpressed in some tumors because of genetic alterations (7, 8), CA9 expression may also be constitutively expressed independently of hypoxia, at least in a subset of tumors. Nevertheless, the strong link of CA9 expression with very poor vascularization, the localization around the areas of necrosis, and the lack of studies confirming a high frequency of genetic events that stabilize the expression of HIFs in squamous carcinomas, suggest that CA9 up-regulation should be a result of hypoxia in this tumor type. Indeed, in a recent study (5), substantial overlapping of CA9 expression and of pimonidazole staining has been noted. Pimonidazole adducts are formed over a short period of time and are long-lived thereafter (9). Considerable fluctuation of the blood flow is well known to occur in tumors (10), and detection of pimonidazole may reflect areas of both acute and chronic hypoxia. On the contrary, CA9 is a stable protein that is induced slowly and accumulates over long periods of hypoxia, which suggests that CA9-stained areas probably reflect chronic hypoxia in tumors (5).

Tumors with CA9 expression had a significantly poorer complete response rate (40% versus 70.9%; P = 0.02; Table 1). This shows that the levels of hypoxia necessary for the induction of CA9 are clinically relevant to resistance to radiotherapy. Looking into the group of tumors with very low vascularization (MVD1; where CA9 was mainly expressed), all of the patients with negative CA9 expression responded completely to radiotherapy. This shows that CA9 expression is a more reliable marker of hypoxia as compared with low MVD. Even cases with a higher MVD (MVD2, 3, and 4) and CA9 expression had a poorer response rate, which shows that hypoxia levels in the range of CA9 induction, which may also occur despite the good vascularization, are related to resistance to radiotherapy.

Patients with CA9-expressing tumors had a significantly poorer LRFS and OS (P = 0.003 and 0.01, respectively). Additional analysis was performed within the CA9-positive group of patients. Using the median percentage of CA9-positive cells (in the CA9+ group, median was 30%) we divided CA9-positive tumors in two subgroups of high versus very high reactivity. No difference in survival was noted between these subgroups of patients. Survival analysis stratifying for MVD showed that highly angiogenic tumors (MVD4 group) had a significantly poorer LRFS and OS (P = 0.006) as compared with all of the other cases, which is in accordance with a previous study of ours (6).

Assuming that CA9 induction is a result of hypoxia in SCHNC, we used the CA9 expression and the angiogenic potential (as assessed with microvessel counting) to distinguish three different groups of tumors: (a) hypoxic tumors (any MVD with CA9 expression); (b) euoxic tumors with low/intermediate angiogenicity (CA9-negative, and MVD1, 2, and 3 cases); and (c) euoxic tumors with very high angiogenic potential (CA9-positive and MVD4 cases). Fig. 3 shows the LRFS and OS curves plotted for these three groups of patients, demonstrating that hypoxia assessed with the CA9 induction relates strongly to poor prognosis after definitive chemoradiotherapy. Moreover, highly angiogenic euoxic tumors similarly share a very poor outcome.

Although the hypoxia-related mechanisms of radioresistance are well studied, the role of angiogenicity in the outcome of radiotherapy is quite unclear. In previous studies, we showed that high angiogenesis and high expression of the angiogenic factor thymidine phosphorylase relates to poor local control rate in head and neck cancer treated with radiotherapy (11). Similarly, Cooper et al. (12) found an inverse association of angiogenesis with LRFS after definitive radiotherapy for cervical cancer. A significant proliferation/apoptosis advantage conferred by angiogenic factor expression on cancer cells may account for this relation (13, 14). Moreover, intensification of the angiogenic process during radiotherapy, which is evident in experimental studies (15, 16), may counteract the efficacy of radiotherapy.

In conclusion, the present study suggests that CA9 is induced in poorly vascularized SCHNC and is located around areas of necrosis. Therefore, additional evidence that CA9 is induced by hypoxia in this tumor type is provided. Moreover, the hypoxia levels required for the induction of CA9 in SCHNC are clinically important, because CA9-expressing tumors were resistant to radiotherapy. Grouping SCHNC patients undergoing radiotherapy according to CA9 expression (used as a marker of hypoxia) and to MVD (used as a marker of tumor angiogenicity), seems to have a strong predictive value. Furthermore, individualized therapeutic strategies may be guided by such a stratification. Combination of radiotherapy with bioreductive drugs, hyperthermia, or even heavy particle irradiation may improve the bad results of radiotherapy in hypoxic (CA9+) tumors. On the other hand, antiangiogenesis policies combined with radiotherapy may be of importance for the effective eradication of highly angiogenic tumors. Euoxic tumors (assessed as CA9-negative cases) sharing a low or intermediate angiogenic potential seem to be highly sensitive to standard chemoradiotherapy schedules. Application of the model in tissue material from patients recruited in large randomized studies that ended in frustration (17, 18) may reveal the subgroups of patients that can benefit from combinations of radiotherapy with various chemical or physical radiosensitizers.